Consumer - Chap 00 Prelims 5/3/04 15:54 Page I

Consumer - Chap 00 Prelims 5/3/04 15:54 Page i

Consumer Acceptance of Genetically Modiﬁed Foods Consumer - Chap 00 Prelims 5/3/04 15:54 Page ii Consumer - Chap 00 Prelims 5/3/04 15:54 Page iii

Consumer Acceptance of Genetically Modiﬁed Foods

Edited by

Robert E. Evenson

Economic Growth Center Department of Economics Yale University Connecticut, USA

and

Vittorio Santaniello

Dipartimento di Economia e Istituzioni Universita degli Studi Roma ‘Tor Vergata’ Rome, Italy

CABI Publishing Consumer - Chap 00 Prelims 5/3/04 15:54 Page iv

CABI Publishing is a division of CAB International

CABI Publishing CABI Publishing CAB International 875 Massachusetts Avenue Wallingford 7th Floor Oxon OX10 8DE Cambridge, MA 02139 UK USA

Tel: +44 (0)1491 832111 Tel: +1 617 395 4056 Fax: +44 (0)1491 833508 Fax: +1 617 354 6875 E-mail: [email protected] E-mail: [email protected] Website: www.cabi-publishing.org

© CAB International 2004. All rights reserved. No part of this publication may be reproduced in any form or by any means, electronically, mechanically, by photocopying, recording or otherwise, without the prior permission of the copyright owners.

A catalogue record for this book is available from the British Library, London, UK.

Library of Congress Cataloging-in-Publication Data Consumer acceptance of genetically modiﬁed foods / edited by Robert E. Evenson and Vittorio Santaniello. p. cm. Includes bibliographical references and index. ISBN 0-85199-747-3 (alk. paper) 1. Genetically modiﬁed foods. 2. Food--Biotechnology. 3. Consumers’ preferences. 4. Food preferences. I. Evenson, Robert E. (Robert Eugene), 1934- II. Santaniello, V. TP248.65.F66C66 2004 381.45664--dc22 2003018019 ISBN 0 85199 747 3

Typeset in 9pt Souvenir by Columns Design Ltd, Reading Printed and bound in the UK by Biddles Ltd, King’s Lynn Consumer - Chap 00 Prelims 5/3/04 15:54 Page v

Contents

Contributors vii Acknowledgements ix Introduction xi

PART I: STUDIES UTILIZING PRICE AND EXPENDITURE DATA 1 Do Agricultural Commodity Prices Respond to GMO Bans? 1 Joe L. Parcell and Nicholas G. Kalaitzandonakes 2 Consumer Acceptance and Labelling of GMOs in Food Products: 9 a Study of Fluid Milk Demand Kristin Kiesel, David Buschena and Vincent Smith 3 Consumer Purchasing Behaviour towards GM Foods in 23 The Netherlands Leonie Marks, Nicholas G. Kalaitzandonakes and Steven Vickner

PART II: STUDIES UTILIZING EXPERIMENTAL METHODS 4. The Welfare Effects of Implementing Mandatory GM Labelling 41 in the USA Wallace E. Huffman, Matthew Rousu, Jason F. Shogren and Abebayehu Tegene 5 Using Simulated Test Marketing to Examine Purchase Interest in Food 53 Products that are Positioned as GMO-free Marianne McGarry Wolf, Angela Stephens and Nicci Pedrazzi

PART III: STUDIES UTILIZING WILLINGNESS-TO-PAY METHODS 6. Measuring the Value of GM Traits: the Theory and Practice of 61 Willingness-to-pay Analysis Simbo Olubobokun and Peter W.B. Phillips 7. Willingness to Pay for GM Food Labelling in New Zealand 73 William Kaye-Blake, Kathryn Bicknell and Charles Lamb

v Consumer - Chap 00 Prelims 5/3/04 15:54 Page vi

vi Contents

8 Contingent Valuation of Breakfast Cereals Made of Non-biotech 83 Ingredients Wanki Moon and Siva Balasubramanian 9 A Comparative Analysis of Consumer Acceptance of GM Foods in 95 Norway and in the USA Wen D. Chern and Kyrre Rickertsen 10 Comparing Consumer Responses toward GM Foods in Japan and 111 Norway Jill J. McCluskey, Kristine M. Grimsrud and Thomas I. Wahl 11 Willingness to Pay for GM Foods: Results from a Public Survey in 117 the USA Hsin-Yi Chen and Wen S. Chern

PART IV: STUDIES OF CONSUMER ACCEPTANCE 12 A Comparison of Consumer Attitudes Towards GM Food in Italy and 131 the USA Marianne McGarry Wolf, Paola Bertolini and Jacob Parker-Garcia 13 Consumer Attitudes Towards GM Food in Ireland and the USA 143 Marianne McGarry Wolf, Juliana McDonnell, Christine Domegan and Heidi Yount 14 Attitudes toward GM Food in Colombia 155 Douglas Pachico and Marianne McGarry Wolf 15 Consumer Acceptance and Development Perspectives of Functional 163 Food in Germany Heiko Dustmann and H. Weindlmaier 16 Factors Explaining Opposition to GMOs in France and the Rest of 169 Europe Sylvie Bonny

PART V: STUDIES OF ECONOMIC CONSEQUENCES 17 Introducing Novel Protein Foods in the EU: Economic and 189 Environmental Impacts Xueqin Zhu, Ekko van Ierland and Justus Wesseler 18 Consumer Attitudes Towards GM Foods: the Modelling of Preference 209 Changes Chantal Pohl Nielsen, Karen Thierfelder and Sherman Robinson Index 231 Consumer - Chap 00 Prelims 5/3/04 15:54 Page vii

Contributors

Balasubramanian, S., Department of Marketing, Southern Illinois University, Carbondale, IL 62901, USA. Bertolini, P., Dip. Economia Politica, Facolt`a di Economia, Moderna, Italy. Bicknell, K., Commerce Division, PO Box 84, Lincoln University, Canterbury 8150, New Zealand. Bonny, S., INRA, UMR d’Economie Publique INRA-INAPG, BP1, 78850 Grignon, France. Buschene, D., Department of Agricultural Economics, Montana State University, Bozeman, MT 59717, USA. Chen, H.-Y., Department of Agricultural, Environmental, and Development Economics, The Ohio State University, Agricultural Admin Building, 2120 Fyffe Road, Columbus, OH 43210-1067, USA. Chern, W.S., Department of Agricultural, Environmental, and Development Economics, The Ohio State University, Agricultural Admin Building, 2120 Fyffe Road, Columbus, OH 43210-1067, USA. Domegan, C., National University of Ireland, Galway, Ireland. Dustmann, H., Forschungszentrum für Milch und Lebensmittel Weihenstephan, Technische Universität, München, Germany. Grimsrud, K.M., Department of Food Sciences, University of Guelph, Ontario, Canada N1G 2W1. Huffman, W.E., Department of Economics, Iowa State University, Ames, IA 50011, USA. Kalaitzandonakes, N.G., The Economics and Management of Agrobiotechnology Center (EMAC), University of Missouri-Columbia, Columbia, MO 65211, USA. Kaye-Blake, W., Commerce Division, PO Box 84, Lincoln University, Canterbury 8150, New Zealand. Kiesel, K., Department of Agricultural Economics, Montana State University, Bozeman, MT 59717, USA. Lamb, C., Commerce Division, PO Box 84, Lincoln University, Canterbury 8150, New Zealand Marks, L., The Economics and Management of Agrobiotechnology Center (EMAC), University of Missouri-Columbia, Columbia, MO 65211, USA. McCluskey, J.J., Department of Agricultural Economics, Washington State University, 211J Hubert Hall, Pullman, WA 99163, USA.

vii Consumer - Chap 00 Prelims 5/3/04 15:54 Page viii

viii Contributors

McDonnell, J., National University of Ireland, Galway, Ireland. Moon, W., Department of Agribusiness Economics, Southern Illinois University, Carbondale, IL 62901, USA. Nielsen, C.P., Danish Institute of Agricultural and Fisheries Economics, Rolighedsvej 25, 1958 Frederiksberg C, Denmark. Olubobokum, S., Department of Agricultural Economics, University of Saskatchewan, 51 Campus Drive, Saskatoon, Canada S7W 5AB. Pachico, D., International Center for Tropical Agriculture (CIAT), AA 6713, Cali, Colombia. Parcell, J.L., The Economics and Management of Agrobiotechnology Center (EMAC), University of Missouri-Columbia, Columbia, MO 65211, USA. Parker-Garcia, J., Agribusiness Department, California Polytechnic State University, San Luis Obispo, CA 93407, USA Pedrazzi, N., Agribusiness Department, California Polytechnic State University, San Luis Obispo, CA 93407, USA. Phillips, P.W.B., Department of Agricultural Economics, University of Saskatchewan, 51 Campus Drive, Saskatoon, Canada S7W 5AB. Rickertsen, K., Department of Economics and Social Sciences, Agricultural University of Norway, Aas, Norway. Robinson, S., International Food Policy Research Institute, 2033 K Street NW, Washington, DC 20006, USA. Rousu, M., RTI International, 3040 Cornwallis Road, Research Triangle Park, NC 27709, USA. Shogren, J.F., Department of Economics and Finance, University of Wyoming, Laramie, WY 82070, USA. Smith, V., Department of Agricultural Economics, Montana State University, Bozeman, MT 59717, USA. Stephens, A., Agribusiness Department, California Polytechnic State University, San Luis Obispo, CA 93407, USA. Tegene, A., Food and Rural Economics Division, Economic Research Service, US Department of Agriculture, Washington, DC 20036, USA. Thierfelder, K., US Naval Academy, USA. van Ierland, E., Environmental Economics and Natural Resources Group, Wageningen University, Hollandsweg 1, 6706 KN, Wageningen, The Netherlands. Vickner, S., Department of Agricultural Economics, 400 Charles E. Barnhart Bldg, University of Kentucky, Lexington, KY 40546-0276, USA. Wahl, T.I., International Marketing Program for Agricultural Commodities and Trade (IMPACT) Center, Washington State University, Hulbert Hall, Rm 123, PO Box 646214, Pullman, WA 99164-6210, USA. Weindlmaier, H., Forschungszentrum für Milch und Lebensmittel Weihenstephan, Technische Universität, München, Germany. Wesseler, J., Environmental Economics and Natural Resources Group, Wageningen University, Hollandsweg 1, 6706 KN, Wageningen, The Netherlands. Wolf, M.M., Agribusiness Department, California Polytechnic State University, San Luis Obispo, CA 93407, USA. Yount, H., Agribusiness Department, California Polytechnic State University, San Luis Obispo, CA 93407, USA. Zhu, X., Environmental Economics and Natural Resources Group, Wageningen University, Hollandsweg 1, 6706 KN, Wageningen, The Netherlands. Consumer - Chap 00 Prelims 5/3/04 15:54 Page ix

Acknowledgements

The chapters in this volume were originally presented at the Sixth International Conference of the International Consortium on Agricultural Biotechnology Research (ICABR), held at Ravello, Italy, in July 2002. They have since been edited and revised. The editors acknowledge sponsorship by the following: ● CEIS – University of Rome ‘Tor Vergata’ ● Economic Growth Center, Yale University

ix Consumer - Chap 00 Prelims 5/3/04 15:54 Page x Consumer - Chap 01 19/3/04 9:00 Page 1

1 Do Agricultural Commodity Prices Respond To GMO Bans?

Joe L. Parcell and Nicholas G. Kalaitzandonakes The Economics and Management of Agrobiotechnology Center (EMAC), University of Missouri-Columbia, Columbia, MO 65211, USA

Introduction The most significant early announcement came from a consortium of seven major After almost 20 years in the laboratory and the European food retailers on 17 March 1999.1 experimental fields, agrobiotechnology arrived Led by UK-based Sainsbury’s, the consortium at the market in the mid-1990s and was quickly had formed to source non-GM ingredients for embraced by farmers in key agricultural produc- its members’ private label products. With over ing countries. In 1996, less than 4 million acres $100 billion in combined sales, 40% in private in six countries were planted with bioengineered labels, the consortium’s announcement set the crops. By 2001, worldwide adoption expanded tone for a cascade of similar actions in the to more than 115 million acres (James, 2002). European food industry. By autumn of 1999, For some countries, including the USA, uptake some key European food retailers and manu- of bioengineered crops has been so fast that facturers had announced their intent to remove prior adoption of even dominant agricultural GM ingredients from their branded products. technologies (e.g. hybrid maize) pale in compar- Outside Europe, bans of GM food ingredi- ison (Kalaitzandonakes, 2003). ents by major food companies were more spo- Some consumers in Europe and in other radic. Starting in the autumn of 1999, a small parts of the world, however, have been scepti- number of large food manufacturers and retail- cal and often combative towards the new tech- ers in Japan, Taiwan, the USA and other parts nology (Gaskel et al.). Nothing has of the world announced intent to remove GM underscored the public acceptance woes of ingredients from their branded products. agrobiotechnology more strongly than the Bans against GM food ingredients by major widespread bans imposed by food companies food companies have been broadly reported on genetically modified (GM) food ingredients and have been often taken to signal demand in Europe and elsewhere. GM food bans shifts away from bioengineered commodities began in Europe on 1 May 1998 when Iceland in favour of identity-preserved non-GM crops. – a small UK food retail chain – announced Yet, the impacts of such bans have not been that it had removed all GM ingredients from its quantified and, indeed, the market size for private label products. Within a few months, non-GM crops remains unknown. In this study similar announcements by large food retailers we assess the significance of bans imposed by and food manufacturers poured in. major food companies against bioengineered

1 The consortium consisted of the following retail chains: Sainsbury’s, and Marks and Spencer (UK), Carrefour (France), Delhaize (Belgium), Migros (Switzerland), Effelunga (Italy) and Superquinn (Republic of Ireland). © CAB International 2004. Consumer Acceptance of Genetically Modiﬁed Food (eds R.E. Evenson and V. Santaniello) 1 Consumer - Chap 01 5/3/04 15:54 Page 2

2 J.L. Parcell and N.G. Kalaitzandonakes

commodities by investigating the response of responded to information on production, commodity prices to such bans. inventories, trade, and other demand and Assuming agricultural commodity futures supply determinants for various agricultural markets are efficient, new information suggest- commodities. Colling et al. (1996) investi- ing a possible demand shift should affect these gated the impact of ‘Export Inspections’ markets. Thus, movements in agricultural com- reports on wheat, soybean and soybean modity futures prices in response to reports on futures prices; Patterson and Brorsen (1993) voluntary bans against bioengineered com- analysed the informational content of US modities by major food companies can provide Department of Agriculture (USDA) export an indirect account of the expected impacts of sales reports; Colling and Irwin (1990) such announcements. Within this context, we analysed the response of live hog futures empirically examine whether soybean futures prices to USDA Hogs and Pigs Reports; and market prices react to news about voluntary Schroeder et al. (1990) investigated opportu- bans against bioengineered commodities by nities for livestock futures trading profits food firms.2 The Chicago Board of Trade around USDA Inventory Reports. (CBOT) soybean futures contract specification Other studies have investigated the is for #2 yellow soybean of US origin of spe- impacts of events that could signal long-term cific quality characteristics, e.g. test weight, shifts in the demand or supply of agricultural damaged kernel. The CBOT soybean futures commodities. Lusk and Schroeder (2002) market represents the primary price discovery analysed lean hog and live cattle futures mar- mechanism for soybeans in the USA. Ban ket price response to meat recall announce- announcements should produce sustained ments. They found little to suggest that either impacts on futures prices only if a shift in the the live cattle or lean hog futures price demand for non-GMO soybean is sufficient to responded to meat recalls. negatively influence the market for commod- Robenstein and Thurman (1996) analysed ity-grade soybean. the impact of health-related media announce- We also investigate the impacts of GM ban ments on the percentage return of a portfolio announcements on the returns of non-GM of livestock futures market contracts. Using soybean futures. The offering of the Tokyo daily data between January 1983 and Grain Exchange (TGE) non-GMO soybean December 1990 they found no futures market adjustment due to reports of heart disease futures contract allows analysis on such problems related to red meat consumption. impacts. If the ban announcements signal sig- This study builds on this rich event study nificant demand shifts, then the TGE non- literature and examines how soybean futures GMO soybean futures price should respond prices respond to announcements of bans positively to the news. against GM ingredients by major food firms. It also extends the previous literature in two ways. First, it investigates not only the imme- Previous Research diate and separate impact of ban announcements but also their potential cumulative Research for understanding how markets effects. Specifically, the hypothesis that each respond to information has a long tradition. additional announcement might have a Beginning with Fama’s (1970) efficient mar- greater impact on markets than prior ones by ket hypothesis, considerable information con- signalling a progressively shifting market is tent research has been conducted in various explicitly tested. Second, this study examines markets including agricultural ones. For the response of separate but interdependent instance, a number of studies have investi- market segments to such ban announcements gated how agricultural markets have – a major commodity market that can be

2 We focus on soybeans due to the pervasive adoption of bioengineered soybean varieties in key producing and exporting countries, such as the USA and Argentina. Under such adoption conditions, bans that signal signiﬁcant shifts away from bioengineered commodities should directly affect soybean commodity prices. Consumer - Chap 01 5/3/04 15:54 Page 3

Do Prices Respond to GMO Bans? 3

diminished and a smaller specialized non-GM Model III market that can be enhanced. CBOT Soybean Futures =+ΩΩ Rate of returnt 01 Cumulative ban announcement + Ω t 2 (4) Foodstuffs index market return + Ω Empirical Models t 3 + Ω Futures contract rollovert 4 USDA report dummy + ω . Following Robenstein and Thurman’s (1996) tt methodology, we examine market price The CBOT nearby futures contract was response to key ban announcements by speci- rolled forward to the next deferred futures fying the rate of return to the CBOT soybean contract the first day of the contract expira- futures price and TGE non-GMO soybean tion month. This practice was followed in futures price as the difference between the order to avoid introducing noise to the model opening futures price on the day t + 1 after caused by delivery notices served. A futures the ban announcement and the soybean contract rollover binary variable (‘Futures con- futures settlement price on day t 1. That is: tract rollover’) was included to account for the Rate of returnt = 1n(priceopening t+1) dependent variable differencing across futures (1) 1n(priceclosing t–1). contract month. This period is chosen to account for our Lence and Hayes (2001) argued that small inability to determine when exactly the ban demand shifts in niche markets with limited announcement officially became public rela- size would result in only a small price impact tive to the trading period. For example, a firm on the conventional commodity. However, if ban announcement may have occurred at 3 the demand shift in the niche market is signifi- p.m. on Tuesday and the information would cantly large, then the price impact will be not enter the market until the open more meaningful. In this study we hypothesize Wednesday morning. A similar procedure was that ban announcements decrease the used by Lusk and Schroeder (2002). demand for commodity soybeans that cannot We then specify the following empirical be guaranteed to be non-GM and increase the models: demand for non-GM soybeans. Because the size of the non-GM soybean market is Model I unknown, the expected impact on the change CBOT Soybean Futures =+ΩΩ in soybean futures from a firm ban announce- Rate of returnt 01 5 ment is left to be determined empirically. ++∑σ Firm ban announcementti Beforeti− A total of eight key ban announcements i=1 5 (2) were used to create the firm ban announce- ∑ ΦΩ+ 3 itiAfter+ 2 Foodstuffs index market ment variable. We estimate two empirical i=1 return++ΩΩ Futures contract rollover models to account for the impacts of individ- tt34ual firm announcements and a third to USDA report dummy + ω . tt account for possible cumulative impacts of such announcements. In Model I we specify Model II the firm ban announcement variable (‘Firm CBOT Soybean Futures =+Ω ban announcement’) as a 0–1 dummy and Rate of returnt 0 8 hence we measure the average impact of ban ΩΩ+ ∑ iitFirm ban announcement 9 announcements. In Model II, we specify the i=1 firm ban announcement variable in such a Foodstuffs index market return + Ω (3) i 10 way so that the individual impact of each Futures contract rollover+ Ω USDA t 11 announcement is captured separately, i.e. 1 report dummy + ω . tt when a specific ban is announced, 0 other-

3 Following Robenstein and Thurman’s (1996) approach, only key ban announcements were used in the empirical analysis. These announcements came from leading food companies or groups of companies and included a consortium of seven leading food retailers from ﬁve different EU countries, leading food manufacturers (e.g. Nestlé, Unilever, Frito Lay and Gerber) as well as leading grain merchants (e.g. ADM). Consumer - Chap 01 5/3/04 15:54 Page 4

4 J.L. Parcell and N.G. Kalaitzandonakes

wise. In Model III we specify the firm ban this model is May 2000 to December 2001. announcement variable (‘Cumulative ban May 2000 is when the TGE non-GMO soy- announcement’) so that the progressive and bean contract was initially offered. cumulative effect of each additional firm A two-limit tobit model estimation proce- announcement relative to the previous ones is dure is typically used to account for limit captured. Accordingly, we specify this variable moves associated with futures prices, i.e. max- so that it takes on a value of 1 for the first imum allowed rate of return under CBOT reg- announcement, a value of 2 for the second ulations. Alternatively, if no limit moves occur, announcement, to a value of 8 for the final then ordinary least squares estimation is suffi- announcement. cient. Because limit moves occurred for a very A foodstuffs index market return is low percentage of the trading days analysed in included to account for normal expected mar- this study, those observations were dropped in ket returns in the grain and oilseed markets, lieu of using a two-limit tobit model. i.e. all other market movements (‘Foodstuffs Additionally, futures markets are represented index market return’). The ‘USDA report by periods of varying volatility in the market. dummy’ variable is specified as 0 or 1 binary Patterson and Brorsen (1993) suggested the variables with a 1 assigned to the day of the GARCH(1,1) model to account adequately for respective USDA announcements, zero other- the periods of varying volatility. Thus, we esti- wise.4 This variable is included to account for mate the Models I–IV following a movements in the soybean futures market due GARCH(1,1) framework. We estimate all to regularly scheduled USDA news releases. empirical models using SHAZAM 9.0 (2001). The announcements are expected to vary in impact depending on the announcement type. We also specify a fourth model using the Data TGE non-GMO soybean futures price as the dependent variable in the following fashion: The period January 1995 to December 2001 Model IV was used for estimation of Models I, II and III. − Daily CBOT soybean futures prices and the Rate of return TGE non GMO Soybean Futures =+ΩΩ t 01 foodstuffs index were obtained from the 5 ++σ Commodity Research Bureau. Dates of firm Firm ban announcementti∑ Beforeti− i=1 ban announcements were obtained through a 5 (5) ΦΩ+ database maintained by the Economics and ∑ iAfterti+ 2 TGE conv. soybean i=1 Management for Agrobiotechnology Center. contract market return+ Ω Futures t 3 USDA report release dates were gathered contract rollover + ω . tt through National Agricultural Statistical There are three primary differences between Service publications. The TGE non-GMO soy- Model IV and Models I, II and III. First, the bean futures prices were obtained from the dependent variable is specified as the percent- Tokyo Grain Exchange website. age market return between the close of the day prior and close of the day after the announcement is made. This is done because Results of the operational structure of the exchange (see Parcell, 2001, for details). Second, a TGE Empirical results for Models I, II and III are conventional soybean contract market return reported in Tables 1.1, 1.2 and 1.3 respec- variable (‘TGE conv. soybean contact market tively. The estimated models have little return’) is used in place of the food index mar- explanatory power, which is common when ket return. Third, the time period covered by analysing daily futures price changes.5

4 USDA releases included Crop Progress, Crop Production, Weather, Export Intentions, Hogs and Pigs and Cattle on Feed (1995–2000). 5 It is difﬁcult to capture all the factors that lead to between-day futures market changes, e.g. speculative traders taking proﬁts or minimizing losses. Consumer - Chap 01 5/3/04 15:54 Page 5

Do Prices Respond to GMO Bans? 5

Table 1.1. Results of empirical Model I.

Coefﬁcient SE

Constant 0.0056 0.0004 Firm ban announcement 0.0426** 0.0041 Sum of ban announcement coefﬁcients 0.0053 before, at and after announcement F-stat (joint) 19.8243** 0.047 (P-value) F-stat (sum) 0.1020 0.749 (P-value) Foodstuffs index rate of return 0.3169*** 0.0254 Futures contract rollover 0.0140*** 0.0012 USDA reports 0.0002 0.0005 R 2 0.1045 No. of observations 1760 Mean of dependent variablea 0.001

a Dependent variable is the percentage rate of return of future price between open on day t + 1 and settlement on day t 1 *** and ** represent statistical signiﬁcance at the 99% and 95% level, respectively.

Table 1.2. Results of empirical Model II.

Coefﬁcient SE

Constant 0.0008** 0.0004 Firm ban announcement Announcement 1 0.0005 0.0148 Announcement 2 0.0074 0.0111 Announcement 3 0.0050 0.0111 Announcement 4 0.0034 0.0143 Announcement 5 0.0185 0.0171 Announcement 6 0.0047 0.0109 Announcement 7 0.0019 0.0205 Announcement 8 0.0109 0.0080 Foodstuffs index rate of return 0.3184*** 0.0265 Futures contract rollover 0.0106*** 0.0013 USDA reports 0.0140*** 0.0012 R 2 0.1045 No. of observations 1760 Mean of dependent variablea 0.001

a Dependent variable is the percentage rate of return of future price between open on day t + 1 and settlement on day t 1. *** and ** represent statistical signiﬁcance at the 99% and 95% level, respectively.

The foodstuffs index rate of return coefficient is also positive, which is consistent with cient and contract rollover coefficient were the theory of cost of carry in the futures mar- both statistically significant and of the ket. The statistical significance of the USDA expected sign. A change in the foodstuffs reports coefficient varied by model, and the index rate of return is positively related to the interpretation of this coefficient is difficult rate of return in the soybean futures contract. because of the various reports co-mingled to The soybean futures contract rollover coeffi- derive the variable. Consumer - Chap 01 5/3/04 15:54 Page 6

6 J.L. Parcell and N.G. Kalaitzandonakes

Table 1.3. Results of empirical Model III.

Coefﬁcient SE

Constant 0.0058 0.0004 Cumulative ﬁrm ban announcements 0.0001 0.0041 Foodstuffs index rate of return 0.3193*** 0.0256 Futures contract rollover 0.0142*** 0.0012 USDA reports 0.0002 0.0006 R 2 0.1007 No. of observations 1760 Mean of dependent variablea 0.001

a Dependent variable is the percentage rate of return of future price between open on day t + 1 and settlement on day t 1. *** represents statistical signiﬁcance at the 99% level.

Empirical results from Model I suggest that GMO soybean contract, the TGE conven- soybean futures prices did indeed respond neg- tional soybean contract rate of return coeffi- atively to ban announcements. The size of the cient was statistically significant and positively parameter estimate indicates that, on average, related to the TGE non-GMO soybean con- ban announcements resulted in a 0.043% tract. Neither the firm ban announcement decrease in the per bushel soybean futures coefficient nor the summation of coefficients price (Table 1.1). During 2001, the average accounting for the rate of return in the 5 days daily volume of the CBOT nearby soybean prior to and the 5 days after the announce- contract was 48,500 contracts. Thus, the aver- ment were statistically significant. This indi- age economic impact on the CBOT soybean cates that the impact of ban announcements futures contract would be nearly $570,000 per by key food companies, as a proxy for the announcement. However, the joint F-test on size of the non-GMO market, was considered the summation of the coefficients for the 5 large enough by the market to matter. days prior to and after the announcement was not statistically significant.6 This further finding suggests that while there was an initial soybean Discussion futures price reaction, the market readjusted to ultimately discount the information. The empirical results of the previous section Results for the firm ban announcement suggest that soybean futures markets variables in Models II and III indicate that responded little to ban announcements by there is no significant difference in the impact major food companies against bioengineered depending on which firm made the announce- commodities. While there was an initial nega- ment (Table 1.2) and there is no evidence to tive response in the CBOT soybean futures support a progressive and cumulative impact market prices after such announcements, it was of the ban announcements (Table 1.3). short lived as the market readjusted and quickly Results for the TGE non-GMO soybean discounted such information. Furthermore, contract model (IV) are presented in Table there was no indication of progressive or cumu- 1.4. The R2 for this model was substantially lative effects from sequential bans. higher than for Models I, II, and III. This Complementary analysis of the TGE non-GMO difference is likely to be due to the similarities soybean contract found that, once again, there between the non-GMO soybean contract rate is no indication that ban announcements signif- of return and the TGE conventional soybean icantly increased the value of non-GMO soy- contract rate of return. For the TGE non- bean futures contract price.

6 Individual coefﬁcients were not reported because the summation of impacts around the events is the relevant hypothesis to test. Consumer - Chap 01 5/3/04 15:54 Page 7

Do Prices Respond to GMO Bans? 7

Table 1.4. Results of empirical Model IV.

Coefﬁcient SE

Constant 0.0008 0.0007 Firm ban announcement 0.0092 0.0077 Sum of ban ban announcement 0.0390 coefﬁcients before, at, and after announcement F-stat (joint) 16.6020 0.12012 (P-value) F-stat (sum) 1.1470 0.2841(P-value) TGE conventional rate of return 0.5574*** 0.0252 Futures contract rollover 0.0100*** 0.0040 R 2 0.5206 No. of observations 406 Mean of dependent variablea 0.0004

a Dependent variable is the percentage rate of return of future price between settlement on day t + 1 and settlement on day t 1. *** represents statistical signiﬁcance at the 99% level.

These empirical results suggest that futures al., 2000; Kalaitzandonakes, 2002), markets did not perceive the bans imposed by announced bans by food companies could major food companies as signals of signiﬁcant have had only a limited effect on the demand demand shifts that would affect the use of for commodity soybeans. Markets appear to commodity soybeans or the size of the have also discounted the possibility that bans demand for identity-preserved non-GM soy- against bioengineered commodities could beans. Some of the details of the bans might extend to animal feed. Given that such bans explain this market response. With few excep- would likely result in signiﬁcant cost escalation tions, the bans announced by key food com- (Kalaitzandonakes et al., 2001), they might panies have focused on ingredients used in have been considered unlikely. food manufacturing. Only a handful of bans It is possible that commodity prices would have focused on bioengineered commodities respond strongly to bans that affect the pri- directed to animal feeds and even those have mary markets of bioengineered commodities. been rather limited in scope (e.g. focused on a Under current parameters, however, market single animal species and a limited geographic response suggests that GM bans imposed by market). Since feed use dominates the various major food companies have had limited markets for soybeans and other key commodi- impacts on commodity markets. They also ties such as maize and canola (Ballenger et indicate that non-GM markets remain thin.

References

Ballenger, N., Bohman, M. and Gehlhar, M. (2000) Biotechnology: implications for US maize and soybean trade. Agricultural Outlook April, 24–28. Colling, P.L. and Irwin, S.H. (1990) The reaction of live hog futures prices to USDA Hogs and Pigs Reports. American. Journal of Agricultural Economics 72, 84–94. Colling, P.L., Irwin, S.H. and Zulauf, C.R. (1996) Reaction of wheat, maize, and soybean futures prices to USDA ‘Export Inspections’ Reports. Review of Agricultural Economics 18, 127–136. Commodity Research Bureau (CD-ROM) Commodity Research Bureau, 330 S. Wells Street, Suite 1112, Chicago, IL 60606–7104, USA. Fama, E.F. (1970) Efﬁcient capital markets: a review of theory and empirical work. Journal of Finance 25, 383–417. Consumer - Chap 01 5/3/04 15:54 Page 8

8 J.L. Parcell and N.G. Kalaitzandonakes

Gaskel, G., Bauer, M. and Durant, J. (1998) Public Perceptions of Biotechnology: Eurobarometer 46.1. In: Durant, J., Bauer, M. and Gaskel, G. (eds) Biotechnology in the London Sphere. Science Museum, London. James, C. (2002) Global review of commercialized transgenic crops: 2001. International Service for the Acquisition of Agribiotechnology and Applications (ISAAA) Briefs, No. 21. Kalaitzandonakes, N. (2002) Agrobiotechnology and competitiveness. American Journal of Agricultural Economics 82, 1224–1233. Kalaitzandonakes, N. (ed.) (2003) Economic and Environmental Impacts of AgBiotech. Kluwer Academic–Plenum, New York. Kalaitzandonakes, N., Maltsbarger, R. and Barnes, J. (2001) The costs of identity preservation in the global food system. Canadian Journal of Agricultural Economics 49, 605–615. Lence, S. and Hayes, D. (2001) Response to an asymmetric demand for attributes: an application to the market for genetically modiﬁed crops. In: Schroeder, T.C. (ed.) NCR-134 Conference on Applied Commodity Price Analysis, Forecasting, and Market Risk Management. Dept of Agricultural Economics, Kansas State University, pp. 1–31. Lusk, J.L. and Schroeder, T.C. (2002) Effects of meat recalls on futures market prices. Agricultural and Resource Economic Review 31, 47–58. Parcell, J.L. (2001) An initial look at the Tokyo Grain Exchange non-GMO soybean contract. Journal of Agribusiness 19, 85–92. Patterson, P. and Brorsen, B.W. (1993) USDA export sales report: is it news? Review of Agricultural Economics 15, 367–378. Robenstein, R. and Thurman, W.N. (1996) Health risk and the demand for red meat: evidence from futures markets. Review of Agricultural Economics 18, 629–641. Schroeder, T., Blair, J. and Mintert, J. (1990) Abnormal returns in livestock futures prices around USDA inventory report releases. North Central Journal of Agricultural Economics 12, 293–304. SHAZAM (2001) Econometrics Computer Program Users Reference Manual, Version 9.0. McGraw Hill, New York. Tokyo Grain Exchange. Non-GMO and conventional soybean price quotes. Available at www.tge.com, (accessed May 2002). US Department of Agriculture, National Agricultural Statistical Service (NASS). Crop Progress. Various issues, 1995–2000. US Department of Agriculture, National Agricultural Statistical Service (NASS). Crop Production. Various issues, 1995–2000. US Department of Agriculture, National Agricultural Statistical Service (NASS). Weather. Various issues, 1995–2000. US Department of Agriculture, National Agricultural Statistical Service (NASS). Hogs and Pigs. Various issues, 1995–2000. US Department of Agriculture, National Agricultural Statistical Service (NASS). Cattle on Feed. Various issues, 1995–2000. US Department of Agriculture, National Agricultural Statistical Service (NASS). Export Intentions. Various issues, 1995–2000. Consumer - Chap 02 5/3/04 15:54 Page 9

2 Consumer Acceptance and Labelling of GMOs in Food Products: a Study of Fluid Milk Demand1

Kristin Kiesel, David Buschena and Vincent Smith Department of Agricultural Economics, Montana State University, Bozeman, MT 59717, USA

Introduction While there is no clear international consensus about whether biotechnology labelling should Innovations through biotechnology enable be mandatory or voluntary,2 quantitative evi- agricultural producers to reduce production dence about the effects of labelling on market- costs and/or enhance product quality for live- level patterns of consumption of biotechnology stock and crop commodities such as milk and food products is important in determining the maize, but at the same time may affect the economic value of labelling to agricultural pro- demand for the products that utilize those ducers, food processors and consumers. commodities. Many consumers, for example, Previous studies of the effects of labelling are concerned about potential risks to human have presented theoretical analyses of the possi- health although by no means all consumers ble effects of voluntary labelling on consumer have such concerns (Burton et al., 2001). The demand, in some cases in the context of house- use of the recombinant bovine growth hor- hold production models (Smallwood and mone (rBGH) in the production of milk has Blaylock, 1991; Caswell and Padberg, 1992; been a particular concern for some consumers Teisl and Roe, 1998; Golan et al., 2000). Teisl as approximately one-third of the US dairy and Roe have emphasized the role of cognitive herd, about 3 million dairy cows, currently abilities, information and time in speciﬁcally receive rBGH supplements (Monsanto, 2000). deﬁning the process by which labelling informa- Product labelling, particularly with respect tion is translated into consideration of product to the provision of health and environmental attributes and Teisl et al. (2000) have adjusted information, is increasingly being used to pro- Stigler and Becker’s (1977) model of advertising vide information about product characteristics to incorporate the effects on consumer know- such as biotechnology content that cannot oth- ledge about product attributes on the demand erwise be observed (Teisl and Roe, 1998). for a product. The theoretical model of the

1 This chapter was supported through a cooperative agreement with USDA-ERS, and by funding through the Montana State University Agricultural Experiment Station. We would like to thank Elise Golan of USDA-ERS and Wendy Stock of Montana State University for helpful comments. The views expressed in this chapter are strictly the authors’, and do not necessarily reﬂect those of Montana State University or of USDA-ERS. 2 The USA, for example, supports voluntary labelling while the EU supports mandatory labelling. © CAB International 2004. Consumer Acceptance of Genetically Modiﬁed Food (eds R.E. Evenson and V. Santaniello) 9 Consumer - Chap 02 5/3/04 15:54 Page 10

10 K. Kiesel et al.

effects of labelling presented in the next section tion information. Ippolito and Mathios (1990) of this chapter is innovative in that it incorpo- found that such labelling had significant effects. rates the consumer’s information search deci- However, Mojduszka and Caswell (2000), in a sion within a random utility specification of a test of Grossmann’s (1981) model of voluntary household production model that reflects the quality signalling, suggested that consumers uncertain nature of product information both in viewed voluntary labelling information provided the absence and presence of labelling. The by firms to be incomplete and not necessarily model provides clear predictions about the reliable. Empirical studies of the effects of impact of improved labelling on the demands labelling on milk demand are mainly limited to for a product that has a desirable but costly to the analysis of survey responses (McGuirk et al., observe characteristic and a competing product. 1992; Grobe and Douthitt, 1995; Misra and This study examines the predictions of this Kyle, 1998). Aldrich and Blisard (1998) utilized model by econometrically examining the effects monthly, pooled time series and regional (cross- of voluntary labelling about the biotechnology section) data for the period 1978–1996 to used in the production of food products on the examine whether the introduction of rBGH milk aggregate level and composition of consump- reduced aggregate fluid milk consumption, but tion of fluid milk in major US markets. The US found no evidence of such an effect. The fluid milk market provides an appealing case econometric results presented in this chapter study for examining the effects of biotechnology indicate that voluntary labelling does affect the labelling for several reasons. First, rBGH has composition of milk consumption in important been used in US milk production since 1994, ways and that some consumers are willing to providing one of the earliest examples of the pay substantial premiums for products labelled use of biotechnology in agricultural production. as free from biotechnology. Thus, it is possible to incorporate some longitu- The chapter is organized as follows. A the- dinal data into the analysis of consumption oretical model of the effects of labelling on behaviour, a facet that is especially important consumer choice is presented in the next sec- since market adjustments to labelling initiatives tion. Then, alternative specifications of appear to occur slowly over time (Teisl et al., econometric estimation models are presented 2000). Second, fluid milk is a relatively stan- and the data used in the study are described. dardized and ubiquitous processed commodity. Next, econometric estimation issues are dis- Third, and perhaps most importantly, fluid milk cussed and results presented and described. A consumption patterns involve cross-sectional summary and conclusion are presented in the differences across markets within the USA with final section of the chapter. respect to both rBGH-free labelled and unlabelled products, and conventional fluid milk products that include milk from dairy cows Theoretical Model receiving rBGH supplements. National-level supermarket scanner data compiled by Product attribute models (Becker, 1965; Information Research Inc. are available for the Rosen, 1974) are combined with models of period 1995–1999 that provide quantitative advertising and search (Stigler, 1961; Stigler information on these patterns of consumption. and Becker, 1977; Teisl and Roe, 1998; Teisl These data, made available to the authors et al., 2000) within a random utility frame- through a cooperative agreement with the US work (McFadden, 1974; Thompson and Department of Agriculture (USDA) Economic Kidwell, 1998) to assess consumer choice Research Service (ERS), are combined with over milk products. We assume that con- information on product brands compiled by the sumers receive utility from milk produced authors to create a data set that is used to esti- without rBGH through subjective attributes mate the effects of voluntary labelling on US regarding health risks, allergies, environmen- milk consumption patterns. Previous empirical tal impacts and ethical beliefs.3 Additional studies of the effects of food product labelling product label information enables a more defi- have tended to focus on the provision of nutri- nite product choice regarding these attributes.

3 The accuracy of these subjective assessments remains an open discussion, somewhat parallel to that for the beneﬁts of organic foods (see the discussion in Tweeten, 2000, and the following commentary). Consumer - Chap 02 5/3/04 15:54 Page 11

Consumer Acceptance and Labelling of GMOs in Food 11

The level of search over product attributes and m2 will be influenced by a random com- is integrated as a choice variable in the ran- ponent, e(M,L,H,t). This random component dom utility model. The randomness in utility is a function of the market share of rBGH-free arises from variation in perception or uncer- brands (M), labelling information (L), previ- tainty about product attributes that can be ously acquired human or consumption capital reduced by search. Consistent with search (H) and search time (t); all of which are models, an increase in the market share of assumed to be negatively related to the vari- products with the desired attributes, labelling ance of the random component. information about these attributes and previ- The first-order derivative for the con- ously acquired human capital reduce the vari- strained optimum in equation (1) with respect ance of the random component. to time spent searching for information is: To focus on the choice between different ∂EUx[(,,)] me +−=λ()w 0 . (2) fluid milk products, the constrained utility ∂t maximization problem is defined as: Two additional conditions further deter- maxEUxme[] ( , , ) subject to Y = xmt,, mine the benefits of search for households of n 2 (1) primary interest and their final product ∑∑px++ pm wt. i ijj choice: i==11j U(x,m ) – U(x,m ) = δ, The vector x in equation (1) includes all con- rBGH–free rBGH (3) sumption goods except fluid milk products δ ≥ 0, sold at market prices p. The consumption of E[U(x,m ,e) – U(x,m )] = δ˜. (4) specific brands of fluid milk is denoted by 1 2 vector m. The household selects search time Condition (3) describes the underlying non- (t) over fluid milk products, where the stochastic difference in utilities for a given opportunity cost of this search time is w. household between consuming milk that (with For reasons of clarity, the household is certainty) does not include rBGH and consum- assumed to search only on the absence of ing conventional milk, holding constant the the genetically modified bovine growth hor- choice of x, the vector of consumption goods. mone (rBGH) and not on any other milk This difference is assumed to be known to product attributes or over characteristics of each consumer and to vary between them other purchased products. based on their consumption of other goods, To simplify the exposition, consider an perceived health risks, environmental con- environment in which only two branded milk cerns and ethical beliefs. Consumers who

products, m1 and m2, are available where m1 consider favourably the use of rBGH in their denotes the ﬂuid milk product produced with- purchase decision have a positive δ. If a con-

out rBGH and m2 denotes a conventional sumer is indifferent between rBGH-free and 4 δ 6 ﬂuid milk product. . Suppose also that con- conventional milk products, equals zero. sumers purchase either brand m1 or brand Consumers are generally uncertain about 5 m2, but not both. The absence of rBGH in the rBGH-characteristic of m1. Equation (4) δ˜ m1 is not known to the consumer with cer- deﬁnes as the expected utility difference tainty, so the consumer’s choice between m1 between m1 and m2 given this uncertainty.

4 While this model is designed to analyse a stage of the market where consumers have already made a decision about whether or not to purchase milk, it could alternatively be applied to the initial market

participation decision. For this purpose, the deﬁnition of m2 would be extended to include no milk purchase or non-dairy substitutes such as soy-based drinks. 5 The choice of a ﬂuid milk product in this model has a discrete and continuous component. The house-

hold faces a discrete choice between m1 and m2. The household will also choose the optimal amount of m1 or m2 that maximizes utility. 6 The potential for households to prefer milk produced with rBGH (δ < 0) is quite interesting, but is abstracted from here. To our knowledge, very few, if any, consumer products are labelled touting the use of biotechnology in their production. Consumer - Chap 02 5/3/04 15:54 Page 12

12 K. Kiesel et al.

Given that second-order conditions are sat- density function for V* in the absence of isfied for the maximization problem in (1), product labelling is denoted by CDFno label, and optimal values for the choice variables x, m CDFlabel represents the cumulative density and t can be found and the following equation function when milk is labelled as rBGH-free. can be derived using the dual of (4): In practice, labelling information is usually V *(Y,p,w,δ,M,L,H) – V *(Y,p,w, discrete. In the data we consider, rBGH-free 1 2 (5) δ,M,L,H) = V*(Y,p,w,δ,M,L,H). milk may be unlabelled (L1), or it may be voluntarily labelled as rBGH-free (L2). Another The probability that milk product m will be 1 labelling regime entails certification by an selected over m given their optimal values 2 independent agency (L3). The increased ‘qual- m *, m * defined through (5) is: 1 2 ity’ of labelling increases the likelihood of pur- > ≡ > P(V* 0) P(m1 0) (6) chase for brand m1: >  < >  The following prediction can be derived P(m1 0 L1) P(m1 0 L2) < >  (8) from this framework by differentiating (6) with P(m1 0 L3). respect to labelling (L):7 If income effects are relatively small, ∂>Pm()0 ∂>PV()* 0 ∂V * 1 ≡ *.> 0 (7) Marshallian demand functions may not differ ∂L ∂V * ∂L significantly from Hicksian demand functions. Both terms on the right-hand side of equation Small income elasticities for fluid milk esti- (7) can be signed for households that have a mated in previous studies (e.g. Heien and positive δ. The probability that V* is greater Wessels, 1988) therefore suggest that own- than zero increases with V*. Ceteris paribus, price effects for branded fluid milk products for consumers with δ > 0 an increase in the are negative for the Marshallian demand func- amount of labelling information about the use tions derived from this model. of rBGH increases V* through a decrease in An additional prediction can be derived. If the error variance of δ. This effect is illus- the difference between the received utilities trated using a mean preserving reduction in from the rBGH-free and the conventional spread in the cumulative density functions for δ brand, , increases, the probability that m1 is V* in Fig. 2.1. In Fig. 2.1, the cumulative chosen will increase.8 For example, new scien-

F(V*)

1 CDFno label

Pno label (V* < 0) CDFlabel

Plabel (V* < 0)

0V* Fig. 2.1. Cumulative density functions for V*.

7 Income effects from an increase in labelling are assumed to be zero in this derivation. Although provision of labelling information may decrease search time for some households, income effects due to saved search time are likely to be very small. This assumption is supported by very small income elasticities for milk estimated in previous studies (e.g. Heien and Wessels, 1988). ∂>PV()()**00∂> PV ∂V * ≡ *. 8 This prediction can be seen from the differentiation of (6) with respect to δ: ∂δδ∂V * ∂ Consumer - Chap 02 19/3/04 9:02 Page 13

Consumer Acceptance and Labelling of GMOs in Food 13

tific information that portrays rBGH negatively production and the milk processor’s labelling (positively) in regard to health and environ- practice during the time period considered. mental risks would increase (decrease) the This additional information was obtained expected individual difference (V*). The pre- through direct contact with processors. Eleven dicted probability increase (decrease) should milk processors met the criteria for inclusion in occur in particular for households at the mar- the IRI database and also provided reliable gin for which this information would be suffi- information about rBGH and labelling charac- cient to change their search behaviour. teristics.11 These 11 firms produced 176 different milk products. They also accounted for 3.69% of nationwide supermarket skimmed Data and low fat milk sales and 2.23% of whole fat milk (as reported by IRI). Two of these 11 National-level supermarket scanner data for processors sold rBGH-free milk that was not fluid milk demand were combined with infor- labelled as such, seven sold fluid milk products mation about the use of rBGH in milk produc- that were labelled as rBGH-free and five sold tion and product-specific labelling to evaluate conventional milk products.12 None of the labelling effects.9 Over 13,000 supermarkets processors changed their policy with regard to that either belong to national chains or oper- rBGH use or labelling over the period ate independently in one of 64 metropolitan 1995–1999. Milk products from all three cate- areas around the country were tracked by gories of products – conventional, rBGH-free Information Resources, Inc. (IRI) over 13- non-labelled and rBGH-free labelled milk – week periods from January 1995 to were available over the entire time period. The December 1997 and over 4-week periods categorical data on product rBGH-characteris- from January 1998 to December 1999.10 IRI tics were coded using two mutually exclusive records sales quantity at the product code dummy variables (rBGHfreenonlabelled and (UPC) level aggregated nationwide. Prices are rBGHfreelabelled). Both of these dummy vari- temporally aggregated (within either a 13- or ables equal zero for conventional milk a 4-week period) and spatially aggregated and products.13 are based on list prices that do not take adver- A market size variable (marketsize) that tised sales into account. accounts for differences in the size of the mar- The analysis focuses on beverage milk, ket served by the 11 milk processors in this excluding buttermilk and flavoured milk and aggregated national-level data set was calcu- only considers half-gallon and gallon contain- lated. Annual state population estimates from ers. Prices and unit sales for fluid milk products 1995 to 1999 by the US Census Bureau offered by 11 different milk processors (1999) for states in which the product is avail- obtained from the IRI database were combined able were used to capture the number of with information about the use of rBGH in milk potential consumers for a given milk product.14

9 A cooperative agreement with the USDA-ERS provided access to a commercial database. 10 IRI uses the food industry’s deﬁnition of a supermarket: A grocery store with dairy, produce, fresh meat, packaged food and non-food departments, and annual sales of $2 million or more. Sales from health food stores, food cooperatives or natural food stores are not included. 11 Not all available milk processors are included in the IRI database, possibly because of limited size, local scope or unavailability in recorded supermarkets, and not all of the contacted milk processors were willing to provide reliable information about their rBGH policy. 12 Two milk processors produce items in more than one category, for example, rBGH-free labelled and conventional milk products. 13 Data on organic milk were initially also included in the analysis. However, organic milk sales were almost certainly affected by supply shocks associated with market penetration over the estimation period, resulting in implausible parameter estimates in the organic milk models. Thus, organic milk sales are excluded in the estimation models presented in the results section of this study. 14 Population estimates are expressed in terms of millions of people. Consumer - Chap 02 5/3/04 15:54 Page 14

14 K. Kiesel et al.

The data were organized into eight differ- simple price and unit sales averages. The ent fat content and container size categories generic brands used as reference brands are to permit comparisons of homogeneous prod- generally thought to include a substantial pro- ucts. For instance, demands for whole milk portion of milk produced using rBGH, but are products in gallon containers are estimated not labelled as such. separately from demands for 2% milk in gal- The panel data set used in the econometric lon containers and from demands for cate- analysis consists of 5293 observations. Each gories of milk in half-gallon containers. Two observation corresponds to a specific fluid additional variables, the logarithm of quantity milk product identified by its UPC that was ratios between each milk product and its ref- sold in a specified 13-week or 4-week period erence brand, and the price difference from 1995 to 1999. Table 2.1a provides between each milk product and its reference descriptive statistics for the variables included brand, are also computed within each fat con- in the data set. Total sales of each brand (unit-

tent and container size category. The refer- salesmi) and total sales of reference brands ence brands used in the computation of these (unitsalesmr) form the log quantity ratio, and variables were derived generic private label prices for brand i (Pmi) and for the reference product entries from the IRI database that brand (Pmr) form the price difference variable. were not identified with specific milk proces- Table 2.1b provides market share data for sors.15 Within each fat content and container these fluid milk products across fat content size category, the available generic private and container size categories. Milk sales are label entries were aggregated by computing distributed across these categories.

Table 2.1a. Descriptive statistics.

Variable Observations Mean SD Minimum Maximum

unitsalesmi 5293 142,311 228,063 1 2,030,569 7 7 unitsalesmr 312 8,351,017 1.18 10 988,116 6.56 10 ln (unitsalesmi /unitsalesmr) 5293 4.90 2.34 14.91 1.29 Pmi 5293 2.14 0.69 0.95 4.29 Pmr 5293 2.03 0.47 1.36 2.87 Pmi Pmr 5293 0.21 0.43 0.89 2.13 marketshare 5293 130.33 103.61 12.58 272.70 rBGHfreenonlabelled 5293 0.10 0.30 0 1 rBGHfreelabelled 5293 0.32 0.47 0 1

Table 2.1b. Market share of ﬂuid milk products across fat content and container size. Observations Mean unit sales % of total unit sales

Half-gallon 3216 120,335.5 51.38 Fat-free 893 135,936.89 16.12 1% 799 94,011.279 9.97 2% 810 143,545.82 15.44 Whole 714 103,949.88 9.85 Gallon 2077 176,337.54 48.62 Fat-free 629 160,145.3 13.37 1% 372 148,949.87 7.36 2% 536 238,416.82 16.97 Whole 540 152,446.12 10.93

15 Supermarket-speciﬁc private brands (e.g. Albertsons or Safeway store brands) could not be used as reference brands because they were not separately identiﬁed in the IRI database. Consumer - Chap 02 5/3/04 15:54 Page 15

Consumer Acceptance and Labelling of GMOs in Food 15

Econometric Specification Equation (13) forms the basis for the estimation equations used in the empirical analy- Following McFadden (1973) and Mathios sis. The right-hand side of equation (13) is (2000), conditional logit models were used to transformed into a linear function of the para- examine consumer purchases of fluid milk. meters for estimation.17 In this formulation, The representative consumer’s direct utility the attribute difference vector (Ai Ar) for purchase and consumption of fluid milk i denotes differences in attributes between the given generally in equation (5) is specified lin- ith fluid milk product and the reference brand. early as:16 This attribute difference vector includes price differences as well as information about V * = A β – ε . (9) i i i whether a brand is conventional (and assumed

In equation (9), the vector Ai indicates the to be produced with rBGH), produced without attributes of milk brand i, and the vector β rBGH but not labelled as such or labelled as represents the weights the household places rBGH-free. on them. The error term in equation (9) is Milk products labelled as rBGH-free are assumed to arise mainly from randomness in more likely to be chosen by those consumers attribute perception. Purchase and consump- who, ceteris paribus, have a difference in util- tion of fluid milk i over alternative milk prod- ity between rBGH-free and conventional milk. ucts indicates that: The sign of the coefficient for milk that is rBGH-free but is not labelled as such is pre- A β – ε > A β – ε for all j ≠ i. (10) i i j j dicted to be smaller in magnitude than the Under an iid logistic distributional assumption coefficient for rBGH-free labelled products (Mathios, 2000), the probability that the ith due to search costs. If search costs are higher

fluid milk product (mi) is purchased can then than the difference in utility between rBGH- be written as: free and conventional brands, then unlabelled eAi rBGH-free brands are only chosen by chance Pm()>=0 , i I and the coefficient associated with the dummy A (11) ∑ e i variable indicating that they are rBGH-free but i=1 where e the exponential function. The relative unlabelled should not be significantly positive. odds of the representative consumer choosing The econometric estimation of fluid milk product i over some reference brand, m , is: demand was carried out separately for each r fat content (fat-free, 1%, 2% and whole) and Pm()> 0 Ai i = e container size (half-gallons and gallons) to (12) > Ar Pm()r 0 e allow for varying levels of substitutability In the data set utilized in this study, the aggre- between these products. Sample sizes for gate number of units of each fluid milk prod- each of these fat content and container spe- uct sold in selected supermarkets is observed cific estimations were over 200. over either a 13-week or a 4-week period nationwide. Therefore, the left-hand side of equation (12) is redefined as unit sales of Diagnostics product i divided by unit sales of a reference brand. Redefining the left-hand side variable There may have been some important struc- this way, and taking its logarithm, equation tural changes in the fluid milk data we evalu- (12) can be rewritten as: ate. Organic fluid milk in gallons started to  unitsales  appear in supermarkets in April 1998, with  mi  = ln (AAir – ) (13)  unitsales  steadily increasing aggregate sales. mr Additionally, data collection by IRI changed

16 The focus of this analysis is the choice of ﬂuid milk brands based on their attributes. This focus, plus the assumption that the choice of x will be unchanged for different indirect utility functions, allows suppres- sion of a constant term that relates to other goods consumed in this speciﬁcation. 17 See also Mathios (2000) for a linear transformation of the conditional logit model. Consumer - Chap 02 5/3/04 15:54 Page 16

16 K. Kiesel et al.

from a 13-week to a 4-week period in 1998; enforcing rBGH-free labels. These costs are this collection method change may affect expected to arise at the milk-processing firm demand estimates even though the dependent level, and are likely to differ across firms. variable in the estimation equations is relative Although we do not have useful measures for quantity. Chow tests (Chow, 1960) were con- these firm-specific costs for fluid milk supply ducted to investigate structural change in the at the retail level, we instrument for them in market in 1998. The absence of structural the estimated demand models through firm- change was rejected at conventional levels for specific dummy variables for the milk proces- approximately one-half of the fat content/con- sors included in the sample. tainer size categories. The regression results There are several alternatives to the instru- reported below split the sample time period mental variables approach that differ in the for both half-gallon and gallon fluid milk treatment of specific firm effects on the quan- demand estimation for consistency. Coefficient tity of fluid milk purchases, including fixed estimates for the split sample do not differ effects models. Fixed effects models for greatly for estimates over the entire sample. demand without instrumenting for firm cost Several tests for heteroskedasticity were effects yielded very similar qualitative results to performed on the data set. A general test for the instrumental variables approach. There heteroskedasticity that does not specify partic- were also generally no significant improve- ular variables (Breusch and Pagan, 1979) ments in model fit when a random effects detected heteroskedasticity for all fat content approach was utilized instead of fixed effects. levels and for both container sizes. More Parameter estimates based on the instrumen- restricted tests for heteroskedasticity, using tal variables approach are reported because of specific variables to define heteroskedasticity their potential to account for firm-specific cost (White, 1980) failed to reject the null hypoth- differences in monitoring and enforcement for esis.18 Heteroskedasticity in the data appears rBGH-free labels. to be introduced by a number of factors that cannot be easily separated. Consequently, the regressions were estimated in a generalized Regression Results least squares (GLS) form and White-corrected standard errors are reported. Finally, no sig- Tables 2.2–2.5 summarize instrumental vari- nificant autocorrelation problems were ables regression results for the two separate detected in the sample, suggesting that the time periods (1995–1997 and 1998–1999) tests for structural change indicating a change and each fat content and container size com- in structure in 1998 result from the change in bination. Estimated coefficients for price dif- data collection procedures that occurred in ference variables are negative and statistically 1998 rather than shifts in consumer behav- significant in all fat content and container size iour over time. combinations and both time periods. These The firms considered in the analysis com- coefficient estimates range from 5.56 to prise less than 5% of nationwide supermarket 2.32 for half-gallons and 3.93 to 1.0 fluid milk sales in the IRI database. The share for gallons, indicating that larger price differ- of one of these particular brands in a given ences between the product of interest and the region may be considerably greater than this reference brand leads to a decrease in the log- nationwide proportion. Although the fluid arithm of the ratio of the quantity of its sales milk supply curve at the (assumed to be per- to the quantity of sales of the reference brand. fectly competitive) farm level is expected to be Estimates for the marketsize variable coeffi- quite flat, the derived retail fluid milk supply cients are also positive and significant in most curves may not be. Of particular interest are regressions, suggesting that, ceteris paribus, price/quantity relationships for retail supply as the potential market for a product that relate to the costs of monitoring and increases, sales increase.

18 In applying White’s approach, the population variable, month dummies and year dummies were used. Consumer - Chap 02 5/3/04 15:54 Page 17

Consumer Acceptance and Labelling of GMOs in Food 17

Table 2.2. Regression results for half-gallon, fat-free and 1% milk relative quantities. Fat-free 1% Independent variable 1995–1997 1998–1999 1995–1997 1998–1999

constant 4.22*** 4.32*** 3.73*** 4.28*** (0.23) (0.09) (0.14) (0.08) marketsize 0.002* 0.0007 0.004*** 0.002*** (0.001) (0.0006) (0.0007) (0.0006) rBGHfreenonlabelled 0.72* 1.67*** 2.41*** 1.60*** (0.40) (0.34) (0.32) (0.24) rBGHfreelabelled 0.083 0.66*** 0.05 0.91*** (0.35) (0.20) (0.23) (0.17) Pmi Pmr 2.32** 3.34*** 3.08*** 2.96*** (1.10) (0.33) (0.73) (0.21) Sample size 279 614 236 563 Degrees of freedom 274 609 231 558 F-statistic 2.49 35.78 51.65 92.72

Standard errors are corrected for heteroskedasticity and reported in parentheses. *, ** and *** denote coefﬁcients that are statistically different from 0 at the 10%, 5% and 1% level, respectively.

Table 2.3. Regression results for half-gallon, 2% and whole milk relative quantities. 2% Whole Independent variable 1995–1997 1998–1999 1995–1997 1998–1999

constant 3.76*** 4.70*** 4.56*** 4.41*** (0.18) (0.16) (0.23) (0.11) marketsize 0.002 0.004*** 0.004*** 0.002** (0.001) (0.0007) (0.001) (0.0008) rBGHfreenonlabelled 2.88*** 1.98*** 2.0*** 2.71*** (0.45) (0.38) (0.52) (0.44) rBGHfreelabelled 0.07 0.82*** 0.37 0.26 (0.20) (0.18) (0.30) (0.17) Pmi Pmr 4.65*** 3.75*** 5.56*** 4.17*** (0.55) (0.27) (0.67) (0.11) Sample size 274 536 227 487 Degrees of freedom 269 531 222 482 F-statistic 24.56 59.21 24.24 62.23

Standard errors are corrected for heteroskedasticity and reported in parentheses. ** and *** denote coefﬁcients that are statistically different from 0 at the 5% and 1% level, respectively.

The parameter estimates for rBGH- content categories for the second time period, freelabelled are of central interest in this study. when the data were collected over more fre- While these coefficients are only statistically sig- quent intervals. Significant positive coefficient nificant for fat-free and whole milk in gallons estimates range from 0.66 to 0.91 for half-gal- for the first time period, they are significant and lons and from 0.57 to 1.57 for gallons. These positive for almost all container sizes and fat results are consistent with the predictions of the Consumer - Chap 02 19/3/04 9:58 Page 18

18 K. Kiesel et al.

Table 2.4. Regression results for gallon, fat-free and 1% milk relative quantities.

Fat-free 1% Independent variable 1995–1997 1998–1999 1995–1997 1998–1999

constant 6.59*** 7.04*** 4.05*** 3.51*** (0.32) (0.17) (0.47) (0.12) marketsize 0.005*** 0.006*** 0.004*** 0.001 (0.002) (0.0009) (0.002) (0.0009) rBGHfreenonlabelled 2.01*** 2.02*** 0.61 1.51*** (0.31) (0.17) (0.85) (0.30) rBGHfreelabelled 1.41*** 1.45*** 0.31 0.57** (0.47) (0.28) (0.42) (0.23) Pmi Pmr 3.54** 3.43*** 2.87*** 3.93*** (0.68) (0.27) (1.05) (0.38) Sample size 184 445 121 251 Degrees of freedom 179 440 116 246 F-statistic 74.93 144.39 9.21 34.78

Standard errors are corrected for heteroskedasticity and reported in parentheses. ** and *** denote coefﬁcients that are statistically different from 0 at the 5% and 1% level, respectively.

Table 2.5. Regression results for gallon, 2% and whole milk relative quantities.

2% Whole

Independent variable 1995–1997 1998–1999 1995–1997 1998–1999

constant 5.35*** 5.05*** 4.78*** 4.47*** (0.41) (0.22) (0.31) (0.21) marketsize 0.005*** 0.006*** 0.003*** 0.003** (0.001) (0.001) (0.001) (0.0007) rBGHfreenonlabelled 0.78 0.39 0.09 1.04*** (0.71) (0.38) (0.49) (0.33) rBGHfreelabelled 0.13 1.16*** 0.96*** 1.57*** (0.34) (0.18) (0.33) (0.22) Pmi Pmr 1.0* 3.08*** 1.69*** 2.52*** (0.52) (0.27) (0.54) (0.29) Sample size 183 353 164 376 Degrees of freedom 178 348 159 371 F-statistic 6.50 51.19 4.01 25.46

Standard errors are corrected for heteroskedasticity and reported in parentheses. *, ** and *** denote coefﬁcients that are statistically different from 0 at the 10%, 5% and 1% level, respectively.

theoretical model. They indicate that labelling, their purchase decisions in response to changes by reducing information search costs, improves in labelling policies over time, although these the quality of information about product char- differences may well be associated with the acteristics and increases the quantity demanded change in the frequency with which the data for rBGH-free milk. The qualitative differences were collected in 1998. There is no evidence, in the coefﬁcient estimates for the two time however, that consumer preferences for rBGH- periods suggest that consumers may adjust free milk have declined over the period in Consumer - Chap 02 5/3/04 15:54 Page 19

Consumer Acceptance and Labelling of GMOs in Food 19

response to information that milk containing variable were negative and significant, ranging rBGH has few, if any, harmful health effects. from 1.51 to 0.09. These results indicate Positive effects on quantities demanded that the provision of relevant information on were not consistently obtained for rBGH-free a product label is required if market segmenta- products that were not labelled as such. The tion is to take place between conventional and estimated coefficients for rBGHfreenon- rBGH-free products. labelled are negative and significant for half- These results also suggest that adding a gallons with parameter estimates ranging label on the package enables consumers to from 2.88 to 0.72. Estimated coefficients make an improved product choice and, as a were only positive and significant for fat-free result, increases consumer surplus, ceteris milk in gallon containers, but the paribus. Figure 2.2 illustrates these effects. rBGHfreenonlabelled coefficient estimates Starting from a demand curve for undiffer-

were significantly greater in magnitude than entiated conventional milk products (Dconv) in the coefficient estimates for the Fig. 2.2, the provision of labelling information rBGHfreelabelled variable. The coefficient shifts the demand for differentiated rBGH-free

estimates for the rBGHfreenonlabelled vari- milk products (Dlabel) to the right. The lighter able for fat-free gallons might be influenced by shaded area illustrates the resulting increase in firm-specific factors because only one of the consumer surplus when both products have two milk processors that offered rBGH-free the same price, P. non-labelled milk sold fat-free milk over the The model specifications permit the com- analysed time period. In all other regressions putation of own price demand elasticity esti- for milk in gallon containers, the estimated mates for rBGH-free labelled and coefficients for the rBGHfreenonlabelled conventional fluid milk.19 Table 2.6 presents

Dconv Dlabel

Qconv Qlabel Q Fig. 2.2. Demand consumer surplus effects for rBGH-free labelled ﬂuid milk.

19 The following regression was estimated using an instrumental variables approach:  unitsales  1n mi  =+ααmarketshare + β rBGHfreelabeled + β() P − P rBGHfreelabeled + β () P − P conventional .   01 1 2mi mr 3 mi mr unitsalesmr The price difference and rBGH-characteristic interaction terms were instrumented using the same exogenous variables as in the primary model. Carrying out appropriate transformations, price elasticities for

β P Pmi ηβ= unitsales ***e 2 mi . mr 2 unitsales rBGH-free labelled milk can be derived from: mi For conventional milk, β β the same equation is used, substituting 3 for 2. Consumer - Chap 02 5/3/04 15:54 Page 20

20 K. Kiesel et al.

Table 2.6. Price elasticities computed at the sample means. 1998–1999 rBGH-free labelled Conventional

Half-gallon Fat-free 0.28 0.14 1% 0.002 0.68 2% 0.14 – Whole 0.05 2.35 Gallon Fat-free 0.95 0.00002 1% 0.36 0.0002 2% 0.56 0.0004 Whole 0.12 1.57

Estimated price elasticities reported in the table are signiﬁcant at the 1% signiﬁcance level.

the price elasticity estimates for rBGH-free Conclusion labelled and conventional milk for the second time period.20 Both of these categories This study is innovative in several respects. A only include branded milk products and the household production model of the effects of elasticity estimates reported here are spe- labelling has been developed that accounts cific to fat levels and container sizes. While for search costs and uncertainty about prod- some of these elasticity estimates are lower uct attributes and the quality of information in absolute magnitude than previously within a random utility framework. The reported price elasticities, in the aggregate, model implies that the provision of more reli- the results do not differ markedly from pre- able information about product quality will, vious studies (e.g. Heien and Wessells, ceteris paribus, increase consumption of the 1988; Gould, 1996; Green and Park, commodity with a desirable but costly to 1998, Glaser and Thompson, 2000). The observe characteristic and reduce consump- elasticity estimates as a whole suggest no tion of a competing commodity with an clear pattern in response to price changes undesirable characteristic. The predictions of between rBGH-free labelled and conven- the model were tested utilizing a new data set tional milk. If only the estimated price elas- on actual purchases of fluid milk produced ticities for gallon milk products are using rBGH and rBGH-free milk. The econo- considered, consumers appear more respon- metric results of the study indicate that the sive to price changes in rBGH-free labelled provision of labelling information increases milk than in conventional milk, but for half- the quantity demanded of rBGH-free milk, a gallon products there is no evidence of simi- result consistent with the predictions of the lar effects. It is important to note, however, theoretical model. This result also confirms that different fluid milk products do have dif- the findings of previous studies based on sur- ferent price elasticities of demand. veys of consumer attitudes (but not consumer

20 β β Price elasticity measures were only computed if 2 and/or ( 3 were statistically significant in the regression specification. For the first time period a number of the coefficients were insignificant, making comparisons between the elasticities impossible. In addition, using this criterion precluded estimating elasticities for unlabelled rBGH-free products. Consumer - Chap 02 5/3/04 15:54 Page 21

Consumer Acceptance and Labelling of GMOs in Food 21

behaviour in the marketplace) that indicate processors that postulate that consumer con- some consumers have a preference for milk cerns over GMO products will diminish over and other foods that are not produced with time in response to a lack of evidence about biotechnology. adverse health effects may be wrong. Another finding of interest in this study is Finally, the results of the study are also that there is no evidence that consumer pref- important in that the findings indicate that erences for rBGH-free milk products have own-price elasticities of demand for different diminished since the introduction of rBGH categories of fluid milk are substantially differ- milk products in 1994. The positive effects of ent. While the elasticity estimates reported labelling on rBGH-free fluid milk demand here are similar to those reported for aggre- appear, if anything, to have increased in the gated fluid milk demand in other studies, they period 1998–1999 as compared to the do differ between commodity milk and milk period 1995–1997. This suggests that food labelled as rBGH-free.

References

Aldrich, L. and Blisard, N. (1998) Consumer Acceptance of Biotechnology. Lessons from the rBST Experience. Agriculture Information Bulletin No. 747–01, Economic Research Service, US Department .of Agriculture, Washington, DC. Becker, G.S. (1965) A theory of allocation of time. The Economic Journal September, 493–517. Breusch, T.S. and Pagan, A.R. (1979) A simple test for heteroskedasticity and random coefficient variation. Econometrica 47, 1487–1494. Burton, M.P., Metcalfe, J.S. and Smith, V.H. (2001) Innovation and the demand for food and drug labeling regulation in an evolutionary model of industry dynamics. Structural Change and Economic Dynamics 12, 457–477. Caswell, J.A. and Padberg, D.I. (1992) Toward a more comprehensive theory of food labels. American Journal of Agricultural Economics 74, 460–468. Chow, G.C. (1960) Tests of equality between sets of coefficients in two linear regressions. Econometrica 28, 591–605. Glaser, L.W. and Thompson, G.D. (2000) Demand for organic and conventional beverage milk.’ Paper presented at the Western Agricultural Economics Association Annual Meetings, Vancouver, British Columbia, 29 June–1 July . Golan, E., Kuchler, F. and Mitchell, L. (2000) Economics of Food Labeling. Agricultural Economic Report No. 793. Economic Research Service, US Department of Agriculture, Washington, DC. Gould, B.W. (1996) Factors affecting US demand for reduced-fat fluid milk. Journal of Agricultural and Resource Economics 21, 68–81. Green, G.M. and Park, J.L. (1998) New insights into supermarket promotions via scanner data analysis: the case of milk. Journal of Food Distribution Research 29, 44–53. Grobe, D. and Douthitt, R. (1995) Consumer acceptance of recombinant bovine growth hormone: inter- play between beliefs and perceived risks. The Journal of Consumer Affairs 25, 128–143. Grossmann, S.J. (1981) The informational role of warranties and private disclosure about product quality. Journal of Law and Economics 24, 461–483. Heien, D.M. and Wessells, C.R. (1988) The demand for dairy products: structure, prediction, and decom- position. American Journal of Agricultural Economics 70, 219–228. Ippolito, P.M. and Mathios, A.D. (1990) Information, advertising and health choices: a study of the cereal market. RAND Journal of Economics 21, 459–480. Mathios, A.D. (2000) The impact of mandatory disclosure laws on product choices: an analysis of the salad dressing market. Journal of Law and Economics 43, 651–675. McGuirk, A.M., Preston, W.P. and Jones, G.M. (1992) Introducing foods produced using biotechnology: the case of bovine somatotropin. Southern Journal of Agricultural Economics July, 209–223. McFadden, D. (1973) Conditional logit analysis of qualitative choice behaviour. In: Zarembka, P. (ed.) Frontiers in Econometrics. Academia Press, New York. McFadden, D. (1974) The measurement of urban travel demand. Journal of Public Economics 3, 303–328. Consumer - Chap 02 5/3/04 15:54 Page 22

22 K. Kiesel et al.

Misra, S.K. and Kyle, D.C. (1998) ‘Demand for milk produced with and without recombinant bovine somatotropin. Journal of Agribusiness 16, 129–140. Mojduszka, E.M. and Caswell, J.A. (2000) A test of nutritional quality signaling in food markets prior to implementation of mandatory labeling.’ American Journal of Agricultural Economics 82, 298–309. Monsanto (2000) Status Update: Posilac Bovine Somatotropin. Available at http://www.monsanto dairy.com/updates/Bovine.htm#general (accessed March 2002). Rosen, S. (1974) Hedonic prices and implicit markets: product differentiation in pure competition. Journal of Political Economy 82, 34–55. Smallwood, D.M. and Blaylock, J.R. (1991) Consumer demand for food and food safety: models and applications. In: Caswell, J.A. (ed.) Economics of Food Safety. Elsevier Science Publishing Co., New York, 4–27. Stigler, G.J. (1961) The economics of information. Journal of Political Economy 69, 213–225. Stigler, G.J. and Becker, G.S. (1977) De gustibus non est disputandum. American Economic Review 67, 76–90. Teisl, M.F. and Roe, B. (1998) The economics of labeling: an overview of issues for health and environmental disclosure. Agricultural and Resource Economics Review October, 140–150. Teisl, M.F., Roe, B. and Hicks, R.L. (2000) Can eco-labels tune a market? Evidence from dolphin safe labeling. Working paper, University of Maine. Thompson, G.D. and Kidwell, J. (1998) Explaining the choice of organic produce, cosmetic defects prices and consumer preferences. American Journal of Agricultural Economics 80, 277–287. Tweeten, L. (2000) Coexisting with alternative agriculture advocates (guest editorial). Choices Second Quarter, 3. US Census Bureau (1999) State population estimates; annual time series, July 1, 1990 to July 1999. Available at http://www.census.gov/population/estimates/state/st.-99–3.txt (accessed September 2001). White, H. (1980) A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 48, 817–838. Consumer - Chap 03 5/3/04 15:54 Page 23

3 Consumer Purchasing Behaviour Towards GM Foods in The Netherlands

Leonie A. Marks,1* Nicholas G. Kalaitzandonakes1 and Steven S. Vickner2 1The Economics and Management of Agrobiotechnology Center (EMAC), University of Missouri-Columbia, Columbia, MO 65211, USA; 2Department of Agricultural Economics, 400 Charles E. Barnhart Bldg, University of Kentucky, Lexington, KY 40546-0276, USA

Introduction respond if faced with such a decision. Multiple opinion polls (e.g. European Commission It is June 1997 and Johanna van Buren1 is at (EC), 1997, 2000; Durant et al. 1998; her local supermarket in the outskirts of Gaskell et al. 1999) have consistently indi- Amsterdam. It’s Wednesday afternoon and cated that a majority of Europeans would she is doing the weekly shop for her family of avoid GM foods in the market place. Yet few four. As usual, she is short on time so she is European consumers have ever been con- hastily walking up and down the aisles check- fronted with such a decision. Since 1999, ing items off her list. When she reaches the food manufacturers and retailers in Europe aisle with the canned soups she is faced with have imposed voluntary bans against GM an array of brands to choose from. Johanna ingredients in their branded foods and have already knows what she is looking for. She avoided mandated GM labelling altogether quickly picks up her favourite can of tomato (Kalaitzandonakes, 2000). soup; but as she puts the can in her shopping Under these circumstances, no study has cart she notices the words genetisch gemodi- ever answered the question of how European ficeerd (genetically modified) on the ingredi- consumers would actually behave in the ents list. She hesitates as she considers the presence of positively labelled GM foods. new information. What will Johanna do next? Instead, much of what is known today about Will she leave the soup in her cart or put it consumer purchasing intentions towards GM back on the shelf and look for another one foods in Europe is derived from attitude sur- that does not contain genetically modified veys. Divergence between attitudes and pur- (GM) ingredients? chasing behaviour, however, is not The question of how consumers in Europe uncommon and in the case of GM foods it would respond to foods with labels indicating has been observed in the past. As bovine the presence of GM ingredients has been on somatotropin (rBST) was being introduced the minds of biotechnology and food firms, in the US market place in 1995, a survey researchers and policy makers alike. A large conducted by Douthitt et al. (1996) found volume of research has been generated to that 74% of consumers expressed concern gauge how European consumers might about possible long-term health effects from

* Senior authorship is not assigned. 1 Johanna is purely ﬁctional, used in this context to represent a typical Dutch consumer. © CAB International 2004. Consumer Acceptance of Genetically Modiﬁed Food (eds R.E. Evenson and V. Santaniello) 23 Consumer - Chap 03 5/3/04 15:54 Page 24

24 L. Marks et al.

consuming milk from rBST-treated cows and 1997, 2000; Durant et al., 1998; Gaskell et reluctance to purchase it. Hindsight being al., 1999), to up-to-the-minute polls carried 20–20, we now know that such attitudes did out by the media. A range of questions has not translate into significant changes in pur- been asked at different locations and points in chasing behaviour – or avoidance – on the time. Responses vary considerably depending part of US consumers (Aldrich and Blisard, on how the questions are framed, the kind of 1998). The vast majority of US consumers sample used (e.g. size, demographics, loca- purchased milk from rBST-treated cattle even tion), and over time. when it was offered side-by-side with ‘non- Universally, Europeans have been more rBST’ milk (Runge and Jackson, 2000). negative about GM foods than other applica- In this study we depart from the previous tions of biotechnology, e.g. medical or envi- literature and focus on revealed rather than ronmental ones (Durant et al., 1998; Gaskell stated consumer preferences towards GM et al., 1999; EC, 2000). Likewise, public sen- foods in Europe. Specifically, we examine timent towards GM foods has been more neg- how consumers actually behaved when they ative in Europe than in North America and could choose between GM-labelled and unla- elsewhere (Hoban, 1996, 1998; Gaskell et belled food products in supermarkets across al., 1999). Even within Europe, however, atti- The Netherlands. We analyse national weekly tudes have varied by country and over time. point-of-purchase data for four separate food For instance, southern European countries product categories that include products with were more accepting of GM foods than GM labels. Our observations began in mid- northern countries up until 1997 1997. Three months after our initial observa- (Zechendorf, 1998). By 1999, Italy and tion, GM labels were introduced on a number Greece exhibited the most negative attitudes of products that contained GM ingredients towards GM foods in Europe (EC, 2000). while the rest remained unlabelled. Almost 3 Attitude surveys can capture public senti- years later, GM labels were removed as manu- ment towards GM foods and biotechnologies, facturers sourced non-GM ingredients. Hence, and investigate the sources of such sentiment our data set allows us to investigate the behav- and other associations. They are relatively iour of consumers when GM labels were intro- easy to execute, cost effective and can be gen- duced and then when they were removed eralized to a population. Attitude surveys are from selected products. constrained, however, by their hypothetical structure, especially since they do not account for price and income effects on consumer- Gauging Consumer Purchasing stated preferences. Attitude surveys may also Intentions Towards GM Foods in Europe: engage their subjects as citizens rather than What Do We Know? strictly as consumers (Noussair et al., 2002). Subjects may use information and beliefs to There is a growing literature that is investigat- respond to the survey that they would not use ing consumer attitudes and stated preferences while making narrower purchasing decisions towards GM foods in Europe. What follows is (Noussair et al., 2002). Accordingly, attitudes a brief discussion of the main approaches may or may not be effective proxies of con- used to date and their potential advantages sumer market behaviour. and pitfalls in anticipating consumer prefer- Importantly, attitude surveys can be subject ences and behaviour in the market place. to significant biases. How questions are framed, the order in which information is presented, the degree of knowledge and under- Attitudinal surveys standing of the respondent, are just some of the potential sources of bias and error (Tolley Opinion and attitude surveys about GM foods and Randall, 1983; Kahneman and Tversky, range from the in-depth studies carried out by 1984; Sterngold et al., 1994). Biases, can be various academics, government agencies and minimized but not eliminated through careful organizations (e.g. Hoban, 1996, 1998; EC, design and statistical analysis. Consumer - Chap 03 5/3/04 15:54 Page 25

Purchasing of GM Foods in The Netherlands 25

Willingness-to-pay studies get a hypothetical answer. Responses on hypothetical scenarios run the risk of giving While the bulk of current research has focused unreliable results particularly when respon- on attitudinal surveys, some researchers have dents are not well informed. This criticism is utilized the concept of ‘willingness to pay’ in pertinent to GM food studies where con- order to capture how consumers might sumers can exhibit a high level of unaware- respond to GM foods if faced with realistic ness about the technology. food choices. Consumers may be willing to The approach is also susceptible to two pay more for GM foods exhibiting desirable other important types of bias – strategic bias attributes (e.g. Boccaletti and Moro, 2000; and hypothetical bias. Strategic bias can occur Burton and Pearse, 2002). Alternatively, con- when consumers deliberately understate or sumers may be willing to pay more to avoid overstate the true value they place on an them altogether (Moon and Balasubramanian, attribute – for example, if they believe that by 2001; Lusk et al., 2003). so doing they might influence a policy out- Contingent valuation is the best-known, come. Hypothetical bias, on the other hand, most frequently used ‘willingness to pay’ typically occurs when consumers are unable to method. Willingness-to-pay measures are esti- accurately assess their willingness to pay. mated from direct consumer responses to a Hypothetical bias is possible even in well- set of hypothetical questions. Consumers are designed surveys, particularly when consumers given a hypothetical scenario and are asked to have limited prior experience with the attribute make a hypothetical choice. Surveys usually (in this case GM versus non-GM). Lack of also gather socioeconomic data about the actual purchasing choices can make it very dif- respondents, ask additional attitudinal ques- ficult for consumers to become aware of their tions, and use follow-up questions to evaluate own preferences so that they can place a value whether the respondents understood the sce- on changes in price, quantity and quality. nario presented. Finally, some studies have shown that even the Using willingness-to-pay surveys, Lusk et order in which questions are asked can affect al. (2003) and Moon and Balasubramanian willingness-to-pay measures by a significant (2001) concluded that many European con- magnitude (Tolley and Randall, 1983). sumers would need a discount to consume foods with GM ingredients. Specifically, Lusk et al. (2003) compared consumer valuations Experimental auction studies of beef steaks from cattle fed with GM maize across four countries – France, Germany, the A handful of recent studies have used experi- UK and the USA. They found that European mental auction markets to directly elicit con- consumers placed a higher value than US sumer preferences towards GM foods (Buhr et consumers on beef raised without GM maize. al., 1993; Huffman et al., 2001, 2002; Lusk Moon and Balasubramanian (2001) compared et al., 2001; Noussair et al., 2004). Such US and UK consumer valuations of GM and experimental approaches can produce more non-GM foods with specific attributes. They detailed results on consumer-stated prefer- concluded that UK consumers were more will- ences than attitude surveys and willingness-to- ing to pay price premiums to avoid GM foods pay studies. than US consumers. Noussair et al. (2004) investigated Willingness-to-pay surveys can yield impor- European consumer response to GM-versus tant information about individual consumer- non-GM-labelled foods in an experimental lab- stated preferences. They can also usefully oratory setting in Grenoble, France. The study combine such information with demographic consisted of a representative sample of 97 data and consumer attitudes. However, they consumers. The consumers compared various are limited by their hypothetical design, and biscuit brands, some of which carried a GM by the constraining complexity of GM food label, while others carried an organic label or choices. This criticism is often stated as fol- no label at all. Noussair et al. (2004) found lows – if you ask a hypothetical question you that 35% of consumers boycotted GM-labelled Consumer - Chap 03 5/3/04 15:54 Page 26

26 L. Marks et al.

foods, 40% were willing to purchase products exhibited in a store environment. For example, containing GM ingredients if they were suffi- Shogren et al. (1999) examined consumer ciently inexpensive, and 25% of the partici- response to irradiated chicken cuts in three dif- pants were indifferent and would purchase ferent settings – an actual retail store where them regardless.2 the irradiated food item was clearly labelled as Experimental auction markets provide a such, an experimental auction market and a more realistic environment than surveys for hypothetical market survey. They found that eliciting consumer preferences. Indeed, econ- the acceptability of the product was greatest in omists have developed approaches that offer the hypothetical setting and least in the retail incentives for consumers to bid their true outlets. Shogren et al. (1999) hypothesized reservation value (Menkhaus et al., 1992). that differences in acceptance were due to the Likewise, experimental auction markets allow importance of information provided in a leaflet researchers to ask attitudinal questions of con- accompanying the irradiated food in the sumers and, in turn, relate them to stated experimental setting. In the store, consumers preferences. Finally, experimental markets did not read the leaflet and were more nega- allow researchers to introduce different types tive towards food irradiation. of information shocks, and to observe Generalization from experimental auction changes in participants’ behaviour (Shogren market research can also be problematic. et al., 1999). Given that most such studies are carried out in Of course, experimental auction market specific locations and/or with small samples analysis is not without limitations. Though of consumers, demonstrating the representa- elaborate, experimental auction markets are tiveness and generality of the results can often still artificial environments. The range of items be challenging. for purchase is much more limited than in an actual retail store. Furthermore, participants are typically asked to bid, paying ‘real’ money Measuring Consumer Behaviour towards for the purchase of ‘real’ goods at the end of GM Labels in the Netherlands: Approach the experiment. However, participants still and Procedures know that they are being observed and can deviate from normal behaviour. For instance, The literature review in the previous section participants may fall foul of what is known as suggests that substantial research effort has the ‘Hawthorne effect’ where they can over- been focused on eliciting consumer attitudes state their bids to please the monitor of the and stated preferences towards GM foods in experiment (Shogren et al., 1999, p.1192). Europe and elsewhere. Yet attitudes and Accordingly, consumer-stated preferences stated preferences are only proxies to actual elicited in experimental auction markets can be consumer behaviour towards GM foods. Such different from normal purchasing behaviour behaviour has not been directly analysed in

2 A few additional studies have used experimental auction markets to investigate stated consumer preferences outside of Europe. Buhr et al. (1993) used a split-valuation experimental auction to evaluate consumer response to porcine somatotropin (PST)-treated pork products in the USA. Despite early opinion surveys that indicated that consumers would avoid such products (Hoban and Burkhardt, 1991), Buhr et al. found that consumers were willing to pay a premium for leaner meat with fewer calories produced using PST. More recently, Huffman et al. (2002) used experimental auction markets to investigate consumer willingness to pay for both positively and negatively labelled GM foods in the USA. The experiments were conducted in two mid-western US cities across a random sample of 142 residents. They chose three different types of products, vegetable oil, tortilla chips, and a bag of potatoes, in order to capture how consumers might react to GM content in different types of foods. They found that consumers were able to distinguish between GM and non-GM labels and responded to such information (Huffman et al., 2002, p. 21). Lusk et al. (2001) used experimental auction markets to examine willingness to pay for non-GM maize chips among college students in a particular US location. They found that 20% were willing to pay some premium for such chips. Consumer - Chap 03 5/3/04 15:54 Page 27

Purchasing of GM Foods in The Netherlands 27

Europe.3 In this study, we examine revealed set spans 247 consecutive weeks, from the instead of stated consumer preferences by Sunday ending 13 April 1997 to the Sunday directly analysing consumer behaviour ending 30 December 2001. Within this time towards GM foods in one European country – frame we capture two important natural The Netherlands. experiments of consumer response to GM The Netherlands provides an excellent case labelling. Across those relevant products con- study for examining consumer response to taining GM ingredients, the labels were intro- GM food labels. First, a meaningful number of duced in the 11th week of our data set, the food products carried positive GM labels for week ending on Sunday 22 June 1997. The an extended period of time allowing detailed labels remained on relevant products for 151 analysis of actual consumer behaviour. consecutive weeks and were removed the Second, Dutch consumers have consistently week ending Sunday 14 May 2000 as non- exhibited higher overall levels of awareness of GM ingredients were sourced. biotechnology and GM foods in opinion sur- The products in each product category were veys (Hoban, 1998; Hamstra, 1993). Third, aggregated into two groups: GM-labelled and The Netherlands has a long history of active unlabelled. Our analysis therefore compares consumer involvement in food and nutrition consumer behaviour towards GM-labelled prod- policy, with over 10% of Netherlands house- ucts and unlabelled products sold side-by-side. holds belonging to consumer organizations Our non-linear conditional expenditure share (Hillers and Lowik, 1998). model accounts for the separate influences of own price, the price of substitutes, per capita real expenditure for the product category, holi- Model development day effects, and the addition and removal of GM labels. Consumer behaviour could also For our analysis, we use a statistical model of change over time in response to new informa- consumer response, consistent with the con- tion about GM foods and biotechnology. To temporary time series demand literature control for such effects, we include a variable (Swartz and Strand, 1981; Smith et al., that measures the amount of relevant informa- 1988; Teisl et al., 2002). The underlying tion that could have been received by consumers framework is the well-known AIDS model from media sources. Accordingly our AIDS (Deaton and Muellbauer, 1980). Other studies model is specified as follows: 2 have used the logit-based, random utility =+αγ + β + wpxPit i∑ ijlog jt i log( i / i ) framework (e.g. Mathios, 2000). This latter j=1 ηφ+++ φ λ + λ + approach employs an arbitrary Gumbel error itzh i11 t i 22 h t i 1 onon424 t i 2 t λλoff4244++⋅+ off θ m on distribution and often violates the itititt34 1 (1) θθθ⋅+⋅++ Independence of Irrelevant Alternatives (IIA) it234m on24 t it m off 4 t it m assumption (McFadden, 1973). ⋅+ε off24t it A national-level, syndicated point-of-pur- where chase scanner data set is used for the analysis. 2 2 2 =+αα +1 γ Four product categories are considered: logPptk0 ∑∑ log kt∑ jk ==2 = canned soup, frozen processed meat, frozen k 1 j 1k 1 (2) logpp log . pizza and frozen processed fish. In each prod- jt kt uct category some products contained GM i = 1 indicates GM-labelled products, i = 2 ingredients and carried a GM label while oth- represents all unlabelled products and t indi- ers did not and remained unlabelled. The data cates time measured in weeks.

