An Empirical Analysis of Rice Demand in

THESIS

Presented in Partial Fulfillment of the Requirements for the Degree Master of Science in the Graduate School of The Ohio State University

By

Edith Ezra Lazaro, B.A.

Graduate Program in Agricultural, Environmental and Development Economics

The Ohio State University

2014

Master's Examination Committee: Abdoul Sam -Advisor

Stan Thompson- Member

Copyrighted by

Edith Lazaro

2014

Abstract

This study analyzes the consumer side of the Tanzania rice market. The study’s primary objective is to estimate own-price and cross-price elasticity of imported and domestically-produced rice.

Previous studies into the Tanzanian rice market have claimed that consumer preference protects the local rice market; this line of research implies that the small amount of imported rice allowed on the market cannot disrupt the local rice market because consumers prefer domestic varieties.

However, in practice, little has been observed in terms of the natural protection of local rice through consumer preference. In fact, rice traders have increasingly complained about the impact of imported rice, which has continuously dragged down the prices of domestic rice. To date, there are conflicting views regarding the extent of substitutability between imported and domestic rice when accounting for price and consumer preference. Therefore, this study pays particular attention to the question of substitutability for these two varieties. Using consumer data from a survey conducted by the researcher, the Linear Approximate Almost Ideal Demand System

(LA/AIDS) provides estimates of price and income elasticities for different rice varieties and maize. The results show that consumers have a high preference for domestic rice varieties with elasticity estimates indicate weak substitutability between domestic and imported rice varieties.

The results of the study provide little justification for the country’s rice import tariff. However, the study does not account for the possibility of rice mixing, whereby local varieties are mixed with imported rice. When prevailing market issues, such as rice mixing, are included in the analysis, the country’s import tariff may be justified. ii

Acknowledgments

Foremost, I would like to express my deep gratitude to almighty God who has been my rock and strength throughout the undertaking of this study. I would like to extend my sincere appreciation to Professor Abdoul Sam for his guidance and constructive criticisms, suggestions and comments, enabling the successful completion of this work. I am also grateful to Professor Stan Thompson for his invaluable advice and input from the first moment I arrived in the department. Dr. Fulgence Mishili, my local advisor, is to be thanked for his guidance and support in conceptualizing the study from its early stages to completion.

My sincere appreciation goes to the iAGRI Project led by the Office of International Programs at Ohio State University for granting me sponsorship and unconditional support throughout my master’s program. I thank Professor David Kraybill, Director of iAGRI; Professor Isaac Minde, Deputy Director of iAGRI; all iAGRI staff members; and the Office of International Programs at The Ohio State University for their inspiration and support.

Special thanks to Temeke, Ilala, and Kinondoni district authorities, the Municipal Council Authority, and the Authority for their cooperation during my field work, which enabled the smooth collection of my data. I remain indebted to my data collection managers (Dr. J. Jeckoniah and Ms. E. Mshote) and the whole data collection team for their hard work and resilience throughout the data collection exercise.

Last but not least, my heartfelt gratitude goes to my father, mother, and brothers for their unwavering support and encouragement during the completion of this work.

iii

Vita

February 2008 ……………………………Advanced Level Secondary Education, Mary Goreti 2011………………………………B.A. Rural Development, Sokoine University of Agriculture

Field of study

Major Field: Agricultural, Environmental and Development Economics

iv

TABLE OF CONTENTS

Abstract ...... ii

Acknowledgments...... iii

Vita ...... iv

List of Tables ...... vii

CHAPTER 1: INTRODUCTION ...... 1

1.1 Background information ...... 1

1.2 Problem Statement and Justification ...... 4

1.3 Objectives of the Study ...... 5

1.3.1 General Objective ...... 5

1.3.2 Specific Objectives ...... 5

1.4 Hypotheses ...... 5

CHAPTER 2: LITERATURE REVIEW ...... 7

2.1 Estimating Price Elasticities from Cross-Sectional Household Data ...... 7

2.2 Tanzania and International Trade in Rice ...... 7

2.3 Nature of the Tanzanian Rice Market ...... 8

2.4 Rice and Agricultural Policies in Tanzania ...... 9 v

2.4.1 The Import Tariff ...... 9

2.4.2 Export Bans ...... 10

CHAPTER 3: METHODOLOGY AND DATA ...... 12

3.1 Methodology ...... 12

3.2 Data ...... 13

3.3 Estimation...... 15

3.3.1 Zero Consumption Households...... 18

3.3.2 The Missing Prices Problem ...... 20

3.3.3 Elasticities ...... 20

CHAPTER 4: RESULTS AND DISCUSSION ...... 22

4.1 First Stage Estimation Results ...... 22

4.2 Second Stage Results ...... 26

4.3 Elasticity Results ...... 28

CHAPTER 5: CONCLUSION AND POLICY IMPLICATIONS ...... 34

BIBLIOGRAPHY ...... 36

vi

List of Tables

Table 1. Variable Definition and Sample Statistics (Sample Size: 1064) ...... 17

Table 2. Percentage of Zero Consumption for each Cereal Category ...... 18

Table 3. Parameter Estimates for the First Stage Estimation...... 25

Table 4. Linearized Almost Ideal Demand System (LAIDS) Parameter Estimates ...... 27

Table 5. Expenditure Elasticity and Budget Shares ...... 29

Table 6. Compensated and Uncompensated Elasticity Estimates ...... 33

vii

CHAPTER 1: INTRODUCTION

1.1 Background information

Rice is one of the most widely-grown crops in Tanzania, after maize. It is the second most important food crop in terms of household consumption, area planted, and production volume

(European Cooperative for Rural Development EUCORD 2012). The average consumption of rice in Tanzania is around 1,024,000 metric tons (MT) annually (2007 – 2013) (United States

Department of Agriculture USDA 2013) with a consumption per capita of around 25kg in 2007.

Rice production in the country covers approximately 681,000 ha, representing 18% of the cultivated land (Ministry of Agriculture Food Security and Cooperatives MAFC 2009).

According to the Agricultural Census of 2004, 17% of all agricultural households grow rice. Almost 90% of these farming households are smallholder farmers using traditional technology in their cultivation. In 2007, total production was 818,000 tons, with a productivity of between 1.0 to 1.2 tons of milled rice per ha (MAFC 2009). In comparison to other countries in the East Africa region, Tanzania has the lowest yield in rice production (Barreiro-Hurle 2012), in part because of the country’s primary reliance on traditional technology in rice cultivation

(Agricultural Council of Tanzania ACT 2010; MAFC 2009).

According to the National Agricultural Census of 2002-03, 42% of rice production is marketed, compared to 28% of maize and just 18% of sorghum. Small-scale farmers sell about

13% of the marketed rice with the rest sold by large-scale farmers (Minot 2010). The Tanzanian rice market is partially open to imports, allowing consumers the option to purchase rice imported 1 from different countries (Barreiro-Hurle 2012). Different types of rice compete in the rice market: imported rice (mostly from Pakistan, India, Thailand and China) and domestic rice varieties (the high quality aromatic rice comes from the Mbeya region (Kyela district), while other semi- aromatic and non-aromatic rice from come from Mbeya (Mbarali district), Morogoro, and

Shinyanga regions). Dar es Salaam is Tanzania’s largest rice market, representing up to 60% of the national rice consumed (Bill and Melinda Gates Foundation BMGF 2012). With 30% of the total national urban population and a GDP per capita of Tsh.1.7 million in 2010, it is the region with the highest population and GDP per capita (BMGF 2012). Other regions with high consumption include Mwanza, Arusha, Mbeya, Morogoro, Kilimanjaro, and Tanga.

