Understanding Consumer Preference for Common Beans from Manifold Viewpoints of Attributes in Dar es Salaam, Tanzania

Common beans appeal to consumers in different ways. One important dis-tinction of this nature is with respect to colour, size, cooking time and gravy quality. When multiple common bean varieties are exposed, consumers normally select several varieties on the same occasion while rejecting some of the offerings. Studies that have explicitly assessed factors underlying such a decision making are confined to demographic and socio-economic factors while ignoring societal and cultural factors. Ignoring these factors distort the measured effects and contribute to the failure of interventions aimed at altering food preferences. This study investigated the factors incorporative for a better understanding of consumers’ preferences for common beans. Discrete Choice Experiment was employed in order to elicit individual preference and uncover how individuals selected common beans with varying attribute levels using a random sample of 732 respondents. Using Poisson Regression Model, the results showed that probability of choosing two common bean types was the highest, although for some consumers, the number of choices ranged from zero to eight. Highlighted findings are essential for breeders, farmers and sellers of common beans to become certain on their decisions. The study re-commends that breeding and market development efforts should primarily focus on unique preferences of consumers whose choices are predominantly within a narrow range of common beans and hence meeting their varied demand.


Introduction
Common bean (Phaseolus vulgaris L.) is an important component of the traditional cropping system in developing countries, especially in sub-Saharan African (SSA) (Nassary et al., 2020). More than 101 million smallholder farmers in SSA grow at least one tropical legume (Abate, 2012). The global production of all food legumes has increased at the rate of more than 1% per annum since 1980 (Nedumaran et al., 2013). Between 1994 and 2019, the growth rate of common bean production in SSA was estimated to be about 4% per annum although this growth was mainly attributed to an increase in acreage and not crop productivity. However, advances of science and technology have a great role to play in food production. With respect to consumption, the common bean has been reported to exhibit an upward trend in all regions of the world except in Central and East Asia (Nedumaran et al., 2015). Accordingly, SSA is the second largest consumer of common bean, after Latin America and Caribbean region, although its per capita consumption remained constant in the past three decades (FAOSTAT, 2020).
Common bean is regarded as a subsidiary-crop to be relied upon during food shortage or supplement the diet as it is drought resistant and well adapted to the semi-arid regions of the tropics (Akibode and Maredia, 2012). In Tanzania, beans are regarded as important for food and nutrition security as well as income generation (Nassary et al., 2020;Leterme, 2002;Hella et al., 2013). Tanzania is ranked seventh among top producers of common beans worldwide and production is dominated by smallholder farmers who consume part of the product and sell the surplus (Ronner & Giller, 2013). Farmers with entrepreneurship skills have what it takes to pursue their farming objectives which include producing products that are demanded. However, smallholder farmers in Tanzania have limited entrepreneurial skills and poor access to information about market requirements (Mishili et al., 2011). This information asymmetry limits these farmers from fully benefitting from the increased production because they cannot bargain effectively when they market their produce and locate lucrative outlets for their produce.
For many years, researchers have identified different technologies such as improved seeds for increasing productivity and production of tropical legumes including common bean (Abate & Orr, 2012;Ronner & Giller, 2013;Nedumaran et al., 2015). These supply side interventions have paid little attention to market needs, although it is known that attributes embedded in the various varieties of legumes are important in shaping consumer preferences 1 . In essence, breeders have focused almost entirely on producer needs without appreciating the needs of consumers. Abate & Orr (2012) reveal that absence of direct link between farmers and buyers create demand mismatch which impacts domestic demand in urban markets. Therefore, providing demand-side information is essential for 1 Consumer preference for common bean. E. Swema, L. Mwinuka DOI: 10.4236/tel.2021.115066 1040 Theoretical Economics Letters farmers and breeders to tailor their production and breeding plans to meet consumers' needs, respectively.
The bean market in Tanzania offers various types of beans that differ by colour, shape, size, cooking time and digestibility (Mishili et al., 2011). These characteristics have been reported to affect consumers' preferences for common beans. Katungi et al. (2011) reported that wealthier households in Kenya preferred low flatulence and large grain size of common beans while poor households were indifferent to size. A study conducted in Tanzania showed that low income earners in Dar es Salaam preferred common bean variety locally known as Soya-kablanketi because of its short cooking time compared to other varieties (Mishili et al., 2011). However, Mundua (2010) found negative relationship between income and willingness to pay for different bean varieties in Uganda, where large grain size and white colour were the preferred bean attributes. A similar study in Kenya reported that women were more conversant with bean attributes than men and made decision about the type of beans to be eaten at home (Gitonga, 2015).
When multiple common bean varieties with varying attribute levels are exposed, consumers normally select multiple varieties on the same occasion while rejecting some of the offerings. Studies that have explicitly assessed factors underlying such a decision making have generally been rare and confined to demographic and socio-economic factors only. Riet et al. (2011) revealed that knowledge from psychology, dietetic and nutritional disciplines is equally important in shaping consumers' food preference. Chadwick et al. (2013) argue that eating habits are partly a reflection of cultural and societal factors that are reinforced through social interactions, community taboos, tastes and preferences. Glanz et al. (1998)

