WILLINGNESS TO PAY FOR PROCESSED GRAINS IN DAKAR SENEGAL: AN ANALYSIS USING DISCRETE CHOICE EXPERIMENTS By Sarah Victoria Chase - Walsh A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Agricultural, Food, and Resource Economics Master of Science 201 9 ABSTRACT WILLINGNESS TO PAY FOR PROCESSED GRAINS IN DAKAR SENEGAL: AN ANALYSIS USING DISCRETE CHOICE EXP ER IMENTS By Sarah Victoria Chase - Walsh This article studies consumer preference for processed traditional and non - traditional grains in Dakar, Senegal. While much attention has focused on substitution between traditional and nontraditional grains, less has shown how consumers make tradeoffs among processed products . Using an exit - interview method and two discrete choice experiments, I obtain marginal values of willingness to pay for processed grains. In this paper, I measure willingness to pay for domestically produced millet, maize, rice, and sorghum. I also measur e willingness to pay for a second stage processed millet product. The results of this study show that consumers are willing to pay a premium for domestically produced processed grains, both traditional and non - traditional, save sorghum. Consumers are only willing to pay a positive premium for imported rice. The results also suggest that consumers are willing to pay a premium for fresh and bul k second stage processed millet iii ACKNOWLEDGEMENTS I would like to acknowledge my friends and family for their continued support. Thank you to my AFRE cohort and professors. I would also like to acknowledge my collaborator, Anta Ngom, for her assistance and friendship, and my committee f or their support, feedback, and guidance. iv TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ........................ v LIST OF FIGURES ................................ ................................ ................................ ..................... vi 1. INTRODUCTION ................................ ................................ ................................ ..................... 1 2. CONTEXT AND METHO DS ................................ ................................ ................................ .. 5 2.1 DATA ................................ ................................ ................................ ................................ .... 6 2.1.1 Sample ................................ ................................ ................................ ............................ 6 2.1.2 Sample Characteristic ................................ ................................ ................................ .... 7 2.2 SURVEY DESIGN AND EXPERIMENTAL PROCEDURES ................................ .......... 11 2.2.1 Inter and Intra Grain Choice Experiment Designs ................................ ...................... 11 2.3 EMPIRICAL MODEL: THE RANDOM PARAMETER LOGIT ................................ ...... 14 3. RESULTS AND DISCU SSION ................................ ................................ ............................. 17 3.1 INTER - GRAIN RESULTS ................................ ................................ ................................ . 17 3.2 INTRA - GRAIN RESULTS ................................ ................................ ................................ 24 4. CONCLUSION ................................ ................................ ................................ ....................... 30 APPENDIX ................................ ................................ ................................ ................................ .. 32 REFERENCES ................................ ................................ ................................ ............................ 43 v LIST OF TABLES Table 1 : Socio - D emographic C haracteristics of the S ample, sample size (N=596) Table 2: Overall Millet Purchasing History (N=596) Table 3: Thièré Purchasing Habits (N=596) . 10 Table 4: Coarse Grain Attributes and Levels Table 5: Second Stage Processed Millet Attributes and Levels Table 6: Results from Inter - Grain Random Parameters Logit Table 7 : Total Willingness to Pay for the First Stage Processed Coarse Grains (500g) Table 8 : Willingness to Pay for Grains Produced Domestically by Sample Demographics Table 9: Willingness to Pay for Imported Grains by Sample Demographics Table 10: Results from Intra - Grain Random Parameters Logit Table 11: Willingness to Pay for Second Stage Processed Millet Thièré (500g) Table 12: Willingness to Pay for Thièré by Sample Demographics vi LIST OF FIGURES Figure 1: Sample Inter - Grain C hoice S et .12 Figure 2: Sample Intra - Grain C hoice S et ..14 1 1. INTRODUCTION Between 2010 and 2016, Senegal's per capita gross national income increased by 18%, one of the highest rates in the region (World Bank 2018). Most of this growth is happening in the capital of Dakar, which is growing at a ra te of 4 . 4% (ibid). In rapidly growing Sub - Saharan African cities like Dakar, the increasing opportunity cost of time is changing the way households buy and prepare food. As incomes and the opportunity cost of time increase, demand for processed foods also increase (Reardon and Timmer 2012). Even poorer households are demanding more convenient forms of food as their time becomes more valuable. There are three main waves of literature that have focused on the rise of processed foods in developing countries. The early debate developed the theory for household time allocation and - traditional grains, mainly rice and wheat, and raw traditional grains , such as millet, processed at home. A need for f urther understanding of the debate prompted the second wave, which focused on substitution between processed non - traditional grains and first stage processed coarse grains. The recent debate, now focuses on substitution among (a) processed non - traditional grains, (b) first stage processed coarse grains, and (c) the second stage processed coarse grains. Th ese three main categories of processed grains, using millet as an example, are defined as follows. First, raw millet, refers to harvested, threshed, clean ed and sorted. Millet flour is first stage processed. It has been cleaned and milled. Second stage processed millet refers to the transformation of millet flour into different products . These forms of processed grains are critical to the analysis of the pr ocessed food literature. Additionally, inter - grain, refers to the study among various grain types. Intra - grain refers to the study within a specific grain type. 2 The first wave of the debate focused on substitution between processed non - traditional grains, mainly rice and wheat and coarse grains processed at home. The opportunity cost of time is key to understanding processed food consumption, first incorporated into the literature by Becker in "Theory of the Allocation of Time" (1965). The theory of opportu nity cost of time was applied to the developing country context by Thompson and Schuh (1975). The connection between the opportunity cost of time and processed food was found by Senauer, Sahn, and Alderman when they studied rice and wheat consumption in Sr i Lanka (1986). The debate was further extended when Kennedy and Reardon determined that poorer households sought more processed rice than wealthier households, in Burkina Faso (1993). These studies show the complexities between rising opportunity cost of time and the demand for processed food. The earlier debate noted that in Asia and Africa, the substitution was between coarse grains that traditionally had not been purchasable in a processed or packaged form and new, non - traditional grains. These coarse g rains required more time at home to process and prepare, compared to the new, non - traditional products, mainly rice and wheat. In Burkina Faso, workers ate prepared rice at lunch from street vendors and purchased wheat for at - home consumption due to its sh orter preparation time (Reardon, Thiomiano, & Delgado 1989). growing demand for convenience inspired a second wave of literature. The second wave of the literature developed as millet processors began selling processed and packaged millet. First , a three - way substitution debate emerged. Pitted against each other were: (1) processed non - traditional grains, rice and wheat; (2) first stage processed millet; (3) buying raw millet and processing at home. The third option was eventually eliminated in u rban , the literature targeted the rise of imported rice and wheat, and governments pushed the consumption of local 3 grains and domestic rice production (Demont & Neven 