What are the key factors influencing consumers’ preference and willingness to pay for meat products in Eastern DRC?

Abstract Dietary patterns for consumers among the elite and middle‐income classes in developing countries are shifting rapidly toward the consumption of more animal‐based products. Although this shift presents opportunities, there are significant market failures affecting their preferences and willingness to pay (WTP). This study used a multistage sample survey of 309 consumers from three different communities of Bukavu, Eastern DRC, to examine the effect of socioeconomic/socio‐demographic characteristics and quality attributes on consumers’ purchasing decisions and WTP for meat products. The results suggested that about 53% of the respondents were dissatisfied with meat products in the market due to their high price, low quantity, unhealthiness, and harmful effects. Older female respondents living in urban areas were more likely to purchase meat products. Their WTP was significantly determined by attributes such as color, in‐mouth texture, and availability. Nutrition, harmful effects, and availability of meat products are the important factors that influence purchasing decisions among higher income groups. Addressing these market failures could have an impact on the meat market, improving the nutrition of low‐income consumers and ensuring food safety standards in DRC and other developing countries with similar challenges.

The projected increase in meat consumption is a sign of a better future with regard to malnutrition levels among the poor in lower-income countries who suffer from micronutrient deficiencies and mainly depend on high fiber and phytate plant-based staples (Neumann et al., 2003). The impact of malnutrition is globally estimated to be as high as US$3.5 trillion per year or US$500 per individual (FAO, 2013). The costs are opportunity costs of economic growth foregone and lost investments in human capital resulting from infections, impaired child development, and mortality (Hoddinott, 2016). In the Democratic Republic of Congo (DRC), over 3.6 million children under five are affected by acute malnutrition annually and 2 million of them suffer from its most severe form (OCHA, 2016). This country is estimated to be losing more than a billion dollars a year to the effects of child undernutrition, which is equivalent 4.5 percent of GDP. Therefore, consumption of meat products could be one of the keys to reducing malnutrition costs in the DRC. However, as argued by Randolph et al. (2007), the negative publicity on livestock and their products is driven by health and food safety concerns related to outbreaks of diseases like avian influenza and the continued debates on the association between the saturated fats and cholesterol found in animal food sources and chronic diseases like heart disease and cancer, contributing to consumer nervousness about meat products. Consumer nervousness affects their WTP, purchase, and consumption of meat products, thus exacerbating the malnutrition level and related costs in developing countries. Nevertheless, consumers' choices are influenced by many factors that ultimately shape purchasing decisions. Font-i-Furnols and Guerrero (2014) identified consumers' behavior as depending on interrelated factors that included psychological influences (willingness, risk, expectations, sociocultural factors, lifestyle, and values), sensory qualities (visual appearance, texture, flavor, and odor), and marketing factors (price, label, brand, and availability). In addition, Grunert, Bredahl, and Brunsø (2004) used the Total Food Quality model to analyze consumers' perception and decision-making in determining meat quality. The model showed that consumers form expectations about quality at the point of purchase, based on their own experience and informational cues available in the shopping environment.
These preferences are influenced not only by quality and consumer-related factors but also by context, culture, and information (Kanerva, 2013;York & Gossards, 2004). Alemu, Olsen, Vedel, Pambo, and Owino (2017) showed that preferences in Kenya are also influenced by context and information in addition to product attributes.
Van Wezemael, Verbeke, de Barcellos, Scholderer, and Perez-Cueto (2010) also reported that European consumers considered label, brand, freshness, and leanness of beef as cues to indicate quality to purchase, whereas safety in Ghana and hygiene in Rwanda were purchasing attributes in purchasing meat products (Niyonzima et al., 2017;Owusu-Sekyere, Owusu, & Jordaan, 2014). However, most of the studies on consumers' preferences for meat products focus on developed countries (Tonsor et al., 2005;Reicks et al., 2011;Schumacher, Schroeder, & Tonsor, 2012;Zimmerman et al., 2012;Hung, de Kok, & Verbeke, 2016;Shan et al., 2017). Only a few studies focus on the African  1961 1963 1965 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999  context where food quality and malnutrition remain huge challenges (Niyonzima et al., 2017;Owusu-Sekyere et al., 2014). Increasing incomes in developing countries together with the inherent market failures makes it vital to understand the factors driving consumers' meat consumption patterns and their WTP for such food products. Failure to understand the key determinants of consumers' preferences could lead to further market failure and the consumption of unwholesome meat products (Mockshell, Ilukor, & Birner, 2014).
The overall objective of this study is to evaluate the preferences for meat and meat products and WTP among consumers in Eastern DRC.
Specifically, this study aims at: (a) identifying consumer and household characteristics influencing consumer preferences and WTP; (b) examining consumers' preferences for meat products; and (c) analyzing the effect of socio-demographics and product attributes on purchasing decisions and WTP by using linear and ordered multinomial logistic regression models. The rest of the paper is structured as follows: Section 2 presents the structure of the meat market in the DRC, and Section 3 presents the materials and methods. The results are presented in Section 4 and discussed in Section 5; the paper presents the conclusions in Section 6.

