JOBNAME: hortte 19#1 PAGE: 1 OUTPUT: November 20 01:32:19 2008 tsp/hortte/176989/01138
an alternative crop to growers and a ‘‘new’’ product to consumers. Chestnuts have favorable nutri- tional characteristics. High moisture and low oil content make chestnuts Production virtually fat free. Chestnuts have a high concentration of complex carbo- hydrates, a low glycemic index, are cholesterol free, and contain only and Marketing one-third the calorie content of pea- nuts (Arachis hypogaea) and cashews (Anacardium occidentale) and as much ascorbic acid as their equivalent Reports weight of lemons (Citrus limonium). Chestnut flour is sweet, nutty, and gluten-free, the latter making it Exploratory Assessment of Consumer appropriate for consumption by peo- ple with celiac disease (Chestnuts Preferences for Chestnut Attributes in Missouri Australia Inc., 2008; Erturk et al., 2006; University of Missouri Center 1,3 2 for Agroforestry, 2006). Francisco X. Aguilar , Mihaela M. Cernusca , and The market for edible chestnuts Michael A. Gold2 has considerable potential for increase in production and demand given growing consumer interest in alter- ADDITIONAL INDEX WORDS. Castanea spp., conditional logit model, conjoint native and healthy foods (Gold et al., analysis, chestnut market development 2004a). Consumers across Europe, SUMMARY. This article explores consumers’ preferences for different chestnut Australia, New Zealand, and the (Castanea spp.) attributes and studies differences across potential market segments. United States have an increasing The study was conducted between 2003 and 2007 during the Missouri Chestnut interest in chestnuts (Kelley and Roast festival. The festival, held annually in October during the chestnut harvest Behe, 2002). Current demand for season, is one of mid-Missouri’s premier family-oriented events. A longitudinal chestnuts in the United States study completed among festival visitors in 2003, 2004, and 2006 to identify exceeds national production, which chestnut characteristics that influence purchasing decisions was complemented with is offset by imports. According to the a conjoint analysis in 2007. The conjoint analysis used a conditional logit model to U.S. Department of Agriculture investigate responses from pairwise product profile comparisons. The attributes (USDA), chestnut imports have investigated include chestnut size (small, medium, and large), price ($3, $5, and $7 per pound), production process (organic and conventional), and origin (Missouri, grown steadily in value since 2003, United States, and imported). Results suggest a strong preference for locally and reaching $11.6 million in 2006 U.S.-grown compared with imported chestnuts. Local growers that provide the (USDA, 2007). Chestnuts are mostly market with medium-size chestnuts that carry organic certification could command imported from Europe (82% of total a market premium compared with imported/nonorganic certified chestnuts. imports in 2006). According to the USDA (2007), the main exporters to the United States are Italy (55%), dible chestnuts were well chestnut blight (Cryphonectria para- China (23%), and Korea (15%). known in the United States sitica) from Asia in the early 1900s In the United States, chestnut Ebefore their near extinction at almost eradicated american chestnut cultivation can be an attractive enter- the beginning of the 20th century. trees (Castanea dentata)(Anagnosta- prise due to high product demand, The accidental introduction of the kis, 1987). Research efforts are cur- favorable prices, and relatively low rently being conducted in Missouri, initial investment requirements (Gold This work was funded through the University of Michigan, Tennessee, and other states et al., 2006). However, aside from a Missouri Center for Agroforestry under cooperative agreement AG-02100251 with the USDA ARS Dale to identify improved cultivars and to small number of successful enter- Bumpers, Small Farms Research Center (Boonville, develop management practices suit- prises, the U.S. chestnut industry is Arkansas). The results presented are the sole responsibility able for commercial chestnut produc- in its initial stages of development, of the P.I.s and/or MU and may not represent the policies or positions of the ARS. tion. These initiatives aim to revitalize with the majority of producers in We thank Dr. Carla Barbieri and Dr. Laura McCann, the chestnut industry and to provide business for less than 10 years and University of Missouri, Columbia, for their review and insightful comments to an earlier version of this manuscript. 1Department of Forestry, University of Missouri, 203 ABNR Building, Columbia, MO 65211 Units 2Center for Agroforestry, University of Missouri, 203 To convert U.S. to SI, To convert SI to U.S., ABNR Building, Columbia, MO 65211 multiply by U.S. unit SI unit multiply by 3Corresponding author. E-mail: aguilarf@missouri. 2.54 inch(es) cm 0.3937 edu. 0.4536 lb kg 2.2046