20 13). The first stage processed and packaged millet emerged to compete with processed rice and wheat products (Bricas 2008). The demand for processed traditional grains led to the rise of small - scale millet vendors who sold directly in food insecure neighb orhoods where incomes were slowing increasing (Bricas 2008). Higher rates of employment and increasing incomes led to the third wave of the literature. The third and current wave of the literature indicates that grain substitution now includes another ele ment. Consumers substitute between: (1) processed non - traditional grains, such as rice and wheat; (2) first stage processed millet; and now; (3) second stage processed millet. Second stage processed millet refers to any product using transformed millet flo ur. The first and second stages remain the same but adding this third wave to the debate changes views of these products. For instance, in Ethiopia, teff, a traditional grain used to prepare second stage products (most commonly a bread called en jera) , was found to be consumed by the relatively well - off households and its purchase increases with income (Alem & Söderbom 2018). Moreover, households headed by individuals with better labor market status consumed relatively more teff than those with po or labor market status (ibid). Lastly, the literature has incorporated different research methods to understand how consumers substitute among food products. Through case studies of micro - enterprises and hypothetical discrete choice experiments, s tudies ha ve assessed how consumers in developing countries substitute among various foods. Initially, cases studies provided a cost - effective way to understand food preference changes. For example, case studies of small - scale African food producers have shown that consumer demand for processed foods has increased food security and female employment ( Bricas 2008 ). Second ly , discrete choice experiments (DCE) help researchers in developing countries affordably collect unique datasets to interpret these changes. For ins tance, 4 i n India, urban consumers' preferences for food safety and quality revealed that wealthier, more educated consumers with children prefer safer and higher quality food (Roy et al. 2010). Additionally, DCEs revealed that rural Ugandan banana consumers were, on average, willing to accept GM varieties of bananas (Kikulwe, Birol, Wesseler, & Falk - Zepeda, 2008). The literature thus far leaves three gaps. The first is the issue of substituting non - traditional grains and traditional processed grains. While studies have looked at the initial substitution issues among traditional and non - traditional grains, few have studied this when incorporating processed traditional grains. The second gap is understanding how consumers substitute among products of the same processed traditional grain. Again, there has been little work into the tradeoffs for processed traditional grains, such as millet. The third gap is a methodological gap of using DCE DCE have rarely been used to determine the demand for processed food products in a developing country. This paper contribu tes to the se gaps in the literature in two main ways . First, it contributes findings about how consumers in an urban, developing country context make tradeoffs among processed millet, maize, rice, and sorghum. This is done using discrete choice experiment to find willingness to pay (WTP) values for processed grains. The second contribution is made through studying the new debate about tradi tional grains like millet. This is achieved using a discrete choice experiment to determine willingness to pay for attributes of a second stage processed millet product called thièré. The rest of the article proceeds as follows. Section 2 presents the met hods used to collect data and the experimental procedures. Section 3 reports the results of the two DCEs and the discussion. The article concludes with Section 4. 5 2. CONTEXT AND METHODS As indicated earlier, there are few critical analyses of consumption of processed traditional and non - traditional grains in Africa. In this study, I used a mixed method approach to assess - interviews to gain demographic and purchasing habits. Then, through the use of DCE , I calculate willingness to pay for different grains. I also used DCE to determine willingness to pay for second - stage processed millet. The literature suggests that exit interviews have a greater external valid ity than studies in laboratories due to the priming of prior food purchases (Minten, Reardon & Sutradhar 2010). This study sampled consumers in Dakar for two reasons. First, Dakar is the largest region in Senegal and the economic and political capital (AN SD, 2018). The food system in the city is changing rapidly as the city is growing at a rate of 4 . 4% per year (World Bank 2018). Second, with wide distribution of incomes, consumers in Dakar have more diverse consumption patterns than the rest of the popula also has the various social classes necessary to capture variability among consumers. These reasons indicated that Dakar was the most efficient area to sample consumers. Lastly, studying consumer pr eferences among millet varieties is critical for two main reasons. First, while v arious processed coarse grains are found in Dakar markets, including millet, maize, rice, and sorghum, millet is a widely consumed cereal. There are three main categories of p rocessed millet. First, raw millet, which refers to harvested, threshed, cleaned and sorted. Millet flour is first stage processed. It has been cleaned and milled. Second stage processed millet refers to the transformation of millet flour into different pr oducts. The second reason for studying millet specifically is to understand demand for traditional grains with new processing techniques. 6 Learning how consumers purchase and make tradeoffs among millet - based products is necessary to understanding the large r changes in the food - system. 2.1 DATA 2.1.1 Sample The sampling framework was set up as follows: First, communes were chosen for sampling areas. Second, retail types were chosen. Third, enumerators sampled respondents at the retail locations. The retail exit survey of 597 consumers was conducted in January and February of 2018 . The survey was collected in 12 communes in Dakar. Commune level population statistics were not available at the time of sampling, so geography experts were consulted to label a d by population density into four groups: dense, moderately dense, less dense, and not densely populated communes. Three communes were then randomly selected from within each density stratum. Within each sampled commune, three retail types were sampled: supermarkets, boutiks, and street - vendors. Boutiks are small neighborhood shops, often located on a corner. Street - vendors are generally women who sell cereal products in their neighborhood. Sam pling at these three retail locations was necessary in order to capture the variability among consumers. All supermarkets in the Dakar region were sampled due to their limited number. Boutiks and street vendors were sampled using a street - by - street approac h to avoid sampling bias. Enumerators first found the most densely populated corner of the commune. They then walked down a street , stopped at a boutik , and sampled the first two customers who had finished shopping. Then , they continued down the street, sa mpling from every other boutik they encountered until they sampled the required amount of respondents. The same method was used for the street vendor sample. Within each 7 commune, enumerators started at the same starting point for both the boutik and street - vendor sample. The questionnaire was structured as follows. First, respondents were asked about household demographics, consumption patterns, and mealtime habits. Second, respondents participated in two DCEs . The first experiment asked respondents to make hypothetical choices among millet, maize, rice, and sorghum as a function of different scenarios of price and production origin. The second experiment asked respondents to make hypothetical choices about a second stage processed millet product called thiè ré, with different processing techniques and packaging . Third, respondents answered questions regarding their purchase history and product preferences. The interview and choice experiment were conducted as follows. First, one enumerator approached the firs t consumer he saw leaving the retail outlet. Then, the enumerator read the questionnaire aloud and filled in the responses on the tablet. Next, the enumerator explained the choice experiment procedure to the respondent and proceeded. The enumerator read th e choice tasks aloud for all participants since it was understood that some respondents could not read French. Then, the enumerators concluded the interview with questions about purchase history, again reading the questions aloud and filling in the respons es. Purchase history was defined by enumerators as physically making the purchase, not simply consuming. Lastly, the enumerator thanked the participant and either continued with another customer or left for another retail outlet. See Appendix for the quest ionnaire. 