| THE ME AT MARK E T IN THE DEMO CR ATI C REPUB LI C OF CONG O
The agricultural sector is an important sector in the economy of the  (FAO, 2005). Livestock populations have suffered significantly since the civil war, when many farms were looted and the animals stolen. As an important source of dietary protein, consumption and sale of wild animals ("bushmeat"), including some primates, are widespread. This has been fueled partly by poor living conditions and the rise in the number of internally displaced people (IDPs) fleeing regional conflicts. As shown in Figure 1, wild meat is the most produced meat product in the DRC followed by pork and beef.
The consumption of meat is higher than the production of meat in DRC, so the country is a net importer of food products Union (EU) countries like the Netherlands that is the largest exporter of pork together with Belgium and Germany. As shown in Figure 2, the main imported meat is chicken followed by pork.
Meat imports have generally declined from the 1980s to date.
The decline in beef imports can be linked to the deliberate effort of government to promote cattle production through the rehabilitation of some cattle farms that were destroyed by the wars, particularly in Katanga Province and North and South Kivu (Goma and Masisi) in the Northeast bordering Rwanda and Burundi as well as the increased consumption and preference of game (Yamaguchi, 2015). Another contributory factor is increased concerns of meat quality especially contamination with salmonella, which is a threat to human health ( and rabbits are generally sold live and slaughtered at home. Highand middle-income households purchase beef or goat meat but lower-income households often choose smaller animals such as pigs, poultry, and rabbits, as their coping strategy . However, small animals are sold in the markets only when household needs arise, and the money raised is mostly invested in school fees (Zozo et al., 2012).  Respondents were asked who was the main breadwinner in their household and who decides which food to purchase. In the third module, respondents were asked about their consumption and purchasing frequencies of meat products including beef, pork, goat, chicken, and rabbit. In the fourth module, respondents were asked how satisfied they were with the meat products and the factors influencing their purchasing decisions and WTP.

| Sampling and survey design
To evaluate the willingness of consumers to pay, the revealed preference method was applied. The method was chosen because data obtained from revealed preference methods more truthfully reflect preferences and choice in the real market when compared to stated preference methods (Howard & Allen, 2008). Respondents were given the average prices based on the different markets for each meat product, and then they were asked to score the influence of product attributes (nutrition, color, texture, smell, harmful effect, price, availability, and quantity) on their perception (no = 0, yes = 1).
In addition, they were asked to rank the importance of these attributes on their purchasing decisions by using a five-point Likert scale (not important/definitely would not pay = 1, least important/ probably would not pay = 2, moderately important/might pay = 3, important/probably would pay = 4, and most important/definitely would pay = 5).

| Data analysis
Data analysis was performed using R software (version 3.2.3, R Core, 2015). Basic statistics (means, standard deviation, and frequencies) were computed to describe the responses. Chi-square (χ 2 ) and analysis of variance (ANOVA) were used to examine the differences in the responses. In order to fit linear regression assumptions for ANOVA, BoxCox power transformations were applied to the continuous variables; the transformed variables were analyzed using ANOVA, and the mean comparisons were done on the back-transformed values (Box & CoX, 1964). Significantly different means were separated using least significant difference (LSD) with the appropriate error terms and a significance level at p < 0.05.
To investigate the factors determining purchasing decisions and WTP among respondents, a logistic regression analysis was performed following a generalized linear regression with probit link.
When Y is the dependent or response variable as Y is dichotomous, the use of probit link, f(Y), leads to the transformation of the response into a continuous variable, Y. The link function then maps the (0, 1) range probabilities onto (−∞, +∞), the range of linear predictors (Agresti, 2002;Fox, 2008). We then have a probit model as: Note. *, **, and *** indicate statistical significance at the 0.1, 0.05, and 0.01 levels, respectively.