3 Outside Europe, two studies have focused on revealed consumer preferences towards GM foods. James et al. (2002) set up a limited market experiment and observed consumer purchasing patterns towards GM and non-GM sweetcorn they placed in a few grocery stores in a college town in Pennsylvania, USA. Kiesel et al. (2002) examined a national dataset of actual consumer purchases of ﬂuid milk produced with rBST and rBST-free milk. Thus, they examined consumer response to negative (‘does not contain’) labels. Their results indicate that a small segment of consumers responded positively to such labels. Consumer - Chap 03 5/3/04 15:54 Page 28

28 L. Marks et al.

To clarify the specification above, con- Testing consumer response to labelling sider the GM-labelled soup market share information equation (w ) in week t. It is a function of 1t The main empirical question addressed in this own price (logp ) expressed as an index 1t paper is the degree to which Dutch con- (Moschini, 1995), the price of unlabelled sumers avoided labelled GM food products soup (logp ) expressed as an index, per 2t when they had the opportunity to do so. The capita real expenditure for the entire soup product labelling variables (on4 ,on24 ,off4 , product category (log(x /P )), a linear time t t t t t off24 ) in equation (1) are used to measure trend (z ), holiday effects (h ,h ), product t t 1t 2t consumer response to on-package labelling labelling variables (on4 ,on24 ,off4t, t t information related to GM ingredients. off24 ), external information variables t A number of functional forms have been (m •on4 ,m •on24 ,m •off4 ,m •off24 ) and t t t t t t t t proposed in the literature to model the impact a stochastic error. Equation (2) delineates of labelling information on consumer pur- the construction of the unobservable non- chases over time, including dummy variables, linear price index. linear trends, non-linear trends and S-shaped The holiday effects (h1t,h2t ) are set to one in the week of the calendar holiday and response functions. After some experimenta- zero otherwise. The Queen’s birthday tion with all of these functional forms, we (observed on 30 April), Remembrance Day constructed 4-week and 24-week linear trends (observed on 4 May) and Liberation Day to model potential immediate and/or gradual 4 (observed on 5 May) are combined into one impacts from the introduction of GM labels. Similarly, we constructed 4-week and 24- holiday effect (h1t ) given the closeness in dates. Sinterklaas (observed the eve of week linear trends to model the immediate 5 December and the following day) is given and gradual potential response to the removal of GM labels from those same products. The by (h2t ). The linear time trend (zt ) increments by one unit for each week in the weeks before and after each linear time trend database. The external information and GM were set to zero in each of the four series. labelling variables are discussed in more Our linear response functions allow for detail in the following sub-sections. possible delays in consumer response over 4- The conditional expenditure shares sum to week and 24-week intervals, both when the one when the following ‘adding up’ conditions labels were introduced and when they were α α γ γ γ γ removed. It is expected that labelling informa- hold 1 + 2 = 1, 11 = 21 = 0, 12 + 22 = 0, β β η η φ φ φ tion diffuses among consumers over some 1 + 2 = 0, 1 + 2 = 0, 11 + 21 = 0, 12 φ λ λ λ λ λ period of time. It might take some time for + 22 = 0, 11 + 21 = 0, 12 + 22 = 0, 13 + λ λ λ θ θ θ consumers to recognize that such labels have 23 = 0, 14 + 24 = 0, 11 + 21 = 0, 12 + θ θ θ θ θ 22 = 0, 13 + 23 = 0 and 14 + 24 = 0. been placed on specific food products. Their γ The homogeneity conditions are given by 11 avoidance might take even longer to manifest γ γ γ + 12 = 0 and 21 + 22 = 0, and the symme- itself in the market place. Our specification γ γ try conditions imply that 12 = 21. These rep- accounts for such effects and allows for an resent statistically testable hypotheses immediate (1 month) response and a more regarding the theoretical consistency of the gradual (6 months) consumer response, both empirical non-linear conditional expenditure when the labels were placed on the products share system. In the estimation of each prod- and when they were removed. uct category, one share equation is estimated For each of our four food product cate- with non-linear least squares (Greene, 2000). gories, if consumers adversely react to the The parameter estimates for the other share introduction of the label immediately we equation are then recovered using the adding λ < λ > expect 11 0 in w1t and 21 0 in the w2t up conditions. ceteris paribus. If consumers do not change

4 It should be noted that the choice of response variable affected the overall model ﬁt but did not affect the qualitative results reported in this study. Consumer - Chap 03 5/3/04 15:54 Page 29

Purchasing of GM Foods in The Netherlands 29

their consumption patterns immediately after health studies are diffused over long time peri- λ the introduction of the label, we expect 11 = ods (e.g. food additives, the link between cho- λ 21 = 0 in both share equations ceteris lesterol in meat/eggs and heart disease) paribus. Similarly, if consumers adversely (Brown and Schrader, 1990; van Ravenswaay react to the introduction of the label in a grad- and Hoehn, 1991; Wessels et al., 1995). λ < ual fashion, then we expect 12 0 in the w1t Therefore, both short run and long run effects λ > and 22 0 in the w2t ceteris paribus. If con- are accounted for in our model specification. sumers do not change their consumption pat- We expect consumer reaction to be ampli- terns in response to the introduction of the fied during information-augmenting events, λ λ GM labels in the long run, we expect 12 = 22 which raise awareness about biotechnology. = 0 in both share equations ceteris paribus. Gaskell et al. (1999) and Durant et al. (1998) In a similar line of reasoning, if consumers have found that heightened media coverage favourably react to the removal of the label increases awareness of biotechnology. We within a month’s period, then, a priori, we therefore expect that shifts in consumer λ > λ < expect 13 0 in the w1t and 23 0 in the demand for specific product ingredients (e.g. w2t ceteris paribus. If consumers do not react GM versus unlabelled products), if they do exist, to the removal of the label in the short run, to be more distinguishable around heightened λ λ we would expect 13 = 23 = 0 in both share media coverage (Marks et al., 2002). equations ceteris paribus. Similarly, if con- Accordingly, a media variable (mt), which mea- sumers favourably react to the removal of the sures article frequency in key newspapers, is λ > label in the long run, then we expect 14 0 used interactively with the labelling variables λ < in the w1t and 24 0 in the w2t ceteris (on4t,on24t,off4t,off24t) to capture the poten- paribus. If consumers do not react to the tial role of information received by consumers 5 removal of the label in the long run, we from global media sources. That is, mt was λ λ expect 14 = 24 = 0 in both share equations multiplied by the four label variables to create ceteris paribus. four media–label interaction effects that serve as additional shift variables in the AIDS model. Similar to the label hypothesis tests above, if Modelling information on biotechnology consumers decrease their purchases of GM- through media coverage labelled foods in response to increased media It is possible that over a long period of time, coverage in the short run, then a priori we θ < θ > consumers could change their purchasing expect 11 0 in the w1t and 21 0 in the behaviour towards GM foods in response to w2t or unlabelled share equation ceteris relevant new information. Since our study paribus. If consumers do not react to the label θ θ examines consumer behaviour over a 4-year given media coverage, we expect 11 = 21 = period, influences from external information 0 in both share equations ceteris paribus. must be accounted for. Similarly, if over the long run consumers react Over 90% of consumers receive informa- to the label in response to increased media cov- tion about food and biotechnology primarily erage by decreasing their purchases of labelled θ < θ through the popular press and television products we expect 12 0 in the w1t and 22 > (Hoban and Kendall, 1993). Previous studies 0 in the w2t ceteris paribus. If consumers do have found that media information on food not react to information from the media in the θ θ risks (such as food contaminants) can affect long run, we would expect 12 = 22 = 0 in food demand in the short run (Wessels et al., both share equations ceteris paribus. The 1995; Smith et al., 1998; Dahlgran and same line of reasoning applies once the GM Fairchild, 2002). Cumulative, long run effects label is removed and details on relevant para- have also been observed – particularly when meters need not be presented here.

5 We include such an interaction term to account for the fact that the two effects are not independent. In other words, consumers can act (if they choose to do so) on external information provided by the media only when products are labelled. Consumer - Chap 03 5/3/04 15:54 Page 30

30 L. Marks et al.

Empirical results Tables 3.2a–d catalogue the empirical parameter estimates and standard errors of α, Table 3.1 reports basic statistics on the series γ, β, η, φ, λ and θ from equations (1) and (2). used in the analysis. The average expenditure As with any singular system, (n 1) of the n shares for GM-labelled and unlabelled canned system equations are estimated and the para- soup was 3.93% and 96.07%, respectively. meter estimates for the remaining equation Average expenditure shares for GM-labelled are recovered using the adding up conditions and unlabelled frozen processed meat was and other relevant parameter restrictions of 39.03% and 60.97%, respectively. Average the model. The GM-labelled equation in each expenditure shares for GM-labelled and unla- of the four product categories is estimated belled frozen pizza was 6.05% and 93.95%, using non-linear least squares since equation respectively. Average expenditure shares for (2), the unobservable price index, is itself a GM-labelled and unlabelled frozen processed highly non-linear function. The parameter ﬁsh was 2.68% and 97.32%, respectively. estimates for the labelled share equation in Prices, measured in Dutch guilders, were the four product categories are recovered nearly identical between GM-labelled and from the adding up conditions. For each of unlabelled products across three of the four the four models, Linear Approximate AIDS food categories. In the case of frozen pizza, (LA/AIDS) model parameter estimates were the GM-labelled products enjoyed almost 40% employed as an initial feasible solution (i.e. higher prices over non-labelled products, indi- starting values for the unknown parameter cating premium brands. Per capita total cate- estimates) to the non-linear conditional AIDS gory expenditure was also measured in Dutch model (Alston et al., 1994). Moreover, consis- guilders. The population series used in the per tent with the AIDS literature (Buse, 1994), the α capita series varied annually and was obtained parameter 0 in equation (2) was restricted to from Statistics Netherlands. zero in all estimations. In the speciﬁcation

Table 3.1. Descriptive statistics of selected demand system variables.a

Mean SD Minimum Maximum

Expenditure shares GM-labelled soup 0.0393 0.0101 0.0208 0.0840 Unlabelled soup 0.9607 0.0101 0.9160 0.9792 GM-labelled meat 0.3903 0.0831 0.1503 0.5424 Unlabelled meat 0.6097 0.0831 0.4576 0.8497 GM-labelled pizza 0.0605 0.0183 0.0321 0.1417 Unlabelled pizza 0.9395 0.0183 0.8583 0.9679 GM-labelled fish 0.0268 0.0090 0.0143 0.1361 Unlabelled fish 0.9732 0.0090 0.8639 0.9857 Prices GM-labelled soup 3.8530 0.2279 2.8101 4.2559 Unlabelled soup 3.8809 0.3420 2.1747 4.5619 GM-labelled meat 10.8219 0.5128 9.8323 12.0234 Unlabelled meat 10.0623 0.4623 8.3132 11.2424 GM-labelled pizza 14.9805 0.9747 9.5624 16.4487 Unlabelled pizza 10.9417 0.6106 8.5907 12.2628 GM-labelled fish 12.9019 0.6083 10.9048 13.9418 Unlabelled fish 12.6820 1.0975 10.4326 15.1471 Per capita total category expenditure Soup 0.1847 0.0351 0.1168 0.3044 Meat 0.0080 0.0013 0.0058 0.0117 Pizza 0.1995 0.0459 0.0971 0.3263 Fish 0.1717 0.0239 0.1168 0.2842

aBased on 247 weeks of data. Consumer - Chap 03 5/3/04 15:54 Page 31

Purchasing of GM Foods in The Netherlands 31

Table 3.2a. Estimated conditional expenditure share equations: canned soup.

GM-labelled Unlabelleda Parameter Parameter estimate SE estimate

Intercept (α ) 0.0088*** 0.0047 0.9912 Log of price (γ ) GM-labelled 0.1061* 0.0058 0.1061 Unlabelled 0.0193* 0.0045 0.0193 Per capita Real expenditure (β ) 0.0208* 0.0022 0.0208 Linear time trend (η) 0.0001* 0.00002 0.0001 Holidays (φ ) Queen’s Day 0.0002 0.0013 0.0002 Sinterklaas 0.0029** 0.0014 0.0029 Labels (λ ) on4 0.0044 0.0044 0.0044 on24 0.0001 0.0002 0.0001 off4 0.0009 0.0015 0.0009 off24 0.0001 0.0001 0.0001

Media–label interaction (θ ) m•on4 0.0067 0.0053 0.0067 m•on24 0.00004 0.0002 0.00004 m•off4 0.0005 0.0004 0.0005 m•off24 0.00004 0.00005 0.00004 Adjusted R 2 0.8589 – –

aThe parameter estimates in the unlabelled soup share equation are recovered using the adding up restrictions. *, ** and ***denote signifance at the 0.01, 0.05 and 0.10 level, respectively.

testing phase, the two homogeneity restric- of GM-labelled soup price and unlabelled soup γ γ tions and the symmetry condition were price (i.e. 11 and 12) were both statistically rejected and so were not imposed on the two- significant (P < 0.01). From the two adding up γ γ γ γ equation system. However, when imposed, restrictions 11 + 21 = 0 and 12 + 22 = 0, γ γ the qualitative findings of this study remained we found 21 and 22 to be 0.1061 and unchanged. Finally, in each estimated equa- 0.0193, respectively. The parameter esti- β tion, an AR(1) term was used to successfully mate for per capita real expenditure ( 1) and η purge first-order autocorrelation from the the linear time trend ( 1) were also statistically empirical residual series. significant (P < 0.01). The latter indicated an

In the GM-labelled equation (w1t) in the exogenous downward trend in w1t not canned soup model (Table 3.2a), the intercept accounted for by the other observable effects α parameter ( 1) was found to be 0.0088 and controlled for in the AIDS model. Finally, one statistically significant (P < 0.10). From the of the holiday effects, Sinterklaas, was statisti- α α α adding up restriction 1 + 2 = 1, we find 2 cally significant in the model (P < 0.05). equals 0.9912. Although these deviate slightly Since the dependent variables in an AIDS from the average soup expenditure shares model are expenditure shares, not quantities found in Table 3.1, which is very common demanded, the γ parameters do not have a in applied demand analysis, the model fits direct interpretation as an own- or cross-price the underlying data generation process well elasticity. Similarly, the β parameters do not with an adjusted R2 value of 0.8589. The have a direct interpretation as a conditional parameter estimates on the natural logarithm expenditure elasticity (Deaton and Muellbauer, Consumer - Chap 03 5/3/04 15:54 Page 32

32 L. Marks et al.

Table 3.2b. Estimated conditional expenditure share equations: frozen processed meat.

GM-labelled Unlabelleda

Parameter Parameter estimate SE estimate

Intercept (α) 0.1970** 0.0792 0.8030 Log of price (γ ) GM-labelled 0.6770* 0.1787 0.6770 Unlabelled 0.2022* 0.0586 0.2022 Per capita Real expenditure (β ) 0.0457* 0.0152 0.0457 Linear time trend (η) 0.0001 0.0002 0.0001 Holidays (φ ) Queen’s Day 0.0128 0.0088 0.0128 Sinterklaas 0.0004 0.0092 0.0004 Labels (λ ) on4 0.0219 0.0286 0.0219 on24 0.0008 0.0014 0.0008 off4 0.0011 0.0099 0.0011 off24 0.0004 0.0011 0.0004 Media–label interaction (θ ) m•on4 0.0237 0.0349 0.0237 m•on24 0.0008 0.0010 0.0008 m•off4 0.0001 0.0025 0.0001 m•off24 0.0003 0.0003 0.0003 Adjusted R 2 0.8969 – –

aThe parameter estimates in the unlabelled meat share equation are recovered using the adding up restrictions. * and ** denote signiﬁcance at the 0.01 level and 0.05 level, respectively.

1980). Own- and cross-price elasticities are a rithm of prices are taken at their sample means somewhat complex function of the estimated δ and ij is the Kronecker delta. Finally, the condi- and recovered α, γ and β parameters as well tional expenditure elasticity (Ei,X) is given by: as the average expenditure shares and aver- β age natural logarithm of prices. In the case of E =+1 i (5) iX, w Marshallian or uncompensated price elastici- i U where β is defined in equations (1) and (2) and ties (E ij ), we use the expression: 1   2  the expenditure shares are taken at their sam- U =−δγβαγ + − +∑  Eij ij ij i j kjlog p k (3)   =  ple means. wi k 1 In Table 3.3a, the respective elasticities where α, γ and β are defined in equations (1) and (2), expenditure shares and the natural from equations (3), (4) and (5) may be found logarithm of prices are taken at their sample for canned soup. Uncompensated and com- δ pensated own-price elasticities are, as means and ij is the Kronecker delta which equals 1 when i = j and zero otherwise. In the expected a priori, negative indicating usual case of Hicksian or compensated price elastic- downward sloping canned soup demand C equations. For example, a 1% increase in ities (Eij ), we use the expression: 1   2  own price results, on average, in a 3.6915% EC =−δβαγ +γ − +∑ log pw −  ij ij  ij i j kj k j  decrease in quantity demanded of GM-   =  wi k 1 (4) labelled soup. Also noteworthy, the GM- +w j labelled soup demand equation is more elastic where α, γ and β are defined in equations (1) than the unlabelled soup demand equation and (2), expenditure shares and the natural loga- regardless of which elasticity measure was Consumer - Chap 03 5/3/04 15:54 Page 33

Purchasing of GM Foods in The Netherlands 33

Table 3.2c. Estimated conditional expenditure share equations: frozen pizza.

GM-labelled Unlabelleda

Parameter Parameter estimate SE estimate

Intercept (α ) 0.1032* 0.0143 0.8968 Log of price (γ ) GM-labelled 0.1063* 0.0113 0.1063 Unlabelled 0.0912* 0.0152 0.0912 Per capita Real expenditure (β ) 0.0065 0.0069 0.0065 Linear time trend (η) 0.0002* 0.00003 0.0002 Holidays (φ ) Queen’s Day 0.0013 0.0040 0.0013 Sinterklaas 0.0007 0.0043 0.0007 Labels (λ) on4 0.0116 0.0141 0.0116 on24 0.0001 0.0005 0.0001 off4 0.0068 0.0044 0.0068 off24 0.0001 0.0003 0.0001 Media–label interaction (θ ) m•on4 0.0155 0.0164 0.0155 m•on24 0.0003 0.0005 0.0003 m•off4 0.0025** 0.0012 0.0025 m•off24 0.0002 0.0002 0.0002 Adjusted R 2 0.6803 – –

aThe parameter estimates in the unlabelled pizza share equation are recovered using the adding up restrictions. * and ** denote signiﬁcance at the 0.01 level and 0.05 level, respectively.

U < U C < C capita real expenditures for soup grew. For used (i.e. E11 E22 as well as E11 E22). This is sensible given there exist more substitution example, a 1% increase in the per capita real possibilities for unlabelled soups as evidenced expenditure for soup resulted in a 0.4699% in the relationship between average expendi- increase in the quantity demanded of GM- ture shares in Table 3.1. Off diagonal, in the labelled soup. Price and expenditure elastici- uncompensated case, the GM-labelled and ties of this magnitude are commonplace in unlabelled soups have the a priori expected empirical demand studies based on weekly substitute relationship given the positive point-of-purchase scanner data (Cotterill, U U > 1994; Vickner and Davies, 1999). cross-price elasticities (i.e. E12, E21 0). A 1% increase in the GM-labelled soup price results The empirical results for the remaining in a 1.0160% increase in quantity demanded three consumer food product categories very of unlabelled soup. In the compensated case closely parallel those found for canned soup (i.e. in the absence of conditional expenditure and so will be overviewed briefly here. In the C < frozen processed meat product category effects), the peculiar result that E12 0is α γ γ β probably due to shoppers in aggregate pur- (Table 3.2b), the parameters 1, 11, 12 and 1 chasing both GM-labelled and unlabelled were found to be statistically significant (P < α soups on the same purchase occasions, with- 0.01, except in the case of 1 where P < out the intention of using them in a com- 0.05). The empirical model fits the data quite plementary fashion. Finally, the conditional well with an adjusted R2 of 0.90. In fact, this expenditure effects are both positive indi- was the highest of the four adjusted R2 cating the quantity demanded of both GM- values, possibly because the GM-labelled meat labelled and unlabelled soups grew as per products made up, on average, 39% of the Consumer - Chap 03 5/3/04 15:54 Page 34

34 L. Marks et al.

Table 3.2d. Estimated conditional expenditure share equations: frozen processed ﬁsh.

GM-labelled Unlabelleda

Parameter Parameter estimate SE estimate

Intercept (α) 0.0297* 0.0097 0.9703 Log of price (γ) GM-labelled 0.1119* 0.0129 0.1119 Unlabelled 0.0219 0.0142 0.0219 Per capita Real expenditure (β ) 0.0008 0.0050 0.0008 Linear time trend (η) 0.00004** 0.00002 0.00004 Holidays (φ ) Queen’s Day 0.0002 0.0028 0.0002 Sinterklaas 0.0027 0.0032 0.0027 Labels (λ) on40.0023 0.0114 0.0023 on24 0.0002 0.0003 0.0002 off4 0.0003 0.0032 0.0003 off24 0.0001 0.0001 0.0001 Media–label interaction (θ ) m•on4 0.0020 0.0129 0.0020 m•on24 0.00003 0.0004 0.00003 m•off4 0.0003 0.0010 0.0003 m•off24 0.0001 0.0001 0.0001 Adjusted R 2 0.2825 – –

aThe parameter estimates in the unlabelled ﬁsh share equation are recovered using the adding up restrictions. * and ** denote signiﬁcance at the 0.01 and 0.05 level, respectively.

Table 3.3a. Estimated price and expenditure elasticities for canned soup.

GM-labelled soup Unlabelled soup

Uncompensated GM-labelled soup 3.6915 1.0160 Unlabelled soup 0.1101 1.0416 Compensated GM-labelled soup 3.7516 0.4539 Unlabelled soup 0.07165 1.9814 Expenditure 0.4699 1.0217

Table 3.3b. Estimated price and expenditure elasticities for frozen processed meat.

GM-labelled meat Unlabelled meat

Uncompensated GM-labelled meat 2.7146 0.6131 Unlabelled meat 1.0976 1.3925 Compensated GM-labelled meat 3.1506 0.0680 Unlabelled meat 0.7366 1.9565 Expenditure 0.8829 1.0750 Consumer - Chap 03 5/3/04 15:54 Page 35

Purchasing of GM Foods in The Netherlands 35

Table 3.3c. Estimated price and expenditure elasticities for frozen pizza.

GM-labelled pizza Unlabelled pizza

Uncompensated GM-labelled pizza 2.7675 1.4108 Unlabelled pizza 0.1138 1.0909 Compensated GM-labelled pizza 2.8214 0.5728 Unlabelled pizza 0.0529 2.0369 Expenditure 1.1080 0.9930

Table 3.3d. Estimated price and expenditure elasticities for frozen processed ﬁsh.

GM-labelled ﬁsh Unlabelled ﬁsh

Uncompensated GM-labelled fish 5.1773 0.8478 Unlabelled fish 0.1150 0.9767 Compensated GM-labelled fish 5.2033 1.7915 Unlabelled fish 0.0882 1.9507 Expenditure 1.0303 0.9992

expenditure share in that category, and in 1 Consumer response to GM labels week 54% of the market share – far in excess of the other three food categories. The price The most significant result in our empirical and expenditure elasticities for frozen model is the lack of any statistically significant processed meat were reasonable and consis- change in consumer response towards foods tent with theory. with GM labels. Indeed, we cannot detect any In the frozen pizza category (Table 3.2c), immediate or gradual consumer response to α γ γ η the parameters 1, 11, 12, and 1 were the introduction of GM labels or to their found to be statistically significant (P < 0.01). removal. Specifically, none of the other para- The empirical model fits the data adequately meter estimates in the model, such as the with an adjusted R2 of 0.68. The own- and label and media–label interaction variables, cross-price elasticities, both uncompensated were statistically significant (P > 0.10), except • and compensated, as well as expenditure elas- for the parameter estimate on the mt off4t θ ticities appear reasonable. variable ( 13). This parameter was found to be In the frozen processed fish category (Table –0.0025 and statistically significant at the α γ 3.2d), the parameters 1 and 11 were found 0.05 level. However, the result is indubitably to be statistically significant at the 0.01 level, spurious as the economic relationship it indi- η and 1 was found to be significant at the 0.05 cates – namely that the GM-labelled frozen η level. Since 1 was positive it indicates an pizza product is penalized in the first 4 weeks exogenous upward trend in w1t not accounted after the label is removed – is not sensible. for by the other observable effects controlled This result is in stark contrast with the bulk for in the AIDS model. The empirical model of the existing literature that anticipates Dutch fits the data rather poorly with an adjusted R2 – and most other European – consumers to of 0.28. However, most of the own- and actively discriminate against foods with GM cross-price elasticities, both uncompensated labels in the market place, when given the and compensated, as well as expenditure elas- opportunity. Indeed, a February 2002 con- ticities are reasonable. sumer survey in The Netherlands directly Consumer - Chap 03 5/3/04 15:54 Page 36

36 L. Marks et al.

asked consumers how they would respond if Irrespectively of motives, however, the key they noticed GM labels on food products they result is that Dutch consumers, in aggregate, regularly purchase. It concluded that a large did not alter their behaviour towards positively majority would stop purchasing them labelled foods with GM ingredients. If the (Environics International, 2002). Our empiri- results obtained here could be generalized cal results suggest that Dutch consumer pref- across products and markets in Europe, they erences revealed in the market place were would call into question the current European drastically different from stated preferences policy of mandatory labelling of GM foods elicited through surveys.6 and ingredients. External information on GM foods and Protecting consumers’ ‘right to know’ and agrifood biotechnology did not have any sig- the ‘right to choose’ is advanced as the main nificant impact on consumer response either. reason for the current European policy stance. Despite substantial media coverage, no signifi- In principle, there can be little objection to the cant influences could be empirically identified. argument that consumers should be able to exercise such rights. Market transparency is the linchpin of well-functioning markets. Discussion and Concluding Comments However, mandatory labelling is not the only option that would allow consumers a choice. Our results indicate that, in aggregate, Dutch Indeed, given that mandatory labelling systems consumers did not significantly alter their pur- are costly to implement (Kalaitzandonakes et chasing behaviour in the presence of foods al., 2001) costs and benefits associated with positively labelled as containing GM ingredi- such labelling regimes must be carefully ents. Nor did they alter their purchasing weighted in order to decide their optimality behaviour towards such foods after the labels (Giannakas and Fulton, 2002). In this context, were removed nearly 3 years later. the proportion of the consumers that would We do not know why Dutch consumers effectively discriminate between GM and con- did not alter their purchasing patterns in the ventional foods in the market place is a key presence of positive GM labels. Our data do parameter (Giannakas and Fulton, 2002). not allow such insight. It could be that a Indeed, Caswell (1998, 2000) and Giannakas majority of Dutch consumers are more and Fulton (2002) have argued that a volun- accepting of the technology (Hamstra and tary labelling programme may better serve a Smink, 1996; Hoban, 1997; Zechendorf, country where only a minority of the popula- 1998). Or it could be that Dutch consumers tion is interested in separating GM from non- have a high level of trust in their food supply GM foods. Mandatory systems, on the other (Hamstra, 1993) and were therefore less con- hand, may better serve countries where a siz- cerned about purchasing GM foods. It could able percentage of the population would differ- also be that Dutch consumers were reassured entiate between genetically modified and by the brand identity of each of the labelled conventional foods in the market place. products (Noussair et al., 2004). In other Understanding whether a majority of con- words, it is possible that a majority of consumers, irrespective of motives, would use sumers read the labels, understood them and GM labels to discriminate against relevant kept on purchasing them regardless. products in the market is essential for effective Alternatively, it could be that Dutch con- policy decisions. Our results are national in sumers did not see or read the labels. scope, cover multiple product categories, and

6 While our empirical results differ from results obtained through surveys and experimental auctions, they are consistent with the limited empirical evidence that is available on revealed consumer preferences and market behaviour. The vast majority of US consumers purchased milk from rBST-treated cows despite stated preferences for the opposite (Aldrich and Blisard, 1998). Similarly, anecdotal evidence from the UK indicates that a GM tomato paste offered in UK stores up until 1999 apparently outsold competing non-GM brands (Nunn, 2000). Consumer - Chap 03 5/3/04 15:54 Page 37

Purchasing of GM Foods in The Netherlands 37

suggest that at least in The Netherlands such across Europe. At a minimum, however, they majority response could not be detected raise questions about the existing conventional despite the fact that consumers could ‘vote wisdom on potential consumer behaviour, with their wallets’ against GM foods in the which is based on stated preferences, and beg presence of alternatives. It is unclear to what for further research that focuses on revealed extent these results could be generalized consumer preferences and actual behaviour.

References

Aldrich, L. and Blisard, N. (1998) Consumer Acceptance of Biotechnology. Lessons from the rBST Experience Agricultural Information Bulletin No. 7417–01, Economic Research Service, US Department of Agriculture, Washington, DC. Alston, J.M., Foster, K.A. and Green, R.D. (1994) Estimating elasticities with the linear approximate almost ideal demand system: some Monte Carlo results. Review of Economics and Statistics 76, 351–356. Boccaletti S. and Moro, D. (2000) Consumer willingness to pay for GM food products in Italy. AgBioForum 3(4), 259–267. Brown, D.J. and Schrader, L.F. (1990) Cholesterol information and shell egg consumption. American Journal of Agricultural Economics 72(3), 548–555. Burton, M. and Pearse, D. (2002) Consumer attitudes towards genetic modification, functional foods and microorganisms: a choice modeling experiment for beer. AgBioForum 5, 51–58. Buhr, B.L., Hayes, D.J., Shogren, J.F. and Kliebenstein, J.B. (1993) Valuing ambiguity: the case of genetically engineered growth enhancers. Journal of Agricultural and Resource Economics 18, 175–184. Buse, A. (1994) Evaluating the linearized Almost Ideal Demand System. American Journal of Agricultural Economics 76, 781–793. Caswell, J.A. (1998) Should use of genetically modified organisms be labeled? AgBioForum 1(1), 22–24. Caswell, J.A. (2000) Labelling policy for GMOs: to each his own? AgBioForum 3(3), 53–57. Cotterill, R.W. (1994) Scanner data: new opportunities for demand and competitive strategy analysis. Agricultural and Resource Economics Review 23, 125–139. Dahlgran, R.A. and Fairchild, D.G. (2002) The demand impacts of chicken contamination publicity – a case study. Agribusiness 18(4), 459–474. Deaton, A. and Muellbauer, J. (1980) An Almost Ideal Demand System. American Economic Review 70, 312–326. Douthitt, R., Zepeda, L. and Grobe, D. (1996) Comparison of National and Poor Households: Results of a Survey of Consumer Knowledge and Risk Perception of Food-related Biotechnologies. Special Report No. 68, Institute for Research on Poverty, University of Wisconsin-Madison. Durant, J., Bauer, M.W. and Gaskell, G. (eds) (1998) Biotechnology in the Public Sphere: a European Source Book. Science Museum Press, London. Environics International Ltd. (2002) CIAA European Food Survey. Available on the World Wide Web: http://www.ciaa.be/ciaa_summit/pages/hetherington,pdf. European Commission (EC) (1997) The Europeans and modern biotechnology. Eurobarometer 46(1). European Commission, Brussels. European Commission (EC) (2000) The Europeans and modern biotechnology. Eurobarometer 52(1). European Commission, Brussels. Gaskell, G., Bauer, M.W., Durant, J. and Allum, N.C. (1999) Worlds apart? The reception of genetically modified foods in Europe and the U.S. Science 285, 384–387. Giannakas, K. and Fulton, M. (2002) Consumption effects of genetic modification. What if consumers are right? Agricultural Economics 27, 97–109. Greene, W.H. (2000) Econometric Analysis, 4th edn. Prentice Hall, Upper Saddle River, New Jersey. Hamstra, A.M. (1993) Consumer Acceptance of Food Biotechnology: The Relation Between Product Evaluation and Acceptance. Research Report 137. SWOKA, Leiden. Hamstra, A.M. and Smink, C. (1996) Consumers and biotechnology in the Netherlands. British Food Journal 98(4), 34–38. Hillers, V.N. and Lowik, M.R.H. (1998) Incorporation of consumer interests in regulation of novel foods produced with biotechnology: what can be learned from the Netherlands? Journal of Nutrition Education January/February, 2–7. Consumer - Chap 03 5/3/04 15:54 Page 38

38 L. Marks et al.

Hoban, T.J. (1996) Trends in consumer attitudes about biotechnology. Journal of Food Distribution Research 27, 1–10. Hoban, T.J. (1997) Consumer acceptance of biotechnology: an International perspective. Nature Biotechnology 15, 232–234. Hoban, T.J. (1998). Trends in consumer attitudes about agricultural biotechnology. AgBioForum 1(1), 3–7. Hoban, T.J. and Burkhardt, J. (1991) Biotechnology control of growth and product quality in meat production: implications and acceptability. In: Van der Wal, P. (ed.) Proceedings of Determinants of Public Acceptance in Meat and Milk Production: North America Conference. Wageningen Agricultural University, The Netherlands. Hoban, T.J. and Kendall, P.A. (1993) Consumer Attitudes about Food Biotechnology. North Carolina Cooperative Extension Service, Raleigh, North Carolina. Huffman, W.E., Shogren, J.F., Rousu, M. and Tegene, A. (2001) The value to consumers of GM food labels in a market with asymmetric information: evidence from experimental auctions. Paper presented at the 5th International Consortium on Agricultural Biotechnology Research (ICABR) Meetings, Ravello, Italy. Huffman, W.E., Rousu, M., Shogren, J.F., and Tegene, A. (2002) Should the United States initiate a mandatory labeling policy for genetically modified foods? Paper presented at the 6th International Consortium on Agricultural Biotechnology Research (ICABR) Meetings, Ravello, Italy. James, J., Parker, T., Fleischer, S. and Orzolek, M. (2002) Consumer acceptance of GMOs revealed: a market experiment with Bt sweet corn. Paper presented at the Northeastern Agricultural and Resource Economics Association Meetings, Camp Hill, Pennsylvania, USA. Kalaitzandonakes, N. (2000) Agrobiotechnology and competitiveness. American Journal of Agricultural Economics 82, 1224–1233. Kalaitzandonakes, N., Maltsbarger, R. and Barnes, J. (2001) Global identity preservation costs in agricultural supply chains. Canadian Journal of Agricultural Economics 49, 605–615. Kahneman, D. and Tversky, A. (1984) Choices, values, and frames. American Psychologist 39, 341–350. Keisel, K., Buschena, D. and Smith, V. (2002) Consumer acceptance and labeling of biotech in food products: a study of fluid milk demand. Paper presented at the 6th International Consortium on Agricultural Biotechnology Research (ICABR) Meetings, Ravello, Italy. Lusk, J.L., Daniel, S., Mark, D. and Lusk, C. (2001) Alternative calibration and auction institutions for predicting consumer willingness to pay for non-genetically modified corn chips. Journal of Agricultural and Resource Economics 26, 40–57. Lusk, J.L., Roosen, J., and Fox, J.A. (2003) Demand of beef from cattle administered growth hormones or fed genetically modified corn: a comparison of consumers in France, Germany, the United Kingdom, and the United States. American Journal of Agricultural Economics 85, 16–29. Marks, L.A., Kalaitzandonakes, N. and Zakharova, L. (2002) On the media roller-coaster will GM foods fin- ish the ride? Choices Spring, 6–10. Mathios, A.D. (2000) The impact of mandatory disclosure laws on product choices: an analysis of the salad dressing market. Journal of Law and Economics 43, 651–675. McFadden, D. (1973) Conditional logit analysis of qualitative choice behavior. In: Zarembka, P. (ed.) Frontiers in Econometrics. Academic Press, New York, pp. 105–142. Menkhaus, D.J., Borden, G.W., Whipple, G.D., Hoffman, E. and Field, R.A. (1992) An experimental application of laboratory experimental auctions in marketing research. Journal of Agricultural and Resource Economics 17, 44–55. Moon, W. and Balasubramanian, S.K. (2001) Public perceptions and willingness-to-pay a premium for nonGMO foods in the US and UK. AgBioForum 4(3–4), 221–231. Moschini, G. (1995) Units of measurement and the stone price index in demand system estimation. American Journal of Agricultural Economics 77, 63–68. Noussair, C., Robin, S. and Ruffieux, B. (2004) Do consumers really refuse to buy genetically modified foods? The Economic Journal 114, 102–120. Nunn, J. (2000) What lies behind the GM label on UK foods. AgBioForum 3, 250–254. Runge, C.F. and Jackson, L.A. (2000) Negative labeling of genetically modified organisms (GMOs): the experience of rBST. AgBioForum 3, 58–62. Shogren, J.F., Fox, J.A., Hayes, D.J. and Roosen, J. (1999) Observed choices for food safety in retail, survey, and auction markets. American Journal of Agricultural Economics 81, 1192–1199. Consumer - Chap 03 8/3/04 10:22 Page 39

Purchasing of GM Foods in The Netherlands 39

Smith, M.E., van Ravenswaay, E.O. and Thompson, S.R. (1988) Sales loss determination in food contamination incidents: an application to milk bans in Hawaii. American Journal of Agricultural Economics 70, 513–520. Sterngold, A., Warland, R. and Herrman, R. (1994) Do surveys overstate public concerns? Public Opinion Quarterly 58, 255–263. Swartz, D.G. and Strand Jr., I.E. (1981) Avoidance costs associated with imperfect information: the case of Kepone. Land Economics 57(2), 139–150. Teisl, M.F., Roe, B. and Hicks, R.L. (2002) Can eco-labels tune a market? Evidence from dolphin-safe labeling. Journal of Environmental Economics and Management 43, 339–359. Tolley, G.S. and Randall, A. (1983) Establishing and Valuing the Effects of Improved Visibility in the Eastern United States. Report of the US Environmental Protection Agency, Washington, DC. van Ravenswaay, E.O. and Hoehn, J.P. (1991) The impact of health risk information on food demand: a case study of alar in apples. In: Caswell, J.A. (ed.) Economics of Food Safety. Elsevier Science Publishing Co., New York, pp.155–174. Vickner, S.S. and Davies, S.P. (1999) Estimating market power and pricing conduct in a product-differentiated oligopoly: the case of the domestic spaghetti sauce industry. Journal of Agricultural and Applied Economics 31, 1–13. Wessels, C.R., Miller, C.J. and Brooks, P.M. (1995) Toxic algae contamination and demand for shellﬁsh: a case study of demand for mussels in Montreal. Marine Resource Economics 10, 143–159. Zechendorf, B. (1998) Agricultural biotechnology: why do Europeans have difﬁculty accepting it? AgBioForum 1(1), 8–13. Consumer - Chap 03 5/3/04 15:54 Page 40 Consumer - Chap 04 5/3/04 15:55 Page 41

4 The Welfare Effects of Implementing Mandatory GM Labelling in the USA1

Wallace E. Huffman,1 Matthew Rousu,2 Jason F. Shogren3 and Abebayehu Tegene4 1Department of Economics, Iowa State University, Ames, IA 50011, USA; 2RTI International, 3040 Cornwallis Road, Research Triangle Park, NC 27709, USA; 3Department of Economics and Finance, University of Wyoming, Laramie, WY 82070, USA; 4Food and Rural Economics Division, Economic Research Service, US Department of Agriculture, Washington, DC 20036, USA

Introduction nomic research has examined the merits and pitfalls of a new regulation that requires manda- Although genetic engineering is a promising tory labelling for GM foods in the USA. tool for crop varietal development, genetically This chapter examines the potential wel- modified foods continue to be controversial. fare effects of imposing a mandatory GM- Many groups that oppose these new goods are labelling policy in the USA. We first discuss supporting a public policy of mandatory when a mandatory labelling policy is likely to labelling for GM content. Debate, however, con- benefit consumers. We then describe an tinues over whether the USA should impose a experimental auction designed to provide data mandatory labelling policy for genetically modi- that are needed to test whether consumers fied (GM) foods. Groups that favour a manda- will benefit from a mandatory GM-labelling tory labelling policy for GM foods include policy. For a sample of adult consumers living Greenpeace International (1997) Friends of the in two major Midwestern cities, our results do Earth (2001), and the Consumers Union not contradict the hypothesis that consumers (Consumer Reports, 1999). Groups opposing interpret voluntary and mandatory market sig- mandatory GM labels include the Council for nals identically. These findings suggest that it Biotechnology Information (2001) and the US would be more efficient or welfare improving Food and Drug Administration (FDA) (2001). for the USA to continue its voluntary labelling This contentious issue has engaged debate from policy and resist calls for new regulations that all sides of the spectrum, yet only modest eco- mandate labelling of GM foods.

1 The authors gratefully acknowledge assistance from Daniel Monchuk and Terrance Hurley in conducting the auctions and assistance from Monsanto in providing some of the products used in the experiment. Authors received helpful comments on the chapter from participants at the ICABR Conference, Ravello, Italy, July 2002. This work was supported through a grant from the US Department of Agriculture Cooperative State Research, Education, and Extension Service, under Agreement 00-52100-9617 and from the US Department of Agriculture, Economic Research Service, under Agreement 43-3AEL-8- 80125, and by the Iowa Agriculture and Home Economics Experiment Station. Views presented in this chapter are the authors and do not represent those of ERS or USDA. © CAB International 2004. Consumer Acceptance of Genetically Modiﬁed Food (eds R.E. Evenson and V. Santaniello) 41 Consumer - Chap 04 5/3/04 15:55 Page 42

42 W.E. Huffman et al.

Background on Labels tries around the world have mandatory labelling policies for GM foods, including Caswell (1998, 2000) emphasizes that the Australia, Japan and New Zealand. For a list of potential GM labelling policies is large, detailed review of labelling policies, see and includes mandatory labelling of GM foods Rousu and Huffman (2001) or Phillips and voluntary labelling of GM foods and bans on McNeill (2000). all labelling. An informed decision about Some groups think that mandating GM labelling policies for GM foods should only be labels would improve a society’s welfare. made after a careful benefit–cost analysis. Many environmental and consumer advo- Caswell points out that a voluntary labelling cacy groups call for mandatory labelling of programme is likely to be a better policy GM foods, which they believe benefits option for a country that has only a small consumers (Greenpeace International, segment of the population that is concerned 1997; Friends of the Earth, 2001; about GM foods, but a mandatory labelling Consumer Reports, 1999). Greenpeace system is likely to be the best policy option in and Friends of the Earth both advocate countries where a large share of the popula- labels on GM foods to give consumers the tion wants to know if their food is GM. A opportunity to choose whether to con- model by Kirchhoff and Zago (2001) reached sume GM foods. Other benefits of labelling a similar conclusion – voluntary GM-labelling are that labels make it easier to find infor- policies may be better for a country that has mation on food products, can increase more consumers who are concerned with consumer information and can improve cost savings, while mandatory GM-labelling product design. policies may be better for more GM-averse Relatively few estimates of the costs of consumers. GM food labelling exist. The The USA does not require mandatory accounting/consulting firm KPMG was labelling for most GM foods. In January commissioned for a study in Australia and 2001, the FDA issued a ‘Guidance for New Zealand to examine the costs of comply- Industry’ statement for labelling GM products, ing with new labelling laws. They estimated which stated that the only GM foods that need that the costs of the labelling laws could to be labelled are ones that have different mean an increase in consumer prices from characteristics from their non-GM versions. 0.5 to 15%, and that firms could also receive Labelling is not required for any other GM lower profits (Phillips and Foster, 2000). Even foods, but firms in the USA do have the though they commissioned the study, the option of voluntarily indicating whether their Australia New Zealand Food Authority (2001) food is GM. Canada also has a similar volun- disregarded KPMG’s study, citing two flaws. tary labelling policy. Whether this council had legitimate problems The European Union (EU) requires that with the study or were responding to political all foods have the label ‘genetically modified’ pressure, we do not know. Smyth and if any ingredient in the food is at least 1% Phillips (2002) estimated that a voluntary GM. The European Parliament voted for identity-preserved production and marketing stricter regulations in early 2001.2 The new system in Canada cost from 13 to 15% regulations call for stricter labelling and mon- during 1995–1996. In a related study, itoring of GM products and allow for the Wilson and Dahl (2002) estimate that identity tracing of GM products through the food preservation costs would be $3.50 per bushel chain (CNN, 2001). The EU standards are for GM wheat, assuming a 1% tolerance the minimum standards that member coun- level. The Philippine Chamber of Food tries must adhere to, although countries can Manufacturers warned that mandatory GM have stricter standards. Several other coun- food labels would increase production costs

2 New regulations passed in 2002 will move this threshold down to 0.5% before the product must be labelled as GM. Consumer - Chap 04 5/3/04 15:55 Page 43

Effects of Mandatory GM Labelling 43

by 15%, and that the increased costs would Experimental Design be passed on to consumers (AgBiotech Reporter, 2001). One issue seems apparent: We designed two experimental auction mar- implementing a labelling policy for GM foods kets, one emulating a market that has manda- is costly, even if the exact magnitude of the tory labelling in place and another which costs is unknown. emulates a market that has voluntary labelling in place. We then test for similarity of consumer bids for three different food products – When Would Consumers Benefit from a vegetable oil, tortilla chips and Russet potatoes. Mandatory Labelling Policy? Experimental units are randomly assigned labelling treatments. Some consumers bid on With asymmetric information between food foods with positive GM labels – the labels that suppliers and consumers, consumers regu- would arise in a mandatory labelling regime; larly must read signals about product quality. others bid on food with negative GM labels – For brand-name products, consumer pur- the labels that would arise in a voluntary chases are higher for some brands than oth- labelling regime.4 If bidding behaviour for GM ers, and consumers are frequently willing to and GM-free food differs across the two mar- try a product based on external signals about kets, this would indicate that the mandatory the product (e.g. packaging, labelling, adver- market presents different signals than a volun- tisements, etc.). The question we examine is tary market. If bidding behaviour does not dif- how would mandatory labelling of GM foods fer, then we would find no evidence that a help consumers in purchasing food mandatory labelling policy assists consumers in products?3 accomplishing their main objective – informing The key benefit that a mandatory labelling consumers. However, finding a change in bid- policy could have is if it were to help con- ding behaviour is a necessary but not a suffi- sumers distinguish genetically modified from cient condition for a mandatory labelling policy non-genetically modified food products. This to be welfare-improving. is why many groups call for mandatory The experimental design consisted of four labelling of GM foods (e.g. see Greenpeace biotech information labelling treatments with International, 1997). But without mandatory each treatment replicated at least four times. labelling there are still both GM and non-GM The treatments were randomly assigned to food products sold – so an ideal test of ten experimental units, each consisting of whether consumer welfare would improve 13–16 consumers drawn from the households from a mandatory labelling policy is to test if a of two major urban areas and who were paid market that has mandatory GM labelling to participate. makes it easier for consumers to distinguish We now describe the four elements in our GM food products from non-GM food prod- GM-labelling experiments – the GM food, the ucts (relative to voluntary GM labelling). The auction mechanism, the experimental units next section outlines how we test this using and the specific steps in the experiment, which experimental auctions. includes the detailed information on the labels.

3 In this chapter we present the intuition behind when a mandatory labelling would benefit consumers, and that a voluntary labelling policy is more efficient than a mandatory labelling policy if consumers can distinguish between GM and non-GM foods identically in either market. A model which derives this for- mally is presented in our technical counterpart to this chapter (Huffman et al., 2002). 4 These experimental markets were chosen to emulate the mandatory and voluntary GM-labelling regimes currently in place throughout the world. Our mandatory regime reflects the labels consumers might find in Europe, where foods that are GM must be labelled as such. Our voluntary labelling regime captures the labels that consumers might see in the USA, where food manufacturers can label their products as non- genetically engineered if they choose. We do not examine several other potential but currently non- implemented labelling policies, including a mandatory labelling policy that requires all non-GM foods to be labelled or a policy that requires every food product in a market to be labelled as GM or non-GM. Consumer - Chap 04 5/3/04 15:55 Page 44

44 W.E. Huffman et al.

The food products and auction participate in a group session that related ‘to mechanism how people select food and household products’, and they were informed that the session Participants in our auction bid on three unre- would last about 90 minutes.7 They were also lated food items: a 32-ounce bottle of veg- told that at the end of the session each partici- etable oil,5 a 16-ounce bag of tortilla chips pant would receive $40 in cash for his/her (made from yellow maize) and a 5-pound bag time. From the initial sample of usable ran- of Russet potatoes. They bid on these items domly selected numbers, the percentage of using the random nth-price auction. people who accepted the offer to participate We chose the random nth-price auction for and then showed up at the auction was our GM food experiments because it is approximately 19%. designed to engage both the on- and off-the- Our total sample size of participants is margin bidders (see Shogren et al., 2001).6 142, and Table 4.1 summarizes the character- This is an aid in identifying the whole demand istics of the auction participants: 60% are curve (rather than a short segment) for a new female, mean age is 51 years and mean good. The random nth-price works as follows. household income is $51,600. Each of k bidders submits a bid for one unit of a good; then each of the bids is rank-ordered from highest to lowest. The auction monitor Sequence of steps in the experiments then selects a random number – the n in the nth-price auction, which is drawn from a uni- Figure 4.1 shows the ten steps in each experi- form distribution between 2 and k, and the mental unit. In Step 1 when participants auction monitor sells one unit of the good to arrived at the experiment, they signed a con- each of the n 1 highest bidders at the nth- sent form agreeing to participate in the auc- price. For instance, if the monitor randomly tion. After they signed this form, they were selects n = 4, the three highest bidders each given $40 for participating and an ID number purchase one unit of the good priced at the to preserve their anonymity. The treatments fourth highest bid. Ex ante, bidders who have were randomly assigned to each experimental low or moderate valuations now have a non- unit, so the observed and unobserved charac- trivial chance to buy the good because the teristics of observations are uncorrelated with price is determined randomly. This auction the treatments. The participants then read brief increases the probability that insincere bidding instructions and ﬁlled out a questionnaire. The will be costly. questionnaire was purposefully given to con- Auctions were planned and conducted in sumers before the experiment to elicit demo- two Midwestern US cities: Des Moines, Iowa, graphic information and to capture consumers’ and St Paul, Minnesota, in 2001. Consumers prior perception of GM foods before bidding, were contacted through a random digit which allowed us to compare their prior beliefs dialling method and were asked if they would to their posterior beliefs after the experiment.

5 For the oil, soybean oil was used for the mandatory labelling trials, and canola oil was used for the voluntary/labelling trials. The soybean oil was initially used in the April experiments. We then tried to purchase non-GM soybean oil in 32-ounce bottles and were unsuccessful. The bids for the vegetable oil follow the same trend as the other products and are discussed in the results section. The other products (and packaging) were absolutely identical, except for the presence or absence of genetic modiﬁcation. 6 The auction combines elements of two classic demand-revealing mechanisms: the Vickrey (1961) auction and the Becker–DeGroot–Marschak (1964) random pricing mechanism. The key characteristic of the random nth price auction is a random but endogenously determined market-clearing price. Randomness is used to give all participants a positive probability of being a purchaser of the auctioned good; the endogenous price ensures that the market-clearing price is related to the bidders’ private values. 7 We considered the possibility that the demand for GM foods may change over the 8 months between auctions, so we replicated two experimental units, using the exact same procedures. We found no evidence that willingness to pay for GM-labelled foods had changed over time. Consumer - Chap 04 5/3/04 15:55 Page 45

Effects of Mandatory GM Labelling 45

Table 4.1. Characteristics of the auction participants.

Variable Deﬁnition Mean SD

Gender 1 if female 0.60 0.49 Age The participant’s age 51.40 18.1 Married 1 if the individual is married 0.65 0.48 Education Years of schooling 14.74 2.36 Household Number of people in participant’s household 2.56 1.49 Income The household’s income level (in thousands) 51.60 33.40 White 1 if participant is white 0.92 0.27 Read_L 1 if never reads labels before a new food purchase 0.02 0.14 1 if rarely reads labels before a new food purchase 0.11 0.32 1 if sometimes reads labels before a new food purchase 0.32 0.47 1 if often reads labels before a new food purchase 0.36 0.48 1 if always reads labels before a new food purchase 0.18 0.39

In Step 2, participants were given detailed After the two practice auction rounds were instructions (both oral and written) about how completed, the binding round and the binding the random nth-price auction works, including nth-prices were revealed in Step 5. All bid an example written on the blackboard. After prices were written on the blackboard, and the participants learned about the auction, a the nth-price was circled for each of the three short quiz was given to them to ensure that products. Participants could see immediately everyone understood how the auction what items they won, and the price they worked. All experimental instructions are would pay. The participants were told that the available from the authors on request. exchange of money for goods was in another Step 3 was the first practice round of bid- room nearby and would take place after the ding, in which participants bid (in a real auc- entire experiment was completed. tion) on a brand-name candy bar. The In Step 6, participants received one of two participants were all asked to examine the potential info-packets that provided non-food product and then to place a bid on the candy label information about biotechnology (for a bar. The bids were collected and the first detailed look at how information affected the round of practice bidding was over. demand for foods labelled as GM, see Rousu Throughout the auctions, when the partici- et al., 2002). These info-packets were pro- pants were bidding on items in a round, they duced as follows. We created three informa- had no indication of what other items they tion sources: (i) the industry perspective – a may be bidding on in future rounds. collection of statements and information on Step 4 was the second practice round of genetic modification provided by a group of bidding, and in this round the participants leading biotechnology companies, including bid separately on three different items. The Monsanto and Syngenta; (ii) the environmen- products were the same brand-name candy tal group perspective – a collection of state- bar, a deck of playing cards and a box of ments and information on genetic modification pens. Participants knew that only one of the from Greenpeace, a leading environmental two rounds would be chosen at random to group; and (iii) the independent third-party be binding, which prevented anyone from perspective – a statement on genetic modifica- taking home more than one unit of any tion approved by a third-party group, consist- product. By using only one binding round, ing of a variety of people knowledgeable about we avoided problems of demand reduction GM goods, including scientists, professionals, that can occur in multi-unit auctions (List and religious leaders and academics, who do not Lucking-Reiley, 2000). The consumers first have a financial stake in GM foods. We limited examined the three products and then sub- each information source to one full page, mitted their bids. organized into five categories: general infor- Consumer - Chap 04 5/3/04 15:55 Page 46

46 W.E. Huffman et al.

Step 1 Step 2 Step 3 Completes consent form and questionnaire, nth-price auction is Candy bar auction receives $40 and ID explained number

Step 4

Auction of a candy bar, a deck of cards and a box of pens

Step 5 Step 6

Binding practice round Both pro- and and binding nth-prices anti-biotechnology are revealed

Both pro- and anti-biotechnology and third-party information

Step 7 Step 8

First round of bidding on Second round of bidding food products on food products

Step 9 Step 10

Binding food round and Post-auction binding nth-prices are questionnaire, winning revealed people purchase goods

Fig. 4.1. Steps in the experiment.

mation, scientific impact, human impact, biotechnology, anti-biotechnology, and inde- financial impact and environmental impact. pendently verifiable.8 These info-packets were The information sheets are available in Rousu then randomized among all ten experimental et al. (2002) or by request from the authors. units, with each info-packet going to four These information sources were then ran- experimental units. By giving all participants domized to create the two info-packets: (i) both both positive and negative information on GM pro- and anti-biotechnology and (ii) pro- foods, and by giving some participants a third-

8 The order of the positive information and negative information was, also, randomized across consumers. Participants who received the third-party, veriﬁable information always received it after the other information sources. Consumer - Chap 04 5/3/04 15:55 Page 47

Effects of Mandatory GM Labelling 47

party perspective on GM foods, we could told them they were next going to look at determine the willingness to pay for individuals another group of food items. Table 4.2 sum- who received all perspectives on the GM food marizes the four treatments. debate. Step 8 had participants examine the same Once we distributed the appropriate info- three food products, each with a different label packet to the participants in a given unit, we from Round 1.10 Again the participants exam- then conducted two auction rounds. The ined the products and bid on the three products rounds were differentiated by the food label – separately. The bids were then collected from all either the food had a standard food label or a of the individuals. In contrast to early experi- label that indicated the status of genetic modi- mental auction work using repeated fication (e.g. see Fig. 4.2).9 In one round (which could be Round 1 or 2 depending on the experimental unit), participants were bidding on the three food products each with the standard food label. We made these labels as Vegetable Oil plain as possible to avoid any influence on the bids from the label design. In the other round, Net weight 32 oz. participants were bidding on the same three food products with either a GM label or a non-GM label. The GM and non-GM labels differed from the standard label only by the inclusion of one extra sentence. The GM label said ‘This product is made using genetic modification (GM)’, while the non-GM label said Vegetable Oil ‘This product is made without genetic modification’. For each experimental unit, partici- Net weight 32 oz. pants knew that only one round would be chosen as the binding round that determined auction winners. This product is made without In Step 7, participants bid on three differ- using genetic modification ent food products: a bottle of vegetable oil, a bag of tortilla chips and a bag of potatoes, either with the standard label or the label indicating the product’s GM status. Six groups bid on foods with plain labels and foods with Vegetable Oil labels saying ‘made using genetic modification (GM)’. Four groups bid on foods with plain Net weight 32 oz. labels and foods with labels saying ‘made without using genetic modification’. The participants were instructed to examine the three This product is made using genetic modification (GM) products and then to write down their sealed bid for each of the three goods. Participants bid on each good separately. The monitor Fig. 4.2. The three types of labels used for the then collected the bids from the people and vegetable oil.

9 Note that our labels are clearly displayed on the front of the package, where consumers would see them. See Noussair et al. (2002) for evidence of how consumers are not always likely to read food labels that are on the back of packages. 10 We randomize the order the participants were presented the food products across groups. The null hypothesis that the round the consumer bid on foods led to the same bids could not be rejected at a 5% level for any of the three goods under both a t-test and a Wilcoxon rank-sum test. Hence, the order in which consumers saw the items did not appear to matter. Consumer - Chap 04 5/3/04 15:55 Page 48

48 W.E. Huffman et al.

Table 4.2. Information and labelling given to the four treatments.

Number or trials Treatment Labelling regime type Third party per treatment

1 Voluntary regime No 2 2 Voluntary regime Yes 2 3 Mandatory regime No 4 4 Mandatory regime Yes 2

trials, this auction used only two rounds to avoid mean and median bids in the two markets any chance of afﬁliation of values and changes are reported in Table 4.3. Eighty-six partici- in willingness to pay due to the posted-market pants were in treatments that bid on the behaviour of other bidders (see List and plain-labelled and GM-labelled food products Shogren, 1999; Knetsch et al., 2001). (the mandatory GM-label market), and 56 Step 9 selected the binding round, and the participants were in treatments that bid on binding random nth-prices for the three the plain-labelled and non-GM-labelled food goods. The winners were notiﬁed. In Step 10, products (the voluntary GM-label market). each participant was asked to complete a For the participants who bid on the GM- brief post-auction questionnaire, and then the labelled and plain-labelled foods, consumers monitors dismissed the participants who did discounted the GM-labelled oil by an average not win. The monitors and the winners then of 11 cents, the GM-labelled tortilla chips by exchanged money for goods, and the auction 8 cents and the GM-labelled potatoes by 8 winners were also dismissed. cents. The participants who bid on the plain- labelled food and the non-GM-labelled food discounted the plain-labelled oil by an aver- Data and Results age of 4 cents, the plain-labelled tortilla chips by 7 cents and the plain-labelled pota- The statistical analysis of our experimental toes by 9 cents. data supports the hypothesis that consumers Our main goal is to determine whether read similar signals in the two markets. The consumers can accurately decipher which

Table 4.3. Mean bids: markets with mandatory and voluntary labels.

n Mean bid SD Median Minimum Maximum

Mean bids for the mandatory GM-labelling market GM oil 86 0.63 0.65 0.50 0 2.75 Oil 86 0.74 0.75 0.50 0 3.29 GM tortilla chips 86 0.61 0.70 0.43 0 3.25 Tortilla chips 86 0.69 0.72 0.50 0 2.89 GM potatoes 86 0.59 0.54 0.50 0 2.00 Potatoes 86 0.67 0.54 0.50 0 2.25 Mean bids for the voluntary GM-labelling market Non-GM oil 56 0.80 0.80 0.50 0 4.75 Oil 56 0.76 0.68 0.50 0 3.00 Non-GM tortilla chips 56 0.75 0.81 0.50 0 4.00 Tortilla chips 56 0.68 0.77 0.50 0 4.00 Non-GM potatoes 56 0.84 0.75 0.75 0 4.00 Potatoes 56 0.75 0.70 0.68 0 4.00 Consumer - Chap 04 5/3/04 15:55 Page 49

Effects of Mandatory GM Labelling 49

food is GM irrespective of the labelling Consumers discounted perceived GM treatment. The size of the discount for the food the same, irrespective of whether the perceived GM food provides evidence market had mandatory or voluntary GM about consumers’ perception of the signals labelling. This result provides evidence from the two labelling regimes. We tested that consumers receive the same signals null hypotheses that consumers did not dis- under either regime. By not rejecting the count the perceived GM food in the two thesis that consumers know GM from non- markets differently. Table 4.4 provides GM food regardless of the labelling regimes, these results. The first column shows the we have no evidence that the necessary difference in bids in the mandatory condition of consumers reading signals dif- labelling trials; the second column shows ferently in a mandatory GM-labelling policy the difference in bids in the voluntary than in a voluntary GM-labelling policy is labelling trials. The third column is the dif- met. Without speculating beyond the reach ference between these columns. The of the laboratory, this finding supports those absolute difference is an average of 7 cents who believe the USA has been prudent in for vegetable oil, 1 cent for the tortilla avoiding calls to initiate a mandatory GM- chips and 1 cent for the potatoes. At the labelling policy. 10% significance level, the tests show that one cannot reject the null hypothesis that the difference in bids is zero for any of the Conclusion three food products.11 Although none of the differences are statistically significant, GM food labelling remains a controversial at first glance it is curious that the mean and an important issue in the USA. Some discount under mandatory and voluntary groups have called for mandatory labelling of labelling regimes is virtually identical for GM foods, but others want to keep labelling the tortilla chips and potatoes, yet it is con- voluntary. The benefit of mandatory labelling siderably larger for the vegetable oil. A that is cited by its supporters is that it will possible explanation for the vegetable oil help consumers to choose between GM and having an average of 7 cents difference is non-GM food products. We designed two the fact that we used two different types of experimental auction markets, one emulated vegetable oil.12 a market that had a mandatory labelling pol-

Table 4.4. t-Test of null hypothesis that differences in bid differences are equal under two labelling regimes.

Difference in bids for Difference in bids for the plain-labelled GM and plain labelled – and non-GM – mandatory regime voluntary regime (n = 86) (n = 56) Difference t-Test statistic

Oil 0.11 0.04 0.07 0.90 Tortilla chips 0.08 0.07 0.01 0.03 Potatoes 0.09 0.08 0.01 0.20

11 Regression models were also ﬁtting to test whether demographic characteristics made a difference on the discount for the perceived GM food. No demographic characteristic affected signiﬁcantly the discount for the perceived GM food. Also, one could not reject the null hypothesis that third-party information did not affect the difference in the discount for the perceived GM food. 12 We also ran Wilcoxon rank-sum tests to see if one could reject that consumers had different behaviour for the different label types. The results for the Wilcoxon rank-sum tests are similar to those of the t-test results, showing that one cannot reject the null hypothesis that consumers perceive the signals from the two labelling policies the same. Consumer - Chap 04 5/3/04 15:55 Page 50

50 W.E. Huffman et al.

icy in place, the other emulated a market GM food labels, say in Europe or Australia. that had a voluntary labelling policy in For example, do consumers in those coun- place. We found no evidence that con- tries read the same signals of genetic modiﬁ- sumers could more easily distinguish which cation in voluntary labelling markets as in product was GM or non-GM in the manda- mandatory labelling markets? If people can tory labelling market. This provides evidence read the signals for which food is GM accu- that the voluntary labelling policy in the rately in either mandatory or voluntary GM- USA is the best policy. labelling regimes, this calls into question the One further avenue for research would be relevance and usefulness of the mandatory to examine the international dimension to labelling policies throughout the world.