2.1.2 Sample Characteristic The socio - demographic characteristics of the sample are reported in Table 1. Half of the sample (5 0 %) were male. This is consistent with other consumption studies in the region (Bello & Awudu 2016). The sample is representative of households since the majority of the sample were heads of 8 their households (44%) or wives (31%). Additionally, two - thirds of respondents (67%) reported purchases was defined as physically selecting food at a retail location, not simply financing the purchases. Of the remaining respondents, 20% were adult children of the household head, two percent were domestic workers, and thre e percent were other family members. Urban households are generally smaller than rural households. The average household in Dakar has six members, while the national average is eight (ANSD 2014). The sample mean and median household sizes were six and fiv this study, household per capita weekly food expenditure is used as a proxy for household income. The mean household weekly food expenditure per capita was 12 USD, and the mean total weekly food expenditure was 56 USD per household . Table 1: Socio - Demographic Characteristics of the Sample, sample size (N=596 ) Socio - demographics characteristics % of total Male Female 50 5 0 Household position relative to household head Head of household Married to head of household Adult child Domestic worker Other 44 31 20 2 3 Highest level of education for all household heads None Primary Secondary University 24 23 26 27 Responsible for household food purchasing 67 Mean household size Median household size 6 5 9 Table 1 Mean per capita household weekly food expenditure 11.27 USD Note. 1 USD equals 517 FCFA in Jan 2018 Consumers in Dakar frequently purchase millet based products. Table 2 reports respondent - level millet purchasing habits . 45% (N=268) of respondents had personally purchased a millet - based product at the survey sampling location. For example, of those respondents that were sampled at the supermarket, 23% had just purchased a millet b ased product. Additionally, 100% of street vendor consumers reported purchasing a millet based product as the street vendors sampled only sold millet. To further understand the consumption history of the sample, I expanded the time recall period. Consideri ng that 55% (N=328) of the sample had not purchased a millet based product on the day of the sample, I asked about those respondents for their millet purchase history over the past three months. Of the 55% (N=328) who did not purchase millet the day of, 87 % (N=285) had purchased a millet - based product in the past three months. This shows that only 13% of the sample had not purchased millet products recently. Table 2: Overall Millet Purchasing History (N=596 ) Purchased millet - based product at the time of in terview 45% Supermarket Boutik Street vendor 23% 12% 100% Purchased millet - based product in past three months (Excluding day - of customers) 87% Now I move from the general millet purchasing history to specific data on purchasing history of the second - purchase of thièré. More than half of the sample had purchased thièré in the p ast week. This is significant finding that suggests that second stage millet is a frequently purchased dish. Despite 10 this finding and the fact that 87% had purchased a general millet based product in the past three months, as mentioned above, 21% of respon dents reported that they had n ever personally purchased thièré. This can be explained by the fact that, for some, thièré is still prepared and consumed at home . This shows that a gap still exists between those who purchase and those who consume millet - based products. Table 3 also reports where respondents made their most recent thièré purchase. These findings are consistent with my previous observations in the field. Most consumers purchase their second stage millet products, like thièré, at str eet vendors with some purchases taking place at supermarkets. Table 3 shows that most respondents purchased their thièré at a street vendor (85%) and some of the sample (13%) purchased their thièré at the supermarket. Very few purchased their thièré at a b outique (1%). Table 3: Thièré Purchasing Habits (N=596) most recent purchase of Thièré (percent of sample) Today 13% Within Last week 53% Two weeks ago 4% Three weeks ago 3% One month ago 2% More than one month ago 4% Never 21% L ocation of most recent purchase of Thièré (percent of sample given past millet purchase) Supermarket 13% Boutique 1% Street vendor 85% 11 2.2 SURVEY DESIGN AND EXPERIMENTAL PROCEDURES 2 .2.1 Inter and Intra Grain Choice Experiment Designs I designed two DCEs as follows. In the first experiment, respondents were asked to make choices among four alternatives represented by second - stage processed grains commonly found in Dakar markets: millet, m aize, rice, and sorghum. Table 4 outlines the attributes and attribute levels selected for this study. The products were hypothetically offered as 500 grams of grain at various price levels which represent real prices found in the market. The price levels were selected based on my market research at the three sampling locations. The product origin was included because some grains, like rice, are produced domestically and imported. The literature suggests that imported rice from Asia is preferred to domesti cally produced rice (Demont 2013). An origin attribute determines consumer preference for imported grains. Table 4: Coarse Grain Attributes and Levels Attributes Levels Product Millet, Maize, Rice, Sorghum Price Millet (200, 250, 300) Maize (250, 300, 350) Rice (300, 450, 500) Sorghum (300, 450, 500) (FCFA) Origin Domestic, Imported The experimental design was critical to creating choices that elicited consumer preferences. Considering every first - stage processed grain at every combination of prices and product of origin levels, a full factorial labeled design would have resulted in 1,296 (3 4 2 4 ) choice questions. To reduce the number of questions and thus avoid respondent fatigue, we generated an orthogonal fractional factorial labeled design. In this design, which resulted in six choice questions, the prices 12 and country of origin levels of each first - stage processed grain are uncorrelated with the prices of each of the other three grains. A no - buy alternativ e was added to each choice question to mimic a more realistic shopping experience ( Caputo et al. 2018 ). A sample choice experiment question is illustrated in Figure 1. Figure 1 : Sample Inter - Grain Choice Set Which Product Would You Choose? Millet : Import ed 400 FCFA Maize : Domestic 200 FCFA Rice : Imported 500 FCFA Sorghum : Domestic 300 FCFA No buy The second choice experiment targets preferences within a specific millet - based product, thièré. During the experiment consumers were asked to make choices between two alternative thièré products with differing attribute levels, as reported in Table 5 . Ta ble 5: S econd Stage P rocessed Millet Attributes and Levels Attributes Levels Price 400, 500, 600 (FCFA) Preparation Dried, Fresh Packaging Bulk , Packaged 13 Bulk " or "Packaged". The two types of preparation were selected to reflect actual product availability. Fresh refers to millet that has been steam cooked and is ready to eat. Dried millet is millet that has been steam cooked and then dried for preservation. Fr esh and dried are the only two ways thièré is found in the market. Similarly, the two packaging levels were included after observing how different vendors sell thièré, regardless of the preparation. Bulk refers to millet that is purchased at any quantity. Packaged refers to millet that has been packaged prior to the purchase. It is important to note that all combinations of thièré preparation and packaging levels exist in Dakar. The price levels were determined by finding average prices of thièré, then crea ting a 100 FCFA minimum price and a 600 FCFA maximum price. The highest and lowest prices found were 750FCFA/500g and 300FCFA/500g, respectively. Most prices were around 450/500 FCFA. Using a difference of 100FCFA between prices reduced the likelihood of t he experimental design producing choices that were either obviously appealing or unappealing. Additionally, keeping the prices within 200 FCFA required respondents to seriously consider the other two attributes. Given these attributes, their levels, and th e number of alternatives in each choice task, there were 144 possible total choice questions. To reduce the number of choice questions, I generated an orthogonal optimal in the difference fractional factorial design (Street and Burgess 2007; Van Loo et al. 2014) . Overall, during the survey, respondents were faced with six choice questions, each represented by two alternatives for processed and packaged thièré and a no - purchase option, as displayed in Figure 2. 