216 • January–March 2009 19(1) JOBNAME: hortte 19#1 PAGE: 2 OUTPUT: November 20 01:32:19 2008 tsp/hortte/176989/01138
just beginning to produce at a com- and that could help in fostering prod- survey that was administered over 3 mercial scale (Gold et al., 2006). uct demand and developing new mar- years (2003, 2004, and 2006). Be- At the other end of the market kets. Specifically, this research aimed to cause very similar results were value chain, U.S. consumers are unfa- 1) identify how much attributes such as obtained in 2003 and 2004, 2005 miliar with chestnuts, are not fully size,price,quality,locallygrown,and was skipped, but the survey was re- aware of their healthful properties, in nutrition-diet-health, influence con- taken in 2006. The selected attributes what form and where to buy them, or sumers’ decisions to purchase chest- (price, quality, locally grown, and how to prepare them (Gold et al., nuts, 2) generate information on the nutrition-diet-health) were identi- 2005). In regions surrounding success- relative importance that consumers fied after a review of the literature ful growers or active research centers, place on particular chestnut character- (Jekanowski et al., 2000; Frank et al., consumers’ familiarity with chestnuts, istics (price, size, production process, 2001). The influence on consumer and knowledge and frequency of con- and label of origin), 3) determine preference for each attribute was sumption is continually increasing salient product characteristics that cur- evaluated on a five-point Likert-type (Carlman, 2007; Cernusca et al., rent and future chestnut producers scale ranging from ‘‘not at all’’ (1) to 2008). This situation represents a mar- should evaluate and adopt for effective ‘‘very strongly’’ (5). Each participant ket opportunity for expansion of chest- marketing purposes, and 4) identify took between 5 and 10 min to nut production in the United States differences in product preferences by complete the questionnaire. No that demands a better understanding demographic characteristics such as incentive in the form of monetary of consumer preferences. urban and rural consumers and or in-kind reward was offered to study Market knowledge is fundamen- gender. participants. tal for the successful establishment of CONJOINT ANALYSIS. CA was agroforestry enterprises (Gold et al., Materials and methods used to design and analyze surveys 2004a). Gold et al. (2006) consider A two-step research approach was distributed during the Missouri that one of the biggest barriers to used to assess consumer preferences Chestnut Roast festival in 2007. The success in the chestnut industry is the for chestnut attributes. The first step CA helped validate and quantify the lack of information for producers, addresses objective 1 and it involved results obtained in the previous sur- retailers, and consumers. Over the the study of perceptions among a veys and provided responses to exam- last two decades there have been a sample of consumers over 3 years. ine the remaining research objectives. limited number of studies focused on The second step corresponds to the CA is based on the premise that improving understanding of chestnut application of a conjoint analysis (CA) consumers can judge the value of markets (Gold et al., 2004b, 2006). A for the exploratory study of chestnut several hypothetical products that study by Wahl (2002) assessed the attributes. All information reported are described by different attributes interest of upscale restaurant chefs in in this article was gathered during the that make up product profiles and value-added chestnut products and annual Missouri Chestnut Roast fes- choose the one that gives them the found that product freshness and tival in New Franklin, MO. The fes- most utility (Green and Srinivasan, quality were the main attributes that tival held annually in October during 1978; Carson et al., 1994). influenced consumers’ and chefs’ the chestnut harvest season is one of Under a random utility frame- interest in purchasing food products mid-Missouri’s premier family-ori- work, a consumer (i)hasautility and ingredients. Wahl (2002) recom- ented events. The one-day event is function defined over an array of J mended that growers should pro- not dedicated solely to chestnuts. It is possible choices of chestnut products. mote chestnuts on the basis of an outstanding opportunity to intro- It is further assumed that a consumer quality, uniqueness, and local pro- duce participants to the broad range utility function contains a deterministic duction. Another study developed of possibilities and benefits that can component and a set of unobservable by the Midwest Nut Producers Coun- result from implementing agrofores- variables that introduce a random error cil and Michigan State University try practices or consuming specialty element (Hanemann, 1984). For the identified marketing opportunities products. Festival participants include ithconsumerfacedwithatotalofJ for chestnuts in upscale restaurants consumers and nonconsumers of choices, the assumed utility of choice j in Michigan (Kelley and Behe, 2002; chestnuts. As a case in point, 67% of can be represented by:
Smith et al., 2002). Chefs that par- participants in 2003, 46% in 2004, Uij = Vij + eij ; j =1; .; J [1] ticipated in the study preferred peeled 45% in 2006, and 34% in 2007 had chestnuts and incorporated them in a never consumed a chestnut before . variety of dishes. Gold et al. (2004b, attending the festival. In addition, Vij = xi1b1 + xi2b2 + xij bj ;: 2005) also reported that U.S. con- in 2007, about 75% of partici- sumers prefer to buy chestnuts from pants attended the festival for the first where the Vs represent utility values grocery stores or farmers markets and time. that summarize the preference of the that organic, brand name, and culti- STUDY OF CHARACTERISTICS ith consumer for the jth product var chestnuts can help capture price THAT INFLUENCE PURCHASING alternative. The xs specify values of premiums. DECISIONS. A question about how chestnut attributes, and the bs denote This study was designed to fur- much each of the following chestnut importance weight parameters (part- ther current knowledge of consumer attributes (i.e., price, quality, locally worths) that were used to determine preferences for chestnuts. There was a grown, and nutrition-diet-health) the specific importance placed by need to identify salient product char- influence people’s decision to pur- consumers on each attribute. Vector acteristics that influence consumption chase chestnuts was included in a eij,j=1,.J captures unobservable
• January–March 2009 19(1) 217 JOBNAME: hortte 19#1 PAGE: 3 OUTPUT: November 20 01:32:20 2008 tsp/hortte/176989/01138
PRODUCTION AND MARKETING REPORTS
variables affecting consumer prefer- bell peppers (Capsicum annuum) from was identified. Also, no specific ences, and reflects the idiosyncrasy of (Frank et al., 2001), peanuts (Nelson definition to the concept of organic the individual in taste for the different et al., 2005), tabletop Christmas production was included in the sur- product alternatives. Under the as- trees (Behe et al., 2005), mandarins vey, but if asked by participants, it sumption that random errors are (Citrus reticulata) (Campbell et al., was indicated that chestnuts were independent and identically distrib- 2006), and asparagus spears (Aspar- produced, processed, and certified uted, this model specification corre- agus officinalis) (Behe, 2006). to be consistent with USDA national sponds to McFadden’s conditional The first step in a CA study is organic standards that avoid most logit model (McFadden, 1974, 1986; the identification of key attributes synthetic chemical inputs. Partici- Punj and Staelin, 1978). and their appropriate levels. To quan- pants were presented with a list of According to Arnold et al. (1981) tify the attributes considered in the product profiles as depicted in Fig. 1. the conditional logit model has sev- 2003–06 studies, the following prod- No photographs or other visual aids eral advantages, one of which is good uct attributes and levels were selected: were used in the conjoint analysis model approximation even when price ($3, $5, and $7 per pound), because all visitors were offered sam- working with small samples (50 ob- chestnut size [small (<1 inch in diam- ples of chestnuts as they entered the servations or more). Traditional con- eter), medium (>1 inch but <1-1/4 festival. ditional logit includes product-specific inch diameter), and large (>1-1/4 Differences in preferences among attribute information in the xs, (e.g., inch diameter)], label of origin (pro- groups are explored in the mixed- product price, size, etc.), but cases duced in Missouri, produced in the effects conditional model by allowing that include respondent-specific varia- United States, and imported), and for interactions with the product attri- bles (e.g., region, gender, age, etc.) production process (conventional and butes and the socioeconomic variables can also be incorporated. The use of organic; Table 1). The second column presented in Table 2. product- and respondent-specific vari- in Table 1 provides a brief description Preference elicitation in CA was ables simultaneously results in what is of the attribute level and the last done using the full profile pairwise commonly known as a mixed-effects column indicates expected signs for comparison method as described by conditional model (Hoffman and the utility partworths. The reader Hair et al. (1998). The latter was Duncan, 1988; Long and Freese, should note that no attribute regard- chosen primarily because it adds 2006). Under this model, Eq. [1] ing nutrition-diet-health was in- more realism to the product selection becomes cluded. This is the case because all process. The profiles of hypothetical