References

AgBiotech Reporter (2001) 18(8). Australia New Zealand Food Authority (2001) Genetically modified foods. Available at http://www.anzfa.gov.au/GMO/. Becker, G., DeGroot, M. and Marschak, J. (1964) Measuring utility by a single response sequential method. Behavioral Science 9, 226–236. Caswell, J.A. (1998) Should use of genetically modified organisms be labeled? AgBioForum 1, 22–24. Caswell, J.A. (2000) Labeling policy for GMOs: to each his own? AgBioForum 3, 53–57. CNN (2001) Europe approves tough GM food rules. Available at http://www.cnn.com/2001/ WORLD/europe/02/14/eu.gm/index.html. Consumer Reports (1999) Genetically altered foods, recommendations. Available at http://www. consumerreports.org/main/detail.jsp?CONTENT%3C%3Ecnt_id=19339andFOLDER%3C%3Efolder_id =18151andbmUID=1011290002896. Council for Biotechnology Information. Frequently asked questions. Available at http://www.whybiotech. com/en/faq/default.asp?MID=10 (accessed October 2001). Friends of the Earth. (2001) The need for labelling genetically engineered foods. Available at http://www.foe.org/safefood/factshtgelabel.htm. Greenpeace International (1997) Greenpeace launches genetech labelling policy as European Commission fails to do so. Available at http://www.greenpeace.org/pressreleases/geneng/1997nov3.html. Huffman, W.E., Rousu, M., Shogren, J.F. and Tegene, A. (2002) Should the United States regulate a mandatory labelling policy for GM foods?’ Working paper, Iowa State University, Ames, Iowa. Knetsch, J.L., Tang, F.-F. and Thaler, R.H. (2001) The endowment effect and repeated trial auctions: is the Vickrey auction demand revealing?’ Experimental Economics 4, 257–269. Kirchhoff, S. and Zago, A. (2001) A simple model of mandatory vs. voluntary labeling of GMOs. Working paper, Istituto Nazionale di Economia Agraria. List, J.A. and Lucking-Reiley, D. (2000) Demand reduction in multiunit auctions: evidence from a sports- card field experiment. American Economic Review 90, 961–972. List, J.A. and Shogren, J.F. (1999) Price information and bidding behavior in repeated second-price auctions. American Journal of Agricultural Economics 81, 942–949. Noussair, C., Robin, S. and Ruffieux, B. (2002) ‘Do consumers not care about biotech foods or do they just not read the labels? Economics Letters 75, 47–53. Phillips, P.W.B. and Foster, H. (2000) Labelling for GM foods: theory and practice. Paper presented at the International Consortium on Agricultural Biotechnology Research (ICABR) on ‘The Economics of Agricultural Biotechnology’, Ravello, Italy, 24–28 August. Phillips, P.W.B. and McNeill, H. (2000) A survey of national labeling policies for GM foods.’ Agbioforum 3, 219–224. Rousu, M. and Huffman, W.E. (2001) GM food labeling policies of the U.S. and its trading partners. Staff paper No. 344, Iowa State University, Department of Economics, Ames, Iowa. Rousu, M., Huffman, W.E., Shogren, J.F. and Tegene, A. (2002) The value of verifiable information in a controversial market: evidence from lab auctions of genetically modified foods.’ Staff working paper No. 3, Iowa State University, Department of Economics, Ames, Iowa. Consumer - Chap 04 5/3/04 15:55 Page 51

Effects of Mandatory GM Labelling 51

Shogren, J.F., Margolis, M., Koo, C. and List, J.A. (2001) A Random nth-price auction. Journal of Economic Behavior and Organization 46, 409–421. Smyth, S. and Phillips, P. (2002) Competitors co-operating: establishing a supply chain to manage genetically modiﬁed canola.’ International Food and Agribusiness Management Review 4, 51–66. US Food and Drug Administration (2001) Guidance for industry: voluntary labeling indicating whether foods have or have not been developed using bioengineering. Available at http://vm.cfsan. fda.gov/~dms/biolabgu.html. Vickrey, W. (1961) Counterspeculation, auctions, and competitive sealed tenders. Journal of Finance 16, 8–37. Wilson, W.W. and Dahl, B.L. (2002) Costs and Risks of Testing and Segregating GM Wheat. Agribusiness and Applied Economics Report No. 501, North Dakota State University, Fargo. Consumer - Chap 04 5/3/04 15:55 Page 52 Consumer - Chap 05 5/3/04 15:55 Page 53

5 Using Simulated Test Marketing to Examine Purchase Interest in Food Products that are Positioned as GMO-free

Marianne McGarry Wolf, Angela Stephens and Nicci Pedrazzi Agribusiness Department, California Polytechnic State University, San Luis Obispo, CA 93407, USA

Introduction GMO-free label on purchase interest for two snack products. Simulated test marketing Research has been conducted to examine con- research is a valid methodology that has been sumer attitudes toward genetically modified used by the marketing community since the food. However, attitude research does not 1960s to forecast purchase interest in new address the specific question: ‘Will purchase products and new positionings for existing interest in a specific product be impacted by products. For example, microwaveable, fat- the consumer’s knowledge that a product is a free, low calorie and organic have been exam- genetically modified food product?’According ined through the use of simulated test to the Wall Street Journal on 5 April, 2001, marketing as characteristics of food products there is a segment of the consumer market that that impact purchase interest. Simulated test wants to know of the presence of genetically marketing is a combination of mathematical modified ingredients and wants to avoid them. modelling and a laboratory experiment. The In response to the consumers’ desire to avoid laboratory experiment is used to simulate the genetically modified ingredients in food prod- purchase environment for consumers. It may ucts, numerous products have appeared on the be hypothesized that consumer reaction to a grocery shelves in the USA that bear the label GMO-free positioning will be related to the ‘non-GMO’. The Wall Street Journal indicated type of food product that has such a position- that industry executives believe that the non- ing. Therefore, this research examines con- GMO segment is growing approximately as fast sumer response to the GMO-free positioning as organic food products. The organic market to two types of branded convenience food is a $7.8 billion market that is growing at eight products: a salty snack food and a fresh veg- times the rate of the packaged foods market etable snack food. This research uses the lab- (Callahan and Kilman, 2001). oratory experiment component of simulated test marketing in a four-cell study design. One cell examines the newly positioned GMO-free Methodology salty snack product. The second cell examines the same product without the GMO-free posi- This research uses simulated test marketing tioning at the same price. Cell three examines methodology to examine the impact of a the newly positioned GMO-free vegetable

54 M. McGarry Wolf et al.

snack product. The fourth cell examines the The validation history for year 1 projec- same product without the GMO-free position- tions is very strong for the forecasting sys- ing at the same price. A print advertisement is tems using simulated test marketing the stimulus used to represent the concept in methodology. For 250 cases reported by the laboratory experiment. Bases, in-market sales have been within 10% The research uses a survey instrument and of predicted sales (Clancy et al., 1994). In- a concept exposure that were administered market sales generated by products tested through the use of a personal interview of 558 using the Designator system have been within randomly selected respondents at food stores 9% (Clancy et al., 1994). in February 2001 in San Luis Obispo County, The LitmusR system uses two types of lab- California. San Luis Obispo County was desig- oratory experiment. One uses behaviourally nated the best test market in the USA by generated market response and the other uses Demographics Daily (Thomas, 2001). San attitudinally generated market response. Over Luis Obispo was found to be the best of 3141 2600 cases have been tested using behav- counties to represent a microcosm of the USA iourally generated purchase interest data. based on 33 statistical indicators. More than 700 of these cases have been available for validation and in 92% of the cases the forecasts were within 10% of the Simulated Test Marketing actuals (Clancy et al., 1994). Among the 287 cases that have been tested using attitudinally Simulated test marketing technology has generated trial, 49 were available for valida- evolved over time through a combination of tion. In 85% of the cases the forecasts were methodologies for generating market within 14% of the actuals. response and mathematical models that simu- To forecast accurately the launch of a new late the marketing environment. The result of product or the repositioning of an existing this combination is a reliable and valid brand through a simulated test market method- methodology for forecasting awareness, pene- ology, the actual marketing environment that tration, share, and volume for new and repo- will exist at the time of the launch/restage is sitioned products and services. modelled. The LitmusR system models the There are a number of simulated test market- entire marketing mix. It uses three categories ing systems that are used by marketing research of model input: category data, market companies. Bases, LitmusR, Designator, and FYI response and marketing plans (Clancy et al., are such systems (Clancy et al., 1994). They 1994). Market response is the consumers’ are branded research methodologies. The spe- response to the test product generated by lab- cific types of forecasts provided by simulated oratory experiment. The purchase probability, test marketing research vary between systems. the repurchase probability, and the purchase However, all simulated test marketing systems cycle for the new or repositioned product are provide the option of generating volumetric the key components of consumer response. forecasts for year 1 sales. Year 1 sales are a The category data and marketing plans pro- function of the awareness of the new product vide the competitive environment that will generated by the marketing plan, the distribu- exist during the launch or restage. tion of the product in the market place and The interview process for most simulated consumer response to the new product. The test marketing methodologies occurs in three LitmusR system provides both awareness and stages. The initial interview stage generates volume forecasts. In particular, the LitmusR sys- purchase interest. The second phase occurs tem forecasts total brand awareness for new after the consumer has taken the product products, campaign awareness in the case of home to use. The second phase generates the the repositioning of an established brand, pene- first repeat purchase probability and the pur- tration (percentage of consumers purchasing), chase cycle. The third phase generates loyalty repeat buyers, unit sales and share for each and refines the purchase cycle. The repeat month of the introductory year for new and purchase probability, the purchase cycle and repositioned products. loyalty are generated attitudinally. Consumer - Chap 05 5/3/04 15:55 Page 55

Purchase Interest in GMO-free Foods 55

The initial phase of research is a one-to- Some possibility will buy (30 chances in 100) one interview. This interview screens the con- Slight possibility will buy (20 chances in 100) sumers to confirm qualifications and collects Very slight possibility (10 chances in 100) background information concerning attitudes will buy and practices in the category. This informa- No chance will buy (0 chances in 100) tion is used for two purposes: to simulate the The 11-point behaviour probability scale was shopping experience by having the consumer developed by Dr Thomas Juster and published think about the category; and to collect infor- in different forms during the 1960s (Clancy et mation that is used to profile the consumer. al., 1994). The 11-point scale couples word For example, the number of units purchased meanings with probability estimates to in the test product’s category identifies heavy, enhance serious thinking. It is more discrimi- medium and light purchasers (Clancy et al., nating than the traditional five-point or seven- 1994). The brand share of requirements for point scale that is used by forecasting systems the category identifies brand purchasers. such as Bases. It has been employed by mar- After the background information is col- keting researchers such as Yankelovich lected, the respondent is exposed to the con- Partners since the early 1960s in numerous cept in a competitive context. The concept consumer behaviour studies including pack- exposure is done through a simulated store aged goods, durables, financial products and environment and an advertisement. In the services, and other categories. absence of a finished product, a printed con- Yankelovich Clancy Shulman found that, cept is used as the exposure. In the LitmusR like all self-reported measures of consumer system the concept is always priced, accom- behaviour, the 11-point scale overstates buy- panied by a competitive array with the key ing (Clancy et al., 1994). Much of the over- competitors priced appropriately for the mar- statement is a result of the 100% awareness ket. The respondent purchases the product in and distribution in the research environment the simulated store environment or evaluates which is never realized in the real world. the concept for potential trial through a model Even taking this into account by factoring in the absence of a simulated store. responses by forecasted awareness and distribution, people are more likely to say they will buy than in fact do buy. This was found to be Market Response true in all product categories examined (Clancy et al., 1994). This research is conducted through the use of a concept exposure phase only. Purchase interest in this research uses an 11-point pur- Concept Exposure chase intent scale. Each of the 11 points is coupled with a verbal anchor from: ‘Certain This research uses simulated test marketing will buy – 99 chances in 100’ to ‘No chance methodology to examine the impact of a will buy – zero chances in 100’: The following GMO-free label on purchase interest for two question is used after concept exposure in the snack products. It may be hypothesized that competitive context: consumer reaction to a GMO-free positioning If you find this new product in a store where you will be related to the type of food product that shop, how likely would you be to purchase this has such a positioning. Therefore, this new product in the next 12 months? research examines consumer response to the GMO-free positioning to two types of branded Certain will buy (99 chances in 100) convenience food products: a salty snack Almost sure will buy (90 chances in 100) Very probably will buy (80 chances in 100) food, Ruffles Snack Pack, and a fresh veg- Probably will buy (70 chances in 100) etable snack food, Cool Cuts. Ruffles Snack Good possibility will buy (60 chances in 100) Pack is a 5 oz container of Ruffles brand Fairly good possibility (50 chances in 100) potato crisps and onion dip in a convenience will buy package. Ruffles is the number two brand of Fair possibility will buy (40 chances in 100) potato crisps in the USA (Gale Research Inc., Consumer - Chap 05 5/3/04 15:55 Page 56

56 M. McGarry Wolf et al.

2000). The Rufﬂes Snack Pack is sold in the indicated a 90% chance or higher probability crisps aisle of the supermarket. Cool Cuts is a are designated likely purchasers of the prod- three-pack of carrots and ranch dip with a uct. Respondents with lower than a 90% picture of Bugs Bunny on the package to probability of purchasing the product are des- attract the attention of consumers. Cool Cuts ignated non-purchasers. There is no differ- is sold in the value-added section of the pro- ence in the purchase interest between the duce department. products or positionings. Therefore, the GMO- This research uses the laboratory experi- free positioning did not have an impact on ment component of simulated test marketing purchase interest (Table 5.1). in a four-cell study design. One cell examines the newly positioned GMO-free salty snack product. The second cell examines the same Positioning of a New or Existing Product product without the GMO-free positioning at the same price. Cell three examines the newly A successful product positioning is based on positioned GMO-free vegetable snack prod- the factors that motivate consumers to pur- uct. The fourth cell examines the same prod- chase one product versus other products. uct without the GMO-free positioning at the The products that are examined here are same price. A print advertisement is the stim- branded snack food products. In order to ulus used to represent the concept in the labo- develop a successful positioning, the charac- ratory experiment. The retail price for all four teristics that are desirable to consumers when cells is $1.99. they shop for a snack food must be identiﬁed. Consumers in all cells were shown the The characteristics that consumers want same competitive board. The competitive when they purchase snack foods are exam- board displayed pictures of crisps, vegetable ined by desirability ratings (Clancy et al., snacks, snack mixes and both of the new 1994). The most desirable characteristics products. Each product was priced based on should be used in the development of a prod- current market conditions. Ritz Snack Mixers uct positioning since those are the most were $1.99 for a 7 oz. container. Pringles important to consumers when they purchase were $1.99 for a 6 oz. container. Carrot a new product. The product positioning Dippers were $1.99 for a three pack. After should also stress the characteristics that the the respondent had time to review the concept consumers perceive the product to have rela- and the competitive boards, purchase interest tive to the competition. was evaluated using the 11-point scale. In order to understand how consumers perceive the products in the competitive array, each product is rated on the charac- Consumer Purchase Interest in the teristics that were evaluated for desirability. It GMO-free Positioning is important to note that consumers develop perceptions about products, in this case The 11-point purchase interest scale was used snack foods, from the concept exposure, to determine purchase interest in the GMO- experience, seeing the products in the store, free positioning compared to the positioning advertisements, word of mouth, public rela- that did not discuss GMOs. Respondents who tions and the media. The perceptions about

Table 5.1. Rufﬂes Cool Cuts Rufﬂes not Cool Cuts not GMO-free GMO-free GMO-free GMO-free (n = 129) (n = 129) (n = 155) (n = 141) Chi squared

Likely purchasers 14.0% 11.6% 8.4% 13.5% 2.70 Non-purchasers 96.0% 88.4% 91.6% 86.5% Consumer - Chap 05 5/3/04 15:55 Page 57

Purchase Interest in GMO-free Foods 57

a product provide the consumer with the desirable characteristics and somewhat to very information they use to decide to purchase a desirable characteristics (Table 5.2). The product. It is the responsibility of the promo- attributes that are very to extremely desir- tional campaign for a product to communi- able to consumers are tasty, flavourful, satisfy- cate the appropriate information to ing and healthy. The somewhat to very consumers who have not had experience desirable characteristics to consumers are with the product. The promotional campaign natural product, contains vegetables, good for also reinforces the perceptions of the con- children and free of genetically engineered sumers that have had experience with the ingredients. Therefore, when marketing these product. products, the characteristics concerning tasty, flavourful, satisfying and healthy are important to the consumer. The characteristic, free Desirability Ratings of New Product of genetically engineered ingredients, is only Characteristics somewhat desirable. Therefore, it is not an important positioning characteristic for the Before the respondents were exposed to the typical snack consumer. new products, GMO-free Ruffles Snack Pack, Ruffles Snack Pack, GMO-free Cool Cuts and Cool Cuts, they rated eight characteristics Product Ratings that describe the products on a five-point desirability scale. The purpose of the rating is In order to understand how consumers per- to identify the characteristics of the products ceived the new products and the conventional that impact a consumer’s purchase decision. products, the products were rated on the Characteristics of the products concerning characteristics that were also rated for desir- taste, natural product, contains vegetables, ability. Respondents answered the following good for children, satisfying, free of geneti- question: cally engineered ingredients, healthy and Based on your perceptions, please use the flavourful were rated. Consumers were asked following scale to describe how these the following question: characteristics describe GMO-free Ruffles, Please rate the following characteristics you look GMO-free Cool Cuts, Ruffles, and Cool Cuts for when shopping for snack foods where: where: 5 = Describes completely; 4 = Describes 5 = Extremely desirable; 4 = Very desirable; very well; 3 = Describes somewhat; 2 = 3 = Somewhat desirable; 2 = Slightly desirable; Describes slightly; 1 = Does not describe at all. 1 = Not at all desirable. The mean rating of 1.95 generated for the Analysis of the mean ratings of the interval characteristic, the product is free of geneti- data indicates that the characteristics are cally modified ingredients, for the GMO-free divided into two groups: very to extremely Ruffles indicates that the respondents that

Table 5.2. Desirability of characteristics (n = 558).

Characteristic Mean SE

Very to extremely desirable Tasty 4.62 0.03 Flavourful 4.56 0.03 Satisfying 4.32 0.04 Healthy 4.06 0.04 Somewhat to very desirable Natural product 3.48 0.05 Contains vegetables 3.27 0.05 Good for children 3.25 0.06 Free of genetically engineered ingredients 3.06 0.06 Consumer - Chap 05 5/3/04 15:55 Page 58

58 M. McGarry Wolf et al.

were exposed to the GMO-free Ruffles did not product that is free of genetically engineered perceive the product as being free of geneti- ingredients either extremely or very desirable cally modified ingredients (Table 5.3). and consumers that find a snack product that Similarly, the mean rating 2.85, generated for is free of genetically engineered ingredients to the GMO-free Cool Cuts indicates that be either somewhat, slightly or not at all desir- respondents only somewhat agreed that the able. Not surprisingly, those that think being product is free of genetically modified ingredi- free of genetically engineered ingredients is an ents. The term, GMO-free, does not appear important attribute are less likely to purchase to mean free of genetically modified ingredi- a genetically modified food (Table 5.4). ents to the average consumer. However, there is no difference in purchase interest in the GMO-free positioning for the snack products examined here between Free of Genetically Engineered the segment that indicated that free of geneti- Ingredients Segmentation cally engineered ingredients is an important attribute and those that do not think it is an Consumers were segmented based on the important attribute (Table 5.5). This is further desirability of snack products to be free of evidence that the term, GMO-free, does not genetically engineered ingredients. Customers appear to mean free of genetically modified are segmented into those that find a snack ingredients to the average consumer.

Table 5.3. Mean product ratings.

Free of genetically engineered ingredients Mean SE

GMO-free Rufﬂes (n = 130) 1.95 0.10 GMO-free Cool Cuts (n = 140) 2.85 0.13

Table 5.4. Purchase interest in a genetically modiﬁed food product. GMO-free Not GMO-free (n = 197) (n = 357) Chi squared

Deﬁnitely 4.1% 3.7% Probably 10.7% 26.1% Maybe 31.6% 44.4% Probably not 25.5% 19.7% Deﬁnitely not 28.1% 6.2% 65.1*

*Signiﬁcance at the 0.10 level.

Table 5.5. Purchase interest in GMO-free products. GMO-free Not GMO-free Chi squared

Both GMO-free products n = 90 n = 178 Likely purchasers 13.3% 11.8% 0.13 Non-purchasers 86.7% 88.2% GMO-free Rufﬂes n = 41 n = 87 Likely purchasers 12.2% 10.3% 0.75 Non-purchasers 87.8% 89.7% GMO-free Cool Cuts n = 49 n = 91 Likely purchasers 14.3% 13.2% 0.86 Non-purchasers 85.7% 86.8% Consumer - Chap 05 5/3/04 15:55 Page 59

Purchase Interest in GMO-free Foods 59

Conclusions eight characteristics described the product. Consumers who rated the salty snack that was The GMO-free positioning had no impact on positioned GMO-free indicated that the purchase interest for either of the products phrase free of genetically modified ingredi- examined, a salty snack food and a fresh ents described the product slightly. vegetable snack food product. Consumers Consumers who rated the salty snack that was who indicated that free of genetically modi- positioned GMO-free indicated that the fied ingredients was extremely or very desir- phrase free of genetically modified ingredi- able in their purchase decision for a snack ents described the product somewhat. product did not have a higher purchase Therefore, it appears that many consumers interest for the products labelled GMO-free. do not understand the term GMO-free. The same consumers indicated that they The GMO-free positioning for a snack were less likely to purchase genetically modi- food does not impact the purchase interest for fied food products. consumers. This may indicate that while con- Positioning research was conducted to sumers report in attitudinal research a likeli- determine the characteristics of snack food hood of not purchasing genetically modified products that are important to consumers food products, when deciding to purchase when purchasing snack products. Eight char- branded and priced products in a competitive acteristics were examined. The least impor- context, the presence of genetically modified tant of the eight characteristics was free of ingredients is a factor of low importance. In genetically modified ingredients. Each prod- addition, many consumers are not familiar uct was also rated on how well each of the with the meaning of GMO-free.

References

Callahan, P. and Kilman, S. (2001) Seeds of doubt. Wall Street Journal, 5 April. Section A, p. 1. Center for Food Safety (2000) Why voluntary labelling of genetically engineered foods won’t help consumers. Available at http://www.centerforfoodsafety.org/facts&issues/VoluntaryLabelingMemo.html. Clancy, K.J. Shulman, R.S. and Wolf, M.M. (1994) Simulated Test Marketing: Technology for Launching Successful New Products. Lexington Books, New York, 1994. Consumer International Briefing Paper. Genetically modified foods: magic solution or hidden menace? Available at www.consumerinternational.org/campigns/biotech/breifing.html. Deis, R.C. (2000) Tortilla chip. Food Product Design. Available at http://www.foodproductdesign.com/ srchive/2000/100ap.html. Gale Research Inc. (2000) Market Share Reporter. Gale Research, Detroit, Michigan. Nuffield Council on Bioethics. Genetically modified crops: the ethical and social issues. Available at www.nuffield.org/bioethics/publication/modifiedcrops/rep0000000082.html. Thomas, G.S. (2001) Playing in San Luis Obispo, Demographics Daily, 6 February. Available at wysi- wyg://44/http://bizjournals.bcentral.com/journals/demographics/. Whitman, D.B. (2000) Genetically modified foods: harmful or helpful? Cambridge Scientific Abstract. Available at www.csa.com/hottopics/gmfood/oview.html. Consumer - Chap 05 5/3/04 15:55 Page 60 Consumer - Chap 06 5/3/04 15:55 Page 61

6 Measuring the Value of GM Traits: The Theory and Practice of Willingness-to-pay Analysis

Simbo Olubobokun and Peter W.B. Phillips Department of Agricultural Economics, University of Saskatchewan, 51 Campus Drive, Saskatoon, Canada S7W 5AB

Introduction measure of the benefit (the value) of the product to the individual. WTP studies have As scientists continue to do research and generally been conducted in the area of develop new crops, along with studies on the environmental economics using tools such food safety of such crops, it is also important as option price, option demand and option to conduct economic studies on how con- value to measure consumers’ willingness to sumers perceive the benefit–cost of the poten- pay for certain resources. This chapter tial traits. reviews and assesses six theoretical Presently, there is growing attention approaches used to measure the willingness towards genetically modified (GM) foods. of a consumer to pay for the perceived ben- While there are a wide range of citizen and efits of a good. consumer surveys that show varying degrees of support or antagonism to GM foods, it is not clear how these views would or could Background influence the operation of the market in the absence of regulation. The introduction of GM crops has economic Consumer welfare is measured by the significance. The 17 crops that have been amount of money that must be paid (or genetically modified and approved for produc- received) by the consumer facing a change tion somewhere in the world are reputed to be in some parameter in order to keep him or potentially used in 70% or more of the her at the particular level of utility. Welfare processed foods sold in developed countries. of an agent is usually measured based on Consumers have expressed a number of con- the fact that value is derived from the utility cerns about this development, ranging from that individuals derive from satisfying their fears of new or enhanced health and environ- wants (according to their preferences). This mental safety risks to questions about the eco- value is usually measured in terms of what nomic, commercial, ethical and social impacts people are willing to pay. Thus, an individ- of the technology. Governments and regulators ual’s willingness to pay (WTP) for an item or are seeking ways to address consumer con- product can be used as a measure of the cerns, using partial or total bans on the tech- utility they derive from the product or as a nology, voluntary or mandatory labelling rules,

62 S. Olubobokun and P.W.B. Phillips

or trade embargoes. Each of these has signifi- Modelling Consumer Preferences cant potential to complicate the already difficult international trading relationships around food. Theoretical background Many surveys show that consumers are beginning to differentiate their views about dif- Using the framework of utility theory, this ferent GM foods. The majority of consumers in chapter will review six theoretical approaches most markets are not keen on GM traits that used to measure the willingness of a con- simply change agronomic practices and sumer to pay for the perceived benefits of a increase yields, such as herbicide tolerance. good. We will show that while there are differ- There is somewhat greater support for input ent methods of evaluating the benefits to a traits that reduce the use of chemicals, such the consumer, some of these methods overlap. Bacillus thuringiensis (Bt)-resistant and viral- The price the individual is willing to pay for a resistant crops. Support rises for those products good represents the expected utility of the that are perceived to have greater consumer good1. This value could be greater than, the benefits, such as those that have been modified same as or less than the market price for the to improve health, nutrition or increase shelf- good. The theoretical model shows that if a life. A survey conducted by Environics consumer perceives a health benefit from the International in 2000 showed, for example, that consumption of a particular product, it will be 86% of respondents were strongly in favour of demonstrated by a willingness to pay a pre- biotechnology applications in new medicines, mium above the market price of the product. 69% for more nutritious crops, 66% for pest- Consumer information and knowledge are resistant crops and 33% for farm animal pro- very important in a period when consumers duction. These survey results are borne out in a appear to be changing their attitude towards similar survey conducted by Yann Campbell quality, especially their definition of quality in Hoare Wheeler (1999) and in consumer studies the area of GM foods. An attitude is an overall where GM and GM-free produce in the same evaluation that expresses how much we like or food category are actually offered for sale to dislike an object, issue or action. Attitudes consumers (Powell, 2000). reflect the consumer’s overall evaluation of Generally, the scientific research process something based on the set of associations involves scientists identifying a problem such linked to it. Attitudes are learned, and they tend as variability in income of farmers due to a to persist over time. Attitudes are important short growing season and then developing a because they guide the consumer’s thoughts, crop that is cold-tolerant. The gap so far is feelings and behaviour and ultimately influence that scientists and investors have invested sub- the consumer’s buying behaviour during the stantial resources to develop new agricultural acquisition, consumption and disposition of an crops that address certain agronomic issues offering (Hoyer and Maclnnis, 2001). with very little assessment of consumer accep- Perception of quality could be based on a tance of such crops. number of factors such as taste, knowledge of This rest of this chapter is organized into the fact that the product was genetically modi- three sections. The next section reviews the fied, the level of pesticide used in production theoretical background and empirical mea- and brand name. Usually consumers define a surements of WTP, examines theoretical mod- set of determining attributes for a particular els used to analyse willingness to pay for a product class by acquiring information and product and assesses the appropriateness of learning about the available alternatives. After the various modelling options for measuring comparing available products based on the willingness to pay for a product. The follow- determining attributes, consumers eliminate ing section provides a short discussion of the some alternatives and develop final ‘choice sets’ applications of the theoretical model and the of products from which to choose. Options fac- final section provides some concluding coming the consumer are: purchase, delay purchase ments and policy options. or not to purchase (Louviere, 1988).

1 The authors acknowledge Richard Gray for his suggestions. Consumer - Chap 06 5/3/04 15:55 Page 63

Measuring the Value of GM Traits 63

With the increased awareness of GM pro- Welfare from the stated WTP values can be ducts, various surveys have been conducted measured by the consumer surplus, compensat- and certain issues have been identified to be ing variation, equivalent variation, option price important to consumers. Four broad groups or demand and option value. The approach of concerns are apparent: specific food safety used in most cost–benefit studies when measur- and quality concerns, environmental con- ing consumer welfare is to estimate the future cerns, fear of the unknown, and ethical objec- demand for the goods under consideration and tions (Hobbs and Plunkett, 1999). By the expected consumer surplus value. It should investigating the willingness to pay for specific traits in GM products, this paper will apply be noted that surveys which elicit the WTP of some economic analysis to some of these spe- consumers and estimate future demand obtain cific issues of consumer perception. results that are classified as stated preferences While consumers cannot directly purchase which may not give the same result as actual units of food safety or quality, they can purchase preferences. In the rest of this sec- choose to avoid the goods that they perceive tion, we will discuss six economic tools used to to be unsafe or those that have lower quality. measure the willingness of a consumer to pay They can also choose to pay a higher price for the perceived benefits of a good. for the goods that they perceive to be less risky or have a higher quality (Kuperis et al., 1999). Consumers typically purchase the Consumer surplus attributes that are embodied in a product, rather than purchase the product for itself. Consumer surplus (CS) is the measure of wel- The product, per se, does not give utility to fare change with a given price change. the consumer – the characteristics of the Consumer surplus is usually defined as how product do (Lancaster, 1966). much money a consumer would pay for the van Ravenswaay (1995) identified three right to continue to buy at the current price classes of consumer activities involved in something that he/she is now buying or household production of a healthy state that intends with certainty to buy in the future. influence utility levels and have implications for Expected consumer surplus, on the other welfare analysis and marketing research. Health maintenance, rehabilitation and protec- hand, will apply to the good whose consumption affect current expenditures and future lev- tion is uncertain. Byerlee (1981) modified the els of utility. A consumer who is protecting definition by adding that consumer surplus is his/her health might not consume a good how much money a consumer would pay for because of the perceived health cost that could the right to buy at the current price something result from the consumption of such a good. A that he/she is now buying or may buy in the consumer who is maintaining or rehabilitating future. This definition includes both certain as his/her health might consume a good because well as uncertain situations. of the perceived health benefit that could result Suppose good X is being consumed because from the consumption of such a good. of its perceived health benefits. For simplicity of Goering (1985) allowed the expectation analysis, one can consider the demand for good about product quality rather than taste to vary X to be equal to the demand for the quality of among consumers. The results indicated that good X2. Assuming a given quantity, the WTP the higher an individual’s expectation about values can be interpreted as the WTP for the product quality, the higher the price the individual will be willing to pay for the product. health benefits in good X. To determine the The distribution of consumer expectations welfare gain from a price decrease, one can was made to determine the demand for the determine how much the consumer will be will- product. From this analysis, we can assume ing to pay for the price change from P0 to P1. that at a given quantity, a consumer having a The welfare change has two components. First, perceived health benefit will be willing to pay the consumer will be willing to pay (P0 P1)X0 a higher price relative to a consumer who has which is the savings on total expenditure on the a perceived health cost. health benefits (in good X) at the original

2 The authors acknowledge Mobinul Huq for his suggestions. Consumer - Chap 06 5/3/04 15:55 Page 64

64 S. Olubobokun and P.W.B. Phillips

amount X0. Second, the lower price makes it The consumer surplus, CV and the EV are possible for the consumer to afford to purchase measured off the observed demand curve. As more health benefits of good X. See Varian such, they are ex post assessments in the (1978, pp. 207–209) for the derivations. sense that welfare is evaluated after the true If we assume that the marginal utility of state of the world is known. The ex post income is constant, then the consumer sur- analysis, while useful, is not directly measur- plus is measured by able for potential GM traits in non-market P0 =− goods. For such goods, estimation of CS, CV ∫ XPy(,)d P11 [( VP ,) y VP ( 0 ,)]/ y P1 and EV can be done ex ante by using an ∂∂ [(,)/]VPy y expected demand function obtained through a where the right-hand side of the above expres- contingency valuation method or in an experi- sion is the money equivalent of the indicated mental setting. utility change (given the price change). Uncertainty in the demand for a good could be due to factors such as uncertainty concerning income, uncertainty concerning comple- Compensating and equivalent variation ments or substitutes, or uncertainty concerning Compensating variation (CV) in income is personal preferences. Most research studying defined as the amount of money that the con- potential products are evaluated ex ante in the sense that the compensation paid or received sumer would need to be paid at P1 so as to is measured while the consumer is uncertain maintain the same level of utility (at P0) before the price change (Varian, 1978). Assume a about the future demand for the good. This compensation is sometimes referred to as ‘will- consumer with an initial price vector P0 (with ingness to pay’ or the option price of an item. demand X(P0)) and a final price vector P1 Willingness to pay for the perceived health (with demand X(P1)), has a constant money benefits in a product represents the full value income W0 in each situation. to the individual of the health benefits in the V(P , W +CV) = V(P , W ) 1 0 0 0 product. What a consumer is willing to pay CV = e(P , V ) – e(P , V ) 1 0 0 0 above the market price for a product with per- V = V(P , W ). 0 0 0 ceived health benefits could be a representa- Equivalent variation (EV) is the amount of tion of the CV or EV of the consumer, but income that could be taken away from the probably is not equal to it due to uncertainty. consumer at P0 to make him as well off as he would be at P 1 Certainty equivalence V(P0, W0 EV) = V(P1, W0) Based on expected utility theory, a rational EV = e(P1, V1) – e(P0, V1) agent’s preferences over risky alternatives can V1 = V(P1, W0). be ordered by the mathematical expectations CV or EV can be represented by the amount of utilities for the possible outcomes of the usually referred to as a premium that the con- alternatives (Varian, 1978). In addition, a sumer is willing to pay for the health benefits rational agent will choose among risky alterna- above the market price of goods. In measuring tives so as to maximize expected utility. Let L the welfare benefits to different agents, the be a lottery which results in a known state (W ) Hicksian surplus measure is commonly used. 0 with a probability of P and a known state (W1) Compensating variation (CV) and equivalent with a probability of 1 – P. The expected value variation (EV) are usually estimated off the of the lottery E(L) = PW0 + (1 – P)W1 = W*. Hicksian demand curve. Since Hicksian Certainty equivalence (CE) of a lottery L is demand curves are unobservable, Slutsky’s usually defined as the amount of ‘sure’ wealth, equation can be used to relate the slope of the which gives the same utility as the expected Hicksian to the Marshallian demand curve, utility of a lottery (L). The utility of the cer- which is the observed demand curve. tainty equivalence is equal to the expected ∂ ∂ ∂ ∂ ∂ [ X1(P, W0)/ P1 = h1(P, V(P, W0))/ P1 X1 utility (EU) derived from the consumption of ∂ (P, W0)/ W] X1. the good: U(CE(L)) = EU(L). Consumer - Chap 06 5/3/04 15:55 Page 65

Measuring the Value of GM Traits 65

Risk aversion means that the individual is ence between what a consumer would be willing to accept a lower amount than the willing to pay for the option of using a expected value of the lottery in order to avoid resource and the expected consumer surplus risk. Consequently, a risk-averse individual of that usage. derives less utility from the expected value of One could consider the option value as a a given lottery (E(L)) than having the same premium paid to protect from uncertainty. amount of sure wealth (W*). The difference Zeckhauser (1969) and Cicchetti and between the utility of a sure wealth and the Freeman (1981) interpreted option value as a expected utility of a lottery is the utility loss risk premium for risk-averse consumers, the due to uncertainty (U(W) EU(L)). If U(W) = premium being the amount of money that a EU(L), the individual is said to be risk neutral risk averse individual will pay so as to avoid a with a linear utility function. risk. Cicchetti and Freeman (1981) demon- Based on expected utility theory, the utility strated that option values exist separately of the certainty equivalence is equal to the from consumer surplus when there is uncer- expected utility derived from the consumption tainty and individuals are risk averse. They of the good. By assuming that the consumer defined option value as the difference is rational with an objective of maximizing between maximum option price and the expected utility, the amount that the con- expected consumer surplus: OV = OP sumer is willing to pay for a good can be con- E(CS). Their definition of option value posits sidered a representation of the certainty that an additional measure of benefit is equiv- equivalence. alent to a willingness to pay in excess of consumer surplus. Option value can be interpreted as an ex Option price and value ante allowance for uncertainty that the con- The option price (OP) is usually defined as the sumer has about a future event taking place. maximum willingness to pay in the present An individual facing uncertainty will be willing period for the option of demanding a to pay a little extra now above his/her resource in the future. The option price is an expected value of consumer surplus to assure ex ante value in the sense that it is payable in that the resource would be available in the the present and determined while the con- future if he/she decides to demand the sumer is still uncertain about future prefer- resource. Fisher and Hanemann (1986) ences, income levels, resource availability or applied the concept of option value to situa- other economic parameters. tions wherever a decision has the characteris- Since the individual is paying now for an tics that one of the possible outcomes is expected future benefit, an option price could irreversible and there is some future prospect be considered equal to an ex ante willingness of better information about the benefits and to pay for a product and a representation of costs of these outcomes. the certainty equivalence. Byerlee (1981) defined option demand as the amount a con- Links between the theoretical approaches sumer is willing to pay for the option of consuming a good in the future. This definition is If probability values can be assigned to the the same as the term option price. Long future consumption of a product, then option (1967) and Byerlee (1981) showed that under value and option price (or option demand) can certainty, option demand and consumer sur- be used. However, if it is not possible to plus are identical. assign probability values then the elicited In a general case when there is uncer- amount that the consumer is willing to pay for tainty, option demand may be greater than, a product will represent the value of the prod- equal to, or less than the expected consumer uct to the consumer. surplus. Weisbrod (1964) and Lindsay (1969) Willingness to pay for the perceived health showed that uncertainty provides an addi- benefits in a product represents the full value tional component (option value) to consumer to the individual of the health benefits in surplus. The option value (OV) is the differ- the product. Given the assumption that the Consumer - Chap 06 5/3/04 15:55 Page 66

66 S. Olubobokun and P.W.B. Phillips

consumer is rational with an objective of max- on the Lancaster (1966) approach, the imizing expected utility, the amount that the demand for a good is based on the attributes consumer is willing to pay for a good can be of the good, consequently, the utility that the considered a representation of the certainty consumer derives from the consumption of equivalence. Where the utility of the certainty the good is based on the utility derived from equivalence is equal to the expected utility the different attributes of the good. This derived from the consumption of the good. model utilizing an ex ante welfare measure What a consumer is willing to pay above the was adapted from the example used in market price for a product with perceived Epstein (1975) and Chavas et al. (1986). health beneﬁts could be a representation of Utility is derived from the amount of X the CV or EV of the consumer, but probably consumed, and the wealth of the consumer. In is not equal to it due to uncertainty. When the this model, any health beneﬁt derived from marginal utility of income is constant, the area the consumption of X will result in an addition under the demand curve (measured as con- to wealth whereas any health cost experi- sumer surplus) gives an exact measure of the enced from the consumption of X will result willingness to pay for a price change (the in a reduction of wealth. extra consumption of good X). U = U(X, W) W = Wo – Px + H(x) + B + e

Model where U is the utility obtained from the consumption of the particular good X; Wo is the Economic theory usually assumes that people initial wealth of the consumer; Px is the make choices based on their preferences with expenditure on good X; H, which is an the objective of maximizing utility. This means increasing function of X, could be positive or that the marginal utility (benefit) derived from negative depending on the perception of the the consumption of a product is equal to the consumer, +H is the perceived health benefit marginal cost of the product. while H is the perceived health cost that This section looks at the economic aspect results from consuming good x. B is a mea- of the concerns that consumers have for GM sure that accounts for factors such as environ- goods. It uses a two-period model to explain mental impact and ethical issues, the value the utility maximizing conditions for individu- that B takes will depend on the type of GM als having different concerns for GM foods. In food being considered. For a particular good, period t = 1, the consumer has incomplete B will be zero if the consumer only has health information. Current consumption in time t = concerns (perceives health benefits or costs 1 is decided subject to uncertainty about from the consumption). However, if in addi- future prices, future income, future health tion to health concerns the consumer has impact and future environmental impact. The other concerns such as ethical/religious uncertainty facing the consumer is assumed to and/or environmental then B can be modelled be temporal such that at time t = 2 the con- as follows: sumer is able to select X with certainty. 2 B = B + B (x) Between t = 1 and t = 2, the consumer t n

observes (at zero cost) the realized values of where Bn is an increasing function of X; +Bn the random variables from period t = 1. This is the perceived beneﬁt to the environment model reveals insight into four areas of con- and Bn is the perceived cost to the environ- sumer concerns: health, environmental, ethi- ment that results from the direct use of the cal issues and uncertainty about future crop and/or the agronomic practices used in

preferences. the production of good X; and Bt, which is The decision to purchase or not to pur- ﬁxed, is the measure of ethical concern. chase a GM product is based on factors such When there is a possible concern regarding as price, perception of health beneﬁt or risk, the ethics surrounding the production process ethical issues, environmental factors, budget of a product, a consumer will either have or

constraint and individual characteristics. Based not have some concern. Consequently, Bt is Consumer - Chap 06 5/3/04 15:55 Page 67

Measuring the Value of GM Traits 67

modelled ﬁxed and could be measured using wealth in the future at time t = 2 when the

dummy variables. e is the measure of uncer- consumer chooses X2. The expected utility in tainty about future preferences. Like Bt, e is t = 1 (E1(Ux1/Uw1 ) = P1 – H) incorporates assumed to be ﬁxed. In the ﬁrst order condi- the perceived health cost into the optimization

tion (FOC) conditions, Bt and e do not show condition. A shift from the original utility level up because they are not modelled as increas- to the new utility level, ceteris paribus, will ing functions of X (they are fixed values). change utility by an amount H (perceived health cost). Perceived health impact For some goods, the expected health impact Perceived environmental impact from the consumption of such goods could Some consumers will consider for certain GM increase or reduce the utility derived by the traits, the perceived positive impact on the consumer. In such cases, the expected utility environment as an additional benefit of con- in t = 1 incorporates the perceived health suming the good, while a negative impact on benefits (costs) into the optimization condi- the environment will be considered a reduction. The utility function U is assumed to be a tion in utility. Von Neumann–Morgenstern utility function Max E {U[X, Wo – Px + H(x) + B]}. that is twice continuously differentiable and strictly quasi-concave in X. Max E1U(X1, X2), Max E1[V2] s.t W = P X Period 1 (t = 1). V1 = Max E1U(X1, X2), Max E1[V2] 1 1 1 subject to W1 = P1X1 Period 1 (t = 1). V2 = Max U(X1, X2) V2 = Max U(X1, X2) W2 + Hx +Bnx = P2X2 Period 2 (t = 2). subject to W2 + Hx = P2X2 Period 2 (t = 2). Using backward induction, the FOC is first Hx is positive in cases of perceived health solved in t = 2 and then solved in t = 1. benefits and negative for perceived health FOC in t = 2: costs. E1 is the expectation operator that is conditional on the information available at Ux2 – Uw2 [P + Hx + Bnx] = 0 U /U = P + H + B . time t = 1. At the current time (t = 1), an x2 w2 n amount of X is consumed which will give FOC in t = 1: future health benefits. This future health bene- E1(Ux1 – Uw1[P1 + Hx + Bnx]) = 0 fit is modelled as an addition to wealth in the E (U /U ) = P + H + B . future at time t = 2 when the consumer 1 x1 w1 1 n The expected utility in t = 1 incorporates the chooses X2. Using backward induction, the FOC is first perceived health and environmental benefits solved in t = 2 and then solved in t = 1. into the optimization condition. A shift from the original utility level to the new utility level, FOC in t = 2: ceteris paribus, will change utility by an UX U [P + H] = 0 2 w2 2 amount +H + B [perceived health benefits UX /U = P + H. nx 2 w2 2 and perceived (positive) environmental impact]. FOC in t = 1: For goods with perceived (negative) envi- E1(Ux1 Uw1[P1 + H]) = 0 ronmental impact, Bx is modelled as a deduc- E (U /U ) = P + H. 1 x1 w1 1 tion from wealth in the future at time t = 2

The expected utility in t = 1 incorporates the when the consumer chooses X2. The expected perceived health beneﬁts into the optimization utility in t = 1 (E1(Ux1/Uw1 ) = P1 – H – Bn) condition. A shift from the original utility level incorporates the perceived environmental cost to the new utility level, ceteris paribus, will into the optimization condition. A shift from change utility by an amount + H (perceived the original utility level to the new utility level, health beneﬁts). ceteris paribus, will change utility by an For goods with perceived future health amount H Bnx [perceived health cost and costs, Hx is modelled as a deduction from perceived (negative) environmental impact]. Consumer - Chap 06 5/3/04 15:55 Page 68

68 S. Olubobokun and P.W.B. Phillips

It should be noted that the perceived bene- summation of the individual WTP for each fits may not be symmetrical to the perceived perceived benefit in the good. costs due to the fact that on average, percep- Chavas et al. (1986) analysed the effects tions of positive benefits in a good by some of consumer compensation (option price) on individuals may not be equal to perceptions of the optimum future consumption levels of the costs in the same good by other individuals. goods in the bundle. They decomposed the As a result, the optimizing values for per- option price into two additive terms; the first ceived health (and environmental) benefits term corresponds to the surplus measure should be solved differently from perceived commonly used in a risk-less situation and the health (and environmental) costs. second term (correction factor) involves the covariance between the compensated marginal utility of future income and the compen- Links between the model and the sated future demand evaluated at the current theoretical approaches time. They concluded that the correction fac- From the survey results, the elicited amount tor is the amount by which the discounted that the consumer is willing to pay for a expected value of the ex post welfare triangle product will represent the value of the prod- must be adjusted to obtain an exact ex ante uct to the consumer. Willingness to pay for welfare measure. It reflects the influence of the perceived health benefits in a product temporal uncertainty on ex ante welfare mea- represents the full value to the individual of surement. This interaction factor is beyond the health benefits in the product. Given the the scope of this chapter. However, it is a assumption that the consumer is rational with potential area of future research. an objective of maximizing expected utility, Given the economic analysis of what con- the amount that the consumer is willing to sumers are willing to pay for a good (with per- pay for a good can be considered a represen- ceived health benefits), the next section will tation of the certainty equivalence, where the review three empirical methodologies used in utility of the certainty equivalence is equal to WTP studies. the expected utility derived form the consumption of the good. What a consumer is willing to pay above Empirical measurements of willingness the market price for a product with perceived to pay health benefits could be a representation of the CV or EV of the consumer, but likely is Valuation research has been used by a lot of not equal to it due to uncertainty. When the economists when evaluating policy options or marginal utility of income is constant, the area the value of a product. In most instances, it is under the demand curve (measured as con- easier to measure the cost of a good or sumer surplus) gives an exact measure of the action but it is not always easy to measure willingness to pay for a price change. If the the benefits (van Ravenswaay, 1995). In the consumer is willing to pay a premium for the case of market goods, when consumers are perceived health benefit above the market behaving in an optimizing manner (maximiz- price, this will show that the WTP value ing utility) the marginal benefit is equal to the includes the CS and some measure of per- marginal cost (the market price of such a ceived benefit or cost (premium such as good). For non-market goods however, esti- option value). mated benefits are generally measured by It should be noted that it is possible for an what consumers are willing to pay for the interaction to exist between the different ele- hypothetical good. ments of the model, for example, a consumer having personal health concerns or living with Hedonic pricing a sick person could have a lot of concern for the environment. Consequently, the presence Hedonic pricing involves observing choices of such interaction may cause the total WTP (revealed preference) made in an actual or values for good X to be other than a straight constructed market, and inferring the value Consumer - Chap 06 5/3/04 15:55 Page 69

Measuring the Value of GM Traits 69

of the perceived benefits or risks (cost) of the tingency valuation surveys (Coursey and product. It is based on the assumption that Schulze, 1986). The ex ante auction could be quality of a good is related to measurable used to observe how bidding behaviour is specification variables. That is, quality deter- affected by alternative incentives and by mines price and such prices can be said to repeated market experience. Prior to running be dependent on the characteristics (quality) a survey instrument, an experimental auction of the product. Empirical applications usually could provide the researcher with the oppor- involve the regression of prices (or the loga- tunity to design, test and replicate the prefer- rithm of prices) of the different varieties of a ence revealing the incentives of the elicitation type of good on the specification variables. method (Fox et al., 1995). Hedonic price functions are usually esti- When experimental auctions are used ex mated in an attempt to reveal the behav- post, they could be run independently as a ioural information about the marginal values valuation process or the results from such auc- of the characteristics of a product. The price tions, which revealed the learning and market that people are willing to pay for a product experience from the experimental auction, can be interpreted as the price they are will- could be used to adjust the bids of the respon- ing to pay for the quality attributes of the dents of the contingency valuation method. product. It is also possible to evaluate the If respondents have vague or undefined trade-offs between price and the attributes of incentives when evaluating a (GM) product, the product. results from non-market methods could be In the hedonic model, the increment in inconsistent. The laboratory experimental price due to increases in any characteristics auctions provide participants with a well- will equal the buyers’ WTP for the charac- defined incentive structure that enables the teristics and the marginal cost of producing researcher to elicit more accurately the value the characteristic for sellers. When buyers of a non-market product. and sellers have time to adjust their Experimental auctions do not have the responses, the marginal hedonic price non-response bias that is common in survey equals the marginal value to consumers and techniques. Usually when participants are the marginal cost to suppliers. For non-mar- recruited, they do not have an indication of ket goods, it is not possible to use actual the nature of the experiment they will be hedonic pricing. However, hedonic pricing involved in. As a result, their willingness to can be used for non-market goods in an participate is unrelated to their attitude experimental setting. towards the product being studied (Fox et al., 1995) A general drawback is the fact that if Experimental auctions respondents do not fully understand the ques- Experimental auctions involve the use of real tions posed to them, the responses may not products and real incentives along with some conform to theoretical expectation. The eval- information on the different products being uation and the subsequent survey response to auctioned. By giving the participants repeated a given product will vary among individuals opportunities to participate in the auction due to the fact that individuals have different market, learning is enhanced; hence the par- perceptions about different products in addi- ticipants can show their real preferences for tion to varying levels of familiarity with a par- the products (Fox et al., 1995). The objective ticular product. of the auction is to elicit the value of the good Valuing non-market goods is difficult to the consumer. because there is no formal market to obtain Experimental auctions can be used as a price or other information relevant for eco- complement or alternative to elicitation meth- nomic analysis. While it is possible to simulate ods such as contingency valuation (CV) sur- the actual purchase of certain products in an veys and hedonic pricing. By pre-testing the experimental setting, it is not possible to simu- CV survey design, experimental auctions can late a market experience for issues such as be used ex ante to improve the design of con- food safety or environmental concern. Consumer - Chap 06 5/3/04 15:55 Page 70

70 S. Olubobokun and P.W.B. Phillips

A lot of non-market valuation research has the economic analysis of consumers’ per- been conducted in the area of food safety and ceived value of a GM trait the survey instru- environmental issues. For non-market goods, ment has to be based on an economic the absence of primary market data (e.g. model. An obvious application of this study demand curves) makes it impossible to esti- is as an economic background for a con- mate welfare changes. This has led to the use sumer survey instrument. Results from such of methods such as contingency valuation to surveys can then be used to test the model. estimate the demand for non-market goods. In surveying one has to be mindful of the welfare measure to be used. Generally, WTP applies to the perceived value of the Contingency valuation benefit to the consumer. If the survey WTP has been empirically measured using the instrument is to elicit WTP then welfare contingency valuation (CV) method. The CV measures such as option price, option method involves the use of surveys to elicit demand and certainty equivalence will consumers’ WTP for non-market goods con- apply. On the other hand, if the survey tingent on the specified scenario. Stated WTP instrument is to elicit WTP a premium values can serve as prices and can be used to above the market price, welfare measures determine the relationship between the deter- such as option value, compensating vari- minants of WTP. ation and equivalent variation will apply. A primary advantage of CV over hedonic pricing is the fact that CV is a flexible tool that can be tailored to analyse specific issues Concluding Comments (Buzby et al., 1995). Usually, CV results are comparable to other methods of valuing non- The price the individual is willing to pay for a market goods and are generally less expensive good represents the expected utility of the than actual market experiments. good. This value could be greater than, the CV relies on the subjective responses of same as or less than the market price for the consumers, which could be a source of bias. good. If a consumer perceives a health benefit Generally, elicited values inflate what con- from the consumption of a particular product, sumers would actually pay because consumers this will be demonstrated by a willingness to take hypothetical situations less seriously than pay a premium above the market price of the real-life situations. product subject to a budget constraint. The researcher should recognize the Attitudes of consumers are directly linked to potential for bias when using CV. Bias could the utility function. Generally, consumers will be minimized through careful phrasing and be willing to pay a positive premium for traits testing of scenario statements and WTP ques- they perceive to have positive health benefits, tions. When biases are present in CV studies, which adds to their utility derived from con- the WTP estimates may be overstated and the suming the product. Given information on the comparison of WTP values between different demand for the final product, farm supply of issues may be questionable. the agricultural product along with information In addition, care should be taken when on the type of the processing sector, the distri- interpreting and comparing the results from bution of economic benefits can be predicted CV studies. A distinction should be made for the adoption of a new GM crop. between measures of WTP above the market When consumer attitudes toward quality price (premium) and actual WTP values, change, the demand for such goods is which represent the price of the product. affected. Generally, consumers develop their attitudes towards the quality of a product based on perceptions of benefits and risks. Research Applications These perceptions are usually based on prior knowledge and experience in addition to Although many surveys are conducted on a information such as brand name, retailer rep- regular basis, in order to measure effectively utation and labelling. Consumer - Chap 06 5/3/04 15:55 Page 71

Measuring the Value of GM Traits 71

Producers need adequate consumer infor- tory labelling might be preferred. If the per- mation to be able to provide the quality assur- ceived quality difference between GM foods ance that consumers of GM foods need. If and non-GM foods is sufficiently small, the producers are able to provide the quality labelling and sorting costs of moving to a sep- assurance that consumers need, then produc- arating market situation could exceed the ben- ers can have a high level of confidence that efits, such that mandatory labelling might not the goods being produced will eventually be be required. Consumer preference studies purchased. could be used to document and support both The differentiation of consumer prefer- labelling requirements and import restrictions ences or the degree of perceived quality differ- on GM foods when there are public concerns ences have implications in the separation of based on possible, but yet unknown, environ- the markets of GM foods and non-GM foods mental and health risks. Given the growing through labelling. The superiority of a policy awareness of GM foods, exporting countries of labelled versus unlabelled GM foods will will have to be aware of the different con- depend on consumer perceptions of GM sumer segments in a region and understand foods in addition to the magnitude of the that the veto of acceptance in different coun- labelling and sorting costs. If the perceived tries needs to be carefully considered when quality difference is sufficiently large, obliga- introducing a new GM food or product.

References

Buzby, J.C. (1995) Using contingency valuation to value food safety: a case study of grapefruit and pesticide residues. In: Caswell, J. (ed.) Valuing Food Safety and Nutrition. Westview Press, Boulder, Colorado. Byerlee, D.R. (1981) Option demand and consumer surplus: comment. Quarterly Journal of Economics 85, 523–27. Chavas, J., Bishop, R.C. and Segerson, K. (1986) Ex ante consumer welfare evaluation in cost–benefit analysis. Journal of Environmental Economics and Management, 13, 255–568. Cicchetti, C.J. and Freeman III, A.M. (1981) Option demand and consumer surplus: further comment. Quarterly Journal of Economics 85, 528–539. Coursey, D.L. and Schulze, W.D. (1986) The application of laboratory experimental economics to the contingent valuation of public goods. Public Choice 49, 47–58. Environics International (2000) Global public perception of food biotechnology. Presented at the Convergence of Global Regulatory Affairs: Its Potential Impact on International Trade and Public Perception Conference. Ag-West Biotech Inc., Saskatoon, Canada. Epstein, L.G. (1975) A disaggregated analysis of consumer choice under uncertainty. Econometrica 43, 877–891. Fisher, A.C. and Hanemann, W.M. (1986) Option value and the extinction of species. Advances in Applied Micro-economics 4, 133–152. Fox, J.A., Shogren, J.F., Hayes, D.I. and Kliebensein, J.B. (1995) Experimental auctions to measure willingness to pay for food safety. In: Caswell, J. (ed.) Valuing Food Safety and Nutrition. Westview Press, Boulder, Colorado. Goering, P.A. (1985) Effects of product trial on consumer expectations, demand, and prices. Journal of Consumer Research 12, 74–82. Hobbs, J.E. and Plunkett, M.D. (1999) Genetically modified foods: consumer issues and role of information asymmetry. Canadian Journal of Agricultural Economics 47, 445–455. Hoyer, W.D. and MacInnis, D.J. (2001) Consumer Behavior. Houghton Mifflin Co., Boston, Massachusetts. Kuperis, P., Veeman, M. and Adamowicz, W.L. (1999) Consumers’ response to the potential use of bovine somatrophin in Canadian dairy production. Canadian Journal of Agricultural Economics 47, 151–163. Lancaster, K.J. (1966) The new approach to consumer theory. Journal of Political Economy 74, 132–157. Consumer - Chap 06 8/3/04 10:29 Page 72

72 S. Olubobokun and P.W.B. Phillips

Lindsay, C.M. (1969) Option demand and consumer’s surplus. Quarterly Journal of Economics 83, 344–346. Long, M.F. (1967) Collective consumption services of individual consumption goods: comment. Quarterly Journal of Economics 81, 351–352. Louviere, J.J. (1988) Analysing Decision Making: Metric Conjoint Analysis. Quantitative Applications in the Social Sciences, Vol. 67. Sage Publications, Thousand Oaks, California. Powell, D. (2000) Safe enough. Enhancing consumer conﬁdence in food production technologies. Unpublished report, University of Guelph, Canada. van Ravenswaay, E.O. (1995) Valuing food safety and nutrition: the research needs. In: Caswell, J. (ed.) Valuing Food Safety and Nutrition. Westview Press, Boulder, Colorado. Varian, H.R. (1978) Microeconomic Analysis. W.W. Norton and Co., New York. Weisbrod, B.A. (1964) Collective consumption services of individual-consumption goods. Quarterly Journal of Economics 78, 471–477. Yann Campbell Hoare Wheeler, (1999) Public attitudes towards biotechnology. Biotechnology Australia. Available at http://www.biotechnology.gov.au/library/content_library/BA_pYCHW.pdf Zeckhauser, R. (1969) Resource allocation with probabilistic individual preferences. American Economic Review 56, 546–552. Consumer - Chap 07 5/3/04 15:55 Page 73

7 Willingness to Pay for GM Food Labelling in New Zealand

William Kaye-Blake, Kathryn Bicknell and Charles Lamb Commerce Division, PO Box 84, Lincoln University, Canterbury 8150, New Zealand

Introduction Standards Council (ANZFSC) estimated that its labelling regime would entail start-up costs New Zealand is one of an estimated 28 coun- in New Zealand of NZ$43 million, with ongo- tries, plus the European Union (EU), that have ing costs of $42 million annually (ANZFA, adopted or planned to adopt labelling for 2001). The European Commission estimated genetically modified (GM) food (Phillips and that IP would increase grain farmgate prices McNeill, 2000). The specific legislation gov- by 6–17% (European Commission, 2000). erning labelling varies, so that each country’s The US Department of Agriculture (USDA) experience with labelling is likely to be unique. estimated that segregating non-biotech maize New EU regulations, for example, will require would add $0.22 per bushel (Lin et al., labels on food products containing GM soy- 2001–2002). Generally, the costs of segrega- bean oil and glucose syrup from maize starch tion and IP depend on the definition of GM (European Commission, 2001), products food: the more ingredients that need labelling which do not require labelling in New and the lower the tolerances allowed, the Zealand. In both places, products containing higher the costs (OECD, 2000; Wright, 2000; genetically modified organisms (GMOs) Lin et al., 2001–2002). An estimate of over- require labels, whereas the voluntary pro- all willingness to pay (WTP) for labelling gramme in the USA provides for labelling whether food is GM would provide some idea non-GM food (Phillips and McNeill, 2000). In of an economically efficient level of spending New Zealand’s case, the Australia New on labelling programmes, and thus on the Zealand Food Authority (ANZFA) requires that nature of IP systems and regulations that food sold in supermarkets be labelled if it con- might best suit consumers’ demand. tains GM ingredients. There is evidence to suggest that the gen- All labelling programmes, however, are eral level of support for mandatory labelling of likely to have something in common: added GM food is high. Interestingly, opinion polls costs. For example, as a result of the ANZFA indicate that a larger proportion of the popu- regulations, manufacturers conducted detailed lation desires labelling compared to the pro- surveys of the ingredients in their products portion that wishes to purchase non-GM (Robertson, 2002). The segregation or food. The European Commission (2000) Identity Preservation (IP) necessary for a trust- reported that nearly three-quarters of the worthy label would also add to the cost of pro- European population favours a clear labelling duction. The Australia New Zealand Food of GM food, versus just over one-half who

74 W. Kaye-Blake et al.

would pay more to avoid GM food. The (Gamble et al., 2000). Age has also appeared Commission also reported the results of vari- in some research as an important factor ous opinion polls conducted in the USA since affecting this demand (Sparks et al., 1994; 1995. They indicate that the level of support Gamble et al., 2000). Finally, price sensitivity for mandatory labelling amongst Americans is or the marginal utility of the food dollar is high and stable, ranging from 81% to 93%. important (Bredahl et al., 1998; Burton et However, only 43% are not likely to buy food al., 2001). enhanced through genetic engineering (GE) Research into consumer acceptance of GM (Wirthlin Quorum Poll, 2000, quoted in food has also focused on the importance of Campbell et al., 2000). prior attitudes. This research, often based in Economic efficiency, however, requires sociological or psychological theory, gauges more than a general level of support. If such factors as respondents’ beliefs about mandatory labelling implies higher costs, it is nature or government, or feelings about desirable to determine whether the public is chemicals in agriculture or food additives. In generally willing to pay for the information fact, these factors are often more important contained on the label. The objective of this than socioeconomic variables with which study is to begin to understand the key factors economists are more comfortable (Bredahl et that influence a consumer’s WTP for labelling al., 1998; Senauer, 2001). One behaviour of GM food, and to draw some initial conclu- that seems a good indicator for this ‘lifestyle’ sions about the magnitude of that WTP. factor is buying organically grown food. Burton et al. (2001), for example, successfully used respondents’ preferences for organically Factors Affecting the Demand for grown food to segment their sample. Labelling Discussions about labelling often raise the question of its usefulness (Valceschini, 1998; Demand for labelling the GM status of food Huffman et al., 2001). In particular, if con- has not been widely investigated. Discussions sumers do not know what genetic engineering of labelling have tended to focus on policy is, then the label is potentially meaningless. A alternatives, and demand for labelling is often closely related issue is whether consumers conflated with demand for non-GM food. actually read food labels. Reading labels has Therefore, guidance for the selection of vari- been shown to affect food consumption ables in the current study was sought from lit- (Nayga, 2001–2002), and so could be impor- erature on the demand for non-GM food. tant in determining which consumers care Prior research is not unanimous on which about GM labelling. economic or demographic characteristics There may be other factors affecting the influence attitudes towards GM food or WTP demand for labelling aside from those that for non-GM food. Usually important is gender affect the demand for non-GM food. – women are more likely to have negative atti- Presumably, consumers who wish to refuse tudes towards GM food than men (Sparks et GM food want the information to determine al., 1994; Anon., 2001; Cook, 2001; what food comes from GMOs and then make Gamble et al., 2000). It has been argued that their refusal. As noted before, however, sup- education also affects how people view GM port for labelling exceeds rejection of GM food, but empirical work indicates that the food. This is consistent with findings that con- effect is unclear (Couchman and Fink-Jensen, sumers want more information about GE 1990). Occupation may be important, with (Bredahl et al., 1998). Furthermore, con- those in higher-skilled occupations seeing sumers may be interested in purchasing an more benefits in the technology (Sparks et al., option regarding future consumption. They 1994). Although economic theory suggests may not currently be worried about GMOs, that income and WTP for labelling are likely but they may want to preserve their right to to be positively correlated, the results of refuse them should they receive new informa- empirical research on how income affects the tion (Gollier et al., 2000; Huffman et al., demand for non-GM food are inconclusive 2001). The complexity of consumer reactions Consumer - Chap 07 5/3/04 15:55 Page 75

Willingness to Pay for GM Food Labelling 75

to GM food therefore supports the idea of ‘strongly agree’. In addition, respondents examining the demand for labelling separately were questioned about food-related behaviour, from the demand for non-GM food. such as whether they bought organically grown food. Responses were again recorded on a five-point scale, from ‘never do’ to The Survey Data ‘always do’. Finally, demographic and economic data As part of a larger study that sought to were collected: age, income, occupation, edu- understand better how consumers perceive cational attainment and gender. and attempt to manage food risks, a total of 450 households in Christchurch, New Zealand, were personally interviewed. The Methodology respondents were adults aged 18 or over who were identified as those individuals pri- Respondents to the survey provided an indica- marily responsible for their households’ food tion of the strength of their demand for purchasing and/or food preparation. The labelling by answering questions as to whether sample was drawn by stratifying suburbs into they were willing to pay an additional 2%, 5% groups based on socioeconomic and property or 10% more for their groceries in order to valuation data, then proportionately allocat- know the GM status of their food. The results ing the total sample size across the groups represent the outcome of a decision between relative to the proportions of households in a finite set of alternatives, which generated a each group. Each interview took 30 to 60 discrete dependent variable. As the underlying minutes. After coding, editing and evaluation dependent variable is assumed to be continu- for completeness, there were 441 usable ous, but only a discrete response is observed, questionnaires. it is appropriate to analyse the data using a As part of the survey enumeration process, qualitative response model (Maddala, 1983; respondents were asked the weekly food Greene, 1993). expenditure for the household. Interviewers In this application, a respondent’s choice added 2% to the figure given and asked the falls into one of four categories that are natu- respondent if he or she would be willing to rally ordered. This gives rise to the following spend the resulting dollar amount on weekly latent regression: food expenditure in order to know whether β ε food had been genetically modified. The same y* = x + (1) procedure was followed for 5% and 10% of where y* refers to some unobserved measure the food bill. The order in which the respon- of the respondents’ WTP for labelling, β is a dent was asked about the different levels of vector of parameters that reflect the impact of spending was randomized. changes in the independent variables (x), and In addition, several attitudinal and behav- ε is an unobserved error term. While y* is not ioural questions were asked. At the beginning observed, the following four values of the of the interview, respondents were given the dependent variable (y) are observed: opportunity to describe the most important y = 0 (WTP = 0%) if y* < γ food-related issues facing their household. 0 y = 1 (WTP = 2%) if γ ≤ y* < γ Each respondent could provide up to three 0 1 y = 2 (WTP = 5%) if γ ≤ y* < γ responses, which were used to explore atti- 1 2 (2) y = 3 (WTP = 10%) if y* ≥ γ . tudes such as price consciousness that might 2 affect WTP for labelling. Attitudes were also Here γ is a vector of unknown parameters to assessed by asking respondents whether they be estimated along with β. agreed with a series of statements, for exam- If ε is assumed to be distributed logistically, ple: ‘I believe that there are definite benefits the parameters can be estimated using an to the consumer associated with GE/GMO ordered logit model. Under these circum- food’. They responded using a five-point stances, the probability that a respondent will Likert scale from ‘strongly disagree’ to answer 0% is given by the following: Consumer - Chap 07 5/3/04 15:55 Page 76

76 W. Kaye-Blake et al.

γ −β 0 x non-ordered) multinomial logit model. If the ==e Prob (response 0%) γ −β (3) β 1 + e 0 x data indicate that the parameters do, in fact, Similarly, vary across response categories, then it is γ −βγ −β 1 x 0 x likely that a multinomial logit would be a more ==e − e Prob (response 2%) γ −βγ −β (4) 11+ e 1 x + e 0 x appropriate modelling choice. This approach, however, loses the information contained in γ −βx γ −β x ==e 2 − e 1 (5) the ordered nature of the responses. Prob (response 5%) γ −βγ −β + 2 x + 1 x 11e e Tables 7.1 and 7.2 summarize the vari- γ −βx ==−e 2 ables included in the preferred ordered logit Prob (response11 0%) γ −β (6) 1 + e 2 x model. Dummy variables were used to indi- In the current study, the observed responses (y) cate price consciousness, label awareness and and the survey data (x) were used to estimate γ purchase practices regarding organically and β using a maximum likelihood procedure. grown food. Price conscious respondents A further assumption of the ordered logit is were identiﬁed from their ‘top-of-the-mind’ that the β parameters do not change for each answers to the survey’s open-ended question response category. The responses should rep- regarding the most important food issues fac- resent a true order, drawn from a single under- ing their household. The label variable was lying distribution. The effects of the used to indicate whether or not respondents explanatory variables should therefore be con- used ingredient or nutritional labelling. If the sistent across all categories. The test of parallel respondent either mostly or always purchased lines tests this assumption by comparing the organically grown food, then the variable results of the ordered logit with a general (i.e. ORG took on a value of 1.

Table 7.1. Deﬁnitions of variables. Variable Deﬁnition

L Strength of demand for labelling: none = 0, 2% = 1, 5% = 2, 10% = 3. PC = 1 if the respondent is price conscious, or 2 otherwise (reference category). RDL = 1 if the respondent reads nutrition or processing information, or 2 otherwise (reference category). ORG = 1 if the respondent mostly or always buys organic food, or 2 otherwise (reference category). GE = 1 if the respondent had some knowledge of GE/GM, or 2 otherwise (reference category). BEN In response to the question ‘I believe that there are definite benefits to the consumer associated with GE/GMO food’, = 1 if the respondent agrees or strongly agrees, 2 if disagrees or strongly disagrees, or 3 if neutral or no response (reference category). RSK In response to the question ‘I believe the risks to health from consuming GE/GMO food are low’, = 1 if the respondent agrees or strongly agrees, 2 if disagrees or strongly disagrees, 3 if neutral or no response (reference category). TRST In response to the question ‘I do not trust the large food manufacturing companies’, = 1 if the respondent agrees or strongly agrees, 2 if disagrees or strongly disagrees, 3 if neutral or no response (reference category). OCC = 1 if the respondent’s occupation is professional or managerial, = 2 if the respondent’s occupation is clerical, sales or service, = 3 if the respondent’s occupation is technical or engineering, = 4 if the respondent’s occupation is agricultural or farming, = 5 if the respondent’s occupation is student, = 6 if the respondent’s occupation is unemployed, receiving a benefit or not given, = 7 if the respondent’s occupation is tradesperson or labourer, = 8 if the respondent’s occupation is retired or housewife (reference category). UNI = 1 if the respondent has a university degree or bursary, university entrance or scholarship, or 2 otherwise (reference category). GEN = 1 if the respondent is female, or 2 otherwise (reference category). Consumer - Chap 07 5/3/04 15:55 Page 77

Willingness to Pay for GM Food Labelling 77

Table 7.2. Descriptive statistics for the sample (n = 441).

Percentage Variable Category Count of sample

L None 111 25.2 2% 107 24.3 5% 121 27.4 10% 102 23.1 PC Yes 130 29.5 No 311 70.5 RDL Mostly or always 237 53.7 Other 204 46.3 ORG Mostly or always 51 11.6 Other 390 88.4 GE Yes 360 81.6 No 81 18.4 BEN Agree 123 27.9 Disagree 165 37.4 Neutral/no response 153 34.7 RSK Agree 327 74.1 Disagree 37 8.4 Neutral/no response 77 17.5 TRST Agree 132 29.9 Disagree 135 30.6 Neutral/no response 174 39.5 OCC Professional/managerial 104 23.6 Clerical/sales/service 60 13.6 Technical/engineering 33 7.5 Agricultural/farming 6 1.4 Student 46 10.4 Unemployed/beneﬁciary/not given 29 6.6 Trades/labourer 46 10.4 Retired/housewife 117 26.5 UNI Degree or UE 172 39.0 No degree indicated 269 61.0 GEN Female 285 64.6 Male 156 35.4

In an attempt to measure familiarity with Since trust in institutions and the food sys- genetic modification, a dummy variable was tem has been highlighted as important in the included that indicated some degree of prior demand for non-GM foods (Sparks et al., knowledge. Specifically, respondents who 1994; Bredahl et al., 1998; Campbell et al., could provide a description of genetic engi- 2000; Bredahl, 2001; Cook, 2001), respon- neering or genetic modification, e.g. ‘gene dents were also asked whether they agreed transfer’ or ‘selective breeding’, were consid- with the statement, ‘I do not trust the large ered ‘knowledgeable’. This approach did not food manufacturing companies’. In an assess the quality of respondents’ knowledge, attempt to capture attitudes towards GM food, but such an assessment was considered respondents were asked whether they agreed beyond the scope of this research. However, with two statements: ‘I believe that there are the results are similar to others’ (Couchman definite benefits to the consumer associated and Fink-Jensen, 1990; Macer, 1992), in that with GE/GMO food’ and ‘Risks from consum- 81.6% of respondents were judged to be ing GE/GMO food are unknown’ (on GM knowledgeable. food and risk perception, see Bredahl, 2001). Consumer - Chap 07 5/3/04 15:55 Page 78

78 W. Kaye-Blake et al.

Demographic and economic data were Results also included in this analysis. A simplified educational indicator was developed: whether or The model specified above was estimated with not the respondent had a university degree the ordinal regression command (PLUM) in (this also included those whose highest qualifi- SPSS, using the logit link function. Table 7.3 cation was bursary, university entrance or contains the parameter estimates obtained, as scholarship, i.e. those who were on a univer- well as several statistics evaluating the model. sity track). Respondents were also grouped The significant chi-squared statistic indicates into several occupational categories. Finally, that the null hypothesis that all of the coeffi- gender was incorporated into the current cients of the explanatory variables are zero analysis with a dummy variable. can be rejected. However, the value of the The survey instrument included a question pseudo R2 is disappointingly low, suggesting on income. Unfortunately, a large number of that a large proportion of the variation in respondents refused to answer the question. WTP for labelling is not captured by the inde- Because preliminary analysis indicated that pendent variables. income was not a useful explanatory variable, Another measure of the model’s overall it was omitted from the final analysis. It should strength is its predictive power. Table 7.4 be noted, however, that the variable on price compares the actual responses with the consciousness may be important in reducing model’s predicted responses. The model pre- the impact of income, since lower-income dicted the correct category about 40% of the consumers tend to be more price sensitive time (compared to a naïve or random expec- (e.g. Jones, 1997). tation of 25%), and the average error between

Table 7.3. Estimated model.

Variable Category Parameter SE Signiﬁcance

γ 0 0.846 0.387 0.029 γ 1 2.109 0.397 0.000 γ 2 3.538 0.419 0.000 PC 0.347 0.200 0.083 RDL 0.302 0.186 0.104 ORG 0.897 0.290 0.002 GE 0.603 0.245 0.014 BEN Agree 0.087 0.229 0.705 (ref = neutral) Disagree 0.529 0.219 0.015 RSK Agree 0.772 0.250 0.002 (ref = neutral) Disagree 0.853 0.385 0.027 TRST Agree 0.162 0.222 0.465 (ref = neutral) Disagree 0.539 0.215 0.012 OCC Professional/managerial 0.549 0.260 0.035 (ref = retired/housewife) Clerical/sales/service 0.634 0.301 0.035 Technical/engineering 0.839 0.368 0.023 Agricultural/farming 0.330 0.775 0.671 Student 0.263 0.353 0.456 Unemployed/beneficiary/not given 0.599 0.401 0.135 Trades/labourer 0.759 0.329 0.021 UNI 0.455 0.201 0.024 GEN 0.518 0.192 0.007 McFadden’s pseudo R 2 0.080 Models compared Purpose Chi-squared df Significance Intercept-only vs. model Assess model fit 98.113 19 0.000 Ordered vs. general logit Test of parallel lines 30.447 38 0.803 Consumer - Chap 07 5/3/04 15:55 Page 79

Willingness to Pay for GM Food Labelling 79

Table 7.4. Predicted demand categories versus actual demand categories.

Actual demand categories Predicted categories None 2% 5% 10% Total

None 62 43 26 16 147 2% 18 9 10 6 43 5% 25 42 57 35 159 10% 6 13 28 45 92 Total 111 107 121 102 441 Correct % 55.9% 8.4% 47.1% 44.1% 39.2%

the predicted and actual categories was less labelling were being price conscious about than one category. The model’s predictive food, being involved in an agricultural occupa- ability varied by WTP category, however. In tion and agreeing that GE has definite bene- particular, the model correctly predicted only fits. Respondents who either agreed or 8.4% of the 2% WTP responses. disagreed that they trusted large companies, Finally, the test of parallel lines (Table 7.3) as opposed to those who were neutral or had indicates that the parameters do not vary by no opinion, were also less willing to pay for category, which confirms the choice of an labelling. Presumably, those who trust compa- ordered logit regression. nies do not see a need for labelling, and those The individual estimated parameters pro- who distrust large companies would not find vide a basic level of information: the direction the label trustworthy. Retirees and housewives of the relationship between the independent (the reference occupation category) were less and dependent variables. Parameters with willing to pay for labelling than other occupa- negative signs indicate that the variable is tional groups. associated with a lower WTP for labelling and At the other end of the scale, respondents positive signs indicate the opposite. were most likely to be willing to pay 10% However, the raw parameter does not more for food in order to have labelling if directly indicate the marginal effect of an they were consumers of organically grown explanatory variable. Whilst effects on the food. This variable was only a little more lowest and highest categories are clear (a important than those who disagreed that the positive sign indicates a shift out of the zero risks of GE are unknown. Respondents who category and a shift into the 10%), the effects were female and those who had a higher on the intermediate categories are uncertain degree of education tended to be more will- (Greene, 1993). Greene suggests directly ing to pay for labelling. Information about GE evaluating the magnitude of an explanatory was important: those who had some knowl- variable’s effect on the dependent variable by edge of GE or GM and those who thought calculating the changes in predicted cate- GE did not offer definite benefits were more gories that result from a change in the willing to pay for labelling. Occupationally, explanatory variable. For a dummy variable, those in paid employment (except in agricul- the model is evaluated twice, once with the ture) were more willing to pay for labelling, as variable in question set to 1 and once with it were students. set to 0, with all other variables set to their mean values. Table 7.5 gives the effects of the different explanatory variables on the pre- Discussion dicted categories. Respondents least likely to pay more for Some of the results of this analysis support labelling were unemployed or on public bene- earlier research. Women, for example, con- fits. Other factors decreasing the demand for sistently score lower on measures of the Consumer - Chap 07 5/3/04 15:55 Page 80

80 W. Kaye-Blake et al.

Table 7.5. Marginal effects of explanatory variables on predicted categories.

Demand for labelling

Variable Category None 2% 5% 10%

PC 0.062 0.025 0.035 0.052 RDL 0.052 0.023 0.028 0.047 ORG 0.125 0.089 0.045 0.169 GE 0.114 0.034 0.065 0.084 BEN Agree 0.015 0.007 0.008 0.013 (ref = neutral) Disagree 0.087 0.044 0.045 0.086 RSK Agree 0.146 0.044 0.082 0.108 (ref = neutral) Disagree 0.118 0.086 0.042 0.161 TRST Agree 0.028 0.012 0.016 0.025 (ref = neutral) Disagree 0.098 0.036 0.055 0.079 OCC Professional/managerial 0.087 0.049 0.042 0.093 (ref = retired/housewife) Clerical/sales/service 0.095 0.060 0.042 0.113 Technical/engineering 0.116 0.084 0.041 0.159 Agricultural/farming 0.062 0.020 0.035 0.047 Student 0.042 0.023 0.022 0.044 Unemployed/beneﬁciary/not given 0.117 0.029 0.067 0.079 Trades/labourer 0.109 0.075 0.043 0.140 UNI 0.076 0.037 0.040 0.073 GEN 0.093 0.036 0.051 0.077

acceptability of GM food; it is therefore How GM products should be labelled has unsurprising that they would be more willing been widely discussed. This research indicates to pay for labelling. Organic buyers, as that respondents who read either the nutri- Burton et al. (2001) have shown, are particu- tional or ingredient labelling are more willing larly likely to be willing to pay more for non- to pay for GM labelling. This finding suggests GM food, and the results presented above are that the ANZFA labels, which indicate the GM consistent with this. In addition, these results status of ingredients on the ingredient label, confirm that attitudinal variables are impor- make the information available to one group tant. This fact makes the job of predicting the of interested consumers in a relatively subtle demand for labelling more difficult, because manner. However, label-readers had a small attitudinal variables are difficult to collect and WTP compared to nearly every other charac- provide a less reliable basis for comparison teristic examined. Thus, some sort of promi- with a wider population. nent labelling may be desirable from a welfare Although preliminary analysis indicated perspective. It may also help producers that there was no significant relationship recoup more of the costs of segregation than between income and the demand for labelling, a less visible label. there seems to be a relationship between One of the interesting attitudinal questions WTP and what might be termed an income was whether or not respondents thought that constraint. Whether the respondent feels an the risks of GE were unknown. Those who income constraint, i.e. whether a respondent had an opinion on the risks seemed to be in feels as though there is disposable income to opposition to those who were neutral or did spend on labelling, seems to affect the not know. Those with opinions had approxi- demand for labelling. In the present model, mately the same WTP regardless of whether this is captured both by those who are price- they thought the risks were known or conscious shoppers and by those who are unknown. The results suggest that those who retired, housewives, unemployed or beneficia- think the risks are known think they are bad. ries, i.e. likely to be on a fixed income. They also suggest that those who believe the Consumer - Chap 07 5/3/04 15:55 Page 81

Willingness to Pay for GM Food Labelling 81

risks are unknown are estimating the risks, important as economic or demographic factors anyway. Furthermore, it supports the idea in determining consumers’ demand for that consumers are, in the absence of specific labelling. Willingness to pay for labelling is risk information, applying a sort of standard spread across the economic spectrum, risk discounting (Lamb et al., 2001). although consumers who feel their food bud- The survey results suggest that there is suf- gets are constrained are not as willing to pay. ficient overall WTP for labelling in New The results also suggest that more work Zealand to warrant the type of labelling man- should be done on identifying factors that dated by ANZFA. Consumption of food at affect demand for labelling. The current model home in New Zealand is approximately accounts for some variation in WTP, but more NZ$6.85 billion annually (Statistics New variation remains to be explained. Moreover, Zealand, 2001). If 23.1% of the population is the relationship between occupation, price willing to pay 10% more for food in order to consciousness and income should be explored have labelling, as was the proportion of the further. That being said, the results of this pre- sample, then aggregate WTP for this group is liminary analysis suggest that total WTP seems $158 million (assuming no covariance to be enough to warrant the labelling pro- between WTP for labelling and spending on gramme currently in place in New Zealand. food). If another 27.4% of the population is A future application of this methodology willing to pay 5% more for labelling, then this could involve specific GM food products or group’s aggregate WTP is $93.8 million. biotechnology processes. Although Bredahl Finally, if another 24.3% is willing to pay 2%, (2001) found that consumers tend to reject the aggregate figure is $33.3 million. The total the technology overall rather than evaluate WTP for the three categories is $285 million products on a case-by-case basis, survey annually. As the ANZFA labelling is estimated results indicate that the type of genetic engi- to cost approximately $42 million annually, a neering application affects its acceptability potential Pareto improvement is likely. (Couchman and Fink-Jensen, 1990; Macer, 1994; Sparks et al., 1994). In addition, ANZFA research found that the use of labels Conclusion varies across food categories (Gamble et al., 2000). The present research has examined Drawing firm conclusions from this research is GM food labelling generally, but the costs of problematic, given the low goodness of fit and labelling will vary by product and it is likely strength of predictions from the model. By and that the WTP for labelling will, too. Economic large, the results tend to support other similar efficiency may turn out to be more complex research. Specifically, attitudes are at least as than simply labelling all GM food.