Rice is a crucial staple in the country, commanding special attention from national policies geared to increase production and attain self-sufficiency in food production (MAFC

2009). The government has put in place initiatives to support the rice sub sector, such as the

National Rice Development Strategy,1 which intends to transform the existing subsistence- dominated rice subsector into a commercially viable production system. The sector also enjoys a favorable policy environment with an import duty of 75%2 (East African Community EAC 2012), providing a wide price margin in favor of domestic rice over imported rice (ACT 2007). The policy environment also provides tax exemptions for agricultural inputs, such as machinery and fertilizers. Rice and maize farmers have further enjoyed the support of a government subsidy program (the National Agricultural Input Voucher Scheme), enforced from 2009-2012. The program aimed at improving farmers’ access to critical agricultural inputs (fertilizer and improved seeds).

1 The Tanzanian government published the National Rice Development Strategy in 2009, proposing to enhance the competitiveness of the Tanzanian rice subsector and double rice production by 2018.

2 This is a Common External Tariff of the East African Community. It is a compound tariff of 75% ad valorem rate or US$200/MT, whichever is higher. 2

Despite these benefits, the rice subsector also faces restrictive policies that have a negative impact, including export bans that restrict the export of all grains from the country.

Since 2005, export bans have been a common phenomenon for the Tanzanian grain sub-sector, with a total of four export bans: from March 2006 to January 2007, from January 2008 to May

2008, from January 2009 to October 2010, and from July 2011 to January 2012 (Stryker & Amin

2012). The bans are usually imposed to safeguard food security in the country, particularly in cases of perceived poor harvests or when consumer prices are unusually high (Diao, Mabiso, &

Kennedy 2012) Since 2012, there have not been any cases of government cereal export bans.

Though Tanzania is regarded as the second largest producer of rice in Eastern Africa (ACT

2010), it does not meet its own rice demand; 8% of its total rice consumption is still imported

(Minot 2010), mainly from Vietnam, Thailand, China, Pakistan, and India. In recent years, regional exports have represented about 5% of rice production, principally to the neighboring countries of Malawi, Zambia, Uganda, Rwanda, Kenya, and Burundi (Lewis 2012). A large percentage (81%) of the rice produced in the country is consumed locally (USAID 2010).

Researchers have attributed the high consumption of domestic rice to consumer preference for certain varieties of domestic rice (ACT 2010; Barreiro-Hurle 2012). Consumers prefer local rice mainly due to its aromatic qualities. A study funded by the Bill and Melinda Gates Foundation in

2012 reported that consumers were willing to pay up to a 21% premium for the local aromatic rice due to its freshness, which affects texture and taste.

Consumers’ lack of preference for rice without aromatic qualities suggests that the market for domestic rice should not be significantly affected by imported varieties, which are usually of low quality and lacking in aromatic qualities. If, as suggested (ACT 2010), consumers prefer domestic aromatic rice, the price of Tanzanian rice would not fall if the import tariff on rice was to be removed, since the domestic rice market is naturally protected by consumer

3 preference (BMGF 2012). In reality, however, little has been observed in terms of consumer protection of local aromatic rice on the markets. Markets for local rice suffered tremendously during years in which the government lifted the import tariff and allowed free flow of imported rice (e.g., 2011/2012 financial year). Rice traders reported a 70% drop in domestic rice prices during the tax relief period of 2011/2012 (The Guardian 2013). 3 It is apparent that there remain conflicting views on the extent of substitutability between imported rice and domestic rice when accounting for price and consumer preference. Therefore, the main objective of this study is to generate evidence that will portray the true picture of rice demand in Tanzania. The study aims at providing an analytical base to guide policy formulation. More specifically, it seeks to contribute to food security and agricultural trade policy dialogues. The underlying idea is to analyze own- and cross-price elasticity between imported and domestic rice, and also between low-quality domestic rice and imported rice and maize. The extent of substitutability between these different types of rice and maize will offer evidence to guide agricultural pricing policies in the country, as well as trade and food security policies.

1.2 Problem Statement and Justification

Previous research that has estimated rice price and income elasticities in Tanzania for groups of staple food products, such as cereals, pulses, meat, dairy, edible oils, vegetables and fruits

(Weliwita et al. 2003). Data for this (Weliwita et al 2003) analysis were taken from the household budget survey conducted between December 1991 and 1992 by the Bureau of Statistics in the

United Republic of Tanzania. As is well known, surveys do not usually collect actual market prices; rather, demand analysis from surveys usually uses unit prices as proxies for market price.

3 Prices for domestic rice usually drop following the government’s notice of intention to import, suggesting that the drop in domestic rice prices is less reflective of an impact of imported rice in the market. Rather, it may result from local wholesalers dumping their rice reserves (that were waiting for peak prices) in the market once the government issues a notice to import.

4

Studies have shown that bias may result from using unit values due to quality effects and measurement errors (Deaton 1988). This study differs from earlier studies of Tanzanian food demand in a few key ways. First, it only considers cereal food products, with a special focus on rice at a disaggregate level, the first study of its kind for Tanzania. The disaggregated elasticities will provide policymakers with information to guide trade and food security policies in the country. The second distinctive feature of this study is the use of actual market prices in the analysis; the study uses survey data reflecting the actual market prices observed by consumers.

As such, the results of this study will be free of the biases that may result from using unit prices as proxies.

1.3 Objectives of the Study

1.3.1 General Objective

The main objective of this study is to provide estimates of own-price and cross-price elasticity for imported rice and domestically-produced rice using disaggregated household survey data and a theoretically consistent micro-econometric demand model.

1.3.2 Specific Objectives

Specifically, the study will seek to do the following:

1) Determine the socioeconomic characteristics of rice-consuming households.

2) Compare demand among different categories of domestically-produced and imported rice.

3) Estimate the cross-price and expenditure elasticities of the different rice varieties and maize.

1.4 Hypotheses

1) The price of rice is a key determinant of the kind of rice that consumers demand on the market.

2) Consumer preference protects aromatic domestic rice in the local market; hence, there is little substitution between aromatic domestic rice and imported rice. It has been established that

Tanzanian consumers have a broad taste preference for domestic rice due to its aromatic qualities

5

(Lewis 2012; MAFC 2009; USAID 2010). If true, removal of the import tariff should not affect the price of domestic rice (BMGF 2012).

6

CHAPTER 2: LITERATURE REVIEW

2.1 Estimating Price Elasticities from Cross-Sectional Household Data

In the field of applied microeconomics, estimating elasticity of consumer demand has been an important issue, especially in agricultural economics. Elasticity estimates are essential in providing information for agricultural policy deliberations, such as prediction of future demand and prescription of tax and subsidy amounts. In turn, all of this information feeds into sectorial policies, such as food security and pricing policies. To guide the public policy of developing countries, it is essential to understand how households adjust consumption in response to changes in price and income. To date, several studies have used elasticity estimates to guide policy

(Andreyeva et al. 2010; Hoang 2009; Nzuma & Sarker 2010; Weliwita et al. 2003).

Previous studies on consumer demand were largely based on aggregated inter-temporal data (see Pollak & Wales 1978; Ray 1982). However, more recent work has relied on cross- sectional household data obtained through consumer surveys or scanner data (see, Capps 1989;

Deaton 1990; Heien & Wesseils 1990). This evolution in the empirical literature on consumer demand has been particularly beneficial to developing countries, where time series data are less common and little is known about how consumers respond to changes in price (Deaton 1987).

2.2 Tanzania and International Trade in Rice

Tanzania is a regional player in the rice market; it is a leading exporter of rice in the region (EUCORD 2012) and is the second largest producer of rice in Eastern Africa after

Madagascar. It exports an average of 51,200MT annually to its East African neighbors and the 7 rest to other neighboring African countries (Malawi, DRC, and Zambia). In the case of imports, the country trades beyond the African region, as most of its rice imports come from East Asia, particularly Pakistan, India, and Vietnam (Lewis 2012). Imports from developed countries play an important role in some years (mainly during the period leading up to the food price crisis), with imports coming mainly from Japan and the USA in the form of food aid (Barreiro-Hurle 2012).