14 Figure 2 : Sample Intra - Grain Choice Set Which Product Would You Choose? Thièré 1: Dried Bulk (400 FCFA) Thièré 2: Fresh Packaged (600 FCFA) None of these products 2.3 E MPIRICAL MODEL: THE RANDOM PARAMETER LOGIT The data were analyzed using random parameter logit (RPL) (Train 2009). The RPL is state of the art in the applied discrete choice literature for many reasons. First, it allows for taste parameters to vary across the population and thus accounts for hetero geneity in consumer preferences (Train 2000). Second, it permits unobserved factors to be correlated over time; random - effects introduced b y the panel - nature of the data. In this case, each individual answered six choice questions in both choice experiment s. Third , it relaxes the independence of irrelevant alternatives (IIA) assumption (Revelt & Train 1998). Its advantages are also widely documented in the food choice e xperiment literature (Aprile 2012; Van Loo et al 2011; Ortega et al. 2011 ). In the RPL mo del, the unconditional choice probability can be expressed as follows: (1) where is the density function. Considering a sequence of observed choices by individual n , one for each choice task in the assigned sequence of T choice tasks, the logit probability in (1) takes the form: 15 (2) where i closed form solution in (1), the probabilities are simulated using Halton draws. Halton draws give a more efficient distribution of draws for numerical integration com pared to random draws (Bhat 2003; Train 2009). The observed portion of the utility, , in (2) varies depending on the experiment. In the experiment among traditional and non - traditional grains, it is as follows: (3) where are alternative specific constants indicating the different types of j grains (e.g., millet, maize, rice, and sorghum ) . PRICE is the k j . ORIGIN is the k origin in alternative j , and represents the effect of the k origin on the utility for the j th grains. The price enters in the model as a continuous variable, while the origin is effect coding (e.g. Bech and Gyrd - Hansen 2005 and Louviere et al. 2000). Hence, it is equal to 1 if domestic and - 1 if imported. As a result, the estimated alternative specific constants represent the averaged utility attached to each grain as specified in the experiment, while the estimated coefficients for origin indicate deviations from the estimated alternative constants (Molin and Timmermans 2010) . Finally, we assume that the coefficients of the origin and alternative - speci fic constants are independently normall y distributed in the population, while t he price coefficient is assumed to be fixed in the population. As for the intra - grain experiment, t he observed portion of utility in this choice experiment is: (4) 16 where ASC is the alternative specific constant representing the no - buy option. Price n jt is the price for a package o f 500 grams of second - stage processed millet, thièré ; Preparation n jt is a dummy variable equal to one if the product was prepa Packaging nj t is a exception of the price coefficient, all coefficients are assumed normally distributed. In both studies, t he estimates derived from the RPL were used to calculate the WTP s . In the inter - grain experiment, we calculated the total amount that each respondent is willing to pay for a specific grain and its domestic /imported origin relative to the no - buy option. For the intra - grain experiment , we calculated the marginal WTP which represents the average premium respondents are willing to pay for fresh and bulk thièré. 17 3. R ESULTS AND DISCUSSION The data shows that consumers are willing to pay more for domestically produced cereals and grains. The results also show that customers are willing to pay a premium for unpackaged, bulk thièré. Lastly the results suggest that consumers are willing to pay a premium for fresh thièré. 3.1 INTER - GRAIN RESULTS Table 6 displays the estimates from the RPL model of the inter - grain experiment. Table 6 : Results from Inter - Grain Random Parameters Logit Variables Millet Maize Rice Sorghum Alternative Specific Constants Mean 1.6 4 *** (5.03) 0.01 (0.03) 5.51 *** (10.20) - 2.88 *** (4.64) Sd. 3.6 6 *** (11.91) 2.45 *** (10.30) 5.46 *** (14.51) 1.93 *** (3.48) Origin Domestic (+1); imported ( - 1) Mean 3.55 *** (12.32) 2.48 *** (11.77) 1.35 *** (9.24) 1.42 *** (4.62) Sd. 2.73 *** (12.33) 2.38 *** (9.36) 1.86 *** (10.70) 1.13 ** (2.49) Price Mean - 0.003 *** (3.58) - 0.003 *** (3.58) - 0.003 *** (3.58) - 0.003 *** (3.58) Model statistics Log likelihood - 2759.6704 Number of Choices 3576 Number of participants 596 McFadden Pseudo R - squared 0 .52 1 Note: ***, **, * significance at 1, 5 and 10% level. bers in parenthesis are standard errors Results indicate that the price coefficient is negative and statistically significant at the 0.01 level . Hence, consistent with the economic theory, higher prices are associated with a lower likelihood 18 of p urchase for both traditional and non - traditional grains. The mean coefficients for millet and rice are positive and statistically significant. On the other hand, the mean coefficient for maize is not statistically significant and the mean coefficient for s orghum is negative and statistically significant. These results suggest that ric e and millet at the most preferred grains, while maize and sorghum trail farther behind. This evidence is consistent with recent consumption trends that also suggest rice and millet are the most popular of the grains included in our studying. In 2017, annu al millet consumption in Dakar was 15.3 kg per capita. Average annual rice consumption was 77.9 kg per capita . Annual , per capita maize consumption in Dakar was only 3 kg. Annual sorghum consumption was so low that in 2017 it was reported that only 0.1% of households in Dakar consumed any sorghum. Our results also indicate that the standard deviations of the grains are all statistically significant, suggesting that consumer preferences for different grain - types are heterogeneous. As previously described, an estimated alternative - specific constant is the main utility derived from that alternative averaged across the levels of the origin variable. Hence, to provide more insights we focus the discussion and interpretation of the results on the relative impact of origin on the selection of the traditional and non - traditional grans. The coefficients of the origin are all positive and statistically significant, indicating that consumers tend to prefer domestic grains. Other studies have shown that consumers pref er domestic over imported products (Shelicia Forbes - Brown et al. 2016; Chryssochoidis, Krystallis, & Perreas 2007 ). In addition to the existing literature, this study also shows that consumer preferences for imported versus domestic grains differ among grain type. To illustrate, the estimated coefficient for the effect of the origin attribute on millet is 3.55, while the estimated 19 coefficient on rice is 1.35. This indicates that the utility of purchasing millet and rice increases by 3.55 and 1.35 i f domestic respectively , while it decreases by - 3.55 and - 1.35 if imported respectively . These results are justified in two ways. First, millet sold in Dakar is almost certainly sourced by Senegalese producers, therefore consumers might not consider impor ted millet to be worth any additional amount of money. Moreover, many households in Dakar have a familial connection to agricultural households in the rest of the country. Thus, consumers might consider that purchasing local