References

Anonymous (2001) Women are concerned about genetically modified foods. Marketing to Women: Addressing Women and Women’s Sensibilities 14, 6. Australia New Zealand Food Authority (2001) Genetically modified foods. Available at http://www.anzfa.gov.au (accessed 29 April 2002). Bredahl, L. (2001) Determinants of consumer attitudes and purchase intentions with regard to genetically modified foods – results of a cross-national survey. Journal of Consumer Policy 24, 23–61. Bredahl, L., Grunert, K.G. and Frewer, L.J. (1998) Consumer attitudes and decision-making with regard to genetically engineered food products – a review of the literature and a presentation of models for future research. Journal of Consumer Policy 21, 251–277. Burton, M., Rigby, D., Young, T. and James, S. (2001) Consumer attitudes to genetically modified organisms in food in the UK. European Review of Agricultural Economics 28, 479–498. Campbell, H., Fitzgerald, R., Saunders, C. and Sivak, L. (2000) Strategic issues for GMOs in primary production: key economic drivers and emerging issues. CSAFE Discussion Paper No. 1. Centre for the Study of Agriculture, Food and Environment, University of Otago, Dunedin. Consumer - Chap 07 5/3/04 15:55 Page 82

82 W. Kaye-Blake et al.

Cook, A. (2001) New Zealand consumer reactions to GM food: studies of beliefs, attitudes and intentions to purchase. In: Proceedings of the Seventh Annual Conference of the New Zealand Agricultural and Resource Economics Society. NZARES, Christchurch. Couchman, P.K. and Fink-Jensen, K. (1990) Public Attitudes to Genetic Engineering in New Zealand. DSIR Crop Research Report No. 138. Department of Scientific and Industrial Research Crop Research, Christchurch. European Commission (2000) Economic Impacts of Genetically Modified Crops on the Agri-food Sector: A First Review. Directorate-General for Agriculture, Brussels. European Commission (2001) Commission improves rules on labelling and tracing of GMOs in Europe to enable freedom of choice and ensure environmental safety (press release). Available at http://europa.eu.int/comm/dgs/health_consumer/library/press/press172_en.pdf (accessed 25 July 2001). Gamble, J., Muggleston, S., Hedderley, D., Parminter, T. and Richardson-Harman, N. (2000) Genetic Engineering: The Public’s Point of View. Report to Stakeholders, HortResearch Client Report No. 2000/249, Mt Albert Research Centre, The Horticulture & Food Research Institute of New Zealand Ltd, February. HortResearch, Auckland. Gollier, C., Jullien, B. and Treich, N. (2000) Scientific Progress and irreversibility: an economic interpretation of the ‘Precautionary Principle’. Journal of Public Economics 75, 229–253. Greene, W.H. (1993) Econometric Analysis, 2nd edn. Maxwell Macmillan International Publishing Group, Sydney. Huffman, W.E., Shogren, J.F., Rousu, M. and Tegene, A. (2001) The value to consumers of GM food labels in a market with asymmetric information: evidence from experimental auctions. Paper presented at the annual meeting of the American Agricultural Economics Association, Chicago, 5–8 August. Jones, E. (1997) An analysis of consumer food shopping behavior using supermarket scanner data: differences by income and location. American Journal of Agricultural Economics 79, 1437–1443. Lamb, C., Mollenkof, D.A. and Ozanne, L.K. (2001) An exploratory look at New Zealand consumers’ perceptions of food risks. In: Chetty, S., and Collins, B. (eds) Bridging Marketing Theory and Practice. Proceedings of the 5th Australia New Zealand Marketing Academy Conference, 3–5 December Auckland, New Zealand. Lin, W., Price, G.K. and Allen, E. (2001–2002) StarLink™: where no Cry9C Corn should have gone before. Choices 16, 31–34. Macer, D.R.J. (1992) Attitudes to Genetic Engineerig: Japanese and International Comparisons. Eubios Ethics Institute, Christchurch. Macer, D.R.J. (1994) Bioethics for the People by the People. Eubios Ethics Institute, Christchurch. Maddala, G.S. (1983) Limited-dependent and Qualitative Variables in Econometrics. Cambridge University Press, New York. Nayga, R.M. Jr (2001–2002) Looking for the nutritional label: does it make a difference? Choices 16, 39–42. OECD (2000) Modern Biotechnology and Agricultural Markets: A Discussion of Selected Issues. Directorate for Food, Agriculture and Fisheries, Committee for Agriculture, OECD. Phillips, P.W.B. and McNeill, H. (2000) A survey of national labelling policies for GM foods. AgBioForum 3, 219–224. Robertson, D. (2002) Marking time: Australian rules on genetically modified food labels aren’t as tough as they’re made out to be. Far Eastern Economic Review 165, 41. Senauer, B. ( 2001) The food consumer in the 21st century: new research perspectives. Working paper 01–03, The Retail Food Industry Center, University of Minnesota. Available at http://trfic.umn.edu/. Sparks, P., Shepherd, R. and Frewer, L.J. (1994) Gene technology, food production, and public opinion: a UK study. Agriculture and Human Values 11, 19–28. Statistics New Zealand (2001) Household spending (year ended 30 June 2001) – standard tables. Wellington. Available at http://www.stats.govt.nz/ (accessed 31 May 2002). Valceschini, E. (1998) L’étiquetage obligatoire des aliments est-il la meilleure solution pour les consommateurs? Éléments de théorie économique. In: Organismes Génétiquement Modifiés à l’INRA: Environnement, Agriculture et Alimentation. Institut National de Recherches Agronomiques, Paris, pp. 111–115. Wright, J.C. (2000) The economics of genetic modification. Background paper for the (New Zealand) Royal Commission on Genetic Modification. Available at http://www.gmcommission.govt.nz/ (accessed 30 May 2002). Consumer - Chap 08 5/3/04 15:55 Page 83

8 Contingent Valuation of Breakfast Cereals Made of Non-biotech Ingredients

Wanki Moon1 and Siva K. Balasubramanian2 1Department of Agribusiness Economics, Southern Illinois University, Carbondale, IL 62901, USA; 2Department of Marketing, Southern Illinois University, Carbondale, IL 62901, USA

Introduction production and marketing system. For example, ERS (2000a,b) estimates that segregation Controversy over biotech foods continues to could add about $0.22 per bushel for maize be an issue of profound importance across the and $0.54 per bushel for soybean to market- globe to stakeholders involved in the food ing costs from country elevator to export ele- supply chain including farmers, grain han- vator. Moreover, labelling incurs tangible and dlers, food processors and retailers, regulatory intangible costs to the society in the forms of agencies as well as consumers. A number of regulatory requirements (e.g. testing, stan- major US and European food manufacturers dardization, certification and enforcement). and retailers announced that they would Hence, food supply chain participants would accept only non-biotech crops (Josling et al., want to ensure that market demand for non- 1999; Economic Research Service (ERS), biotech foods is sizeable enough to guarantee 2000a,b). This is coupled with the recent that market prices cover these costs. The key recall of taco shells made of StarlinkTM maize question then is whether instituting segregated which has stirred up a wave of turmoil in the markets for non-biotech foods generates ben- domestic food supply chain as well as in efits greater than those segregation costs. export markets. The uncertain prospect of Benefits associated with segregating and agrobiotechnology is in sharp contrast to the labelling non-biotech foods can be measured initial promise of agrobiotechnology as a by estimating whether and how much con- major technological breakthrough that would sumers would be willing to pay more for non- revolutionize the way crops are produced biotech foods. while enhancing the nutritional value of food Despite the heated debate over the needs products. Growing public concerns appear to of instituting segregation and labelling, we be altering the dynamic path of the progress have little knowledge about how consumers of agricultural biotechnology, raising such would behave if they are given the right to intriguing issues as adoption of identity choose between biotech and non-biotech preservation, market segregation, and foods. This lack of knowledge is detrimental labelling as ways of segregating GMOs from to shaping constructive dialogue among stake- non-GMOs throughout the food supply chain. holders involved in the food supply chain. In Identity preservation and market segrega- practice, the lack of information could criti- tion are not without additional costs to the cally disrupt the basic supply–demand rela-

84 W. Moon and S.K. Balasubramanian

tionships for major crops. For example, a speciality grains case, segregation costs large demand for non-biotech crops relative to include the additional costs of storage, han- the supply is likely to bring about substantial dling, risk management, analysis and testing, premiums for non-GMOs and deep discount and marketing (i.e. expenses associated with for GMOs (Babcock and Beghin, 1999). negotiating contract terms). Theoretically, government can serve market- At this time, several estimates are available facilitating functions such as provision of regarding the additional costs associated with information or regulations that would prevent segregating, certifying and testing of non- such market disruptions from taking place. GMO crops. ERS (2000a,b) estimates that However, if government fails to adapt rules segregation could add about $0.22 per bushel and regulations for biotech foods to evolving to marketing costs of non-GMO maize from consumer preferences, consumers will country elevator to export elevator. increasingly lose confidence in the credibility Segregation of non-GMO soybeans at these of the regulatory agency, which would deepen elevators could add $0.54 per bushel. These instability further in the food marketing sys- are averaged over 84 surveyed elevators. tem. Consequently, the lack of information in According to Bullock et al. (2000), non-GMO regard to the demand for non-biotech foods soybeans average a 50 cent premium per could undercut governmental regulations as bushel on the Tokyo market compared to well as market itself. GMO soybean. He found that most of the The primary objectives of this chapter are: premium resulted from the additional cost of (i) to estimate whether and how much con- segregating GMOs from non-GMO crops sumers would be willing to pay more to pur- within the grain-handling system but not from chase non-biotech foods, and (ii) present at the farmers’ level. empirical insights into the current debate con- A survey conducted in 1999 by Spark cerning the need for redesigning the food Companies disclosed that the non-GMO pre- supply chain to separate non-biotech from miums were estimated by a number of sources biotech crops, and (iii) evaluate how differ- at 10–15 cents per bushel for soybeans and ences in individual characteristics including 5–10 cents for maize (ERS, 2000a). The nationality and perceived attributes of agro- lower end of the premium ranges reflects less biotechnology affect individual valuation of the strict tolerance levels for GMO content. Good non-biotech attribute of foods. In recognition et al. (2000) conducted a survey to estimate of the new product or non-market property of additional marketing costs associated with non-biotech foods, a stated preference speciality grains in Illinois. The survey approach (i.e. contingent valuation) is used to reported an average additional handling cost measure consumer willingness to pay a pre- of 17 cents per bushel for maize and 48 cent mium for non-GMO foods using cross-sec- per bushel for soybeans in 1998. tional data collected in December 2000 in the In conjunction with the segregation USA and UK. throughout the food supply chain, an appropriate labelling system needs to be instituted to communicate the segregation and identity Segregation Costs and Labelling preservation to consumers, the final destination of the food supply chain. What type of Institutional underpinnings fundamental for labelling is used has important implications for the market for non-biotech foods to emerge all participants in the food supply chain. For include segregation or identity preservation example, while mandatory labelling of prod- throughout the food supply chain and ucts that use GMOs has the advantage of giv- labelling of final food products at food manu- ing consumers full information, it would be facturing and retailing levels. Segregation of over-regulating if there is a significant seg- non-GMOs from GMOs is essentially an ment of the population who do not have a extension of the handling process for special- preference for non-biotech foods. As a conse- ity grains and oilseeds, which has been in quence, while mandatory labelling benefits place for some time (ERS, 2000a,b). In the consumers, it may incur a substantial cost to Consumer - Chap 08 5/3/04 15:55 Page 85

Contingent Valuation of Breakfast Cereals 85

society. Besides, it can give the impression FDA approach facilitates consumers’ right to that biotech foods are not safe, potentially informed choices while restricting the scope of worsening the uncertain prospect of agro- the claims by the disclaimer. biotechnology. Voluntary labelling of products that do or do not use GMOs is desirable in the sense that Determination of Premium for it allows consumers to choose products consis- Non-GMO Crops tent with their preferences. That is, it relies on market forces to determine the acceptance of If non-biotech crops are segregated from new technologies. Voluntary labelling can be biotech crops throughout the food supply accompanied by a disclaimer that may be nec- chain and a labelling system is instituted to essary to prevent consumers from being mis- offer consumers the opportunities to choose led about safety differences (Caswell, 2000). between biotech and non-biotech foods in gro- For example, the Food and Drug cery stores, then prices may or may not be dif- Administration (FDA) chose this option in ferentiated between biotech and non-biotech regard to the marketing of dairy products from crops depending upon segregation costs and cows treated with supplemental rBST, allowing the strength of the demand for non-biotech the voluntary labels to include a statement: ‘No crops. Figures 8.1 and 8.2 show hypothetical signiﬁcant difference has been shown between situations concerning the supply of and milk derived from rBST-treated and non-rBST- demand for non-GMO crops and illustrate how treated cows’. Caswell (2000) notes that this premium is determined for non-GMO crops.

Price, segregation costs

Snon-GMO SGMOs

P2 Fixed segregation costs (SE) Quantity of GMO and non-GMO crops Premium Supply schedule of non-GMO

Pr2

Pr1 (SE)

D4 Q1 Q2 Quantity of non-GMO Premium = Price of non-GMO crops –

Price of GMO crops D2 D3

Fig. 8.1. Determination of premium for non-GMO crops when the costs of segregation are constant. Consumer - Chap 08 5/3/04 15:55 Page 86

86 W. Moon and S.K. Balasubramanian

Fixed segregation costs curve of non-biotechs is shown to be steeper than that of biotechs to account for the increas- Figure 8.1 illustrates the important role that ing segregation costs. The lower panel depicts the strength of the demand plays in determin- the supply schedules for segregation as a func- ing the size of segregation and premium for tion of segregation costs. As far as the demand non-biotech crops when there are some costs for non-GMOs lies below the vertical line of the of segregating non-biotech from biotech crops supply curve, the size of premium will be identi- and such costs do not vary across differing cal with the costs associated with segregation, levels of quantity. testing, and certifying. For example, if the The upper panel in Fig. 8.1 shows the sup- demand for non-GMO crops is represented by

ply schedules for biotech and non-biotech D0, the premium would be Pr1. Pr1 coincides crops with the vertical section indicating that with the costs of segregation. Once the the supply for non-biotech crops is limited by demand reaches the maximum the food supply

total acreage planted (Q2) in the short run (e.g. chain can offer, the strength of the demand for in a given year). Based on segregation costs in non-GMOs will be the only determinant of the the upper panel, the lower panel derives the size of premiums. A shift in demand for non-

supply schedule of segregation as a function of biotech crops from D1 to D2 will increase the segregation costs or premiums. In essence, the magnitude of premium from Pr2 to Pr3. lower panel describes three possibilities concerning the role of the demand for non-biotech crops in determining the size of segregation Discounts for biotech foods and premium for non-biotech crops. First, as long as the demand curve is posi- While we analysed several potential outcomes

tioned below D1, the food supply chain would in relation to the market for non-GMO crops, have no incentive to segregate non-biotech there could be parallel developments in the from biotech crops. No segregation would market for GMOs. On one hand, if the take place due to the insufﬁcient strength of demand for non-GMOs lies within the vertical

the demand for non-biotech crops. Second, section of the supply curve such as D2 in Fig. only part of non-biotech crops planted would 8.2, it suggests that there would be a surplus be segregated if the demand curve lies in the market for biotech crops. Any markets

between D1 and D3. For example, if the with a surplus would need to offer discounts demand for non-biotech crops is given by D2, to clear the market. Consequently, discounts the difference between non-biotech crops will emerge in the market for GMO foods. On

planted and demanded as represented by Q1 the other hand, if the demand for non-GMOs Q2 would be commingled with biotech is given by a relatively small magnitude such crops instead of being segregated. Last, when as D0, then, the amount of non-GMO crops the magnitude of the demand is larger than represented by Q3 Q2 would be commin- D3, all non-biotech crops planted would be gled with GMOs rather than being segregated. segregated and the size of premium would There would be no discounts for GMOs in begin to diverge with segregation costs as such a case. In sum, discounts for biotech

illustrated by Pr2 associated with D4. foods are likely to arise only if demand for non-GMO crops lies on the vertical section of the supply curve. Increasing segregation costs

Figure 8.2 exhibits the determination of pre- Survey Design and Data mium for non-biotech crops if segregating non- biotech from biotech crops incurs additional Given such a key role of the demand for non- costs and such costs increase as the scale of biotech crops in determining whether or not segregation expands. Analogous to Fig. 8.1, segregation is needed, a survey instrument the upper panel depicts the supply curves of was designed to shed light on the size of the biotech and non-biotech crops. The supply demand for non-biotech foods and to evaluate Consumer - Chap 08 5/3/04 15:55 Page 87

Contingent Valuation of Breakfast Cereals 87

public attitudes toward agrobiotechnology. The survey method using such an estab- The survey instrument is composed of two lished panel is called ‘permission-based survey’ sections: (i) measuring attitudes and percep- and is increasingly used in exploring various tions as related to agrobiotechnology issues, respects of consumer behaviour for academic and (ii) measuring behavioural intentions with or commercial purposes. Advantages associ- a focus on willingness to pay for breakfast ated with the permission-based survey include: cereals made of non-biotech ingredients. The (i) the response rate is higher than other regu- instrument was tested and question wording lar surveys, and (ii) demographic information was reﬁned in light of the results of three is disclosed for non-returners as well as focus group studies. A mail survey was con- returners, which would permit researchers to ducted in the USA and an online survey in the assess potential non-response bias. UK in December 2000 using household pan- Questionnaires were distributed to 5200 els maintained by the National Panel Diary households selected across the USA by ran- (NPD) group (a marketing consulting ﬁrm spe- dom sampling: about 3000 households cializing in research on consumer behaviour returned completed questionnaires, yielding a and food marketing). response rate of nearly 58%.

Price, segregation costs

SGMOs Snon-GMO

Segregation costs 1 = P2 – P1 P4 Segregation costs 2 = P4 – P3 Segregation P3 costs 2 (SE2) P2 Segregation

costs 1 (SE2) P1

Premium

Premium = Price of non-GMO crops –

Pr3 Price of GMO crops

Pr2

Pr1 (SE2)

SE1

Quantity of non-GMO Q1 Q2 Q3

D2 D1 D0

Fig. 8.2. Determination of premium for non-GMO crops when the costs of segregation increase across the scale of segregation. Consumer - Chap 08 5/3/04 15:55 Page 88

88 W. Moon and S.K. Balasubramanian

The sample is drawn stratified by geo- sus in most of the demographic categories. In graphic regions, market size, household head addition, Table 8.1 shows that the sample is age, education and income to balance with the drawn from four geographic regions US census for adults. Table 8.1 compares the (Northeast, Midwest/North Central, South and sample with the US census based on socioeco- West) quite in proportion with the US census. nomic and demographic profiles. The compar- The only noticeable discrepancy is the moder- ison suggests that the survey sample is ate under-representation of the ‘Not remarkably well representative of the US cen- Employed’ category in the sample as compared with the US census. The same instrument was administered to Table 8.1. Comparison of the survey sample with consumers in the UK. The survey in the UK the US census for adults. used an online method instead of mail. The Survey US census online method uses the Internet as a data collec- sample for adults tion tool and provides the following advantages (%) (%) over the more traditional methods (e.g. mail, Census region mall, telephone): (i) faster response time; (ii) Northeast 21.2 18.9 lower costs versus mall or mail/telephone; (iii) Midwest 28.0 23.6 ability to follow up with panelists quickly and South 34.1 35.8 West 16.8 21.8 inexpensively; and (iv) more current samples. Market size The online method could result in a bias <1,000,000 31.0 28.0 due to the fact that the sampling was 1,000,000–2,499,999 22.6 24.6 restricted to those who have access to a 2,500,000 22.5 28.4 computer and the Internet. A possible con- Non-MSA 23.8 19.0 jecture about the direction of the bias is that Household head age consumers with an access to the Internet would 18–24 years 3.7 6.4 25–34 years 17.9 21.4 be more technologically oriented and more 35–44 years 27.5 27.3 likely to accept new technologies than those 45–54 years 24.2 22.9 without access, leading to an upwardly biased 55–64 years 19.0 15.5 acceptance rate. Accordingly, the results from 65–69 years 7.7 6.6 the UK survey should be interpreted in light of Household head education the potential biases that may be associated with Grade 1.9 4.7 Some High 9.3 8.8 this particular sampling method. Graduated High 36.8 35.6 About 9000 consumers voluntarily regis- Some College 23.4 23.8 tered to participate in the NPD online survey Graduated 18.9 18.2 panel in the UK. For this survey, the NPD Post College 9.6 9.2 group sent all of the 9000 consumers e-mails Household head occupation providing a brief description of the survey and White collar 41.3 44.7 Blue collar 31.9 30.4 leading them to the website exhibiting the ques- Student 1.0 0.9 tionnaire. Nearly 2600 consumers returned Retired 20.5 23.5 completed surveys within the first 7 days after Other 5.2 0.5 they were contacted by the NPD via e-mails.1 Household income <$15,000 14.9 14.7 $15,000–$24,999 11.6 12.2 $25,000–$34,999 13.6 13.1 Contingent valuation of non-biotech foods: $35,000–$49,999 18.2 16.9 payment card format $50,000–$74,999 21.3 20.5 $75,000 + 20.4 22.7 Contingent valuation questions were included Source: NPD Genetically Modified Organism in the survey instrument to evaluate whether (GMOs) Consumer Survey, 2000. and how strongly consumers value the attribute

1 A little fewer than 1000 consumers responded on the ﬁrst day of opening the survey ﬁeld on the Internet. Consumer - Chap 08 5/3/04 15:55 Page 89

Contingent Valuation of Breakfast Cereals 89

that foods are made of non-biotech ingredients. includes various sizes of premium ranging The contingent valuation part asked respon- from $0.00 to $3.00 for a box of breakfast dents to imagine the following situation: cereals (with a base price of $4.00) made of non-biotech crops (see Table 8.2 for distribu- Suppose that you walk into a grocery story and want to buy a box of breakfast cereals. The tion of responses across the range of premi- grocery store carries breakfast cereals (e.g. corn ums). The payment card approach avoids the flakes, frosted flakes, corn pops) of two types: high rate of item non-response on open- (1) made from GMO crops, and (2) made from ended valuation questions. In this approach, conventional nonGMO crops. consumers are asked simply to go over the range of values and to circle the highest With this introduction to contingent valuation, amount they would be willing to pay. the next step is to determine the type of elicitation format. The most widely used methods of eliciting willingness to pay (WTP) were the open-ended and closed-ended (dichotomous Empirical Model Specification choice) questioning techniques. The open- Estimation procedure ended format tends to produce an unaccept- ably large number of non-responses or protest With the contingent valuation question gener- zero responses because of the cognitive diffi- ating value responses in the form of intervals culties associated with choosing a dollar rather than point valuations, mid-points as amount of the value for a public good. approximations to the true unobserved values Besides, it was often associated with strategic may be used in fitting a univariate distribution bias. The closed-ended format avoids these of values. The mid-points can also be used as problems by giving only two choices to the dependent variable in ordinary least respondents, although this format yields less square (OLS) regression. In consideration of information as compared to other formats. the fact that expected values within the inter- The payment card questioning technique vals are not necessarily equal to the interval has been increasingly used in recent years to mid-points, Cameron and Huppert (1989) compromise the advantages and disadvantages associated with the open-ended and closed- ended formats. The payment card method was Table 8.2. Distribution of responses for each size developed as an alternative to the bidding of premium in payment card format: USA and UK. game, the oldest and the most widely used elicitation method until recently (Mitchell and US UK Carson, 1989). The survey instrument for this Interval ($) frequency frequency study included contingent valuation questions 0.00–0.09 701 (22.9%) 433 (16.8 %) in the payment card format: 0.10–0.19 143 (4.67%) 97 (3.77%) 0.20–0.29 150 (4.90%) 118 (4.59%) Consumers might have to pay a higher price for 0.30–0.39 120 (3.92%) 214 (8.33%) nonGM foods due to the costs of segregation in 0.40–0.49 60 (1.96%) 122 (4.75%) the production and marketing system plus the 0.50–0.74 429 (14.0%) 178 (6.93%) additional costs of testing, certification and 0.75–0.99 83 (2.71%) 129 (8.91%) labeling GM foods. Suppose the price of 1.00–1.24 274 (8.95%) 127 (4.94%) breakfast cereals made from GM crops is $4.00 1.25–1.49 21 (0.68%) 66 (2.57%) per box. The price of conventional nonGM 1.50–1.74 58 (1.89%) 168 (6.54%) breakfast cereals will be higher than $4.00, but 1.75–1.99 11 (0.35%) 12 (0.46%) is not determined yet. What is the most above 2.00–2.49 48 (1.56%) 46 (1.79%) the current price of $4.00 you would be willing 2.50–2.99 29 (0.94%) 19 (0.73%) to pay to purchase a box of conventional 3.00 or higher 50 (1.63%) 181 (7.04%) nonGM breakfast cereals? ‘Don’t know’ 883 (28.8% 558 (21.7%) Contingent valuation questions in the form of Number of 3060 (100 %) 2568 (100%) observations a payment card contains an ordered set of threshold values (Cameron and Huppert, Source: NPD Genetically Modified Organisms 1989). The payment card for this study (GMOs) Consumer Survey, 2000. Consumer - Chap 08 5/3/04 15:55 Page 90

90 W. Moon and S.K. Balasubramanian

proposed the maximum likelihood (ML) pro- tance of price (PRICE_2) in food shopping is cedure for estimating WTP valuation equa- anticipated to have a negative effect on WTP. tions measured with interval data. The model The consumption frequency of organic food underlying the ML estimator is given by the products (ORG_3) is hypothesized to be posi- system (Stewart, 1983): tively related to WTP. * β The second subgroup (X ) involves eight Wi = X + Ei 2 * < perceived attributes in relation to agrobiotech- Wi = Pj1, if Pj1 Wi Pj, i = 1, 2, 3 … n; j = 1, 2, 3 … 14 (1) nology: (1) health risks; (2) environmental risks; (3) moral and ethical considerations; (4) where W * is the unobserved true WTP, X is a i the image of multinational corporations as the vector of explanatory variables that are primary beneficiaries of biotechnology; (5) the hypothesized to influence consumer willing- growing control of multinational corporations ness to pay a premium for non-biotech foods, over farming; (6) potential increase in yields; E is a random disturbance normally distrib- i (7) reduced use of chemicals in crop produc- uted; Wi is grouped observed WTP; and Pj represents observed threshold values for each tion; and (8) potential improvement in nutri- WTP category. The likelihood function depict- tional composition. The first four attributes ing the above model is given by describe negative aspects of agrobiotechnology while the latter three capture positive L = [Φ(P β X)/σ ) Φ(P – β X)/σ )]D (2) j j1 ij sides of agrobiotechnology. There were considerable correlations (rang- where Dij is 1 if Wi* falls in the jth category and 0 otherwise. The likelihood function can ing from 0.54 to 0.76) among consumer per- be maximized with respect to the vector of ceptions about the eight attributes of parameters (β ) using non-linear optimization agrobiotechnology. To cope with potential mul- algorithms. ticollinearity problems in specifying the empirical model, the eight perceived attributes were reduced to four variables: (i) an index of risk per- Model specification ception (RISK_1) using stated perceptions about health and environmental risks2; (ii) consumer In equation (1), consumers’ willingness to pay perception about moral and ethical aspects of a premium (W) was hypothesized to be deter- agrobiotechnology (MORAL_2); (iii) an index of mined by the vector (X). In our study, the vec- consumer perceptions about multinational cor- tor X is partitioned into four subgroups X = porations (MULTI_3) based on consumer per-

[X1, X2, X3, X4] including general attitudes ceptions about (4) and (5); and (iv) an index of toward food purchase, perceptions about beneﬁt perception (BENEFIT_4) based on con- attributes speciﬁc to agrobiotechnology, sumer perceptions about (5), (6) and (7).

awareness of the agrobiotechnology issues, The third subgroup (X3) refers to consumer and labelling preference. Table 8.3 presents awareness (AWARE) as measured with a six- the deﬁnitions and descriptions of the vari- point scale ranging from ‘Nothing’ to ‘A Great ables included in the empirical model along Deal’ to the question ‘how much have you read with summary statistics or heard about genetically modiﬁed organisms

The first subgroup (X1) includes attitudes (GMOs)?’ The fourth subgroup (X4) represents toward the importance of food safety consumer perceptions about labelling (LABEL) (SAFE_1) and price (PRICE_2) in food-pur- as reflected in response to the question ‘A chasing decisions and consumption frequency labelling system is necessary to differentiate of organic food products (ORG_3). We expect conventional foods from genetically modified that the importance of food safety (SAFE_1) foods on the supermarket shelf’. Respondents is positively associated with consumer WTP a were given a six-point Likert scale ranging from premium for non-biotech foods. The impor- ‘Disagree Completely’ to ‘Agree Completely’.

2 The index is constructed by adding up the two variables. Hence, the index would range theoretically from 2 to 12 with 12 representing complete agreement with the health and environmental risks of agrobiotechnology. Other indexes are similarly constructed. Consumer - Chap 08 5/3/04 15:55 Page 91

Contingent Valuation of Breakfast Cereals 91

Table 8.3. Deﬁnitions and descriptive statistics of variables used in the empirical model.

Variable Description Mean SD

X1 Subgroup SAFE_1 Importance of food safety in food purchasing decisions 4.89 (4.37) 1.13 (1.43) PRICE_2 Importance of food price in food purchasing decisions 4.95 (4.58) 1.09 (1.25) ORG_3 Consumption frequency of organically grown food 2.46 (2.86) 1.35 (1.48) products (1 = Never; 6 = All the Time)

X2 Subgroup RISK_1 6.89 (8.05) 2.64 (2.50) Health risks Biotech foods pose health hazards Environmental Agrobiotechnology poses hazards on ecosystem risks MORAL_2 It is morally and ethically wrong to use biotechnology 3.11 (3.28) 1.65 (1.75) MULTI_3 7.98 (9.12) 2.96 (2.59) Beneﬁciary Multinational corporations are primary beneﬁciaries of agrobiotechnology, while consumers assume most of the risks Control on Multinational corporations are increasingly controlling farming farming BENEFIT_4 11.8 (10.6) 3.33 (3.50) Increase in yields Agrobiotechnology reduces world food shortages by increasing yields Chemical use Agrobiotechnology reduces the use of chemicals in crop production Nutritional contents Agrobiotechnology enhances nutritional composition

X3 Subgroup AWARE How much have you heard about GMOs (1 = Nothing; 3.21 (3.85) 2.45 (2.11) 6 = A Great Deal)

X4 Subgroup LABEL Labelling is necessary to differentiate non-GMO from 4.88 (5.24) 1.34 (1.28) GMO foods in grocery stores

Unless otherwise indicated, questions are measured with a seven-point scale ranging from 1 = Disagree Completely; 6 = Agree Completely; 7 = Don’t Know. Respondents who selected Don’t Know are dropped from empirical estimations. Numbers in the parentheses represent means and standard deviations for UK responses.

Empirical Results tistics are 772, 134.6 and 573 for the pooled, US and UK data, respectively. Given Three models were estimated based on the the critical value (25.19) at a 0.01 probability likelihood function for interval data as shown level with 10 degrees of freedom, the in equation (2). The first model is based on hypotheses were decisively rejected across the pooled data between the USA and UK, three models, suggesting that the specified whereas the second and third models are esti- models have the capabilities to explain the mated with regard to the US and UK data, variation in the WTP a premium for non- respectively. Table 8.4 presents estimated biotech foods. parameters and asymptotic t-values for the The results from pooled data clearly indi- three models along with other summary statis- cate that both general food shopping habits tics. The hypothesis that all coefficients in the and and risk and benefit perceptions specific valuation models using payment card interval to agrobiotechnology play an important role data are simultaneously equal to zero was in determining behavioural intentions as mea- tested using χ2 statistics. The calculated χ2 sta- sured with WTP a premium to purchase non- Consumer - Chap 08 5/3/04 15:55 Page 92

92 W. Moon and S.K. Balasubramanian

Table 8.4. Maximum likelihood estimates for willingness to pay a premium for non-biotech foods: payment card interval data approach.

Pooled data USA UK

Variable Parameter Asymptotic Parameter Asymptotic Parameter Asymptotic name estimates t-values estimates t-ratios estimates t-ratios

Constant 0.0592 0.252 0.4391 1.428 0.2532 0.762 SAFE_1 0.0622** 2.936 0.0512* 1.630 0.0375 1.344 PRICE_2 0.1671*** 7.984 0.0591* 1.860 0.2174*** 7.938 ORG_3 0.1155*** 5.800 0.0665** 2.403 0.1292*** 4.750 RISK_1 0.0707*** 4.027 0.0560** 2.318 0.0843*** 3.509 MORAL_2 0.0565** 2.500 0.0017 0.049 0.0815** 2.792 MULTI_3 0.0295** 2.016 0.0336* 1.674 0.0437** 2.142 BENEFIT_4 0.0460*** 4.996 0.0191 1.540 0.0558*** 4.353 AWARE 0.0310 1.578 0.0604** 2.323 0.0085 0.031 LABEL 0.1134*** 4.430 0.0785** 2.229 0.1394*** 3.911 COUNTRYa 0.2648*** 4.613 – – – – σ 0.9908*** 45.41 0.882*** 28.65 1.02*** 35.196 Log L 4269.9 1672.5 2560.3 Log L (β = 0) 4656.1 1739.8 2846.8 χ 2b 772.4 134.6 573.0 No. of 1887 785 1102 observations

*P < 0.1; ** P < 0.05; and ***P < 0.01. aThe variable, COUNTRY, is a binary variable equal to one if the respondent is from the USA. bCritical value of chi-squared with 10 degrees of freedom is 25.19 at α = 0.01.

biotech foods. SAFE_1 had a positive effect (BENEFIT_4), they were not likely to pay a on WTP a premium indicating that if con- premium to purchase non-biotech foods. This sumers agree that food safety is an important result is significant in consideration of the fact consideration in making food purchasing deci- that consumers are not direct beneficiaries of sions, they were likely to pay a higher pre- those benefits and some of them are yet to be mium than those who do not agree. The realized. If the second-generation biotech food importance of price (PRICE_2) had a theoreti- products with substantive private and social cally correct sign, confirming the negative benefits are materialized and those benefits linkage between price sensitivity and WTP. As outweigh the risks posed by agrobiotechnol- hypothesized, consumption frequency of ogy, the magnitude of the market for non- organically grown food products (ORG_3) was biotech foods is clearly influenced. positively linked to the size of premium con- Along with risk and benefit perceptions, sumers would be willing to pay for non- moral and ethical considerations (MORAL_2) biotech foods. and perceptions about multinational corpora- As expected, risk and benefit perceptions tions had a measurable impact on shaping about agrobiotechnology proved to be major behavioural intentions with regard to purchas- factors shaping consumer willingness to pay a ing non-biotech foods. If consumers perceive premium for non-biotech foods. If consumers that it is ethically and morally wrong to use perceived risks on human health or environ- biotechnology to genetically modify food ments from the use of biotechnology in crop products, they tended to express more willing- and food production (RISK_1), then they ness to pay a premium for non-biotech break- were more likely to pay a premium to avoid fast cereals. Likewise, consumers were more consuming breakfast cereals made of biotech willing to pay a premium for non-biotech ingredients. In contrast, if consumers associ- foods if they believe that corporations are the ate agrobiotechnology with various benefits main beneficiaries from agrobiotechnology, Consumer - Chap 08 5/3/04 15:55 Page 93

Contingent Valuation of Breakfast Cereals 93

while consumers assume most of the risk or if a premium in both countries but had a statisti- they are concerned about the growing influ- cally significant effect on the USA consumers ence of multinational seed companies on only. Benefit perception (BENEFIT_4) was farming. These results support the notion that positively linked to WTP a premium in the consumer behaviours in regard to non-biotech USA but not statistically significant. This result food consumption are influenced not only by may be associated with the fact that the bene- direct and indirect risk and benefit perceptions fits are mostly yet to be materialized. In con- but also by other considerations including trast, it was a major factor driving down the moral values and beliefs about the social value size of premiums that UK consumers would of farming. be willing to pay for non-biotech foods. Consumer awareness of GMO issues (AWARE) was negatively associated with WTP but the association was not statistically signifi- Predicted mean WTP for breakfast cereals of cant. Labelling preference (LABEL) was posi- non-biotech ingredients tively and statistically significantly linked to the amount of premium consumers would be will- Using estimated parameters from the three ing to pay for non-biotech foods. If consumers models, fitted values (Wˆ ) were calculated con- agree that a labelling system is needed to seg- ditional on (X βˆ). First, based on the pooled regate non-biotech from biotech foods in gro- data, the US mean WTP a premium for non- cery stores, they were likely to pay a higher biotech breakfast cereals was $0.3134, while premium to purchase breakfast cereals made consumers in the UK on average were willing of non-biotech ingredients. Finally, consumers to pay a substantially higher premium in the USA (COUNTRY) were willing to pay a ($0.5782). While the model using pooled data significantly lower premium for non-biotech allows us to calculate mean WTPs for each foods as compared to the UK consumers, country based on the binary indicator (COUN- holding other factors constant. This result TRY), separate models would produce mean may be reflective of structural differences in WTPs more consistent with the preferences of attitudes toward foods between US and UK each country. Using separate models, the consumers. Further, food scares such as mad mean WTP was $0.3838 and $0.7316 for cow disease (BSE) and foot and mouth disease the US and UK consumers, respectively. that have swept European nations over the These comparisons indicate that the demand last few years may have made biotech foods a for non-biotech foods would be greater in the disproportionately sensitive issue. UK than in the USA. The mean size of pre- If consumers in the USA and UK struc- mium ($0.3838) for the US consumers is turally differ in perceiving risks, benefits and translated into about 9.5% of the base price other attributes of agrobiotechnology and in ($4.00) of breakfast cereals, while $0.7316 willingness to pay for non-biotech foods, pool- for the UK consumers is about 18% of the ing the data from both countries could gener- base price. ate misleading results. Therefore, this chapter These fitted values provide information splits the data into USA and UK subsets and useful in assessing the magnitude of potential estimates the valuation model for each coun- demand for non-biotech foods. The mean try. Comparison of the estimated results from size of premium of 9.5% in the USA suggests the two countries reveals important differ- that the strength of demand for non-biotech ences in terms of factors influencing behav- foods is neither immense nor negligible. As ioural intentions with regard to non-biotech discussed earlier, estimates of segregation foods. For example, moral and ethical consid- costs available to date range from 5 to 20% eration (MORAL_2) did not play an important per bushel for maize or soybean. Assuming role in shaping behavioural intentions with that the share of grain costs in manufacturing regard to non-biotech foods in the USA, a box of breakfast cereals is 30%, 20% of whereas it did for UK consumers. The degree additional segregation costs translate into a of awareness of the agrobiotechnology issues 6% increase in the price of the final products. (AWARE) was negatively associated with WTP The average size of premiums that certain Consumer - Chap 08 5/3/04 15:55 Page 94

94 W. Moon and S.K. Balasubramanian

segments of the US population would be will- respondents who selected the ‘Don’t Know’ ing to pay seems to be enough to cover addi- option to various questions formulate their tional costs associated with segregating attitude toward agrobiotechnology in the non-biotech from biotech foods. The com- future. Second, the explanatory variables parison indicates that a niche market for non- included in the WTP valuation model are biotech foods could emerge if consumers are mostly subjective perceptions about private or given the right to choose between biotech social risks and benefits associated with agro- and non-biotech foods. For the UK con- biotechnology. If consumer perceptions about sumers, the estimated mean size of premium various attributes of agrobiotechnology of 18% indicates that the strength of demand change over time, the estimates would change for non-biotech foods is considerably greater accordingly. More fundamentally, the esti- than that in the USA. mated size of market is linked to uncertainties inherent with the new gene splicing techniques in crop production. Paraphrasing, reli- Concluding Remarks able scientific evidence either confirming alleged risks or establishing the equivalence The WTP estimates are contingent upon sev- between biotech and non-biotech foods would eral factors that were built into the specifica- dramatically change consumer perceptions tion of the empirical models. First, the and consequently the strength of the demand estimates will differ depending on how the for non-biotech foods.

References

Babcock, B. and Beghin, J. (1999) Potential market for non-GMO corn and soybeans. Economic Perspectives on GMO Market Segregation. Staff Paper No. 298, Iowa State University, Ames, Iowa. Bullock, D.S., Desquilbet, M. and Nitsi, E.I. (2000) The Economics of Non-GMO Segregation and Identity Preservation. Working Paper, November 2000, Department of Agricultural and Consumer Economics, University of Illinois, Urbana-Champaign. Cameron, T.A. and Huppert, D.D. (1989) OLS versus ML estimation of non-market resource values with payment card interval data. Journal of Environmental Economics and Management 17, 230–246. Caswell, J.A. (2000) An evaluation of risk analysis as applied to agricultural biotechnology (with a case study of GMO labeling). Agribusiness: an International Journal 16, 115–123. Economic Research Service (ERS) (2000a) Biotechnology: U.S. grain handlers look ahead. Agricultural Outlook. US Department of Agriculture, Washington, DC. Economic Research Service (ERS) (2000b). Biotech corn and soybeans: changing markets and the government’s role. ERS Issues Center, US Department of Agriculture, Washington, DC Good, D., Bender, K. and Hill, L. (2000) Marketing of Speciality Corn and Soybean Crops. Value Project Publication AE-4733, March 2000, University of Illinois, Urbana-Champaign. Josling, T., Unnevehr, L., Hill, L. and Cunningham, C. (1999) Possible policy and market outcomes in WTO 2000. The Economics and Politics of Genetically Modiﬁed Organisms in Agriculture: Implications for WTO 2000. Bulletin 809, University of Illinois, Urbana-Champaign. Mitchell, R. and Carson, R.T. (1989) Using Surveys to Value Public Goods: the Contingent Valuation Method. Resources for the Future, Washington, DC. Stewart, M.B. (1983) On least squares estimation when the dependent variable is grouped. Review of Economic Studies 50, 737–753. Consumer - Chap 09 5/3/04 15:55 Page 95

9 A Comparative Analysis of Consumer Acceptance of GM Foods in Norway and the USA1

Wen S. Chern1 and Kyrre Rickertsen2 1Department of Agricultural, Environmental, and Development Economics, The Ohio State University, Columbus, OH 43210-1067, USA; 2Department of Economics and Social Sciences, Agricultural University of Norway, Ås, Norway

Introduction rapidly in the US (Darr, 2001; Darr and Chern, 2002). However, over this short time Genetically modified organisms (GMOs) have period, the use of these GM products has been been developed from advanced biotechnology regulated in the EU, Japan and other coun- to achieve certain desirable traits in agricultural tries. In 1997, the EU imposed a mandatory production such as weed and pest resistance. labelling of GM foods with a 1% tolerance However, the adoption of genetically modified level, while Japan followed suit in 2001 with a (GM) crops has been controversial because of 5% GM content limit. The debates on con- the opposition by some consumer and environ- sumer acceptance and labelling regulations mental groups. If there are no direct tangible have attracted much interest concerning con- benefits to the consumer, the foods produced sumer attitudes toward GMOs and GM foods. with GMO ingredients may be perceived as There have been several consumer surveys being inferior to their non-GM counterparts. conducted in the USA (Hoban, 1999; This perception is critical for the acceptance of Hallman and Metcalfe, 2001; Moon and GM foods (Caulder, 1998; Hoban, 1998; Balasubramanian, 2001; Mendenhall and Hallman and Metcalfe, 2001). There have been Evenson, 2002), Europe (Boccaletti and Moro, concerns about the consumer’s acceptance of 2000, for Italy; Burton et al., 2001, for the GM foods in many countries of the world such UK; Spetsidis and Schamel, 2001, for as those in the European Union (EU) and Germany; and Verdurme et al., 2001, for Japan, as no food manufacturers have dared to Belgium) and Japan (Macer and Ng, 2000; Ng test the markets with specifically labelled GM et al., 2000). Most of these studies are foods under the mandatory labelling regulations. descriptive in nature and few deal with the esti- Since the first commercialization of GM mation of the willingness to pay (WTP) for GM crops in 1996, the adoption of Roundup foods. Moon and Balasubramanian (2001) Ready soybeans and Bt maize has increased estimated the WTP for breakfast cereals made

1 The authors would like to thank Arild Skogmo, Frode Alfnes and Dadi Kristofersson for help with the Norwegian survey, and Naoya Kaneko, Lewis Horner and the Center for Survey Research of The Ohio State University for research assistance and funding support for the US survey. © CAB International 2004. Consumer Acceptance of Genetically Modiﬁed Food (eds R.E. Evenson and V. Santaniello) 95 Consumer - Chap 09 5/3/04 15:55 Page 96

96 W.S. Chern and K. Rickertsen

of non-GM ingredients in the USA and the tives and regulations do not come into effect UK. But their samples are not representative. until the member states enact them as Boccaletti and Moro (2000) also attempted to national laws. Some member states also want quantify the WTP for generic GM products to go beyond the base requirements. For with different hypothetical attributes in Italy example, Austria prefers a ban on GM foods and Burton et al. (2001) calculated the WTP (Phillips and McNeill, 2000). for such products in the UK. Our study Most recently, on 3 July 2002, the attempts to extend these previous works to European Parliament backed the European design a survey instrument for eliciting the Commission’s two new proposals on GMOs WTP for different GM foods based on national which will establish ‘a sound community sys- representative samples. Speciﬁcally, since tem to trace and label GMOs and to regulate 2000, a joint research project has been under- the placing on the market and labelling of taken to conduct a multi-country analysis on food and feed products derived from GMOs’ consumer attitudes toward GM foods and on (European Commission, 2002). The propos- eliciting the consumer’s WTP for GM versus als ‘will require the traceability of GMOs non-GM foods in Japan, Norway, Taiwan and throughout the chain from farm to table and the USA. In this chapter, we only report the provide consumers with information by results from pilot national telephone surveys labelling all food and feed consisting of, con- using a revised uniform questionnaire containing or produced from a GMO’ (European ducted in Norway and the USA in 2002. Commission, 2002). Under these proposed For the remainder of this chapter, we will regulations, the most notable changes from ﬁrst compare the GM food regulations the existing provisions include that products between the EU and the USA. We will then such as soybean and rapeseed oils which were present the survey results and compare the exempted from the existing labelling regula- estimated WTP for GM soybean oil, salmon tion due to their undetectibility of the pres- fed with GM soybeans and GM salmon. ence of transgenic DNA or protein, are now subject to the newly proposed labelling regulation. Furthermore, the feeds containing Consumer Concerns and Labelling GMOs such as GM soy meal are required to be labelled as such. However, the Parliament There is substantial resistance to GM crops in rejected the more extreme amendments that Europe and other parts of the world. Consumer would have required labelling of products like organizations have expressed concerns regard- meats or eggs produced with GM feed as GM ing antibiotic-resistant marker genes, potential products. The proposals were later adopted allergic reactions, ethical and religious concerns, by the European Council in 2002. and the lack of consumer choice due to inade- Norway is a member of the European quate labelling (Franks, 1999). Economic Space and is in many cases bound by Most national labelling systems are still EU directives and regulations. Norway has under development and different countries adopted somewhat stricter requirements than have taken different approaches. As noted those established in the EU’s Novel Food earlier, the EU has imposed mandatory Regulation. One major difference is that labelling systems. In the EU, a number of labelling is mandatory even if GM foods do not directives set the framework for the labelling differ. Due to consumer opposition, none of the systems in the member states. Directive major Norwegian food retailers sells GM foods. 90/220 from 1990 establishes requirements In the USA, the government made a deci- for labelling GM crop varieties for seeds, the sion on 3 May 2000 to reject biofood labels Novel Food Regulation 258/97 from 1997 on the grounds that from a health and safety sets a 1% tolerance level for whole and standpoint, these foods do not differ from their processed foods, and Regulation 1139/98 conventional counterparts. Since GM foods from 1998 covers GM varieties of maize and such as GM soybeans are nutritionally equiva- soybeans that were released before Regulation lent to the conventional ones, the Food and 258/97 was adopted. However, the EU direc- Drug Administration (FDA) does not require Consumer - Chap 09 5/3/04 15:55 Page 97

Acceptance of GM Foods in Norway and the USA 97

labelling of GM foods (Vogt and Parish, 1999). important features of the survey is that we did So far, there has not been notable consumer not assume a priori that GM foods are inferior opposition to GM foods in grocery stores to the conventional counterparts. Therefore, despite opposition from consumer groups who the respondents could state their preferences have accused the US government of catering for GM foods even if they were priced higher to the biotechnology industry and ignoring the than the conventional products. A copy of the consumer’s right to know. questionnaire is available upon request. The nationwide US survey consisted of 250 respondents aged 18 and over. The survey The Survey was conducted by telephone with the random digit dialling method. Even though the sample Two telephone surveys were conducted during was relatively small, it covered 43 states in the March and April 2002 in Norway and the continental USA. This pilot survey was funded USA. We asked similar questions, however, and conducted by the Center for Survey the surveys were conducted in different lan- Research (CSR) of The Ohio State University. guages creating some differences regarding We went through many rounds of revisions. the exact wording of questions. Some ques- Among them was a pre-test by a group of tions from the original English questionnaire graduate students. The final version was given were also omitted from the Norwegian survey. to the CSR for conversion to a telephone For example, adjectives like ‘extremely’ were interview format. The CSR also conducted toned down in the Norwegian translation and another pre-test and the feedback was used to questions concerning ‘race’ or ‘religion’ change some of the questions and wordings. (Norway is 95% white and Protestant) were The US survey was conducted within a 3-week omitted. We have used the US wordings of period in April 2002, with a mix of daytimes the alternatives and questions in the tables and evenings. The overall response rate was presented later. There are many advantages 28.7% while the cooperation rate (among of doing a telephone survey. One is that the households in which we could speak with the alternative choices of several questions can be eligible respondent) was 80.6%. Average age randomly selected for each interview. The of the US survey respondents is 47 while 77% interviewers were trained to answer questions are females. Note that in the US survey, we of the respondents and thus the quality of the required the respondents to be a food shopper responses should be higher than a typical mail in the household. There are 4.3% of the survey. With the random digit dialling method respondents who are vegetarians. used in the US survey, we can reach the The Norwegian survey was conducted by entire country very easily. Skogmo (2002) and the Norwegian results The survey contained seven sections and from the public survey are based on his included many questions on general food pur- results. In the Norwegian survey, 100 respon- chasing habits, attitude and perception dents aged 18 and over living in Oslo (the towards GM foods, information sources, capital) and 100 respondents living in knowledge, willingness to buy GM foods with Nordland (a county without any major cities in varying attributes, labelling regulation, contin- the northern part of Norway) were randomly gent valuation (CV) for GM versus non-GM selected from the telephone directory and food products, and demographic information. interviewed. The directory covers about 97% The US survey contained 61 questions while of Norwegian households. The sample con- the Norwegian survey had fewer questions sists of 46% male and 54% female respon- because it included only soybean oil and dents. The average age of the respondents salmon, but not cornflake cereal as in the US was 49 years or about 4 years above the survey. The CV part of the survey consisted of national average for the age group 20–80 sequential closed-ended binary choice ques- years. The high mean age was partly a result tions (Carson and Mitchell, 1995). The specific of 40% of the interviews being conducted dur- design of these WTP questions will be ing daytime when many retired people described in more detail later. One of the most answered the telephone. Furthermore, four Consumer - Chap 09 5/3/04 15:55 Page 98

98 W.S. Chern and K. Rickertsen

out of five calls were rejected, pointing to a Less than a third of Norwegian (30.5%) potential self-selection problem with less par- and 43.0% of American respondents ticipation among people with valuable time. claimed that they were willing to consume foods produced with GM ingredients. The American resistance is unexpected given that Comparison of Survey Results about 70% of the foods on the retail food store shelves are said to contain some form The results in Table 9.1 show that about 45% of GMO ingredient in the USA (Kinsey, of the respondents considered themselves ‘not 2001). A larger proportion of the informed’ and about 45% considered them- Norwegian than the US respondents were selves ‘somewhat informed ‘ about GM foods either ‘extremely unwilling’ (45.5%) or more or GMOs. A somewhat larger percentage of surprisingly ‘extremely willing’ (13.0%) to Americans (14.1%) than Norwegians (8.0%) consume GM foods. claimed to be ‘very well informed’. The opposition against GM foods was The high proportions of not informed reduced when some benefits associated with respondents correspond well with the propor- them were explicitly mentioned in the question of correct answers to our two knowledge tions suggesting that GM foods can grow in statements. Only 37.5% of the Norwegian and popularity when consumers become aware of 43.8% of the American respondents thought it the potential benefits. Benefits offered in our was false that ‘Non-genetically modified soy- surveys were reduced use of pesticides, beans do not contain genes while genetically improved nutritional qualities or lower price. modified soybeans do’ while 36.0% of Close to 40% of Norwegians and around Norwegians and 61.3% of Americans believed 70% of Americans were willing to consume it was false that ‘By eating GM foods, a per- GM foods conditional on those benefits. son’s genes could be altered’. The responses to When we asked which of these potential ben- the knowledge questions are used to identify the efits was the most important about 65% of knowledgeable consumers as those answering the Norwegian and 55% of the American both questions correctly. These questions pro- respondents answered reduced use of pesti- vide important explanatory variables used in the cides and below 10% answered reduced econometric models discussed later. price. More than half of Norwegians found The results in Table 9.2 show that a major- reduced price to be ‘extremely unimportant’ ity of Norwegians (59.5%) and close to half of for their decision to buy or not to buy GM Americans (48.9%) believed that GM foods foods. The insensitivity to price may be were risky to human health while 23.5% of caused by the hypothetical nature of the Norwegians and 20.7% of Americans thought choice (i.e. no real goods or payments) as dis- they were safe. A third of the Norwegians cussed in much of the experimental econom- considered them extremely risky. ics literature (e.g. List and Shogren, 1998).

Table 9.1. Consumer information and knowledge, percentage distribution for each question.

Question Alternative Norway USA

Before this survey, how well were you informed Very well 8.0 14.1 about GM foods or organisms? Somewhat 45.0 41.0 Not informed 47.0 44.9 Non-genetically modiﬁed soybeans do not contain True 16.0 23.4 genes while genetically modiﬁed soybeans do. False 37.5 43.8 Don’t know 46.5 32.8 By eating GM foods, a person’s genes could be True 28.0 22.3 altered. False 36.0 61.3 Don’t know 36.0 16.4 Consumer - Chap 09 5/3/04 15:55 Page 99

Acceptance of GM Foods in Norway and the USA 99 Alternatives Extremely Somewhat NeitherDon’t Somewhat Extremely Consumer attitudes toward GM foods, percentage distribution for each question. Table 9.2. Table QuestionHow risky would you say GM foods are in terms of risk to human health? 1, 2 = Risky and 4, 5 Safe NorwayHow willing are you to consume foods produced with GM ingredients? 1, 2 = Willing and 4, 5 Unwilling 33.5 NorwayHow willing would you be to consume GM foods if they reduced the amount of pesticides applied to crops? 1, 2 = Willing and 4, 5 Unwilling 13.0How willing would you be to purchase GM foods if they were 26.0 Norwaymore nutritious than similar foods that are not GM? Norway1, 2 = Willing and 4, 5 UnwillingHow important is the price factor when you decide whether or 17.5 17.0not to buy GM foods? Norway 17.5 8.01, 2 = Important and 4, 5 UnimportantHow willing would you be to purchase GM foods if it posed USA 16.0 21.5a risk of causing allergic reactions for some people? USA 4.0 19.51, 2 = Willing and 4, 5 Unwilling Norway 13.0How important are ethical or religious concerns when you Countrydecide whether or not to consume GM foods? USA 9.4 20.01, 2 = Important and 4, 5 Unimportant 1.5 18.0 9.5 Norway 4.7How important is it to you that food products are speciﬁcally (1) 10.5 7.5labelled as GM or non-GM? USA Norway USA1, 2 = Important and 4, 5 Unimportant 13.7 39.5 21.5 6.0 38.3 11.5 45.5 9.0 8.5 10.0 (2) 94.0 18.0 54.7 29.7 16.0 2.0 8.0 35.5 USA 13.7 7.0 USA 2.0 39.0 4.5 53.9 (3) 37.5 15.2 5.0 9.4 3.5 USA 23.8 50.5 6.5 3.5 12.5 4.0 0.5 5.1 (4) 11.3 7.0 5.5 0.5 16.4 58.6 2.5 21.5 23.8 83.5 14.5 0.0 9.4 (5) 12.1 3.1 9.0 28.5 62.5 0.5 15.2 5.9 know 10.9 2.0 1.0 12.5 2.0 4.3 18.0 26.2 2.7 0.0 1.2 28.9 5.9 41.4 1.6 1.6 1.6 1.2 Consumer - Chap 09 5/3/04 15:55 Page 100

100 W.S. Chern and K. Rickertsen

We also asked about some potential sources for sale. Nevertheless, there is a considerable of concern. More than 80% of Norwegians and interest for consumers’ acceptance and WTP 40% of Americans were ‘extremely unwilling’ to for these qualities in the aquaculture sector. purchase GM foods if it posed a risk of causing To calculate WTP we use a stated choice allergic reaction for some people. Only 10.0% method (SCM), which is based upon buyers’ of Norwegians and 25.0% of Americans were hypothetical choice for GM food purchases. willing to take such a risk. Ethical and religious We used a simple design developed for the concerns were important for 29.5% of telephone survey and the only attributes Norwegians and 36.3% of Americans while included were prices of GM versus non-GM such concerns were ‘extremely unimportant’ for implying that attributes like reduced use of as much as 62.5% of Norwegians and 28.9% pesticides or improved nutritional values were of Americans. not considered. Note that in the survey we The majority of Norwegian (98.5%) and defined non-GM products as those with GM American (87.1%) consumers demand content of less than 3% since, as stated, ‘it is labelling. These results are in line with the nearly impossible to ensure 100% purity’. results in the Eurobarometer (European A disadvantage of the SCM (and other Commission, 2001) where 94.6% of the stated preference methods) is that peoples’ 16,029 respondents in the 15 member states behaviour in a hypothetical setting may not of the EU wanted to have the right to choose fully reflect actual behaviour, i.e. the respon- between GM and non-GM foods. Support for dents may not act on their stated choices. labelling was reduced when the respondents However, given that none of the GM qualities are reminded that labelling may increase food of salmon are available, we could not use prices, however, 55% of Norwegians support experimental auctions or other incentive com- labelling even if prices were increased by 5% patible techniques. or more. The insensitivity to price may again In the Norwegian survey, we had two alter- be partly explained by the hypothetical nature natives of soybean oil and three alternatives of of the question. salmon. The choice experiment consisted of The results indicate more favourable atti- two steps and each step consisted of one tudes to GM foods in the USA than in binary choice for soybean oil and two binary Norway; however, the opinions in the USA choices for salmon. In step one, we asked the are also quite mixed. This general conclusion respondents if they would choose: (i) non-GM is consistent with Priest (2000) who found or GM-fed salmon, (ii) non-GM or GM that the USA increasingly resembles Europe salmon, and (iii) non-GM or GM soybean oil in having significant amounts of reservation given identical prices for each of the two towards biotechnology. choices. In the US survey, we also included cornflake cereal. Therefore, we had four versions of the questionnaire with each including Methodology only two of the three products at a time. In this chapter, we analyse only the results per- The main objective of this study is to estimate taining to soybean oil and salmon. the willingness to pay for different qualities of The base prices we used reflected prices soybean oil (non-GM and GM) and salmon found for the non-GM products in stores (i.e. (non-GM, GM-fed and GM). GM foods are not located in Oslo, Norway and Columbus, sold in Norway but GM soybean oil is com- USA). The percentage distributions of the monly sold in the USA. Salmon can potentially respondents’ choices are shown in Table 9.3. be fed with GM soybeans (GM-fed salmon) and More than 80% of Norwegians chose the a GM salmon has been developed by the non-GM alternative for each of the three Canadian company Genesis. The GM salmon choices. For the American respondents, grows faster than wild salmon (but not neces- 45.1% chose non-GM soybean oil, 59.2% sarily faster than farmed salmon) and the feed- chose non-GM salmon (over GM-fed) and ing costs are lower (Aftenposten, 9 September 68.9% chose non-GM salmon (over GM 2001). None of these qualities of salmon is yet salmon). Not for any of the choices did more Consumer - Chap 09 5/3/04 15:55 Page 101

Acceptance of GM Foods in Norway and the USA 101

Table 9.3. Stated choices at identical prices, percentage distribution for each choice.

First choice

Choices Country Non-GM GM GM-fed Indifferent None Don’t know

Salmon Norway 81.8 1.0 8.6 8.1 0.5 Non-GM/GM-fed USA 59.2 6.5 24.9 8.3 1.2 Salmon Norway 86.4 1.0 4.0 7.6 1.0 Non-GM/GM USA 68.9 3.6 21.0 5.4 1.2 Soybean oil Norway 85.4 2.5 7.0 4.5 0.5 Non-GM/GM USA 45.1 8.7 24.9 19.1 2.3

than 10% of the respondents prefer a GM normalized to zero. Letting the scale parame- product but in the USA close to a quarter of ter µ = 1, the probability of choosing alterna- the respondents are indifferent between the tive i for respondent n is estimated by the GM and non-GM alternatives. For estimation, logit model: V the choices of indifferent respondents are π = e in n()i weighted with a half on each of the two indif- ∑ eVjn (2) ferent alternatives. j In step two, each respondent was given For soybean oil we use a binary (i = 1 is non- the same choices as in step one but offered GM and i = 2 is GM) and for salmon a multin- price reductions for the commodity he did not omial (i = 1 is non-GM, i = 2 is GM-fed and i choose. The price reductions were in the = 3 is GM) logit model. interval 5–50% for GM soybean oil and GM- The estimated parameters can be com- fed salmon and 10–60% for GM salmon. bined to identify monetary values associated Respondents that were indifferent between with changes in each attribute and characteris- some alternatives in step one were randomly tic level. Since the utility of the non-GM alter- β ε offered a reduced price for one of the alterna- native (i = 1) is V1n = 1p1n + 1n, the WTPin tives. Note that the choices made in both for the GM alternatives (i = 2,3) can be calcu- steps are taken as two separate observations lated from the expression: β ε β β β in the econometric analysis. 1p1n+ 1n= i0+ 1(pin+WTPin)+ i2xn2+ … β ε (3) Following Ben-Akiva and Lerman (1985) + ikxnk+ in. and Haab and McConnell (2002), we spec- ε ε ε Assuming that E( 1n) = E( 2n) = E( 3n) = 0, ify a random utility model that is linear in the average consumer’s willingness to pay for parameters: each alternative is V =β +β p +β x + … +β x +ε , (1) 1 in i0 1 in i2 n2 ik nk in WTP=+++()ββ xK β x , iiiikkβ 022 (4) 1 where Vin is respondent n’s utility of choosing where —x denotes the mean value of the indi- alternative i, pin is the price offered to respon- k vidual specific characteristic k. The marginal dent n for alternative i, xn2 … xnk are the individual specific characteristics (e.g. gender change in WTP for alternative i associated or education) of respondent n, and the error with a change in characteristic k is terms ε are assumed to be independently, ∂WTP β in i =− ik . (5) ∂ β identically and extreme value (Gumble) distrib- xk 1 uted. The estimated parameters, except the β utility of money ( 1), are allowed to vary across the alternatives allowing the personal Regression Results characteristics to have non-constant effects for the alternatives and thereby an impact on The models are first estimated using the the choices made. For identification, the para- Norwegian data and the LIMDEP program ver- β meters of the first equation (except ( 1) are sion 7. Specifically, the binary logit model is Consumer - Chap 09 5/3/04 15:55 Page 102

102 W.S. Chern and K. Rickertsen

used for soybean oil while the multinomial that people in Nordland have a higher proba- logit model is employed for salmon. We esti- bility of not choosing the GM soybean oil, mate models using different sets of character- compared to people in Oslo. Since there are istics; however, the average consumer’s WTP limitations in a sample this small, no general for each alternative is reasonably robust for conclusions can be drawn, but there is a dif- choice of variables. Later we use the same ference between respondents in this sample. specification and estimate the corresponding The ORDER variable tells us that more peo- models with the US data. The definitions and ple chose to go for the GM alternative if this sample means of the variables used are pre- product (soybean oil) was the second product sented in Table 9.4. Since each stated choice asked about after salmon in the survey. This in the survey is taken as one observation in result suggests that the order of products may the regression analysis, the number of obser- affect the respondent’s choice of GM vs. non- vations (n) is larger than the sample size from GM products in the survey. Only one of two the survey. The number of observations knowledge variables (KNOW2) is significant. varies slightly among the products depending This result is interesting, since the variable is upon the stated choices. For example, constructed from an objective test of knowl- respondents with ‘indifferent’ between the edge. Respondents who answered question alternatives would result in two observations A4 in the survey correctly are more positive while a decisive choice is counted as only one towards genetically modified soybean oil. It is observation. The list of variables includes very logical that you are more negative if you many typical demographic variables such as believe consuming GM oil will affect your age, gender, education, income and region. In this specification, knowledge about GMOs genes, but it can also be interpreted in a more is among the key determinants of the WTP general fashion: the more you know about for GM vs. non-GM products. There are GMO, the more likely you are to be positive many other variables which can be con- towards it. This ‘fear of the unknown’ is a structed from the survey data and we con- quite common phenomenon with humans, tinue to search for a more refined and clearly also valid in the context of GM specification with the data we have. foods. The variable OIL (coded 1 if the Table 9.5 compares the regression results respondent believes GM oil to be on the mar- for soybean oil. The coefficient estimates from ket already) is not significant at any Norway are much more satisfactory than respectable probability level. Furthermore, those obtained from the US survey. Consider AGE negatively affects the utility gained from first the results from Norway. The price coeffi- the GM alternative. In other words, young cient has a negative sign, as expected, and it people are more likely to buy GM soybean oil. is significant. It basically means that the prices The education variable (EDU) is also signifi- had impact on which alternative the respon- cant at the 5% level. A higher level of educa- dent chose (i.e. price significantly influenced tion will positively affect utility gained from the utility gained from each alternative). A sig- GM oil. The education level is expectedly nificant price coefficient is, of course, a neces- highly correlated with age, as obtaining higher sity if WTP values are to be accurately education has been an increasing trend in calculated. If it is not significant, it can be Norway. It is also common to suspect that age interpreted as the respondents being price and education are correlated with knowledge insensitive. level, but this need not necessarily hold true. The constant term is the ‘starting point’ of The fit of the model for soybean oil with utility gained from the alternative, and in this the US data is considerably poorer as the case it is negative. The GM alternative has on McFadden R2 is only 0.07 compared with average a ‘negative’ utility, other variables not 0.303 for the Norwegian model. As shown in taken into account, indicating that it is of Table 9.5, the price coefficient in the US lesser value to most consumers as compared model is not significant, making it difficult to to the non-GM oil, whose constant term is produce reliable estimates of the WTP for a normalized to zero. Both ZONE and ORDER premium to avoid GM soybean oil. The only are significant at the 10% level. The finding is significant variables in the US model for Consumer - Chap 09 5/3/04 15:55 Page 103

Acceptance of GM Foods in Norway and the USA 103

Table 9.4. Deﬁnition and sample mean of variables.

Norway sample mean US sample mean

Oil Salmon Oil Salmon Variable Deﬁnition (n = 384) (n = 374) (n = 245) (n = 168)

PRICE Price of the product a 35.80 71.74 1.79 5.56 MIDWEST 1 if Midwest; 0 otherwise NR NR 0.22 0.22 SOUTH 1 if South; 0 otherwise NR NR 0.32 0.31 WEST 1 if West; 0 otherwise NR NR 0.21 0.26 NORTHEAST 1 if Northeast; 0 otherwise (dropped) NR NR 0.25 0.21 ZONE 1 if Nordland; 0 if Oslo 0.52 0.51 NR NR VER1b 1 if survey version 1 (oil and cereal); NR NR 0.30 NR 0 otherwise. Irrelevant for salmon VER2b 1 if survey version 2 (cereal and oil); NR NR 0.26 NR 0 otherwise. Irrelevant for salmon VER3b 1 if survey version 3 (oil and salmon); NR NR 0.44 0.40 0 otherwise. Dropped for soybean oil. Irrelevant for cornflake cereal VER4b 1 if survey version 4 (cornflakes cereal NR NR NR 0.60 and salmon); 0 otherwise. Dropped fpr salmon. Irrelevant for soybean oil ORDERb 1 for survey version 1 (salmon and oil); 0.49 0.52 NR NR 0 for version 2 (oil and salmon) for Norway KNOW1 1 if wrong answer to A3; 0 if don’t 0.20 0.21 0.30 0.43 know; 1 if correct answer to A3c KNOW2 Defined similarly to KNOW1 but 0.11 0.08 0.47 0.54 related to A4d OIL 1 if one believes that GM ingredients 0.43 NR 0.58 NR are used in the production of soybean oil; 0 otherwise. AGE 0.1 (age – mean age) 0.007 0.01 0.27 0.02 GENDER 1 if male; 0 if female (USA); 0.047 0.07 0.31 0.24 1 if female; 1 if male (Norway) EDU Education level coded from 1 (elementary 3.297 3.33 3.94 4.03 school) to 8 (doctoral/advanced degree) in the USA and from 1 to 7 in Norway KIDS 1 if living with children (17 years old or 0.45 0.45 0.39 0.36 younger) ln (INCOME) Logarithm of income categories from 1.40 1.39 1.54 1.61 1 to 10 measured in increments of US$10,000 in the USA; from 1 to 11 in increments of NOK 100,000 in Norway FREQ Frequency of salmon consumption NR 2.53 NR 2.46 coded 1–4 (1 if once a week; 2 if once a month; 3 if once every 3 months; 4 if once a year)

n, number of observations; NR, not relevant. aPrices are in US$ per 32 ﬂ. oz for oil, and per pound for salmon in the US survey and NOK per litre for oil and per kilo for salmon in the Norwegian survey. bWe have four versions of questionnaires in the US survey. The Norwegian survey has only two versions for oil and salmon. All order effects are taken care of by the version variables (VER1, VER2, VER 3 and ORDER). cA3 states ‘Non-GM soybeans do not contain genes while genetically modiﬁed soybeans do’. dA4 states ‘By eating GM foods, a person’s genes could be altered’. Consumer - Chap 09 5/3/04 15:55 Page 104

104 W.S. Chern and K. Rickertsen

Table 9.5. Regression results for soybean oil. Norway US Variable Coefﬁcient estimate t-Ratio Coefﬁcient estimate t-Ratio

PRICE 0.187*** 5.61 0.543 1.05 CONSTANT 3.038*** 4.88 1.900*** 2.85 MIDWEST NR NR 0.955* 1.89 SOUTH NR NR 0.694 1.51 WEST NR NR 0.070 0.13 ZONE 0.810* 1.86 NR NR VER1 NR NR 0.763* 1.91 VER2 NR NR 0.208 0.46 ORDER 0.831* 1.96 NR NR KNOW1 0.172 0.58 0.163 0.71 KNOW2 0.880*** 2.92 0.329 1.49 OIL 0.614 1.57 0.389 1.06 AGE 0.269* 1.85 0.085 0.66 GENDER 0.851*** 3.61 0.463 1.18 EDU 0.317** 2.04 0.048 0.45 KIDS 0.198 0.43 0.400 1.03 ln (INCOME) 0.848** 2.18 0.373 1.47 McFadden R 2 0.303 0.07 n 384 245

n, number of observations; NR, not relevant. *Significant at 10%; **significant at 5%; ***significant at 1%.

soybean oil are CONSTANT and VER1. mies, two knowledge variables and gender are These results show that American consumers all significant determinants for choosing GM have a negative utility associated with GM vs. non-GM salmon. The Norway model for soybean oil without taking into account any salmon also shows very similar results as the other factors. Furthermore, the order of the knowledge variables and gender are statisti- products asked in the survey affects the valua- cally significant. The price variable is highly tion or the WTP for GM vs. non-GM attrib- significant. In addition, income is found to be utes. However, the US model shows the significant for GM-fed salmon. The salmon opposite results from the Norwegian model as models with significant price coefficients pro- placing soybean oil first in the questionnaire vide a much stronger bases for computing the appears to increase rather than decrease the WTP estimates than the soybean oil model for probability of choosing GM soybean oil. the USA. Table 9.6 presents and compares the regression results for salmon from the USA and Norwegian survey data. The multinomial Estimated Willingness to Pay model results for the US are much more satisfactory than those for soybean oil. The price The WTP estimates are obtained from the coefficient is statistically significant at the 5% estimated regression models according to level and the McFadden R2 is considerably equation (4). The sample means of the higher (0.16). The estimated coefficients for explanatory variables are used in these com- non-GM vs. GM salmon are much stronger putations. We may also compute the WTP by than those for non-GM vs. GM-fed salmon, respondents and investigate the distribution of indicating that American consumers are more estimated WTP among respondents. The concerned or fearful about GM salmon than computed WTP may be interpreted as the GM-fed salmon. The results for GM salmon WTP a premium for a non-GM product or as show that the constant term, regional dum- the WTP to avoid a GM product. In other Consumer - Chap 09 5/3/04 15:55 Page 105

Acceptance of GM Foods in Norway and the USA 105

Table 9.6. Regression results for salmon. Norway US Variable Coefﬁcient estimate t-Ratio Coefﬁcient estimate t-Ratio

PRICE 0.068*** 7.42 0.421** 2.71 GM-fed salmon CONSTANT 2.130** 2.46 1.496 1.45 MIDWEST NR NR 0.149 0.23 SOUTH NR NR 0.357 0.62 WEST NR NR 0.184 0.28 ZONE 0.630 1.62 NR NR VER3 NR NR 0.493 1.15 ORDER 0.102 0.27 NR NR KNOW1 0.118 0.44 0.371 1.25 KNOW2 0.858*** 3.32 0.737** 2.20 FREQ 0.115 0.51 0.195 1.00 AGE 0.107 0.87 0.020 0.12 GENDER 0.536*** 2.67 0.062 0.11 KIDS 0.163 0.39 0.340 0.65 EDU 0.144 1.00 0.154 1.10 ln (INCOME) 0.926*** 2.64 0.167 0.39 GM salmon CONSTANT 3.565*** 3.59 5.778*** 3.79 MIDWEST NR NR 2.057** 2.35 SOUTH NR NR 2.015** 2.48 WEST NR NR 0.222 0.25 ZONE 0.366 0.86 NR NR VER3NRNR0.308 0.57 ORDER 0.381 0.93 NR NR KNOW1 0.059 0.20 0.821** 2.32 KNOW2 1.207*** 4.10 1.367*** 2.94 FREQ 0.021 0.09 0.694*** 3.03 AGE 0.105 0.76 0.309 1.60 GENDER 0.583*** 2.63 1.389** 2.02 KIDS 0.173 .0.38 0.461 0.75 EDU 0.190 1.18 0.255 1.59 ln (INCOME) 0.505 1.33 0.829 1.53 McFadden R 2 0.261 0.159 n 374 + 364 = 738 168 + 165 = 333

n, number of observations; NR, not relevant. *Significant at 10%; **significant at 5%; ***significant at 1%.

words, the WTP may be interpreted as the price of GM-fed salmon has to be reduced by amounts that we would have to reduce the NOK 43.42 and GM salmon by NOK 53.96 price of the non-GM alternative to let the from the base price of NOK 80. This corre- average consumer be equally well off. sponds to price reductions of 55%, 54% and The results are shown in Table 9.7. 67% for GM soybean oil, GM-fed salmon and Consider ﬁrst the results for Norway. The GM salmon, respectively. As expected, the price of non-GM soybean oil was NOK 40 required reduction in price is larger for GM and the price of GM soybean oil has to be salmon than for the other GM alternatives. reduced by NOK 22.13 to NOK 17.87 per There is a distinction between direct and indi- liter to make the average Norwegian con- rect GM consumption and there is also a dif- sumer equally well off. In a similar way, the ference between plant and animal genes. Consumer - Chap 09 5/3/04 15:55 Page 106

106 W.S. Chern and K. Rickertsen

Table 9.7. Estimated WTP values to avoid GM alternatives. Alternative Country Item GM soybean oil GM-fed salmon GM salmon

Norway Mean, NOK 22.13 43.42 53.96 Mean, US$a 2.77 5.43 6.75 % reduction 55 54 67 USA Mean, US$ 1.82 2.75 4.49 % reduction 84 46 71

aThe exchange rate is set to NOK 8.00 per US$.

The WTP values to avoid GM products for The marginal WTP values reported in American consumers are estimated to be $1.82 Table 9.8 show how much a change in one per 32 fl. oz for soybean oil, $2.75 per pound of the individual specific characteristics will for GM-fed salmon and $4.49 for GM salmon. affect the WTP to avoid the different GM The corresponding percentages of price reduc- alternatives. For most parts, the effects of tion are 84%, 46% and 71% for soybean oil, the characteristics are consistent across the GM-fed salmon and GM salmon, respectively. various GM alternatives. Since many coeffi- These figures are much higher than expected. cient estimates are not statistically signifi- With respect to the WTP for soybean oil, the cant in the US models, the signs of the estimate is based on the price coefficient which effects are not entirely consistent. In is not statistically significant as noted earlier. Norway, the age effects are always positive Thus we cannot place any confidence on this and significant for GM soybean oil and GM- estimate. The unexpectedly high value of the fed salmon. If the age of the respondent WTP for soybean oil in the USA may be caused increases by 10 years, then the respondent by the high standard errors associated with the demands an extra price reduction of NOK price coefficient and other parameters in the 1.88, 3.52 and 3.54 for GM soybean oil, model. The high errors may cause the esti- GM-fed salmon and GM salmon, respec- mated WTP to be either extremely large or tively. The gender effects are always nega- extremely small. The price coefficient for tive and significant. Females are coded as salmon is statistically significant and thus the 1 and males as 1 implying that female resulting WTP estimates should be more credi- consumers demand price reductions of NOK ble. The results show that American consumers 4.48, 9.32 and 11.72, as compared with are willing to pay substantial amounts of pre- the average Norwegian consumer, for GM mium to avoid GM-fed or GM salmon. soybean oil, GM-fed salmon, and GM

Table 9.8. Marginal WTP values to avoid GM alternativesa. Alternative Country Variable GM soybean oil GM-fed salmon GM salmon

Norway Age 1.88 3.52 3.54 Gender 4.48 9.32 11.72 Education 2.87 5.29 5.85 Income 1.05 3.71 3.03 USA Age 0.16 0.05 0.73 Gender 0.85 0.15 3.30 Education 0.09 0.37 0.61 Income 0.19 0.10 0.49

aNOK for Norway and US$ for the USA. Consumer - Chap 09 5/3/04 15:55 Page 107

Acceptance of GM Foods in Norway and the USA 107

salmon respectively. The effect of education Conclusions is always negative and significant. The more education the less price reductions are This chapter presents survey results and needed. If the educational level (from one to analyses from a joint research project to con- six) increases by one, then the respondent duct a multi-country study on consumer requires NOK 2.87, 5.29 and 5.85 less acceptance of GM foods. The results indicate compensation for consuming GM soybean more favourable attitudes to GM foods among oil, GM-fed salmon and GM salmon, respec- US than Norwegian consumers. However, the tively. The effect of income is always posi- opinions in the USA are also quite mixed and tive and significant implying that only 43% of the American respondents respondents with higher incomes demand claimed that they are willing to consume foods larger price reductions. Since the log of produced with GM ingredients. The opposi- income is used as a variable, there is always tion against GM foods is reduced when some a positive but decreasing effect of income, benefits associated with them are introduced and the estimates reported are for changes into the questions, suggesting that GM foods from mean income. If the mean household have a potential to become more popular. income increases by one class (or NOK Reduced use of pesticides and improved nutri- 100,000), then the respondent demands an tional qualities are perceived as more impor- additional price reduction of NOK 1.05, tant potential benefits than reduced price. 3.71 and 3.03 for GM soybean oil, GM-fed Health concerns are apparently more impor- salmon and GM salmon, respectively. tant than ethical or religious concerns in In the USA, the only significant demo- explaining the negative attitudes towards GM graphic variable is gender (coded 1 for male foods. The support for mandatory labelling is and 0 for female). The results are very similar overwhelming in the surveys even when to those found in Norway, indicating that labelling may increase food prices. females are willing to pay more than males in There is a substantial WTP to avoid GM order to avoid GM products. Specifically, alternatives. In the surveys, 80% of the females are willing to pay $0.85, $0.15 and Norwegian respondents chose the non-GM $3.30 more than males to avoid GM soybean alternatives in each case and for the American oil, GM-fed salmon and GM salmon, respec- respondents 45% chose non-GM soybean oil, tively. Even though the estimated coefficients 59% non-GM salmon over GM-fed salmon, and of education are not statistically significant, 69% chose non-GM salmon over GM salmon. the results show the opposite effects as com- These figures indicate that there are differences pared with those from Norway. It appears between direct and indirect GM consumption that more-educated American consumers are and between animal and plant genes. willing to pay more premiums for non-GM The WTP for avoiding the GM alternatives salmon than less-educated consumers. In indicates that the average Norwegian con- other words, more-educated people would sumer demands price reductions of 55%, demand higher price reductions for consum- 54% and 67% for GM soybean oil, GM-fed ing GM products than less-educated ones. salmon and GM salmon, respectively, as com- The reported WTP figures are quite sub- pared with the conventional alternatives. For stantial, indicating a strong opposition to American consumers, the estimated price GM foods in Norway and the USA. Given reductions demanded for GM alternatives are the potential hypothetical bias mentioned 84%, 46% and 67% for soybean oil, GM-fed above they must be interpreted as upper salmon and GM salmon, respectively. It is sur- bounds. However, we may note that the prising to have such a high WTP for vegetable reported WTP values are identically and oil in the USA. As noted earlier, since the US inversely related to the estimated price para- estimate is based on an insignificant price β meter, 1, implying that any hypothetical coefficient for soybean oil, the WTP estimate bias affects the levels of the WTP and not is not credible for vegetable oil. Overall, the the relative price effects between the GM high percentage premiums may, at least to and GM-fed salmon. some extent, be due to the hypothetical Consumer - Chap 09 5/3/04 15:55 Page 108

108 W.S. Chern and K. Rickertsen

nature of the choices without any real pay- to construct our econometric models and to ments. Note also that the estimated WTP val- test econometrically the structural differences ues obtained in this study are substantially between Norway and the USA. The surveys higher than those estimated by Chen and reported in this chapter were pilot surveys. Chern (2002) using a mail survey conducted These pilot surveys were also used as a basis in Columbus, Ohio, USA. It is important to for re-designing our survey questionnaire for validate these estimates with larger samples. a large-scale public survey with 1000 tele- Future research will focus on searching for phone interviews conducted in 2003. We a better modelling speciﬁcation so that we also plan to do similar surveys in Japan, can use more information from the surveys Spain and Taiwan.