2.3 Nature of the Tanzanian Rice Market

Rice is increasingly becoming one of Tanzania’s important commercial food crops.

According to the National Agricultural Census of 2002-03, 42% of rice production is marketed, compared to 28% of maize and just 18% of sorghum. Small-scale farmers sell about 13% of marketed rice, and the rest is sold by large-scale farmers (Minot 2010). The Tanzanian rice market is characterized by competing varieties that can be placed into two major categories: imported and domestic rice. Domestic rice can be further categorized into different classes based on its quality; this distinction is usually based on the place of origin, the region of cultivation.

Rice is often labeled as being from particular regions that consumers perceive as offering particular qualities (BMGF 2012). Common rice varieties in the country include Kyela rice, which is viewed as the best quality aromatic rice, followed by Mbeya rice (from Mbarali) and

Morogoro rice which are viewed as high-quality semi-aromatic rice. Meanwhile, Shinyanga rice is viewed as low quality because it is not aromatic, and historically, it contained a significant amount of foreign matter (Lewis 2012).

A large portion of imports in the country’s rice market are low-quality non-aromatic rice varieties. High-quality imported rice varieties, such as Basmati and Jasmine, are also available in high-end market chains. However, their consumption and availability are negligible when analyzing the Tanzanian rice market. Currently, no formal statistics are available for consumption of this high-end imported rice in the country; however, in a wider context, Africa

8 consumed only about 0.5% of total Basmati rice exported from India in 2000/01 (Directorate

General of Commercial Intelligence and Statistics India DGCIS 2002). This statistic confirms that a large portion of imported rice in the country is not the high quality imported rice but rather the low quality imported rice. This is also reflected in the price of imported rice, which is lower than domestic rice prices (Minot 2010). For example, price data for 2010 show that high-quality domestic rice is sold at a premium of about 28% over imported rice (USAID 2010).

2.4 Rice and Agricultural Policies in Tanzania

2.4.1 The Import Tariff

In 2005, East African Community (EAC) member states opted for a 75% ad-valorem common external tariff on rice to protect local rice farmers from lower-cost producers in countries outside the EAC. For intra-EAC trade, the protocol splits traded products into category A and B goods.

Category B goods (which includes agricultural products) face a 10% import tariff among members. However, this percentage was subject to a transition period of five years following

2005, and the agreement allowed for an annual reduction of 2% so that the 10% tariff was to be phased out by 2010 (Khorana et al. 2007). In theory, the 75% ad-valorem could keep rice prices high as the tariff would be expected to reduce the availability of rice on the market. Conversely, a tariff on imported rice could trigger increased production as a response to increased market prices.

However, studies show varying findings regarding the tariff’s impact on the rice subsector. Theoretically-grounded studies have found that the tariff in place protects the domestic rice subsector(USAID 2010). Such research primarily argues that the rice subsector is a price- sensitive market, suggesting that the domestic rice would lose out with price when compared to low-quality imported rice. The study provided 2010 price data for import rice (Thai al Super rice)

9 bearing a CIF4 price of $445/mt, compared to high-quality domestic rice selling at $970/mt and low-quality domestic rice selling at $750/mt. In this case, the 75% import duty raises the Thai al

Super to $756/mt, thus enabling the domestic market to compete (USAID 2010).

Other studies have reported no effects on price changes as a result of the import tariff. As quoted from a report by the Ministry of Finance, “...the closing down of the imports of pre-2005 magnitudes following the introduction of CET should have resulted in major rice shortages or rapid price increases. There is no indication of the former during the 2001 to 2007 period, and rice prices did not increase rapidly” (United Republic of Tanzania URT 2010). This finding could indicate that imports have little impact on the local rice market, perhaps because of consumer preference for local varieties.

2.4.2 Export Bans

Grain export bans are among the most frequent policy interventions for food security in Tanzania

(Morrissey & Leyaro 2009). While the export of almost all agricultural products is liberalized, maize and rice face occasional export bans. These periodic bans pose disincentives to local producers and deny them of lucrative market opportunities that may be available in neighboring countries. Bans also negatively affect consumers because producers respond to export bans by reducing production (Brigham 2011).

Many stakeholders in the rice subsector have questioned the information used by the government to initiate these bans. In most cases, the government relies on the food security information provided by regional offices to make these decisions. Food insecurity in a few regions can lead to nationwide export ban regardless of the availability of excess supplies in high- producing regions. Between 2005 and 2012, there was a total of four cereal export bans in the

4 Cost Insurance and Freight: This is the price of a good delivered at the border of the importing country, including any insurance and freight charges incurred to that point but prior to payment of any import duties, taxes on imports or trade, or transport margins within the country. 10 country (Stryker & Amin 2012). Poor transportation infrastructure in production areas has been the main reason for food insecurity alarms in the country. The inability to transport food from surplus regions to low-production regions has resulted in false alarms of food insecurity, leading the government to enforce export bans (The East African 2013). Other government policies assist the rice subsector, including an agriculture subsidy program that ended in 2012. This program subsidized inputs for maize and rice farmers. As another incentive, the government provides tax exemptions on agriculture inputs, such as machinery and equipment.

Tanzania’s protection policies in the agricultural sector are put in place in the name of i) ensuring food security through self-sufficiency (EUCORD 2012) and ii) protecting the domestic rice sector from cheaply imported rice. These protectionist policies come at a price: they cause distortions in prices, resource allocation and trade patterns, thus reducing the welfare of individual consumers and producers as well as the country as a whole.

11

CHAPTER 3: METHODOLOGY AND DATA

3.1 Methodology

This study adopts the Linear Approximate Almost Ideal Demand System (LA/AIDS) (Deaton &

Muellbauer 1980) in estimating consumer demand in the Tanzanian rice subsector.

 x  n wi  i  i log *    ij log p j  ui1  p  j 1 i 1,2...n,where * log p  wj logp j  ...(2) j

Equation (1) is a linear approximation of the AIDS where wi = the budget share of good i. x = the household’s per capita food consumption expenditure.

pj = price of commodity j.

i, i, and ij = parameters to be estimated. ui = random disturbance.

Equation (2) is the Laspeyres price index, where 푤̅ i is the mean observed budget share of good i for all households. This weighted index is an improved substitute to the commonly-used Stone price index, which has proven vulnerable to units of measurement error (Moschini 1995). The

Laspeyres price index is invariant to the units of measurement; thus, it is proposed as a better alternative to the Stone index. Theoretical restrictions imposed on the model include adding up

(3), homogeneity (4), and symmetry (5):

12 n n n i 1,  i  0,  i j  0 ... (3) i1 i1 i1 n

 i j  0 ...(4) j1    ... (5) i j ji

3.2 Data

This study uses cross-sectional data collected from a household consumer survey conducted in

Tanzania. The consumer survey focused primarily on cereal consumption at the household level with an intentional focus on rice consumption. The survey covered two of the major rice regions in the country, Dar es Salaam and Morogoro. The Dar es Salaam region merits selection because it is where most of the rice trading and consumption takes place. The was key to this survey because it is one of the leading regions in rice production; it also had the merit of providing clear disaggregation between urban and rural communities, offering a solid platform for comparing rice consumption trends in different settings.

A multi-stage population-proportionate random sampling procedure was used to select the 1064 participating households. The first study area, Dar es Salaam, was divided into three clusters based on administrative setting: Temeke, Kinondoni, and Ilala. The sampling frame for these areas was obtained from the 2012 census data. From each administrative area, three wards were randomly selected based on household economic category (high-income households, average households, and low-income households). Economists and statisticians from the district offices assisted with this disaggregation based on attributes they normally use in classifying the households. At the ward level, representative households were systematically chosen to be interviewed from a sampling frame provided by ward administrative officers.