TA B L E 4 Preference of respondents on all meat products in the market
The probit link function is given by Faraway (2006) as: where Φ -1 is the inverse normal cumulative distribution function, such as N (0, 1) (Agresti, 2002).

And the regression equation becomes:
The model parameters were estimated using the maximumlikelihood method, with chi-square test of significance (Dodge, 2008). The following vector of independent variables was considered for their socio-demographic effects: These are standard socio-demographic variables such as living area, gender, current age, marital status, education level, and employment status of household head, household size, and income. The effects of product attributes (nutrition, color, texture, harmful effect, price, availability, and quantity) on consumers' purchasing decisions and WTP were determined by performing an ordered multinomial logistic regression model, as the above dependent variables were nominal and polytomous, i.e. had more than two categories with an ordered structure (Engel, 1988;Menard, 2002).
When the following ordered probit model estimated using maximum-likelihood (ML) method is considered, we have with y * n is the unobserved dependent variable, x ′ n is the vector of independent variables, and β is the vector of regression coefficient to estimate. The latent random variable y * n for individuals n = 1,2,3…N, linearly depends on the independent variables x n and ε n is the error term. Therefore, If the errors ε n are logistically distributed, with distribution function Λ( i ) = 1 1+e − i produces an ordered logistic model given by Akshita, Ramyani, Sridevi, and Trishita (2013) as: With regard to household income, the influence of product attributes on purchasing decisions and WTP was explained by the gg plots.

| Consumer and household characteristics
Most of the respondents (86%) lived in urban areas, and the majority were female (56%) with an average age of 37 years ( Table 2). The average household size was 6 persons, and the composition is characterized by 54% of children, 18% of the household head, 15% of the spouse. In this study, 87% had attained at least primary school, with an average of 8 years of formal education. Most of the respondents in Ibanda had completed higher education when compared to those in Kadutu and Bagira. The main occupation of respondents varied among communities. Self-employed business/services (26%) was observed as a main occupation in Ibanda, whereas many respondents in Kadutu (20%) and Bagira (20%) were unemployed.
Household income and expenditure profiles varied substantially (Table 2). Relating this to household size, the average per capita income was about US$1,039 in Ibanda, US$397 in Kadutu, and US$368 in Bagira. The main source of income in Ibanda was permanent employment (48%), whereas petty trading (37%) was reported  Although the main source of fuel for cooking was charcoal (86%), more households (16%) in Ibanda had access to electricity than in other communities. About 98% of the main household water supply was from RIGIDESO, the water supply authority in Bukavu.

| Household consumption of meat products
In terms of frequency of meat consumption, results showed that beef was the most consumed product, with 83% of the household consuming it at least weekly (Table 3). Goat meat and pork were widely consumed too, with between 66% and 71% of the respondents, respectively, consuming these products weekly. The products least consumed were chicken and rabbit since these are less often produced and available. On average, 68% of the respondents consumed milk in a week, followed by sausage (53%), yogurt (48%), and cheese (45%). Households in Ibanda purchased more fresh meat and meat products than those in Kadutu and Bagira. The average daily consumption in Ibanda was 1.9 times higher for beef than in Kadutu and Bagira, 1.5 times higher for goat meat, 1.5 times higher for pork, and 3.5 times higher for chicken. In the study, it was also found that the price of meat products varied by communities. For example, the price of processed products (sausage, milk, yogurt, and cheese) seemed to be higher in Ibanda than in Kadutu and Bagira.

| Preference of meat products
Only 47% of the respondents were satisfied with the meat products in the market (Table 4). When asked about the criteria that caused dissatisfaction, 24% claimed unhealthiness and high price as the main criteria, followed by low quantity (18%) and harmful effect (11%). It could be seen that the dissatisfaction can be divided into two groups. The respondents, especially in Kadutu and Bagira, used high price and low quantity as extrinsic criteria; unhealthiness and harmful effect were mainly perceived as intrinsic attributes by the respondents in Ibanda.

| Social factors influencing purchasing decision
Regarding the association between socio-demographic and socioeconomic factors on purchasing decisions and WTP for meat products, it was observed that living area and gender have a positive significant effect on purchasing decisions but a negative significant effect on WTP (Table 5). Results of the logit model also indicate a negative correlation between age and purchasing decisions; a positive correlation was observed between age and WTP.
Although other variables were not found to affect purchasing decisions and WTP significantly, when the education level of the household head changes from low to high, the estimated coefficients of purchasing decisions increase by 2.3 times and of WTP by 2.9 times. Marital status and intrahousehold sharing of information were not found to affect purchasing decisions and WTP. Similarly, the employment status of a household head, household size, and the presence of children did not have a significant influence. Surprisingly, household annual income did not play a significant role.