References

Ben-Akiva, M. and Lerman, S.R. (1985) Discrete Choice Analysis: Theory and Application to Travel Demand. The MIT Press, Cambridge, Massachusetts. Boccaletti, S. and Moro, D. (2000) Consumer willingness to pay for GM food products in Italy. AgBioForum. 3, 259–267. Burton, M.L., Rigby, D., Young, T. and James, S. (2001) Consumer attitudes to genetically modified organisms in food in the UK. European Review of Agricultural Economics 28, 479–498. Carson, R.T. and Mitchell, R.C. (1995) Sequencing and nesting in contingent valuation surveys. Journal of Environmental Economics and Management 28, 155–173. Caulder, J. (1998) Agricultural biotechnology and public perceptions. AgbioForum 1, 38–39. Chen, H.-Y. and Chern, W.S. (2002) Willingness to pay for GM foods: results from a public survey in the U.S. Paper presented at the 6th International Conference on ‘Agricultural Biotechnology: New Avenues for Production, Consumption, and Technology Transfer’, Ravello, Italy, 11–14 July 2002. Darr, D. (2001) Econometric analysis of genetically modified organism adoption: a study of Ohio grain farmers. Unpublished Masters’ thesis, Department of Agricultural, Environmental, and Development Economics, The Ohio State University, Columbus, Ohio, USA. Darr, D. and Chern, W.S. (2002) Estimating adoption of GMO soybeans and corn: a case study of Ohio, U.S.A. In: Santaniello, V., Evenson, R.E. and Zilberman, D. (eds) Market Development for Genetically Modified Foods. CAB International, Wallingford, UK, pp. 141–158. European Commission (2001) Europeans, Science and Technology. Eurobarometer 55(2). European Commission, Brussels. European Commission (2002) Press release, 3 July. Available at www.europa.eu.int under the category of ‘environment’. Franks, J.R. (1999) The status and prospects for genetically modified crops in Europe. Food Policy 24, 565–584. Haab, T.C. and McConnell, K.E. (2002) Valuing Environmental and Natural Resources: the Econometrics of Non-market Valuation. Edward Elgar Publishers, Cheltenham, UK. Hallman, W.K. and Metcalfe, J. (2001) Public perceptions of agricultural biotechnology: a survey of New Jersey residents. US Department of Agriculture. Available at http://www.nalusda.gov/bic/pubpercept. Hoban, T., J. (1998) Trends in consumer attitudes about agricultural biotechnology. AgBioForum 1, 3–7. Hoban, T.J. (1999) Public perceptions and understanding of agricultural biotechnology. US Information Agency. Available at www.usia.gov/journals/ites/1099/ijee/bio-toc.htm. Kinsey, J.D. (2001) The new food economy: consumers, farms, pharms, and science. American Journal of Agricultural Economics 83, 1113–1130. List, J.A. and Shogren, J.F. (1998) Calibration of the difference between actual and hypothetical valuations in a field experiment. Journal of Economic Behavior and Organization 37, 193–205. Macer, D. and Ng, M.A.C. (2000) Changing attitudes to biotechnology in Japan. Nature Biotechnology. 18, 945–947. Mendenhall, C.A. and Evenson, R.E. (2002) Estimates of willingness to pay a premium for non-GM foods: a survey.’ In: Santaniello, V., Evenson, R.E. and Zilberman, D. (eds) Market Development for Genetically Modified Foods. CAB International, Wallingford, UK, pp. 55–61. Consumer - Chap 09 5/3/04 15:55 Page 109

Acceptance of GM Foods in Norway and the USA 109

Moon, W. and Balasubramanian, S.K. (2001) Estimating willingness to pay for nonbiotech foods: a comparison across US and UK consumers.’ Paper presented at the Annual Meeting of the American Agricultural Economics Association, Chicago, Illinois, 5–8 August. Ng, M.A.C., Takeda, C., Watanabe, T. and Macer, D. (2000) Attitudes of the public and scientists to biotechnology in Japan at the start of 2000. Eubios Journal of Asian and International Bioethics 10, 106–113. Phillips, P.W.B. and McNeill, H. (2000) A survey of national labeling policies for GM foods. AgBioForum 3, 219–224. Priest, S.H. (2000) U.S. public opinion divided over biotechnology? Nature Biotechnology 18, 939–942. Skogmo, A. (2002) Consumer attitudes toward genetically modiﬁed foods: a study of consumer acceptance in two regions of Norway. MSc thesis, Department of Economics and Social Sciences, Agricultural University of Norway, Aas, Norway. Spetsidis, N.M. and Schamel, G. (2001) A survey over consumer cognitions with regard to product scenarios of GM foods in Germany. Paper presented at the 71st EAAE Seminar on ‘The Food Consumer in the Early 21st Century’, Zaragoza, Spain, 19–20, April. Verdurme, A., Gellynck, X. and Viaene, J. (2001) Consumer’s acceptance of GM food. Paper presented at the 71st EAAE Seminar on ‘The Food Consumer in the Early 21st Century, Zaragoza, Spain, 19–20 April. Vogt, D.U. and Parish, M. (1999) Food biotechnology in the USA: science, regulation, and issues. CRS Report to Congress, 2 June, Available at http://www.usinfo.state.gov/topical/global/biotech/crs- food.htm. Consumer - Chap 09 5/3/04 15:55 Page 110 Consumer - Chap 10 5/3/04 15:55 Page 111

10 Comparing Consumer Responses towards GM Foods in Japan and Norway1

Jill J. McCluskey,1 Kristine M. Grimsrud2 and Thomas I. Wahl3 1Department of Agricultural Economics, Washington State University, 211J Hubert Hall, Pullman, WA 99163, USA; 2Department of Economics, University of New Mexico, Albuquerque, NM 87131-0001, USA; 3International Marketing Program for Agricultural Commodities and Trade (IMPACT) Center, Washington State University, Hulbert Hall, Rm 123, PO Box 646214, Pullman, WA 99164-6210, USA

Introduction Europeans simply do not want genetically modified organisms (GMOs). The introduction of genetically modified (GM) Consumer attitudes may be just as important crops to world markets has created a new as consumer knowledge. The Eurobarometer division between the crop trading countries. survey showed that moral doubts were more The USA and Canada have great economic important than health risks in shaping public interests in exporting their transgenic crops, acceptance of gene technology (Biotechnology however, lack of public acceptance of GM and the European Public Concerted Action food products in the European Union (EU) Group, 1997). Environmental concerns are also and Japan has already resulted in reduced or important-in the 2000 Eurobarometer survey, curbed demand for GM food products. Many 59.4% of EU citizens interviewed said that they European and Japanese consumers believe thought GMOs could have a negative impact on GM foods pose a threat to human health. the environment. They fear short- and long-term consequences Macer and Ng (2000) report that support for their own health and their offspring. for biotechnology and genetic engineering in A suggested remedy has been consumer Japan is decreasing, especially for agricultural education about GM food safety. However, applications. They used a mail survey the results of the 1996 Eurobarometer survey spanning the years 1991, 1993, 1997 and suggested that more knowledgeable people do 2000 and found that Japanese interest in not necessarily have a more positive opinion; science and biotechnology increased from they just have a more definite opinion about 30% in 1991 to 47% in 2000. From the biotechnology (Biotechnology and the 2000 survey, 97% of respondents are familiar European Public Concerted Action Group, with the term ‘biotechnology’, which implies 1997). A later survey, the 2001 that awareness of biotechnology has increased Eurobarometer survey, showed that 70.9% of significantly among the Japanese public. Also

1 The authors wish to thank Hiromi Ouchi for excellent research assistance. Further, the authors thank without implicating Phil Wandschneider for helpful discussion and advice. The authors gratefully acknowledge ﬁnancial support from the IMPACT Center at Washington State University. © CAB International 2004. Consumer Acceptance of Genetically Modiﬁed Food (eds R.E. Evenson and V. Santaniello) 111 Consumer - Chap 10 5/3/04 15:55 Page 112

112 J.J. McCluskey et al.

from the 2000 survey, only 31% are likely to tion in a given country prefers to avoid GM support GM foods, and only 20% say they are foods. willing to buy GM fruits. They conclude that Mandatory labelling forces US producers although the majority of Japanese consumers to segregate crops to claim food products are optimistic about biotechnology, there are are ‘GM-free’. This would be difficult and increasingly negative views towards its appli- costly. For example, many grain elevators cation for agriculture. are not physically equipped to segregate Consumer attitudes and behaviour towards crops. US producers may lose market share GM food products are complex and differ because consumers can reject their GM across cultures. A better understanding is crops. essential for designing market strategies. We investigate factors that affect consumer acceptance of GM food in Japan and Norway and Data estimate the discounts necessary for consumers to be willing to purchase GM food or In August 2001, we conducted 400 in-person the premium consumers are willing to pay for interviews in Japanese at the Seikatsu Club GM food. We compare consumer preferences Consumer Cooperative (Seikyou), a grocery across countries. store-like setting, in Matsumoto City, Japan. Mandatory labelling of GM foods has Matsumoto is a relatively agricultural area with obvious implications for trade. The EU has about 13% of the population coming from imposed mandatory labelling for some foods farm households compared with 2% in all of that contain GM ingredients. In October Japan. Consumer cooperatives usually focus 1999, the EU gave preliminary approval to on offering ‘safe foods’, as a marketing strat- a law that requires labels on all foods con- egy, and target members who are more willing to purchase safe foods. The Seikyou has taining more than 1% GM ingredients. In significant market power in the Japanese mar- Japan, authorities have ordered mandatory ket place. labelling for 29 categories of food if they In January 2002, we conducted 400 in- contain any GM ingredients. The USA has person interviews in Norwegian at the RIMI argued that there is no health-related or sci- Liertoppen grocery store in the Oslo region of entific reason to reject GM commodities and Norway. This region is the most populated food products and has challenged the EU’s part of Norway and one of the Norwegian mandatory GM labelling as a non-tariff trade centres of economic activity. The RIMI chain barrier. of grocery stores has chosen a low-price/lim- The Codex committees of the World Trade ited selection niche in the market and has in Organization (WTO) are working on harmo- this way gained significant power in the nizing international standards and resolving Norwegian market place. trade disputes associated with food labelling to The surveys solicited demographic infor- promote fair trade of foods while protecting mation, respondents’ attitudes about the consumer health. Since different countries environment and food safety, and their self- have different attitudes towards GM food reported knowledge and perceptions about products, the Codex frameworks allow each biotechnology. Our surveys included con- country to develop its own standards. The tingent valuation questions regarding willing- challenge of Codex is to set international stan- ness to accept a discount to purchase food dards for GM food labelling that both pro- products made from GM wheat. The hypo- motes fair trade and allows for consumer thetical market for the good in question must choice. An important issue in GM labelling be as close as possible to a real market in policy is scientific versus consumer sover- order to reveal people’s true preferences if eignty. Although the scientific consensus may an actual market existed (Pearce and Turner, be that GM foods are completely safe for con- 1990). In Japan, we asked about noodles sumption aside from potential allergens, it made with GM wheat, and in Norway we may be the case that a majority of the popula- asked about bread made with GM wheat. Consumer - Chap 10 5/3/04 15:55 Page 113

Responses to GM Foods in Japan and Norway 113

The food products (bread and noodles) used buy bread with GM wheat flour when offered in this study are appropriate for examination no discount compared to conventional since they are frequently consumed food bread. Further, 39% of consumers in the products. sample replied that they would be willing to First, consumers were asked if they were purchase the GM product if it was cheaper willing to purchase the GM food product if it than the conventional product. Of the 400 was offered at the same price as the corre- Japanese respondents, only 3% said that sponding, non-GM product. If the respon- they would be willing to purchase the GM dent’s answer to this question was ‘no’, a noodles at the same price as the correspond- follow-up question was asked if the respon- ing non-GM product. Further, only 17% of dent was willing to purchase the GM food consumers in the sample stated that they product if offered a percentage discount would be willing to purchase the GM product compared with the corresponding non-GM if it was less expensive than the correspond- product. The discount was set at one of the ing non-GM product. The rest of the following levels: 5%, 10%, 25%, 40% and Japanese respondents, that is 80%, were not 50%. Each level of discount was used for willing to purchase the GM noodles even one-fifth of the surveys. That is, 80 of the with the discount. 400 surveys had a 5% discount for GM bread, another 80 surveys had a 10% discount for GM bread, and so on. The assign- Empirical Results ment of survey version (and thus, discount) was random to the respondent. The maxi- The model that is applicable for examining mum discount was set at 50%, assuming the outcomes of our survey can be consid- that people who would not prefer the GM ered a special case of the double-bounded food product at such at large discount would logit model (Hanemann et al., 1991). In this not choose the product at any discount. No model, the initial bid (B0) equals zero and follow-up question was asked if the cus- implies no price difference between the GM tomer’s answer was ‘yes’ to the initial ques- food product and the non-GM food product. tion, and he/she was willing to purchase the The second bid (B ), is the GM food product GM food product at no discount. The ratio- D offered at a random percentage discount nale for no follow-up to a ‘yes’ response is relative to the non-GM food product. This that the type of genetic modification associ- bid is only given to individuals who answer ated with these GM food products is a that they would not buy a GM food product process attribute, which reduces production at the same price as a non-GM food costs – as opposed to a product-enhancing product. attribute. An example of a GM product with Let WTA denote an individual’s willingness a product-enhancing attribute is the Flavr to accept (WTA) compensation (or bid func- Savr tomato. Proponents claim that the GM products with process attributes are identical tion) for GM food products, relative to the non-GM food products, and let B # 0 denote to non-GM products. Opponents view D genetic modification as a negative attribute. the discount bid on GM food products relative Therefore, it would not make economic to non-GM food products. The following dis- sense after an initial ‘yes’ to respond with a crete outcomes of the bidding process are follow-up question that involves paying a observable: premium for these GM products that only 1 WTA≤ B have cost-reducing attributes.  0 D = 2 B<≤ WTA B . (1) Although Norwegian consumers were  0 D 3 B< WTA generally sceptical of GM foods, more of the  D Norwegian consumers chose the GM prod- The WTA function for each food product for ucts compared with the Japanese con- individual i is sumers. Of the 400 Norwegian respondents, α ρ λ ε more than one-quarter of the sample would WTAi = + Bi + zi + i i = 1, … ,n (2) Consumer - Chap 10 5/3/04 15:55 Page 114

114 J.J. McCluskey et al.

where Bi is the ultimate discount bid individual i feed to their children. Therefore, it is not sur- faces, zi is a column vector of observable char- prising that most Japanese and Norwegian ε acteristics of the individual, i is a random vari- consumers want to avoid GM foods. A better able accounting for random noise and possibly understanding of Japanese and Norwegian unobservable characteristics. Unknown parame- consumers’ attitudes and behaviour towards ters to be estimated are α, ρ and λ. Linearity in GM food products and how these attitudes z and ε is assumed for all individuals. affect the purchasing choices for such food The bid information and other demo- products is essential for marketing GM food graphic information were used to estimate the products in those countries. We found that magnitude of factors that affect consumers’ although the majority of Norwegian consumers WTA for GM food products and how much of want to avoid GM food, they are not as nega- a relative discount consumers will require to tive as the Japanese. purchase GM food products. Our estimation The Norwegian policy restricting GM results for Japan show that variables repre- food products and the results of surveys senting food safety and environmental atti- show that there is considerable scepticism in tudes, self-reported knowledge about the Norwegian population towards GM biotechnology, self-reported risk perceptions foods. The results of this study suggest that towards GM foods, income, and education all our sample of Norwegian consumers is, on signiﬁcantly increase the necessary discount average, willing to purchase GM bread with required for Japanese consumers to choose a 49.5% discount compared to the corre- GM foods. With the Norwegian data, increas- sponding non-GM product. Consumers’ per- ing self-reported risk perceptions toward GM ceptions and attitudes towards GM food and foods and preferences for domestically pro- respondents’ age increase WTA (that is, a duced food both signiﬁcantly increase the dis- greater discount would be required) for GM count required for Norwegian consumers to food. choose GM foods. The strong scepticism in the Norwegian The mean willingness to accept for both population may be fading, leaving a potential GM food products, WTA, was estimated by future market for GM foods. People younger restricting λ = 0 (Hanemann et al., 1991). than the average age of the sample were The empirical mean WTA can then be calcu- willing to purchase the GM food products lated as –α˜/ρ˜. Our results indicate that with an average discount that is half (or less) Japanese Seikyou members, on average, of what the customers above the average age want a 60% discount on GM noodles com- of the sample needed. The gap in between pared to non-GM noodles. Our results indi- generations for WTA indicate that younger cate that, on average, the Norwegian people may be more open to GM foods and consumers in our sample want a 49.5% dis- that it may a question of time before the count on GM bread compared to conven- Norwegian market may be more open to tional bread. (Further details of each country GM food products. analysis are presented in Grimsrud et al., More Japanese respondents rejected the 2003, and McCluskey et al., 2003.) GM food products compared with the Norwegians. An implication of these studies is that there is an opportunity to market food Conclusions segregated from GM products in both countries, but especially in Japan. Japanese The Japanese and Norwegian cultures both consumers need to be convinced of the safety place a great deal of value on tradition. This of GM foods if they are to be marketed worldview extends to the food they eat and successfully there. Consumer - Chap 10 5/3/04 15:55 Page 115

Responses to GM Foods in Japan and Norway 115

References

Biotechnology and the European Public Concerted Action Group (1997) Europe ambivalent on biotechnology. Nature 387, 845–847. Grimsrud, K.M., McCluskey, J.J., Loureiro, M.L. and Wahl, T.I. (2003) Consumer attitudes toward genetically modified food in Norway. IMPACT Center Technical Working Paper. Hanemann, W.M., Loomis, J. and Kanninen, B.J. (1991) Statistical efficiency of double-bounded dichotomous choice contingent valuation. American Journal of Agricultural Economics 73, 1255–1263. Macer, D. and Ng, M.A.C. (2000) Changing attitudes to biotechnology in Japan.’ Nature Biotechnology 18, 945–947. McCluskey, J.J., Grimsrud, K.M., Ouchi, H. and Wahl, T.I. (2003) Consumer response to genetically modified food products in Japan. Agricultural and Resource Economics Review 32(2), 222–231. Pearce, D.W. and Turner, R.K. (1990) Economics of Natural Resources and the Environment. Harvester Wheatsheaf, London. Consumer - Chap 10 5/3/04 15:55 Page 116 Consumer - Chap 11 5/3/04 15:56 Page 117

11 Willingness to Pay for GM Foods: Results from a Public Survey in the USA

Hsin-Yi Chen and Wen S. Chern Department of Agricultural, Environmental, and Development Economics, The Ohio State University, Agricultural Admin Building, 2120 Fyffe Road, Columbus, OH 43210-1067, USA

Introduction countries like Japan and Taiwan. In addition, the potential environmental endanger of GM Introducing biotechnology into agricultural products appears to be one of the major production is one of the most prominent hindrances for consumers to accept GM foods. benchmarks in the history of agricultural Consumer acceptance of GM products, development. The application of genetic mod- therefore, has become a vital factor on how ification technology on agricultural crops and prosperous the market for GM foods will be in the resulting genetically modified organisms the future. It will affect the future course of (GMOs) are considered one of the most private and public investment in the develop- important yet controversial advancements in ment and use of GM technology. Thus, con- science and technology. Despite all the sumer perception of and acceptance of GM promises and benefits proclaimed by many technology and GM foods are crucial for the biotech companies and the governments such global market of GM products, agricultural as reduced pesticide usage, higher crop yields, trade and the future development of agricul- enhanced nutritional values and many more, tural biotechnology. the controversy surrounding its application to Despite the seemingly ample information food production persists in many countries. available, research on consumer perception Studies have shown that many consumers in and attitudes towards biotechnology in gen- countries like those in the European Union (EU) eral or GM foods in particular, has been lim- and Japan have difficulties accepting genetically ited. In most cases, the methodology adopted modified (GM) products (Macer and Ng, 2000). relied heavily on qualitative questions and Consumers are hesitant to buy GM foods descriptive comparisons. Few attempted to largely because of concerns about the uncertain quantify the consumer’s purchase behaviour effects of GM foods on human health. There and to investigate the willingness to buy GM are also religious and ethical concerns about the foods. Also, the constantly changing attitudes possible intake of genes from animals contained expressed by consumers signify the need to in GM foods, and consumers have no means to investigate and obtain more up-to-date data identify these products. These concerns in on consumer attitudes toward GM foods. turn have generated a strong demand for the Furthermore, few studies have attempted to labelling of GM products in the EU and estimate the consumer’s willingness to pay

118 H.-Y. Chen and W.S. Chern

(WTP) for GM or non-GM foods. This study, while 11% of the respondents would not try therefore, is aimed at providing the WTP esti- GM foods, 42% say they might still try, indi- mates based on consumer attitude towards cating that UK consumers are not highly GM foods and other characteristics. Actually opposed to GM foods (Loader and Henson, this study is part of a research project 1998). Another study conducted by the attempting to provide a more global perspec- Nordic Industrial Fund (2000) for Denmark, tive by conducting a multi-country survey in Finland, Norway and Sweden also suggests four countries, Japan, Norway, Taiwan and that Nordic consumers are united in their neg- the USA. This chapter, however, presents ative attitudes toward GM foods. Another only the analysis of a survey conducted in the study by Burton et al. (2001), based on a sur- Columbus Metropolitan Area, Ohio, under vey of residents in Manchester, UK, shows this project. that the average willingness to pay to achieve Specifically, the objectives of this study are a reduction in risk associated with GM prod- to analyse an Ohio survey on the consumer ucts is 9.8% of the respondent’s food expen- acceptance of GM foods and to conduct a diture, and the average willingness to pay to contingent valuation of the willingness to pay achieve a GM-free diet is 13% of their food for three selected products with and without expenditure. Their study also finds that gender GMO ingredients, i.e. vegetable oil, salmon is a significant determinant of attitudes and cornflake breakfast cereal. towards GM technology and female respondents are willing to pay more to reduce risk. In Asian countries, a study by Hoban Background and Literature (1996) concludes that Japanese consumers have optimistic attitudes towards GM biotech- Consumer acceptance towards GM foods nology similar to American consumers, as varies greatly among countries. Studies in the 87% of the Japanese respondents in his study USA mostly show that its consumers have a were positive about the use of biotechnology. higher acceptance rate towards biotechnology However, a recent study by Macer and Ng and GM foods than those in other countries. (2000) indicates that a small majority of One of Hoban’s studies (1998) indicates that Japanese respondents in the past 3 years two-thirds of American consumers are posi- have been favourably inclined to GM technol- tive about plant biotechnology, and this sup- ogy and consider it a means of improving the port is the highest among men and people quality of life (54% in 1997 and 59% in with more formal education. National surveys 2000), implying the changing attitudes among conducted in recent years also show that the Japanese consumers. In Taiwan, even roughly between 35 and 45% of American though knowledge of GM foods is not perva- consumers consider they have heard or read a sive, a Gallup Poll (2000) shows that lot or some about biotechnology (International Taiwanese consumers are not against the use Food Information Council Foundation, 2001). of GM technology in food production and are In Europe, on the other hand, the support for willing to purchase GM foods. GM technology in general and GM foods in Studies on other agricultural technologies particular are relatively low as compared to have been done. Wang et al. (1997) the USA. Based on a European consumer attempted to measure the WTP for bovine survey conducted in 1999, Gaskell (2000) somatotropin (rBST)-free milk in Vermont and shows that Europeans are mostly neutral their results show that for rBST-free milk, about agricultural biotechnology but opposed 37.4% of the respondents do not want to to both GM foods and cloning of animals, pay any premium, 50.6% are willing to pay a especially in countries such as Greece, Austria premium up to 40 cents per gallon and and Luxemburg. Specifically, the survey indi- 12.0% would pay a premium of 41 cents or cates that only 22% of the European respon- more. Also, the authors find that the con- dents are supporters of GM foods, 25% are sumer WTP is affected by several demo- risk-tolerant supporters and up to 53% are graphic variables, such as income, education opponents. A study in the UK shows that and gender. Halbrendt et al. (1995) con- Consumer - Chap 11 5/3/04 15:56 Page 119

Willingness to Pay for GM Foods 119

ducted a nationwide contingent valuation this study contains both dichotomous choice (CV) survey to measure the consumer willing- and polychotomous choice questions. Food ness to purchase pork with lower saturated products used in the survey include vegetable fats, and their results show that the survey oil, salmon and cornﬂake breakfast cereal. respondents are willing to pay an average Vegetable oil and cornﬂake breakfast cereal 16–23 cents more per pound for fresh pork used a dichotomous response question, as the with reduced levels of saturated fats. Buzby et respondents were asked to choose between al. (1995) investigated the consumer willing- GM and non-GM alternatives given the price ness to pay for grapefruit with reduced chem- scenario. For salmon, a polychotomous ical residue. Their results indicate that response question was designed, as surveyed respondents are willing to pay an average of respondents were asked to rank within three 31% more for grapefruit under the 50% risk different types of GM, non-GM but fed with reduction scenario, and 38% more if there is feed containing GM maize or soybeans, and a 99% reduction in the risk associated with non-GM fed with non-GM feed alternatives. In chemical residue. this study, we only analyse a model for the dichotomous choice between GM and non- GM products. Thus, the GM salmon and GM- Methodology fed salmon are combined as the same category in this study. CV has been used to elicit consumer willingness to pay for non-market goods, such as water quality improvement (Carson and The random utility model Mitchell, 1981) or air pollution control (Loehman and De, 1982). CV is also widely We adopt a random utility model for used in evaluating consumer willingness to analysing dichotomous CV responses. pay for food safety, such as reduced food- Following Haab and McConnell (2002), the borne risks (Hammitt, 1986). Although there indirect utility function for respondent j can are several economic tools to value non-mar- be written as: ket goods, CV is generally considered to be U = u (y , Z , ε ) (1) the most appropriate choice for measuring ij j j ij food safety (Buzby et al., 1995). where i is the dichotomous choice (1 as the Among the most important tasks for a CV preferred state and 0 the status quo) and j analysis are questionnaire design and survey refers to the respondent. The determinants

procedure (Haab and McConnell, 2001). The of utility are yj, the jth respondent’s income, CV method uses surveys in which people are Zj, a vector of respondent characteristics ε asked how much they are willing to pay for a and attributes of the choice, and ij, a com- change in the condition of some environmen- ponent of preferences known to the individ- tal resources or a service that is meaningful to ual respondent but not observed by the the respondent in a hypothetical situation researcher. (Diamond et al., 1993; Haab and McConnell, Based on this model, respondent j chooses 2002). Early CV designs involve open-ended the non-GM food if the utility of non-GM food questions such as ‘What is the maximum exceeds the utility of the status quo (GM amount you would pay for…?’ More recently, food), given prices: the commonly used methods include iterative U (Z , y – Pngm ,1, ε ) > bidding, payment cards and dichotomous 1j j j j 1j U (Z , y – Pgm , 0, ε ) (2) choice questions. Several studies reveal that 0j j j j 0j different techniques of asking CV questions where 1 denotes the respondent j choosing provide signiﬁcantly different estimates of the non-GM food, 0 denotes the respondent j

Hicksian surplus (Boyle and Bishop, 1988). choosing the status quo (GM food), Pngmj is In this study, the CV is employed to esti- the price of non-GM food, Pgmj is the price mate the WTP for non-GM food products. of GM food. All other variables were deﬁned The CV scenario in the survey conducted in previously. Consumer - Chap 11 5/3/04 15:56 Page 120

120 H.-Y. Chen and W.S. Chern

Therefore, the probability that the respon- Furthermore, with a logistic distribution, ε has dent thinks he/she is better off by choosing a mean of zero and variance π 2σ 2/3. the non-GM food, given its price can be Normalizing by σ creates a logistic variable expressed as: with mean zero and variance π 2/3. Equation (8) becomes: Prob (Respondent j chooses non-GM food) ε > Prob (non-GM) = Prob [θ < (α Z – β∆P)/σ ] = Prob [U1j(Zj, yj – Pngmj,1, 1j ) j (9) ε = Ψ [α Z /σ – (β∆P/σ )] U0j (Zj, yj – Pgmj,0, 0j )]. (3) j where θ = ε/σ, σ is the standard error and Ψ is the cumulative distribution function. The logistic model Therefore, by using a logistic distribution, the probability of choosing the non-GM product is: With a further assumption of the linear form Prob (non-GM) = [1 + exp ((α Z /σ – j (10) for the utility function, the utility of respon- β∆P/σ))]1. dent j choosing the non-GM food can be speciﬁed as: U = α Z + β (Y – Pngm ) + ε (4) 1j 1 j 1 j j 1j Calculating willingness to pay And the utility of respondent j choosing the GM food is: For calculating the WTP, we need to estimate the parameters α and β for the vector of U = α Z + β (Y – Pgm ) + ε (5) 0j 0 j 0 j j 0j explanatory variables (Haab and McConnell, If respondent j chooses the non-GM food, it 2002). A CV question induces the respondent implies that the utility of choosing the non- to choose between the proposed condition at GM food is greater than that of choosing GM the required payment, and the current state. food: The required payment therefore states the respondent’s willingness to pay in order to > U1j U0j (6) achieve the proposed scenario. In our case, the By assuming the marginal utilities of money WTP is the proposed price of a non-GM prod- (income) for non-GM food and GM food are uct that would make the respondent indifferent β β β between consuming GM (paid with the current identical, i.e. 1 = 0 = , the probability of choosing non-GM food is: price of the GM product) and the non-GM product. Based on this principle, the WTP for α β Prob (non-GM) = Prob [ 1Zj + 1 (Yj – Pngmj) – the non-GM food product can be deﬁned as: α Z – β (Y – Pgm ) > 0] 0 j 0 j j α Z + β (y – WTPngm ) + ε = α Z + = Prob [(α – α ) Z – β (Pngm – 1 j j j 1j 0 j 1 0 j j β ε (11) ε ε > (yj – WTPgmj ) + 0j. Pgmj) + ( 1j – 0j) 0] Solving equation (11) for WTP yields: This can be written more compactly as: WTPngm – WTPgm = αZ /β + ε /β (12) α β ∆ ε > j j j j Prob (Non-GM) = Prob [ Zj ( P) + j 0] (7) where: where, α = α – α α α α 1 0 = ( 1 – 0) ε = ε – ε . ∆P =(Pngm – Pgm) j 1j 0j ε ε ε However, the parameters are unknown and j = ( 1j – 0j). therefore must be estimated. In the expression Assume further that the error term has a for mean, only the ratio of parameter esti- logistic distribution and it is symmetrical. mates is required. Relying on Slutsky’s theo- Therefore, we can derive the probability of rem on consistency, the logit maximum choosing non-GM food as: likelihood estimates for θ = {α/σ, β/σ} are con- α β∆ ε sistent (Haab and McConnell, 2002). Prob (non-GM) = Prob [ Zj – P + > 0] α β∆ < ε Therefore, a consistent estimate of expected = Prob [ ( Zj – P) ] α β∆ > ε willingness to pay for a non-GM food product = 1 – Prob [ ( Zj – P) ] ε < α β∆ = Prob [ ( Zj – P)] (8) derived from equation (12) is: Consumer - Chap 11 5/3/04 15:56 Page 121

Willingness to Pay for GM Foods 121

α β α β E (WTPngmj – WTPgmj| , , Zj) = Zj / (13) Survey questionnaire where α is the vector of the estimated coefficients of the explanatory variables and β is the In the first part of the questionnaire, con- estimated coefficient of the price difference sumer knowledge and awareness of biotech- between a non-GM and GM food product. nology and GM foods are being elicited. Note that the estimated price coefficient Next, respondents were asked about their obtained from equation (9) is (β), and there- attitudes and acceptance toward GM foods, fore the calculation of WTP needs to reverse as well as other GM food-related issues such the sign for β. as environmental concern and pesticide By adopting the logistic model to estimate usage. For most of these questions, five the probability of choosing the non-GM food, options for the response are typically given the econometric model can be specified as the along with an option of ‘Do not know’. For logit model: example, in the question, ‘To what extent do you feel that GM foods are risky, or safe, y = αk + βp + ε (14) for human health?’, the respondent was 1 if the respondent chooses the where y = non-GM food product given the choices of ‘Extremely risky’, ‘Very { 0 otherwise. risky’, ‘Somewhat risky/Somewhat safe’, ‘Very safe’, ‘Extremely safe’ and ‘Do not Also, k is a vector of explanatory variables know’. Thirdly, respondents were asked and p is price factor. In our empirical model, about their support for GM food labelling the price factor is defined as the ‘price differ- and type of labelling. ence’ between non-GM and GM food in order Afterwards, a CV scenario was presented to capture the price effect and WTP can be along with the food products and price com- estimated as the expected premium for non- binations. Surveyed respondents were first GM food. asked about their consumption habit and frequency of the food product. Then, they were asked to select or rank the food prod- The Survey and Data ucts according to different GM contents, given the prices. In designing the price A mail survey was conducted in the Columbus matrix, we assumed that GM food products Metropolitan Area, Ohio, in March 2001. A are cheaper than their non-GM counter- three-wave procedure combined with mailing and telephone was used in order to maximize parts. Therefore, we specified the prices of the response rate. The sampling frame was GM food products by taking a discount of obtained from the Center for Survey Research those of non-GM food products, which were at The Ohio State University. The Center ran- based on the market prices. The discount domly selected 650 telephone subscribers in ranged from 10% to 25%. Table 11.1 pre- the Columbus Metropolitan Area based on the sents the four price scenarios used in the zip code. The questionnaires and postage- survey. Note that the market prices paid return envelopes were mailed to these observed in Columbus, Ohio, at the time of randomly selected households. In total, 141 survey were used as the base prices. In two completed survey questionnaires were versions, these base prices were changed returned, along with 120 undeliverable return- slightly to provide more variations in prices. ing questionnaires, yielding an overall Even though this range of differences response rate of 26.6%. Four versions of the between the GM and non-GM products questionnaire with different prices in the CV seems reasonable, the specific price varia- section were equally distributed among the tions chosen are somewhat subjective. 650 mailings. Among the 141 returned The last part of the questionnaire con- respondents, we collected 39 copies of ver- tained the demographic information such as sion 1, 28 of version 2, 26 of version 3 and age, sex, race, income, education, religion, 48 of version 4. (These four versions of prices occupation, etc. (A copy of the questionnaire will be discussed later.) is available upon request.) Consumer - Chap 11 19/3/04 9:06 Page 122

122 H.-Y. Chen and W.S. Chern

Table 11.1. Price matrix for CV design. Cornﬂake Vegetable oil price Salmon price cereal price ($ per 32 ﬂ oz) ($ per pound) ($ per 18 oz) Version Non-GM GM Non-GM GM-fed GM Non-GM GM

1 (10% difference) $2.49a $2.24 $6.99a $6.29 $5.66 $4.39a $3.95 2 (20% difference) 2.49a 1.99 6.99a 5.59 4.47 4.39a 3.51 3 (15% difference) 2.19 1.86 5.99 5.09 4.33 3.79 3.22 4 (25% difference) 2.19 1.64 5.99 4.49 3.37 3.79 2.84

aThese are the observed reference market prices of the products during the survey period.

Variables show that the variables related to Attitude, Perception, Labelling, Demographic, and Price In the logistic model, the dependent variable is have significant effects on consumer choices a binary variable being one if the respondent between GM and non-GM food products. The chose the non-GM food and zero otherwise. knowledge and awareness variables, however, Note that in the case of salmon, our survey appear to be not statistically significant. Let us collected data on the ranking of the three discuss these findings in more detail. types of salmon, i.e. GM salmon, non-GM but fed with GM feed, and non-GM fed with non- GM feed. However, in the regression model, Attitude we simply estimated the probability of choosing non-GM versus GM salmon. We grouped Results indicate that the risk perception of GM GM salmon and non-GM salmon but fed with foods places an important impact on GM food GM feed together as GM salmon in the analy- consumption, as higher risk perception gener- sis. An extension of the model dealing with ates lower GM food consumption. The per- three separate choices in a multinomial logit centage of organic food purchase, used as an model would be desirable for future research. indicator of attitude towards risk, is insignifi- From the survey data, various explanatory cant in the cornflake cereal model but signifi- variables can be grouped into six categories: cant in the vegetable oil and salmon models. ‘Knowledge and Awareness’, ‘Attitude’, Environmental concern of GM foods is also a ‘Perception’, ‘Labelling’, ‘Demographic’ and significant factor determining GM food con- ‘Price’. Variable definitions and the sample sumption, so is religious or ethical concern. means used in the econometric model are pre- Further, the perceived difference between GM sented in Table 11.2. Since the models for veg- and non-GM food affects the willingness to etable oil and cornflakes are based on a slightly consume GM foods, implying that if the per- different sample size than the model for ceived difference is not huge, consumers are salmon, different descriptive statistics are more willing to consume GM foods. shown. It is interesting to note that only 58% of the respondents considered themselves as either ‘very well’ or ‘somewhat’ informed about Perception GMOs or GM foods. Furthermore, a majority of the respondents thought GM foods are Price attribute is a significant determinant for ‘somewhat risky’ to human health (53%), while willingness to consume GM foods, as the only 19% replied ‘extremely or very safe’. respondent’s concern on price tends to induce them to consume more GM food, which is assumed to be cheaper than their tra- Empirical Results ditional counterpart. Interestingly, those who think it is most important to reduce saturated Table 11.3 presents the regression results for fats in GM vegetable oil still tend to consume the three food products: vegetable oil, salmon more non-GM vegetable oil. On the other and cornflake breakfast cereal. The results hand, those who believe the most important Consumer - Chap 11 5/3/04 15:56 Page 123

Willingness to Pay for GM Foods 123

Table 11.2. Variable definition and sample means. Sample mean (vegetable oil Sample and cornflake mean Variable/category Definition and coding cereal) (salmon)

Knowledge and awareness KNOW 1 if very well/somewhat informed of GMOs or GM foods; 0.58 0.59 0 otherwise OIL 1 if aware of GM vegetable oil; 0 otherwise 0.35 CF 1 if aware of GM vegetable cornflake cereal; 0 otherwise 0.36 GMS Percentage of GM foods sold in the market place (guessed) 38% 38% Attitude RP1 1 if GM foods are extremely/very risky; 0 otherwise 0.09 0.08 RP2 1 if GM foods are somewhat risky; 0 otherwise 0.53 0.57 RP3 1 if GM foods are extremely/very safe; 0 otherwise 0.19 0.20 (Focus group: do not know) O1 1 if 1–20% organic food purchase; 0 otherwise 0.57 0.58 O2a 1 if more than 20% (or 21–40%) organic food purchase; 0.10 0.06 0 otherwise purchase; 0 otherwise O3b 1 if more than 40% organic food purchase; 0 otherwise 0.06 (Focus group: 0% and do not know) EN1 1 if GM technology is extremely/very beneficial to the 0.21 0.22 environment; 0 otherwise EN2 1 if GM technology is extremely/very risky to the environment; 0.11 0.12 0 otherwise (Focus group: Somewhat risky) REL 1 if religious concerns are extremely/very important; 0.18 0.21 0 otherwise PESTICID 1 if large/some pesticide decrease after applying GM 0.43 0.48 technology; 0 otherwise DIF1 1 if GM and non-GM foods are extremely/very different; 0.39 0.39 0 otherwise DIF2 1 if GM and non-GM foods are not very/not at all different; 0.26 0.28 0 otherwise (Focus group: I have no idea) CON 1 if excellent/good government performance in food safety; 0.38 0.38 0 otherwise Perception SATFB 1 if the respondent believes the potential to reduce saturated 0.17 fats in foods is the most important benefit of GM foods; 0 otherwise PESTB 1 if the respondent believes the potential to reduce pesticides 0.67 0.74 in foods is the most important benefit of GM foods; 0 otherwise PRICE 1 if the respondent ranks ‘price’ as the first and second 0.25 important food attribute; 0 otherwise

SAFETY 1 if the respondent ranks ‘safety’ as the ﬁrst and second 0.45 important food attribute; 0 otherwise TASTE 1 if the respondent ranks ‘taste’ as the ﬁrst and second 0.56 important food attribute; 0 otherwise Labelling LABEL 1 if labelling of GM foods is extremely/very important; 0.67 0.65 0 otherwise

Continued Consumer - Chap 11 5/3/04 15:56 Page 124

124 H.-Y. Chen and W.S. Chern

Table 11.2. Continued. Sample mean (vegetable oil Sample and corn- mean Variable/category Deﬁnition and coding ﬂake cereal) (salmon)

Demographic AGE1 If <34 years old; 0 otherwise 0.11 0.11 AGE2 If 35–60 years old; 0 otherwise 0.57 0.65 (Focus group: >60 years old) GENDER 1 if males; 0 otherwise 0.44 0.44 MARITAL 1 if married; 0 otherwise 0.53 0.56 EDU1 1 if some college and associate degree; 0 otherwise 0.27 0.25 EDU2 1 if bachelor degree, some graduate school and graduate degree; 0 otherwise 0.44 0.46 (Focus group: No high school, some high school and high school diploma) IN1 1 if income $30,001–$50,000; 0 otherwise 0.27 0.27 IN2 1 if income $50,001–$70,000; 0 otherwise 0.24 0.25 IN3 1 if income more than $70,001; 0 otherwise 0.21 0.23 (Focus group: Less than $30,000) RACE 1 if Caucasian; 0 otherwise 0.86 0.86 RELIGION 1 if Protestant; 0 otherwise 0.45 0.42 CHD 1 if there is one or more children under 18 years old in the 0.31 0.33 household; 0 otherwise Price POIL Price difference between non-GM and GM oil 0.42 PSAL Price difference between non-GM and GM salmon 1.62 PCF Price difference between non-GM and GM cornﬂake cereal 0.72 Sample size 105c 92c

aO2 = 1 if 21–40% organic food purchase in the salmon model. bO3 = 1 if more than 40% organic food purchase in the salmon model. c The usable sample sizes are smaller than 141 because of missing data in the CV section of the survey.

benefit of GMOs is to reduce pesticide usage cation in the salmon model. Income dummies tend to consume more GM salmon and GM are highly significant in the cornflake break- cornflake breakfast cereal. fast cereal model and have negative effects, implying that the people with higher income tend to consume more GM cornflake break- Labelling fast cereal. This result is somewhat surprising. Note that the number of children within the The opinion on labelling is a significant factor household has a significant negative effect on in the salmon and maize flake breakfast cereal respondents’ willingness to consume GM models, showing that the more important the foods, as the concern for younger children in respondents think that GM food labelling is, the household would certainly decrease the the more non-GM salmon and maize flake consumption of GM foods. breakfast cereal they are going to consume.

Price Demographic Price is highly signiﬁcant in the three mod- Demographic characteristics turned out to be els, suggesting that lower prices of GM insigniﬁcant with respect to age, gender, mari- foods encourage consumers to consume tal status and education, except age and edu- more GM products. Consumer - Chap 11 5/3/04 15:56 Page 125

Willingness to Pay for GM Foods 125

Table 11.3. Regression results.

Vegetable oil Salmon Cornﬂake cereal

Variable/category Coeff. t-Ratio Coeff. t-Ratio Coeff. t-Ratio

Constant 0.2420 0.069 10.6209 1.806 8.6643 1.744 Knowledge and awareness KNOW 0.8311 0.758 1.6416 1.008 0.9009 0.662 OIL 0.3423 0.331 CF 0.0907 0.082 GMS 0.0322 1.059 Attitude RP1 6.0563 2.073*** 2.7625 0.756 9.3633 1.879** RP2 3.9027 1.985** 1.3237 0.870 1.7853 1.173 RP3 0.6010 0.260 6.0145 1.518* 2.4177 0.916 O1 0.2036 0.172 0.4022 0.250 1.3369 1.074 O2 2.9776 1.370* 10.9137 2.631*** 1.8885 0.894 O3 0.6134 0.205 EN1 8.5130 2.051*** 4.3476 1.512* 4.1532 1.697* EN2 3.7732 1.444* 0.0407 0.014 0.1628 0.079 REL 7.5893 2.148*** 5.2668 1.972** 2.6617 1.165 PESTICIDE 1.2412 1.021 1.9062 1.312* 0.2328 0.249 DIF1 0.3020 0.277 0.5385 0.387 0.5233 0.465 DIF2 2.4975 2.011** 2.7618 1.515* 3.8577 1.767** CON 0.5099 0.523 1.4127 1.003 0.1416 0.140 Perception PRICE 5.3223 2.591*** 6.2551 2.317*** SAFETY 2.5639 1.525* 0.2956 0.208 TASTE 0.6826 0.624 0.3861 0.268 SATFB 2.8779 2.016** PESTB 2.8351 1.551* 1.7446 1.487* Labelling LABEL 0.6757 0.641 3.8607 1.871** 2.6327 2.085*** Demographic AGE1 1.3395 0.592 4.4977 1.033 0.2374 0.070 AGE2 0.3099 0.173 2.6670 1.509* 1.7804 0.955 GENDER 1.3487 1.157 0.5547 0.313 0.3989 0.341 MARITAL 0.5832 0.592 1.3714 0.816 1.2198 1.071 EDU1 0.8810 0.522 1.8517 0.826 0.2328 0.151 EDU2 1.1396 0.823 2.2280 1.324* 1.2022 0.866 IN1 1.7423 1.127 0.2659 0.147 6.8875 2.247*** IN2 2.8542 1.646* 0.4146 0.166 5.7335 2.021** IN3 1.1893 0.753 1.8905 0.861 5.6907 1.940** RACE 0.7852 0.499 0.7364 0.446 1.6685 1.004 RELIGION 1.4551 1.254 5.1121 2.044** 1.4260 1.056 CHD 2.6626 1.969** 3.9305 2.491*** Price POIL 7.7657 1.775** PSAL 5.8891 2.215*** PCF 5.8173 1.702** Number of observations 105 92 105 McFadden R 2 0.5944 0.6537 0.6177

Coeff., coefficient. ***2.5% significance level; **5% significance level; *10% significance level. Consumer - Chap 11 5/3/04 15:56 Page 126

126 H.-Y. Chen and W.S. Chern

McFadden R2 values in these models range are simply means or averages of all house- from 0.5944 to 0.6537, which are actually holds in the sample. The WTP for a non-GM quite high for this type of cross-sectional data. product reflects the premium for the non-GM In general, the results indicate that the willing- food that the consumer is willing to pay. We ness to purchase GM foods is heavily influ- also compute the percentage of premium enced by the risk perception of GM foods to using the price of the GM food as the base. human health, environmental concern and Since there are different prices for GM foods religious concern when consuming GM foods, used in the four versions of price scenarios, as well as the perceived difference between the percentage figures vary depending on the GM and non-GM foods. Also, the importance base price. The results show that the survey of food characteristics such as ‘price’ will respondents are willing to pay a premium of affect consumers on their GM food consump- 5–8% for non-GM vegetable oil, 15–28% for tion. The sensitivity to price is also reflected non-GM salmon and 12–17% for non-GM by the significance of the price factor, show- cornflake breakfast cereal. ing that more GM food products will be cho- Table 11.5 shows the computed WTP pre- sen if the price difference between non-GM miums for various demographic groups by and GM foods increases. sex, age and race. It is interesting to observe Demographic variables are not very signifi- that the WTP premiums for non-GM foods cant. Only income and the number of children vary by demographic group. Note that even in the household affect the consumer’s pur- though some demographic variables may not chase decision. Surprisingly, the respondents be significant in the logit model, the computed with higher income tend to consume more WTP premiums can still be different among GM cornflake breakfast cereal, implying that demographic groups. This is because the wealthy people are more confident on this WTP is based on the entire model and the GM product, and would not view it as particu- entire set of estimated parameters, not just larly risky. We are not sure whether or not the coefficient related to a particular demo- this result is caused by the fact that higher- graphic variable. The results are very telling income households in the USA tend to con- that female respondents are always willing to sume more breakfast cereals. Furthermore, pay a higher premium for non-GM food prod- breakfast cereal is considered to be a relatively ucts than male respondents, especially in the expensive food item. case of vegetable oil and cornflake breakfast cereal. This finding is in accordance with previous studies regarding consumer WTP on Willingness to Pay for Non-GM Foods organic food produce (Huang, 1993). Survey respondents between 35 and 60 Based on the methodology described above, years old tend to pay higher premiums for the WTP for the three non-GM foods can be non-GM salmon and non-GM vegetable oil computed for the entire sample and the than those who are younger than 35 or older results are presented in Table 11.4. Note that than 60. Furthermore, the respondents we compute first the WTP household by younger than 35 are willing to pay more for household. The figures presented in the table non-GM cornflake breakfast cereal than the

Table 11.4. Estimated WTP premiums for non-GM food products. Cornﬂake Vegetable oil Salmon breakfast cereal Item ($ per 32 ﬂ oz) ($ per pound) ($ per 18 oz)

WTP premium $0.13 $0.96 $0.49 Percentage of premiuma 5~8% 15~28% 12~17%

aPercentage of premium = (WTP premium/price of GM food product) 100. Consumer - Chap 11 5/3/04 15:56 Page 127

Willingness to Pay for GM Foods 127

Table 11.5. Estimated WTP premiums for non-GM food products by demographic groups.

Sex Age Race

<35 years 35–60 60 years Product/item Female Male old years old >old Non-white White

Vegetable oil (32 ﬂ oz) WTP premium $0.34 $0.16 $0.27 $0.39 $0.31 $0.54 $0.08 Percentage of premium 15~21% (7) ~ (10)% (12)~(16)% 17~24% (14)~(19)% 24~33% 4~5% Salmon (per pound) WTP premium $1.13 $0.74 $0.51 $1.05 $0.90 $1.53 $0.86 Percentage of premium 18~34% 12~22% 8~15% 17~31% 14~27% 24~45% 14~26% Cornﬂake breakfast cereal (18 oz) WTP premium $0.70 $0.16 $0.66 $0.59 $0.15 $1.28 $0.38 Percentage of premium 18~25% 4~6% 17~23% 15~21% 4~5% 32~45% 10~13%

other age groups. The senior respondents, Implications and Discussion however, are the least willing to pay higher prices for non-GM vegetable oil and cornflake The purposes of this study are to conduct an breakfast cereal, but willing to pay more analysis on GM food consumption and to mea- premiums for non-GM salmon than those who sure consumer willingness to pay for non-GM are younger than 35. This finding suggests vs. GM foods. The empirical results show that that middle-aged consumers tend to put more the willingness to consume GM foods depends concern on food safety issues than those who on risk perception, environmental concern, reli- are younger or older and therefore are willing gious and ethical concern of GM foods, opinion to pay more for non-GM foods. Besides, their on labelling of GM foods, and perceived differ- income sources are more stable as compared ence between GM and non-GM foods. Also, the to younger and older generations, and there- number of children within the household is a fore middle-aged consumers are more willing key determinant on GM food consumption. In to pay a premium for non-GM food products. addition, the price factor of GM foods is fairly On the other hand, senior citizens are less will- significant to the respondents in the survey, sug- ing to pay more for non-GM vegetable oil and gesting that by advertising GM food products cornflake breakfast cereal than that for non- with a lower price, the consumption of GM GM salmon which indicates that senior citizens foods might be substantially increased. are less sensitive to food safety, especially These results imply that in order to gain those food products that are less relevant to consumer acceptance of GM foods, it is their consumption compared to younger gen- important to change their risk perception of erations, such as cornflake breakfast cereal. GM foods and to deviate their other concerns. Results also show that non-White respon- The survey results show that only 59% of the dents are more likely to pay a premium for respondents indicated that they are either very non-GM food products than White respon- well or somewhat informed on GMOs or GM dents. Note that the difference in WTP foods. In fact, the majority of consumers are between the two racial groups is dramatic. still not very well informed about GM foods. Non-White respondents are willing to pay a Therefore, how can we change the con- premium of at least 24% for the three food sumer’s perception? The government, food products, while White respondents are only industry and consumer groups have to provide willing to pay a maximum premium of 26%. unbiased information to the consumer. If the This finding is somewhat surprising and it may information can change consumers’ percep- suggest that the non-White respondents may tions, then the willingness to buy GM foods lack confidence on food safety and therefore would increase. Therefore, the effectiveness are willing to pay a higher premium for non- of the information is very crucial to the suc- GM foods than the White respondents. cess of GM foods in the future. Consumer - Chap 11 5/3/04 15:56 Page 128

128 H.-Y. Chen and W.S. Chern

The econometric results also show that the foods as perceived by the respondents is found respondents are willing to pay a premium, to be the main hindrance to the consumer’s ranging from 5% for non-GM vegetable oil to acceptance of such foods, which reinforces the 28% for non-GM salmon. Clearly, these necessity to educate the general public to be results imply that the consumer must see a more aware of GM foods with more unbiased tangible benefit in order for them to buy GM scientific information. Also, the result points to foods. Therefore, the future of GM foods is the importance of GM food labelling, implying critically dependent upon the ability to reduce the need to provide the consumer with more the price for GM foods as compared to their information on GM foods so that consumer traditional non-GM counterparts. Therefore, confidence can be established. Moreover, the the stress on the indifference between GM price factor is significant in determining the and non-GM foods is unlikely to induce the purchase of GM foods, suggesting that lower consumer’s willingness to buy GM foods. price can be a useful tool to stimulate GM food Our results also show that the consumer consumption. WTP for non-GM foods varies among demo- The results of WTP indicate that the survey graphic characteristics. Specifically, female respondents are willing to pay a premium in respondents, those aged between 35 and 60, order to differentiate between GM and non- and non-White respondents are willing to pay GM foods. This implies that producers of non- a higher premium for non-GM foods than GM foods might benefit from the labelling other groups. This finding is useful to the gov- policy. If consumers are willing to bear the ernment, food industry and consumer groups premium for non-GM foods, producers do not for designing appropriate programmes to edu- need to fully absorb the cost of segmenting cate the consumer about GMOs and GM the market. From the government standpoint, foods targeted to different demographic labelling of GM foods might cause a warfare groups. loss to the society in the long run if the market is not competitive for both GM and non- GM food products. That is if non-GM food Conclusions products were produced by a few producers. Consumers would pay a higher price in order In this study, we attempt to investigate con- to avoid GM foods, but the prices of non-GM sumer attitudes towards GM foods and to elicit foods were higher than those in a competitive WTP for non-GM foods. The empirical results market. The welfare loss in the long term indicate that the consumer acceptance towards might discourage the government from GM foods is affected by attitudinal factors, enforcing a mandatory labelling policy regard- such as risk perception, environmental ing GM foods. Therefore, it is crucial to edu- impacts, opinion on GM food labelling, per- cate the general public about the ceived difference between GM and non-GM characteristics of GM foods so that the risk foods, and the potential benefits of GM foods. perception associated from consuming such Overall, the high risk associated with GM foods can be mitigated.

References

Boyle, K.J. and Bishop, R.C. (1988) Welfare measurements using contingent valuation: a comparison of techniques. American Journal of Agricultural Economics 70, 20–28. Burton, M., James, S., Ridby, D. and Young, T. (2001) Consumer attitudes to genetically modiﬁed organisms in food in the UK. Paper presented at the 71st EAAE Seminar ‘The Food Consumer in the Early 21st Century’, Zaragoza, Spain, 19–20 April. Buzby, J.C., Skees, J.R. and Ready, R.C. (1995) Using contingent valuation to value food safety: a case study of grapefruit and pesticide residues. In: Caswell, J.A. (ed.) Valuing Food Safety and Nutrition. Westview Press, Boulder, Colorado, pp. 219–256. Consumer - Chap 11 5/3/04 15:56 Page 129

Willingness to Pay for GM Foods 129

Carson, R.T. and Mitchell, R.C. (1981) An experiment in determining willingness to pay for national water quality improvements. Office of Policy Analysis Draft report, US Environmental Protection Agency, Washington, DC. Diamond, P.A., Hausman, J., Leonard, G.K. and Denning, M.A. (1993) Does contingent valuation measure preferences? Experimental evidence. In: Hausman, J.A. (ed.) Contingent Valuation: A Critical Assessment. North Holland, New York, pp. 41–89. Gallop Market Survey Corp., Taiwan (2002) Public Opinion Survey of Genetically Modified Foods. Report publishe by Department of Health, The Executive Yuan, Republic of China (Taiwan), August 22. Gaskell, G. (2000) Agricultural biotechnology and public attitudes in the European Union. AgBioForum 3, 87–96. Haab, T. and McConnell, K.E. (2002) Valuating Environmental and Natural Resources: the Econometrics of Non-Market Valuation. Edward Elgar, Cheltenham, UK. Halbrendt, C., Sterling, L., Snider, S. and Santoro, G. (1995) Contingent valuation of consumer willingness to purchase pork with lower saturated fat. In: Caswell, J.A. (ed.) Valuing Food Safety and Nutrition. Westview Press, Boulder, Colorado, pp. 313–339. Hammitt, J.K. (1986) Estimating Consumer Wilingness-to-Pay to Reduce Foodborne Risk. Prepared by Rand Corporation for US Environmental Protection Agency, R-3447-EPA, Washington, DC. Hoban, T.J. (1996) How Japanese consumers view biotechnology. Food Technology July, 85–88. Hoban, T.J. (1998) Trends in consumer attitudes about agricultural biotechnology. AgBioForum 1, 3–7. Huang, C. (1993) A simultaneous system approach for estimation of consumer risk perceptions, attitudes, and willingness to pay for residue-free produce. Paper presented at the American Agricultural Economics Association Meeting, Orlando, Florida. International Food Information Council Foundation (IFIC) (2001) More U.S. consumers see potential benefits to food biotechnology.’ Wirthlin Group Quorum Surveys, January 2001. Available at http://www.ific.org/proactive/newsroom/release.vtml?id=19241 (accessed March 2001). Loader, R. and Henson, S. (1998) A view of GMOs from the UK. AgBioForum 1, 31–34. Loehman, E. and VoHu, D. (1982) Application of stochastic choice modeling to policy analysis of public goods: a case study of air quality improvements. Review of Economics and Statistics 54, 474–480. Macer, D. and Ng, M.A.C. (2000) Changing attitudes to biotechnology in Japan’. Nature Biotechnology 18, 945–947. Nordic Industrial Fund (2000) Negative attitude to gene-modified food. Available at http://www.nordicin- novation.net. Wang, Q., Halbrendt, C., Kolodinsky, J. and Schmidt, F. (1997) Willingness to pay for rBST-free milk: a two-limit tobit model analysis’. Applied Economics Letters 4, 619–621. Consumer - Chap 11 5/3/04 15:56 Page 130 Consumer - Chap 12 5/3/04 15:56 Page 131

12 A Comparison of Consumer Attitudes towards GM Food in Italy and the USA

Marianne McGarry Wolf,1 Paola Bertolini2 and Jacob Parker-Garcia1 1Agribusiness Department, California Polytechnic State University, San Luis Obispo, CA 93407, USA; 2Facoltà di Economia, Modena, Italy

Introduction items genetically to: be more resistant to plant disease and less reliant on pesticides, 70%; In 1973, science made a very significant help prevent disease, 64%; improve nutri- advancement when a gene was transferred tional value, 58%; improve flavour, 49%; and between plants for the first time. Just 14 extend shelf life, 48%. years later, the first outdoor genetically engi- By contrast, in the European Union (EU) neered plants were grown. By 1995, crops of the consumer generally views GM foods as these plants were cultivated on commercial unhealthy. The politically active ‘Green acreage (Lamb, 2000). Between 1995 and Movement’ has done much to publicize and 1998, biotech companies introduced 16 new put the issue of genetically altered food on genetically modified (GM) crops into the USA the European continent in a negative light. for use by farmers. In the USA, GM foods are For example, it has derisively nicknamed very much a part of American nutrition. genetically altered food ‘Frankenfoods’. A According to the US Department of survey cited by the EU found that most Agriculture (USDA), one-third of the maize Europeans see GM food as health hazards, and more than half the soybeans and cotton despite assurances from producers (Wielaard, grown in the USA are in some way products 2001). In November 1999, the European of biotechnological applications. Total Commission passed a law requiring all hectares of biotech crops grew worldwide European retailers to label food containing from 39.9 million in 1999 to 44.2 million in more than 1% genetically modified ingredi- 2000 (Stickman, 2001). The US share of ents. The Commission also required restau- biotech crops was 69% in 2000. rants to inform consumers if meals contained To date, the USA has managed to avoid GM ingredients. Wolf et al. (2001) found food scares such as mad cow disease and the that the Irish consumer was more likely than other food safety scares that have plagued the US consumer to indicate that mandatory Europe. Americans are much more confident labelling of GM food was very important in about the safety of their food supply and trust their first phase of research, October 1999 government regulation more. This confidence and January 2000. However, after the media has led American consumers to be more coverage of the recall of 2.5 million boxes of accepting of GM foods. A recent study by The Taco Bell brand taco shells produced by Packer (2001a) found that American con- Kraft Foods that contained StarLink™ maize sumers felt it was appropriate to modify food in the USA in September 2000 (Copple,

132 M. McGarry Wolf et al.

2000), Wolf et al. (2001) found that the pro- Methodology portion of consumers in the USA that indicated that mandatory labelling of GM food This research uses a survey instrument that was was very important grew to a level similar to administered through the use of a personal that in Ireland. interview during the autumn of 2001. The ran- Although both the US and Irish consumer dom sample of 232 food shoppers for the indicated that mandatory labelling of GM food USA was collected in San Luis Obispo County, was very important, Wolf et al. (2001) also California. San Luis Obispo County was desig- found that the Irish consumer is less likely to nated the best test market in the USA by purchase food that has been genetically modi- Demographics Daily (Thomas, 2001). San fied than the US consumer. Almost one-half Luis Obispo was found to be the best of 3141 of the Irish respondents indicated they were counties to represent a microcosm of the USA likely to purchase food that has been geneti- based on 33 statistical indicators. The random cally modified, more than two-thirds of the US sample of 252 food shoppers for Italy was col- consumers indicated they were likely to pur- lected in Modena, Italy, during the autumn of chase food that has been genetically modified. 2001. Modena, in Emilia Romagna, is a rich These findings are similar to the results industrial area that represents one of the most reported by The Times in April 2001. A sur- important areas of food production in Italy for vey in the UK indicated that 48% of respon- both industrial food and for typical traditional dents would eat GM food (The Times, 2001). quality food such as parmesan cheese, Modena The results for the USA are similar to the ham, Parma ham and Modena vinegar. In addi- findings of The Packer’s survey of 1000 US tion, Modena is important to the food distribu- consumers (The Packer, 2001a). The tion system of Italy since the largest distribution Packer’s survey found that 60% of consumers group, Coop, resides in Modena. These char- were extremely, very, or somewhat likely to acteristics make Modena an important area to purchase a fresh produce item that has been represent consumers’ attitudes towards food in genetically modified. Italy, especially in northern Italy. Attitudes in Italy appear to be much the This research examines differences in the same as those of other EU countries. In some following between the US and Italian respon- cases, Italian law is stricter towards GM dents: attitudes towards science, food pur- foods. For example, Italian law forbids the chasing behaviour, organic food purchasing use of GM seeds in open fields. They are behaviour, knowledge of organic foods, famil- considered possible health hazards. In April iarity with GM food, and consumer attitudes 2001, a fire destroyed a Monsanto factory in towards GM food and labelling. Consumer northern Italy. Authorities know it was an attitudes toward purchasing GM food are arson attack but are not sure who set the fire. examined based on the purpose for the use of There has been speculation that places biotechnology: to help plants withstand weed- responsibility on the anti-GM, Green Party killers, to improve nutrition, to kill pests and activists. Before the fire, protesters had gath- allow farmers to use less pesticide, and to ered at the Monsanto factory protesting improve taste. Attitudes towards government about the production of genetically modified agencies and food safety and attitudes towards seeds. This incident caused Italian Farm food producers and food safety and the envi- Minister Alfonso Pecoraro Scanio, a member ronment are examined in relation to con- of the Green Party, to seize some 400 tons sumers’ attitudes toward GM food. of suspect Monsanto soybean and maize seeds. Tests revealed the seeds were genetically modified and that they had been Attitudes Toward Science and imported from the USA and were not Italian Technology Use (Deutsche Presse-Agentur, 2001). The objective of this research is to compare con- Both the US and Italian consumers agree that sumer attitudes towards GM food in the USA scientific research is an important factor in the and Italy. quality of life (Table 12.1). However, most US Consumer - Chap 12 5/3/04 15:56 Page 133

Attitudes to GM Food in Italy and the USA 133

Table 12.1. Scientiﬁc research is an important factor in the improvement of the quality of life. USA Italy (n = 232) (n = 252) Chi squared

Strongly agree 45.7% 50.0% Agree 46.6% 44.4% Disagree 4.7% 4.4% Strongly disagree 3.0% 1.2% 2.58

consumers are using the Internet and e-mail In research concerning organic lettuce, at home and less than half of the Italian con- Wolf et al. (2002) have shown that there sumers are using these technologies at home appears to be confusion in consumers’ under- (Table 12.2). standing of the properties of organic food in the USA. For example, in the examination of organic lettuce, it was found that consumers Food Purchasing Behaviour value the organic characteristics of lettuce such as environmentally friendly as somewhat to Approximately two-thirds of consumers in very desirable, while they rate organically Italy and the USA have purchased organic grown and certiﬁed as only slightly to some- food within the past year (Table 12.3). what desirable. Thus, Wolf hypothesized that However, a higher proportion of consumers consumers do not understand the properties of in the USA have purchased organic milk, organic foods (Wolf et al., 2002). This fresh fruits, fresh vegetables and wine. research has attempted to address the possible

Table 12.2. Technology usage at home. USA Italy (n = 232) (n = 252) Chi squared

Internet 81.6% 40.2% 85.72** E-mail 76.8% 31.9% 97.11** None 22.2% 57.5% 69.77**

**Signiﬁcant difference at 0.05 level.

Table 12.3. Organic food consumption within past year. USA Italy Phase 1 (n = 196) (n = 252) Chi squared

Any organic food 67.6% 64.6% 0.475 Meats 20.7% 15.4% 2.35 Milk 31.2% 16.1% 15.29** Other dairy products (excluding milk) 24.2% 21.3% 0.614 Fresh fruits 57.8% 44.9% 8.10** Fresh vegetables 57.8% 44.9% 8.10** Wine 9.1% 4.7% 3.64* Bakery items including bread 20.3% 20.5% 0.003 Other 12.6% 13.4% 0.064

*Signiﬁcant difference at 0.10 for all tests; **signiﬁcant difference at 0.05 level. Consumer - Chap 12 5/3/04 15:56 Page 134

134 M. McGarry Wolf et al.

misconceptions of consumers by examining A higher proportion of consumers in Italy their responses to the question: ‘How strongly plan to increase their purchase of organic do you agree or disagree that all produce sold food in the next year. The majority of con- at a farmers’ market is organic?’ The farmers’ sumers in the USA, 54.7%, expect the quan- markets in the research region in the USA tity of organic food purchased to stay the were observed to sell primarily conventionally same (Table 12.5). grown produce. Therefore, respondents that In the USA, over three-quarters of food pur- either agree or strongly agree are consumers chasers read the label for nutritional informa- that are likely confused about the attributes of tion very or somewhat often when making a organic produce. Over a quarter of consumers purchase decision (Table 12.6). However, in the USA agree that all produce sold at farm- slightly more than half of Italian purchasers ers’ markets is organic (Table 12.4). Only read the label for nutritional information very 12.4% of consumers in Italy agree that all pro- or somewhat often when making a purchase duce sold at farmers’ markets is organic. decision. Approximately two-thirds of respon- Therefore, it appears that the Italian consumer dents in both countries read the label for ingre- has a better understanding of organic food dient information very or somewhat often than the consumer in the USA. when making a purchase decision (Table 12.7).

Table 12.4. All produce sold at a farmers’ market is organic.

USA Italy (n = 232) (n = 252) Chi squared

Strongly agree 2.6% 2.0% Agree 24.5% 10.4% Disagree 52.0% 80.0% Strongly disagree 21.0% 7.6% 43.35**

**Signiﬁcant difference at 0.05 level.

Table 12.5. In the next year, the quantity of purchases of organic food products.

USA Italy (n = 232) (n = 252) Chi squared

Stay the same 54.7% 33.5% Increase 26.7% 44.5% Decrease 4.3% 2.8% Will not purchase organic food products 14.2% 19.3% 25.89**

**Signiﬁcant difference at 0.05 level.

Table 12.6. Nutritional label readership and purchase decision.

USA Italy (n = 232) (n = 252) Chi squared

Very often 45.3% 25.6% Somewhat often 33.2% 29.5% Not very often 15.9% 28.3% Not at all 5.6% 16.5% 35.04**

**Signiﬁcant difference at 0.05 level. Consumer - Chap 12 5/3/04 15:56 Page 135

Attitudes to GM Food in Italy and the USA 135

Table 12.7. Ingredient label readership and purchase decision.

USA Italy (n = 232) (n = 252) Chi squared

Very often 34.5% 32.3% Somewhat often 29.7% 36.6% Not very often 27.5% 21.3% Not at all 8.3% 9.8% 4.166

Familiarity with GM Foods Television is the number one source of information about GM food (Table 12.9). Approximately one-half of consumers in the More than half of Americans and Italians get USA are familiar with GM food (Table 12.8). information about GM food from newspapers However, less than a third of Italian con- and television. Consumers in the USA are sumers indicated that they are familiar with more likely to get information about GM food GM food. These awareness levels for Italy are from radio news; news magazines; discussions signiﬁcantly lower than those found by Wolf with family, friends or colleagues; the Internet; et al. (2001) in their examination of familiar- and by working in farming or food processing. ity with GM food in the USA and Ireland. Television news reporters are considered Approximately half of the respondents in the the most appropriate source of information USA and 40% of respondents in Ireland were about GM food in the USA and Italy (Table at least somewhat familiar with GM food. 12.10). However, there are very different

Table 12.8. Familiarity with genetically modiﬁed food.

USA Italy (n = 232) (n = 252) Chi squared

Very familiar 8.3% 2.4% Somewhat familiar 41.3% 25.2% Not very familiar 34.8% 55.5% Not at all familiar 15.7% 16.9% 29.14**

**Signiﬁcant difference at 0.05 level

Table 12.9. Sources of genetically modiﬁed food awareness.

USA Italy (n = 232) (n = 252) Chi squared

Television news 62.8% 78.3% 14.23** Newspaper 54.3% 51.6% 0.373 Radio news 21.6% 9.4% 13.92** News magazines 31.9% 13.0% 25.24** Consumer Reports magazine 12.2% 17.0% 2.23 Discussion with family, friends or colleagues 38.4% 16.9% 28.15** Internet 14.7% 3.9% 16.92** Employment, work in farming or food processing 7.8% 2.8% 6.22** Other 3.0% 2.0% 0.565

**Signiﬁcant difference at 0.05 level. Consumer - Chap 12 5/3/04 15:56 Page 136

136 M. McGarry Wolf et al.

Table 12.10. Appropriate sources of information concerning genetically modiﬁed food.

USA Italy (n = 232) (n = 252) Chi squared

Television news reporters 62.80% 55.10% 2.92* Newspaper reporters 60.20% 29.90% 44.86** Local government agencies 44.20% 22.00% 26.92** Farmers 41.40% 6.30% 84.14** Discussions with family, friends, or colleagues 35.90% 5.90% 67.64** Education seminar 34.50% 0% 104.85** University professors 29.40% 40.60% 6.55** Radio news reporters 29.30% 5.90% 46.90** Internet websites 25.00% 4.30% 42.52** Science teachers 24.10% 0% 69.30** Representatives from food processors 22.40% 4.70% 33.18** Visits to food production facilities 17.70% 0% 49.25** Local politicians 15.50% 0% 42.57** Reports from a seed producer 12.50% 0% 33.77** E-mail 8.70% 0% 23.14** Other 4.7% 0.4% 9.52**

*Signiﬁcant difference at 0.10 level; ** signiﬁcant difference at 0.05 level.

opinions between the consumers in the USA past’; ‘global food producers are producing and Italy concerning the appropriateness of food using environmentally safe methods’; other sources of information. The second and ‘computer technology is an important fac- most important source of information for the tor in the improvement of the quality of life’. consumers in Italy is the university professor, The following rating scale was used to evalu- the seventh most important source of infor- ate these statements: strongly agree = 4; mation for those in the USA. Less than a agree = 3; disagree = 2; strongly disagree = 1 third of the consumers in Italy indicated that (Table 12.11). newspaper reporters are appropriate sources Respondents in the USA are more likely of information, while 60% of consumers in than the Italian respondents to agree that: the USA indicated that newspaper reporters ‘government agencies in my country have are appropriate sources of information. done a very good job at ensuring food safety Therefore, an educational campaign concern- in the past’; ‘I trust government agencies in ing GM food will need to include television my country to ensure food safety in the news in both countries. However, other future’; ‘global food producers have done a sources of information will differ between the very good job at ensuring food safety in the two countries. past’ and ‘global food producers are producing food using environmentally safe methods’.

Attitudes Towards Government Agencies and Food Producers Attitudes Towards Genetically Modiﬁed Food Respondents were asked how strongly they agree or disagree with the following state- Most consumers in both countries indicated ments: ‘government agencies in my country that government imposition of mandatory have done a very good job at ensuring food labelling is important, 98% in Italy and 89.6% safety in the past’; ‘I trust government agen- in the USA (Table 12.12). This shows that cies in my country to ensure food safety in the most Italians agree with the current law of future’; ‘global food producers have done a mandatory labelling GM foods and Americans very good job at ensuring food safety in the want the government to label if the foods are Consumer - Chap 12 5/3/04 15:56 Page 137

Attitudes to GM Food in Italy and the USA 137

Table 12.11. Mean ratings. USA Italy Statement (n = 232) (n = 252) t-Test

Government agencies in my country have done a very 2.99 2.12 14.41** good job at ensuring food safety in the past I trust government agencies in my country to ensure 2.94 2.57 6.02** food safety in the future Global food producers have done a very good job at 2.44 2.04 6.5** ensuring food safety in the past Global food producers are producing food using 2.29 2.10 3.06** environmentally safe methods

**Signiﬁcant difference at 0.05 level.

Table 12.12. Government imposition of mandatory labelling for genetically modiﬁed food.

USA Italy (n = 232) (n = 252) Chi squared

Very important 61.3% 84.35% Somewhat important 28.3% 13.4% Not very important 10.0% 2.0% Not at all important 0.4% 0.4% 35.19**

**Signiﬁcant difference at 0.05 level.

genetically modiﬁed. More Americans felt consumers, half indicate that they would at mandatory labelling was not very important, least maybe buy GM foods. The consumers in 10%, compared to Italians, 2%. A compari- the USA observed in the research generated son of the results observed by Wolf et al. by Wolf et al. (2001) and in this research are (2001) in Ireland and the USA indicates that more likely to consume GM food than their the Italian consumer rates the importance of European counterparts. labelling higher than both the Irish and US consumers. Most Americans, 71.6%, indicate that they Attitudes Towards Different Attributes of would at least maybe buy GM foods (Table Genetically Modiﬁed Food 12.13). Only 43.1% of Italians will possibly buy GM foods. More than half of Italians Respondents were asked how likely they would not purchase GM foods. The results were to purchase GM food to improve nutri- observed by Wolf et al. (2001) indicate that tion, kill pests allowing farmers to use less the consumers in Ireland are slightly more pesticides, improve taste and help plants likely to purchase GM food than the Italian withstand weedkillers. The rating scale used

Table 12.13. Likelihood to buy genetically modiﬁed food.

USA Italy (n = 232) (n = 252) Chi squared

Deﬁnitely, probably, maybe 71.6% 43.1% Probably not, deﬁnitely not 28.4% 56.9% 39.95**

**Signiﬁcant difference at 0.05 level. Consumer - Chap 12 5/3/04 15:56 Page 138

138 M. McGarry Wolf et al.

to evaluate purchase interest is: 5 = defi- cides. Over 40% of the Italian consumers nitely; 4 = probably; 3 = maybe; 2 = proba- expect to increase their purchases of organic bly not; and 1 = definitely not. The US foods in the next year. The consumers in respondents were more likely to purchase Italy are less willing to purchase a GM food GM food products overall and for the pur- product that helps plants withstand weed- poses of improving nutrition, helping plants killer, improves nutrition, or improves taste. withstand weedkillers and improving taste However, the consumer in the USA, that is (Table 12.14). The US and Italian consumers more likely to read a nutrition label, rates evaluated the importance of using biotechnol- improved nutrition the highest attribute for a ogy to kill pests and allow farmers to use less GM food product. The second most impor- pesticides similarly. A similar proportion of tant attribute is to kill pests and allow farm- consumers in both countries purchased ers to use less pesticides for the consumer in organic food in the past year. the USA. A comparison of the importance of specific uses of biotechnology is given in Table 12.15 for each country. The attributes are Attitudes Towards Government Agencies listed from high to low based on their means and Producers, and Willingness to and paired to examine differences between Purchase GM Foods the means for each country. The only attribute that achieves a maybe will pur- The responses to the questions concerning chase among the Italian consumers is that to trust in government and global food producers kill pests and allow farmers to use less pesti- to ensure food safety were compared with the

Table 12.14. Likelihood to buy – attribute mean rating. USA Italy Statement (n = 232) (n = 254) t-Test

How likely are you to purchase a food product that has 2.98 2.26 7.50** been genetically modiﬁed? To kill pests and allow farmers to use less pesticides? 3.26 3.20 0.53 To help plants withstand weedkillers? 3.00 2.71 2.88** To improve nutrition? 3.43 2.70 6.96** To improve taste? 3.16 1.83 13.23**

**Signiﬁcant difference at 0.05 level.

Table 12.15. Likelihood to buy – attribute mean rating. Mean rating Paired t

Italy (n = 254) To kill pests and allow farmers to use less pesticides? 3.20 To help plants withstand weedkillers? 2.71 7.38** To improve nutrition? 2.70 0.012 To improve taste? 1.83 12.72** USA (n = 232) To improve nutrition? 3.43 To kill pests and allow farmers to use less pesticides? 3.26 3.63** To improve taste? 3.16 1.66* To help plants withstand weedkillers? 3.00 2.74**

*Signiﬁcant difference at 0.10 level; **signiﬁcant difference at 0.05 level. Consumer - Chap 12 5/3/04 15:56 Page 139

Attitudes to GM Food in Italy and the USA 139

willingness of a respondent to purchase GM the belief that government agencies ensured food. The results show that for the US food safety in the past and willingness to pur- respondents, those who believe the govern- chase a GM food product. Most Italian ment and global food producers ensure food respondents evaluated the government and safety are more willing to purchase GM food global food producers lower than the US (Tables 12.16 and 12.17). The Italian respon- respondents. Both the Italian and US con- dents who believe the global food producers sumers are more likely to purchase a GM food ensure food safety are more willing to pur- product when they agree that global food pro- chase GM food. However, for the Italian ducers are using environmentally safe meth- respondents there is no relationship between ods (Table 12.18).