A similar procedure was carried out in the second study area, the Morogoro region, where two administrative areas the and the Morogoro rural district were chosen 13 to carry out the survey. Each selected area was to represent a different setting for the survey. In the case of the urban setting, we randomly selected three wards based on household economic category, per the district office’s disaggregation. At the ward level, interviewed households were systematically selected from a sampling frame available in the ward administrative offices. In the case of the rural setting, two districts were chosen to represent rural areas. We used the 2014 census data as a point of reference for all study areas in order to compute the number of households that were to participate in the survey.

The main goal of the survey was to collect data on consumption and expenditure patterns for different cereals with a focus on disaggregating different varieties of rice available in the

Tanzanian rice market. The survey collected consumption data on four common domestic varieties of rice, most which are identified by the name of the region/location of production. The survey included rice from Kyela, commonly known as Kyela rice. It is considered grade-one rice on the market as it is 100% aromatic; it is also very clean and less than 20% broken. Another variety included in the survey was Mbeya rice from Mbarali, an average quality variety characterized by semi-aromatic qualities. It is usually comparatively clean and 20% broken. A similar quality variety captured in the survey was Morogoro rice from Ifakara, which is also characterized by a semi-aromatic quality; it is usually clean and available with grains that are only 20% broken. The consumption of other domestic rice varieties was also captured in the survey, including commonly-consumed low-quality domestic varieties, such as Shinyanga rice, which is usually regarded as the lowest quality rice on the market. These varieties are non- aromatic rice varieties that have a great deal of foreign matter in them, and their grain quality is poor usually more than 20% broken.

The survey also collected data on imported rice, more specifically low quality imported rice that has a considerable amount of consumption on the market. The imported rice varieties

14 collected by the survey are composed of all long grain and short grain low-quality imported rice varieties available on the Tanzanian rice market. These varieties tend to fetch the lowest prices on the market compared to domestic varieties, and in most cases, consumers can’t differentiate the varieties. It is important to note that the 75% rice import tariff was in effect during the survey.

Therefore, the price of imported rice reflects the current tariff in place.5 Consumption of other cereals collected by the survey includes maize cereal and other common cereals, such as millet, sorghum, and wheat. The survey also collected data on the demographic characteristics of rice- consuming households.

3.3 Estimation

In this study, estimations are based on 1064 observations collected in the survey; however, some rice varieties recorded very low consumption rates, which necessitated the aggregation of these varieties into several categories to enable estimation of the model. Domestic rice varieties were categorized into three categories based on the price and quality acquired by these varieties at market: a high-quality domestic category comprised of high-quality Kyela rice, an average- quality domestic category including Mbarali Mbeya rice and Morogoro rice, and an “other” domestic rice category composed of Shinyanga rice and other domestic rice varieties. Table 1 provides other variables included in the estimations, including demographic variables.

Correct specification of these demographic variables is crucial in order to estimate income and expenditure elasticities correctly (Blow 2003). Household income is an important factor in determining the consumption and expenditure patterns of any household. It provides a sound reflection of a household’s purchasing power, and thus, the quality and quantity of food

5 It is not possible to say with certainty that the price of imported rice collected reflects the 75% import tariff, this due to the presence significant smuggling of imported rice from Zanzibar to the mainland because of a large tariff differential between Tanzania mainland (import tariff of 75%) and Zanzibar (import tariff of 12.5%) (Therkildsen, 2011).

15 consumption (Roos et al. 1998; Sekhampu 2012). The head of the household’s education level was included in the model estimation as a control variable for income; this variable emerged because most respondents were reluctant to provide actual information on their income during the survey. The sex of the head of the household is an important determinant of expenditures and of the nature of food consumption. Female-headed households tend to have lower purchasing power and more likely to purchase more affordable food commodities than male-headed households

(Ogundele 2014). Household size is meant to capture the nature of household composition in relation to food expenditures on rice (Babalola & Isitor 2014; Chern 2003). Marital status has also been observed to influence the nature of food consumption in a household (Roos et al. 1998).

Female partners are more likely to be conscious of the quality of the food prepared in their home, and thus, more likely to influence the purchase of quality food commodities. The age of the head of household also plays a crucial role in influencing household food consumption. Studies have shown that older people are more likely to pay attention to food quality; therefore, the age of the head of household is likely to influence the quality of food consumed (Lallukka et al. 2006).

Older people also tend to be more conservative about the kind of food they consume, indicating that households with older heads are more likely to prefer more traditional food varieties over less traditional foods. Location is a dummy variable that captures the setting of the household (1 for households located in urban areas).

16

Table 1. Variable Definition and Sample Statistics (Sample Size: 1064)

Variable Mean Std Dev

Quantities in Kgs (Per Household) High quality domestic rice 5.647957 3.89844 Average quality domestic rice 5.627696 3.762522 Other domestic varieties 5.84879 4.493086 Maize 5.204386 4.041546 Imported rice 6.134513 3.949989 Other cereals 8.642994 107.7293 Price per Kg (Tanzanian Shillings) High quality domestic rice 1667.98 328.1884 Average quality domestic rice 1505.94 272.9665 Other domestic varieties 1482.25 311.2121 Maize 1010.83 277.4889 Imported rice 1031.59 91.34343 Other cereals 1434.49 541.6731 Expenditure (Tanzanian Shillings) High quality domestic rice 9162.98 6161.04 Average quality domestic rice 8391.94 5652.48 Other domestic varieties 8489.92 6394.4 Maize 5322.14 4091.35 Imported rice 6331.19 4066.5 Other cereals 3950.65 4728 Income (log of household income) 705057.4 1541994 Location (dummy=1 if location is urban) 0.81109 0.391621 Noeduc (dummy=1 if head of household has no formal education) 0.069549 0.254505 Prieduc (dummy=1 if head of household has primary school education) 0.066729 0.24967 Seceduc (dummy=1 if head of household has secondary education) 0.485902 0.500036 Otheduc (dummy=1 if head of household has other form of education) 0.115602 0.319897 Agehhl (dummy=1 if age of household head is less than 31 years) 0.18985 0.392367 Agehha (dummy=1 if age of household head is 31- 50 years) 0.50282 0.500227 Maritmar (dummy=1 if head of household is married) 0.744361 0.436425 Sizesml (dummy =1 if household size less than 4 persons) 0.335526 0.472396 Sizeav(dummy=1 if household size 4-6 persons) 0.495301 0.500213 SexHH(dummy=1 if household head is male) 0.75846 0.428219

Exchange rate during survey 1USD = 1658TZS (Source: Bank of Tanzania)

17

Other key estimation strategies used in this study are explained below.

3.3.1 Zero Consumption Households

The presence of zero expenditures is a common phenomenon in survey data; it often appears that some households do not consume some of the commodities in question. This lack of consumption usually results from imperfect recall or infrequency of purchase resulting from a relatively short period of data collection. This study is no exception; small percentages in the number of household consumption are observed in some of the cereal categories (see Table 2).

Table 2. Percentage of Zero Consumption for Each Cereal Category

Cereal categories % of zero consumption

Imported rice6 90% High quality domestic 78% Average quality domestic rice 42% Other domestic rice 88% Other cereals 68% Maize 2%

To offset the sample selection problem that may arise with the presence of zero consumption households, this study adopts the popular two-step approach (Shonkwiler & Yen 1999) for estimating elasticities in different cereal categories. Following Shonkwiler and Yen, the zero expenditure can be stated as follows: Let the demand equation (1) be denoted by f(xij,βi) with random disturbance ϵi.