millet is supporting the countr y as well as family. Second, local rice consumption might follow similar logic. Our results suggest that consumers are willing to pay for domestic rice. This finding contributes to the literature that consumers prefer local rice (Demont, 2013). For the rea sons mentioned above, as well as increased effort by the government to support domestic rice production, consumers might feel that by consuming local rice they are supporting the goal to reduce dependence on imported rice. Table 7 reports the total mean W TP value for each of the first stage processed coarse grains , both domestic and imported. Table 7 : Total Willingness to P ay for the F irst S tage P rocessed C oarse G rains (500g) Grains Total WTP (in FCFA ) Domestic Imported Millet 1737 *** (412) - 642** (277) Maize 827 *** (166) - 834** (337) Rice 2262 *** (516) 1361 *** (274) Sorghum - 489* (266) - 1436 *** (541) 20 Note: ***, **, * significance at 1, 5 and 10% level. bers in parenthesis are standard errors The results indicate that consumers are willing to pay the highest price for domestic rice (2262 FCFA ). Interestingly, rice also represents the only imported grain for which consumers exhibited a positive willingness to pay (1361 FCA). This finding mirrors the current market where imported rice makes up a substantial proportion of purchased rice in Dakar. Domestically produced millet is the second most preferred grain, with a total WTP of 1737 (FCFA). Total WTP for imported millet, maize, and sorghum are negative which suggests that consumers would need to be compensated in order to consume these imported grains. Additionally, even domestic sorghum is not preferred by consumers with the only negative domestic total WTP. Lastly, the total WTP values also indicate that the preferences are heterogeneous across grains. Domestic rice is nearly 20% more preferred than even domestic millet. H owever, due to the reasons mentioned above, millet still has a higher relative willingness to pay value than maize, Domestic maize, despite having a pos itive and statistically significant total WTP value, is still 50% less desirable than domestic millet. Consumers are willing to pay a statistically significant premium for domestically produced grains , even though the preferences vary across grain types . T Due to high imports of rice, the government of Senegal is pushing for rice import substitution (Demont & Rizzoto 20 12). Senegalese rice production is growing to meet the demand for local rice. However, due to marketing and scale issues, imported rice remains cheaper than local rice (ibid). My results suggest that consumers in Dakar are willing to pay a positive premium for local rice. Therefore, the government of Senegal could continue to support local rice production as well as the post farm - gate value chain. Moreover, consumers are willing to pay for local millet as well so investing in similar improvements to the millet value chain would also benefit millet consumers. 21 T ables 8 and 9 r eport mean total willingness to pay values for each grain produced domestically and imported among different consumer groups. Consumer groups are defined based on various sample characteristics including gender, level of food expenditure per capita, frequency of millet purchases, and where the respondent was sampled , among others. The statistical significance of differences in WTP across consumer groups are tested using t - tests assuming unequal group variances. Table 8 : Willingness to Pay for Grains Produced Domestically by Sample Demographics Variable Obs Millet Maize Rice Sorghum Gender Male 278 2094 986 2483 - 629 Female 316 2078 922 2614 - 566 Low vs Mid and High Food Expenditure (Per Capita) Low 402 2136 911* 2638 - 613 Mid and High 194 1982 1034 2389 - 560 Low vs Mid Food Expenditure (Per Capita) Low 194 1982 1034 2389 - 560 Mid 201 2155 869 2742 - 603 Mid vs. High Food Expenditure (Per Capita) Mid 201 2155 869 2742 - 603 High 200 2112 955 2532 - 622 Low vs High Food Expenditure (Per Capita) Low 194 1982 1034 2389 - 560 High 200 2112 955 2532 - 622 Bought Millet Today (Per Capita) Yes 323 2087 960 2471 - 560*** No 272 2081 942 2658 - 638 Supermarket versus boutique sample location Supermarket 197 223** 883 2154*** - 609** Boutique 198 1971 1020 2688 - 524 Supermarket versus street sample location Supermarket 197 2239 883 2154*** - 609 Street 200 2043 951 2823 - 653 Boutique versus street sample location Boutique 198 1971 1020 2688 - 524*** Street 200 2043 951 2823 - 653 Household head versus other Head 335 2065 917 2609 - 570** Other 260 2109 996 2489 - 629 22 Table 8 Wife versus other Wife 188 2195 956 2682 - 609 Other 407 2033 950 2499 - 589 Household head versus wife Head 260 2109 996 2489 - 629 Wife 188 2195 956 2682 - 609 Household head has at least a primary school education versus lower than primary >= Primary 450 2147 914 2573 - 600 < Primary 145 1888 1069 2505 - 583 Household head has at least a secondary school education versus lower than secondary >=Secondary 314 2114 861*** 2429* - 614 < Secondary 281 2050 1053 2700 - 575 Household head has a university education University 160 1972 921 2179*** - 636 < University 436 2127 962 2695 - 581 Household has six or more members >= 6 284 2120 960 2389 - 574 < 6 311 2052 944 2709 - 616 Note: Means of a two - sample t - test with unequal varia nces. *=.10, **=.05, ***=.01 Table 9 : Willing ness to Pay for Imported Grains by Sample Demographics Variable Obs Millet Maize Rice Sorghum Gender Male 278 - 765 - 1029 1250** - 1726** Female 316 - 698 - 1006 1648 - 1677 Low vs Mid and High Food Expenditure (Per Capita) Low 194 - 854* - 1102*** 1217** - 1689 Mid and High 402 - 672 - 976 1582 - 1705 Low vs Mid Food Expenditure (Per Capita) Low 194 - 854* - 1102* 1217*** - 1689 Mid 201 - 626 - 1018 1764 - 1709 Mid vs. High Food Expenditure (Per Capita) Mid 201 - 626 - 1018 1764* - 1709 High 200 - 714 - 933 1408 - 1701 Low vs High Food Expenditure (Per Capita) Low 194 - 854 - 1102*** 1217 - 1689 High 200 - 714 - 933 1408 - 1701 Buy Millet Today Yes 323 - 688 - 978* 1426 - 1676** No 272 - 780 - 1063 1513 - 1728 Responsible for Food Purchase Yes 400 - 731 - 992 1521 - 1709 23 Table 9 No 195 1892 1004 2305 - 551 Share of household member who ate breakfast at home yesterday 100% 232 - 818 - 1042 1158*** - 1712 <100% 363 - 674 - 1000 1663 - 1692 Share of household member who ate lunch at home yesterday 100% 377 - 696 - 1017 1617** - 1698 <100% 218 - 789 - 1016 1205 - 1703 Share of household member who ate dinner at home yesterday 100% 536 - 723 - 1014 1575*** - 1699 <100% 59 - 795 - 1039 478 - 1704 Supermarket versus boutique sample location Supermarket 197 - 596 - 874*** 1288 - 1667 Boutique 198 - 754 - 1101 1517 - 1687 Supermarket versus street sample location Supermarket 197 - 596** - 874*** 1288 - 1667*** Street 200 - 838 - 1074 1591 - 1745 Boutique versus street sample location Boutique 198 - 754 - 1101 1517 - 1687*** Street 200 - 838 - 1074 1591 - 1745 Household head versus other Head 335 - 683 - 998 1652*** - 1677** Other 260 - 791 - 1041 1226 - 1729 Wife versus other Wife 188 - 630 - 926** 1749** - 1702 Other 407 - 776 - 1059 1335 - 1699 Household head versus wife Head 260 - 791 - 1041* 1226*** - 1729 Wife 188 - 630 - 926 1749 - 1702 Household head has at least a primary school education versus lower than primary >= Primary 450* - 654** - 998 1529 - 1698 < Primary 145 - 966 - 1074 1269 - 1704 Household head has at least a secondary school education versus lower than secondary >=Secondary 314 - 645* - 965** 1441 - 1700 < Secondary 281 - 825 - 1075 1494 - 1700 Household head has a university education University 160 - 722 - 869*** 1376 - 1692 < University 436 - 735 - 1072 1495 - 1703 Household has six or more members >= 6 284 - 815 - 1075 1352 - 1693 < 6 311 - 653 - 963 1570 - 1706 Note: Means of a two - sample t - test with unequal varia nces. *=.10, **=.05, ***=.01 24 We first focus our discussion on differences in w illingness to p ay for g rains p roduced d omestically across groups, reported in Table 8. The key finding from Table 8 is that preferences for all domestically produced grains are not statistically significantly different across income groups. This indicates that preferences for domestic grains are relatively homogenous across incomes (per capita household weekly food expenditure is a proxy for income). Moreover, this su ggests that for consumers in Dakar, domestic grains, both traditional and non - traditional, are normal goods. The only exception is domestically produced maize. The statistically significant result shows that lower income households are willing to pay less for domestic maize than mid or high income households, at the 10% level. Next, Table 9 reports total WTP values for imported g rains across groups. The same groups were used in Table 8 and Table 9. The key finding in Table 9 is that, relative to domestic gr ains, there are more statistically significant differences between income groups across imported grain types. For example, low income consumers are willing to pay less than mid and high income households for imported millet, maize, and rice. This is an imp ortant finding because it suggests that imported millet, maize, and rice are not inferior goods. The positive WTP values for imported rice allows me to suggest that imported rice is at least a normal good. This is consistent with much of the literature sur rounding rice consumption (Demont 2013). I cannot suggest that millet and maize are normal goods because the total WTP values are still negative and statistically significant. 