| Product attributes influencing purchasing decisions
The results showed that although the respondents were dissatisfied about unhealthiness, harmful effect, high price, and low quantity, TA B L E 6 Ordered probit regression for product attributes determining consumer purchasing decision and willingness to pay for meat products these attributes did not exhibit a significant influence on their purchasing decisions and WTP for meat products but color, in-mouth texture, and availability were identified as significant attributes. The respondents selected availability as the only significant attribute for WTP (Table 6).

| D ISCUSS I ON
Consumers' preferences, behavior, and perception of meat and meat products depend on many factors, sensory (product-specific factor), psychological (individual factor), and marketing (environmental factor). These aspects might be altered owing to individual behavior, context, culture, available information (Font-i-Furnols & Guerrero, 2014), concerns, lifestyles, and socio-demographic characteristics (Bernués, Olaizola, & Corcoran, 2003;Grunert et al., 2004). Among sociodemographic variables, our findings demonstrated that, as expected, living area and gender had a positive significant effect on purchasing decisions but a negative significant effect for WTP. The positive significant effect of living area on purchasing decisions and WTP for meat products indicated that people living in rural areas make a decision to purchase meat products differently from those living in urban areas. While a higher rate of WTP among respondents for meat products was found in urban areas, price alone cannot be used to infer the actual WTP of respondents because they were aware of the artificial purchase situation. Consumers often claim that they would pay higher prices for certain product attributes than they actually do in real purchase situations (Feldmann & Hamm, 2015). For gender effect, Croson and Gneezy (2009) stated that men and women apparently vary in their emotional response to uncertain situations and this difference results in dissimilarities in risk taking. In food purchasing, women are more selective and tend to integrate multiple cues in the household more than men. In contrast, men are generally more confident and more willing to take risks in purchasing complex products/services than women (Erasmus, Donoghue, & Dobbelstein, 2014). Cavaliere, Ricci, and Banterle (2015) reported that women are more concerned about a healthy diet and have high levels of personal knowledge on food characteristics, and thus, they are more careful than men about what they eat. Dibb and Fitzpatrick (2014) also showed that men tend to consume more meat than women and are less willing to consider reducing their consumption.
A negative correlation between age and purchasing decision suggests that younger people were less concerned in making decisions to purchase than older people. In contrast, a positive correlation between age and WTP shows that older and more experienced people tend to be more conscious about the meat products they buy. Although household income did not play a significant role in this study, pork and poultry products were mostly demanded by respondents from Kadutu and Bagira, while those from Ibanda rated beef and goat meat highly. This result can be explained by the fact that beef and goat are sold in large portions that require refrigeration: Pork and poultry are mostly sold in smaller portions that do not need it. People in Ibanda who have access to more electricity are likely to purchase and consume goat meat. Likewise, higher income and more educated consumers in Ibanda may prefer quality rather than quantity of products when compared to consumers in Kadutu and Bagira. This could be explained by the budget constraints of lower-income households that may be limited to cheaper choices (Morales & Higuchi, 2018). This is in agreement with the findings of Jolly, Bayard, Awuah, Fialor, andWilliams (2009) andSabran, Jamaluddin, Abdul Mutalib, andAbdul Rahman (2012) who mentioned that wealthier consumers are more likely to take precautions about food and are more willing to pay for high-quality products than those with lower incomes. Additionally, Silva, Caro, and Magana-Lemus (2016) also found that food-secure households with higher incomes purchase a wider variety of highquality food items than food-insecure households with lower incomes. However, this finding contrasts with the studies reported by Robert, Manolis, and Tanner (2003) who reported that lowerincome consumers are more concerned about the value of money and with not wasting their money on goods and services that do not meet their basic needs (Erasmus et al., 2014).
Moreover, it could be seen that the more educated people in Ibanda generally have higher incomes; thus, they might have more options than less educated people when purchasing meat products. Also, people in Ibanda might be sensitive to quality since meat products can be a risk factor for their health. In Ibanda, high blood pressure (57%), high cholesterol levels (21%), and incidence of diabetes (20%) were reported as a cause of specific dietary requirements, while in Kadutu and Bagira, the averages reported were 41% for high blood pressure, 16% for high cholesterol, and 14% for diabetes (data not shown). A study by Chen, Anders, and An (2013) showed that consumer willingness to purchase also increased with level of education; and the education level was positively linked to consumers' willingness to adopt new products (Huotilainen, Pirttila-Backman, & Tuorila, 2006 (Pfeiler & Egloff, 2018). Luning, Marcelis, and Jongen (2002) mentioned that quality represents the features/properties of a product that result in satisfying the consumers' physiological and/or psychological needs. Dransfield (2005) also suggested that at least two attributes of appearance are normally used by consumers in quality judgements on meat. For instance, cut type, color, and fat structure and levels have been observed as influential in calculating quality expectations (Grunert et al., 2004). When the influence of product attributes on purchasing decisions in this study is considered, quality aspects such as color and in-mouth texture cannot be ignored.
Color as an intrinsic quality attribute influences consumers' expectations of meat quality at the moment of purchase (Carpenter, Cornforth, & Whittier, 2001;Font-i-Furnols & Guerrero, 2014;Gracia & de Magistris, 2013;Verbeke et al., 2005;West, Larue, Touil, & Scott, 2001), probably because consumers normally use color to indicate wholesomeness or contamination of meat products (Mancini, 2009;Owusu-Sekyere et al., 2014). On the other hand, eating quality and in-mouth texture are found to be highly correlated with the overall experienced quality, attitude to purchase, and WTP for meat products (Lusk et al., 2001;Banović, Grunert, Barreira, & Aguiar Fontes, 2009). Robbins et al. (2003) reported consumers were most concerned with color, fat content, price, and type of cut when purchasing beef, whereas texture and flavor were most important in determining eating satisfaction.
The findings from this study also suggest that availability (marketing factor) is one of the most important attributes that influences purchasing decisions and WTP of meat products. Availability is one reason that can explain, for instance, the lack of access to markets and market information that had a negative influence on consumers' WTP and purchase behavior toward food products (Zundel and Kilcher, 2007;Young, Hwang, McDonald, & Oates, 2010). Young et al. (2010) also mentioned that consumers generally do not like to spend much time searching for food products although perception, a psychological motivator for purchasing meat products, affects the process for consumers in selecting, organizing, and interpreting information related to meat products (Kotler, Armstrong, Harris, & Piercy, 2013). This factor is important in shaping consumers' acceptance, purchase, and future consumption, as stated by Grunert, Verbeke, Kügler, Saeed, and Scholderer (2011). The results in this study exhibited a sig-  Figures 3,4), it appears that the higher the income, the better the consideration that is given to nutrition, harmful effect, and availability as important factors on purchasing decisions and WTP. This result agrees with the findings of Henchion, McCarthy, Resconi, and Troy (2014) who pointed out that the influence of factors such as income and price are likely to decline over time so that other factors, such as quality, will become more important in purchasing meat products.