6 Low preference for imported rice in the country could be regarded as the main reason for the high percentage of zero consumption in this category. However, it should be noted that other factors may also have a role in this phenomenon, including the timing of the survey. The survey was administered during the harvest season (June-July) when prices for domestic rice are very low; thus, consumers are more likely to consume domestic rice varieties during this period. Another factor is rice mixing. There is a tendency for rice traders to mix domestic rice varieties with imported rice and to sell the product as domestic rice. This practice could affect the number of households reporting the quantity of imported rice consumed. 18

The following illustrates the sample selection system:

 f xi j , i  i j if di j  1, i 1,2, ...n wi    ... (6) 0 if d  0, j  1,2, ....n  i j 

di j  Izi ji  vi j  0 ....(7)

2 2 Where  v  N 0,0,  2 , v 2 i j i j  i i 

f = functional form of LA-AIDS. dij = dichotomous variable indicating consumption.

I(.) = binary indicator function.

푧́ = vector of demographic variables.

αi= parameter vector.

The two-step estimation procedure proposed by Shonkwiler and Yen (1999) is based on the assumption that the joint distribution of the error terms (ϵi,vi) is bivariate normal with a covariance parameter θi , leading to the following censored demand system:

 z    E(w )   z   f x   i j i   , ...(8) i  i j i   i j i  i i j  zi i   j  i 1,2,...n j 1,2... j

In the equation above, ϕ(.) and Φ(.) are standard normal cumulative distribution and probability density functions, and E(wi) are the unconditional means of the expenditure shares. The procedure involves two steps. First, we estimate αi by the probit model to obtain 훼̂푖 i= 1…n. In this step, a probit regression model is estimated for all cereal categories. The parameter estimates of the probit regressions are then used to compute the inverse Mills ratios for each of the censored cereal categories. In the second step, we estimate the equation system (8) and evaluate at 훼̂푖 as a system of seemingly unrelated regressions (SUR). In the second step, each censored demand

19 equation is augmented with an inverse-Mills ratio constructed from the probit estimates and then weighted by the normal cumulative distribution function to correct for zeros in the dependent variable. Following Sam and Zheng (2010), we used total income as the exclusion restriction in our estimation.

3.3.2 The Missing Prices Problem

Generally speaking, the problem of zero expenditures usually goes hand-in-hand with the issue of missing prices: households with zero consumption provide no information on quantities consumed or commodity prices. As a result, the data bears zeros not only in the quantity variable but in the price variable as well. This omission may, in turn, affect model estimation, rendering it difficult to attain convergence, especially when there is a large number of zero expenditures in a data set. To manage this problem, we obtained prices for observations with zero expenditure under the assumption that the household face average prices prevailing in their location. Mean prices at ward level were used in place of missing prices.

3.3.3 Elasticities

As for the adding-up restrictions for the LA/AIDS, equation (3) cannot be guaranteed in the presence a censored demand system (Yen and Lin 2006). To accommodate adding up, Engel

푛 푛 푛 (∑푖=1 푤iei=1), Cournot (∑푖=1 푤ieij+wj=0), and Euler (∑푗=1 푒ij+ ei=0) aggregation can be imposed after estimation to ensure that the elasticity estimates are consistent with theory (Sam and Zheng

2010). Estimates are made for the first (n-1) goods, treating the nth good as a residual good. The uncompensated price and expenditure elasticities for the nth good are recovered using the theoretical aggregation above, and the compensated elasticities are obtained using the Slusky equation (Yen and Lin 2006). In the case of (n-1) non-residual goods, the uncompensated price and expenditure elasticities are computed using elasticity formulas (9) and (10):

20

 Ew   1  e   i     ...(9) i j    i j   log p j  E(wi ) 

 Ew  1  i   ei     1 ...(10)   log x  Ewi 

21

CHAPTER 4: RESULTS AND DISCUSSION

4.1 First Stage Estimation Results

The first stage estimations involved estimating a probit regression for each commodity (Table 3), the dependent variable presents a dichotomous choice problem equal to one if the respondent consumes a commodity and zero otherwise. The parameters of the probit model are then used to compute the inverse Mills ratio entered in the second stage estimation as an instrument correcting zeros in the dependent variable.

The results in Table 3 show that only 23 out of 72 coefficients estimated are significant.

This finding is not surprising due to the cross-sectional nature of the data. Most of the signs of the coefficient tally with expectations; the income coefficient shows that an increase in household income causes a decrease in the household’s probability of consuming imported rice and other domestic rice (Shinyanga and other domestic rice varieties). These varieties of rice are usually the most affordable on the market, and in most cases, considered inferior by consumers. Rice categories that showed a positive relationship with income were the high-quality domestic category (Kyela rice) and the average-quality domestic category (combination of Mbarali Mbeya rice and Ifakara Morogoro rice). These varieties fetch high prices on the market; thus high and average-income households can afford them. The income coefficient for other cereals is positive, while maize bears a negative coefficient, suggesting that an increase in income would reduce the probability of a household consuming maize. Other studies have also found that consumers are more likely to shift from maize to rice consumption with increased income (BMGF 2012).

22

Results regarding the influence of gender in the consumption of different rice categories were significant for two rice categories: high-quality domestic rice and average-quality domestic rice. The direction of the coefficients showed a higher probability for male-headed households to consume rice from the average-quality domestic rice category, while female-headed households were more likely to consume the high-quality domestic category. In some ways, these results are unexpected. The expectation was for male-headed households to be more likely to consume rice from the high-quality domestic category, which is usually the most expensive on the market, since male-headed households generally have higher purchasing power than female-headed households. Nevertheless, these results make more sense when the issue of food quality is considered; female-headed households are more concerned with food quality compared to male- headed households, making them more likely to consume rice from high-quality domestic varieties. The gender coefficient was significant for the other cereals categories with a negative sign, indicating that it was more likely to be consumed by female-headed households. Again, food quality concerns could explain these results because female-headed households are more likely to offer quality food than male-headed households.

The location coefficient is significant in two out of four rice categories. As expected, the sign of the coefficients indicates that households located in urban areas have a higher probability of consuming the high-quality domestic category, while households located in rural areas are more likely to consume the other domestic rice category. This category (other domestic rice category) is comprised of less expensive rice varieties that are primarily consumed in their area of production. The coefficient is significant in the other cereal category with a positive sign indicating that urban households have a higher probability of consuming other cereals. This finding is realistic because products made of cereals, such as bread, cakes, and cornflakes are usually expensive, and their availability is generally limited to urban areas.

23

The head of household’s education greatly impacts the quality and type of food that a household consumes. Though only four out of the 16 coefficients for education were significant, the influence of education on food consumption was evident: a highly-educated head of household has a higher probability of consuming rice from the high and average quality categories compared to heads of household with little education. Meanwhile, heads of household with a low education level were more likely to consume from the low-quality rice, imported rice, and other domestic rice categories. This finding was expected because education level was regarded as a proxy for income in the study, as an indicator of purchasing power, the higher the education level, the higher the chances of one earning a good income (other factors constant). The influence of education on rice consumption could result from the social context as well; educated heads of household may feel prestige by consuming high-quality rice varieties.

The age coefficient was significant for imported rice; older heads of household were more likely to purchase rice from the imported category when compared to heads of household that are young or middle aged. This finding was not expected, as older heads of household are expected to be more conservative, thus preferring to consume traditional varieties of rice. We thus expected that they would be more likely to consume rice from domestic rice categories than from the imported rice category. An alternative explanation to these results could be that older heads of household tend to have lower purchasing power because most of them are retirees. This socioeconomic factor may force them to opt for cheaper varieties of rice.

Married heads of households are more likely to consume rice from the high-quality domestic category when compared to single heads of household. This finding was expected because married individuals are more likely to care about food quality than single heads of household.