3.2 INTRA - GRAIN RESULTS Estimates from the RPL are presented in Table 10 . 25 Table 10: Results from Intra - Grain Random Parameters Logit Variables Coefficients Estimates Preparation (dried (+1); fresh (0) Mean - 26.66 *** (4.67) Standard Deviation 38.17 *** (4.87) Packaging (bulk (+1); packaged (0) Mean 19.82*** (4.47) Standard Deviation 34.99*** (5.36) Price Mean - 0.002*** (2.31) NOBUY Mean - 2.45 (5.59) Model Statistics Log likelihood - 1620.926 Number of Choices 3582 Number of participants 597 McFadden Pseudo R - squared 0.5880986 Note: ***, **, *significance at 1, 5 and 10% level. ber in parenthesis are |t - stats| The estimated means for all attributes are statistically signifi cant, and the signs are as expected. For instance , the price coefficient is negative and statistically significant, indicating that as price increases, consumers demand less. The alternative specific constant indicates that the no buy option is also negative and statistically signif icant, suggesting that our sampled consumers preferred one of the product profiles as compared to the no - buy option. The coefficient for the bulk packaging attribute is positive and statistically significant , while the coefficient of preparation is negativ e and statistically significant thièré is fresh and unpackaged (bulk). The estimated standard deviations of the preparation and 26 p ackaging attributes are also statistically significant, indicating that consumer preferences for those attributes are heterogeneous. The estimated means and standard deviations of these c oefficients provide information on the share of the population that places positive value s on each of those attribute s and the share that pl aces negative value s (Train 2009). In this case, for example, the distribution of the p ackaging coefficient obtain ed an estimated mean of 19.82 and estimated standard deviation of 34.99 , such that 71% of the distribution is above zero and 29 % is below. In other words, about 71% of our sampled consumers prefer bulk t hièré , and about 29 percent prefer unpackaged t hièré . Similarly, the results indicate that 24% of the sampled population prefer dried, while the vast majority, 76% prefer fresh. Table 11 presen ts marginal willingness to pay values for dried and bulk thièré. Table 11: Willingness to Pay for Second Stage P rocessed M illet Thièré (500 g) Marginal WTPs Mean (in FCFA) Preparation (Dried) - 13273.8 Packaging (Bulk) 9866.04 As household sizes are quite large (average six members), meals are likely produced in large quantities. Thus, for ease of preparation, most food may be purchased fresh. The results also indicate t hat households prefer to purchase fresh thièré in bulk. By buying in bulk , consumers are in control of the quantity purchased , compared to a pre packaged product. Additionally, consumers are more accustomed to purchasing bulk . A fresh and bulk product provide s convenience through easily prepared second - stage millet based product in quantities that households demand. 27 Table 12: Willingness to Pay for Thièré by Sample Demographics Variable Observations Mean Willingness to Pay for Dried Preparation Mean Willingness to Pay for Bul k Gender Man 278 - 5491 1685 Woman 316 - 4578 2813 Low vs Mid Weekly Food Expenditure (Per Capita) Low 194 - 4623 2062 Mid 201 - 4236 3057 Mid vs High Weekly Food Expenditure (Per Capita) Mid 201 - 4236 ** 3057 High 200 - 6099 1621 High vs Low Weekly Food Expenditure (Per Capita) High 200 - 6099 1621 Low 194 - 4623 2062 Bought Millet Today No 323 - 4793 1327 Yes 272 - 5221 334 6 Bought Millet Past Three Month s No 40 - 5441 2564 Yes 283 - 4701 1876 Responsible for Food Purchase No 195 - 5057 573 Yes 400 - 4955 3067 Share of household member who ate breakfast at home yesterday 100% 363 - 4619 1753 <100% 232 - 5567 302 8 Share of household member who ate lunch at home yesterday 100% 377 - 4932 1815 <100% 218 - 5087 3002 Share of household member who ate dinner at home yesterday 100% 536 - 4786 2180 <100% 59 - 6828 2884 Supermarket versus boutique sample location Supermarket 197 - 5219 1405 Boutique 198 - 4400 1262 28 Table 12 Supermarket versus street sample location Supermarket 197 - 5219 140 5 Street 200 - 5344 4060 Boutique versus street sample location Boutique 198 - 4400 1262 Street 200 - 5344 4060 Household head versus other Head 260 - 5387 1948 Other 335 - 4679 2484 Wife versus other Wife 188 - 4331 375 5 Other 407 - 5292 1555 Household head versus wife Head 260 - 5387 194 8 Wife 188 - 4331 3755 Household head has at least a primary school education versus lower than primary >/= Primary 450 - 5169 2285 < Primary 145 - 4427 2141 Household head has at least a secondary school education versus lower than secondar y >/=Secondary 314 - 5307 1802 < Secondary 281 - 4633 2750 Household head has a university education University 160 - 5420 1186 < University 436 - 4826 2686 Household has six or more members >/= 6 284 - 4591 2378 < 6 311 - 5351 2133 Table 12 compares the WTP values between demographic groups. Groups were determined by n atural breaks in the data. For example, gender was separated by male and female members ate breakfast at home versus households where at least one memb er ate breakfast away from home. Comparing low and mid dle weekly per capita food ex penditure households, shows 29 that there is not a statistically significant difference in mean WTP between the two for dried thièré or for buying thièré in bulk. Comparing mid and high incomes is there only grouping that shows a statistically significant dif ferent between preferences for fresh millet. Middle income household dislike dried thièré less than middle income households. This is understandable because these household s are more likely to be prefer buying first stage processed millet and prepare thièr é at home instead of purchasing it from a processor . Also, respondents who purchased a millet - based product on the day of the interview are willing to pay a statistically significant premium for bulk thièré. This is most likely due to the fact that a major ity of the people who purchased their millet product today were sampled at the street vendor location Street vendors often sell their products unpackaged meaning that these consumers are accustomed to purchasing their products in bulk. The marginal willing ness to pay value is also positive and statistically significant for all respondents who had purchased a millet - based product within three months of the interview. These results suggest that consumers prefer fresh thièré and they prefer bulk thièré. The implications of these findings are already observed in the market. Consumers overwhelming ly prefer unpackaged and fresh thièré . Supermarkets understand that consumers prefer to pu rchase only the quantity needed as indicated by the rise of bulk bins in supe rmarkets and selling fresh (frozen and packaged) thièré in Dakar. 