Theoretical Framework
The study employed the demand theory which asserts that goods as such, are not the immediate objects of preference or utility or welfare, but have associated characteristics which are directly relevant to consumers (Lancaster, 1966). If the consumer is required to make independent choices among various common bean types, the choice to be made reflects unobserved utility. This utility is derived from the attributes' embedded in the chosen common bean type which varies across individuals depending on their demographic, socio-economic, cultural and societal characteristics.
The utility function is specified as, where, ij U is the utility of the i th individual derived from j th type of common beans, j Att is the j th specific attributes embedded in, i S represents individual specific characteristics and i ε is the stochastic error for the i th individual.
However, the random utility model (Equation (1)) is appropriate when individuals are able to rank the choices based on perceived satisfaction derived from each type of common bean (Louviere et al., 2000). If an individual is not required to rank the possible choices, the random utility model is replaced by the latent model specified in Equation (2). The attributes component in Equation (1) is left out in Equation (2) because individuals are exposed to specific type of common bean at a time.
( ) where, * i y reflects the perceived net benefits from a given common bean type, If the perceived net expected benefits are greater than zero, the specific common bean type is selected and if the perceived net expected benefits are less or equal to zero the specific common bean type is rejected.

Data Collection
Data were collected from Dar es Salaam city to represent beans consumers in Tanzania. The selection of this city was purposeful because it is the major market for common beans in the country and business centre (market share of about 26%). Moreover, Dar es Salaam has diverse population to account for variation The second stage entailed a proportionate sampling of 100 EAs from the three strata. The third stage involved a random selection of eight households from each of the 100 EAs resulting into a sample of 732 households. This sample is less than the desired one of 800 due to incompleteness and poor quality of information that were solicited from 68 targeted household heads. A household survey was conducted in all three districts of Dar es Salaam city, Tanzania namely Kinodoni, Temeke and Ilala. A household survey was accompanied with a Discrete Choice Experiment to gather information on a household and respondent specific factor that could influence consumers' preferences for common beans' types. The choice task was first comprehensively introduced to the respondents (household heads) to make sure the task of choosing common beans with specific attributes were properly understood. A pilot test before the actual choice experiment confirmed the respondent understanding of the choice task.

Discrete Choice Experiment
The discrete choice experiment (DCE) was employed in order to elicit individual preference and uncover how household heads selected common bean types by asking them to state their choice over different hypothetical attributes combinations. Four different common beans' attributes were identified to include colour, grain size, cooking time and gravy quality. The colour attribute involved four colours namely pure grey locally known as soya-supa, grey locally known as soya-kawaida, yellow locally known as soya-njano and mottled-red. The grain size attribute involved three classes of size namely small, medium and large. The cooking time attribute involved common beans that cook fast and those that cook slowly. Finally, the gravy quality attribute was specified in two levels namely poor and good gravy quality. Among the four attributes, common bean color was considered as the very first factor attribute underlying consumers' choices followed by other attributes of beans such grain size, cooking time and gravy quality. The common bean colors involved in this experiment are presented in Plate 1.
The experiment generally intended to capture trade-offs among the attributes in the common bean types. Therefore, specific common bean with specific color was physically presented to household heads and traded off with other attributes such as grain size, cooking time and gravy levels to produce the preference. Following the given four attributes with their levels, its combination (4 × 3 × 2 × 2) gives the rise of 48 common bean types. From the 48 common bean types, 6 unique choice sets were created each containing 8 different types of common bean. The total sample of respondents was divided into 6 sub-samples and each was subjected to only one of the 6 choice sets.