24

Looking at variables related to the size of the household, large-sized households have a greater probability of consuming rice from the high-quality domestic category compared to the small-sized and average-sized categories. This result was unexpected because large households were expected to opt for cheaper varieties due to their size. Nevertheless, these results could be explained in the context of economies of scale. Large households would be able to afford the expensive high-quality variety as they obtain the relatively affordable wholesale price due to buying the rice in large quantities. Large-sized households also had a higher probability of consuming maize compared to small-sized households, an expected finding because maize is the main staple and one of the most affordable cereals on the market. It would be economical for a large-sized household to consume more of it.

Table 3. Parameter Estimates for the First Stage Estimation

Average Other Variables High quality quality Other Imported rice domestic Maize domestic rice domestic cereals rice rice

Intercept 1.6246 -1.6571 -0.8379 -0.9718 -4.3893 6.3110 (0.0353) (0.0091) (0.1372) (0.1893) (<.0001) (<.0001) ltotal_ic -0.2231* 0.0635 0.0665*** -0.0084 0.2935* -0.2961* (0.0002) (0.1777) (0.1137) (0.8797) (<.0001) (0.0045) SexHH 0.1102 -0.2987*** 0.2793*** -0.1270 -0.3606** 0.1889 (0.6416) (0.1073) (0.0915) (0.5368) (0.0371) (0.6446) location -0.1174 0.3796* 0.1479 -0.4334* 0.4305* -0.2884 (0.3871) (0.0047) (0.1600) (0.0006) (0.0005) (0.4668) noeduc -0.0055 -0.1403 0.0274 0.2646 -0.1120 0.1964 (0.9822) (0.5004) (0.8784) (0.2345) (0.5702) (0.6953) prieduc 0.3106 -0.2688 0.0412 -0.0014 -0.2604 3.2989 (0.1673) (0.1949) (0.8173) (0.9954) (0.1897) (0.9597) seceduc 0.2379*** -0.3128* 0.0595 0.2005*** -0.1158 0.2374 (0.0981) (0.0043) (0.5484) (0.1402) (0.2568) (0.3193) otheduc -0.1565 0.3016 -0.4073* 0.1784 -0.0351 0.2450 (0.5363) (0.0445) (0.0050) (0.3712) (0.8130) (0.3944)

25 Continue

Continue Average Other Variables High quality quality Other Imported rice domestic Maize domestic rice domestic cereals rice rice agehhl -0.5388* 0.0214 0.1124 0.2269 0.1365 0.5385 (0.0027) (0.8811) (0.3676) (0.1509) (0.3121) (0.1184) agehha -0.3739* -0.0356 0.2182** 0.0158 0.1731*** 0.1486 (0.0029) (0.7381) (0.0215) (0.9015) (0.0880) (0.5057) marit 0.1063 0.4358** -0.3754** -0.0465 0.0504 -0.3000

(0.6495) (0.0189) (0.0221) (0.8206) (0.7686) (0.4642) sizesml -0.1020 -0.2484*** 0.0412 0.1013 -0.0321 -0.6559** (0.5479) (0.0726) (0.7413) (0.5549) (0.8106) (0.0490) sizeav 0.0158 -0.2553** 0.0047 0.2206 0.0604 -0.1970 (0.9146) (0.0357) (0.9666) (0.1547) (0.6116) (0.5399)

(...) are P-values. Coefficients in bold are significant at 1% (*), 5% (**), 10% (***)

4.2 Second Stage Results

In the second stage, we estimate the Linearized Almost Ideal Demand System (LAIDS) equation (1) with the inverse Mills ratio from the first stage results (see Table 4). The coefficient estimates for the inverse Mills ratio show significant t-values, suggesting that there is evidence of sample selection problem in estimating cereal demand. This issue justifies the two-step sample selection correction procedure adopted by the study. Almost half of the price coefficient estimates are significant at 5%; most cross-sectional data sets rarely produce a large number of significant results in estimations. Nonetheless, the large number of significant price coefficients in Table 4 indicates some degree of sensitivity of the cereal budget share to prices. Among the four rice categories, only two categories (the high quality domestic category and the average quality domestic category) have significant own-price effects, indicating that these two rice categories are more sensitive to own-price changes than other categories of rice. This sensitivity implies that price changes in these rice varieties will have a greater impact in terms of changes in quantity

26 demanded. Similarly, both maize and the other cereals categories had significant own-price coefficients.

The expenditure coefficients for all categories are significant, with the exception of the average-quality domestic rice category. All of the significant coefficients bear a positive sign, indicating that these are luxury categories. Because rice in these categories is usually of low quality and the cheapest on the market, this result was unlikely for two of these categories, the imported rice and other domestic rice categories. However, low-income consumers regard rice from these categories as luxuries because they normally can only afford to buy maize for their consumption. The other cereals category was positive, indicating that they are a luxury, while the maize category has a negative sign, indicating a necessity.

Table 4. Linearized Almost Ideal Demand System (LAIDS) Parameter Estimates

Category Imported High Average Other Other Maize rice quality quality domestic cereals domestic domestic rice rice rice γ (price coefficient)

Imported rice -0.0363 (0.5958) High quality domestic 0.0278 0.1326* rice (0.3666) (0.0001) Average quality 0.0922* 0.0899* 0.0752 domestic rice (0.0034) (0.0001) (0.0019) Other domestic rice 0.0392 -0.0044 0.0937 -0.0235 (0.4040) (0.8977) (0.0032) (0.7432) Other cereals -0.0183 -0.0027 -0.0065 0.0137 -0.0305* (0.3383) (0.8424) (0.5877) (0.4745) (0.0071) Maize -0.0579** -0.0837* -0.0798* -0.0563** 0.0171*** 0.0725* (0.0265) (0.0001) (0.0001) (0.0222) (0.0774) (0.0001) α -0.3608 -1.0097*** -1.6257* -0.7715 0.1245 1.8811* (0.6004) (0.0167) (0.0001) (0.2503) (0.5780) (0.0001)

Continue 27

Continue β (expenditure coefficient) 0.0329* 0.0144*** -0.0037 0.0408* 0.0117** -0.0169** (0.0015) (0.1367) (0.6674) (0.0011) (0.0447) (0.0341) Inverse Mills ratio 0.1163* 0.1003* 0.4629* 0.5266* 0.0789* - (34.13) (31.39) (10.7) (38.57) (20.23) -

(...) P-values coefficients (except for the inverse Mills ratio coefficients, which present t-values) in bold are significant at 1% (*), 5% (**), 10% (***)

4.3 Elasticity Results

Expenditure coefficients for all cereal categories are presented in Table 5. The expenditure elasticities are all positive, indicating that all cereal categories are normal goods with coefficients ranging from 1.07 to 0.18. The smallest expenditure coefficient is for maize (0.18), indicating that an increase in income leads to a very small change in the demand for maize; this result is not surprising, as maize is the country’s staple food. A normal Tanzanian lunch or supper is comprised of Ugali, a stiff porridge made out of maize flour. A close to unitary expenditure elasticity is seen for the average-quality rice category (0.99). Expenditure elasticities greater than one, which indicates luxury goods, is observed for high-quality domestic rice (1.02), imported rice (1.07), other domestic rice (1.07), and other cereals (1.04). These results imply that rice is regarded as a luxury good by all income levels. Thus, increased income would likely lead to a shift in consumption from maize to rice. Similarly, the other cereal category was observed to be a luxury good; this category includes cereal products from wheat, sorghum, and millet. Products made from these materials, such as bread, cakes, and biscuits, are usually expensive and regarded as luxuries.