30 4. CONCLUSION In this study, I use two choice experiments to analyze consumer preferences for various grains and a second - stage processed millet product, thièré , in Dakar, Senegal . My res ults show three main points. First, consumers in Dakar are willing to pay a premium for domestically produced g rain, in the order of; domestic rice, millet, maize, sorghum. Second, consumers are not willing to pay for imported millet, maize, or sorghum. Th ey are however willing to pay for imported rice. Third, consumers are willing to pay a premium for fresh and unpackaged thièré. First, customers are willing to pay a premium for domestically produced cereals. A positive coefficient for domestic products does support previous literature suggesting that consumers make tradeoffs among traditional and non - traditional grains (Kennedy & Reardon 1994). Additionally, poorer households have a positive and statistically significant preference for millet th an other households. Second, consumers are only willing to pay a statistically significant premium for imported rice compared to imported millet, maize, or sorghum. Moreover, consumers are heterogeneous in their willingness to pay for various grains depending on their income level. Poorer households are willing to pay statistically significantly less for imported rice than richer households. This adds to the existing literature on rice consumption by providing WTP values by income level. Third, consumers are willi ng to pay a premium for fresh and bulk second - stage millet, thièré . Conventional wisdom suggests that consumers demand packaged food since western consumers associate packaged foods with convenience . H owever, I find that fresh and bulk attributes may repre sent another form of conv enience for a household in Dakar. First, households find it convenient to buy in bulk because it allows them to purchase exactly the quantity they demand. Additionally , packaged millet products ( for both fresh and dried) are a rela tively recent 31 phenomenon. Therefore, the sample might be more accustomed to purchasing in bulk and thus report that they prefer it. Finally, households may prefer buying fresh because it frees up time cooking the second stage millet. These findings culmin ate to suggest that consumption habits are complex. Consumers are willing to pay a statistically significant premium for domestically produced grains , even though the preferences vary across grain types . Demand for traditional grains is being over taken by non - traditional grains, like rice. Understanding the nuances and complexities of the food system is critical to supporting the future transformation of food systems in growing cities, like Dakar. The f u ture of work on food systems and the rise of processed foods in developing countries consider two main points, hypothetical b ias and food system monitoring. First, the hypothetical bias surrounding discrete choice experiments in developing countries This p aper shows that hypothetical bias in discrete choice modelling can report values that are simply too high. I found extremely high values of willingness to pay for processed products. Future work in the discrete choice modeling literature will need to addre ss and cont rol for these high value. I suggest finding geographically and culturally specific ways to avoid hypothetical biases . Second, continued reporting on trends in rapidly changing food systems, like Dakar, are critical to keeping a pulse on demand f or processed foods. A s shock waves of rising incomes and access to technology continue , researchers should investigate behavior changes surrounding the food system and their outcomes. These behavior changes increasingly include packaged foods, food delivery systems, and online marketing of food products. Only future research will tell what outcomes are to be expected. 32 APPENDIX 33 Questionnaire: Consumption of Millet Based Product January 30 2018 You are being asked to participate in a research study of urban millet consumption patterns. You will first be asked general questions regarding your household. The next section is a recall of your consumption patterns from the day before. You will then be asked to make hypothetical choices about food consumption options. You will not have to give your name or the name of family members. You will not have to divulge any personal financial information. You must be at least 18 years old to participate in this research. Participation in this research questionnaire is completely voluntary. You have the right to say no, or withdraw at any time. You will not be compensated for this research project. This project will take approximately 20 minutes to complete. I f you have concerns or questions about this st udy, such as scientific issues, or to report an injury, please contact the researcher s ( Sarah Chase - Walsh +221 782938251 or Anta Ngom +221 772024672). If you have questions or concerns about your role and right s as a research participant, would like to obtain information or offer input, or would like to register a complaint about this study, you may contact, +1 - 517 - 355 - 2180, Fax +1 - 517 - 432 - 4503, or e - mail irb@msu.edu or mail to 4000 Collins Rd, Suite 136, Lansing, MI 48910. You indicate your voluntary agreement to participate by completing and returnin g this survey by clicking on the button below, you indicate your voluntary agreement to participate in this online survey. Instructions The respondents will be infor med that this survey is anonymous and no individual information will be taken. The survey is composed of four parts. Part A consists of general questions. Part B is a series of questions concerning the purchase of food and the current behavior of purchases of processed millet products. Part C consists of questions to evaluate the hypothetical c hoices of participants among millet - based products. Part D deals with recent purchases and preferences. Part A General Questions 1. Neighborhood _________ Town_______________ Date___________ 2. Type of vendor or enterprise? 1=Corner store, 2=Woman selling on corner, 3=Supermarket, 4=Fastfood shop 5=Wetmarket 3. Did you buy a millet based product today? 1=Yes 2= No 4 Have you purchased a millet based product in the past three months? 1=Yes 2=No 5. Are you responsible for food purchases in your household? 1=Yes 2=No 6. What is your position in the household? 1= Head of household 2.Spouse 3 Child 4 House - keeper 7. How many members are in the household Children Adults 34 8. What is the highest level of education received by the head of the household ? 9. Respondent Gender 1=Male 2=Female Part B Breakfast 1. What did your household eat for breakfast yesterday? CODES 1= Rice, 2=Bread, 3=Millet 4= Sorghum 2. Did your family eat together yest erday ? CODES 1= Yes 2= No 3. How many children ate breakfast with the adults ? 4. What breakfast prepared in the house or purchased outside of the home? CODES 1= Prepared at home, 2= purchased outside of the home 5. If you purchased a millet based Made at home? Purchased ready to eat Purchased as an intermediate product Freshl y packa ged Fresh and unpack aged Packa ged and dried Unpack aged and dried Purchase location 1=Corner store, 2=Street vendor, 3=Superma rket, 4=Fast food , 5=Market Freshl y packa ged Fresh and unpack aged Packa ged and dried Unpack aged and dried Purchase location 1=Corner store, 2=Street vendor, 3=Superm arket, 4=Fast food , 5=Market Unit Price Unit Price Unit Price Unit Price Place Type of enterp rise1= Small 2= Me dium 3= Larg e Unit Price Unit Price Unit Price Unit Price Place Type of enter prise Thia kry Thiè ré Nga lakh 35 6.Did anyone in the household eat anything else for breakfast inside the home ? Who Which product? (1=Thiakry 2= Thièré 3=Ngalakh 4=Fonde 5=Lakh Sankel) Where was it purchased (1=prepared in the house, 2= Purchased at corner store , 3= Purchase from street vendor , 4= Purchased at supermarket 5= Market) Husband Spouse Children Other 7. Did anyone in the household eat anything else for breakfast outside the home Who Which product (1=Thiakry 2= Thiere 3=Ngalakh 4=Fondé 5=Lakh Sankel) Where was it purchased (1=prepared in the house, 2= Purchased at corner store , 3= Purchase from street vend or , 4= Purchased at supermarket 5= Market Husband S pouse Children Other Fon dé Lak h 36 Lunch 1. What did your household eat for lunch yesterday? CODES 1= Rice, 2=Bread, 3=Millet 4= Sorghum 2. Did your family eat together yest er day? CODES 1= Yes 2= No 3. How many children ate lunch with the adults ? 