Table 12.16. Mean willingness to purchase a genetically modiﬁed food product: government agencies in my country have done a very good job at ensuring food safety in the past. Mean t statistic

Italy (n = 252) Strongly agree or agree 2.46 1.62 Strongly disagree or disagree 2.19 USA (n = 232) Strongly agree or agree 3.08 2.83** Strongly disagree or disagree 2.57

**Signiﬁcant difference at 0.05 level.

Table 12.17. Mean willingness to purchase a genetically modiﬁed food product: Global food producers ensure food safety. Mean t statistic

Italy (n = 252) Strongly agree or agree 2.70 2.89** Strongly disagree or disagree 2.16 USA (n = 232) Strongly agree or agree 3.31 4.90** Strongly disagree or disagree 2.69

**Signiﬁcant difference at 0.05 level.

Table 12.18. Mean willingness to purchase a genetically modiﬁed food product: Global food producers are producing food using environmentally safe methods. Mean t statistic

Italy (n = 252) Strongly agree or agree 2.59 2.75** Strongly disagree or disagree 2.14 USA (n = 232) Strongly agree or agree 3.34 4.60** Strongly disagree or disagree 2.77

**Signiﬁcant difference at 0.05 level. Consumer - Chap 12 5/3/04 15:56 Page 140

140 M. McGarry Wolf et al.

Conclusions The only attribute that achieves a maybe will purchase among the Italian consumers is The objective of this research is to use a case the attribute, to kill pests and allow farmers to study to compare consumer attitudes toward use less pesticides. Approximately 40% of the GM food in the USA and Europe using two consumers in Italy expect to increase their communities. The results of this research indi- organic purchases in the next year. The con- cate that half of the consumers in the USA sumers in Italy are less willing to purchase a are familiar with GM food, while only 28% of GM food product that helps plants withstand the consumers in Italy are familiar with GM weedkillers, improves nutrition or improves food. This level of familiarity in Italy is lower taste. However, the consumer in the USA, than that observed in Ireland by Wolf et al. in who is more likely to read a nutrition label, 2001. Both the consumers in the USA and rates improved nutrition the highest attribute Italy agree that the most appropriate source for a GM food product. The second most for information about GM food is television important attribute for the consumer in the news reporters. The consumers in the USA USA is to kill pests and allow farmers to use indicate that there are many appropriate less pesticides. sources of information for GM food. The consumers responded to questions However, the Italian consumers indicate that concerning trust in government and global there are few appropriate sources and rate food producers to ensure food safety. These university professors high relative to the US questions are compared with the willingness consumers. of a respondent to purchase GM food. The The Italian consumer is less likely to read results show that for the US respondents, nutrition labels, equally likely to read ingre- those who believe the government and global dient labels and believes it is more important food producers ensure food safety are more than the US consumer to label GM food. willing to purchase GM food. The Italian Most Americans, 71.6%, indicate that they respondents who believe the global food pro- would at least maybe buy GM foods. Only ducers ensure food safety are more willing to 43.1% of Italians will possibly buy GM food. purchase GM food. However, for the Italian The results observed by Wolf et al. (2001) respondents there is no relationship between indicate that the consumers in Ireland are the belief that government agencies ensured slightly more likely to purchase GM food food safety in the past and willingness to pur- than the Italian consumers; half indicate that chase a GM food product. Both the Italian they would at least maybe buy GM foods. and US consumers are more likely to pur- The consumers in the USA observed in the chase a GM food product when they agree research generated by Wolf et al. (2001) that global food producers are using environ- and in this research are more likely to mentally safe methods. consume GM food than their European This research shows that the consumers in counterparts. the USA and Italy have different attitudes Respondents are asked how likely they are toward GM food. An educational programme to purchase GM food to improve nutrition, kill concerning GM foods must be different for pests allowing farmers to use less pesticides, the consumer in the USA and Italy. The con- improve taste and help plants withstand weed- sumers in both countries have different opin- killers. The US respondents are more likely to ions concerning the appropriate sources of purchase GM food products overall and for information for GM food. Further, they have the purposes of improving nutrition, helping different opinions concerning the attributes of plants withstand weedkillers and improving GM food. The consumers in the USA indicate taste. The US and Italian consumers evaluate that improved nutrition is the most important the importance of using biotechnology to kill attribute. The consumers in Italy indicate that pests and allow farmers to use less pesticides reduced pesticides is the most important similarly. There are also similar proportions of attribute and are more likely to increase their consumers that purchased organic food in the purchases of organic foods than the consumer past year in both countries. in the USA. Consumer - Chap 12 5/3/04 15:56 Page 141

Attitudes to GM Food in Italy and the USA 141

References

Bates, B. (2001) Will Frankenfood feed the world? Time Magazine 19 June. Berger, M. (2000) Public health and agricultural biotechnology: a review of the legal, ethical, and scientific controversies presented by genetically altered foods. Published dissertation, Emory University. Bocker, A. and Hanf, C.-H. (2002) Is European consumers’ refusal of GM food a serious obstacle or a tran- sient fashion? In: Santaniello, V., Evenson, R.E. and Zilberman, D. (eds) Market Development for Genetically Modified Food. CAB International, Wallingford, UK. Copple, B. (2000) Scientist, activist, yogi? Forbes 30 October, pp. 54–56. Cray, C. (2000) Biosafety truce reached. Multinational Monitor 21, pp. 6–8. Deutsche Presse-Agentur (2001) Italian Monsanto seed test positive to GM. Deutsche Presse-Agentur, 5 April. Foltz, T. (2002) Organic enter mainstream in San Francisco region. The Packer, 28 January, p. B8. Hayden, T. (2002) Bad seeds in court. U.S. News and World Report, 4 February. Lamb, M. (2000) Brave new food. Mother Earth News, pp. 54–62 179. Moore, S.K. and Scott, A. (1999) Waging a war for public approval. Chemical Week, 15 December, pp. 23–39. Spanier, O. and Tamura (1993) Food Flavor and Safety. American Chemical Society, San Francisco. Stickman, A. (2001) New markets for biotech. Technology Review, July/August. Thayer, A.M. (2002) Ag-biotech industry is gambling on an information campaign, continued farmer acceptance, and promises for the future. Chemical and Engineering News, 2 October, pp. 21–29. The Packer (2001a) Fresh Trends 2001 profile of the fresh produce consumer. The Packer (2001b) EU likely to retain biotech food ban. 31 December, p. B3. The Times (2001) GM tide turns. (London), 3 April. Thomas, G.S. (2001) Playing in San Luis Obispo. Demographics Daily, 6 February. Wielaard, R. (2001) Wary of industry law suits, EU to move this month on biotech. BC News, 8 March. Wolf, M.M., McDonnell, L. and Domegan, C. (2001) A comparison of consumer attitudes toward genetically modified food in Ireland and the USA. Paper presented at the International Conference on Agricultural Biotechnology Research, Ravello, Italy. Wolf, M.M., Johnston, B., Cochran, K. and Hamilton, L. (2002) Consumer attitudes toward organically grown lettuce. Journal of Food Distribution Research Volume 32 (1). Consumer - Chap 12 5/3/04 15:56 Page 142 Consumer - Chap 13 5/3/04 15:56 Page 143

13 Consumer Attitudes Towards GM Food in Ireland and the USA

Marianne McGarry Wolf,1 Juliana McDonnell,2 Christine Domegan2 and Heidi Yount1 1Agribusiness Department, California Polytechnic State University, San Luis Obispo, CA 93407, USA; 2National University of Ireland, Galway, Ireland

Introduction after experimental design was used to conduct this research to eliminate the impact of Biotechnology is being used to produce genet- pre-measurement error. The first phase of ically modified organisms (GMOs) that are this research examines 882 randomly used in food production. Food producers have selected food purchasers interviewed in used the new biotechnologies. Negative con- October 1999 and January 2000. The sec- sumer response to these products in Europe, ond phase of research commenced in Japan and Australia has caused farmers to October 2000 with a random sample of question whether or not to adopt the new 324 respondents. All phases of research technologies. There is concern that American were conducted in San Luis Obispo County, consumer attitudes may follow those of the California, and Galway, Ireland. San Luis Europeans (The Packer, 2001). The Obispo County was designated the best test European Union, Japan and Korea have market in the USA by Demographics Daily labelling laws (Olson, 2000). Senator Barbara (Thomas, 2001). San Luis Obispo was Boxer has introduced a GMO food-labelling found to be the best of 3141 counties to law for the USA (Gregerson, 2000). represent a microcosm of the USA based on The objective of this research is to use a 33 statistical indicators. case study to compare consumer attitudes towards genetically modified food in the USA and Europe using two communities and two Attitudes towards Science and Food time periods. The results of the first phase Purchasing Behavior were presented in Ravello, Italy, during August 2000 at the 4th International Conference on During both phases of the research US the ‘Economics of Agricultural Biotechnology’. respondents appear to have more positive attitudes toward science (Table 13.1). During the first phase US consumers were more Research Methodology likely to have purchased organic food in the past year than Irish respondents (Tables 13.2). The research uses a survey instrument that However, during the second phase organic was administered through the use of a per- purchasing increased and is similar in both sonal interview. A simulated before and countries.

144 M. McGarry Wolf et al.

Table 13.1. Scientiﬁc research is an important factor in the improvement of the quality of life. Ireland USA Chi squared

Phase 1 n = 197 n = 682 Strongly agree 39.6% 48.5% Agree 54.8% 43.3% Disagree 4.6% 6.0% Strongly disagree 1.0% 2.2% 8.754** Phase 2 n = 100 n = 224 Strongly agree 32.00% 45.50% Agree 60.00% 45.50% Disagree 4.00% 7.10% Strongly disagree 4.00% 1.80% 8.435**

**Signiﬁcant difference at 0.05 level.

Table 13.2. Organic food consumption within past year. Ireland USA Chi squared

Phase 1 n = 196 n = 678 Yes 52.6% 62.9% No 47.4% 30.8% 18.579** Phase 2 n = 100 n = 224 Yes 70.10% 72.30% No 29.90% 27.70% 6.945

Food labelling appears to be more impor- Familiarity with GM Foods tant to US respondents than Irish respondents when purchasing food since The results of the ﬁrst phase indicated that nutritional labels and ingredient information there was a similar level of familiarity with GM are read more often by US respondents dur- food in Ireland and the USA (Table 13.5). ing both phases of the research (Tables 13.3 Approximately 43% of respondents in both and 13.4). countries indicated that they were familiar with

Table 13.3. Nutritional label readership and purchase decision. Ireland USA Chi squared

Phase 1 n = 197 n = 683 Very often 27.9% 47.7% Somewhat often 28.4% 30.9% Not very often 28.9% 14.8% Not at all 14.7% 6.9% 41.755** Phase 2 n = 97 n = 224 Very often 18.00% 46.90% Somewhat often 27.00% 33.90% Not very often 37.00% 14.30% Not al all 18.00% 4.90% 46.211**

**Signiﬁcant difference at 0.05 level. Consumer - Chap 13 5/3/04 15:56 Page 145

Attitudes to GM Food in Ireland and the USA 145

Table 13.4. Ingredient label readership and purchase decision. Ireland USA Chi squared

Phase 1 n = 195 n = 681 Very often 22.6% 38.3% Somewhat often 31.8% 30.8% Not very often 30.3% 21.3% Not at all 15.4% 9.5% 20.860** Phase 2 n = 99 n = 224 Very often 16.00% 38.80% Somewhat often 27.00% 33.00% Not very often 42.00% 21.00% Not at all 14.00% 7.10% 29.021**

**Signiﬁcant difference at 0.05 level.

Table 13.5. Familiarity with genetically modiﬁed food. Ireland USA Chi squared

Phase 1 n = 196 n = 681 Very familiar 5.1% 7.0% Somewhat familiar 38.3% 36.6% Not very familiar 39.3% 39.4% Not at all familiar 16.8% 17.0% 1.076 Phase 2 n = 98 n = 224 Very familiar 6.10% 12.90% Somewhat familiar 34.30% 37.50% Not very familiar 45.50% 33.00% Not at all familiar 13.10% 16.50% 8.837*

*Signiﬁcant at 0.10 level

GM food. This level of familiarity is similar to varieties of tacos, tortillas, tostadas and chips that observed by the Wirthlin Group Quorum made by Mission Foods on 13 October 2000 Surveys conducted for the International Food (The Associated Press and Reuters, 2000). Council in the USA in May 2000. The An examination of the sources of informa- Wirthlin Group found that 43% of adults were tion about GM food among respondents who aware that there were foods produced through indicated they were at least somewhat familiar biotechnology in the supermarket now. An shows that the Irish respondents have learned increase in familiarity was observed in the about GM food from a wide variety of sources USA during the second phase of this research, (Table 13.6). Almost all of the familiar Irish with half of the respondents indicating famil- respondents have heard about GM food from iarity with GM food. However, the level of the newspaper or television news in both familiarity in Ireland remained at a level similar phases; while only two-thirds of familiar US to that observed during the previous phase. respondents had heard about GM food from the It is possible that the consumers in the newspaper or television news. Slightly over one- USA became more familiar during the second quarter of familiar US respondents had heard phase due to the media coverage of the recall about GM food from the radio in the first of 2.5 million boxes of Taco Bell brand taco phase. Radio increased as a source of informa- shells produced by Kraft Foods that contained tion in the USA to over a third of respondents. StarLink™ maize in September 2000 Over four-fifths of the familiar Irish respondents (Copple, 2000). The recall of the taco shells had heard about GM food from radio news dur- was followed by a recall of approximately 300 ing the first phase and two-thirds during the Consumer - Chap 13 5/3/04 15:56 Page 146

146 M. McGarry Wolf et al.

Table 13.6. Sources of genetically modiﬁed food awareness among very or somewhat familiar. Phase 1 Phase 2 Ireland USA Chi Ireland USA Chi (n = 86) (n = 297) squared (n = 40) (n = 113) squared

Newspaper 97.6% 68.3% 28.93** 90.0% 70.8% 5.94** Television news 97.6% 64.2% 34.88** 95.0% 61.9% 15.55** Radio news 86.3% 27.9% 90.82** 62.5% 36.3% 8.28** Discussion with family, friends, etc. 71.8% 36.7% 30.43** 37.5% 38.9% 0.026 News magazines 52.2% 40.4% 3.63** 35.0% 45.1% 1.24 Employment, work in farming or 27.6% 16% 4.35** 15.0% 15.9% 0.02 food processing Consumer Reports magazine 25.4% 20.5% 10.40** 15.0% 20.4% 0.03 Internet 22.8% 15.1% 2.04 20.0% 26.5% 0.68 Other 0% 11.3% 6.01** 5.0% 15.0% 2.74*

*Signiﬁcant difference at 0.10 level; **signiﬁcant difference at 0.05 level.

second phase. It is clear that GM food was an Attitudes Towards Government Agencies, important issue to the familiar Irish respondent Producers and Technology during the ﬁrst phase because almost three- quarters indicated that they engaged in discus- Respondents were asked how strongly they sions with family, friends, and colleagues. agree or disagree with the following state- However, only one-third of familiar US respon- ments: ‘government agencies in my country dents indicated they have engaged in discus- have done a very good job at ensuring food sions with family, friends and colleagues about safety in the past’; ‘I trust government agen- GM food. During the second phase fewer Irish cies in my country to ensure food safety in the respondents indicated discussions. A similar future’; ‘global food producers have done a proportion of familiar US and Irish respondents very good job at ensuring food safety in the indicated that they became aware of GM food past’; ‘global food producers are producing from discussions with family, friends and col- food using environmentally safe methods’; leagues. News magazines became a more and ‘computer technology is an important fac- important source of information for US respon- tor in the improvement of the quality of life’. dents and a less important source of information The following rating scale was used to evalu- for Irish respondents. Leading magazines in the ate these statements: strongly agree = 4; US such as Time discussed the use of biotechnology in food production. Time’s article, agree = 3; disagree = 2; strongly disagree = 1 ‘Grains of hope’ was a high-proﬁle front-page (Table 13.8). article that discussed the biotechnology debate. Respondents in the USA are more likely The second phase asked respondents to than the Irish respondents to agree that: ‘gov- indicate appropriate sources of information ernment agencies in my country have done a concerning GM food. The Irish indicated more very good job at ensuring food safety in the sources were appropriate to provide informa- past’; they trust ‘government agencies in my tion concerning GM food (Table 13.7). The country to ensure food safety in the future’; top two sources that respondents in both and ‘global food producers have done a very countries thought appropriate were TV news good job at ensuring food safety in the past’. and newspaper reporters. Less than half of the However, respondents in both countries are respondents from both countries indicated that more likely to disagree that ‘global food pro- experts such as science teachers, representa- ducers are producing food using environmen- tives from food processors or university profes- tally safe methods’. Computer technology is sors were appropriate sources of information perceived to improve the quality of life in both concerning GM food. the USA and Ireland. Consumer - Chap 13 5/3/04 15:56 Page 147

Attitudes to GM Food in Ireland and the USA 147

Table 13.7. Appropriate sources of information concerning genetically modiﬁed food. Ireland USA (n = 93) (n =223) Chi squared

TV news reporters 88.20% 55.60% 30.673** Newspaper reporters 76.30% 53.80% 13.936** Radio news reporters 69.90% 36.30% 25.040** Local government agencies 55.60% 36.30% 8.322** Educational seminars 55.20% 34.40% 8.404** Internet websites 48.30% 30.50% 6.680** Science teachers 42.60% 28.60% 3.978** Representatives from food processors 39.70% 25.00% 4.907** University professors 38.00% 34.50% 0.216 Discussion with friends, family, etc. 37.10% 30.80% 0.881 Farmers 33.30% 41.10% 1.04 Local politicians 24.60% 12.50% 5.2 Visits to food production facilities 24.40% 17.40% 1.225 E-mail 20.80% 13.80% 1.51 Reports from a seed producer 15.90% 16.10% 0.001 Other 7.50% 9.90% 0.221

**Signiﬁcant difference at 0.05 level.

Table 13.8. Mean ratings. Ireland USA Statement (n = 100) (n = 223) t-Test

Government agencies in my country have done a very 2.37 2.77 4.13** good job at ensuring food safety in the past I trust government agencies in my country to ensure food 2.42 2.68 2.69** safety in the future Global food producers have done a very good job at ensuring 2.15 2.38 2.60** food safety in the past Global food producers are producing food using environmentally 2.13 2.27 1.62 safe methods Computer technology is an important factor in the improvement 3.02 3.04 0.17 of the quality of life

**Signiﬁcant difference at 0.05 level.

Attitudes Towards GM Food of the Irish respondents indicated that mandatory labelling of GM food was very important, Most consumers in both countries indicated that while approximately one-half of US respondents government imposition of mandatory labelling is indicated that mandatory labelling of GM food important, 95% in Ireland and 81% in the USA was very important. During the second phase during the first phase (Table 13.9). Although the US consumers felt mandatory labelling was as Irish respondents were less likely than the US important as the Irish consumers. respondents to read nutritional labels or ingredi- The Irish consumer is less likely to purchase ent labels when making a food purchase deci- food that has been genetically modified than the sion, they were more likely than the US US consumer (Table 13.10). Almost one-half of respondents to indicate that government imposi- the Irish respondents indicated they were likely tion of mandatory labelling of GM food is impor- to purchase food that has been genetically mod- tant during the first phase. Almost three-quarters ified, more than two-thirds of the US consumers Consumer - Chap 13 5/3/04 15:56 Page 148

148 M. McGarry Wolf et al.

Table 13.9. Government imposition of mandatory labelling for genetically modiﬁed food. Ireland USA Chi squared

Phase 1 n = 197 n = 681 Very important 70.6% 52.3% Somewhat important 25.4% 28.9% Not very important 3.6% 12.2% Not at all important 0.5% 6.6% 31.709** Phase 2 n = 100 n = 223 Very important 68.0% 65.0% Somewhat important 27.0% 23.3% Not very important 4.0% 6.7% Not at all important 1.0% 4.9% 4.222

**Signiﬁcant difference at 0.05 level.

Table 13.10. Likelihood to buy genetically modiﬁed food. Ireland USA Chi squared

Phase 1 n = 197 n = 679 Definitely, probably, maybe 47.7% 72.8% Probably not, definitely not 27.0% 23.3% 43.38** Phase 2 n = 100 n = 223 Definitely, probably, maybe 49.5% 62.5% Probably not, definitely not 50.5% 37.5% 4.78**

**Signiﬁcant difference at 0.05 level.

indicated they were likely to purchase food that food to improve nutrition, kill pests allowing has been genetically modified. The proportion farmers to use less pesticides, improve taste, of Irish consumers that indicated a positive likeli- and help plants withstand weedkillers. The US hood to purchase remained stable during the respondents were more likely to purchase GM two periods, however, the proportion of US food products overall and for the purposes of respondents that were positive declined improving nutrition and improving taste (Table between the two phases. The findings of this 13.11), while the US and Irish consumers research are similar to the results reported by evaluated the importance of using biotechnol- The Times in April 2001. A survey in the UK ogy to kill pests and allow farmers to use less indicated that 48% of respondents would eat pesticides and help plants withstand weed- GM food (The Times, 2001). The results for killers similarly. The rating scale used to evalu- the US are similar to the findings of The ate purchase interest is: 5 = definitely; 4 = Packer’s survey of 1000 US consumers (The probably; 3 = maybe; 2 = probably not; and Packer, 2001). The Packer’s survey found that 1 = definitely not. 60% of consumers were extremely, very or A comparison of the importance of specific somewhat likely to purchase a fresh produce uses of biotechnology is given in Table 13.12 item that has been genetically modified. for each country. The attributes are listed from high to low based on their means and paired to examine differences between the means for Attitudes Towards Different Attributes of each country. For the Irish respondents using GM Food biotechnology to kill pests and allow farmers to use less pesticides and improve nutrition are Respondents during the second phase were the top reasons. Nutrition is the most impor- asked how likely they were to purchase GM tant reason to use biotechnology followed by Consumer - Chap 13 5/3/04 15:56 Page 149

Attitudes to GM Food in Ireland and the USA 149

Table 13.11. Likelihood to buy – attribute mean rating. Ireland USA (n = 100) (n = 223) t-Test

How likely are you to purchase a food product that has been genetically modiﬁed? 2.53 2.80 2.10** To improve nutrition? 3.03 3.30 1.82* To kill pests and allow farmers to use less pesticides? 3.05 3.10 0.33 To improve taste? 2.56 2.91 2.50** To help plants withstand weedkillers? 2.64 2.76 0.94

*Signiﬁcant difference at 0.10 level; **signiﬁcant difference at 0.05 level.

Table 13.12. Likelihood to buy – attribute mean rating. Mean rating Paired t

Ireland (n = 98) To kill pests and allow farmers to use less pesticides? 3.05 To improve nutrition? 3.03 0.29 To help plants withstand weedkillers? 2.64 3.58** To improve taste? 2.83 0.46 USA (n = 223) To improve nutrition? 3.36 To kill pests and allow farmers to use less pesticides? 3.17 3.57** To improve taste? 3.02 2.89** To help plants withstand weedkillers? 2.83 2.47**

**Signiﬁcant difference at 0.05 level.

the use of less pesticides for the US consumers. ensuring food safety and willingness to pur- Improving taste and helping plants withstand chase a GM food product. Most Irish respon- weedkillers were beneﬁts that were rated lower. dents evaluated the government and global The Packer (2001) also found that consumers food producers lower than the US respon- evaluated improved nutritional content as a dents. Both the Irish and US consumers are more appropriate reason to genetically modify more likely to purchase a GM food product food items than improving ﬂavour. when they agree that global food producers are using environmentally safe methods (Tables 13.13–13.21). Attitudes Towards Government Agencies and Producers, and Willingness to Purchase GM Foods Conclusions

The responses to the questions concerning The objective of this research is to use a case trust in government and global food producers study to compare consumer attitudes toward to ensure food safety were then compared GM food in the USA and Europe using two with the willingness of a respondent to pur- communities and two time periods. The chase GM food. The results show that for the research used a survey instrument that was US respondents, those who trust the govern- administered through the use of a personal ment and global food producers to ensure interview. The ﬁrst phase of this research food safety are more willing to purchase GM examines 882 randomly selected food pur- food. However, for the Irish respondents chasers interviewed in October 1999 and there is no relationship between trust in gov- January 2000. The second phase of research ernment agencies and global food producers commenced in October 2000 with a random Consumer - Chap 13 5/3/04 15:56 Page 150

150 M. McGarry Wolf et al.

Table 13.13. Mean willingness to purchase a genetically modiﬁed food product: ‘Government agencies in my country have done a very good job at ensuring food safety in the past’. Mean F statistic

USA (n = 222) Strongly disagree 2.06 16.09** Disagree 2.18 Agree 2.97 Strongly agree 3.53 Ireland (n = 99) Strongly disagree 2.42 2.15 Disagree 2.45 Agree 2.67 Strongly agree 2.00

**Signiﬁcant difference at 0.05 level.

Table 13.14. USA (n = 222): ‘Government agencies in my country have done a very good job at ensuring food safety in the past’. Mean difference

Strongly disagree Disagree 0.1244 Agree 0.9111** Strongly agree 1.4739** Disagree Strongly disagree 0.1244 Agree 0.7867** Strongly agree 1.3494** Agree Strongly disagree 0.9111** Disagree 0.7867** Strongly agree 0.5627** Strongly agree Strongly disagree 1.4739** Disagree 1.3494** Agree 0.5627**

**Signiﬁcant difference at the 0.05 level.

Table 13.15. Mean willingness to purchase a genetically modiﬁed food product: ‘I trust government agencies in my country to ensure food safety in the future’.

Mean F statistic

USA (n = 222) Strongly disagree 2.21 37.11** Disagree 2.29 Agree 2.98 Strongly agree 3.48 Ireland (n = 98) Strongly disagree 2.08 5.77 Disagree 2.78 Agree 2.50 Strongly agree 2.00

**Signiﬁcant difference at the 0.05 level. Consumer - Chap 13 5/3/04 15:56 Page 151

Attitudes to GM Food in Ireland and the USA 151

Table 13.16. USA (n = 222): ‘I trust government agencies in my country to ensure food safety in the future’. Willingness to purchase a GM food product Mean difference

Strongly disagree Disagree 0.007 Agree 0.7578** Strongly agree 1.2641** Disagree Strongly disagree 0.007 Agree 0.6811** Strongly agree 1.874** Agree Strongly disagree 0.7578** Disagree 0.6811** Strongly agree 0.5063 Strongly agree Strongly disagree 1.2641** Disagree 1.874** Agree 0. 5063

**Signiﬁcant difference at the 0.05 level.

Table 13.17. Mean willingness to purchase a genetically modiﬁed food product: ‘Global food producers have done a very good job at ensuring food safety in the past’. Mean F statistic USA (n = 222) Strongly disagree 2.25 31.31** Disagree 2.67 Agree 3.14 Strongly agree 4.14 Ireland (n = 99) Strongly disagree 2.11 5.88 Disagree 2.64 Agree 2.55 Strongly agree 3.50

**Signiﬁcant difference at the 0.05 level.

Table 13.18. USA (n = 222): ‘Global food producers have done a very good job at ensuring food safety in the past’. Mean difference

Strongly disagree Disagree 0.4167 Agree 0.8851** Strongly agree 1.8929** Disagree Strongly disagree 0.4167 Agree 0.4685** Strongly agree 1.4762** Agree Strongly disagree 0.8851** Disagree 0.4685** Strongly agree 1.0078 Strongly agree Strongly disagree 1.8929** Disagree 1.4762** Agree 1.0078

**Signiﬁcant difference at the 0.05 level. Consumer - Chap 13 5/3/04 15:56 Page 152

152 M. McGarry Wolf et al.

Table 13.19. Mean willingness to purchase a genetically modiﬁed food product: ‘Global food producers are producing food using environmentally safe methods’. Mean F statistic USA (n = 222) Strongly disagree 2.59 30.04** Disagree 2.72 Agree 3.10 Strongly agree 3.75 Ireland (n = 99) Strongly disagree 2.00 7.68* Disagree 2.54 Agree 2.81 Strongly agree 3.00

*Signiﬁcant difference at the 0.10 level; **signiﬁcant difference at the 0.05 level.

Table 13.20. USA (n = 222): ‘Global food producers are producing food using environmentally safe methods’. Mean difference

Strongly disagree Disagree 0.1300 Agree 0.5060** Strongly agree 1.160** Disagree Strongly disagree 0.1300 Agree 0.3760** Strongly agree 1.031** Agree Strongly disagree 0.5060** Disagree 0.3760** Strongly agree 0.6538 Strongly agree Strongly disagree 1.156** Disagree 1.031** Agree 0.6538

**Signiﬁcant difference at the 0.05 level.

Table 13.21. Ireland (n = 99): ‘Global food producers are producing food using environmentally safe methods’. Mean difference

Strongly disagree Disagree 0.5385 Agree 0.8148** Strongly agree 1.000 Disagree Strongly disagree 0.5385 Agree 0.2764 Strongly agree 0.4615 Agree Strongly disagree 0.8148** Disagree 0.2764 Strongly agree 4615 Strongly agree Strongly disagree 1.0000 Disagree 0.4615 Agree 0.1852

**Signiﬁcant difference at the 0.05 level. Consumer - Chap 13 5/3/04 15:56 Page 153

Attitudes to GM Food in Ireland and the USA 153

sample of 324 respondents. All phases of kill pests and allow farmers to use less pesticide, research were conducted in San Luis Obispo, and to improve taste. Respondents in the USA California, and Galway, Ireland. San Luis indicated that the most important reason to pur- Obispo has a population of approximately chase a GM food product is to improve nutri- 42,000 and Galway has a population of tion followed by modifying it to kill pests and approximately 57,000. allow farmers to use less pesticides. Genetically The results of the ﬁrst phase indicated that modifying food products to improve taste and there was a similar level of familiarity with GM help plants withstand weedkillers were less food in Ireland and the USA. Approximately important criteria when purchasing a GM food 43% of respondents in both countries indi- product. The Irish respondents indicated that cated that they were familiar with GM food. using biotechnology to kill pests and allow farm- The second phase of research was conducted ers to use less pesticides and to improve nu- shortly after the recall of many food products trition are equally important. Using in the USA containing maize that was grown biotechnology to help plants withstand weed- using the StarLink™ seed. An increase in killers and improve taste were less important familiarity was observed in the USA, with half to the Irish respondent. of the respondents indicating familiarity with Questions were added during the second GM food. However, the level of familiarity in phase of research to examine attitudes toward Ireland remained at a level similar to that government agencies and food safety and atti- observed during the previous phase. tudes toward food producers and food safety Results of the ﬁrst phase indicated a differ- and the environment. Perceptions of govern- ence in attitudes between the Irish consumer ment agencies and food safety and perceptions and consumers in the USA toward GM food. of food producers and food safety and the envi- The familiar US respondents perceived GM ronment are more positive in the USA than in food to have neutral or positive attributes. The Ireland. Perceptions of government agencies Irish consumer attributed more negative attrib- and food safety and perceptions of food pro- utes to GM food. Further, the Irish consumers ducers and food safety are related to a con- were more likely to indicate that mandatory sumer’s attitudes toward GM food in the USA. labelling is important and less likely to purchase However, such attitudes are not related to a a GM food product. During the second phase consumer’s attitudes toward GM food in of research, the Irish consumer continued to be Ireland. Consumers in the USA that have more less likely to purchase a GM food product. positive attitudes toward government agencies However, the Irish and consumers in the USA and food safety and of food producers and indicated similar attitudes toward the labelling food safety are more likely to purchase GM of GM food during the second phase of food. Such a relationship does not exist in research since more respondents in the USA Ireland. However, there are less positive atti- indicate that mandatory labelling is important. tudes toward government regulators and global Additional questions were added during the food producers in Ireland. It was found that a second phase of research to examine consumer positive relationship exists between perceptions attitudes toward GM food based on the purpose of global food producers using environmentally for the use of biotechnology: to help plants safe methods and the purchase probability for withstand weedkillers, to improve nutrition, to GM food for consumers in both countries.

References

Alexander, N. and Toner, C. (2001) More consumers see potential benefits to food biotechnology. International Food Information Council. Available at www.ific.org/procative/newsroom/ release.vtml?id=19241. Central Statistics Office Home Page, Principal Statistics (1999) Available at www.cso.ie/principalstat/pris- tat2.html. Copple, B. (2000) Scientist, activist, yogi? Forbes, 30 October, pp. 54–56. Consumer - Chap 13 8/3/04 10:33 Page 154

154 M. McGarry Wolf et al.

Gregerson, J. (2000) What do we know about GMO’s? Food Engineering, March. Hoban, T. (2000) U.S. food consumption is largely unaffected by StarLink corn recall. News Services, 27 November. Nash, J.M. (2000) Grains of hope. Time 156 (5). Olson, J. (2000) GMO free zone. Farm Industry News, February. The Associated Press and Reuters (2000) Scope of biotech corn product recall revealed. CNN.com, 2 November. The Packer (2001) Fresh Trends 2001 proﬁle of the fresh produce consumer. The Times (London) (2001) GM tide turns. 3 April. Thomas, G.S. (2001) Playing in San Luis Obispo. Demographics Daily, 6 February. Available at wysi- wyg://44/http://bizjournals.bcentral.com/journals/demographics/. Turcsik, R. (2001) Still life in biotech. Progressive Grocer, April, pp. 16–22. US Bureau of the Census (1991) State and Metropolitan Data. Consumer - Chap 14 5/3/04 15:56 Page 155

14 Attitudes Towards GM Food in Colombia1

Douglas Pachico1 and Marianne McGarry Wolf2 1International Center for Tropical Agriculture (CIAT), AA 6713, Cali, Colombia; 2Agribusiness Department, California Polytechnic State University, San Luis Obispo, CA 93407, USA

Transgenic or genetically modified crops are the potential risks of GM foods than do US widely grown, covering over 50 million hectares consumers. Little if any similar research has in 2001 (James, 2002), and transgenic food is examined consumer attitudes towards GM food widely consumed, entering an estimated 60% of in low-income countries where hunger and processed foods in the USA (Hopkins, 2001). malnutrition are most common and where, Some see genetically modified (GM) crops as therefore, GM crops might have their greatest critical to improving agricultural productivity and contribution to the welfare of consumers. ensuring food supplies especially for the poor This chapter makes an initial examination and malnourished in developing countries of consumer attitudes in Colombia toward GM (Evans, 1998; Oxfam Policy Department, food. It follows on from research previously 1999). Others argue that this is a myth and that conducted in the USA and Ireland. First, the there are significant health and environmental methods of the study are briefly described and risks from GM crops and food (Altieri, 2001). In some characteristics of the sample population this context, consumer attitudes towards GM noted. Second, some general background atti- foods have become a factor both in the market tudes of the Colombian consumer with demand for GM foods, and in their regulation. respect to food safety, science and govern- In some markets there is no doubt that con- ment regulation are reviewed. Third, levels sumer attitudes have slowed the utilization of and sources of consumer knowledge about GM crops (Charles, 2001). GMO food are presented. Fourth, attitudes Because consumer attitudes have become towards GMO foods, including likelihood of such a key factor in the acceptance of GM purchase, are analysed. Finally, the major food, and because these attitudes seem to vary implications of the study are reviewed and so substantially between countries, increasing some areas for further research are noted. attention has been paid to understanding consumer attitudes towards GM food (Bredahl et al., 1998; Sheehy et al., 1998). Particular Methods and Data attention has been paid to understanding differences in consumer attitudes between the USA This study largely followed the approach and and Europe (Nelson, 2001; Wolf et al., 2001; utilized a modified questionnaire that had Wolf and Domegan, 2002). In general, been previously used in Ireland and the USA European consumers have a stronger sense of (Wolf et al., 2001; Wolf and Domegan,

1 Research assistance provided by Katie Canada, California Polytechnic State University, California, USA. © CAB International 2004. Consumer Acceptance of Genetically Modiﬁed Food (eds R.E. Evenson and V. Santaniello) 155 Consumer - Chap 14 5/3/04 15:56 Page 156

156 D. Pachico and M. McGarry Wolf

2002). This both facilitates international com- Attitudes to Food Safety and Science parisons and provides a research instrument that has been validated in previous studies. There is a high level of awareness among Cali The questionnaire is largely comprised of consumers of possible food risks (Table 14.1). questions scaled around different degrees of There is strong agreement among 63.3% that frequency or different degrees of agreement. pesticides in food are dangerous and among A Spanish translation was developed and pre- 64% that food additives are dangerous. A tested to ensure understanding. A few addi- majority of 52.7% strongly agrees that foods tional questions on attitudes were added. are adulterated with false ingredients. This study was conducted in Cali, However, even though (or perhaps because) Colombia, an urban centre with a metropoli- foot and mouth disease is endemic among tan population of around 2.5 million. Cali can cattle in some regions of Colombia, 65.3% of reasonably be considered as a typical South the sample disagrees or strongly disagrees American city in a region where over 70% of that foot and mouth disease is a food risk for the total population is urban. As in the Irish humans. Clearly the sample exercises discrim- and US studies, the questionnaire was ran- ination about what it considers to be real food domly applied to people approaching or risks. It is sensitive to some potential food departing from points of food purchase. These risks, like pesticide residues or food additives, included supermarkets and open air markets in but it is prepared to discount other factors, six different neighbourhoods selected accord- like foot and mouth disease, that could have ing to general indicators of economic status. A been perceived as a food risk. total of 150 questionnaires were conducted Overall, the Cali food purchasers have a among food purchasers in March 2001 by a positive view of science and technology. single experienced sociologist. There is strong agreement among 68% of the Females comprised 89.3% of respondents, sample that science improves the quality of 51.7% were between 25 and 44 years old, life while 56% strongly agree that computers 68% were married or lived with partners and improve the quality of life. Thus, the sample 69% were members of dual-income households. would not appear to have a prior predisposi- The respondents were almost evenly divided tion to be sceptical of scientific innovations among those who work full time (34.7%), those such as GM food but rather might even be employed part time (34.0%) and those not predisposed to associate new scientific discov- employed (31.3%). The sample was compara- eries with something positive. tively well educated with 33.5% having attended There is also a fairly high level of confi- university. Some 71.8% of respondents have dence in the government assuring food safety. children under the age of 18 living at home. While 75.3% agree or strongly agree that the

Table 14.1. Consumer attitudes to food, safety, science, government and food producers, Cali, Colombia, 2001 (n = 150). Strongly Strongly agree Agree Disagree disagree Not sure

Pesticides dangerous to health 63.3 28.0 6.7 2.0 0.0 Foot and mouth disease a food risk 24.0 10.7 53.3 12.0 0.0 Food additives dangerous to health 64.0 29.3 6.0 0.7 0.0 False ingredients put in food 52.7 37.3 3.3 2.0 4.7 Science improves quality of life 68.0 26.0 3.3 2.0 0.7 Computers improve quality of life 56.0 30.7 10.7 2.7 0.0 Government assures food safety 38.0 37.3 15.3 6.7 2.7 Global producers assure food safety 49.3 24.7 17.3 4.0 4.7 Global food producers environmentally safe 30.9 27.5 30.2 2.0 9.4 Household food supply adequate 24.8 36.9 37.6 0.7 0.0 Price most important in choosing food 31.3 24.0 41.3 3.3 0.0 Consumer - Chap 14 5/3/04 15:56 Page 157

Attitudes towards GM Food in Colombia 157

government assures food safety, only 22.0% in their examination of familiarity with GM disagree or strongly disagree. Similarly, food in the USA and Ireland. Approximately 74.0% agree or strongly agree that food pro- half of the respondents in the USA and 40% ducers assure food safety, but only 21.3% dis- of respondents in Ireland were at least some- agree. There is less confidence in the what familiar with GM food. environmental safety of food production, with Television news has been the main source 32.2% disagreeing or strongly disagreeing of information in Colombia, reaching 10% of that food production is environmentally safe. the sample, while 6% had discussed it with One major distinguishing characteristic of acquaintances, 5.3% had heard about it over this sample is that nearly two-fifths, 38.3%, the radio, 4.0% had read about it in news- disagree or strongly disagree that there is papers and 1.4% had read about it in maga- always enough food to eat in their family. zines. Given the very low levels of familiarity, Likewise, a majority, 55.3%, agree or obviously many of those with some familiarity strongly agree that price is the most impor- had had access to information about trans- tant factor in purchasing food. These two atti- genic foods from more than one source. tudes would support the hypothesis that for Given the low levels of familiarity and many people in low-income countries, they access to information about GM food, it is are unable to afford the quantity of food that probable that many of the attitudes towards they desire. This being the case, Colombian GM food reported in this chapter may not be consumers may be less sensitive to potential strongly held. The attitudes of the Colombian but as yet unidentified food risks than con- consumer could be subject to significant sumers in high-income countries like those in change in the light of additional exposure to Europe where food is abundantly available information in the future. and quality issues come to the fore. These data suggest that in contrast to high-income countries, inadequate availability of food may Attitudes to GM Food be the most pressing food-related health issue for many people, thus decreasing the likeli- Attitudes towards GM foods among hood of resistance to transgenic food. Colombian consumers are mixed. There is widespread agreement that some GM foods may be unsafe, but none the less only a Knowledge of GM Food minority would be unwilling to buy GM food. Nearly three-quarters of Colombian con- There is a very low level of familiarity with sumers strongly agree or agree that some GM food in Colombia. The vast majority of foods produced by genetic engineering may the sample, 77.6%, reports that they are not be unsafe (Table 14.2). Less than one-quarter at all familiar with GM food. Only 5.4% indi- would disagree. On the other hand, cate they are very familiar with transgenic Colombian consumers are split quite evenly food and 7.5% say they are somewhat famil- into three groups in terms of their willingness iar. These awareness levels are significantly to purchase GM food. Some 33.6% would lower than those found by Wolf et al. (2001) definitely or probably buy GM food, 32.9%

Table 14.2. Attitudes to GMO food, Cali, Colombia, 2001 (n = 150). Strongly Strongly Not agree Agree Disagree disagree Sure

GMO food unsafe 36.0 39.3 20.7 2.0 0.7 Probably Deﬁnitely Deﬁnitely Probably Maybe not not

Willingness to buy GMO food 15.1 18.5 32.9 18.5 15.1 Consumer - Chap 14 5/3/04 15:56 Page 158

158 D. Pachico and M. McGarry Wolf

might buy it and 33.6% would probably or that they appreciate (Table 14.3). For exam- definitely not buy GM food. The Colombian ple, consumers indicated that pesticides are probability of purchasing GM food is similar dangerous to their health. The use of genetic to that observed by Wolf et al. (2001) in the modification to reduce the use of pesticides USA. However, it is higher than that observed generated the highest purchase interest. for Ireland where only 17.2% would definitely Further, willingness to buy is significantly or probably buy GM food. higher for characteristics that would be Thus, nearly three-quarters of consumers desired by consumers like improved nutrition perceive potential risks with GM food, but or taste than for a characteristic like resistance two-thirds would be willing to purchase GM to weedkillers that does not directly benefit foods. There are some possible explanations consumers. Consistent with the finding for this apparent inconsistency. In the first reported above that Colombian consumers place, the widespread belief that some GM are aware of the risks of pesticide residue on foods may be unsafe does not preclude the food, this genetically modified pest resistance simultaneous belief that some GM foods may which would reduce the use of chemical pesti- be safe. Given the previously reported high cides also gives a higher willingness to buy. level of reported confidence in the food regu- Further research might attempt to elucidate latory system, consumers may simply trust whether a lower cost of GM food would simi- that some GM foods are safe, and those that larly elicit a greater willingness to purchase. are not would be excluded from the food sup- Attitudes of Colombian consumers to ply by the regulatory authorities. Another labelling may also have some relevance to explanation could be that consumers might be these questions. A majority of Colombian willing to absorb the risk of GM food if it met consumers read food ingredient labels very or other important criteria for them. somewhat often (Table 14.4). This could indi- The characteristics of the GM food would cate that they rely on the content and even have an influence on consumers’ likelihood to the mere presence of food labels as a warrant purchase GM food. Using a five-point scale of of food safety. Moreover, 68% of consumers willingness to buy (definitely = 5; probably = report that they think that mandatory labelling 4; maybe = 3; probably not = 2; definitely not of GM food is very important and 22.7% = 1), it can be seen that consumers are more think it is somewhat important. It is possible willing to buy GM food if it has characteristics that this implies that with a system of labelling

Table 14.3. Likelihood to buy – attribute mean rating (n = 150). Attribute means Paired t

To reduce the use of pesticides? 3.43 To improve nutrition? 3.39 0.513** For improved taste? 3.16 2.44** To resist weed killers? 2.84 3.14**

**Signiﬁcant difference at 0.05 level.

Table 14.4. Practices and attitudes towards food labelling, Cali, Colombia, 2001 (n = 150). Very Somewhat Not very Not at often often often all Read food labels for ingredients 38.1 26.5 22.4 12.9 Very Somewhat Not very Not at all important important important important Importance of mandatory labelling of GMO food 68.0 22.7 4.7 4.7 Consumer - Chap 14 5/3/04 15:56 Page 159

Attitudes towards GM Food in Colombia 159

most Colombians would be willing to pur- food purchase. Those who are less sensitive chase GM food even though they believe that to price have a lower willingness to buy GM some GM foods might be unsafe. food, 2.40. This relationship of higher willing- Consequently, in order to be able to have the ness to purchase GM food with higher sensi- assurance that they can consume GM foods tivity to food price is consistent across the safely, Colombians think that mandatory categories of opinion with respect to price labelling is very important. and is statistically significant. This would be This is not really contradicted by the fact consistent with the hypothesis that higher- that in practice far fewer consumers, 38%, income people, for whom the cost of food is often read food ingredient labels than think less important, are more influenced by possi- that mandatory labelling of GM foods is very ble food quality characteristics and for this important, 68%. To some extent, the mere reason are less willing to purchase GM food. presence of labels may be a sufficient indica- In contrast, for those consumers for whom tor for many consumers that appropriate food prices are a major criterion in food pur- authorities are monitoring food safety. In addi- chase, they may be more disposed to pur- tion, for items that are consumed regularly, chase GM food as long as it is cheaper. This people may not expect constant changes in suggests that poor consumers could benefit ingredients and therefore do not need to read disproportionately from cheap GM food so the labels of regularly purchased food on a long as it was indeed safe, but on the other frequent basis. hand if it really was not safe, then they could To understand better the attitudes of be more vulnerable to any risks associated Colombian consumers to GM food, Table 14.5 with consuming GM food. shows mean willingness to buy scores for peo- Similarly, among those who strongly dis- ple holding different opinions. Thus, it is clear agree that the quality and variety of food in that among those who more strongly disagree the family is good, that is, among those con- that engineered foods are unsafe, that is sumers in families where the quality and vari- among those who perceive less chance of risk ety of food is less than desired, the willingness from GM foods, willingness to buy is higher, to buy GM food is high, 3.5. In contrast, 4.33, than it is among those who strongly among those families where they agree that agree that GM foods are not safe. Those the quality and variety of food is already good, strongly agreeing that GM foods are unsafe willingness to buy GM food is low, 2.57. have a lower willingness to buy, 2.84. Although this relationship is not statistically Perceptions of the risks of GM food thus have significant for this sample, it does consistently the expected relationship with willingness to suggest that the less adequate the quality of buy. This relationship is consistent across the current food consumption, the more willing opinion categories and is statistically significant. people are to buy GM food. The better the Furthermore, for those who strongly agree current quality of food, the less willing are that low price is important in the food pur- people to buy GM food. This finding is again chase decision, willingness to buy is higher, quite consistent with the previous result on 3.41, than for those who strongly disagree the relationship between food price and will- that price is the main decision criterion for ingness to purchase GM food.

Table 14.5. Mean willingness to buy genetically modified food by food attitude groups (5 = definitely willing; 1 = definitely not willing), Cali, Colombia, 2001 (n = 150). Strongly Strongly disagree Disagree Agree agree F statistic

Genetic engineered foods not safe 4.33 3.35 2.88 2.84 2.491* Low price important to buy food 2.40 2.90 2.75 3.41 2.606** Family food supply adequate 3.00 3.15 2.68 3.19 1.714 Family food quality good 3.50 3.18 3.10 2.57 1.856

*Signiﬁcant difference at 0.01 level; **signiﬁcant difference at 0.05 level. Consumer - Chap 14 5/3/04 15:56 Page 160

160 D. Pachico and M. McGarry Wolf

There is, though, not a clear relationship About three-quarters of Colombian con- between the adequacy of current diets in terms sumers agree that some GM foods may be of quantity and willingness to buy GM food. It unsafe. Nevertheless, some two-thirds of con- would have been hypothesized that consumers sumers would be willing to buy food with GM without an adequate quantity of food would ingredients. This result is similar to that have been more willing to purchase GM food, observed in the USA, but lower than the but there is no evidence for this. probability of purchase for consumers in Ireland. Concerns about safety do affect the willingness to buy GM food. There is a statisti- Summary and Suggestions for Further cally significant relationship between percep- Research tion of genetically engineered food as unsafe and the willingness to buy GM food: the Although transgenic crops are being widely stronger the safety concern, the lower the grown worldwide, consumer attitudes towards willingness to purchase. them have been found to vary substantially Nevertheless, many consumers who per- between Europe and the USA. This study of a ceive some safety risks in GM food would still sample of 150 food purchasers in Cali, be willing to buy it. Economic factors may be Colombia, is believed to be one of the first important in this regard. Those for whom low studies to examine consumer attitudes price is the most important factor in the food towards GM food among consumers in tropi- purchase decision are significantly more willing cal or low-income countries. to buy GM food. Likewise, those for whom the Consumers in Colombia are aware of pos- current quality and variety of food is less than sible food risks, with about two-thirds agreeing desired are also more willing to buy GM food. These findings suggest that for resource-con- that residues of pesticides or food additives strained food consumers, ill-defined or uncer- are dangerous. However, other factors that tain risks would not necessarily be highly could have been perceived as a food risk, like dissuasive of GM food consumption, especially foot and mouth disease, were not considered if it were cheap. Thus, if GM food risks are dangerous by nearly two-thirds of consumers, indeed low or non-existent, then poor con- indicating that Colombian consumers do not sumers would be most likely to reap the bene- simply accept as dangerous any hypothetical fits of GM foods that reduce the price of food. risk factor. Finally, though highly suggestive, these Well over half of consumers in Colombia results must still be taken as a very tentative appear to have positive views of science and picture of the attitudes of Colombian con- technology and a surprisingly high level of sumers to GM food. Familiarity with GM food three-quarters of consumers have confi- is still very low and current attitudes could dence in government regulation of food shift with increased familiarity. safety. Holding these positive attitudes Several further extensions to this initial towards the benefits of science and the research could be considered. It would be useful effectiveness of food safety regulations is to more directly assess whether a lower cost of likely more consistent with less concern GM food would elicit a greater willingness to about the risks of GM food. purchase. It could be useful to more purpo- Economic factors seem to affect the access sively sample among consumers with a higher to food of a significant number of Colombian degree of familiarity with GM food to attempt consumers. Nearly two-fifths sometimes do to project what likely attitudes might be with not have enough food to eat in their family increased familiarity in the future. The survey and for nearly one-half a low price is the most approach could be supplemented with a focus important factor in buying food. For con- group approach to probe more into people’s sumers such as these, for whom the absolute attitudes and to see how additional information quantity of food is a pressing concern, quality might shape these attitudes. Further research is factors such as potential but unidentified food planned to contrast the results of this survey risks from GM foods may not play a major with those of surveys in high-income countries role in food purchase decisions. to compare and contrast the differences. Consumer - Chap 14 5/3/04 15:56 Page 161

Attitudes towards GM Food in Colombia 161

References

Altieri, M. (2001) Genetic Engineering in Agriculture: Myths, Environmental Risks and Alternatives. Food First Books, Oakland, California. Bredahl, L., Grunert, K. and Frewer, L. (1998) Consumer attitudes and decision-making with regard to genetically engineered food products. Journal of Consumer Policy 21, 251–277. Charles, D. (2001) Lords of the Harvest: Biotech, Big Money and the Future of Food. Perseus, New York. Evans, L.T. (1998) Feeding the Ten Billion. Cambridge University Press, Cambridge. Hopkins, K. (2001) The risks on the table. Scientific American 284(4), 60–61. James, C. (2002) Global Hectarage in GM Crops 2001. International Service for the Acquisition of Agri- biotech Applications, Ithaca, New York. Nelson, C.H. (2001) Risk perception, behavior, and consumer response to genetically modified organisms: toward understanding American and European public reaction. American Behavioral Scientist 44, 1371–1388. Oxfam Policy Department (1999) Genetically modified crops, world trade and food security. Available at http://www.oxfam.org.uk/policy/papers/gmcrop.htm. Sheehy, H., Legault, M. and Ireland, D. (1998) Consumer and biotechnology: a synopsis of survey and focus group research. Journal of Consumer Policy 21, 359–386. Wolf, M.M. and Domegan, C. (2002) A comparison of consumer attitudes toward genetically modified food in Europe and the USA: a case study over time. In: Santaniello, V., Evenson, R.E. and Zilberman, D. (eds) Market Development for Genetically Modified Food. CAB International, Wallingford, UK, pp. 25–38. Wolf, M.M., McDonnell, J. and Domegan, C. (2001) A comparison of consumer attitudes toward genetically modified food in Ireland and the USA. Paper presented at the International Conference on Agricultural Biotechnology Research, Ravello, Italy. Consumer - Chap 14 5/3/04 15:56 Page 162 Consumer - Chap 15 5/3/04 15:56 Page 163

15 Consumer Acceptance and Development Perspectives of Functional Food in Germany

Heiko Dustmann and H. Weindlmaier Centre of Life and Food Sciences Weihenstephan Professorship for Dairy and Food Industry Management, Technische Universität, München, Germany

Functional Food Market Facts functional food market. Functional foods are considerably more spread with a turnover of The association for consumer research (GfK) €3.5 billion in Japan and €11 billion in the estimates that the turnover of functional food USA (at the end of the 1990s) (Menrad et al., in the food retail trade amounts to €920 mil- 2000, pp. 154–157). lion (2000) in Germany.1 This is proportion- Nationally as well as internationally probi- ally slightly less than 1% of the total turnover otic dairy products and functional drinks in the food retail trade. According to a press together make up the biggest part of the func- release of 29 November 2001, A.C. Nielsen tional food market with more than 50% (Fig. assigns 1.5% of the total food market to the 15.1) (A.C. Nielsen GmbH, 2001, p. 1).

Fruit drinks 16% Others incl. sweets 29% Fruit juices 6%

Isotonic drinks 5%

Butter/margarine 4%

Yoghurt 19% Cereal products 14% Cream-cheese 7% Fig. 15.1. Product groups of functional food in Germany in 2001. Source: Based on A.C. Nielsen GmbH (2001, appendix).

1 See GfK Panel Services Consumer Research GmbH, Birnbaum quoted in Soßna (2001, p. 22).

164 H. Dustmann and H. Weindlmaier

Consumer Trends definite consumer habit resulting from the trends plays the decisive part. A written survey In order to classify development of potential as well as ten nationwide group discussions consumer acceptance for functional food in (with 102 participants in total) were carried out Germany, consumer trends in the total system in 2001 by the Institute for the Management of of food demand and supply were analysed as Dairy Companies to investigate the potential the basis for the further steps of the project acceptance of functional food. (Fig. 15.2). The trends towards health and con- The members of the group discussions venience seem to be particularly decisive for who also participated in the written survey the successful positioning of functional food. exclusively consist of housekeeping persons of The functional components add to the food all Nielsen-fields. For the selection of partici- promise an additional positive effect on health. pants two facts were considered most impor- A health-orientated composition of food is sub- tant: on the one hand, a rather representative stituted by functional foods incorporating differ- distribution of age and on the other hand, an ent healthy components; though flops in this appropriate distribution of urban and rural field showed that consideration of the consumer areas. A selection of the informative results trends of health and convenience are not suffi- for the development of acceptance of these 2 cient for the market success of functional food. products is presented here. It is necessary to pay special attention to, for The participants of the group discussions example, sensory qualities in contrast with pre- associate with ‘functional food’ various prod- scription drugs or over-the-counter business. ucts, especially probiotic yoghurts, physical fitness, health, activity and added value. In the discussion of the question: ‘What do you Consumer Motivation to Buy Functional make of functional food?’ there were different Food answers, for example: In order to bring the current consumer trends ● ‘If I get a balanced diet, I don’t need func- and consumer motivation to buy into line, it is tional food.’ necessary to analyse the trends in detail. The ● ‘Unsureness about health value.’

Moral food Environmentally Secure/ Natural/ Functional Minerals/ Low-fat (fair trade) friendly food controlled organic food vitamins with good taste

Freshness Products for families 5 2 Professional Products Security/ Healthy food from star for elderly respons- life cooks people ibility Life style Consumer Pleasure Modern Products needs Trends traditions for children Trends 4 1 Cost- No time/ Food for effectiveness convenience Ethnic food singles 63 Absolute sinful, Fun food sensorial pleasure

Basic Take away/ Snacking Convenience products fast food Fig. 15.2. Most important consumer trends observed in the German food market. Source: Weindlmaier et al. (2001, p. 68).

2 See Biester (2001, p. 33). A popular ﬂopped functional food brand is ‘Aviva’ (Novartis). Consumer - Chap 15 5/3/04 15:56 Page 165

Acceptance and Development of Functional Food 165

● ‘In support of a varied nutrition are these With the aid of tasting some products it was products reasonable, especially when it has possible to substantiate the comments of the to go quickly.’ participants. The evaluation of the different ● ‘It tastes. But I can’t say that I feel better product groups varied signiﬁcantly (Table 15.1). after eating these products.’ ● ‘A little bit of the content of the advertising must be true. There is no chance. You Perspectives of Functional Food have to trust in the products as you can’t control the effectiveness by yourself.’ In order to determine the future perspectives ● ‘Only wheeling and dealing.’ of functional food a Delphi-Study was carried ● ‘I like tasting new products, if I believe in out. Sixty-one experts from the food industry, them.’ the food trade, public and science were asked Figure 15.3 presents the most important in two stages about subjects related to the aspects of consumer acceptance of functional whole value-added chain. food. Generally, good taste, high quality, rea- The experts of the Delphi-Study expect a sonable prices, freshness, and a proven health continuous growth of functional food during value are the most important demand drivers the next 10 years. Within the next 5 years a for functional food. growth of 3% is forecasted (Fig. 15.4).3

Taste

66%

Health value 46% Quality of ingredients 60%

48%

53%

Freshness Price Fig. 15.3. Most important demand drivers for functional food.

Table 15.1. Evaluation/comments by the majority of consumers.

Probiotic ♦ Efficient completion of the range of traditional yoghurt. yoghurt ♦ A surcharge of about 10% is acceptable. Functional ♦ Healthy effects are accepted. However, the diversity of bread bread offered in Germany already stands for wellness. Necessity of bread with functional ingredients is not given. ACE drinks ♦ Not really perceived as a healthy food – no alternative to fresh pressed juice or fruits. Functional fat ♦ Criticized with respect to taste, however, belief in health value. Functional ♦ Product is evaluated to be contradictory (health versus sweet). confectionery but still seen as an alternative to normal sweets for children.

3 Median of expert opinion from the second inquiry of the Delphi-Study. Consumer - Chap 15 5/3/04 15:56 Page 166

166 H. Dustmann and H. Weindlmaier

3.5

2.5

1.5

Market share (%) 1

0.5

0 1997 1999 2001 2003 2005 2007 2009 2011 Fig. 15.4. Growth of functional food during the next 10 years.

Further Results of the Delphi-Study R&D in the German food economy and the inadequate perception of the value Further results of the Delphi-Study can be added to functional food by the consumer summarized as follows: are the main restrictions for a fast market growth. ● Actual trends in consumer behaviour such ● The different actors in the value-added as pleasure seeking, the trend to fitness, chain will not equally participate in the the preference for convenient products, development of further functional food the polarization of markets, and the obso- products. Important will be a closer vertical lescence of society are important drivers cooperation. for the demand for functional food. ● Heavy users of functional food are The results of the Delphi-Study and the con- expected to be athletes and women in spe- sumer survey allow the conclusion that the cific stages of life. market entry of a functional food product will ● Additional functional properties like the positively affect the total demand for the prevention of cancer and cardiovascular product group (Fig. 15.5). diseases and the compensation of malnutrition are important. Research efforts are necessary to incorporate some of the Conclusion properties of functional food into the raw materials, especially into plants. Functional food presents chances for all actors ● For a further growth of the market the posi- in the value-added chain, if the consumer is tioning as premium brands, marketing orien- correctly addressed. The right target group, tation on specific target groups and intensive clear and simple information about the health promotional activities are important. effects, tastefulness and an emotional commu- ● Legal barriers for advertising the health- nication will be the basics for the acceptance related added value, shortcomings in of functional food. Consumer - Chap 15 5/3/04 15:56 Page 167

Acceptance and Development of Functional Food 167

Price P Supply

PHL + FF

PHL

Demand HL + FF

Demand HL Volume X XHL XHL + FF HL = conventional food FF = functional food Fig. 15.5. Demand development at market entry of functional food into the product group. Source: Based on own data collection of the Delphi-Study and group discussions.

References

A.C. Nielsen GmbH (2001) Functional Food weiter im Aufwind. A.C. Nielsen press report, 29 November. Frankfurt am Main. Biester, S. (2001) Verhaltene Stimmung. Der Markt für Functional Food kann bisher nicht halten was vielversprechende Prognosen glauben machen. Lebensmittel Zeitung 53, 33. Menrad, K., Reiss, T., Hüsing, B., Menrad, M., Beer-Borst, S. and Zenger, A. (2000) Functional Food. TA Publikationen 37/2000. Zentrum für Technikfolgen-Abschätzung (Hrsg.), Bern, Switzerland. Soβna, R. (2001) ‘Health food’: an opportunity in stagnating markets. European Dairy Magazine 11, 22. Weindlmaier, H., Fallscheer, T. and Dustmann, H. (2001) Dem Trend auf der Spur. Weiβe Linie: Perspektiven und Erfolgspotentiale. Milch-Marketing 18, 66–71. Consumer - Chap 15 5/3/04 15:56 Page 168 Consumer - Chap 16 19/3/04 9:08 Page 169

16 Factors Explaining Opposition to GMOs in France and the Rest of Europe

Sylvie Bonny INRA (French National Institute of Agricultural Research), UMR d’Economie Publique INRA-INAPG, BP 1, 78850 Grignon, France

Introduction several surveys, which gave insights into the reasons for the opposition. The aim of this The agricultural applications of biotechnol- chapter is thus to contribute to a better ogy were generally seen as highly promising understanding of this opposition movement, in the 1980s. Yet, at the end of the 1990s its determinants and its processes. Over the and in the early 2000s a strong movement past few years extensive literature has been of opposition to genetically modified organ- published on GMO issues. Thus, this text isms (GMOs) has developed throughout the does not address certain topics already stud- world, particularly in some countries. How ied in depth, such as the segmentation of can its rapid growth be explained? Asking opinions according to various categories of this question does not mean that such oppo- consumers or the economic and political sition is unjustified or, conversely, that it is effects of this opposition (e.g. agenda set- legitimate but expressed with great differ- ting, decision making, public policy, regula- ences in emphasis according to the country. tory aspects concerning GMOs, etc.). The aim of this article is to analyse various factors explaining this opposition in France, a country in which it is particularly strong. Growing Suspicion and Concern About This seems especially important in view of GMOs Throughout the World the current deadlock. The French case is fairly representative of various European Various recent opinion polls reveal fairly countries in this respect: even if differences strong and increasing rejection of GMOs exist, depending on cultural characteristics throughout the world. Indeed, the opposi- and economic situations, a number of fac- tion movement intensified after 1998; prior tors of opposition are found throughout. We to that, opinions were on the whole more studied the processes and mechanisms of favourable (Hoban, 1997). To be sure, the the development of the opposition move- results of such surveys have to be inter- ment. We also examined recurring topics in preted with caution, for they have various discussions and debates on the subject, in limits. These limits are above all: the risk of the discourse of opponents and in articles artefacts when the respondent has to on GMOs in the media. We furthermore choose, in a short space of time and out of monitored and observed initiatives by the context, an answer in a series of items pro- various actors, and drew on the results of posed on a complex subject; the influence of

170 Sylvie Bonny

the formulation of the questions and their International level interaction on the answers obtained; the impact of the context or of recent events; At the international level, a survey carried out in and, lastly, the risk of superﬁciality of the early 2002 showed that concerns about GMOs approach compared to more in-depth inter- have spread around the world, even to develop- views. Moreover, a stated opinion may differ ing countries and the USA (Table 16.1). In from an effective opinion, and from February–March 2002, the Ipsos-Reid Group, behaviour in a real situation where other an opinion survey institute, carried out a poll factors play a part. In spite of all, polls dealing with some food issues in 12 countries prove to be a useful source of information, (500 people questioned per country, 1000 in to be used with other methods; they have the USA). One question asked regarding the the advantage of providing indicators on level of concern about the safety of GM foods vast samples representing the group under (Ipsos-Reid, 2002). Across the 12 countries, an study. Bearing these limits in mind, we here- average of 76% of those polled said they were after present the data of some recent sur- worried about GM food safety. Almost half veys of perceptions of GM foods. (48%) ‘strongly agreed’ and 28% ‘somewhat

Table 16.1. Concern about GM food in various countries throughout the world by country and socio- demographic characteristic (Ipsos-Reid, 2002): ‘Please tell me if you agree strongly, agree somewhat, disagree somewhat or disagree strongly with the statement: I am concerned about the safety of genetically modiﬁed foods?’ (% of answers).

Agree strongly Agree somewhat Total agree

Japan 54 37 91 South Korea 55 27 82 Germany 62 18 80 Brazila 58 21 79 France 54 23 77 Canada 46 29 75 USA 44 28 72 Indiaa 44 28 72 South Africaa 41 30 71 Russiaa 40 31 71 UK 45 22 67 Chinaa 34 29 63 Overall 48 28 76

aBrazil, China, India, Russia and South Africa data represent urban- only samples.

Overall strong Gender Age Relative income agreementa Men Women 18–34 35–54 55+ Lower Middle Higher

North America 44 37 51 41 47 42 49 43 40 Europe 50 45 55 46 54 51 53 50 50 Asia-Paciﬁc 51 50 51 44 52 59 47 50 59 Overall 48 44 52 45 51 49 49 48 49

Europe includes France, Germany, urban Russia, UK. Asia-Paciﬁc includes urban China, urban India, Japan, South Korea. Income groupings are nationally relative, and represent the bottom, middle and top income strata within each country. aPercentage of respondents who agree strongly. Consumer - Chap 16 5/3/04 15:56 Page 171

Opposition to GMOs in France and Europe 171

agreed’ when they were asked if they agreed or USA disagreed, strongly or moderately, with the statement ’I am concerned about the safety of In the USA, the International Food Information genetically modified foods’. Only one in five dis- Council (IFIC) commissioned surveys on a sam- agreed ‘somewhat’ or ‘strongly’ with the state- ple of about 1000 people in 1997, 1999, ment. From country to country, the proportion 2000, 2001 and 2002 (IFIC, 2002). It enables of people who are concerned ranges from a a follow up of opinion several years in a row low of 63% (in urban China) to a high of 91% because the same questions were asked several (in Japan). The percentage who ‘strongly years in a row. Various questions on attitudes agreed’ is the highest (more than 54%) in toward food biotechnology were formulated by Germany, Brazil, South Korea, France and referring mainly to its advantages. Therefore Japan. In many countries women are more the survey results can be used more as regards likely than men to be concerned about this the trends than in absolute value compared to safety aspect (52% versus 44%, respectively) others which have more neutral formulation. (Table 16.1). According to the chosen delimita- Americans seem to be favourable to biotech- tion of age groups and large geographical nologies. However, the feeling that biotechnol- zones, the gender difference appears quite pro- ogy will be beneficial diminished between 1997 nounced in North America, less so in Europe, and 2000, particularly in 1999, but tends to be but hardly perceptible in the other countries sur- quite stable later (Table 16.2a). veyed. Old people seem to be more disquieted, In January 2002, the Pew Initiative on particularly in Asia, but elsewhere age seems to Food and Biotechnology commissioned a sur- be less a factor of differentiation as far as GM vey dealing with environmental issues of food safety is concerned. In North America genetic engineering (Pew, 2002). People people from lower-income households are more were asked ’Overall, which do you feel is worried than the wealthier about the safety of greater – the environmental risks or the envi- GM food. It is the opposite in Asia-Pacific, per- ronmental benefits of using biotechnology to haps because people with lower income are less genetically modify plants, animals, fish or educated in such zones and less aware of this trees?’ This question was asked twice: on the issue. In addition, in China and India, con- one hand, at the beginning of the survey, on sumers from lower-income households may the other hand, at its end (Table 16.2b). The have more urgent concerns and so be less likely rest of the questionnaire mainly addressed to pay attention to these questions. several other points dealing with more

Table 16.2a. Benefits expected from biotechnology in the USA (IFIC, 2002): ‘Do you feel that biotechnology will provide benefits for you or your family within the next five years?’ (% of answers). March 1997 Feb. 1999 Oct. 1999 May 2000 Jan. 2001 Sept. 2001 Aug. 2002

Yes 78 75 63 59 64 61 61 No 14 15 21 25 22 17 18 Don’t know/ 8 10 16 16 14 21 21 refused

Table 16.2b. Perception of environmental risks and benefits of genetic engineering (Pew, 2002) ‘Overall, which do you feel is greater – the environmental risks or the environmental benefits of using biotechnology to genetically modify plants, animals, fish or trees?’ (% of answers). Which do you feel are greater? Risks Benefits About the same Not sure

Pre-statement (ﬁrst question) 40 33 19 9 Post-statement (last question) 38 38 21 3

US nationwide survey of 1214 adults conducted by Zogby International in January 2002. Consumer - Chap 16 5/3/04 15:56 Page 172

172 Sylvie Bonny

detailed and precise aspects of environmental on about 16,000 people in the EU1. It risks or benefits of GM. Thus, at the end, the enables a comparison between the different respondents were more aware of GMOs; EU countries because the same questionnaire however, there could be a bias in the answer has been used in the 15 EU countries. The to the last question according to the charac- questionnaire, filled in during face-to-face teristics of the examples mentioned in the rest interviews, addressed various issues related to of the interview. In this list of possible envi- technological and scientific progress ronmental risks and benefits of GMOs, the (European Commission, 2001). Among them, interviewed people had to grade each of them some questions dealing with biotechnology according to the importance of effect. show a high level of mistrust of GMOs. The specific environmental risks on the poll were: Previously, in the 1990s, other drifting genes, creating ‘superweeds’, increasing Eurobarometer surveys were specifically pest resistance, affecting non-target organisms, devoted to biotechnology and tackled many reducing biodiversity, or changing the ecosystem. aspects of opinion about it (European Benefits listed were: engineering plants to clean Commission, 2000; Gaskell et al., 2000). up toxic waste, reducing soil erosion, reducing From the 2001 Eurobarometer, the results of run-off, needing less water to grow crops, saving only one question are presented here: the endangered or threatened species, reducing the feeling of danger about food based on GMOs need to log in native forests, or reducing pesticide use. Asked to rank these 13 items in terms of (Table 16.3). personal importance, the environmental benefits A majority of Europeans (56.4%) believe scored significantly higher than any of the risks that food based on GMOs is dangerous, as listed, with the exception of the non-target opposed to 17.1% who do not; however, this organism issue nationally … Prior to reading a is an open question for more than a quarter series of informational statements about the of Europeans (26.5% of ‘don’t know’). possible benefits and risks of biotechnology, Variations by country are significant: in The respondents nationwide were more likely to say Netherlands, Finland, the UK and Sweden that the risks of biotechnology outweighed the benefits (40% to 33%), while 19% thought the less than 46.5% of people are wary about the benefits and risks were about the same, and 9% danger of GM food. However, in Greece, were unsure. However, after being read a series France and Luxembourg, more than two- of questions about specific environmental risks thirds of the inhabitants believe the same. and benefits (without specifically identifying which This greater level of variation by country than were risks or benefits), respondents were exactly by usual socio-demographic variables can be evenly divided, with 38% saying the risks linked to the importance of cultural aspects, outweigh the benefits and another 38% saying as well as to the differences in public debate, the benefits outweigh the risks. (Pew, 2002) government intervention, history of economic development and industrial situation between the various European countries (Zechendorf, European Union 1998; De Cheveigné et al., 2002; Springer et al., 2002). However, variations per socio- In the European Union, Eurobarometer sur- demographic characteristics are smaller than veys reveal high suspicion towards biotechnol- those per country. The feeling of danger is a ogy. The most recent Eurobarometer survey little lower among managers, students, high- on this topic was carried out in spring 2001 income people, educated people (i.e. those

1 More precisely, a total of 16,029 people was questioned between 10 May and 15 June 2001. In each EU Member State a representative sample of the national population aged 15 and over was taken, with an average of some 1000 people per country, except in Germany (1000 in the new Länder and 1000 in the former Länder), in the UK (1000 in Great Britain and 300 in Northern Ireland) and in Luxembourg (600). This opinion poll, managed and organized by the EC Directorate-General for Press and Communication, Public Opinion Sector, was carried out at the request of the Directorate-General for Research. It was conducted under the general coordination of EORG, the European Opinion Research Group, a consortium of market study and public opinion agencies. Consumer - Chap 16 5/3/04 15:56 Page 173

Opposition to GMOs in France and Europe 173

Table 16.3. Opinion of GMOs in the EU, by country and by socio-demographic group (European Commission, 2001): ‘Do you think it is true or false that food based on GMOs is dangerous?’ (% of answers).

Country True False Don’t know

Netherlands 37.9 31.7 30.4 Finland 43.0 34.3 22.8 UK 44.8 23.9 31.3 Sweden 46.6 25.0 28.4 Denmark 48.,9 25.8 25.4 Belgium 51.4 21.3 27.2 Germany (E) 53.8 15.1 31.0 Germany 53.8 15.8 30.4 Germany (W) 53.8 16.0 30.2 Total EU 15 56.4 17.1 26.5 Portugal 57.2 10.5 32.3 Ireland 58.3 13.4 28.3 Italy 59.8 14.6 25.6 Spain 60.8 14.2 25.0 Austria 64.4 15.0 20.6 Luxembourg 66.6 10.9 22.5 France 67.6 12.4 20.0 Greece 88.8 3.2 8.0

Germany (E), new Länder; Germany (W), former Länder. Countries are ranked by increasing level of perception of GMOs as being dangerous.