U ij = V ij + Si + eij [2] chestnuts products are assumed to products were generated using the provide the same level of nutrition. Bretton-Clark designer program fol- where Si captures respondent-specific No particular country or region from lowing a fractional design. This pro- demographic information. where the chestnuts might be imported gram produces a subset of hypothetical In this study, various models were tested. First, a fixed model with only attribute-specific characteristics Table 1. Chestnut attributes, corresponding attribute levels, and expected was estimated. Differences in part- partworth signs used in the conjoint analysis of consumer preferences worth parameters were then explored for chestnuts in a survey distributed in 2007 during the Missouri between place of residence (urban Chestnut Roast festival. and rural), gender, age, and income Product attributes Expected for market segmentation purposes. and attribute levels Descriptionz partworth sign Brown (2003) found that rural resi- Chestnut size dents in Missouri are more willing to Small <1 inch diameter (base level variable) NAy seek out locally produced food than Medium >1 inch but <1-1/4 inch diameter + are urban residents, but household (1 = medium size, 0 = other) location did not significantly influ- Large >1-1/4 inch diameter + ence the price consumers are willing (1 = large size, 0 = other) to pay for locally grown food. Aguilar Pricex and Vlosky (2007) identified gen- $3.00/lb Price per pound (continuous variable) – der and income differences when elic- $5.00/lb iting willingness to pay premiums $7.00/lb for environmentally certified wood Production process products. Conventional Conventional use of agrochemicals NA CA helps determine a set (base level variable) of partial utilities (partworths) for Certified organic Certified to organic production standards – the individual attributes that are con- (1 = organic, 0 = other) sistent with the respondent’s overall Label of origin preferences (Green and Wind, 1975). Imported Imported chestnuts (base level variable) NA The partworths identify the relative Missouri Produced in Missouri (1 = Missouri, 0 = other) + importance that consumers place on United States Produced in the U.S. (1 = U.S., 0 = other) + each particular chestnut attribute. z1 inch = 2.54 cm. CA has been applied to a number yNot applicable for base level variables. of horticultural products, including y$1.00/lb = $2.2046/kg.
218 • January–March 2009 19(1) JOBNAME: hortte 19#1 PAGE: 4 OUTPUT: November 20 01:32:21 2008 tsp/hortte/176989/01138
The generated profiles were included in a survey questionnaire together with questions about fre- quency of consumption, familiarity with chestnuts, and demographic information (i.e., age, gender, marital status, level of education, total house- hold income, and ethnic group). For the CA, respondents were asked to review the 10 pairs of hypothetical chestnut products and to select one product (A or B) that they would be most likely to purchase (Fig. 1). A note also mentioned that all prod- Fig. 1. Example of the conjoint analysis pairwise instrument used to gather ucts meet a standard of quality consumer preferences for chestnuts’ attributes (i.e., size, price, production that guarantees freshness, good fla- process, and label of origin) during the Missouri Chestnut Roast festival vor, and nuts free of fungal- or in 2007 (small = <1 inch diameter, medium = >1 inch but <1-1/4 inch bacteria-caused damage. The survey diameter, large = >1-1/4 inch diameter; 1 inch = 2.54 cm, $1.00/lb = $2.2046/kg). was pretested on 20 subjects and received required human subject- Table 2. Socioeconomic variables, corresponding levels, and expected partworth approval before being distributed to signs used in the conjoint analysis of consumer preferences for chestnuts in a participants. survey distributed in 2007 during the Missouri Chestnut Roast festival. DATA COLLECTION AND ANALYSIS. Expected Respondents for each of the surveys Socioeconomic variables partworth were selected following the tailored and corresponding levels Description sign design procedure for in-person deliv- ered questionnaires recommended by Urban and rural consumers z Dillman (2000). Interviewers ran- Urban setting Lives