Data Analysis
The data was analyzed by a combination of descriptive statistics and Poisson re- Because the Poisson mean on the left-hand side represents an expected count and must be non-negative, a simple linear model shown in Equation (5) was modelled by applying logarithm in both side to get a linear model as: Since for poison distribution the mean equals to the variance, this gives rise to Equation (7).
The i y′ are independently distributed as Poisson random variables with mean i λ for each individual. So, the Poisson regression model may be written as: where i λ is the response variable indicating the expected number of common bean type chosen by i th consumer, i X is a vector of independent variables influencing the number of common bean varieties chosen by the i th consumer and β represents a vector of parameters estimated.

Variable Estimations and Diagnostic Tests
Marginal effects provide a way of measuring the effect of each explanatory variable on the dependent variable. The marginal effect of one explanatory variable is the expected instantaneous rate of change in the dependent variable as a function of the change in that covariate while keeping other covariates constant for PRM from Equation (6).
Therefore, the marginal effect of the change in regressor , Because the derivative in Equation (9) is with respect to a small change, it is not appropriate to apply for the effect of a change in a dummy variable, or change of state. The appropriate marginal effect for the dummy variable, say, d, would be; Simple comparisons based procedure for testing the hypothesis (Equation (10)) is carried out by comparing the values for mean and variance of the expected number of choices. If the expected variance happens to be greater than the expected mean, over dispersion in the model is confirmed. If it happens that the expected variance is equal to expected mean or the expected mean are greater than expected variance with small variations (less than 0.5) the model is confirmed not have over dispersion.

Most and Least Chosen Common Bean Types
On the basis of the available data and design of the choice experiment a list of most and least chosen common bean types are presented in Table 1 and Table 2 respectively. For example, according to Table 1, common bean type with grey colour, large grain size, fast cooking and whose gravy quality is good was chosen by majority of consumers in the first choice set. The rest of most and least chosen common bean types with specific attributes and specific choice set are as presented in Table 1 and Table 2.
The common bean attributes such as yellow colour, large size and good gravy quality appeared to dominate the list of most chosen common bean types by consumers in Dar es Salaam. Evidence from several studies concurs with this observation (Mishili et al., 2011;Katungi et al., 2011;Gitonga, 2015). *Sets of common bean types exposed to consumers. Each set comprised of 8 common bean types. The common bean types that shown are those selected by majority of consumers in each choice set.  *Sets of common bean types exposed to consumers. Each set comprised of 8 common bean types. The common bean types that shown are those selected by least number of consumers in each choice set.
beans with yellow colour are preferred because they are associated with lowest cooking time (about 75 minutes) compared to beans with other colours (Mishili et al., 2011).
Similarly, the majority of low income consumers prefer fast cooking beans to save on cooking fuel (Katungi et al., 2011). Therefore, the results potentially explain why the majority of consumers will choose common beans with yellow colour which are associated with fast cooking attribute. In comparison with other colours involved in the choice experiment, yellow colour was dominant in the list of most chosen common bean types as indicated in Figure 1.
Literature suggests that large sized grains expand more when cooked than the small sized grains of common beans (Gitonga, 2015). Saba et al. (2015) studied on swelling capacity of common bean, they found that common bean with large size expand/swell more when cooked than those with small size. Therefore, it makes sense that fewer amount of common bean with large grain size would be required to make a meal compared to when the common bean is of small grain size. This is however the reason why common bean with large grain size are chosen most by consumers compared to common bean with small grain size. In terms of gravy quality common bean with good gravy quality were observed to dominate the list of most preferred common beans' types. Several studies have identified good gravy to influence food preference (Wahl et al., 2017;Cohen and Babey, 2012;Chadwick et al., 2013). Table 2 shows that common bean attributes such as pure grey colour, small size and poor gravy quality observed to be chosen by least number of consumers. According to Ronner and Giller (2013) common beans with this character (pure grey colour) are associated with higher prices compared to beans with other colours. These type of beans are mainly