Socio-demographic variables True False Don’t know

Gender Male 54.0 20.3 25.7 Female 58.7 14.1 27.2 Age (years) 15–24 54.6 20.8 24.6 25–39 56.4 19.3 24.3 40–54 58.4 16.9 24.7 55+ 55.9 13.5 30.6 Education level 15 58.0 11.7 30.4 (age when left school) 16–19 57.9 17.5 24.6 20+ 53.2 20.8 26.0 Still studying 53.0 23.4 23.7 Area Rural/village 59.0 15.3 25.7 Small town 55.6 17.9 26.5 Large town 55.0 17.9 27.1 Don’t know 49.9 16.9 33.1 Profession Self-employed 61.1 15.6 23.3 Managers 48.8 24.1 27.1 Employees 57.1 19.1 23.8 Manual workers 58.0 17.5 24.5 House-persons 59.3 12.5 28.3 Unemployed 58.4 15.5 26.1 Retired 55.7 13.2 31.1 Students 52.3 23.6 24.1 Income level – – 57.7 13.7 28.6 – 56.6 16.6 26.8 + 56.6 18.8 24.5 + + 53.1 22.5 24.4 Don’t know 57.3 15.5 27.2 EU 15 EU 15 56.4 17.1 26.5 Consumer - Chap 16 5/3/04 15:56 Page 174

174 Sylvie Bonny

who have studied beyond age 20) and men: because of their control over timing, their ability among them, less than 54% think that it is to have their position treated as credible (at least true that ‘food based on GMOs is dangerous’. parts of the media), and their ability to However, among self-employed, house-per- command an agenda and debate. In all of these sons, rural area or village inhabitants, and elements they have frequently been more successful than the risk experts and managers. women, more than 58.7% are worried and (Petts, 2002) the feeling of danger is higher. Thus, people who are a little more vulnerable or fragile In the GMO field, the media played a sig- appear to be a little more worried about the nificant role as has been shown (Petts et al., potential risks of GMOs. 2000; Ipsos Corporation unpublished study, 2002; Frewer et al., 2002; Petts, 2002). In addition, GMOs surfaced in force at The Focus on Potential Risks and the about the same time as the public’s confi- Extensive Publicity Given to Them dence in institutions and certain technological advances had been shaken by several Thus in France, opposition to GMOs and con- safety affairs, in particular the issues of cont- cern about their risks appear to be among the aminated blood, mad cow disease, asbestos, strongest, particularly when one compares the etc. To make an extremely brief summary of opinions in France with those expressed in these affairs, one can say that in the case of most other European countries. What factors ‘contaminated blood’, through blood transfu- explain this increasing hostility in public opin- sions patients received blood products conta- ion? A number of processes commonly used minated by the AIDS virus when in fact the and a number of products currently on the state of knowledge at the time could have market present drawbacks and offer only illu- allowed this risky practice to be limited. In sory benefits. They are, however, well-estab- the case of ‘mad cow disease’, despite pre- lished and generate only a small degree of sumptions of risks, stringent measures on rejection. In the same way, in technical fields cattle feed and meat imports were some- numerous process innovations are hardly times taken with much delay – or were not noticed or are known only in limited circles. complied with – primarily to protect eco- However, the controversy over transgenic nomic interests in the sector. About plants drew a wide audience and received asbestos, although its risks had been known extensive concern, especially in the late for a long time, it continued to be used, 1990s. The intensity of the GMO debate especially to protect the interests of this heightened and opinions became increasingly industry which was an influential player in radical. One important factor is the frequently the official body responsible for evaluating critical, or even negative, information spread and managing risks. These events led to defi- about them, in particular by opponents and nite distrust of firms and public authorities the media (see below). Of course, the influ- and increased the public’s attention to criti- ence that the information received from the cal voices, and so the principle of precaution media has on people’s opinions is not linear became an omnipresent reference. and direct: people are not merely passive spectators and ‘recipients’ of information, particularly if significant risks are involved. The strong influence of associations that However, the media play a significant role in focus on risks particular when several actors are in contest to advance their views. In France GMOs have been strongly opposed by various associations. Initially these con- The media are active mediators through (i) their focus on particular issues and neglect of others, sisted essentially of ecologist associations (ii) the process of legitimising certain points of (Greenpeace, Friends of the Earth, etc.) and view – i.e. who gets to speak, and (iii) their groups of various tendencies (e.g. Ecoropa, significant ability to resonate with the public the Natural Law Party), as well as supporters mood. They compete effectively on the field of the Green political parties and organic agri- Consumer - Chap 16 5/3/04 15:56 Page 175

Opposition to GMOs in France and Europe 175

culture associations. This movement progres- various groups or associations to enhance sively expanded from environmentalist circles their renown, recognition and resources, and towards groups active in the economic to acquire a degree of legitimacy by present- domain including, for example, a farmer’s ing themselves as defenders of consumers and union – the Confédération Paysanne – anti- of their health and interests, but also of the globalization organizations (ATTAC, the environment and of the interests of develop- Association for the Taxation of Financial ing countries or of future generations. Since it Transactions for the Aid of Citizens), LETS has proved to be so fruitful, this encourages (Local Exchange Trading Systems), etc. them to pursue their militancy in this field and Finally, small circles of associations were cre- to devote more resources to it. ated for the very purpose of fighting against Many actors are involved more or less GMOs, for example OGM Danger!, OGM directly in the GMO field but their respective Dangers, Terre sacrée, etc. influence varies widely. For the entire EU, The impact of these associations has been the Eurobarometer survey in late 1999 strong, owing to the dynamism of their action showed that the actors who were judged which gave them extensive publicity: numer- most often by respondents as ‘doing a good ous strongly worded press releases, the job for society’ as regards GMOs were pri- repeated mass dissemination of alerts and marily consumers’ unions, doctors, the warnings, petitions, leaflets, standard letters to media and environmentalist groups. By con- send to elected representatives, agro-food trast, industry is the only actor judged most firms or food retailing companies, lawsuits, often as not doing ‘a good job’ for society in demonstrations, and so on. In particular, these this respect (Table 16.4). The countries most groups took advantage of the new communi- hostile to industry’s biotechnology activities cation technologies: multi-transmission of are Sweden, Greece, France, Denmark, information via electronic mailing lists, forums Ireland, Italy, Austria, Luxembourg and the of discussion on the internet, very well-docu- UK. This sheds light on certain determinants mented and frequently updated websites used of opposition to GMOs and on the respec- extensively by many as sources of informative impact of the actors involved. Industry tion, etc. The endless reuse of certain infor- seems to have little credit and its arguments mation (sometimes very partial or biased) gave are therefore taken into consideration rela- it credibility due to multiple repetition that tively little or are even discredited. By con- ended up making it seem reliable (since it was trast, other actors that are often opposed to frequently mentioned, it was corroborated). In GMOs – consumer unions, environmentalist addition, the influence of groups that had associations and the media – have more taken a stand against GM extended way legitimacy and are therefore taken into beyond their own supporters to many sympa- account and quoted more often. thizers or people close to them. The mobilization of the staff and members of many associations on this issue was Behaviour of other actors involved in intense, not only because they felt strongly publicizing information on risks about it but also because it helped to establish their audiences and legitimacy, especially in The publicity given to various associations’ the case of diverse associations that were for- denunciation of GMOs has been noteworthy, merly in a tiny minority. For example, its anti- particularly in the case of the media which GMO action was instrumental in have played a significant part in making strengthening Greenpeace-France which had GMOs widely known and in highlighting their been in serious financial straits and was expe- potential dangers, especially from 1999 riencing a relative drop in its membership when many journalists became increasingly compared to other north European countries. opposed to GMOs. Whereas previously – Greenpeace now has sound legitimacy and is especially in the early 1980s when there invited to many debates and conferences. were few articles on the subject – the media More generally, the GMO issue has enabled presented biotechnology as a promising inno- Consumer - Chap 16 5/3/04 15:56 Page 176

176 Sylvie Bonny

Table 16.4. How the activity of the actors involved in biotechnology in the EU and in France is perceived (European Commission, 2000) (% of answers). Tend to Tend to agree disagree Don’t know Do you think they are doing a good job for society: EU France EU France EU France

Consumer organizations checking products of 70 73 12 14 19 12 biotechnologya Medical doctors keeping an eye on the health 69 69 11 15 20 16 implications of biotechnology Newspapers and magazines reporting on biotechnology 59 56 18 25 23 19 Shops making sure our food is safe 59 48 21 35 20 17 Environmental groups campaigning against biotechnologyb 58 66 18 18 24 16 Farmers deciding which types of crop to grow 55 53 20 27 25 20 Ethics committees looking at the moral aspects of 53 55 18 23 29 22 biotechnology Our government making regulations on biotechnology 45 40 29 39 26 21 The churches giving their points of view on biotechnology 33 20 31 44 35 36 Industry developing new products with biotechnologyc 30 25 38 51 32 25

aThe countries most favourable to the consumerist associations are The Netherlands, Finland, Greece, Denmark and Austria. bThe countries most favourable to the ecologist associations are Greece, Austria, Denmark, France and Finland cThe countries most favourable to industrial activity in biotech are Finland, The Netherlands, ex-East Germany, Belgium and Portugal. Countries which are the most opposed to industrial activity: Sweden, Greece, France, Denmark and Ireland.

vation, in 1999 and 2000 increasingly critical Furthermore, the communication methods opinions were often expressed. A number of of associations opposed to GMOs often guar- journalists focused on risks and expressed anteed them a strong impact in the media. standpoints opposed to GMOs, sometimes These associations focused on spectacular entering into opposition movements them- actions announced in advance. Pictures of selves (Durant and Lindsey, 2000; activists chained to or climbing on to strategic Kassardjian, 2002). This can be explained by or symbolic places, photographs of large various factors. Initially, the subject of protest banners, destruction of transgenic biotechnology was treated by scientific jour- crops, and so on, had every chance of receiv- nalists who were relatively in favour of it. ing extensive media coverage due to their char- Later, when the topic became more politico- acteristics and attractiveness. This is precisely economic, it was also covered by other jour- one of the aims of this type of action (Ruckus nalists, for example those who had worked Society, no date). Likewise, their press commu- on the issues of contaminated blood, mad niqués were particularly lively, stimulating and cow disease, etc., and who drew parallels clear, and their websites well documented. between these issues. Another explanation On the other hand, the firms involved have lies in the characteristics of the journalistic often maintained a more traditional type of profession and the strong competition within communication, strongly influenced by their the media sector. Shocking headlines reveal- usual clientele – the agricultural sector, not the ing hidden dangers and dramatic presentation public at large. Moreover, until 1997–1998 of issues guarantee wider audiences and have they often underestimated suspicion of GMOs, more impact than more moderate, qualified considering it to be the product of irrational articles; hence, this tendency to overstate and and somewhat residual fears that would pro- try to outmatch one another. gressively disappear as more information Consumer - Chap 16 5/3/04 15:56 Page 177

Opposition to GMOs in France and Europe 177

became available. They often seemed to con- magazines for the general public or popular- sider that the rejection of GMOs was mainly ized science magazines, and to participate in the result of poor knowledge of biotechnology scientific conferences rather than in debates and that the public only needed to be with the general public. In fact, the latter educated. But their promotion of the advan- forms of publicizing results are even fre- tages of GMOs did not convince the public. quently discredited in the scientific world. As for the public research organizations, on Even if researchers have participated in public the whole they did relatively little public rela- debates, in total these have reached only a tions work on the subject in France, especially very small audience. Thus, on the whole little compared to the associations involved. but silence can be heard from public research. Institutional communication often remained So, views against genetic engineering voiced focused on the presentation of important by certain scientists are quite frequently cited. results obtained by research teams. No state- These scientists are judged as saying aloud ments were issued to clarify the matter when what others dare only to think (for fear of los- facts or controversies on specific points con- ing their contracts with private firms, or dis- cerning GMOs were mentioned in the media suading public financing). Public research has, (which was very often). As a result, explana- moreover, published relatively few books or tions and interpretations disseminated very statements for the general public on GMOs. It widely among the general public were often has participated in many fairly specialized sci- those from associations opposed to GMOs. entific conferences on this theme, but these Researchers from public research organiza- have received little attention outside scientific tions were interviewed, but they were often circles. In contrast, the book Tais-toi et quoted too selectively (or partially) in articles mange! L’agriculteur, le scientifique et le that mainly reflected the viewpoints of associa- consommateur [Shut-up and eat! The tions opposed to GMOs. In addition, the views farmer, the scientist and the consumer] expressed by scientists tend to be complex (Paillotin and Rousset, 1999), written by the while those expressed by opponents are very president of INRA, was judged simply on the loud and clear: ‘GMOs are dangerous, we basis of its title as legitimizing reservations. must ban them’. We note that the opinions of Various consumers’ unions also became researchers in the life sciences on genetic engi- strongly involved in the GMO controversy, neering vary, depending essentially on their without being opposed from the outset. They specific discipline, but that the vast majority of stressed the need to take into consideration those working in molecular biology risks and the principle of precaution. In late believe that recombinant DNA techniques 2000, delegates at the International constitute powerful and safe means for the Consumers’ Organization conference in modification of organisms and can contribute Durban called for a moratorium: substantially in enhancing quality of life by governments and international institutions improving agriculture, health care, and the should require full pre-market evaluation and environment. … We … express our support for social and safety impact assessments of GM the use of recombinant DNA as a potent tool for foods and the products of other new food the achievement of a productive and sustainable technologies to ensure that they are safe, agricultural system. (AgBioWorld, 2002) environmentally sustainable and acceptable to To be sure, the point of view of consumers, and impose a moratorium on the researchers working in the environmental sci- cultivation and marketing of new GM foods until this is done. (Consumers International, 2000) ences is often somewhat more reserved, and many scientists are concerned about patents In the EU the European Consumers’ or other economic aspects, but generally Organization (BEUC) emphasizes that the researchers working in the life sciences think recent years have shown that GM food will that genetic engineering is a useful tool. never be accepted without consumer choice. However, evaluation methods in public orga- A press release from July 2002 entitled nizations urge them to publish in highly spe- ‘GMOs or No GMOs: the choice should be cialized scientific journals far more than in ours!’ reminds of its demand ‘to ensure that Consumer - Chap 16 5/3/04 15:56 Page 178

178 Sylvie Bonny

the key consumer rights to information and tion are often considered to be weak or non- choice are met’ and claims ‘only clear existent, while their risks are considered to be labelling will ensure that consumers can substantial (Gaskell et al., 2000; European choose whether or not to buy GM foods’. Commission, 2000, 2001; De Cheveigné et The question of labelling of GM products al., 2002). helped to radicalize the debate. In 1998 Greenpeace launched an ‘Info-conso network’ with the slogan ‘no GMOs on my plate’, list- Advantages of GMOs judged weak by many ing products and brands according to whether or not they contained GM ingredients, and Opponents of GMOs presented them as a stigmatizing those that did. It urged con- technology with high potential risks and with sumers to ask producers or distributors to no advantages except for the firms that devel- adopt measures necessary to ‘preserve oped them. They strongly emphasized two Europe and the food chain from contamina- arguments: tion from GMOs against consumers’ will’. 1. GMOs comprise many risks, from which This movement was strongly relayed. To avoid no one can escape since they concern daily a loss of market share, one by one many food and the immediate environment. They agro-food or mass distribution groups commit- also comprise other, more global, potential ted themselves to excluding GMOs in France, dangers and risks for farmers in developing in Europe and sometimes also in the USA. countries and for biodiversity, which legit- In this context in which many influential imizes opposition and actions against them actors (the media, associations) denounced the and even makes this opposition essentially risks of GMOs, hostility towards them seemed ethical. to many to ‘stand to reason’, simply on the 2. Their possible benefits will go primarily basis of information received or because that to the firms that produce them and not to was the standpoint of the ideological move- society as a whole or to consumers. On the ment to which they felt closest. contrary, consumers’ and society’s safety is sacrificed. Many risks or negative effects are suspected It is not surprising that, thus presented, GMOs in a very wide field were met with suspicion, especially since these arguments of distrust of GMOs were We have established a typology of these risks, perceived as credible in a context where agri- fears and reasons for refusal, on the basis of cultural productivism is strongly called into the subjects mentioned repeatedly in debates, question and suspicion reigned in the after- articles and arguments against GMOs (Table math of several public health affairs. On the 16.5). This focus on the risks of a technical other hand, arguments which tried to present innovation is nothing new. At the time of the potential advantages of genetic engineer- their introduction many innovations were vio- ing were often rejected because they were lently opposed, e.g. industrial mechanization, perceived as hypocritical. the railway, the potato, etc. (Salomon, 1984). In the USA, where opinions were more But in the case of GMOs this opposition has favourable to GMOs, people often thought been particularly strong and widespread, so that opposition in Europe came from its rela- that in Europe this innovation has comparative backwardness in this field, and that argu- tively few advocates or supporters. ments on risks concealed a form of protectionism aimed at avoiding the disman- tling or buy-out of the European seed indus- A Risk–Benefit Assessment of GMOs try. Another frequent US interpretation is that Perceived as Very Unbalanced GMO refusal aimed to protect the European market from US grain importations. But, One of the causes of European opposition to while this fear might sometimes have had GMOs is that their advantages in food produc- an influence, it did not stem mainly from Consumer - Chap 16 5/3/04 15:56 Page 179

Opposition to GMOs in France and Europe 179 st he themes repeatedly ntamination’ of grain rms opinion: ‘They’re hiding something from us’ → GMOs accumulate products of degradation the desire to return true nature (growing interest in organic products) → impure harvests, ’contaminated’ → → ‘superweeds’, invasive plants, accelerated decrease in biodiversity → consuming continuously secreted insecticide toxins → hypocrisy of saying ’Genetic engineering is necessary to feed humanity’ → insufficient knowledge of the genome to authorize such tinkering with transfer foreign genes (living organisms are not ju insufficient ‘building blocks’) ’Such progress, why bother?’ (a certain loss of faith in science and progress) through gene flow Media showing scientists (or associates) opposed to GMOs } Problem of volunteer plants in the following crop (rapeseed) interesting molecules for use in other agricultural sectors Risk of a drop in Bt or glyphosate efficiency, Increasingly dependent agriculture (farmers must buy seeds every year) for developing countries to access such technology (patents) Difficulty in the positions taken by public authorities Vacillation Perception ’Everything is messed with more and more’ GMOs symbolize development towards a type of society which is perceived negatively } Appropriation of genetic resources by a few large multinationals GMOs = symbol of privatization all resources, now even genetic resources ’Imperialist’ (gene flow) technology because coexistence with non-transgenic production is difficult Motives put forward for GMO rejection: risks, fears and reasons refusal (typology developed by the author on basis of t Bt maize Gene coding for Bt toxin glyphosate-tolerant soya Gene coding for the enzyme which degrades glyphosate ● ● for example safety tests: ‘consumers = guinea pigs’ Insufficient Table 16.5. Table treated in debates, articles and declarations made by opponents). of riskTypes violent gene Troublesome, transfer process = transgression of the barrier between species Transgenesis Health, Fears and perceived risks Risk engendered by troubling the ‘order of genome’, which may appear only later Allergies, long-term toxicity EnvironmentalAgro-economic Gene flow towards related wild species Gene flow towards nearby crops of the same species Economic Of little interest to consumers, ‘product imposed’ by the multinationals Agricultural and food production modelMore socio-political motives (value systems and beliefs) Innovation neither asked for nor desired, but set up solely the profits of some multinational fi Reinforcing of the industrialized model, limits which have already been critically portrayed No respect for consumer free choice due to the presence of GMOs in many additives and fortuitous ‘co Consumer perception: ‘They’re playing with our health to make more money’ (cf. BSE and contaminated blood). Consumer - Chap 16 19/3/04 9:07 Page 180

180 Sylvie Bonny

economic protectionism since the French and there is a missing fundamental element in this other European people’s rejection affected progress which constitutes the king-pin between the biotechnology activities of their own coun- science and society: the utility function … One of tries just as much. the basic springboards towards acceptance of GMOs thus lacked supporters and allies in innovation is the risk/utility equation … If an invention arouses incomprehension as to its many European countries, including France. usefulness while presenting a potential risk factor it Moreover, in the late 1990s the public author- is doomed to a motivated rejection. (Boy, 1999). ities adopted a hesitant attitude in this respect, often backtracking and procrastinating, which In other words, even if GMOs can have heightened confusion and perplexity (De advantages for society as a whole and for all Cheveigné et al., 2002). Thus, faced with actors, the actors situated downstream from strong denunciation by various associations, production – who today have considerable extensively relayed by the media, there were weight – judged them as being of negligible few actors to present GMOs in a favourable and often even of no interest compared to light: firms were judged as having little credi- their potential and unknown risks. Nothing bility and public research organizations made justified the use of GMOs, perceived as serv- few public and official statements on the sub- ing only the interests of the firms involved; ject; the few scientists or their allies inter- any risk-taking seemed unjustified. viewed by the media were in some cases against GMOs; and, lastly, the authorities seemed confused and hesitant. The European Some words on risk perception and French situation is, in this respect, very Concern as regards GMOs cannot simply be different from that in the USA where GMOs imputed to a lack of knowledge in biology, as enjoyed extensive support. many actors arguing for better education of Yet an abundant scientific literature has the public have done (Miller and Conko, been published on the potential benefits of 2000). In addition, the public at large cannot GMOs, generally evaluated as being greater be accused of irrationality, as research con- than the foreseen risks: more efficient agri- ducted in several European countries shows cultural production with reduced losses; (Marris et al., 2001). Various studies have increase in productive capacities in difficult enabled us to better understand risk percep- conditions; improvement in various qualita- tion. Experts evaluate it in relation to two tive characteristics; diversification of uses of components: the probability of an undesirable plants with the possibility to produce diverse event actually happening, and the seriousness molecules, etc. We mention only a few refer- of its consequences. The public, on the other ences here on this topic which alone necessi- hand, takes into account a set of other factors tated a considerable amount of synthesis in its assessment of risks, as many studies work (Conway and Toenniessen, 1999; have shown (Slovic, 1987; Morgan, 1993; Borlaug, 2000; Interacademies, 2000; Slovic et al., 1995; Powell, 1998; Siegrist, AgBioWorld, 2002; ASPB, nd; ACN, 2002; 2000). These factors are in particular: SOT, 2002). Biotechnology opens the possibility of a new path for technological devel- 1. The knowledge of and familiarity with opment, based more on living processes and the specific risk: household accidents and on information (knowledge) than on chemi- automobile accidents generate less worry cal inputs. But in Western Europe, in a con- than the potential dangers of GMOs. The text of agricultural overproduction and invisible or uncontrollable is especially prone extensive calling into question of agricultural to provoke anxiety. productivism, these aspects hardly aroused 2. The delay before the appearance of bad much public interest. Moreover, the very first consequences: some important risks (such as products commercialized (transgenic soy- heavy smoking or sun-tanning without precau- beans and maize) seemed to consumers to tion) are quite often taken deliberately with have little interest, especially in Europe. As lack of concern for the consequences because Boy puts it they will appear only in the distant future. Consumer - Chap 16 5/3/04 15:56 Page 181

Opposition to GMOs in France and Europe 181

3. The semblance of catastrophe: an acci- a very wide field, including many socioeco- dent affecting several people at the same time nomic or political aspects. Hence consumer and place usually has more impact than indi- suspicion is not rooted solely in a lack of vidual accidents spread across time and space, knowledge of production and risk. The under- even if the total number of people affected is lying issues are much more complex: much lower. The argument that scientific literacy, or more 4. The possibility for those exposed to risk to knowledge about a technology, will increase control the risk: the feeling of mastery is the support assumes that there is no reasonable essential point. basis for opposition. It assumes that technology 5. The voluntary or involuntary nature of the is objectively a desirable thing and that event. One is more angry about being opposition must stem from ignorance of its true exposed to an inescapable risk than to a risk benefits and costs. This approach not only fails from which one can escape (or choose for to receive empirical support, but adhering to it, conceptually, severely jeopardizes policy oneself); one tolerates deliberate risk best. development and communication strategies … 6. Risk-related advantages for the person Proponents of technology often argue that exposed to or taking the risk: a risk which public perceptions of risk are irrational. Yet, brings profits to the person responsible for its public risk calculations are rational, given the creation but not to the person exposed to it socio-ethical perspective from which they are induces a high level of indignation. People are derived. In fact, expert risk assessments also much more shocked by the impact of acci- stem from a specific socio-ethical perspective. dents to children because of the ‘innocence’ The difference between expert and public risk of the victims. perceptions should be seen as the difference in 7. Scientific uncertainty and controversy: the socio-ethical perspective which defines the calculation, not as the difference in rational poorly understood risks make everyone ner- ability. (Espey, 1998) vous. In the case of controversy the public suspects those who minimize the risk of having vested interests in the affair or of being obliged to take their position by people who Diverse Opposition to and Concerns want to avoid a crisis (as for BSE in the UK at About the Functioning of Society and its the end of the 1980s). Evolution Crystallized Around GMOs 8. Confidence in institutions. The perceived risk of biotechnology will be Limited trust in the institutions and firms significantly influenced by trust in the system involved that produces it … Components of the relationship that builds trust, or distrust, is the Despite parliamentary debates in 1992 at the extent to which an individual feels affiliated with time of the transposition of European direc- the system, agrees with the distribution of tives concerning the dissemination of GMOs, decision-making power, and shares the values and various articles in the media on biotech- enabled by the system. (Espey, 1998) nology, discussion on the subject remained Thus, some individual practices which rep- limited in the early 1990s to a fairly small cir- resent a true danger, such as heavy smoking cle. It started to spread in the public at large or driving a car, arouse less worry than mainly from late 1996 when the very first genetic engineering which is less well known, imports of transgenic oilseeds from the USA unobservable, difficult to control and can lead arrived in Europe and animated debate sur- to risk exposure which is not a question of rounded authorization of Bt maize from the personal choice. So, acceptability depends on firm Novartis. At that stage public opinion numerous factors in relation to risk percep- was strongly marked by various affairs, espe- tion. In addition, this acceptability is highly cially ‘contaminated blood’ and BSE, which contingent on the consumer assessment of led to strong distrust and caused people to the expected benefits justifying risk taking and think that firms and public authorities some- offsetting potential bad effects. But the con- times disregard certain health risks in order to sidered risks of GMOs have been extended to protect certain economic or political interests. Consumer - Chap 16 5/3/04 15:56 Page 182

182 Sylvie Bonny

In the period 1998–2000, debate on GMOs ing the policies of the public authorities and (authorization, importation, labelling, impact, ﬁrms involved in the commercialization of etc.) was situated in a context strongly inﬂu- GMOs increased sharply. It was, moreover, enced by food safety issues (BSE, listeriosis, fuelled by the many turnarounds and etc.) that had been widely publicized. As a instances of procrastination which could give result, GMOs were perceived as an additional the impression that ‘they’re hiding something indication of negligence when it came to from us’ or that too many unknowns still health hazards. The precautionary principle existed (Table 16.6). In July 2000 a majority therefore became extensively invoked and (58%) of respondents said they tended to dis- adduced. agree with the opinion that ’the public author- One of the arguments often put forward by ities can be trusted to make good decisions on the promoters of biotechnology to justify its GMOs’ while 40% tended to agree (IFOP and development is that it is necessary for feeding Libération, 2000). the world’s population, particularly in the coming decades. Although it is valid in several respects, this argument has frequently been GMOs – symbol of negatively perceived perceived as highly hypocritical when used by trends multinationals. Indeed, these corporations frequently adopted a policy of patenting and Biotechnology is often seen as an ultimate prohibition on the free reuse of saved seeds reinforcement of highly industrialized agricul- by farmers, that is, commercial policies that ture that has been the object of much criti- could strongly limit poor farming communi- cism in the past few years (Bonny, 2000a). It ties’ access to biotechnology. Furthermore, is blamed for deterioration in the quality of non-governmental organizations stress that foods, damage to the environment, an accel- genetic engineering is likely to increase the erated reduction in the number of farms, etc. risk of food dependence on major agro- This mistrust generated by the modernization exporting countries. Hence, mistrust regard- of agriculture appears in a survey carried out

Table 16.6. Trust in French public authorities to protect people in several areas, and the belief that the truth is told on related dangers (Charron et al., 2000). Respondents having Respondents thinking conﬁdence (%) the truth is told (%) No Yes No Yes (not at all More or (strongly or (not at all More or (strongly or Area or not really) less somewhat) or not really) less somewhat)

Nitrates and pesticides 53.3 30.2 13.5 55.6 26 16.1 Genetically modiﬁed crops 49.3 28.6 18.4 62.7 23 10.8 Young people’s smoking 49.0 26.8 23.4 28.3 24.5 46.5 Lake, river and sea pollution 47.4 32.7 19.2 52.7 31.1 15.8 Genetic manipulations 46.1 28.0 20.7 57.3 26.9 12.9 Radioactive waste 45.4 31.5 21.9 64.4 21.1 13.8 Chemical waste 44.8 33.1 20.4 60.8 25.7 12.2 Air pollution 43.5 34.8 21.1 45.3 31.8 22.3 Alcoholism 39.8 30.4 29.0 25.5 21.4 52.5 Nuclear plants 38.7 27.2 32.7 55.2 26.9 17.2 Food products 37.8 31.9 30.2 47.9 27.8 24.0 Chemical plants 35.0 36.4 25.2 54.4 30.6 12.3 AIDS 28.7 27.0 43.8 24.9 20.9 24.9 Road accidents 28.0 30.4 41.2 18.2 23.3 58.3 Water from the tap 26.5 32.1 40.9 36.2 29.2 33.8

BVA poll carried out on 16–31 October 2000, representative sample of 1000 people. Consumer - Chap 16 5/3/04 15:56 Page 183

Opposition to GMOs in France and Europe 183

in late 2000 and early 2001. In late 2001 For some people, especially many activists, over one-third of respondents considered that biotechnology also symbolizes the negative ’agricultural use of scientiﬁc and technological aspects of globalization and economic liberal- innovations’ is ’a bad thing for the con- ism: destruction of local cultures and sumer’, while a quarter thought that ’it is a economies, growing trend of commodifying good thing’ (Table 16.7a) (UNCAA-SIGMA, everything including genetic resources, and 2001; Union InVivo 2002). The aggravated competition often perceived as dis- Eurobarometer survey carried out in 2001 loyal due to the rivalry created between (European Commission, 2001) also shows a economies with different levels of development. relative scepticism in France about the So, certain surveys reveal that economic impacts of science and technology on agricul- motives have become a signiﬁcant cause of tural and food production. The French are opposition to GMOs – at least for some groups the most sceptical of the Europeans particu- – (Table 16.8). Arguments put forward by larly when compared with northern European active opponents show that they often perceive countries (Table 16.7b). this struggle as a form of opposition to extreme

Table 16.7a. Opinion about the agricultural use of scientiﬁc and technological progress (UNCAA- SIGMA, 2001; Union InVivo, 2001) (% of answers). ’What do you think of the agricultural use of scientiﬁc and technological progress?’ Would you say that it is… December 2000 November 2001

A good thing for the consumer? 34 25 A bad thing for the consumer? 31 35 A neither good nor bad thing? 32 32 Don’t know. 3 8

UNCAA-SIGMA-SOFRES poll carried out in late 2000 (from 27 December 2000 to 3 January 2001) and late 2001 (in November 2001); representative sample of 1000 people.

Table 16.7b. Opinion about the impacts of science and technology on agricultural and food production (European Commission, 2001): ‘Do you think it is true or false that science and technology will improve farming and food production?’ (% of answers). True False Don’t know

France 49.2 31.6 19.2 Luxembourg 51.6 28.9 19.5 Italy 52.0 23.1 25.0 Austria 53.1 25.2 21.7 Spain 55.4 21.4 23.2 Portugal 58.0 14.7 27.3 Germany (former Länder) 58.5 19.9 21.6 Belgium 58.6 23.1 18.3 European Union EU 15 59.0 20.7 20.3 Germany total 60.1 18.5 21.4 Ireland 62.4 13.8 23.8 UK 65.5 16.6 17.9 Germany (new Länder) 66.1 13.5 20.4 Greece 69.5 14.9 15.6 The Netherlands 75.7 12.0 12.4 Sweden 75.7 12.4 11.9 Denmark 78.0 12.5 9.5 Finland 78.3 9.7 11.9 Consumer - Chap 16 5/3/04 15:56 Page 184

184 Sylvie Bonny

Table 16.8. Reasons for the opposition to GMOs in 1999–2000, according to surveys aimed at explaining purchasing behaviour towards such products by consumers, in experimental economics (non- representative sample) (Rufﬁeux and Robin, 2001) (% of answers). Reasons mentioned in the questionnaire: Tend to Tend to Don’t Producers and food distributors sell products … agree disagree know

Which are above all proﬁtable for themselves 90 5 5 Which are not safe for my health 64 22 14 Which are too far removed from nature 61 22 17 Which are not environmentally sound 53 27 20 Which are not in line with my moral standards and my principles 39 42 19

economic liberalism. Militancy in this respect is Conclusion in a sense a sort of metamorphosis of anti-capi- talist militancy, or at least of protest against its It has been quite often said that the rejection of excesses. Since the collapse of the communist GMOs was the result of poor knowledge in biol- ideal has made direct opposition to capitalism ogy and that the public mainly needed to be more difficult today, it seems to have found educated. However, it is necessary to analyse new forms of expression including, in particu- the causes of GMO refusal more deeply and fur- lar, criticism of globalization, certain aspects of ther: such was the goal of this chapter, which consumption, technical developments, etc. presented various factors, actors and mecha- For the general public, GMOs are per- nisms of this movement. Opposition to GMOs ceived above all as hardly useful, non-natural stems from the many potential risks particularly and risky. This suspicion, along with limited highlighted by associations and many media, trust in the institutions and firms concerned, and from a stigmatization of their possible often leads to the suggestion that greater par- advantages. By presenting themselves as ticipation of citizens in scientific and techno- defenders of consumers’ interests and health, logical choices would be desirable and useful. the opposition rallied a substantial proportion of Some people believe that it would help to the Western public who saw no advantages in solve the current deadlock regarding accep- GMOs. GMOs thus seem to have become a tance. Others believe in the need for a symbol for many negative aspects of global eco- renewal in democracy, as this extract from an nomic development when in fact they are by no editorial vehemently illustrates: means the only forms or embodiment of that development. In this respect they differ from People no longer automatically accept that many other innovations that also strongly repre- scientific development is necessarily beneficial to humanity. Particularly because that progress has sent general economic development but the become inextricably tied up with money, advantages of which are judged more clearly hijacked by companies greedy for profit … In apparent by those who have access to them, addition, our decision-makers have developed a and which are therefore the focus of little oppo- bad habit of mortgaging our collective futures sition. The quite low trust in companies per- without first asking us, the people. The basis of forming gene technology and selling GM seeds the democratic pact has thus been altered. As a – mainly agro-chemical companies – has an result, people have become more and more important effect on the benefit and risks per- suspicious. They are increasingly unwilling to ceived and thus on the acceptance of GMOs. give the powers-that-be the authority to play On the opposite side, organizations or media with our collective futures by rubber-stamping putting forward GMOs’ risks receive more trust scientific innovations that are risky and insufficiently tested. A new spirit of distrust is and their risk alerts are therefore valued more abroad among the sorcerer’s apprentices of and listened to. neo-scientism … Shouldn’t we all have a say in A large part of denunciations and economic defining what is acceptable risk, and not just criticisms of GMOs are not in fact on specific leave it to the ‘experts’? (Ramonet, 2000). dimensions of these products but concern Consumer - Chap 16 5/3/04 15:56 Page 185

Opposition to GMOs in France and Europe 185

techno-economic aspects affecting many goods In this context a change of attitude towards and sectors. In other words, the economic criti- GMOs seems difficult to achieve at present in cism of GMOs could apply to many other prod- France and the rest of Western Europe due to ucts (which are spared the same opprobrium). the strength of the current opposition (Bonny, In fact, they should target not GMOs them- 2003). On the contrary, the opposition move- selves, but rather the context and economic ment is spreading around the world. A change conditions of their production and use. Thus, of attitude would require that GMOs no longer for example, the large concentration of firms in be considered the symbol of various unpopular large multinational groups exists in many sec- trends but rather for themselves, in relation to tors, as does the commodifying of new activi- their potential and the objectives to be set for ties; underprivileged populations in developing them. However, the history of techniques countries are exposed to and will probably con- shows that many innovations, after strong ini- tinue to be exposed to difficulties of access to tial rejection, were subsequently widely diffused many goods requiring resources to obtain them but with considerable improvement, especially or the infrastructure to produce them; patents as regards risk reduction, improved conve- have existed for a long time for many goods nience of use and usefulness. Changes in the that are sometimes vital. As for the potential general socioeconomic context as well could impact on the environment, it is considerable perhaps play an essential role by allowing for multiple human activities; moreover, biotech- GMOs to be perceived in a different light. For nology can be considered as being able to con- example, other risks, such as global change tribute to greater sustainability, and not the and its impacts, could move into the fore- opposite. Yet these issues are raised most force- ground and make biotechnology seem to be a fully for GMOs – as if they were the only subject possible solution. Transgenic plants are still in to warrant them, perhaps because of the partic- their early stages and various subsequent ular place of food and agriculture in society. developments could reduce their potential risks They therefore seem to be acting as a scape- or highlight more positive aspects of this tech- goat, in a sense. Unlike many other products nique or its products. But could this reversal with identical characteristics, GMOs are take place when some have made GMOs a accused, even when they are not really directly scapegoat that has to be eliminated because it concerned. For example, it is not genetic engi- symbolizes trends perceived as negative? neering itself that imposes patents for technical Another solution may be the development of reasons; it is current economic conditions that other applications of biotechnology and life- lead to the use of patents; however GMOS are science research. The development of biotech often accused of being responsible for the and genomics applications could lead to new patents on living organisms. Finally, GMOs are prospects for plant breeding and farming and suspected due to their very essence, and not so, perhaps, make foreign gene transfer less in relation to the way in which they are used. necessary. However that may be, the debate And yet in fact, the impact of techniques on this subject has had a considerable impact depends on the way and conditions in which by inducing many questions about technical they are used, the purpose given to them, the progress, scientific expertise, trust in private orientation of their applications, etc. (Bonny, companies and public authorities, participation 2000a,b); they are accused because they are of the public, uncertainty, acceptable risk and perceived as having little utility. choice of technological development paths.

References

ACN (American College of Nutrition) (2002) The future of food and nutrition with biotechnology. Journal of American College of Nutrition 21 (suppl. 3). AgBioWorld (2002) Scientists in support of agricultural biotechnology. Declaration in support of agricultural biotechnology signed by more than 3330 scientists around the world until early December 2001. Available at http://www.agbioworld.org/ Consumer - Chap 16 5/3/04 15:56 Page 186

186 Sylvie Bonny

ASPB (American Society of Plant Biologists) (nd) Statement on Genetic Modification of Plants Using Biotechnology. ASPB, Rockville, Maryland. Available at http://www.aspb.org/ Bonny, S. (2000a) Consumer concerns about industrialized agriculture and food safety: importance, origin and possible solutions. Annales de Zootechnie: An International Interdisciplinary Journal in General and Comparative Animal Science 49, 273–290. Bonny, S. (2000b) Will biotechnology lead to more sustainable agriculture? In: Lesser, W.H. (ed.) Transitions in Agbiotech: Economics of Strategy and Policy. University of Connecticut, Food Marketing Policy Center, Storrs, Connecticut, pp. 435–453. Available at http://agecon.lib.umn. edu/cgi-bin/pdf_view.pl?paperid=1986&ftype=.pdf Bonny S. (2003) Why are most Europeans opposed to GMOs? EJB Electronic Journal of Biotechnology 6, (1). Borlaug, N.E. (2000) Ending world hunger: the promise of biotechnology and the threat of antiscience zealotry. Plant Physiology 124, 487–490. Boy, D. (1999) Le progrès en procès. Presses de la Renaissance, Paris. Charron S., Mansoux, H. and Brenot, J. (2000) Perception des risques et de la sécurité: résultats du sondage d’octobre 2000 (Baromètre IPSN). Note SEGR 00/112. IPSN, Clamart. Consumers International (2000) Consumers, social justice and the world market. Statement from Consumers International’s 16th World Congress, Durban, November 2000. Available at http://www.consumersinternational.org/homepage.asp Conway, G. and Toenniessen, G. (1999) Feeding the world in the twenty-first century. Nature 402 (suppl. 2), C55–C58. De Cheveigné, S., Boy, D. and Galloux, J.-C. (2002) Les biotechnologies en débat: pour une démocratie scientifique. Balland, Paris. Durant, J. and Lindsey, N. (2000) The Great GM Food Debate – a Survey of Media Coverage in the First Half of 1999. Parliamentary Office of Science and Technology, Report 138, London. Available at www.parliament.uk/post/home.htm Espey, J. (1998) Socioethical Implications of Biotechnology. Industry Canada, Office of Consumer Affairs, Ottawa, Canada. European Commission (2000) Europeans and biotechnology. Eurobarometer 52(1). European Commission, Brussels. European Commission (2001) Europeans, Science and Technology. Eurobarometer 55(2). European Commission, Brussels. Frewer, L.J., Miles, S. and Marsh, R. (2002) The media and genetically modified foods: evidence in support of the social amplification of risk. Risk Analysis 22, 701–711. Gaskell, G., Allum, N., Bauer, M., Durant, J., Allansdottir, A., Bonfadelli, H., Boy, D., de Cheveigné, S., Fjaestad, B., Gutteling, J.M., Hampel, J., Jelsoe, E., Jesuino, J.C., Kohring, M., Kronberger, N., Midden, C., Nielsen, T.H., Przestalski, A., Rusanen, T., Sakellaris, G., Torgersen, H., Twardowski, T. and Wagner, W. (2000) Biotechnology and the European public. Nature Biotechnology 18, 935–938 Hoban, T.J. (1997) Consumer acceptance of biotechnology: an international perspective. Nature Biotechnology 15, 232–234. IFIC (International Food Information Council) (2002) US Consumer Attitudes Toward Food Biotechnology. Washington: IFIC, press release, September 2002. IFOP and Libération (2000) Les Français et les risques alimentaires. Libération, 3 August. Interacademies (2000) Transgenic Plants and World Agriculture. Report prepared under the auspices of seven Academies of Science (Brazil, China, India, Mexico, Third World, UK, USA). National Academic Press, Washington, DC. Ipsos-Reid (2002) Genetically modified foods and food labeling. Ipsos World Monitor, 2nd Quarter, 28–37. Available at http://www.ipsos-reid.com/index.cfm. Kassardjian, E. (2002) Appropriation de concepts en situation d’éducation non formelle, Cas d’une exposition scientifique sur les OGM. Thesis, Université Claude Bernard, Lyon, France. Marris, C., Wynne, B., Simmons, P. and Weldon, S. (2001) Public Perceptions of Agricultural Biotechnologies in Europe. Final report of the PABE Research Project commissioned by the EC. Available at http://www.pabe.net. Miller, H.I. and Conko, G. (2000) The science of biotech meets the politics of global regulation. Issues in Science and Technology 17, 47–54. Morgan, G. (1993) Risk analysis and management. Scientific American 269, 32–41. Consumer - Chap 16 5/3/04 15:56 Page 187

Opposition to GMOs in France and Europe 187

Paillotin, G. and Rousset, D. (1999) Tais-toi et mange! L’agriculteur, le scientifique et le consommateur. Bayard Centurion, Paris. Petts, J. (2002) Science, society and risk: bridging the gap? Centre for Environmental Research and Training, University of Birmingham, inaugural lecture, 30 May. Petts, J., Horlick-Jones, T. and Murdock, G. (2000) Social Amplification of Risk: The Media and the Public. Contract Research Report no. 329/2001. HSE Books, Health and Safety Executive, Sudbury, UK. Available at http://www.hse.gov.uk/research/crr_pdf/2001/crr01329.pdf. Pew (2002) Environmental Savior or Saboteur? Debating the Impacts of Genetic Engineering. Pew Initiative on Food and Biotechnology, Washington, DC. Available at http://pewagbiotech.org/ newsroom/releases/020402.php3. Powell, D. (1998) Impacts of biotechnology, environment, food safety: communications. Paper presented at the Agriculture Risk Management Conference, Hull, Canada. University of Guelph, Guelph, Canada. Ramonet, I. (2000) Les peurs de l’an 2000. Le Monde Diplomatique 561, p 1. Ruckus Society (nd) The Ruckus Society Media Manual. Ruckus Society, Berkeley, California. Available at http://www.ruckus.org/man/media_manual.html. Ruffieux, B. and Robin, S. (2001) Analyse économique de la disposition à payer des consommateurs pour des produits garantis sans utilisation d’OGM et choix du signal distinctif pertinent. Rapport du contrat ‘Pertinence et faisabilité d’une filière ‘sans OGM’’. IREPD-ENSGI, Grenoble. Salomon, J.-J. (1984) Prométhée empêtré: la résistance au changement technique. Anthropos, Paris (Pergamon, Paris, 1982). Siegrist, M. (2000) The influence of trust and perceptions of risks and benefits on the acceptance of gene technology. Risk Analysis 20, 195–203. Slovic, P. (1987) Perception of risk. Science, 236, 180–285. Slovic, P., Malmfors, T., Krewski, D., Mertz, C., Neil N. and Bartlett, S (1995) Intuitive toxicology. II. Expert and lay judgments of chemical risks in Canada. Risk Analysis 15, 661–675. SOT (Society of Toxicology) (2002) The safety of genetically modified foods produced through biotechnology. Position paper, SOT, Reston, Virginia. Springer, A., Mattas, K., Papastefanou G. and Tsioumanis, A. (2002) Comparing consumer attitudes towards genetically modified food in Europe. In: Proceedings of the Xth European Congress of Agricultural Economists (CD-ROM), Zaragoza, Spain, pp. 053–073. UNCAA-SIGMA (2001) Baromètre UNCAA/SIGMA sur les français et l’agriculture (sondage SOFRES). Media release of 10 January. UNCAA, Paris. Available at http://www.tns- sofres.com/etudes/pol/120101_agri_r.htm Union InVivo (2001) Enquête SOFRES sur les Français et l’agriculture. Press release, December, Union InVivo, Paris. Zechendorf, B. (1998) Agricultural biotechnology: why do Europeans have difficulty accepting it? AgBioForum 1, 8–13. Consumer - Chap 16 5/3/04 15:56 Page 188 Consumer - Chap 17 5/3/04 15:57 Page 189

17 Introducing Novel Protein Foods in the EU: Economic and Environmental Impacts1

Xueqin Zhu, Ekko van Ierland and Justus Wesseler Environmental Economics and Natural Resources Group, Wageningen University, Hollandsweg 1, 6706 KN, Wageningen, The Netherlands

Introduction There is extensive literature (e.g. Carlsson- Kanyama, 1998; Mattsson, 1999; Kramer, Animal protein production, in particular pork 2000) on the relationship between food con- production, has important environmental sumption and the environment. These studies impacts, and animal health problems are a focused on the environmental pressure of the major concern of consumers. These consider- different types of food, and the environmental ations may result in shifting consumers’ pref- assessment of food production and consump- erences for proteins from animal proteins to tion by life-cycle assessment or other environ- novel protein foods (NPFs). To analyse the mental assessment methods. impacts of enhancing the demand for NPFs is Some agricultural and environmental one of the purposes of the research pro- applied general equilibrium (AGE) models gramme PROFETAS2 in The Netherlands. This investigate the impacts of agricultural policies programme was initiated to study whether or and environmental policies. Examples are the not a substantial shift from animal to plant GTAP model (Hertel, 1997), the ECAM model protein foods is ‘environmentally more sus- (Folmer et al., 1995) and the MERGE model tainable, technologically feasible and socially (Manne and Richels, 1995). The GTAP model desirable’. Novel protein foods are plant pro- (global trade analysis project model) uses the tein-based food products, which are devel- general equilibrium modelling framework oped by modern technology (including combined with a huge database on inter- biotechnology) and designed on the basis of national trade. It is a very dis-aggregate model consumers’ preferences for ﬂavour and tex- and it is continuously updated by a team of ture. Examples are ‘vegetable burgers’, high- researchers. Version 5 of the GTAP model quality soybean products or protein products includes 57 sectors and 66 regions (GTAP, made of peas. Baggerman and Hamstra 2002). The ECAM model (EC agricultural (1995) suggest that NPFs can reduce environ- model) distinguishes two sectors (farm sector mental pressure because the conversion of and non-farm sector), two consumers (farmer plant proteins into meat proteins is biochemi- and non-farmer) and two commodities (agri- cally and environmentally inefﬁcient. cultural commodity and non-agricultural

1 The research is financed by the Netherlands Organization for Scientific Research (NWO). The usual disclaimer applies. 2 PROtein Foods, Environment, Technology And Society, see http://www.profetas.nl/ for details. © CAB International 2004. Consumer Acceptance of Genetically Modified Food (eds R.E. Evenson and V. Santaniello) 189 Consumer - Chap 17 5/3/04 15:57 Page 190

190 X. Zhu et al.

commodity) for each country. We mention the environment, we temporarily only consider

MERGE model (a model for evaluating regional the atmospheric emissions of CO2 as an envi- and global effects of greenhouse gas reduction ronmental indicator for several pollutants and policies), because it integrates the AGE frame- environmental effects related to the use of work with climate change and damage assess- energy in the model application. The nitrogen ment sub-models to assess climate change and phosphate emissions from the manure of policy proposals in a multi-regional context. pork production could also be included, but The purpose of this chapter is to study they are not yet considered in the application some of the potential economic and environ- because of data problems. The consumers’ mental consequences of a shift from animal concern for environmental quality is repre- protein foods to NPFs in the European Union sented by the willingness to pay for the envi- (EU). In order to investigate the consequences ronment. To be specific, by the utility elasticity of a shift from animal protein foods to NPFs, for environmental quality (ε ) in the utility func- we apply the AGE framework and include the tions. Since the value for ε is difficult to environmental aspects in the utility function of obtain, we analyse the impacts of NPFs by consumers and in the production function of means of sensitivity analysis for ε over a rela- the producers. The contribution of this chap- tively wide range of values (0.05–0.20). The ter is to present an environmental AGE model new runs for the different values of ε construct that can capture the environmental concerns different scenarios. The comparison between of the consumers, and that is applied for the results of the base run and those of the examining the impacts of the enhanced intro- scenarios provides insights into the potential duction of NPFs. economic impacts of a shift towards the con- The environmental concern of the consumption of NPFs, considering consumers’ sumers is embodied in the utility function of concern for the environment. our AGE model. The consumer’s utility The chapter is organized as follows: the depends not only on the consumption of the next section includes the theoretical back- rival goods but also on the environmental ground, and the motivation for the choice of quality, as a non-rival public good. The intro- the AGE format. It gives the specification of duction of NFPs to society is simulated in the the AGE model considering environmental model by an exogenous shift in consumer pollution, where the environment is viewed demand, i.e. by increasing the expenditure as an input of the production and the con- share of NPFs in the protein budget (δ ) to sumers have to pay for their consumption of partially replace the consumption of pork. We the environmental good. The following use an increasing expenditure share of NPFs section includes the model application and because we simulate a voluntary shift to NPFs, the sensitivity analysis of the utility elasticity which is the central hypothesis in the research for environmental quality. In this section, programme. This shift might be considered to some simulation results are presented and a be the result of consumers’ orientation to brief interpretation of the results of the model ‘green products’ and to the safety of the plant application is also given. The final section protein products. The substitution between gives the preliminary conclusions of the pork and NPFs is represented by the substitu- impacts of NPFs based on the application of tion elasticity (σ ) in the utility function. In the the model, and some discussions of the application of the model, the expenditure model. share of NPFs in the total protein budget of the consumers (δ ) is increased from 0% in the base run to 30% after the enhanced introduc- The Specification of the AGE Model tion of NPFs in the simulation run. The substitution elasticity between pork and NPFs (σ ) is The postulates of the general equilibrium chosen to be 0.8, considering the consumers’ model are that consumers maximize their util- concerns with health and the tendency to the ity subject to their budget constraint and pro- new products on the one hand, and the pre- ducers maximize their profits subject to the sent diet habits on the other hand. For the technological constraint. In an AGE model, all Consumer - Chap 17 5/3/04 15:57 Page 191

Introducing Novel Protein Foods in the EU 191

the goods, services and factors in the econ- (Perroni and Rutherford, 1996). For the omy are called commodities. Industries pro- uniqueness of the equilibrium, the Cobb- duce outputs using factor inputs and/or Douglas functional form for production and intermediate inputs. Products produced in strictly concave utility function (e.g. CES or industries can be used as intermediate inputs Cobb-Douglas utility function) can be chosen for production of other products, or con- (Ginsburgh and Keyzer, 1997). sumed by households as final goods or Based on the theoretical structure of the exported to the rest of the world. In equilib- AGE model with environmental concerns rium the demand for all commodities cannot presented in Appendix 17.I, we have speci- exceed their supply. fied the model for this study. In this section Our central research aim is to analyse the we describe the characteristics of the applied potential impacts of an increase in demand model, and specify the functional forms of for NPFs. We choose the AGE model because the model. AGE modelling is considered the best choice to anticipate adjustments in the economy and it has been used to evaluate a wide range The characteristics of the model of policy issues, including changes in direct taxes, trade policy, income redistribution and In the AGE model applied in this chapter, the public investments (Ginsburgh and Keyzer, world is divided into two regions: the EU and 1997). As the problems of agricultural the rest of the world (ROW). Thus, we have production increasingly became issues of two representative consumers, i = EU and efficiency in allocation, there was a growing ROW. The flow of the commodities in these need for such an embedding of the agri- two regions is shown in Fig. 17.1. Six prod- cultural and food sector in a wider setting (e.g. ucts are distinguished: pork, other food, non- Keyzer, 1989; GTAP, 2002). food, NPFs, peas and feed. The first four Mathematically we can represent the gen- goods – pork, other food, non-food and eral equilibrium models in several formats. NPFs – are the consumption goods of the There are five alternative formats according to consumers. Peas are both direct consumption Gunning and Keyzer (1995) and Ginsburgh goods and intermediate goods for production and Keyzer (1997): (i) excess demand format; of NPFs. Feed is an intermediate input of (ii) Negishi format; (iii) full format; (iv) open pork production. For the production of pork, economy format; and (v) CGE format. Each the factor inputs (labour, capital and land) and format is best suited for specific purposes. It intermediate input (feed) are used, while for should, however, be stressed that all formats the production of other food and feed, only describe the same model and lead to the same the factor inputs are used. NPFs are pro- equilibrium solution. For this study, we have duced by capital, labour and an intermediate chosen the Negishi format because it provides good of peas. The non-food product only a direct link to welfare analysis. It starts with a uses the factors capital and labour. Feed and welfare programme, which is subsequently peas are both produced as intermediate decentralized through commodity and agent- goods in agriculture by the factor inputs specific signals (e.g. prices). labour, capital and land. The environment is According to Ginsburgh and Keyzer specified in two ways. Firstly, the use of envi- (1997), choosing the Negishi format implies ronmental services is included as an input for that only primal forms can be used. Following production. Secondly, the utility of each con- the primal approach, we represent the pro- sumer is related to the consumption of pri- duction technology (production set) by a func- vate goods and services, and to the level of tional form with a finite number of an environmental quality indicator. Thus parameters. The parameters have to be cho- there are nine commodities (pork, other sen so that the production set satisfies the food, non-food, NPFs, peas, feed, labour, condition of strict convexity. In applied gen- capital and land) and one non-rival good eral equilibrium applications, the global prop- (expressed by an environmental quality indi- erties of functional forms become important cator) in the model. All the goods and Consumer - Chap 17 5/3/04 15:57 Page 192

192 X. Zhu et al.

Industries (production) Industries (production) in EU in ROW inputs inputs Factors Factors Products Products Intermediate Exports Intermediate Markets Markets

Imports

Households (consumption) Households (consumption) in EU in ROW

Fig. 17.1. The ﬂows of the two-region AGE model.

services can be exported or imported based The utility function in our model is a on the comparative advantages of each nested function of three levels. The substitu- region under free trade. In our application tion structure of the consumption of goods is the factors of production are immobile shown in Fig. 17.2. At Level 1, it is a Cobb- between the two regions. For simplicity, the Douglas function with substitution between model is comparatively static. the consumption of rival goods and a non- rival good (environmental quality). At Level 2, it is also a Cobb-Douglas function with substi- Objective function and utility functions tution between proteins, other food, non-food and peas for the consumption of rival goods. The objective function of the welfare pro- At Level 3, it is a CES function with substitu- gramme in the Negishi format is: tion between pork and NPFs for the con- W = Max[α •logU + α •logU ] (1) sumption of proteins. EU EU ROW ROW The demand function (Shoven and Whalley, 3 where W is the total welfare, UEU and UROW 1992) for pork and NPFs will then be : are the utility of the EU and ROW, and α EU ()1 − δ E α pr,EU and ROW are the Negishi weights of the EU C = EU,pork pppσ ⋅−[(1 δδ )()11−−σσ + () ] and ROW, respectively. A list of symbols is pork pork NPFs given in Appendix 17.II. For the equilibrium δE C = pr,EU solution of the model, the Negishi weights EU,NPFs σ ⋅−δδ()11−−σσ + () pppNPFs[(1 ) pork NPFs ] have to be found such that the budget con-

straints hold. Analytically in the sequential where CEU,NPFs is the consumption of NPFs, joint maximization (SJM) method, the CEU,pork is the consumption of pork in the Negishi weights are the respective shares in EU, σ is the elasticity of substitution total income in the economy when Cobb- between pork and NPFs, is the expendi-

Douglas utility functions and production ture share of NPFs in protein budget, Epr,EU functions are chosen (Ermoliev et al., 1996; is the expenditure of the consumers on pro- Rutherford, 1999). tein consumption in the EU (the protein

3 The demand function of pork and −NPFs are based on the CES utility function in Level 3 for the protein ()1 () −1 =−⋅ +⋅ −1 consumption: UCC(protein ) [(1 )EU,pork EU,NPFs ] . Consumer - Chap 17 5/3/04 15:57 Page 193

Introducing Novel Protein Foods in the EU 193

Utility

Consumption of rival goods Consumption of non-rival goods Level 1 (proteins, peas, other food (expressed by environmental and non-food) quality indicator)

Protein Other food Non-food Peas Level 2

Pork NPFs Level 3

Fig. 17.2. Nesting structure in utility function in the EU.

budget), ppork and pNPFs are prices of pork of the environmental services in production and NPFs, respectively. Therefore, accord- and by preferences of the consumers for the ing to the substitution effects and expendi- non-rival good ‘environmental quality’. The

ture share of the two proteins, the following utility function ui(xi,gi) is continuous, con- 4 relation exists : cave, increasing in (xi,gi) and satisﬁes: u (0,g )=0, where x is the vector of con- δ  p  σ i i = pork ⋅ sumption goods, g is the non-rival consump- CEU,NPFs  C EU,pork. (2) 1 − δ  p  NPFs tion of environmental quality and i is the The protein consumption in the EU C consumer. This results in the following utility EU,pr functions with C as the consumption of and in the ROW C is as follows: s ROW,pr rival good s, s = proteins (pork + NPFs), CEU,pr = CEU,pork + CEU,NPFs (3) other food, non-food and peas, and g as the C = C (4) non-rival consumption of an environmental ROW,pr ROW,pork good (expressed by the environmental qual- where CROW,pork is the pork consumption in ity indicator): the ROW. −− β ==i 11i isi∏ i For the environment use, we consider UgfCi i(( si )) g i ( C si ) s (5) the simple case in which environmental services are used as input in the production where i indicates the consumer (i = EU and process. The use of environmental inputs ROW), is the utility elasticity for environ- β decreases the utility of the consumers by mental quality, s are the utility elasticities for reducing environmental quality that we consumption of rival goods without consider- express in the model by means of an envi- ing the environment, and Σβ = 1. The utility s si ronmental quality indicator. In this manner functions used in the applied model are given environmental quality is affected by the use in Appendix 17.III.

4 If σ =1.0, this relationship between pork demand and NPFs demand does not hold any more. Then the consumption of both goods is dependent on the protein balance function and the utility function. The consumer will only consume the cheaper one. Consumer - Chap 17 5/3/04 15:57 Page 194

194 X. Zhu et al.

Production functions production function indicating the cost share of the emission EM for production, 0 < ξ < 1, A production function describes the technical LB refers to labour input, LD land input, K relationship between the inputs and outputs of capital input, FD the feed input and PI the a production process represented by a mathe- pea input. One can consider EM as the use of matical function. The production of pork, or ‘environmental services’, which reflects that animal protein products (processed pork), and the firm must release its emissions to the envi- NPFs can be described by the production ronment. We can think of the firm as requir- chains because the agriculture process is very ing EM emission permits in order to produce different from the industrial production. The (Copeland and Taylor, 2003). Therefore when two representative chains are shown in Fig. environmental services are treated in the pro- 17.3a and b. duction function in this way, an emission per- Along the chains, many inputs and outputs mit system reflecting the annual endowment (including the environmental emissions) are of environmental services for each region is involved. It is impossible to include all the necessary for the modelling. Thus the follow- inputs along the chains in the production ing relationship holds function of the pork and NPFs production, ≤ ∑ EMij EM i (6) and simplification is necessary. As we have j noticed, the production processes not only where EM is the use of environmental ser- use production factors as inputs but also gen- ij –––– erate the emissions from production. For vices in region i for good j, and EMi is the technical reasons pollution in our model is number of emission permits in region i. not viewed as a negative externality but as the The production function for good j is then: η ξ η η η 4j η 1−ξj use of a natural resource. The production Y= EMi[( LB )1j ( LD2j ( K )3j () FD ( PI )5j ] (7) inputs of pork include labour, capital, land, jijjjjj η η η η η the intermediate good ‘feed’ and an environ- where 1, 2, 3, 4 and 5 is the cost share mental input (e.g. emission). For the produc- of each input (LB, LD, K, FD, PI) for production of all the goods, an environmental input tion without considering the cost of emissions, η η η η η is also used. The Cobb-Douglas production with 1+ 2+ 3+ 4+ 5=1. function for production of good j with envi- For the parameters of the production ronmental input can in general be presented functions, we use information from other as follows: studies. For example, the feed costs amount ξ ξ to 60% of the total production costs in the Y = EM j F(LB ,LD ,K ,FD ,PI )1– j j j j j j j j Netherlands (Jogeneel, 2000). For the EU an where j is the production good ( j = pork, average of 45% of the feed costs is used in NPFs, other food, non-food, peas and feed), the pork production function. The technolog- Y is the production, EM is the environmental ical parameters in the production functions of input, ξ is the exponent of the emission in the the EU and the ROW are 1.0 and 0.6

(a)

Feed Pig Meat Crop Slaughtering Distributing Consumer industry farming processing

(b) NPFs Peas Distributing Consumer processing

Fig. 17.3. Production and consumption chains of pork (a) and novel protein foods (b). Consumer - Chap 17 5/3/04 15:57 Page 195

Introducing Novel Protein Foods in the EU 195

respectively5. The production functions in model version we approximate this relation by this manner grosso modo reflect the produc- means of a linear function in the use of the tion technology for the region that we distin- environmental services: n guish in our study. The production functions =−Ψφ gEMii∑ ij ()i (8) and balance equations are reported in j=1 Appendix 17.III. n where Ψ is the intercept and ΣEM is the total i j=1 ij emissions of all the producers in region i. This relationship shows that the higher the emis- Environmental quality sions the lower the environmental quality. Since consumers will enjoy and pay for this The balance equation for environmental goods environmental quality, it can be seen as a prod- (e.g. clean air), which are inputs to the produc- uct produced by a certain environmental sector. tion process, is assumed to be determined by the initial stock and production inputs as shown in equation (A2) of Appendix 17.I. But the ini- Budget constraints tial stock of ‘environmental services’ is hard to know and the link between output and the use Under constant return to scale, profits are of environmental goods in the production zero so that income is the value of initial process is hard to establish. With the emission endowments, which are employed in the pro- permit system, we established the relationship duction. According to the endowments of pro- in equation (6) for producers. For consumers, duction factors and emission permits the the environment is valued in terms of environ- income is: mental service which is constrained by equation (A2)’ of Appendix 17.I. For our applied model, hi = rli LDi + wi LBi + rki Ki + pmi EMi we only consider a one-dimensional (e = 1) where rl is the price of land, w is the wage, rk environmental service g, which reflects a num- is the price of capital and pm is the price of ber of environmental issues that are related to emission permit. It should be equal to the the energy use and the release of pollutants like total revenue of the production sectors and NOx and SOx or greenhouse gases. As a proxy the ‘environmental sector’: for energy use and related emissions we use hpYg=⋅+∑ φ the level of CO2 emissions in the respective ijij i i (9) regions. Then we can define the ‘environmen- j

tal quality indicator’ to be determined by the where pj is the price scalar of good j, the ﬁrst level of emissions. If the emissions are above a item of the right-hand side Σp ·Y is the rev- j j ij critical level, the environmental quality indicator enue of the production sectors, and the sec- φ will decrease. We next use the environmental ond item igi is the revenue of the quality indicator as the non-rival consumption ‘environmental sector’ which maintains cer- of environmental goods in the utility function. tain environmental quality demanded by the Of course, the model can be easily expanded to consumer. include more dimensions of environmental Budget constraints say that the expenditure goods g, by explicitly modelling emissions of of the consumer should be equal to his/her nitrogen oxides and other pollutants as long as income. Now that the non-rival environmental the data are available. quality is one of consumption goods, the con- Obviously, the environmental quality that sumer has to pay for his/her consumption. consumers face in region i is determined by Just like the producer has to pay for the emis- the total use of the environmental services of sion permit for production, the consumers all the producers in region i. In the present who simply enjoy the presence of the

5 These technological parameters are chosen to the best of our knowledge but require further research. The model speciﬁcation in GAMS (General Algebraic Modeling System) is available on request from the authors and the impact of different parameter values can be easily established. Consumer - Chap 17 5/3/04 15:57 Page 196

196 X. Zhu et al.

resource or environmental quality should pay interesting issue. We carry out the sensitivity to the ‘environmental sector’ for the environ- analysis for the value of parameter ε. The val- mental services. The budget constraint of the ues of 0.1 and 0.2 for ε are simulated under consumer now looks like: two cases of four runs where (i) different values ⋅+φ = for the EU and ROW and (ii) similar values for ∑ pCs si i g i h i (10) s the EU and ROW are used, respectively. The results of all these scenarios are compared with where ps is the price scalar of good s, s = proteins (pork + NPFs), other food non-food, the results of the base run. The comparison and peas, C is the consumption of good s in gives an impression of some potential impacts Σ si of the enhanced introduction of the NPFs in region i, ps·Csi is the total expenditure on s φ the EU on the economy and the environment. the consumption of all rival goods and igi is the payment by the consumers for the environmental quality g, and h is income. The data In this welfare programme, where both the consumers and producers have to pay for the As stated, the model is applied to the econ- environmental use, the Lindahl equilibrium is omy with two regions: the EU and ROW. The reached (Ginsburgh and Keyzer, 1997). data for labour, land and capital are based on the database of the FAO (2002) and the Penn World Table (Penn World, 2002). The labour The Model Application and Results force in 1998 in the EU is 252 million and 3323 million in the ROW. The total land area The data and scenarios in 1998 in the EU is 313,000 ha and 12,149,000 ha in the ROW. Non-residential The base run and scenarios capital stock per worker in the EU is approximately €30,000 per worker and €5000 per We have applied the model to develop the worker in the ROW according to the Penn base run, a scenario for the enhanced con- World Table. The total capital stock in the EU sumption for NPFs and some scenarios of is €7560 billion and in the ROW it is €16,615 sensitivity analysis. billion. The data for emissions is based on the little Green Data Book (World Bank, 2000). BASE RUN There are no NPFs, the environ- The EU contributes about 12% of the global mental concern is indicated by the utility elas- CO emissions (3000 million tonnes in the ε 2 ticity for environmental quality , which is EU and 22,000 million tonnes in the ROW in assumed to be 0.05 for both regions. 1998). As we have already mentioned, emission permits should be given when the emis- NPF SCENARIO For the simulation of the sions are taken as an input for the production new scenarios, we assume that the substitu- function. In the model run, we initially allocate σ tion elasticity of the NPFs for pork is = 0.8 emission permits to the EU and ROW accord- and we simulate a situation where the expen- ing to the emission levels of 1998. The initial diture share of NPFs in the protein budget is endowments are shown in Table 17.1. These δ increased to 30% ( = 0.3) after enhanced data are used for the model applications. introduction of NPFs. We do not assume NPFs as perfect substitutes of pork (σ =1) because we think in the short run it is impossi- The results ble to replace all the animal protein products by NPFs. In this scenario, we use the same The results for the base run value for the utility elasticity for the environment ( = 0.05) as in the base run. When there are no NPFs, and = 0.05, we run the model as the base case. The results SENSITIVITY ANALYSIS As a consumer-driving for the ‘base run’ are reported in Table 17.2. economy, the sensitivity of the results to the Firstly, for production the table shows that the parameters in the utility functions is a very EU is basically the major producer of pork Consumer - Chap 17 5/3/04 15:57 Page 197

Introducing Novel Protein Foods in the EU 197

Table 17.1. Factor endowments of labour, land, capital and CO2 emission permits. Labour Land Capital Emissions (millions) (ha × 1000) (billion €) (million tonnes)

EU 252 313 7,560 3,000 ROW 3,323 12,149 16,615 22,000

Table 17.2. Baseline: production, consumption, trade, emissions and income.

Production Consumption

Other Other Pork food Non-food Peas Feed Pork food Non-food Peas Feed

EU 304 0 1283 0 0 94 668 1218 3 301 ROW 39 2422 3163 43 340 249 1754 3229 40 39 Total 343 2422 4446 43 340 343 2422 4447 43 340 Income per Utility Trade (+, export; , import) Emissions worker (welfare)

EU +210 669 + 3 301 1162 12.4 779 ROW 210 +669 +3 +301 8188 2.5 2140 Total 0 0 0 0 0 9350 (7.39)

and non-food. It exports pork and non-food Comparing Tables 17.2 and 3, we observe to the rest of the world and imports other the implications of the enhanced introduction food, peas and feed from the rest of the of NPFs in the EU to the economy. The bud- world. Secondly, for the use of environmental get share of 30% for NPFs results in a reduc- services, the entry ‘emissions’ in Table 17.2 tion of consumption of pork in the EU by shows that the EU emits 12% of the global 28%. Pork production in the EU will be emissions, which is consistent with the decreased by 5% (15 units) from 304 to 289 endowment of environmental services that we units. The reduction in consumption of pork used. Pork is, in our analysis, the most pollut- is more than the decrease of the pork producing product with the highest environment tion in the EU because the EU will beneﬁt input in the production function. Pork is more from exporting pork to the ROW. The inter- expensive, because its production needs more national trade of pork is increased by 5% factor inputs, including feed as an extra inter- from 210 to 221 units. mediate input. Finally, Table 17.2 shows that Since the production of NPFs is less pollut- income per worker in the EU is ﬁve times ing than that of pork production, the total higher than the rest of the world. emissions will decrease by about 0.8% in the EU. As for the ROW, the emissions are decreased by 0.2% because they produce less The results for the NPF scenario pork by importing from the EU. The total By introducing an exogenous increase in the emissions are reduced by about 0.2% because consumption of NPFs in the EU by increasing the emissions of the EU are much less than the expenditure share of NPFs in protein bud- those of the ROW. get, with the same environmental concern in For income related to the remuneration of the two regions as the base run (ε = 0.05), a factors, we observe that income for the EU new equilibrium will be reached. The results falls slightly because the production of NPFs are reported in Table 17.3. needs simpler processing than pork and thus Consumer - Chap 17 5/3/04 15:57 Page 198

198 X. Zhu et al.

Table 17.3. NPFs scenario: production, consumption, trade, emissions and income. Production Consumption Other Non- Other Non- Peas + Feed Pork NPFs food food Peas Feed Pork NPFs food food input input

EU 289 52 0 1281 0 0 68 52 662 1205 3 + 66 284 ROW 30 0 2421 3164 109 313 251 0 1759 3240 40 29 Total 319 52 2421 4445 109 313 319 52 2421 4445 109 313 Income Utility Trade (+, export; , import) Emissions per worker (welfare)

EU +221 0 662 +77 69 283 1153 12 794 ROW 221 0 +662 77 +69 +283 8170 2.5 2148 Total 0 0 0 0 0 0 9323 (7.4)

less primary input. Therefore the factors are for the utility elasticity for the environment is less in demand than before the enhanced used for both regions. This means that the introduction of NPFs, and prices of factors are consumers are willing to pay 5% of their lower. Given the fixed volume of factors, the expenditure for a good environment. But in remuneration will be lower. reality different people have different willing- The utility is increased slightly because in ness to pay for the non-rival consumption of our model the utility depends on both the environmental goods. Therefore, it will be consumption of rival goods and the environ- interesting to see how the attitude of the con- mental quality indicator. The environmental sumers will influence their consumption bun- quality indicator is linear and declining in the dle. In the first case, we consider the different level of emissions. The consumers have to environmental concerns in different regions. make a trade-off between more consumption The market for environmentally friendly of the rival goods with lower environmental goods is located mainly in the member coun- quality and better environmental quality with tries of the OECD, where during the last few less consumption. More consumption of the years consumers have started to articulate rival goods means more pollution but more strong environmental concerns. These con- pollution implies lower environmental quality. cerns have been translated into both individual In this manner the preference of consumers purchasing decisions and government regula- for environmental quality gives feedback to tions (Bharucha, 1997). In the second case, consumption of rival and non-rival goods and we will increase the value of ε from 0.05 to then to production. 0.1 and 0.2 for both regions. Therefore, we will carry out sensitivity analysis under these two cases of the four runs which are shown in Sensitivity analysis Table 17.4. As the preferences of consumers for environmental quality will have feedback on production and consumption in a competitive model, Table 17.4. The runs for sensitivity analysis for . the interesting question is how the consumers Model runs Values of value this environmental quality. We carried out some sensitivity analysis for the valuation Case 1 of the consumer for the environmental qual- Run 1 EU = 0.1, ROW = 0.05 Run 2 = 0.2, = 0.05 ity, because little information is available on EU ROW the role of the environment in the utility func- Case 2 Run 3 = 0.1, = 0.1 tion of the consumers. In the above two appli- EU ROW Run 4 = 0.2, = 0.2 cations of the model, a modest value of 0.05 EU ROW Consumer - Chap 17 5/3/04 15:57 Page 199

Introducing Novel Protein Foods in the EU 199

In Case 1, we ﬁx the value of the utility elas- analysis. The model considers both the utility ticity for the environment ε in the ROW at from the consumption of goods and the disutil- 0.05, and increase the value for the EU from ity from environmental pollution. The emis- 0.05 to 0.1 and 0.2. See Tables 17.A1 and sions from production give feedback on utility 17.A2 in Appendix 17.IV for the detailed and on the bundle of rival and non-rival con- results. With the increased value for the EU, sumption, and then indirectly on production. pork production in the EU will decrease. If the For a value of 0.05 for the utility elasticity for value increases to 0.2, pork production in the the environment, the enhanced introduction of EU will be hampered severely and at the given NPFs decreases the emissions from pork pro- technology as described by the present produc- duction in the EU and decreases the total tion functions will eventually disappear, because emissions slightly. The EU will consume less pork is the most polluting product. The price of pork by consuming some NPFs and will export pork decreases because it is demanded less. more pork than before. Moreover, pork pro- Non-food production will increase in the EU duction in the ROW will decrease slightly because the EU has a comparative advantage because slightly more pork can be imported and it is less polluting than the other products. from the EU. Thus, the introduction of NPFs As a result, the export of non-food to the ROW decreases in this setting the emissions in the will increase and the price of non-food falls ROW slightly. As a result, the total emissions because more production takes place. The in the world will decrease slightly too. emissions in the EU decrease with the increase Nevertheless, the model results are sensi- of consumer’s valuation of the environmental tive to the value of the utility elasticity for the goods in the EU, because the EU switches to environment. If the EU has a higher utility produce more non-food and less pork. In con- elasticity for the environment than the ROW trast, pork production in the ROW will increase (0.1 versus 0.05) pork production in the EU as a result of the increase in the value of the will decrease more strongly and the export of environmental goods in the EU, because the pork to the rest of the world will decrease. As EU will reduce production and export of pork. a result, the rest of the world has to increase Since pork and non-food become cheaper with pork production for their high demand for the increase of ε, the ROW is also better off. pork. The emissions in the ROW will increase The emissions in the ROW increase, however, by 1.7% from 8188 to 8328 units. If the util- because the ROW has to produce the polluting ity elasticity for the environment in the EU product ‘pork’ for its own consumption and increases to 0.2, then a stronger trend will export to the EU. occur. The EU will stop producing pork and In Case 2, we have increased the value of will import some pork from the ROW. Then ε for both regions from 0.05 to 0.1 and 0.2. the emissions in the ROW will increase by The simulation results are reported in Tables 2.7% (to 8413 units). To summarize, if only 17.A3 and 17.A4 in Appendix 17.IV. The consumers in the EU increase their environ- results show that pork production for both mental concern, the introduction of NPFs regions decreases and the emissions decrease does not reduce the emissions in the ROW. greatly. But switching to produce more NPFs and less pork in the EU is helpful to reduce the unevenness of the income distribution by Conclusions and Discussion improving the income share of the ROW. If the two regions have the same concern In this chapter we have sketched some impor- for the environment, the increase in the value tant aspects and possible implications of an of ε will limit pork production in both regions enhanced demand for NPFs, by means of an and limit the emissions globally. The model AGE model. Although we are aware that the strongly suggests that the enhanced introduc- model is far from perfect and that it is formal- tion of NPFs is meaningful for global environ- ized at a high level of aggregation, we think it mental improvement by emission reduction, is worthwhile to discuss some of the character- only if both regions increase their preferences istics, the assumptions and the results of the for environmental quality. Consumer - Chap 17 5/3/04 15:57 Page 200

200 X. Zhu et al.

This chapter presents an AGE model that the role of biotechnology in the economy captures the environmental concerns in the and for studies on the environmental con- utility function. The model presented in this cerns of consumers. The inclusion of agricul- chapter shows how the economy can be tural elements, like land use, water use and modelled by general equilibrium modelling agricultural chemicals use, effects of the when facing some changes in preferences. common agricultural policy (CAP) and other Despite its simplicity, it illustrates some of environmental issues (such as environmental the most important fundamental environ- policy measures) are important aspects for mental economic mechanisms that might expansion and application of the model. At occur as a result of the enhanced introduc- the theoretical level, embodying the dynamic tion of NPFs based on the classiﬁcation of properties of the environment and introduc- the goods and their production functions of ing explicit environmental feedback on pro- our model. The model provides a useful duction and consumption in the AGE model framework for further empirical studies on is an interesting challenge.

References

Baggerman, T. and Hamstra, A. (1995) Motives and perspectives from consumption of NPFs instead of meat [Motieven en perspectieven voor het eten van NPFs in plaats van vless]. DTO-werkdocument VN9, DTO, Delft. Bharucha, V. (1997) The impact of environmental standards and regulations set in foreign markets on India’s export. In: Jha, V., Hewison, G. and Underhill, M. (eds) Trade, Environment and Sustainable Development A South Asian Perspective. Macmillan Press, Basingstoke, pp. 123–124. Carlsson-Kanyama, A. (1998) Climate change and dietary choices – how can emissions of greenhouse gases from food consumption be reduced? Food Policy 23, 277–293. Copeland, B.R. and Taylor, M.S. (2003) International Trade and the Environment: Theory, Evidence and Policy. Princeton University Press, Princeton, New Jersey. Ermoliev, Y., Fischer, G., and Norkin, V. (1996) Convergence of the sequential joint maximization method for the applied equilibrium problems. Working paper. IIASA (International Institute for Applid Systems Analysis), Laxenburg. FAO (2002) Land and Population from Data Collections. Available at http://www.fao.org/. Folmer, C., Keyzer, M.A. and Merbis, M.A. (1995) The Common Agricultural Policy beyond the MacSharry Reform. Elsevier Science B.V., Amsterdam. Ginsburgh, V. and Keyzer, M.A. (1997) The Structure of Applied General Equilibrium Models. The MIT Press, London. GTAP (2002) GTAP 5 documentation. Available at http://www.gtap.agecon.purdue.edu/ Gunning, J.W. and Keyzer, M.A. (1995) Applied general equilibrium models for policy analysis. In: Behrman, J. and Srinivasan, T.N. (eds) Handbook of Development Economics. Elsevier Science B.V., Amsteram, pp. 2025–2107. Hertel, T.W. (1997) Global Trade Analysis: Modeling and Applications. Cambridge University Press, Cambridge. Jongeneel, R.A. (2000) EU’s Grains, Oilseeds, Livestock and Feed Related Market Complex: Welfare Measurement, Modelling and Policy Analysis. Social Science Department, Wageningen, Wageningen University. Keyzer, M.A. (1989) Some views on agricultural sector modeling. In: Bauer, S. and Henrichemeyer, W. (eds) Agricultural Sector Modeling: Proceedings of the 16th Symposium of the European Association of Agricultural Economists (EAAE). Wissenschaftsverlag Vauk Kiel KG, Kiel, pp. 23–30. Kramer, K.J. (2000) Food Matters: On Reducing Energy Use and Greenhouse Gas Emissions from Household Food Consumption. Groningen University, Groningen. Manne, A. and Richels, R. (1995) MERGE: a model for evaluating regional and global effects of EGH reduction policies. Energy Policy 23, 17–34. Mattsson, B. (1999) Environmental Life Cycle Assessment (LCA) of Agricultural Food Production. Department of Agricultural Engineering, Swedish University of Agricultural Sciences, Alnarp. Consumer - Chap 17 5/3/04 15:57 Page 201

Introducing Novel Protein Foods in the EU 201

Penn World (2002) Penn World Table 5.6. Available at http://pwt.econ.upenn.edu/ Perroni, C. and Rutherford, T.F. (1996) A comparison of the performance of ﬂexible functional forms for use in applied general equilibrium analysis Available at http://www.gams.com/solvers/mpsge/ domain.htm Rutherford, T.F. (1999) Sequential joint maximization. In: Weyant, J. (ed.) Energy and Environmental Policy Modeling. Kluwer Academic Publishers, Dordrecht. Shoven, J.B. and Whalley, J. (1992) Applying General Equilibrium. Cambridge University Press, Cambridge. Consumer - Chap 17 8/3/04 11:04 Page 202

202 X. Zhu et al.

Appendix 17.I. The Theoretical AGE dimensions, yj is the vector of the net output Model with Environmental Concerns – of a producer j with k + e dimension if each ω The Negishi Format producer produces only one good, and i is the vector of initial endowments (including Consider an economy consisting of m con- the environmental goods) of consumer i with sumers, indexed by i, i = 1, 2, …, m and n k + e dimensions. Positive yj indicates the producers, indexed by j, j = 1, 2, …, n. There output of a production process and negative are r commodities (goods and factors), yj indicates the input of the production indexed by k, k = 1, 2, …, r. Environmental process. A vector of Lagrange multipliers goods indexed by g (g = 1, 2, …, e) are associated with the balance constraints, i.e. a involved in the economy for consumption and vector of the shadow prices of each commod- production. The welfare programme in the ity or environmental good is indicated by p in Negishi format, which allocates the resources the bracket. The commodity can be a final in the economy optimally (Ginsburgh and product, a production factor or an intermedi- Keyzer, 1997), is as follows6 ate good. This equation states that the consumption of a commodity or environmental Max ∑ au(,) x g ≥∀ iii i i (A1) xgii,,,,0 y j ij good must be smaller than or equal to its pro- subject to the balances of rival commodities duction plus its initial endowments. and environmental goods: Equation (A2) is the balance equation of non-rival environmental consumption goods, ∑∑∑+≤ +ω iixxg jj y i i () p (A2) where a consumer’s individual consumption should not exceed the common consumption g ≤ x .(φ ) (A2) i g i of all the consumers. This also makes it possi- ∈ Production technology: yj Yj (A3) ble to obtain explicit Lagrange multipliers for the value that each consumer attributes to the With welfare weights ai, such that environmental consumption xg. The vector of px+=+φωθλ g p∑ ∏ ()( p ) (A4) iii i jijj i the Lagrange multiplier fi in the bracket with e and dimension is the price vector that each consumer has to pay for the consumption of envi- 1 a = . (A5) i λ ronmental goods. i Equation (A3) shows that the production In this model, equation (A1) is the objec- plan must belong to some feasible set, or is tive function of the model, where u is the i constrained by the production technology. Yj utility function of each individual i (i = 1, 2, is the production set of firm j reflecting its fea- … , m), x is the vector of consumption sible technology. goods with k dimension and g is the vector Equation (A4) states that the expenditure of consumption of non-rival environmental of the consumer must be equal to his/her goods with e dimension. The objective of income, where the left-hand side shows the this welfare programme is to maximize the total expenditure and the right-hand side total welfare, which is a weighted sum of the shows the income of the consumer. The total utility of all the m consumers in the econ- expenditure includes the total expenditure on omy, the Negishi weight of consumer i is the consumption of all rival goods, pxi, and given by α . φ i the payment for the environment, igi. The The equations in (A2) are the balance income of consumer i includes the value of ω equations for each commodity k (k = 1, …, r) his/her initial endowments p i and his/her and each environmental good g (g = 1, 2, …, total profit, received from firm j (j = 1,2, …, θ e). In this equation, xg is the vector of con- n). ij is the profit share of consumer i in firm Π sumption of environmental goods with e j, j(p) is the profit of firm (producer) j.

6 In this appendix we follow the original notation of Ginsburgh and Keyzer (1997). Consumer - Chap 17 5/3/04 15:57 Page 203

Introducing Novel Protein Foods in the EU 203

Equation (A5) shows how welfare weights welfare weight that is attributed to consumer i are related to the budget constraints in this such that the equilibrium of the economy welfare programme. The Lagrange multiplier exists. The allocation resulting from the equa- associated with the budget constraint of con- tion system from equations (A1) to (A5) is λ sumer i is indicated by i, its inverse is the called the Lindahl equilibrium. Consumer - Chap 17 5/3/04 15:57 Page 204

204 X. Zhu et al.

Appendix 17.II. A List of the Symbols ξ = cost share of the emission input in the production Variables Π =proﬁt σ = substitution elasticity C = consumption ψ = environmental standard E=expenditure ω = vector of initial endowments. EM = emission –— EM = emission permits (exogenous) FD = feed input in the production of Shadow prices pork h = income λ = Lagrange multipliers associated g = vector of consumption of environ- with the budget constraint of the mental goods in general model, or consumer environmental quality indicator in φ = shadow price of the environmental applied model goods K = capital p = shadow price vector of commodi- — K = capital endowment (exogenous) ties LB = labour p = shadow price (scalar) of production –— j LB = labour endowment (exogenous) good j LD = land p = shadow price (scalar) of consump- –— s LD = land endowment (exogenous) tion good s

PI = peas input for the production of pm = shadow price of emission permit NPFs rk = shadow prices of capital u or U = utility of the consumer rl = shadow prices of land W = total welfare (Negishi welfare) w = shadow prices of labour. X = net export x = vector of consumption goods Y = production sets or production Subscripts quantity y = vector of net production of goods. g=environmental goods, g =1,2, …, e i=consumers, i = 1, 2, …, m for theoretical model, and i = EU and Parameters ROW for applied model j = goods or products, j =1, 2, …, n α = welfare weights for general model, and j = pork, β = parameter in the utility function other food, non-food, NPFs, peas γ = parameters in Cobb-Douglas pro- and feed for the applied model duction function k = commodities, k =1, 2, …, r for δ = expenditure share of NPFs in pro- general model tein budget s=consumption goods in applied = utility elasticity for the environment model, s = proteins (pork + NPFs), (in utility function) peas, other foods and non-food η = parameter in the utility function EU = the European Union θ = proﬁt share ROW = the rest of the world. Consumer - Chap 17 5/3/04 15:57 Page 205

Introducing Novel Protein Foods in the EU 205

Appendix 17.III. Utility and Production =⋅⋅003..0.4 03 YEMLBKROW,feed06.[ ROW,feed ROW,feed ROW,feed Functions, and Balance Equations of the 03.. 097 LDROW,feed ] Applied Model =⋅⋅003....0.4 03 03 097 YEMLBKLDEU,peas EU,peas[] EU,peas EU,peas EU,peas Utility functions =⋅⋅003..0.4 03 YEMLBKROW,peas06.[ ROW,peas ROW,peas ROW,peas =⋅⋅⋅⋅0... 05 0 12 0 299 0 . 001 0 .. 58 0 95 UgEU() EU ( CCC EU,pr EU,otf EU,peas C EU,nf ) 03.. 097 LDROW,peas ]. =⋅⋅⋅⋅0.. 05 0 12 0 . 295 0 . 05 UgROW()( ROW C ROW,pork CC ROW,otf ROW,peas 058.. 095 CROW,nf ) NPFs in EU =⋅⋅0.. 0150.1 0 2 YEMLBKEU,NPFs EU,NPFs[ EU,NPFs EU,NPFs LD07..]. 0985 Production functions EU,NPFs

Here Y indicates the production quantity, LB the labour input, LD the land input, K the Balance equations capital input, FD the feed input and EM the emission input. Feed balance

The production of feed crops Yfeed is equal to Pork the intermediate input for pork production =⋅⋅⋅005...0.2 015 020 FD plus its net export X . YEMLBLDKEU,pork EU,pork[ EU,pork EU,pork EU,pork pork feed 045.. 095 FDEU,pork ] Yfeed = FDPork + Xfeed

=⋅005. 0.2 YEMLBROW,pork06.[ ROW,pork ROW,pork 015..⋅⋅ 020 045 .. 095 Balance of peas LDROW,pork K ROW,pork FD ROW,pork ]. Peas are produced for direct use and production of NPFs, and NPFs are only produced in Other food the EU7 =⋅⋅004....0.30 035 035 096 YEMLBLDKEU,otf EU,otf[] EU,otf EU,otf EU,otf YEU,peas = CEU,peas + PIEU,peas + XEU,peas =⋅⋅004..0.3 04 YEMLBLDROW,otf06.[ ROW,otf ROW,otf ROW,otf YROW,peas = CROW,peas + XROW,peas 03.. 096 KROW,otf ]. where C is the direct consumption of peas, PI is the intermediate input of peas for pro- Non-food duction of NPFs and X is the net export of peas. YEMLBK=⋅002...[]0.45 055 098 EU,nf EU,nf EU,nf EU,nf YEMLBK=⋅06.[002...0.45 055 ]. 098 ROW,nf ROW,nf ROW,nf ROW,nf Balance of pork, other food and non-food

The production of a good in one region Yij Feed or peas equals the consumption of a good Cij plus its net export X . Here feed is the yield of feed crops. The fol- ij lowing production functions are used for feed Yij = Cij + Xij crops and peas: where j = pork, other food, non-food, but j ≠ YEMLBKLD=⋅⋅003....[]0.3 020 04 097 EU,feed EU,ROW EU,feed EU,feed EU,feed feed, peas.