in a city with a population of at least NA domly selected festival attendees, 10,000 (base level variable) explained the purpose of the study, Rural setting Lives in a town or village with a population NA and asked for their participation. of less than 10,000 (1 = rural, 0 = other) Those who agreed to participate com- Annual income Total annual household income: pleted the survey on site. The sample Less than $24,999 Base level variable NA generated by this method is a con- $25,000–$34,999 1 = $25,000 £ income £ $34,999, 0 = other + venience sample and implies caution $35,000–$49,999 1 = $35,000 £ income £ $49,999, 0 = other + in generalizing the results to the $50,000–$74,999 1 = $50,000 £ income £ $74,999, 0 = other + whole population of consumers. For $75,000–$99,999 1 = $75,000 £ income £ $99,999, 0 = other + the surveys distributed in 2003–06 $100,000 or more 1 = income ‡ $100,000, 0 = other + and the first part of the survey dis- Education tributed in 2007, frequencies, cross High School Has completed high school education + tabulation, and correlations were cal- (base level variable) culated to analyze data. Technical school Has completed training at a technical school + The information presented in (1 = technical school, 0 = other) the product profiles corresponded to College Holds a college degree (1 = college, 0 = other) + the chestnut attributes (x). Data were Graduate school Holds a graduate degree +/– entered as a series of dummy variables (1 = graduate, 0 = other) representing the different categories Other Other degrees/diplomas composing a chestnut profile. The (1 = others, 0 = no others) only variable not categorical is price, Gender which is continuous, and the actual Male Male respondent (base level variable) NA price presented on the survey was Female Female respondent (1 = female, 0 = other) + entered in the data matrix. For each z Not applicable for base level variables. of the three categorical variables (chestnut size, production process, and label of origin), the base level profiles based on the attribute levels Bretton-Clark designer program is indicated in Table 1 was left out as the provided by the researcher. The pro- presented in Fig. 1. Eighteen profiles point of comparison in the analysis, gram minimizes the confounding of for a total of nine comparisons were and a ‘‘1’’ or ‘‘0’’ was entered if the attribute main effects by selecting a produced. An additional 10th pair profile contained or lacked informa- subsample of orthogonal product was generated by the authors to com- tion for a corresponding attribute combinations (Harrison and Sambidi, pare attribute levels that were omitted level. As such, after leaving out the 2004). An example of two of the nine in the other nine comparisons and base level in our data, we obtained pairwise profiles generated by the deemed relevant. two categories for chestnut size
• January–March 2009 19(1) 219 JOBNAME: hortte 19#1 PAGE: 5 OUTPUT: November 20 01:32:23 2008 tsp/hortte/176989/01138
PRODUCTION AND MARKETING REPORTS
(medium and large), one category strongly and very strongly influenced of age (10%). Respondents with a for production process (certified or- by quality in their decision to pur- total household income between ganic), and two categories for origin chase chestnuts), followed by locally $25,000 and $50,000 valued locally (produced in Missouri and produced grown (56%), nutrition-diet-health grown products more than respond- in the United States). For the case of (54%), and price (23%). Some of these ents in other income ranges. A cor- the mixed-effects conditional model, results are in accordance with those relation analysis revealed a small, a set of binary variables were created reported by Brown (2003). Brown negative significant correlation between to capture the respondent character- found that quality and freshness were price and age [r (425) = –0.141, P < istics as described in the socioeco- the most important concerns when 0.01], the older the respondents, nomic variables presented in Table shopping for fresh fruits and/or the less influenced they are by price 2, and each one was multiplied by vegetable for 82% of consumers in when making decisions to purchase the product attribute levels to gener- southeast Missouri, while price was chestnuts. ate interaction terms that allow for important for only 8%. CA STATISTICS (2007). A total the study of differences among differ- A crosstab analysis of the 2006 of 104 surveys were collected in ent group segments. data revealed relationships between 2007, of which 54% were completed price and age of participants [c2 (8) = by women and 46% by men. A lower Results and discussion 15.66, P < 0.05]; locally grown and number of surveys reflects inclement C HARACTERISTICS