| CON CLUS IONS
This exploratory study investigated the preference and WTP for meat products of Congolese consumers in Eastern DRC. The study revealed that women and older consumers from urban areas were more likely to purchase meat products. Although the respondents were expected/hypothesized to rate healthiness, quantity, and the low price of products, consumers' decisions to purchase meat products are more often based on sensory factors such as color and in-mouth texture as well as on marketing factors such as availability. Availability played a prominent/key role on their WTP. However, nutrition, harmful effect, and availability tended to be taken into consideration in higher income groups.
This result is related to personal WTP and is a consequence of consumers' poor access to information about meat quality.
Therefore, public efforts are needed to address knowledge gaps through awareness campaigns that promote and disseminate information about meat quality. In summary, the empirical findings presented here reveal new and essential insights into consumers' preferences and their purchase of meat products (in a region where food insecurity is prevalent). These insights provide practical insights for actors in the meat value chain to better satisfy consumers' expectations, demands, and needs. The findings can be used to identify opportunities for livestock farmers to commercialize livestock enterprises for income and employment generation, thus contributing to improving nutrition and alleviating poverty. These insights can also be of relevance to countries with similar socioeconomic characteristics in low-income countries.

ACK N OWLED G M ENTS
The authors gratefully acknowledge the "ILRI/IITA Crop Livestock Integration Project (PJ-002057)" for the opportunity to prepare this article.

CO N FLI C T O F I NTE R E S T
The authors have no conflict of interests.

E TH I C A L S TATEM ENT
This study does not involve any human or animal testing.