28

Table 5. Expenditure Elasticity and Budget Shares

Category Expenditure elasticity Budget shares

Imported rice 1.0736 0.05 High quality domestic 1.0245 0.13 Average quality domestic 0.9934 0.32 Other domestic varieties 1.0725 0.07 Other cereals 1.0497 0.08 Maize 0.1803 0.36

Table 6, presents the Marshallian (Uncompensated) Elasticities’ own- and cross-price elasticity estimates. All own-price uncompensated elasticities are negative, suggesting that these cereal categories conform to the law of demand. Estimates for imported rice, other domestic varieties, and the other cereals categories are all greater than unity, -1.08 -1.04 and -1.13, suggesting that these categories are price elastic. Meanwhile, the own-price elasticities for high- quality domestic category, average-quality domestic category and maize are less than unity, -0.78

-0.86 and -0.24, indicating an inelastic response in the change in demand to the own-price change. Among the rice categories, the least price sensitive rice was the high-quality domestic category. For the remaining categories, the least sensitive cereal to price changes was maize, an expected finding because maize is the country’s food staple.

The magnitude of coefficients for the cross-price elasticities are smaller in absolute terms compared to those of own-price elasticity, suggesting that consumers are more responsive to changes in own-prices compared to cross-price changes. Most of the cross-prices are positive, indicating substitutability between the cereal categories, an expected result because all the categories belong to the same food group. Weak substitutability is observed when looking at the demand of domestic rice categories to changes in the price of imported rice; the cross-price

29 elasticity coefficients are 0.02 for high-quality domestic rice, 0.03 for average-quality domestic rice, and 0.06 for other domestic varieties. The low elasticity estimates could justify claims of little substitutability between domestic and imported varieties due to consumer preference. It is also interesting to note that the other domestic rice category showed the highest substitutability response to changes in the price of imported rice (0.065). This result suggests that imported rice is relatively more substitutable for rice in the other domestic rice category compared to rice from the high-quality domestic and average-quality domestic categories. This finding makes intuitive sense since the other domestic rice category is mainly comprised of low-quality domestic rice varieties, which are generally comparable in quality to the imported rice. Maize was expected to be a close substitute to the imported rice because they fetch nearly the same price on the market; however, it had the lowest elasticity estimate of all cereal categories (0.01), indicating very little substitutability.

The cross-price elasticity among categories of domestic rice is positive, suggesting substitutability within domestic varieties of rice. This finding holds true except for the estimate between the other domestic rice categories and the two domestic varieties; the high-quality domestic rice and average-quality domestic rice had negative estimates (-0.01) and (-0.02), which suggests complementarity between these categories. The substitutability dynamics between domestic categories is a satisfactory reflection of reality. The high-quality domestic and the other domestic category are more responsive to changes in price than the average domestic category

(0.14) and (0.79); thus, greater substitutability exists between average-quality domestic rice and other domestic rice, as they are regarded as close substitutes on the market when compared to other categories of rice or other cereals. Similarly, the demand for the other domestic category is more responsive to changes in the price of average-quality domestic rice. The other cereals category and maize indicate a complementary relationship with the high-quality and average

30 quality domestic categories. Despite the fact that these categories are not usually consumed simultaneously to warrant complementarity, they are usually consumed interchangeably within a day’s meals. For example, an individual might have porridge made of millet (from the other cereals categories) for breakfast, Ugali stiff porridge made out of maize flour (from the maize category) for lunch and a rice dish for supper (from the high-quality domestic and average-quality domestic categories).

Table 6 shows the cross- and own-price elasticities for compensated elasticity. Own-price compensated elasticity shows similar trends to their counterparts’ uncompensated own-price elasticities; they all bear a negative sign conforming to demand theory. The own price elasticity estimates for compensated elasticity are smaller in magnitude compared to uncompensated own price elasticity estimates, this suggests that the income effect on the quantities demand for the different categories is an important factor for consumers. The compensated own-price elasticity for imported rice (-1.08), other domestic rice (-1.04), and other cereals (-1.13) were elastic, while high-quality domestic rice (-0.78), average- quality domestic rice (-0.86), and maize (-0.24) are price inelastic. The compensated cross-price elasticity shows substitutability for all cereal categories except for maize, high-quality domestic rice, and average-quality domestic rice. The compensated cross-price elasticities between imported and domestic rice categories are higher compared to the uncompensated cross price elasticities, suggesting that there is little substitutability between these two rice categories. The change in sign from negative to positive in some of the compensated cross-price estimates further emphasizes the presence of greater income effects on consumers.

The compensated demand estimates offer the best estimates when looking at the substitutability between different categories of goods, this is because these estimates offer a

31 purely substitution effect unlike the uncompensated elasticity estimate which constitute both the substitution effect and the income effect.

In a scenario where prevailing market issues like rice mixing are taken into account, the obtained elasticity results would imply that the high cross-price estimates observed among domestic rice categories partly include the substitutability effect between domestic and imported varieties. This suggests that, if the mixing were to be removed, the coefficient for cross-price elasticity between domestic and imported rice categories would increase in magnitude, thus showing greater substitutability between domestic and imported rice varieties.

32

Table 6. Compensated and Uncompensated Elasticity Estimates

Imported High Average Other Other Maize rice quality quality domestic cereals domestic domestic rice rice rice varieties

% ∆ in quantity 1 % ∆ prices demanded

Uncompensated elasticities

Imported rice -1.0788 0.0553 0.1819 0.0903 -0.0444 -0.2779 High quality 0.0225 -0.7772 0.1444 -0.0066 -0.0058 -0.4272 domestic rice Average quality 0.0305 0.0622 -0.8642 0.1663 -0.0113 -0.6080 domestic rice Other domestic 0.0653 -0.0194 0.7937 -1.0392 0.0209 -0.2537 rice varieties Other cereals -0.0265 -0.0132 -0.0662 0.0196 -1.1322 0.1590 Maize 0.0054 -0.0459 -0.2881 0.0119 0.1000 -0.2433 Compensated elasticities

Imported rice -1.0279 0.1951 0.5293 0.1608 0.0373 0.1054 High quality 0.0711 -0.6438 0.4759 0.0606 0.0721 -0.0615 domestic rice Average quality 0.0776 0.1915 -0.5427 0.2315 0.0643 -0.2534 domestic rice Other domestic 0.1162 0.1203 1.1408 -0.9688 0.1025 0.1292 rice varieties Other cereals 0.0233 0.1235 0.2734 0.0885 -1.0523 0.5337 Maize 0.0140 -0.0224 -0.2297 0.0238 0.1137 -0.1789

Bolded estimates are own-price elasticities

33

CHAPTER 5: CONCLUSION AND POLICY IMPLICATIONS

Tanzanian rice consumers show a greater preference for domestic rice varieties compared to imported varieties. Although imported rice is a third cheaper than domestic rice, the majority of consumers still opts to consume the relatively more expensive domestic rice. Preference for domestic rice is further underscored by the low substitutability between domestic and imported rice categories.

Weak substitutability observed between domestic and imported rice categories offers little justification for country’s rice import tariff. However it is impractical to completely disregard concerns about the effects of imported rice on the domestic rice market. Especially when prevailing issues like rice mixing are considered, imported rice has the potential to impact the domestic rice market. In the face of these market issues, the presence of a rice import tariff is essential in keeping the inflow of imports in check, thus protecting the domestic rice subsector in a rather complex rice market. The prevailing protective environment for the Tanzanian rice subsector and increased government and private sector production initiatives seem to be insufficient in boosting the Tanzanian rice market. There is a need for further research in the subsector that has the potential to identify strategic marketing initiatives that would enable the unique identification of domestic rice varieties through branding while guaranteeing markets for the farmers.

The demand for major rice categories (high-quality and average-quality domestic rice) is not quite responsive enough to changes in own-prices, income, and other household demographics, such as education, age, and location. Thus for any food policy to be effective, they

34 must consider these factors of utmost importance. Cereal categories with the highest budget share were average-quality rice and maize, which suggests that a substantial price decrease in these two categories as a result of increased production would be beneficial to the majority of households; thus, food security policies in the country should focus on enabling increased production in maize and rice varieties in average- quality categories.

35

BIBLIOGRAPHY

ACT. "Report on Market Output Support: Business Care Services and Centre for Sustainable Development Initiatives," 2007.