THAT age [c2 (8) = 24.01, P < 0.01], locally weather throughout the 2007 festi- INFLUENCE PURCHASING DECISIONS grown and income [c2 (12) = 23.36, val. Because of the limited sample size (2003–06 SURVEY). We collected P < 0.05], nutrition-diet-health and available for analysis, this research is 232 questionnaires in 2003, 217 in age [c2 (8) = 28.19, P < 0.001], and deemed exploratory. Regarding age, 2004, and 487 in 2006. In 2003, nutrition-diet-health and income [c2 respondents were classified in six cat- quality (69% of respondents were (12) = 22.06, P < 0.05]. Participants egories. Seven percent were under 25 strongly and very strongly influenced over 55 years of age (21%) tended to years old, 12% were between 26 and by quality in their decision to buy be less influenced by price than people 35 years of age, 13% were in the 36- to chestnuts) and nutrition-diet-health under 35 years of age (14%) or people 45-year range, 29% were in the 46- to (55%) were the most important between the ages of 36 and 55 years 55-year range, 24% were in the 56- attributes mentioned. Price was listed (10%). Participants under 35 years of to 65-year range, and 15% were over as the least important attribute (26%) age (14%) were less influenced by 65 years old. Regarding place of (Gold et al., 2004b). Similar results locally grown than participants from residence, 60% of the study partici- were obtained from the 2004 data 36 to 55 years old (28%) or older than pants lived in towns or cities with a (Gold et al., 2005). In 2006, an 55 years of age (23%). Participants population of at least 10,000. Forty increase was noted in the preference between 35 and 55 years old (29%) percent came from rural areas and for locally grown chestnuts. As shown were more influenced by nutrition- towns with less than 10,000 people. in Fig. 2, quality remained the top diet-health when making purchase Twenty-eight percent of respondents attribute (72% of respondents were decisions than people under 35 years had high school or technical school education, 39% were college edu- cated, and 30% had a graduate degree. Thirteen percent of survey partici- pants had a total household income of less than $25,000, 21% between $25,000 and $50,000, 32% between $50,000 and $100,000, and 20% over $100,000 per year. In terms of ethnicity, 95% were Caucasian, 4% were Asian, and 1% were Hispanic. Compared with the demographic characteristics of Missouri, the fol- lowing population segments were overrepresented: people between 46 and 55 years old (53% for our sample vs. 26% for the Missouri population), people with more than a high school education (69% vs. 31%), the Caucasian population (95% vs. 83% in Missouri), and people representing households with more than $100,000 annual in- come (20% vs. 13%), while people with an annual income under $50,000 were underrepresented (34% in our sample Fig. 2. Attributes that influence consumer decisions to purchase chestnuts as compared with 57% in Missouri). resulted from surveys distributed between 2003 and 2006 to Missouri Chestnut Thirty-four percent of respond- Roast participants. ents have never eaten a chestnut, 43%
220 • January–March 2009 19(1) JOBNAME: hortte 19#1 PAGE: 6 OUTPUT: November 20 01:32:28 2008 tsp/hortte/176989/01138
consumed them once per year or less, and the base level, while an odds ratio size chestnuts were preferred over and 23% consumed chestnuts two to of 5 suggests that consumers like a small ones and the effect was about three times per year or more. Fifty- given attribute five times better than the same compared with medium-size two percent of respondents were not the base level. Table 3 also presents chestnuts (1.85). This finding trans- at all familiar with roasting chestnuts the results for the models that allow lates into recommendations for chest- and 51% had never bought chestnuts for market segmentation allowing nut producers to plant proven before the event. Compared with data preferences to differ between men chestnut cultivars instead of seedlings collected during previous annual Mis- and women, and urban and rural to obtain medium to large chestnuts, souri Chestnut Roasts festivals in settings. Additional models were consistent quality, and more con- 2003, 2004, and 2006, we noted a tested, but no statistical significant sistent yields (Hunt et al., 2006). continuous increase in the frequency differences between previous chest- Consumers also were 1.31 times of consumption from year to year nut consumers, age, income, and more likely to select chestnuts that (Cernusca et al., 2008). education groups were detected. are organically certified compared The estimated attribute part- These results of the latter are omitted, with conventionally produced chest- worth coefficients