7 We assume that NPFs are particularly developed in the European market and that in the short run they will mainly be produced within Europe. Consumer - Chap 17 5/3/04 15:57 Page 206

206 X. Zhu et al.

Balance of NPFs Balance of factors

The production of the NPFs equals its con- ≤ ∑ LBij LBi sumption. j ∑ LD≤ LD YEU,NPFs = CEU,NPFS. ij i j ≤ ∑ KKij i Trade balance j = where j includes pork, other food, non-food, ∑ Xij 0, for j = pork, other food, non-food, i feed, peas and NPFs, LB is the labour usage, feed and peas, but j ≠ NPFs. LD is the land usage and K is the capital ––– ––– –– usage for production. LBi, LDi and Ki are the factor endowments of region i. Consumer - Chap 17 5/3/04 15:57 Page 207

Introducing Novel Protein Foods in the EU 207

Appendix IV. Results for Sensitivity Analysis

Table 17.A1. Production, consumption, trade, emissions and income ( EU = 0.1, ROW = 0.05). Production Consumption Other Non- Other Non- Peas + Feed Pork NPFs food food Peas Feed Pork NPFs food food input input

EU 223 49 0 1334 0 0 64 49 619 1133 3 + 62 207 ROW 95 0 2400 3103 105 301 254 0 1781 3304 40 + 0 94 Total 318 0 2400 4437 105 301 318 49 2400 4437 105 301 Income Utility Trade Emissions per worker (welfare)

EU 159 0 618 201 65 207 669 10.7 802 ROW 159 0 618 201 65 207 8328 2.4 2181 Total 0 0 0 0 0 0 8997 (7.43)

Table 17.A2. Production, consumption, trade, emissions and income ( EU = 0.2, ROW = 0.05). Production Consumption Other Non- Other Non- Peas + Feed Pork NPFs food food Peas Feed Pork NPFs food food input input EU 0 35 0 1410 0 180 46 35 445 827 2 + 45 0 ROW 319 0 2354 3001 90 138 273 0 1909 3584 43 + 0 318 Total 319 35 2354 4411 90 318 319 35 2354 4411 90 318 Income Utility Trade Emissions per worker (welfare)

EU 46 0 445 583 47 180 311 7.0 719 ROW 46 0 445 583 47 180 8413 2.3 2345 Total 0 0 0 0 0 0 8724 (7.54)

Table 17.A3. Production, consumption, trade, emissions and income ( EU = 0.1, ROW = 0.1). Production Consumption Other Non- Other Non- Peas + Feed Pork NPFs food food Peas Feed Pork NPFs food food input input EU 280 51 0 1268 0 0 66 51 648 1193 3 + 65 280 ROW 28 0 2371 3131 107 308 242 0 1723 3206 39 + 0 28 Total 308 51 2371 4399 107 308 308 51 2371 4399 107 308

Income Utility Trade Emissions per worker (welfare)

EU 214 0 648 75 68 280 702 12 837 ROW 214 0 648 75 68 280 4883 2.5 2385 Total 0 0 0 0 0 0 5585 (7.50) Consumer - Chap 17 5/3/04 15:57 Page 208

208 X. Zhu et al.

Table 17.A4. Production, consumption, trade, emissions and income ( EU = 0.2, ROW = 0.2). Production Consumption Other Non- Other Non- Peas + Feed Pork NPFs food food Peas Feed Pork NPFs food food input input EU 369 50 0 1250 0 0 63 50 631 1177 3 + 64 274 ROW 27 0 2307 3088 105 302 233 0 1676 3161 38 + 0 28 Total 296 50 2307 4338 105 302 296 50 2307 4338 105 302 Income Utility Trade Emissions per worker (welfare)

EU 205 631 73 274 66 274 380 12 945 ROW 205 631 73 274 66 274 2525 2.5 2994 Total 0 0 0 0 0 0 2905 (7.69) Consumer - Chap 18 5/3/04 15:57 Page 209

18 Consumer Attitudes Towards GM Foods: The Modelling of Preference Changes

Chantal Pohl Nielsen,1 Karen Thierfelder2 and Sherman Robinson3 1Danish Institute of Agricultural and Fisheries Economics, Rolighedsvej 25, 1958 Frederiksberg C, Denmark; 2US Naval Academy, 589 McNair Road, Annapolis, MD 21402, USA; 3International Food Policy Research Institute, 2033 K Street NW, Washington, DC 20006, USA

Introduction consequences for the structure and pattern of world food trade. Regardless of whether a The sharp reaction against the use of geneti- country is a net exporter or net importer of cally modified organisms (GMOs) in food pro- agricultural and food products, it will be duction by some consumers has already affected to some extent by the changing con- initiated the creation of differentiated market- sumer attitudes toward GMOs in the devel- ing systems for genetically modified (GM) and oped world. Some countries are highly conventional maize and soybeans in the USA. dependent on exporting particular primary Consumer attitudes will be an important agricultural products to GM-critical regions. determinant for the profitability and hence the Depending on the strength of opposition viability of markets for non-GM varieties in towards GM products in such regions, the the longer term. For producers it is a matter costs of segregating production, and the rela- of assessing the benefits and costs of gaining tive productivity difference between GM and access to niche markets for non-GM crops rel- non-GM production, such countries may ben- ative to the benefits of lower production costs efit from the establishment of segregated agri- associated with cultivating GM crops. cultural markets for GM and non-GM Furthermore, many consumers are not only products. In principle these countries may critical of the use of genetic engineering tech- choose to grow GM crops for the domestic niques in the production of bulk commodities market and for exports to countries where such as soybeans and cereal grains; they are consumers are indifferent as to GMO content, also concerned about GM ingredients in ani- and to supply GMO-free products to countries mal feed and processed foods. To the extent where consumers are willing to pay a pre- that consumers are in fact willing to pay the mium for this characteristic. Such a market additional costs of having these preferences, development would be analogous to the niche identity preservation systems will develop so markets for organic foods. Other countries that these demands can also be satisfied. are net importers and can benefit from the These divergent consumer attitudes toward widespread adoption of GM technology. To GM foods and the increasing demand to be the extent that consumers in those countries informed about production processes through are not opposed to GM products, they will identity preservation systems etc. will have benefit from lower world market prices.

210 C. Pohl Nielsen et al.

This chapter analyses the price, production of land – making up 3.4% of the world’s total and trade consequences of changing con- agricultural area and representing a consider- sumer preferences regarding the use of GMOs able expansion from less than 3 million in food production. The analytical framework hectares in 1996.1 Cultivation of transgenic used is an empirical global general equilibrium crops has so far been most widespread in the model, in which the two primary GM crops, production of soybeans and maize, accounting soybeans and maize, are specified as either for 54% and 28% of total commercial trans- GM or non-GM. This GM vs. non-GM split is genic crop production in 1999, respectively. maintained throughout the entire processing Cotton and rapeseed each made up 9% of chain: GM livestock and GM food processing transgenic crop production in 1999, with the industries use only GM intermediate inputs; remaining GM crops being tobacco, tomato, likewise non-GM livestock and non-GM food- and potato (James, 1997, 1998, 1999). To processing industries use only non-GM inter- date, genetic engineering in agriculture has mediate inputs. This approach is an extension mainly been used to modify crops so that they of the authors’ earlier work, where only the have improved agronomic traits such as toler- primary crop markets were segregated (see ance to specific chemical herbicides and resis- Nielsen et al., 2000). tance to pests and diseases. Development of The following section provides a concise plants with enhanced agronomic traits aims at overview of the current status of GM crops in increasing farmer profitability, typically by food production and briefly discusses selected reducing input requirements and hence costs. issues related to the segregation of GM and Genetic modification can also be used to non-GM marketing systems. Section three improve the final quality characteristics of a presents the empirical analysis of changing product for the benefit of the consumer, food- consumer preferences regarding GMOs. processing industry or livestock producer. Such Consumer reactions against GMOs may be traits may include enhanced nutritional con- interpreted to mean several things. Hence the tent, improved durability and better processing empirical analysis illustrates two different characteristics. approaches and how they have been imple- The USA holds almost three-quarters of mented in the model. This too is an extension the total crop area devoted to GM crops. of the authors’ earlier work: preference Other major GM-producers are Argentina, changes may not only be taken to mean Canada and China. At the national level, the reduced price sensitivity – in this analysis we largest shares of genetically engineered crops also investigate what it means to interpret in 1999 were found in Argentina (approxi- preference changes as a structural shift. The mately 90% of the soybean crop) Canada empirical results are examined in the fourth (62% of the rapeseed crop) and the USA section, and a final section provides some (55% of cotton, 50% of soybean and 33% of concluding remarks. maize) (James, 1999). The US Department of Agriculture (USDA), figures for the USA are similar in magnitude (Economic Research Genetic Engineering in Food Production Service (ERS), 2000a): it is estimated that 40% of maize and 60% of soybean areas har- Current status vested in 1999 were genetically modified. Continued expansion in the use of transgenic Genetic engineering techniques and their appli- crops will depend in part on the benefits cations have developed rapidly since the intro- obtained by farmers cultivating transgenic duction of the first GM plants in the 1980s. In instead of conventional crops relative to the 1999, GM crops occupied 40 million hectares higher cost for transgenic seeds.2 So far the

1 Calculations are based on the FAOSTAT statistical database accessible at www.fao.org. 2 As long as private companies uphold patents on their transgenic seeds they will be able to extract monopoly rents through price premiums or technology fees. Consumer - Chap 18 5/3/04 15:57 Page 211

Consumer Attitudes towards GM Foods 211

improvements have been not so much in in a wide range of processed foods as well increased yields per hectare of the crops, but as in animal feed products. Given that con- rather by reducing costs of production (OECD, sumers in many developed countries are 1999). Empirical data on the economic bene- generally becoming more aware of the profits of transgenic crops are still very limited, duction processes that lie behind their food however. The effects vary from year to year products, they are also increasingly begin- and depend on a range of factors such as crop ning to formulate demands about how these type, location, magnitude of pest attacks, dis- processes should take place. This includes ease occurrence and weed intensity. whether or not the use of genetic engineering techniques is deemed acceptable or desirable. Hence livestock producers and Future market structures food-processing industries also need to consider the consequences of their input As indicated in the previous section, adop- choices both in terms of domestic and for- tion of GM crop varieties has been extremely eign demand. rapid in North America and in some large Current methods of testing a food prod- developing countries such as Argentina and uct for possible GMO content are not com- China. Lacking systematic evidence of the pletely reliable. Heating GM maize, for economic benefits of cultivating GM crops example, eliminates the GM proteins. This relative to conventional varieties, this rapid renders current testing methods unsatisfac- adoption must be taken to reflect real or tory if the information demanded by con- anticipated benefits for farmers. Further- sumers is whether or not genetic more, it may seem that the strong reaction engineering techniques have been used at against GMOs by consumers in, for example, any stage in the production process. Western Europe and Japan has not been Therefore, in order to provide consumers fully anticipated. The lack of consumer with the choice of purchasing guaranteed acceptance in these countries has made the non-GM foods, the principles of identity market for GM crops more uncertain. As preservation (IP) must be followed in the shall be seen below, several of the large pro- food-marketing systems. IP systems are well ducers of GM-potential crops are highly known from existing specialty markets (e.g. dependent on exporting to GM-critical coun- high-oil maize), but are also applied to a tries and hence there are important commer- greater or lesser extent for almost all traded cial interests in maintaining access to these agricultural products. Existing grading sys- markets. As the use of genetic engineering tems based on the type, length, colour, moves into its ‘second stage’ and increas- weight, water content, share of broken or ingly provides quality-enhanced foods (e.g. damaged grains, etc. can be thought of as better nutritional content, improved durabil- basic IP systems. After classification has ity, etc.) consumer attitudes toward GMOs taken place, the subsequent handling, stor- might change. However, as long as environ- age and processing systems must ensure mental and food safety issues remain uncer- that the identity of the product is retained tain there will be some consumers that wish throughout the supply chain – as far and as to avoid GMOs altogether. detailed as the final user or the regulatory The USDA (ERS, 2000b) reports that the authorities require (see Buckwell et al., present demand for non-GM maize and soy- 1999, for a more detailed discussion of the beans in the USA is very limited. Markets economics of IP systems). for non-GM crops have, however, developed Identity preservation adds administration in response to GM labelling requirements in and marketing costs at all stages along the sup- the European Union and to serve a handful ply chain, and they can be considerable. How of niche markets domestically, in Europe these costs are shared between the farmer, sup- and in Japan. This demand for non-GM plier, processor, distributor and consumer varieties could very well expand rapidly. depends on how price sensitive demand is at Maize and soybeans are used as ingredients each stage. The lower the demand elasticity, Consumer - Chap 18 19/3/04 9:29 Page 212

212 C. Pohl Nielsen et al.

the larger the share of the extra costs that must crops that have been genetically modified to be borne by the purchaser. To the extent that date are soybeans and maize. The sectoral demand for non-GM varieties is strong enough aggregation of the database for use in this to support the price differential that will arise analysis therefore comprises a cereal grains between GM and non-GM foods, product and sector (which includes maize but not wheat processing methods will be segregated into GM and rice) and an oilseeds sector (which and non-GM varieties to serve these differenti- includes soybeans) to reflect these two GM- ated demands, and private suppliers of market- potential crops. The livestock, meat and dairy, ing services to enable producers to segregate vegetable oils and fats, and other processed their products will develop. The larger the seg- food sectors are also singled out, since they regated section of the market, the lower these are important demanders of cereal grains and costs need be as economies of scale are real- oilseeds as intermediate inputs in production. ized. In summary, the economic consequences In terms of the importance of the two of changing consumer attitudes toward GMOs GM-potential crops in total primary agricul- in food production will depend crucially on the ture, Table 18.1 shows that the cereal grains relation between three aspects: (i) the nature sector accounts for almost 20% of agricul- and extent of the preference change (willing- tural production in the USA, approximately ness to pay); (ii) the costs of preserving the 11% in South America, but less than 7% in identity of products in completely segregated all other regions. Oilseed production markets; and (iii) the size of the relative produc- accounts for 6–7% of agricultural production tivity difference between GM and non-GM pro- in the Cairns group3, low-income Asia, the duction methods. USA and sub-Saharan Africa, while its share is even smaller in Western Europe, high- income Asia and the South American coun- Empirical Analysis of Consumer tries outside the Cairns group. Preferences As discussed above, maize and soybeans, the two crops that have benefited most from GM-potential foods in world production and genetic engineering techniques to date, are trade widely used both directly as animal feed and as ingredients in processed feed and food prod- The data used in the empirical analysis ucts. With the exception of some final con- described below are from version 4 of the sumption of oilseeds in high-income Asia, Global Trade Analysis Project (GTAP) data- cereal grains and oilseeds are generally not base, which is estimated for 1995 (McDougall used for final consumption in the three high- et al., 1998). As discussed above, the main income regions, high-income Asia, the USA

Table 18.1. Agricultural production structures, 1995. High- Low- Rest of Sub- Cairns income income South Western Saharan Rest of group Asia Asia USA America Europe Africa world

Cereal grains 4.9 1.8 4.9 19.1 10.7 5.4 2.9 6.7 Oilseeds 6.1 0.4 7.2 6.2 1.8 1.9 6.4 2.4 Wheat 4.6 0.8 6.2 4.8 2.0 5.1 4.3 7.8 Other crops 50.4 67.1 52.4 24.2 53.6 33.6 69.1 43.7 Livestock 34.0 29.9 29.3 45.7 31.9 54.0 17.3 39.3 Total agriculture 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Source: Multi-region GMO model database derived from GTAP version 4 data.

3 Consists of Argentina, Australia, Bolivia, Brazil, Canada, Chile, Colombia, Costa Rica, Guatemala, Indonesia, Malaysia, New Zealand, Paraguay, the Philippines, South Africa, Thailand and Uruguay. Consumer - Chap 18 5/3/04 15:57 Page 213

Consumer Attitudes towards GM Foods 213

and Western Europe. In the Cairns group and The importance of trade in these two in the developing regions, cereal grains and commodities varies across the regions. Table oilseeds are to a larger extent used for final 18.2 shows that the value of oilseed exports consumption, although these shares vary sub- relative to total value of production is signifi- stantially across regions. In terms of intermedi- cant for the Cairns group, the USA and the ate use, there are some interesting differences rest of South America. Cereal grain exports regarding the use of these two crops as direct are also moderately large in value terms for feed for livestock or as inputs into the food- the first two regions, but otherwise most of and feed-processing industries (the latter the production value of these two crops is including manufacturers of prepared feeds, captured on the domestic markets. For the e.g. soybean meal, feed pellets, etc.). Within Cairns group, the rest of South America, the the Cairns and high-income Asia groups, by USA and sub-Saharan Africa, the impact of far the largest share of these crops is used in genetic engineering would be much larger if food and feed processing rather than directly these techniques were applicable to the as feed. This is also the case for oilseed use in crops contained in the much larger aggre- both Western Europe and the USA. For cereal gate ‘other crops’ sector. On the import grains, however, three-quarters of the total use side, the value of oilseed imports into in the USA is directly for livestock feed, where- Western Europe amounts to almost 40% of as for Western Europe there is more of a the total value of oilseed absorption. High- 50–50 split for use of cereal grains directly as income Asia is also heavily dependent on livestock feed and as inputs into the processing imports of oilseeds and to a lesser extent industries. These use patterns illustrate how cereal grains. Table 18.2 also shows the important it is for GMO-critical consumers to importance of trade in livestock and be able to trace GM cereal grains and oilseeds processed food products by region. These throughout both the livestock and other food- trade dependencies are generally lower than producing chains. those just described.

Table 18.2. Trade dependence: agricultural and food products, 1995.

High- Low- Rest of Sub- Cairns income income South Western Saharan Rest of group Asia Asia USA America Europe Africa world

Value of exports in % of total production value Cereal grains 9.7 0.2 0.7 16.0 0.7 3.7 4.3 0.7 Oilseeds 15.7 4.1 2.7 28.7 32.4 1.8 5.8 11.2 Wheat 28.5 0.0 0.3 39.2 6.6 6.8 0.1 1.5 Other crops 15.4 0.7 3.5 18.9 29.2 4.7 20.0 6.6 Livestock 7.3 0.2 1.5 2.4 2.9 1.2 2.4 1.7 Vegetable oils and fats 32.8 4.8 3.2 7.2 4.0 4.3 10.3 6.7 Meat and dairy 10.2 0.4 12.6 4.9 1.5 3.1 11.3 1.7 Other processed foods 12.6 0.7 10.3 5.2 10.9 6.2 15.7 4.1 Value of imports in % of total absorption value Cereal grains 7.2 18.3 5.5 0.9 14.8 5.0 7.2 10.3 Oilseeds 6.5 71.1 0.9 2.4 55.2 38.2 0.4 10.6 Wheat 11.9 17.1 10.4 3.4 51.4 3.7 15.5 17.7 Other crops 5.5 6.5 2.3 17.8 5.7 18.3 1.4 8.0 Livestock 0.9 5.4 1.5 2.1 1.6 2.3 0.4 2.4 Vegetable oils and fats 3.1 19.0 17.2 5.0 15.3 4.1 14.5 23.1 Meat and dairy 2.0 9.9 6.4 1.8 8.9 1.5 35.1 10.4 Other processed foods 4.6 4.2 3.5 4.6 5.9 3.6 15.8 10.3

Source: Multi-region GMO model database derived from GTAP version 4 data. Consumer - Chap 18 5/3/04 15:57 Page 214

214 C. Pohl Nielsen et al.

Compared with trade in other crops, meat, income Asia and Western Europe. As men- dairy and other processed foods, the value of tioned above, the USA is a substantial world trade in cereal grains and oilseeds is exporter of both cereal grains and oilseeds. modest, as the net trade values in Table 18.3 Half of the former exports are sold in high- show. The USA is by far the dominant income Asia. For US oilseed exports, 34% exporter of both crops followed by the Cairns are sold in Western Europe and 37% in high- group. Western Europe is the main importer income Asia. A large share of meat and dairy of oilseeds and high-income Asia is the main and other processed food exports from the importer of cereal grains and the second USA goes to the high-income markets of largest importer of oilseeds. In terms of Asia. Cereal grain imports into high-income processed food trade, countries in the Cairns Asia are very narrowly sourced in the sense group and Western Europe are large that more than 90% of this region’s cereal exporters of meat and dairy products and grain imports come from the USA, whilst its other processed foods, of which high-income imports of processed food products come Asia is a major importer. from both the Cairns group and the USA. For In terms of bilateral trade ﬂows, it may be Western Europe, almost 45% of its oilseed noted that processed food exports from the imports come from the USA and another Cairns group are mainly destined for high- almost 35% from the Cairns group. As men-

Table 18.3. Net trade ﬂows and composition of world trade, 1995. High- Low- Rest of Sub- Cairns income income South Western Saharan Rest of group Asia Asia USA America Europe Africa world Total

Net trade (US$ billion) Cereal grains 0.31 4.51 0.88 8.03 1.20 0.13 0.06 1.81 0 Oilseeds 1.64 2.86 0.46 4.57 0.52 3.55 0.23 0.05 0 Wheat 2.71 1.62 2.40 5.35 1.00 1.23 0.55 3.72 0 Other crops 13.37 11.59 2.09 0.62 9.23 20.16 8.42 1.97 0 Livestock 5.84 4.47 -0.11 0.50 0.32 1.62 0.22 0.68 0 Vegetable oils and fats 7.18 1.16 3.02 0.50 0.61 0.16 0.14 2.91 0 Meat and dairy 9.81 11.09 0.82 5.48 1.93 5.53 0.69 7.93 0 Other processed foods 18.09 24.09 6.03 1.17 3.11 11.01 0.46 14.85 0 Value of exports in % of value of world trade Cereal grains 11.29 0.10 1.06 75.88 0.55 9.29 0.71 1.13 100 Oilseeds 26.48 0.48 6.89 49.83 4.18 2.43 2.52 7.20 100 Wheat 31.88 0.01 0.64 48.20 0.86 15.68 0.03 2.69 100 Other crops 28.05 1.83 8.78 16.29 15.01 7.47 12.44 10.13 100 Livestock 40.41 1.27 8.95 17.73 4.06 15.57 1.59 10.41 100 Vegetable oils and fats 55.86 2.16 3.50 11.37 1.30 18.22 1.67 5.91 100 Meat and dairy 34.65 1.05 4.68 24.33 1.01 29.84 0.45 3.98 100 Other processed foods 27.44 3.70 8.90 16.39 6.54 27.83 2.30 6.89 100 Value of imports in % of world trade Cereal grains 8.50 40.82 8.97 3.42 11.42 8.14 1.27 17.46 100 Oilseeds 9.65 29.88 2.20 2.85 9.54 38.99 0.19 6.71 100 Wheat 8.45 13.99 21.40 2.00 9.49 5.10 4.74 34.82 100 Other crops 9.48 17.94 5.88 15.43 2.20 35.47 0.75 12.86 100 Livestock 4.79 28.59 9.62 14.66 2.12 25.43 0.26 14.53 100 Vegetable oils and fats 4.10 10.50 25.26 7.81 5.69 17.04 2.71 26.90 100 Meat and dairy 6.74 32.60 2.36 8.75 6.51 14.10 2.40 26.53 100 Other processed foods 10.63 26.08 3.30 15.31 3.65 17.60 2.73 20.69 100

Source: Multi-region GMO model database derived from GTAP version 4 data. Consumer - Chap 18 5/3/04 15:57 Page 215

Consumer Attitudes towards GM Foods 215

tioned above, Western Europe is a large which demands commodities according to exporter of meat, dairy and other processed fixed expenditure shares, maximizing a Cobb- foods. About one-fifth of these exports are Douglas utility function. destined for the high-income Asian markets. As in other CGE models, it is only relative (Tables containing these data are to be found prices that are determined – the absolute in Nielsen et al., 2001.) price level is set exogenously. In this model, the aggregate consumer price index in each sub-region acts as the numeraire. A conve- Global CGE model with segregated food nient consequence of this specification is that markets solution wages and incomes are in real terms. The solution exchange rates in each region The modeling framework used in this analysis are also in real terms and can be seen as equi- is a multi-region computable general equilib- librium price-level-deflated exchange rates, rium (CGE) model consisting of eight regions, using the country consumer price indices as which are inter-connected through bilateral deflators. The international numeraire is trade flows: the Cairns group, high-income defined by fixing the exchange rate for North Asia, low-income Asia, the USA, the rest of America. World prices are converted into South America, Western Europe, sub-Saharan domestic currency using the exchange rate, Africa and the rest of the world.4 For the pur- including any tax or tariff components. Cross- pose of describing the model,5 it is useful to trade price consistency is imposed, so that the distinguish between the individual regional world price of country A’s exports to country models and the multi-region model system as B are the same as the world price of country a whole, which determines how the individual B’s imports from country A. regional models interact. When the model is Sectoral export-supply and import-demand actually used, the within region and between functions are specified for each region. As is region relationships are of course solved common in other CGE models, the multi- simultaneously. Each regional CGE model is a regional model used in this analysis specifies relatively standard trade-focused CGE model, that goods produced in different countries are with 12 sectors: five of which are primary imperfect substitutes. On the supply side, sec- agriculture, three are food-processing indus- toral output is a constant elasticity of transfor- tries and the remaining four are manufactur- mation (CET) aggregation of total supply to all ers and services. Each regional model has five export markets and supply to the domestic factors of production: skilled and unskilled market. The allocation between export and labour, capital, land and natural resources. For domestic markets is determined by the maxi- each sector, output supply is specified as a mization of total sales revenue. On the constant elasticity of substitution (CES) func- demand side the assumption of product differ- tion over value-added, and intermediate inputs entiation is combined with the almost ideal are demanded in fixed proportions (a Leontief demand system (AIDS) to determine the input specification). Profit-maximization behaviour aggregation equation. Although not used in by producers is assumed, implying that each this application, this specification allows for factor is demanded so that marginal revenue non-unitary income elasticities of demand for product equals marginal cost, given that all imports and pairwise substitution elasticities factors are free to adjust. Each regional econ- that vary across countries (unlike the more omy contains domestic market distortions in typical CES specification). The macro closure the form of sectorally differentiated indirect of the model is relatively simple. First of all, consumption and export taxes, as well as aggregate real investment and government household income taxes. There is a single consumption are assumed to be fixed. representative household in each economy, Secondly, since the trade balances in each

4 Note that the bilateral trade ﬁgures that link these regions are net of trade within the region, and that in the model intra-regional trade is treated as another source of domestic demand. 5 The description of the standard version of the model draws in part on Lewis et al. (1999). Consumer - Chap 18 5/3/04 15:57 Page 216

216 C. Pohl Nielsen et al.

region also are assumed to be fixed with the use in the GM and non-GM varieties are ini- real exchange rates adjusting to equilibrate tially assumed to be identical. The destination aggregate exports and imports, the macro clo- structures of exports are also initially assumed sure of the model is achieved by allowing to be the same, and this determines the result- domestic savings for each region to adjust to ing import composition by ensuring bilateral achieve macro equilibrium. trade flow consistency. The next step is to identify the sectors that use cereal grains and oilseeds as intermediate Data base adjustments and model extensions inputs as GM and non-GM sectors to reflect IDENTITY PRESERVATION: SEGREGATING INTERMEDIATE the concept of identity preservation. The USERS OF GM AND NON-GM CROPS In order to GM/non-GM split is applied to the following operate with segregated GM and non-GM sec- sectors: livestock, vegetable oils and fats, meat tors in the extended model, the base data and dairy, and other processed foods. In the must also reflect this segregation. First of all, base data the GM/non-GM split for these four the base data are adjusted by splitting the sectors is determined residually, based on the cereal grain and oilseed sectors into GM and share of GM inputs of cereal grains and non-GM varieties.6 It is assumed that all oilseeds in total (GM plus non-GM) inputs of regions in the model initially produce some of cereal grains and oilseeds for each sector. both GM and non-GM varieties of cereal These shares are then used to split the data grains and oilseeds. Specifically, the assumed into GM and non-GM varieties of the four shares are as shown in Table 18.4, adapted processing sectors. At this stage, the from estimates provided in James (1999) and described procedure leaves all agricultural and (ERS 2000a)7. In the Cairns group, for exam- food sectors using some of both GM and non- ple, 40% of total cereal grain production is GM inputs. The input–output table is then assumed to be of the GM variety, whilst 60% adjusted so that GM sectors only use GM of oilseed production is assumed to be GM. inputs and non-GM sectors only use non-GM The structures of production in terms of the inputs. As mentioned above, in the descrip- composition of intermediate input and factor tion of the model, intermediate demand is

Table 18.4. Assumed initial shares of GM varieties in cereal grain and oilseed production. High- Low- Rest of Sub- Cairns income income South Western Saharan Rest of group Asia Asia USA America Europe Africa world

GM cereal grains as a % of total (GM + non-GM) cereal grain 40 20 40 40 40 15 15 15 production GM oilseeds as a % of total (GM + non-GM) oilseed production 60 35 60 60 60 15 15 15

6 As will be discussed later, the distinguishing characteristic between these two varieties is the level of productivity. Furthermore, although acknowledging the fact that there may be environmental risks and hence externality costs associated with GM crops, they are impossible to estimate at this time and this chapter makes no attempt to incorporate such effects in the empirical analysis. 7 It is due to technical data limitations that the shares of GM varieties in total crop production in high- income Asia are as high as they are. The important point here is the relative size of the shares between regions, not so much the levels within each region. Producers in high-income Asia and Western Europe are assumed to be reluctant with respect to adopting GM crops due to negative consumer attitudes. GM adoption in sub-Saharan Africa and the rest of world is assumed to be restricted by other factors such as lack of access to the technology. The other regions are assumed to adopt GM crop varieties enthusiasti- cally and in fact be able to do so. Consumer - Chap 18 5/3/04 15:57 Page 217

Consumer Attitudes towards GM Foods 217

characterized by a Leontief specification, i.e. set of equations for each of the six food cate- the input-output coefficients are fixed. Given gories (here shown for one arbitrary food the adjustments of the input–output data product f):

described above, this means that in the 1/ + ρ C(,) gmf k  PC(,) ngf k α (,)gmf k  ((,))1 G fk model, intermediate use in the GM sectors is =⋅ G  (2) Cngk(,) PC(,) gmf k ((,))1 − α gmf k restricted to only GM inputs and intermediate  G  use in the non-GM sectors is restricted to only The adding-up constraints are included as fol- non-GM inputs. This is an important differ- lows, ensuring that expenditure on each com- ence compared with the authors’ earlier work posite good (i.e. GM plus non-GM) remains (Nielsen et al., 2000), where intermediate fixed: users of oilseeds and cereal grains had a choice between GM and non-GM varieties. C0(gmf,k) + C0(ngf,k) = C(gmf,k) + C(ngf,k) (3) where C0(gmf,k) and C0(ngf,k) are the base ENDOGENOUS FINAL DEMAND CHOICE BETWEEN GM level share coefficients for the GM and non- AND NON-GM FOODS In the model con- GM varieties, respectively. sumers have a choice between GM and non- GM varieties. Final demand for each composite good (i.e. GM plus non-GM) is held Design of experiments fixed as a share of total demand, while introducing a choice between GM and non-GM GM and non-GM production technologies varieties. In this way, all the initial expenditure As mentioned above, the distinguishing char- shares remain fixed, but for six of the food acteristic between the GM and non-GM maize product categories (oilseeds, cereal grains, and soybean sectors is the level of productivity. livestock, vegetable oils and fats, meat and The GM cereal grain and oilseed sectors are dairy, and other processed foods), a choice assumed to benefit from increased productivity has been introduced between GM and non- in terms of primary factor use as well as a GM varieties. All other expenditure shares reduction in chemical use.8 The available esti- remain fixed. mates of agronomic and hence economic ben- The choice between GM and non-GM efits to producers from cultivating GM crops varieties is determined by a CES function are very scattered and highly diverse (see e.g. (here shown for an arbitrary food product f in OECD, 1999 for an overview of available esti- country k): mates). Nelson et al. (1999), for example, sug- −ρ =⋅α ⋅G (,)fk + gest that glyphosate-resistant soybeans may Cfk(,) afk (,)[G (,) fk Cgmfk ( ,) 1 (1) −ρ − / ρ generate a total production cost reduction of (1 −⋅α (,))f k C ( ngf ,) k GG(,)fk ] (,)fk G 5%, and their scenarios have GM maize where C(f,k) is the share of food product f in increasing yields by between 1.8% and 8.1%. total final consumption in country k, gmf is For present purposes, the GM-adopting cereal the GM variety and ngf is the conventional grains and oilseed sectors are assumed to variety. The parameter a(f,k) is the CES make more productive use of the primary fac- demand shift parameter. The exponent is tors of production as compared with the non- defined by the elasticity of substitution GM sectors. In other words, the same level of σ between GM and non-GM varieties, G(f,k): output can be obtained using fewer primary ρ σ G(f,k) = [1/ G(f,k)] – 1. The CES share coeffi- factors of production, or a higher level of out- α α cients are G(f,k) and (1 G(f,k)). In the put can be obtained using the same level of model, the following first-order conditions are production factors. In our scenarios, the GM included for the demand for the six food types oilseed and GM cereal grain sectors in all that come in GM and non-GM varieties – one regions are assumed to have a 10% higher

8 Note that this is an asymmetric shock and that it will therefore have different effects in different regions because of different cost structures: the shares of primary factor costs and chemical costs in total production costs are different. Consumer - Chap 18 5/3/04 15:57 Page 218

218 C. Pohl Nielsen et al.

level of factor productivity as compared with that the two varieties are imperfect substitutes their non-GM (conventional) counterparts. with a region-specific elasticity of substitution, σ Furthermore, there seems to be evidence that G(f,k), between the two. Three different con- cultivating GM varieties substantially reduces sumer response scenarios are examined (sum- the use of chemical pesticides and herbicides marized in Table 18.5). In the base case (see e.g. Pray et al., 2000). Hence the use of consumers in all countries are relatively indiffer- chemicals in the GM oilseed and GM cereal ent with respect to the introduction of GM tech- grain production is reduced by 30% to illus- niques in food production, and so they find GM trate this cost-saving effect. and non-GM food varieties highly substitutable. It is assumed that the elasticity of substitution between GM and non-GM varieties is high and σ Consumer preferences equal in all regions. Specifically, G(f,k) = 5.0 for all for all regions k. There are many ways to model changes for- The next two experiments then attempt to mally in consumer preferences. This chapter reflect the fact that citizens in Western Europe illustrates how two such ways can be imple- and high-income Asia dislike the idea of GM mented in a computable general equilibrium foods. In the second experiment this is illus- model. This is done by shifting and altering the trated by lowering the elasticities of substitution curvature of the indifference curve between GM between the GM and non-GM varieties for con- σ and non-GM commodities. Each alternative has sumers in these two regions G(f,k) = 0.5 for k a different interpretation of what consumers = Western Europe and high-income Asia). might mean when they say they disapprove of Consumers in these regions are simply assumed GM foods. The starting point for the consumer to be less sensitive to a given change in the ratio preference experiments is that food products of prices between GM and non-GM varieties. come in two varieties, distinguished by their They are seen as poor substitutes in consump- method of production: GM and non-GM. As tion in these particular regions. Citizens in all described above, the model has the representa- other regions are basically indifferent, and tive consumer demanding these two product hence the two varieties remain highly substi- σ varieties in terms of a CES function, meaning tutable in consumption ( G(f,k) = 5.0).

Table 18.5. Experiment design to illustrate different representations of preference changes. Experiment Description

1. Base case Base model 1: Consumers in all countries ﬁnd GM and non-GM foods highly and equally σ substitutable. Elasticity of substitution between GM and non-GM varieties G(k) = 5.0 for all regions k Shock to base model 1: Factor productivity increase (10%) Reduced chemical use in GM oilseed and GM cereal grains (30%) 2. Price sensitivity Base model 2: Consumers in Western Europe and high-income Asia are less responsive to σ changes in the relative price of GM and non-GM goods. G(k) = 0.5 for k = W. Europe and high-income Asia. Other regions: Same as base case. Shock to base model 2: Same as base case. 3. Structural change Consumers in Western Europe and high-income Asia completely shift their consumption away from GM goods. Other regions: Same as base case. Shock to base model 2: Same as base case plus shock the share of GM varieties in ﬁnal consumption to 0.02 for Western Europe and high-income Asia. Consumer - Chap 18 5/3/04 15:57 Page 219

Consumer Attitudes towards GM Foods 219

Comparing experiments 1 and 2 corre- the same budget line (same expenditure on sponds to altering the curvature of the indif- the composite food product, i.e. GM plus ference curves of consumers in Western non-GM, and hence same level of utility). The α α Europe and high-income Asia as illustrated in CES share parameters G(f,k) and (1 G(f,k)) Fig. 18.1. The two curves in the ﬁgure corre- are unchanged by this experiment.

spond to the same level of utility, U0. When It is not clear, however, whether reduced the relative prices of GM and non-GM foods price sensitivity is an appropriate interpretation change, consumers in Western Europe and of consumers’ critical approach to GM foods. In high-income Asia are, in the second experi- some rich countries, where consumers can ment, assumed to be less inclined to shift con- indeed afford to be critical of these new tech- sumption toward GM varieties as they were in niques in food production, irrespective of how the base case, where substitutability was high. cheap these products may become (relative to In terms of the CES function (equation 1), this non-GM foods), some consumers may simply ρ means that we are changing the G(f,k) para- not want to consume them. In this case, we are meter by altering the substitution elasticity, in changing the ratio of GM to non-GM foods effect, changing the curvature of the CES demanded at a given (constant) price ratio, hold- function. The representative consumer is on ing utility constant. This is illustrated in Fig.

U 0 U 0 GM and non-GM poor substitutes (low price sensitivity) GM foods

X 0 GM and non-GM good substitutes (high price sensitivity)

Non-GM foods Fig. 18.1. Consumer preferences modelled as different degrees of price sensitivity.

U 0 Equal share of GM and non-GM in consumption

= U 1 U 0 GM foods

X 0 Lower share of GM in total consumption X 1

Non-GM foods Fig. 18.2. Consumer preferences modelled as a structural change. Consumer - Chap 18 5/3/04 15:57 Page 220

220 C. Pohl Nielsen et al.

18.2, where the representative consumer in chemicals in total production costs. In sectors Western Europe and high-income Asia is as well and regions where these costs make up a large off as before but now with a lower share of GM share of total costs, the impact of the produc- foods in his/her consumption bundle. tivity shock in terms of lower supply prices will Recalibrating the model around these new be greater than in sectors and regions where shares gives us a new constant and new shares the share is smaller. Intermediate users of GM α α in the CES function, G(f,k) and (1 G(f,k)), inputs (the GM livestock and GM processed determining the composite use of GM and non- food producers) will beneﬁt from lower input GM foods in consumption. In determining these prices. The non-GM product markets will be new values we ensure that the total value of affected by the productivity gain in the GM expenditure on each composite food item sectors in three ways. First, there will be remains the same. In other words, consumers increased competition for primary factors of still spend the same amount on their consump- production and intermediate inputs because tion of food, but the composition hereof is GM production will increase. Second, con- changed in favour of non-GM varieties. In the sumers domestically might change their con- experiment we reduce the GM share of foods in sumption patterns in response to the new consumption in Western Europe and high- relative prices depending on their initial con- income Asia to 2%. The description of the sumption pattern and substitution possibilities. three experiments is summarized in Table 18.5. Third, importers will change their import pattern depending on the relative world prices, their initial absorption structures and the sub- Results of Empirical Analysis stitution possibilities between suppliers. In all three cases, the initial cost, consumption and Price and trade results import structures on the one hand, and the substitution possibilities between products for Base case experiment input use, ﬁnal consumption and imports on The increase in factor productivity and the the other, will determine the net impact of the reduced need for chemicals in the GM cereal productivity experiment. The net effects are grain and oilseed sectors causes the cost-driven theoretically ambiguous and hence must be prices of these crops to decline. The magni- determined empirically. tude of this price decline in the different sec- Figure 18.3 depicts for selected regions tors and regions will differ, depending on the the price wedges that arise between the non- shares of primary production factors and GM and GM varieties in the base case experi-

9 8 7 6 5 4 3 2

Price wedge (% difference) 1 0 Cereal grains Oilseeds Livestock Meat and Vegetable oils Other dairy and fats processed foods High-income Asia USA Western Europe Sub-Saharan Africa

Fig. 18.3. Base case experiment: price wedges between non-GM and GM products (percentage points). Consumer - Chap 18 5/3/04 15:57 Page 221

Consumer Attitudes towards GM Foods 221

ment, where GM and non-GM foods are con- hence the spillover effect is largest here. As sidered to be good substitutes in consumption the base data revealed, a rather large share of in all regions. Generally, the relative price of cereal grains is used as intermediate input to non-GM to GM commodities rises, and the livestock production in the USA compared percentage point differences between the with other regions. For this reason Fig. 18.3 prices of non-GM and GM varieties of cereal shows a relatively large price wedge for live- grains and oilseeds are between 6.3 and 9.4. stock, meat and dairy products in the USA As described above, the price wedges vary compared with the other regions. across the regions in part because they have The lower GM crop prices mean improved different shares of primary factor and chemi- international competitiveness for exporters of cal costs in total production costs. Hence the these crops. Hence, as Table 18.6 shows, extent to which the individual regions beneﬁt the largest exporters of cereal grains and from the productivity increase differs. oilseeds, the Cairns Group and the USA, The lower GM crop prices in turn result in increase their exports of GM crops in this lower production costs for users of GM base case by between 8.6% and 14.9%. Due inputs, thereby reducing those product prices to the reduced relative competitiveness of relative to the non-GM varieties as well. As non-GM crops, exports of this variety can be seen in Fig. 18.3, the price wedges decline. High-income Asia and Western that arise between the GM and non-GM live- Europe increase their imports of the cheaper stock and processed food products are of GM varieties. This is particularly so in the course much smaller than the price wedges case of oilseeds because these two regions between GM and non-GM primary crops are highly dependent on imported oilseeds because the cost reduction concerns only a from countries that are enthusiastic GM- part of total production costs. Relatively adopters. Imports of the non-GM varieties speaking, oilseeds constitute a large share of decline slightly due to the reduced relative production costs in vegetable oils and fats pro- price competitiveness of non-GM products duction (compared with oilseed and cereal when consumers ﬁnd GM and non-GM food grain use in other food production), and varieties to be good substitutes.

Table 18.6. Selected trade results of base experiment (percentage changes). High- Low- Sub- Cairns income income Western Saharan group Asia Asia USA Europe Africa

Exports Non-GM cereal grains 4.7 3.4 8.0 2.4 3.0 4.1 GM cereal grains 14.9 17.4 22.6 9.0 16.5 23.0 Non-GM oilseeds 4.3 3.0 9.1 2.1 2.9 3.0 GM oilseeds 13.0 13.5 16.7 8.6 17.6 20.7 Non-GM vegetable oils and fats 2.0 1.9 2.2 1.3 1.0 0.6 GM vegetable oils and fats 4.4 6.4 3.7 3.4 3.9 3.6 Non-GM other processed food 0.2 0.2 0.4 0.3 0.2 0.1 GM other processed food 0.7 0.8 0.9 0.7 0.8 1.1 Imports Non-GM cereal grains 4.2 0.2 12.3 1.8 0.3 4.8 GM cereal grains 5.6 1.7 19.7 2.7 0.8 32.8 Non-GM oilseeds 9.1 3.0 14.8 4.3 1.7 5.5 GM oilseeds 10.2 10.7 16.4 5.1 9.2 27.4 Non-GM vegetable oils and fats 2.0 1.9 2.2 1.3 1.0 0.6 GM vegetable oils and fats 4.4 6.4 3.7 3.4 3.9 3.6 Non-GM other processed food 0.2 0.2 0.4 0.3 0.2 0.1 GM other processed food 0.7 0.8 0.9 0.7 0.8 1.1 Consumer - Chap 18 19/3/04 9:29 Page 222

222 C. Pohl Nielsen et al.

Price sensitivity experiment experiment, final demand in these regions is not only very insensitive to relative price differ- The price wedges resulting from the price sen- ences between GM and non-GM food varieties. sitivity experiment are not markedly different Consumers in Western Europe and high- from the ones reported in the base case experiment. It may be mentioned, however, that the income Asia are assumed to simply shift their prices for GM cereal grains and especially consumption patterns away from GM varieties oilseeds are slightly lower on the Western and in favour of non-GM varieties, regardless European and high-income Asian markets of the relative price decline of GM foods. This when consumers are critical (i.e. less price sen- shift is measured relative to the experiment, in sitive): larger price reductions are required in which price sensitivity in these regions is low to order to sell GM varieties in GMO-critical mar- begin with. Hence the effects of this structural kets. Conversely, demand for non-GM crops is shock are an addition to the second experi- relatively stronger, and hence the prices of non- ment. The results show that this rejection is GM oilseeds, for example, are higher. Hence clearly a much more dramatic change com- we find that the price wedges for especially pared with reduced price sensitivity. Critical oilseeds, but also cereal grains, are larger in consumers simply do not want to eat GMOs. high-income Asia and Western Europe in the The price of GM varieties in the GMO-critical price sensitivity experiment. In large oilseed- countries declines further because of the almost producing markets such as the USA, the price complete rejection of these products, whereas of the non-GM variety falls slightly more and the price of non-GM foods increases. This the price of the GM variety falls less as com- leads to substantially larger price wedges in the pared with the base case. Compared with the GM-critical regions as compared with the previ- base case, the increase in GM oilseed and ous experiments, as is evident from Fig. 18.4. cereal grain exports from the Cairns group and By the nature of this model, the larger price the USA is smaller when consumers in their wedges between GM and non-GM primary important export markets are less responsive to crops follow through the entire food-processing the GM/non-GM price difference. Conse- chain. The price increase for non-GM foods is, quently, on the import side, the results show however, moderated by the fact that there are that the declines in imports of the more expen- indeed markets for non-GM products in all sive non-GM oilseeds into high-income Asia regions in the model, so these consumers are and Western Europe are smaller. The decreases not closing themselves off to necessary goods in non-GM cereal grain imports have even nor are they required to produce all the non- turned into minor increases. High-income Asia GM goods themselves. The model allows all and Western Europe still increase their GM countries to produce both varieties and hence oilseed imports in this price sensitivity experi- supply both GMO-indifferent and GMO-critical ment (although at lower rates) because of their consumers. Total US GM cereal grain and high dependence on importing from GM- oilseed exports fall by no less than –17% and enthusiastic regions. This result is due to the –33%, respectively (Table 18.7). Instead, fact that there is a symmetry in the trade exports of the non-GM varieties increase by dependence concerning oilseeds: US oilseeds make up a large share of oilseed imports into 10% and 16%, respectively. These changes are high-income Asia and Western Europe, and a direct reaction to the relative prices obtain- these regions make up a large share of US able on their key export markets, namely high- exports. Hence changes in consumer prefer- income Asia and Western Europe. The price of ences in these countries will have an impact on GM cereal grains and oilseeds on these markets the trading conditions for US producers. plummets and the price of non-GM varieties increases slightly. The price decline for GM cereal grains in Western Europe is not as large Structural change experiment as for oilseeds because this region is less dependent on imports of these crops, relatively In this final experiment consumers in Western speaking. This explains the larger price wedge Europe and high-income Asia simply turn for oilseeds compared with cereal grains in against GM foods. Compared with the previous Western Europe as depicted in Fig. 18.4. Consumer - Chap 18 5/3/04 15:57 Page 223

Consumer Attitudes towards GM Foods 223

5 Price wedges (% difference)

0 Cereal grains Oilseeds Livestock Meat and Vegetable oils Other dairy and fats processed foods High-income Asia USA Western Sub-Saharan Africa

Fig. 18.4. Structural change case: price wedges between non-GM and GM products (percentage points).

Table 18.7. Selected trade results of structural shift experiment (percentage changes). High- Low- Sub- Cairns income income Western Saharan group Asia Asia USA Europe Africa

Exports Non-GM cereal grains 1.4 4.0 1.7 8.1 3.8 0.8 GM cereal grains 1.4 42.5 4.1 17.4 30.7 2.0 Non-GM oilseeds 10.7 12.9 1.6 14.5 2.2 5.1 GM oilseeds 31.8 45.8 9.6 33.3 33.7 5.9 Non-GM vegetable oils and fats 5.4 11.3 6.5 3.7 6.5 6.4 GM vegetable oils and fats 13.9 29.7 35.6 10.6 29.2 50.2 Non-GM other processed food 7.4 11.1 7.5 5.4 8.0 6.9 GM other processed food 31.9 39.6 35.3 30.3 37.4 50.6 Imports Non-GM cereal grains 3.9 18.9 12.6 0.1 9.8 4.0 GM cereal grains 4.4 70.7 21.2 0.7 59.1 34.8 Non-GM oilseeds 6.4 23.5 14.1 6.0 10.3 4.2 GM oilseeds 12.0 56.8 28.8 22.7 60.4 40.3 Non-GM vegetable oils and fats 3.0 24.8 0.8 2.6 9.5 5.4 GM vegetable oils and fats 3.3 72.4 1.5 1.7 59.5 11.2 Non-GM other processed food 2.9 15.4 3.4 2.6 8.5 3.9 GM other processed food 2.8 66.8 1.0 0.2 60.2 10.5

Turning to the import results, Table 18.7 increase substantially, at slightly higher shows that imports of GM cereal grain and prices. The sourcing of these non-GM crop oilseeds into Western Europe and high- imports is spread across all regions, because income Asia decline substantially (between in the model all regions are assumed to be –57% and –71%). These decreases in quanti- able to produce both varieties and to be able ties are accompanied by import price to credibly verify this characteristic to declines in the order of –21% to –26%. importers. Clearly, this is a simpliﬁcation of Conversely, imports of non-GM crops reality, and one can easily imagine that for Consumer - Chap 18 5/3/04 15:57 Page 224

224 C. Pohl Nielsen et al.

some regions, living up to the principles of does in e.g. low-income Asia, a region that is identity preservation and verifying this is very not very heavily engaged in international trade costly, thereby putting them at a cost disad- in these particular crops. Figure 18.5 compares vantage. Such effects are not captured in this the impact on production in the USA of the model. The increases in non-GM cereal grain different and changing assumptions made and oilseed imports are supplemented by about consumer preferences in Western increases in own production in both high- Europe and high-income Asia. Since exports income Asia and Western Europe9. make up a relatively large share of the total value of production in these sectors, particularly for oilseeds, we see that there is a marked Production results effect on the composition of production. Production of GM crop varieties increases in Being a major exporter of both crops, the the ﬁrst two experiments, whilst production of increased demand for GM cereal grains and non-GM varieties declines somewhat. The oilseeds in the base case experiment ﬁlters impact is slightly less when consumers in high- through to an increase in production of these income Asia and Western Europe are less sen- crops in the USA. The effect is dampened, sitive to the GM/non-GM price difference. however, by the fact that its major destination In the structural shift experiment, however, regions (high-income Asia and Western the production of GM oilseeds in the USA Europe) have much larger non-GM sectors (rel- declines by 15% in spite of the factor pro- ative to their GM sectors), which are required ductivity gain and the reduced chemical to use only non-GM inputs.10 This also means, requirements. This is because the USA is so for example, that the production of non-GM highly dependent on exporting especially crops does not fall as markedly in the USA as it oilseeds to the GM-critical markets and

12 Structural shift 8

0 Base case Price sensitivity –4

–8

Production (% difference) –12

–16 Non-GM cereal grains GM cereal grains Non-GM oilseeds GM oilseeds

Fig. 18.5. Production effects in the USA (%).

9 Note that Western Europe might be restricted by the Blair House agreement in terms of increasing acreage for oilseed production and so the reported production increase may not be allowed. 10 Comparing these production effects with the results of our previous analysis, which did not have the identity preservation (IP) requirement in place (Nielsen et al., 2000), we see that the effects reported here are substantially smaller. This is precisely because the IP requirement introduces much stronger restrictions on intermediate input choice for livestock producers and food processors. In our previous analysis intermediate users had a free choice between GM and non-GM varieties and could therefore beneﬁt fully from the lower GM prices. In this model, however, intermediate users are required to use only GM or non-GM inputs. Consumer - Chap 18 5/3/04 15:57 Page 225

Consumer Attitudes towards GM Foods 225

because a structural consumer preference importance of exports in total production is change has much more of an impact on this small, and therefore these changes only show region’s trading opportunities compared with up as small compositional changes in the the reduced price sensitivity experiment. The production of livestock, meat and dairy prod- production of non-GM oilseeds, on the other ucts, vegetable oils and fats, and other hand, increases by 10% – another direct processed food products in the USA. Similar reflection of the importance of the GMO-criti- compositional changes are reported for the cal export markets for this region. Cairns group, but because exports of As was discussed above, the Cairns group processed foods are relatively more impor- is not as dependent on its exports of cereal tant to this region, the magnitude of change grains and oilseeds as the USA is. in terms of production changes is slightly Furthermore, its exports of cereal grains are larger than for the USA. more evenly spread across trading partners. An interesting question is whether these For these reasons, the productivity difference changing preferences in Western Europe and between GM and non-GM varieties shows up high-income Asia can open opportunities for more directly in the production results for the developing countries to export non-GM vari- Cairns group in the first two experiments (Fig. eties of cereal grains and oilseeds to these 18.6). The percentage changes are larger in regions. Sub-Saharan Africa has some pro- the Cairns group compared with the USA, duction of oilseeds, for example, and although but these changes are from a smaller initial exports of these crops do not account for a base. Furthermore, the structural shift experi- significant share of total production value at ment does not change the production results present, they might if niche markets for non- for the Cairns group as much as it does for GM crops develop in Western Europe. the USA because the former group of coun- Similarly, low-income Asian countries might tries is relatively less dependent on exports of look into expanding their production of e.g. these particular crops. For cereal grains pro- non-GM oilseeds if nearby niche markets in duction in the Cairns group countries, the high-income Asian countries develop. changes are not large enough to switch the Although the differences are very small, com- signs as they did in the USA because the paring the export and production results of Cairns group exports only very little to the three experiments indicates that this might Western Europe and high-income Asia. be a path to follow if the price premiums Although the export results for processed obtainable for non-GM varieties are large foods show similar compositional changes enough to outweigh the relative decline in between GM and non-GM varieties, the productivity and any identity preservation and

12 10 8 6 Structural shift 4 2 0 –2

Production (% difference) –4

–6 Base case Price sensitivity –8 Non-GM cereal grains GM cereal grains Non-GM oilseeds GM oilseeds

Fig. 18.6. Production effects in the Cairns group (%). Consumer - Chap 18 5/3/04 15:57 Page 226

226 C. Pohl Nielsen et al.

labelling costs. But even more significant in assumption, real absorption is an appropriate value terms for these countries are exports of welfare measure. It indicates the change in processed foods, i.e. vegetable oils and fats, the total amount of goods and services con- meat and dairy products, and other processed sumed following a change in preferences. The foods. Important in that respect are existing results of the experiments show that global trade patterns, proximity of markets, histori- absorption increases by US$7.4 billion in the cal ties, etc. which will determine whether or base case, where consumers are assumed to not producers will choose to forego productiv- find GM and non-GM foods to be good substi- ity increases and lower costs in GM produc- tutes. Increasing the price sensitivity of GM- tion in order to retain access to their critical consumers in high-income Asia and traditional export markets by selling non-GM Western Europe lowers this gain in total products. For a region like sub-Saharan Africa absorption marginally to US$7.2 billion. As with strong ties to Western Europe, changing the previous results have shown, the structural consumer attitudes toward GM foods are shift experiment is a much more dramatic expected to be an important determinant of change in preferences, and hence we find that future decisions regarding genetic engineering the global absorption gain is only US$0.02 in food production. billion in that experiment. The absorption results are reported for selected regions in Fig. 18.7 for the three Absorption results experiments. The changes are reported in billions of US dollars and it should be noted In this modelling framework, where we are that the percentage changes are very small. operating with a representative consumer, we It is clear from this figure that the Cairns are implicitly aggregating over two consumer group, low-income Asia and the USA are the types – those who are indifferent about GM main beneficiaries of the productivity products and those who are concerned about increase given that these are the regions potential hazards of consuming GM products. assumed to be intense adopters of the GM We have considered two changes in prefer- crop varieties. All other regions also experi- ences concerning GM-inclusive foods. First, ence an increase in total absorption, albeit at attitudes harden. The size of the two groups a lower absolute level. Reducing the price does not change, but those who are con- sensitivity of consumers in high-income Asia cerned about GM products become more and Western Europe reduces the increase in price sensitive. As described above, this global absorption only marginally and does changes the curvature of the indifference not change the distribution of the gains curve, as shown in Fig. 18.1. Second, we across regions. Most importantly, all regions have considered the effects of a structural still gain in terms of aggregate absorption preference shift – more people perceive that from the productivity increase and hence there are health hazards from consuming GM lower product prices in spite of the increased foods and choose to consume less, i.e. the aversion towards GM foods in high-income share of consumption of GM foods drops, Asia and Western Europe. regardless of relative price changes. In Interpreting consumer preference changes essence, the group of GM-sensitive consumers as a structural shift, however, alters the expands. This causes the indifference curve to absorption results dramatically. Because our shift, as depicted in Fig. 18.2. model has completely segregated GM and As discussed above, the level of utility stays non-GM production systems restricting input the same when the indifference curve shifts. use to either GM or non-GM varieties, the The representative consumer is on the same structural preference shift has a strong effect budget line with a different combination of on the demand for non-GM intermediates, GM and non-GM foods. We do not assume and not all regions experience increases in that the consumer obtains additional utility total absorption in this experiment. Despite from his/her decision to increase the share of the productivity gain in the large GM crop non-GM products he/she consumes. With this sectors in the Cairns group and the USA, Consumer - Chap 18 5/3/04 15:57 Page 227

Consumer Attitudes towards GM Foods 227

2.2 1.8 1.4 1 0.6 0.2 –0.2 Cairns groupHigh-income Low-income USA Western Sub-Saharan –0.6 Asia Asia Europe Africa –1 –1.4 Absorption (US$ billion) –1.8 –2.2 –2.6

Base case Price sensitivity Structural change Fig. 18.7. Changes in total absorption.

these results reveal that aggregate absorption rejection of GM foods, this means that declines in these regions when consumers in regions that are major exporters of GM- important export markets turn against their potential crops as well as being enthusiastic main product and there is little diversion to GM-adopters risk losing out on overall other markets. Total absorption declines by absorption if consumer attitudes turn against US$2.6 billion in the Cairns group and GM foods. US$0.9 billion in the USA. Although these declines amount to percentage changes of just –0.05% and –0.007%, respectively, they illus- Concluding Remarks trate how different interpretations of preference changes will have very different impacts This chapter has analysed the price, produc- on total absorption results. tion and trade consequences of changing Furthermore, the impact on aggregate consumer attitudes toward the use of genetic absorption in the GM-critical regions is engineering techniques in food production ambiguous. Comparing these results with our using an empirical global general equilibrium previous work (Nielsen et al., 2000), in model, in which the food-processing chain is which market segregation concerned only the segregated into GM and non-GM lines of primary cereal grain and oilseed sectors and production. Clearly, the present analysis preference changes were represented just as relies on some simplifying assumptions, in reduced price sensitivity, we ﬁnd that model- particular those made about the productivity ling preference changes as a structural shift in impact of adopting GM crops. Improved data this model changes not only the magnitude, would ideally provide information about how but potentially also the direction of our the GM productivity effects differ across sec- absorption results. In our previous model, tors and regions. Another limitation of this livestock and food producers that used cereal analysis is that it does not explicitly take grains and oilseeds as inputs had ﬂexibility in account of the costs of having to preserve the their choice of GM and non-GM varieties. In identity of a crop or food product throughout this model, we have restricted production the production and marketing chains, or the decisions substantially by limiting input use to costs of any testing and labelling require- only GM or non-GM varieties throughout the ments at national borders. Experience from food production chain. Combined with a identity preservation of speciality crops today structural shift in preferences in important reveals that these costs can potentially export markets, amounting in effect to a increase the price of such products by Consumer - Chap 18 8/3/04 10:42 Page 228

228 C. Pohl Nielsen et al.

between 5 and 15% (Buckwell et al., 1999). these crops as inputs in production. It is argued by, for example, Frandsen and Interpreting consumer dislike of GM foods as Nielsen (1999) and Runge and Jackson a reduced sensitivity to relative price changes (1999) that such a cost-price premium will – dampens the impact of the productivity differ- in a free market, and in the absence of ence between the two varieties. unsympathetic political reactions – emerge If consumer preference changes are in fact on the guaranteed non-GM products. more a matter of rejection rather than Whether consumers in Western Europe and reduced price sensitivity, the effects on prices, elsewhere are in fact willing to pay such a production and trade flows are much more premium is yet to be determined empirically. dramatic and the direction of effects reverses To the extent that evidence becomes avail- in some cases. Countries that are heavily able of a threshold beyond which consumers dependent on exporting GM-potential crops are unwilling to pay for their non-GM prefer- to the GMO-critical regions find themselves ences, for example, this could be incorpo- increasing exports and hence production of rated explicitly in the model used here. non-GM varieties and reducing production of Whilst keeping these caveats in mind, the GM varieties in spite of the productivity bene- empirical analysis described in this chapter fit. Furthermore, total absorption results are does bring attention to two very important also dependent on how preference changes aspects of the GMO debate: (i) segregation of are interpreted and hence modelled. Clearly, GM and non-GM production and marketing the results depend crucially on the extent of systems, and (ii) the power of consumer senti- GMO rejection by consumers and the size of ment. The analysis has shown that when pro- the productivity gain foregone in comparison duction and marketing systems are segregated with the relative price premium obtainable on into GM and non-GM lines all the way from non-GM varieties. For some countries the primary crops through livestock feed to food development of segregated GM and non-GM processing, changing consumer attitudes food markets is a way of retaining access to towards GMOs will have substantial effects on important export markets if and only if the trade, production and prices not only for the non-GM characteristic can in fact be pre- crop sectors that benefit directly from the new served and verified throughout the marketing technology, but also for the sectors that use system at reasonable costs.

References

Buckwell, A., Brookes, G. and Bradley, D. (1999) Economics of Identity Preservation for Genetically Modiﬁed Crops. Final report of a study for Food Biotechnology Communications Initiative (FBCI). With contributions from Peter Barfoot, Stefan Tangermann and Jan Blom. Mimeo. Economic Research Service (ERS) (2000a) Biotech corn and soybeans: changing markets and the government’s role. Available at http://ers.usda.gov/whatsnew/issues/biotechmarkets/. US Department of Agriculture, Washington, DC. Economic Research Service (ERS) (2000b) Biotechnology: U.S. grain handlers look ahead. Agricultural Outlook. US Department of Agriculture, Washington, DC. Frandsen, S.E. and Nielsen, C.P. (1999) Derfor skal de GMO-frie varer mærkes [This is why non-GMO products should be labelled], Chronicle (in Danish), Jyllandsposten, 20 December. James, C. (1997) Global Status of Transgenic Crops in 1997. ISAAA Briefs No. 5. International Service for the Acquisition of Agri-biotech Applications, Ithaca, New York. James, C. (1998) Global Review of Commercialized Transgenic Crops: 1998. ISAAA Briefs No. 8. International Service for the Acquisition of Agri-biotech Applications, Ithaca, New York. James, C. (1999) Global Status of Commercialized Transgenic Crops: 1999. ISAAA Briefs No. 12: Preview. International Service for the Acquisition of Agri-biotech Applications, Ithaca, New York. Lewis, J.D., Robinson, S. and Thierfelder, K. (1999) After the Negotiations: Assessing the Impact of Free Trade Agreements in Southern Africa. TMD Discussion Paper No. 46. International Food Policy Research Institute, Washington, DC. Consumer - Chap 18 5/3/04 15:57 Page 229

Consumer Attitudes towards GM Foods 229

McDougall, R.A., Elbehri, A. and Truong, T.P. (eds) (1998) Global Trade, Assistance, and Protection: The GTAP 4 Data Base. Center for Global Trade Analysis, Purdue University, West Lafayette. Nelson, G.C., Josling, T., Bullock, D., Unnevehr, L., Rosegrant, M. and Hill, L. (1999) The Economics and Politics of Genetically Modified Organisms: Implications for WTO 2000. With Julie Babinard, Carrie Cunningham, Alessandro De Pinto and Elisavet I. Nitsi. Bulletin 809. College of Agricultural, Consumer and Environmental Sciences, University of Illinois at Urbana-Champaign. Nielsen, C.P., Robinson, S. and Thierfelder, K. (2000) Genetic Engineering and Trade: Panacea or Dilemma for Developing Countries. TMD Discussion Paper No. 55. Trade and Macroeconomics Division, International Food Policy Research Institute, Washington, DC. Nielsen, C.P., Thierfelder, K. and Robinson, S. (2001) Consumer attitudes towards genetically modified foods: the modelling of preference changes. Working paper No. 01/2001. Danish Institute of Agricultural and Fisheries Economics (SJFI), Copenhagen. OECD (1999) Modern Biotechnology and Agricultural Markets: A Discussion of Selected Issues and the Impact on Supply and Markets. Directorate for Food, Agriculture and Fisheries. Committee for Agriculture. AGR/CA/APM/CFS/MD(2000)2, OECD, Paris. Pray, C.E., Ma, D., Huang, J. and Qiao, F. (2000) Impact of Bt cotton in China. Paper presented at the International Food Policy Research Institute (IFPRI), 9 May. Runge, C.F. and Jackson, L.A. (1999) Labelling, Trade and Gentically Modified Organisms (GMOs): A Proposed Solution. Working Paper WP99–4. University of Minnesota, Center for International Food and Agricultural Policy. Consumer - Chap 18 5/3/04 15:57 Page 230 Consumer - Chap 19 Index 5/3/04 15:57 Page 231

Index

activism against GM foods 132, 175, 176 beneﬁts of GM foods, consumer perceptions 98, 171 age and attitudes 106, 126–127 advantages judged as weak 178, 180 AGE (applied general equilibrium) models BEUC (European Consumers’ Organization) 177–178 189–190, 212 bias Negishi format 191, 202–203 in attitude surveys 24 shift from animal protein to novel protein in willingness-to-pay studies 25 foods 191–192 bovine growth hormone (somatotropin) see rBGH balance equations 205–206 (recombinant bovine growth hormone) budget constraints 195–196 breakfast cereals 93–94, 122, 124, 126–127 data and scenarios 196 environmental quality 195 objective and utility functions 192–193, Canada 205 extent of biotechnology use 210 production functions 194–195, 205 labelling costs 42 results 196–198 cereals sensitivity analysis 198–199, 207–208 breakfast cereals 93–94, 122, 124, 126–127 symbols used 204 grains agriculture, attitudes to modernization of 182–183 model absorption results 226–228 AIDS (almost ideal demand system) models 27–29 model of trade and production 215–218 allergies and GM foods 100 model price and trade results 220–223 animal protein, model of shift away to novel protein model production results 224–226 foods see AGE (applied general world production and trade 212–215 equilibrium) models certainty equivalence 64–65 animal vs. plant modiﬁed genes 105 closed-ended questioning 89 Argentina: extent of biotechnology use 210 Codex committees (WTO) 112 attitude surveys see surveys, attitude Colombia auctions see experimental auction studies attitudes to food safety and science 156–157 Australia, labelling costs 42 attitudes to GM food 157–160 Australia New Zealand Food Authority (ANZFA) 73 consumer knowledge 157 Australia New Zealand Food Standards Council commodities see markets (ANZFSC) 73 computable general equilibrium models absorption results 226–228 consumer preferences 219–220 bans (voluntary) on GM food 1 market segregation 215–217 models of market response 3–7 price and trade results 220–224 behaviour probability scale 55 production results 224–226

231 Consumer - Chap 19 Index 5/3/04 15:57 Page 232

232 Index

concept exposure 55–56 organized opposition to GM foods 174–175 consumer organizations 177–178 public research organizations 177 consumer surplus 63–65 public trust in authorities 182 effects of labelling 19 functional foods 163–167 consumption, direct vs. indirect 105 contigent valuation 69, 70, 88–89, 119–121 corporations, multinational see multinationals; gender and attitudes 74, 79–80, 106–107, 126, consumer perceptions 171 genes, plant vs. animal 105 genetic engineering in food production, current demand curves 12, 64 status and future market structures and segregation costs 86 210–211 desirability, characteristics of Germany snack products 56–59 consumer trends in food market 164 developing countries 155, 157 functional foods 163–167 discounts for GMO foods 86 governments, attitudes to 136, 137–138, 146, 156–157 ECAM (EC agricultural model) 189–190 and willingness to purchase GM food 149, education and attitudes 78, 107 150, 151 environment Greenpeace 174, 178 impact of move from animal protein to novel growth hormone, bovine see rBGH (recombinant protein foods see AGE (applied bovine growth hormone) general equilibrium) models GTAP (global trade analysis project) model 189, modelling consumer concerns 67–68 212 risk and benefit perceptions 92, 111, 122, 126, 157, 171–172 see also opposition to GM foods Hawthorne effect 26 ethics and morals 126 health and consumer concerns 92, 100, 111 functional foods 163–167 modelling consumer concerns 66–67 modelling consumer concerns 66, 67 Europe risk and benefit perceptions 63, 92, 98, 100, attitude surveys 24–25, 73–74, 111, 118, 122, 126, 156 172–174, 176 USA vs. Europe 132 experimental auction studies 25–26 see also opposition to GM foods hostility to biotechnology industry 175 hedonic pricing 68–69 image in United States 178, 180 Hicksian demand curves 12, 64 labelling costs 73 Iceland (retailer) 1 labelling policies 42, 96 novel protein foods, model of introduction see AGE (applied general equilibrium) identity preservation systems 211 models models 216–217 willingness-to-pay studies 25 income, disposable 80 see also individual countries Info-conso Network (Greenpeace) 178 European Consumers’ Organization (BEUC) information 177–178 effect on purchase decisions 45–47 expenditure share equations 31–35 importance 127 experimental auction studies 69–70 sources 135–136, 145–146, 147, 157, Europe 25–26 175–178 limitations 26 see also media random nth-price auctions 44 International Consumers’ Organization 177 United States 26n, 43–49 International Food Council 145 interviews in shops 112–113, 143, 156 Food and Drug Administration (US FDA) 85 in simulated test marketing 54–55 food insecurity 157, 159 Ireland 131, 132 France attitude surveys 143–144 influence of media 174 consumer knowledge 145–146 Consumer - Chap 19 Index 5/3/04 15:57 Page 233

Index 233

Italy segregated agricultural markets 209 anti-GM protests 132 computable general equilibrium model attitude surveys 132–140 215–217 organic food 133–134 soybean futures 2 reading of labels 134 Marshallian demand curves 12, 64 media modelling effects of coverage 29 Japan role in opinion formation 174, 175–176 attitude surveys 111–112, 118 see also information interviews in shops 112–113 men see gender and attitudes labelling policies 112 methods: testing for GMO content 211 willingness-to-pay studies 113–114 milk use of bovine growth hormone (somatotropin) Juster, Thomas 55 see rBGH (recombinant bovine growth hormone) models knowledge, consumer 111 AGE (applied general equilibrium) models see Colombia 157 AGE (applied general equilibrium) Ireland 145–146 models Italy 135–136 AIDS (almost ideal demand system) models New Zealand 77 27–29 Norway 98 computable general equilibrium models see organic foods 133–134 computable general equilibrium United Kingdom 93 models United States 93, 98, 127, 135–136, 144–146 conditional logit models 15 effects of voluntary labelling 9–13 logistic model 120 labelling of market response to voluntary bans 3–7 attitude surveys 73–74, 93, 136–137, 147, multiregional models 215–217 148, 158–159 ordered logit model 75–76 costs 42–43, 73, 84 random utility models 11–13, 119–120 effects on consumer surplus 19 limitations 27 effects on US milk purchases 17–21 in willingness-to-pay studies 66–68 factors affecting demand 74–75 Monsanto (Italy) 132 harmonization112 morals see ethics and morals lack of consumer response, The Netherlands multinationals, consumer perceptions 92–93 35–37 mandatory vs. voluntary 36, 42–43, 84–85 experimental auction studies 43–49 Negishi AGE model format 191–196, 202–203 national policies 42, 73, 96–97 Netherlands, The AIDS model of response to GM labelling 27–29 and price insensitivity 100 attitude surveys 35–36 theoretical models 9–13, 27–29 demand system variables 30 usefulness or otherwise 74 expenditure share equations 31–35 labels, reading of 134, 144, 145, 158 price elasticities 34–35 lifestyle and attitudes 74, 79, 80 New Zealand limited market experiments 27n attitude surveys 75–81 R Litmus simulated test marketing system 54 labelling costs 42, 73 logistic model 120 Norway attitude surveys 97–100 interviews in shops 112–113 mail surveys 121–122 labelling policies 96 maize, StarLink(TM) 145 willingness-to-pay studies 100–108, 113–114 markets novel protein foods, model of acceptance see AGE GM crops in world production and trade (applied general equilibrium) models 212–215 niche market behaviour 3 response to information 2 oil, soybean 102–104, 106–108 empirical models 3–7 oil, vegetable 122, 124, 126–127 Consumer - Chap 19 Index 5/3/04 15:57 Page 234

234 Index

oilseeds quality, perception of 62–63 model of trade and production 215–217 design of experiments 217–219 results 220–228 race and attitudes 127 world production and trade 212–215 random nth-price auctions 44 online surveys 88 random utility models 11–13, 101, 119–120 open-ended questioning 89 limitations 27 opinion polls see surveys, attitude rational agent theory 64–65 opposition to GM foods rBGH (recombinant bovine growth hormone) 85 and future markets 211–212 theoretical model of labelling effects 10–13 Monsanto protests (Italy) 132 US milk purchases by organizations and groups 174–175, 176 data characteristics 15–16 reasons for opposition 179, 184 data collection and organization 13–14 as reflection of wider social concerns demand consumer surplus effects 19 181–184 econometric analysis 15 and risk perception by public 180–181 price elasticities 19–20 US perception of European attitudes 178, 180 regression results 16–20 world statistics 170 Vermont 118 option price 65 use in USA 9 option value 65 regulations and consumer confidence 84 ordered logit model 75–76 religion, influence of see ethics and morals organic foods 133–134, 143, 144 retailers removal of GM ingredients 1 risk, perception of 65, 159, 172–174, 180–181 payment card questioning 89 see also environment; health; opposition to permission-based surveys 87 GM foods pesticides, reduced use 98, 138, 148–149, 158 Philippines, The, labelling costs 42–43 plant protein: model of acceptance see AGE Sainsbury (supermarkets) 1 (applied general equilibrium) models salmon 104, 105, 106–108, 122, 124, 126–127 plant vs. animal modified genes 105 San Luis Obispo Country, CA (USA) 54, 132, 143 polls, opinion see surveys, attitude science, attitudes to 132–133, 143, 144, 156, pork: model of shift away to plant protein see AGE 182–184 (applied general equilibrium) models scientists, opinions of 177, 180 preferences, consumer, modelling 218–220, segregation 209 226–228 costs and demand 86 price elasticities simulated test marketing The Netherlands 34–35 concept exposure 55–56 and rBGH in milk 19–20 interview process 54–55 prices methodology 53–54 determination of premium for non-GMO crops systems 54 85–86 Slutsky’s equation 64 and willingness to pay for non-GM products snack products 55–59 122, 157, 159 somatotropin, bovine see rBGH (recombinant pricing, hedonic 68–69 bovine growth hormone) producers soybeans attitudes of consumers 136, 137–138, 146, futures prices 2 157 empirical models of movements 3–7 Europe 175 oil 102–104 and willingness to purchase GM food standards, international 112 151, 152 StarLink(TM) maize 145 communication to public 176–177 stated choice method 100 PROFETAS (PROtein Foods, Environment, surveys, attitude 62, 73–74 Technology And Science, NL) 189 Colombia 156–160 protein, model of switch from animal to plant Europe 24–25, 118 protein see AGE (applied general international 170–171 equilibrium) models Ireland 143–144, 146–153 Consumer - Chap 19 Index 5/3/04 15:57 Page 235

Index 235

Italy 132–140 limited market experiments 27n Japan 118 organic food 133–134 limitations 24 perception of European opposition 178, 180 methodology 86–88, 97–98 reading of labels 134, 144, 145 The Netherlands 35–36 recombinant bovine growth hormone in milk Norway 97–100 see rBGH (recombinant bovine Taiwan 118 growth hormone) United Kingdom 118 San Luis Obispo Country, CA 54, 132, 143 United States 97–100, 118, 132–140, simulated test marketing 55–59 146–153, 171–172 willingness-to-pay studies 87–88, 91–94, see also willingness-to-pay studies 100–108, 121–128 USDA (US Department of Agriculture) effects of announcements on markets 4, 5 Taco Bell recall 145 utility theory 64–65 Taiwan, attitude surveys 118 technology, home use of 132 telephone surveys 97–98 valuation, contingency 69, 70 television as information source 135 variation, compensating and equivalence 64 tests for GMO content 211 trade, world contribution of GM crops 212–213 trade dependence 213–214 welfare analysis, Negishi format AGE model trade flows 214–215 191–196 tradition, influence of 114 willingness-to-pay studies trust, consumer breakfast cereals 93–94 in food producers 77, 79, 136, 149, 151 and consumer surplus 63–65 in government and regulatory agencies 84, demographic and lifestyle influences 79–80, 136, 149, 150, 151, 156–157 92–93, 124, 126 negative influences 174, 181–182 empirical measurements 68–70 Europe 25 Japan 113–114 uncertainty 65 limitations 25 United Kingdom methodology 75–78, 88–89, 100–102, attitude surveys 88, 118, 132 119–122 voluntary bans on GM food 1 models 66–68, 75–76, 89–90 willingness-to-pay studies 88, 91–94 New Zealand 75–81 United States Norway 100–108, 113–114 attitude surveys 74, 87–88, 97–100, 132–140, research applications 70 143–144, 171–172 theoretical background 62–63 consumer confidence in food supply 131 United Kingdom 88, 91–94 consumer knowledge 144–146 United States 87–88, 91–94, 100–108, experimental auction studies 26n, 43–49 122–128 extent of biotechnology use 131, 210–211 Wirthlin Group Quorum Surveys 145 labelling policies 42 women see gender and attitudes and consumer response 48–49 World Trade Organization (WTO) 112 Consumer - Chap 19 Index 5/3/04 15:57 Page 236