4. What dinner prepared in the house or purchased outside of the home? CODES 1= Prepared at home, 2= purchased outside of the home chart) Ma de at ho me ? Purchased ready to eat? Purchased as an intermediate product Freshl y packa ged Fresh and unpack aged Packa ged and dried Unpac kaged and dried Purchase location1 =Corner store, 2=Street vendor, 3=Super market, 4=Fastfoo d 5=Market Freshl y packa ged Fresh and unpack aged Packa ged and dried Unpac kaged and dried Purchase location1 =Corner store, 2=Street vendor, 3=Super market, 4=Fastfoo d 5=Market U ni t Pr ic e Un it Pri ce U ni t Pr ic e Un it Pri ce Plac e Type of ente rpris e U ni t Pr ic e Un it Pri ce U ni t Pr ic e Un it Pri ce Pla ce Type of ente rpris e Thi akr y Thi èré Nga lak h Fon dé Lak h 37 6.Did anyone in the household eat anything else for lunch inside the home ? Who Which product? (1=Thiakry 2= Thièré 3=Ngalakh 4=Fonde 5=Lakh Sankel) Where was it purchased (1=prepared in the house, 2= Purchased at corner store , 3= Purchase from street vendor , 4= Purchased at supermarket 5= Market) Husband Spouse Children Other 7. Did anyone in the household eat anything else for lunch outside the home Who Which product (1=Thiakry 2= Thiere 3=Ngalakh 4=Fondé 5=Lakh Sankel) Where was it purchased (1=prepared in the house, 2= Purchased at corner store , 3= Purchase from street vendor , 4= Purchased at supermarket 5= Market Husband S pouse Children Other 38 Dinner 1. What did your household eat for dinner yesterday? CODES 1= Rice, 2=Bread, 3=Millet 4= Sorghum 2. Did your family eat together yest er day? CODES 1= Yes 2= No 3. How many children ate dinner with the adults ? 4. What dinner prepared in the house or purchased outside of the home? CODES 1= Prepared at home, 2= purchased outside of the home Made at home ? Purchased ready to eat? Purchased as an intermediate product Freshl y packa ged Fresh and unpack aged Packa ged and dried Unpack aged and dried Purchase location 1=Corner store, 2=Street vendor, 3=Superm arket, 4=Fast food , 5=Market Freshl y packa ged Fresh and unpack aged Packa ged and dried Unpack aged and dried Purchase location 1=Corner store, 2=Street vendor, 3=Superm arket, 4=Fast food , 5=Market Unit Price Unit Price Unit Price Unit Price Place Type of enter prise Unit Price Unit Price Unit Price Unit Price Place Type of enter prise Thia kry Thiè ré Ngal akh Fon dé Lak h 6.Did anyone in the household eat anything else for lunch inside the home ? Who Which product? (1=Thiakry 2= Thièré 3=Ngalakh 4=Fonde 5=Lakh Sankel) Where was it purchased (1=prepared in the house, 2= Purchased at corner store , 3= Purchase from street vendor , 39 4= Purchased at supermarket 5= Market) Husband Spouse Children Other 7. Did anyone in the household eat anything else for lunch outside the home Qui Which product (1=Thiakry 2= Thiere 3=Ngalakh 4=Fondé 5=Lakh Sankel) Where was it purchased (1=prepared in the house, 2= Purchased at corner store , 3= Purchase from street vendor , 4= Purchased at supermarket 5= Market Husband S pouse Children Other 40 Choice Experiment Section Choice Experiment 1. Imports vs. Domestic products 1. Which product would you buy? Millet Maize Rice Sorghum No Buy Senegal: (400) Imported: (300) Imported (500) Imported (400) 2. Which product would you buy? Millet Maize Rice Sorghum No Buy Imported: (500) Imported: (250) Imported (600) Senegal (300) 3. Which product would you buy? Millet Maize Rice Sorghum No Buy Imported: (400) Senegal: (500) Imported (500) Senegal (500) 4. Which product would you buy? Millet Maize Rice Sorghum No Buy Senegal: (400) Senegal: (300) Senegal (300) Imported (300) 5. Which product would you buy? Millet Maize Rice Sorghum No Buy Senegal (400) Imported: (500) Imported (500) Senegal (400) 6. Which product would you buy? Millet Maize Rice Sorghum No Buy Senegal (300) Imported: (309) Senegal - (500) Imported (500) Choice Experiment 2 Imports vs. Domestic products 1. Which product would you buy? Thièré 1 Thièré 2 No Buy Fresh Packaged (500) Dried unpackaged (300) 2. Which product would you buy? Thièré 1 Thièré 2 No Buy Fresh Packaged (350) Fresh unpackaged (350) 41 3. Which product would you buy? Thièré 1 Thièré 2 No Buy Fresh unpackaged (500) Dried packaged (300) 4. Which product would you buy? Thièré 1 Thièré 2 No Buy Dried Packaged (300) Dried unpackaged (300) 5. Which product would you buy? Thièré 1 Thièré 2 No Buy Dried unpackaged (500) Dried unpackaged (300) 6. Which product would you buy? Thièré 1 Thièré 2 No Buy Fresh Packaged (400) Dried unpackaged (300) 42 Part D. Recent purchase history and shopping preferences 1. Roughly, what is your weekly budget for food purchases? (In FCFA ) ? 2. When was the last time you purchased ? 3. Where did ? 1=Corner store, 2=Street vendor, 3=Supermarket, 4=Fast food 5=Market 4. Why did you choose this vendor ? 1 = Best price 2 = Best quality 3 = Food safety 4 = For convenience Thiakry Thièré Ngalakh Fondé Lakh 43 REFERENCES 44 REFERENCES Biggest Cash Crop. Washington, DC: International Food Policy Research Institute (IFPRI). Aprile International Journal of Consumer Studies, 36(2), 158 - 165 Bech, M., and D. Gyrd - Hansen . 2005. Effects coding Health Economics 14 (10), 1079 1083. Becker, G. (1965). A Theory of the Allocation of Time The Economic Journal, 75 (299), 493 - 517. Bello, M., and A., Awudu. 2016 . Impact of ex - ante hypothetical bias mitigation methods on attribute non - attendance in choice experiments Blackwell Publishing. Bhat, C. R. 2003 . Simulation estimation of mixed discrete choice models using randomized and scrambled Halton sequences. Transportation Research , 37(9), 837 855. RAD. Bricas N . and C. Broutin . 2008. Food processing and retail micro - activities and poverty reduction in sub - Saharan Africa : Trade as a development tool: partnerships and policies: 1st Conference of the Geneva Trade and Development Forum. GTDF . s.l.: s.n. , 17 p. Conference of the Geneva Trade and Development Forum. 1, Crans Montana, Suisse, 17 September 2008/20 September 2008. Vacuum: The Effects of Extern Journal of Economic Behavior, 145, 335 - 351. Chryssochoidis G ., Krystallis , A. , Perreas, Ethnocentric beliefs and country of origin (COO) effect: Impact of country, product and product attributes on Greek consume rs' European Journal of Marketing 41:1518 - 1544. Development Policy Review. 30 (4), 451 - 47 2. 45 302 - 309 Forbes - Brown, S., Micheels , E., and J., Hobbs . 2016. Consumer Willing ness to Pay for Dairy Products w ith the 100% Canadian Milk Labe Journal of International Food & Agribusiness Marketing , 28:3, 203 - 224. Kennedy, E. and T. Reardon, 1994. Shift to non - traditional grains in the diets of East and West Africa: role of w cost of time. Food Policy, 19(1) 45 - 56. Kikulwe, E., J . Wesseler and J. Falck - Introducing a Genetically Modified Banana in Uganda: Social Benefits, Costs, and Consumer Perceptions , (IFPRI Discussion Paper 767), Washington, DC: International Food Policy Research Institute (IFPRI). Louviere, J.J., Hensher, D.A., and J.D., Swait. 2000. Stated Choice Methods: Analysis and Application . Cambridge, Cambridge University Press Lusk, J. L., & T. C., Schroeder. 2004 . Are choice experiments incentive compatible? A test with quality differentiated beef steaks. American Journal of Agricultural Economics , 86(2), 467 482. McFadden, D. (1974). Conditional logit analysis of qualitative choice beha vior. In P. Zarembka (Ed). , Frontiers in econometrics . pp 105 142). New York: Academic Press. Minten, B., Seyoum Taffesse, E., Alemayehu S., and B , Petra, ed. 2018. The economics of teff: l Food Policy Research Institute (IFPRI). https://doi.org/10.2499/9780896292833 World Development . 38 (12) 1775 - 1787. Food Policy. 36, 318 - 324. Reardon T. , Tim m er The Economics of the Food System Revolution Annu al Review of Resource Economics 2012 4 : 1 , 225 - 264 Reardon, T., Thiombiano, T., and C., Delgado. traditionnelles dans la Connsommation Économie Rural . 190, 9 - 14 Revelt, D., and Train, K. E. 1998 . appliance efficiency level. Review of Economics and Statistics , 80(4), 647 657. Senauer, B, Sahn, D and The effect of the value of time on food consumption patterns in developing cou American 46 Journal of Agricultural Economics, 68, 4 (November). Street, D. J., and L. Burgess. 2007. The Constru ction of Optimal Stated Choice Experiments. Hoboken, N.J.: Wiley. annual meeting, Columbus OH, 10 - 13 Aug. 1975. Train, K. 2009 , Discrete Choice Methods with Simulation , C ambridge: Cambridge University Press. United Nations. 2017. Department of Economic and Social Affairs, Population Division Household Size and Composition Around the World 2017 Data Booklet (ST/ESA/ SER.A/405). - home Prepared Foods: Contribution to Energy and Nutrient Intake of Consumers Living in Low - Income Areas Public Health Nutrition. 5(4), 515 - 522. Van Loo, E.J., Caputo, V., Nayga, R.M., Meullenet, J. - F., S.C., Ricke. 2011. willingness to pay for organic chicken breast: evidence from choice experiment. Food Qual. Prefer . 22 (7), 603 613. Van Loo, E.J., Caputo, V., Nayga, R.M. , and W., V Food Policy, 49 (1), 137 - 150. W orld Bank. 2018 . The World Bank Development Indicators [Database]. Washington DC.