ACT. "Value Chain Analysis of Rice and Maize in Selected Districts in Tanzania, Volume I: Introduction, Context Analysis and Recommended Way Forward Final." Report Match Maker Associates Limited, 2010.

Andreyeva, Tatiana, Long, Michael W, and Brownell, Kelly D. "The Impact of Food Prices on Consumption: A Systematic Review of Research on the Price Elasticity of Demand for Food. American Journal of Public Health, 100 no. 2 (2010): 216.

Babalola, DA, and Isitor, SU. "Analysis of the Determinants of Food Expenditure Patterns among Urban Households in Nigeria: Evidence from Lagos State." Journal of Agriculture and Veterinary Science (IOSR-JAVS) 7 no. 5 (2014): 71-75.

Bill and Melinda Gates Foundation. "Developing the Rice Industry in Africa. Tanzania Assessment," July 2012.

Blow, Laura. "Demographics in Demand Systems: IFS Working Papers." Institute for Fiscal Studies (IFS), 2003.

Brigham, Anne Margrethe. "Agricultural Exports and Food Insecurity in Sub‐Saharan Africa: A Qualititative Configurational Analysis." Development Policy Review, 29, no. 6 (2011): 729-748.

Capps, O. "Utilizing Scanner Data to Estimate Retail Demand Functions for Meat Products." American Journal of Agricultural Economics 71, (1989): 750-760.

Chern, Wen S. "Analysis of the Food Consumption of Japanese Households. (2003).

Deaton, Angus. "Estimation of Own and Cross-Price Elasticities from Household Survey Data." Journal of Econometrics 36, no. 1 (1987): 7-30.

Deaton, Angus. "Quality, Quantity, and Spatial Variation of Price." American Economic Review 78, (1988): 418-430.

Deaton, Angus. (1990). "Price Elasticities from Survey Data: Extensions and Indonesian Results." Journal of Econometrics 44, no. 3 (1990): 281-309.

36

Deaton, Angus, and Muellbauer, John. (1980). "An Almost Ideal Demand System. The American Economic Review (1980): 312-326.

DGCIS. Summary of Trade Data. Directorate General of Commercial Intelligence and Statistics Ministry of Commerce and Industry. Government of India, 2002.

Diao, X, Mabiso, A, and Kennedy, A. Economy-Wide Impact of Maize Export Bans on Agricultural Growth and Household Welfare in Tanzania – A Dynamic CGE Model Analysis USAID Feed the Future Contracted Project to IFPRI, 2012.

EAC. East African Community Customs Union Common External Tariff, 2012.

EUCORD. Rice Sector Development in East Africa: A Desk Study Prepared for the Common Fund for Commodities, 2012.

Barreiro-Hurle,J. "Analysis of Incentives and Disincentives for Rice in the United Republic of Tanzania." Technical note series.Roma: FAO, MAFAP 2012.

The East African. "Dar Forced to Import Rice to Stabilize Price of Local Harvest." January 5th, 2013.

The Guardian. "Government Revokes Tax Relief on Imported Rice." October 13th 2013.

Heien, Dale, and Wesseils, Cathy Roheim. "Demand Systems Estimation with Microdata: A Censored Regression Approach." Journal of Business & Economic Statistics, 8, no. 3 (1990): 365-371.

Hoang, Linh Vu. "Estimation of Food Demand from Household Survey Data in Vietnam." DEPOCEN Working Paper Series 2009.

Khorana, Sangeeta, Kimbugwe, Kato, and Perdikis, Nicholas. "The Impact of Tariff Reductions Under the East African Community Customs Union: Intra-trade Effects on Uganda."2007.

Lallukka, T, Laaksonen, M, Rahkonen, O, Roos, E, and Lahelma, E. "Multiple Socioeconomic Circumstances and Healthy Food Habits." European Journal of Clinical Nutrition 61, no. 6 (2007): 701-710.

Lewis, Ian. "Tanzania’s Rice Sub-Sector and Value Chain- Analysis & Review and Proposed Vision & Strategy for Improved Competitiveness & Growth." 2012.

MAFC. National Rice Development Strategy - Final Draft. May 2009. Ministry of Agriculture, Food Security and Cooperatives. The United Republic of Tanzania, 2009. . Minot, Nicholas. Staple Food Prices in Tanzania. Paper presented at the Contributed Paper Prepared for the COMESA Policy Seminar Maputo, Mozambique, 2010.

37

Morrissey, O, and Leyaro, V. (2009). "Distortions to Agricultural Incentives in Tanzania." In Ole Therkildsen 2011: Policy Making and Implementation in Agriculture: Tanzania’s Push for Irrigated Rice. Edited by K. Anderson and W. Masters, 2009.

Moschini, Giancarlo. "Units of Measurement and the Stone Index in Demand System Estimation." American Journal of Agricultural Economics 77, no. 1 (1995): 63-68.

Nzuma, Jonathan M, and Sarker, Rakhal. "An Error Corrected Almost Ideal Demand System for Major Cereals in Kenya." Agricultural Economics 41, no. 1 (2010): 43-50.

Ogundele, Olorunfemi. "Factors Influencing Consumers’ Preference for Local Rice in Nigeria. African Journal of Marketing Management 6, no. 4 (2014): 49-55.

Pollak, Robert A, and Wales, Terence J. "Estimation of Complete Demand Systems from Household Budget Data: the Linear and Quadratic Expenditure Systems." The American Economic Review (1978): 348-359.

Ray, Ranjan. "The Testing and Estimation of Complete Demand Systems on Household Budget Surveys: An Application of AIDS." European Economic Review 17, no.3 (1982): 349- 369.

Roos, Eva, Lahelma, Eero, Virtanen, Mikko, Prättälä, Ritva, and Pietinen, Pirjo. "Gender, Socioeconomic Status and Family Status as Determinants of Food Behaviour." Social Science & Medicine, 46 no. 12 (1998): 1519-1529.

Sam, Abdoul G, and Zheng, Yi. (2010). "Semiparametric Estimation of Consumer Demand Systems with Microdata." American Journal of Agricultural Economics (2014).

Sekhampu, Tshediso Joseph. "Socio-Economic Determinants of Household Food Expenditure in a Low Income Township in South Africa." Mediterranean Journal of Social Sciences, (2012): 449.

Shonkwiler, J Scott, and Yen, Steven T. "Two-Step Estimation of a Censored System of Equations." American Journal of Agricultural Economics 81, no. 4 (1999): 972-982.

Stryker, D, & Amin, M. Study of Policy Options for Increasing Tanzanian Exports of Maize and Rice (Draft) – Improving Food Security to 2025. USAID Feed the Future contracted project to AIRD, 2012.

Therkildsen, Ole. Policy Making and Implementation in Agriculture: Tanzania's Push for Irrigated Rice: Danish Institute for International Studies, 2011.

URT. Trends in Food Insecurity in Mainland Tanzania. Food Security and Nutrition Analysis of Tanzania Household Budget Surveys 2000/1 and 2007 April 2010. Ministry of Finance and Economic Affairs & National Bureau of Statistics Dar es Salaam, 2010.

38

USAID. MicroCLIR/CIBER ASSESSMENT. The Legal, Policy, Regulatory and Institutional Constraints to the Growth of Maize and Rice in Tanzania. Agenda for Action, August 2010.

USDA. Average Global Rice Production, 2013. http://www.fas.usda.gov/psdonline/psdreport.aspx?hidReportRetrievalName=BVS&hidR eportRetrievalID=681&hidReportRetrievalTemplateID=7.

Weliwita, Anada, Nyange, David, and Tsujii, Hiroshi. (2003). "Food Demand Patterns in Tanzania: A Censored Regression Analysis of Microdata." Sri Lanka Journal of Agricultural Economics 5, (2003): 10- 29.

39