Range of Choice(s) of Common Bean Types
In order to explore the current and future demand of different common bean The number of choices of common bean types were ranked corresponding to probability of choice as presented in Table 3. A ranking results revealed that majority of consumers confined in choosing one type of common bean to four types of common bean. The probability of consumers to stick on choice of only one type observed to be 20%, meaning they are not flexible in their choice. Also, the sum of probabilities of choosing more than one type of common bean observed to be 79%. This is to mean that majority of consumers who were involved   in this study are flexible in their choices with high proximity of trading off between two to four different types of common bean. Factors affecting this type of behaviour are discussed in the next section.   Louviere et al. (2000) the pseudo R-squared should be greater than 0.1 to have a stable or meaningful model whereas a value between 0.2 and 0.4 is considered as an extremely good fit. Poisson regression model is rejected when there is over-dispersion in the dependent variable (i.e. variance>mean). Table 5 indicates non-presence of over-dispersion since mean of the dependent variable is slightly greater than variance. According to Cameron & Trivedi (1986), if the mean is greater than variance with a small variation (<0.5), the means are said to be roughly equal to    (Parraga, 1990;Leterme, 2002). More education is also associated with curiosity and ability to search and try new food products with different but desired attributes (Vartanian et al., 2008;Riet et al., 2011).  With respect to societal/cultural variables, household head domestic place of origin and food purchase decisions observed to significantly influence the number of choices made at 5% level. For the household domestic place of origin, the mean number of common bean type chosen by consumers originated from major bean producing areas is 20% higher than those originated from least or non-bean producing areas. This is to mean that, consumers originated from major bean producing regions are more flexible in their choice of common bean if compared to those originated from least or none bean producing areas. Moreover, joint decision making when purchasing the beans increased the number of choice of common bean types. Based on Gillespie & Johnson-Askew (2009), in a household attributes such as personal food knowledge, skills and other human resources contributes to the food and eating alternatives available at home.

Conclusion and Implications
Findings from this study seek to inform stakeholders and actors with respect to common bean types that are most and least chosen by consumers in the market.
The information is essential for breeders, farmers and sellers of common beans to become certain about consumers' desire and what to breed, produce and sell, respectively. Relevant institutions should therefore ensure that there is a match E. Swema, L. Mwinuka DOI: 10.4236/tel.2021.115066 1052 Theoretical Economics Letters between produced common bean types and consumer preference. This can create a preference oriented market which is beneficial for the respective chain actors in a sub sector.
Also, consumers' flexibility in choices of different common bean types reflects how consumers through household heads are diverse in their choices. Thus, considerable efforts to breed, produce and sell common beans with specific attributes can be the best way to fulfil consumers' desires. Meeting the varied demand of consumers, means increasing the range of choices that appeal to consumers. Moreover, the significant influence of the cultural and societal factors entails how they are important in shaping consumer preferences in choice of common beans.
Although individual domestic place of origin, consumers eating habits and household food purchase decisions were the only cultural and societal factors studied, future studies in this area may include other cultural/societal factor like belief, value, religion and norms by finding appropriate proxies to study consumer preferences on bean sub sector. Also, the common bean attributes considered in this study were colour, grain size, cooking time and gravy quality.
However, it is likely that there are other attributes of common bean that are important to consumers beyond what was considered in this study. Future studies are advised to reshape the attribute levels that considered in this study or include other common bean attributes like flatulence, keeping quality and beans swelling capacity and see if they are pleasant to consumers at household level.