of the conditional as they provided little additional nuts. The most salient product at- logit models, along with P values and information. tribute was the label of origin. odds ratios, are presented in Table 3. Results of the fixed-effects model Information collected indicates that Odds ratios were obtained by expo- show that all product attributes were consumers were 10 times more likely nentiating the b coefficient and can be statistically significant at the a = 0.10 to choose chestnuts grown in Mis- interpreted as changes in the odds of a level and all but the large-size charac- souri, and five times more likely consumer choosing a product with a teristic were significant at a = 0.05. to select chestnuts grown in the given category over the base level. An Odds ratios indicate that consumers United States compared with odds ratio of 1, then, indicates that had a preference for medium-sized imported nuts. These results reflect there is no difference in consumer chestnuts that is 1.89 times higher the support for locally grown prod- preference between that attribute than the small-size base level. Large- ucts reported by Darby et al. (2006)
Table 3. Conditional logit partworth estimates of chestnut product attributes for fixed and mixed-effects conditional models derived from a conjoint analysis of consumer preferences for chestnuts performed in 2007 during the Missouri Chestnut Roast festival. Mixed-effects model Mixed-effects model (market segmentation for (market segmentation for Fixed-effects model gender differences)z rural differences)z Attributes Coefficient P > z Odds ratio Coefficient P > z Odds ratio Coefficient P > z Odds ratio Medium sizey 0.63780 0.009 1.8923 0.50202 0.159 1.6521 0.88191 0.007 2.4155 Large sizey 0.61460 0.080 1.8489 0.50137 0.337 1.6510 0.76953 0.095 2.1588 Organic production 0.26755 0.048 1.3068 –0.19533 0.348 0.8226 0.32761 0.066 1.3876 Produced in Missouri 2.33726 <0.001 10.3528 2.14514 <0.001 8.5433 2.34872 <0.001 10.4721 Produced in the U.S. 1.67419 <0.001 5.3345 1.48537 0.007 4.4166 1.60538 0.001 4.9798 Price –0.18729 <0.001 0.8292 –0.22653 <0.001 0.7973 –0.23925 <0.001 0.7872 Medium size · female — — — 0.27737 0.584 1.3197 — — — Large size · female — — — 0.22504 0.756 1.2524 — — — Organic production · female — — — 0.82183 0.004 2.2747 — — — Produced in Missouri · female — — — 0.45374 0.518 1.5742 — — — Produced in the U.S. · female — — — 0.41911 0.588 1.5206 — — — Price · female — — — 0.05017 0.429 1.0515 — — — Medium size · rural — — — — — — –0.64930 0.199 0.5224 Large size · rural — — — — — — –0.43131 0.550 0.6497 Organic production · rural — — — — — — –0.11114 0.690 0.8948 Produced in Missouri · rural — — — — — — –0.00749 0.991 0.9925 Produced in the U.S. · rural — — — — — — 0.15223 0.844 1.1644 Price · rural — — — — — — 0.13064 0.041 1.1396 n 2,000 1,940 1,980 Likelihood ratio c2 175.58 193.14 185.57 P >c2 <0.001 <0.001 <0.001 zA lower number of observations (n) in the mixed models is indicative of demographic information not provided by respondents. yMedium = >1 inch but <1-1/4 inch diameter, large = >1-1/4 inch diameter (1 inch = 2.54 cm).
• January–March 2009 19(1) 221 JOBNAME: hortte 19#1 PAGE: 7 OUTPUT: November 20 01:32:28 2008 tsp/hortte/176989/01138
PRODUCTION AND MARKETING REPORTS
in Ohio, Gallons et al. (1997) in ‘‘locally grown,’’ as this attribute was ganic certified chestnuts. In partic- Delaware, Jekanowski et al. (2000) showntobethemostimportantto ular, there will be a considerable in Indiana, and Patterson et al. survey respondents. competitive advantage for local growers, (1999) in Arizona. In a study con- as consumers give preference to dom- ducted in New England, Giraud et al. Conclusion estically and locally grown chestnuts. (2005) found that favorable attitudes Results of this and other studies toward local goods are positively (Gold et al., 2005; Cernusca et al., correlated with the probability of 2008) indicate that frequency of Literature cited purchasing the local good. chestnut consumption is increasing. Aguilar, F.X. and R.P. Vlosky. 2007. As would be expected, the model Quality, locally grown, and nutrition- Consumer willingness to pay price premi- suggests that as price increases, the diet-health were consistently per- ums for environmentally certified wood likelihood of purchasing chestnuts ceived as the most important attrib- products in the U.S. For. Policy Econ. 9: decreases. However, the relative utes influencing chestnut purchase 1100–1112. influence of price on consumer choice decisions. Three-year findings (2003, Anagnostakis, S.L. 1987. Chestnut blight: is much smaller than that of the other 2004, 2006) confirmed initial results The classical problem of an introduced attributes (0.83). indicating that consumers who par- pathogen. Mycologia 79(1):23–37. Statistically significant differen- ticipate in the Missouri Chestnut ces were obtained when allowing for Roast festival value product quality, Arnold, S.J., V. Roth, and D.J. Tigert. variability in respondent-specific pref- local production, and nutritional 1981. Conditional logit versus MDA in erences for two sociodemographic value, constantly ranking price as a the prediction of store choice. Adv. Con- sum. Res. 8(1):665–670. characteristics: gender and place of relatively lower priority attribute. residence. When consumers were These results support promotion Behe, B.K. 2006. Conjoint analysis classified according to their gender, focused on local production and qual- reveals consumers prefer long, thin aspar- there was evidence that women are ity (quality chestnuts are firm, heavy agus spears. HortScience 41(5):1259– more likely to select chestnuts certi- for their size, and free of external 1262. fied to organic standards. These defects) to receive a premium price. Behe, B.K., R.M. Walden, M.W. Duck, results are in accordance with findings The exploratory CA suggests B.M. Cregg, K.M. Kelley, and R.D. Line- of other studies (Harris et al., 2000). that medium and large chestnuts were berger. 2005. Consumer preferences for Harris explained these findings by the preferred over small ones by consum- tabletop Christmas trees. HortScience fact that women tend to be the pri- ers in Missouri. There were small 40(2):409–412. mary food shoppers of a household differences between medium and Brown, C. 2003. Consumers’ preferences and may be more aware of food issues. large sizes and a slightly higher pref- for locally produced food: A study in The results of the mixed-effects con- erence for medium- compared with southeast Missouri. Amer. J. Altern. Agr. ditional model controlled for place of large-size chestnuts. This is a finding 18(4):213–224. residence (urban and rural) diverge that actually matches consumer pref- Campbell, B.L., R.G. Nelson, R.C. Ebel, from those reported by Brown erences with agronomic reality. Chi- and W.A. Dozier. 2006. Mandarin attri- (2003) for the state of Missouri who nese chestnut (Castanea mollissima) butes preferred by consumers in grocery found that household location influ- cultivars that reach a medium size are stores. HortScience 41(3):664–670. enced preference for local products already proven producers and are and did not influence the price a recommended to be grown in Mis- Carlman, S.F. 2007. Cherished chestnuts. buyer was willing to pay. In our case, souri and the midwestern United No longer endangered. 7 Oct. 2008. < the price attribute was the only stat- States (Hunt et al., 2006). http://www.pantagraph.com/articles/ 2007/12/10/news/doc475c7f54c1 istically significant variable in the Chestnuts that originate in Mis- ca5818953424.txt>. model, suggesting that urban resi- souri and the United States would be dents are more price sensitive than preferred over imported ones and Carson, R.T., J. Louviere, D. Anderson, rural residents. there was a preference for organic P. Arabie, D. Bunch, D. Hensher, R. Building on a market-based compared with conventional produc- Johnson, W. Kuhfeld, D. Steinberg, J. approach, efforts to encourage the tion. Higher prices had a relatively Swait, H. Timmermans, and J. Wiley. 1994. Experimental analysis of choice. expansion of chestnuts production small negative influence on consumer Mktg. Lett. 5(4):351–367. in the United States should be preference for chestnuts. Looking at focused on the use of chestnut culti- all attributes, the highest importance Cernusca, M.M., M.A. Gold, and L.D. vars that produce a medium- to large- in the purchase decision is the label of Godsey. 2008. Influencing consumer size chestnut. Organic certification origin (produced in Missouri fol- awareness through the Missouri Chestnut would help to increase the likelihood lowed by produced in the United Roast. J. Ext. (Forthcoming: December of purchase. However, ‘‘locally grown’’ States). The next most important issue). is the most powerful attribute. Results attributes are chestnut size (medium Chestnuts Australia Inc. 2008. Nutritious, on price effects suggest that prices of $5 followed by large) and organic pro- delicious Australian chestnuts. 20 Feb. or $7 per pound are attainable in the duction. The least important attrib- 2008
222 • January–March 2009 19(1) JOBNAME: hortte 19#1 PAGE: 8 OUTPUT: November 20 01:32:31 2008 tsp/hortte/176989/01138
produced foods: A customer intercept Green, P.E. and V. Srinivasan. 1978. Con- Frontier in econometrics. Academic Press, study of direct market and grocery store joint analysis in consumer research: Issues New York. shoppers. 20 Feb. 2008
• January–March 2009 19(1) 223