The results of the pooled sample show that about 38% of the sample households sold their maize directly to consumers in the local market (Table 2 in Appendix). About 19% and 20% sold maize through retailer and wholesaler channels, respectively, while 23% sold through the collectors. About 35% of male decision-making households sold maize through collectors whereas 44% and 39% of female and joint decision-making households sold it directly to consumers in the local market, respectively.
On average, the age of household heads was 42.6 years with the highest average age in households selling maize through wholesalers (43.7 years), followed by those through collectors (43.5 years). In comparison, households that directly sell maize to consumers in the local market owned on average fewer livestock, had a lower rate of improved maize seed application, and allocated a smaller area of their farmland to maize production.
The average age of the household head is highest among joint decision-maker households, particularly those selling maize through collectors, followed by those through consumers (Table 3 in Appendix), and is lowest among female decision-making households. The average years of the household head’s education is higher among female decision-making households than those of male and joint decision-makers. The average number of adult male and adult female family members is highest in male decision-making households, followed by joint decision-making households, and lowest in female decision-making households. The average number of livestock owned is highest in male decision-making households, followed by joint and female decision-making households.
On average, households using improved maize seed, the area of farmland for maize, and the amount of maize sold to the market are higher in male decision-making households than female and joint decision-making farm households. However the average unit price received from maize sale is highest in joint decision-making households while the average unit marketing cost incurred by maize farm households is highest in female decision-making households.
To estimate the MNL model first we began by normalizing one category, usually referred to as the base category, in this case ‘consumers in the local market’ since most sampled farm households choose direct sale to consumers in the local market. We then tested for potential endogeneity or any situation in which explanatory variables are correlated with residuals. Other studies have suggested that access to credit service, market information, contact with extension agents, and participation in social events be assumed as endogenous variables in a choice model (e.g., Deressa et al. 2009; Mmbando et al. 2016). For our test we adopted a two-stage approach that involves the use of predicted values of potentially endogenous variables (Wooldridge 2012). Probit models for access to credit, market information, contact with extension agents, and participation in social events are specified in the first stage. We then used the predicted values of these variables in the second stage of estimating factors affecting farm households’ choice of maize marketing channel. The test failed to reject the null hypothesis, suggesting there is no significant correlation between explanatory variables and residuals. Subsequently, we fitted the Ordinary Least Square model and tested for multicollinearity by using the Variance Inflation Factor (VIF). The VIFs for all the explanatory variables are less than 10 (1.02–1.64), which suggests that there is no serious multicollinearity problem among the explanatory variables included in the model. Finally, we ran the model and tested for the validity of the independence of irrelevant alternative (IIA) assumption by using the Hausman specification test. The test failed to reject the null hypothesis of the independence of the maize marketing channel choice options, suggesting that the MNL specification is appropriate for modeling the maize marketing channel choice of the smallholder maize farm households.
We ran pooled and separate sample models to determine the effect of gender on maize marketing channel choice. In both models, the likelihood ratios indicated by statistics are significant at 1.0% probability, suggesting that both models have a strong explanatory power. The pooled model explains 23.32% of the variation in the market choice among the sampled maize producers. The separate model explains 18.56% of variation in the market choice among male, 48.49% among female, and 39.67% among joint decision-making households, respectively.
As indicated earlier, the coefficient estimates of the MNL model provide only the direction of the effect of the regressor variables on the response variable, i.e., estimated coefficients do not represent the actual magnitude of change or probabilities. Thus, we report and discuss the average marginal effects from the MNL, which helps measure the expected change in the probability of a particular market channel choice being made with respect to a unit change in regressor variables. In all the cases, the estimated coefficients should be compared with the base category of direct sale to consumers in the local market.
In the pooled sample model, the gender of female and joint decision-makers is included to examine the relative positions of an individual or a joint decision-making pair within a farm household while the gender of male decision-makers is considered as a reference group for the analysis. The result indicates that female decision-makers reduce the probability that maize producing households would sell maize to collectors at the farm gate by 9.1% (Table 4 in Appendix). Meanwhile, they had a higher probability of selling maize in the local market by 13.7%.
There are three possible explanations for this result in view of the gender-specific constraints and women’s marketing behaviors in the study area. First, women dominate local market sales by negotiating with buyers who are themselves often local women buying maize for their family consumption. The women peddlers in the local village market have strong social bonds with customers living in the neighborhood. They frequently visit a local market (usually once a week) as they do not have time to travel to the main market located far from the local community, given that they spend significant time each day in obligatory household activities. Second, most women prefer to occasionally sell small quantities in the local market while most men prefer to sell one-shot bulk quantities in more distant markets (Aregu et al. 2011). This behavior is mainly linked to the price volatility of maize in the study area as confirmed by agricultural experts, community elders, and farmers; they could be wary of incurring a significant loss by selling in bulk when prices drop. Considering women’s responsibility for family sustainability, they may wish to minimize the risk. This explanation can thus be linked to the generally more risk adverse behavior of women than men (Eckel and Grossman 2008). Lastly, women producers could be less visited by collectors, who may assume that men are primary agricultural producers in the village. This explanation is linked to the notion that men are more likely approached by traders than women for their agricultural products (Barham and Chitemi 2009).
Another result is that joint decision-making households are 13.8% less likely to sell maize to wholesalers in their nearby town. Conversely, they are 13.3% more likely to sell maize in the local market. According to Nyikahadzoi et al. (2010), collective marketing reduces the cost of getting the product to markets and helps improve farmers’ bargaining power. The result may thus suggest that joint decision-making households selling their maize products in the local market tend to incur lower transaction costs than do either male or female decision-making households. Another explanation may be related to the tendency for women to play a leading role when they make local marketing decisions. This is exemplified by the following interview narrative provided by a male farmer:
When I sell maize in the local market, I always go there with my wife to help her for transportation and, most importantly, security. I always prefer not to engage in sales activities because most of the buyers there are women and they always charge a lower price for me by saying, ‘you are man, you are the main producer, so do not get involved in this women’s activity’. As a cultural norm, it is no good for us to argue with a woman in the local market. Hence, I let my wife talk to them and she easily negotiates with them for better prices.
His wife in turn confirmed her husband’s view: “I am in charge of selling maize in the local market. However, we [wife and husband] are handling money from the sales together.” A further explanation of women’s relatively louder voice in the joint decision-making households could be linked to the price volatility of maize as well as quantitative requirements of the wholesale market. Wholesalers usually require large quantities of produce for purchase although the future price of maize is usually unpredictable for smallholder farmers. Thus, men and women who make joint decisions tend to sell maize in the local market as it caters for much smaller quantities than the wholesale market. In this connection, women who have stronger bonds with customers in the local market are better positioned to take advantage of such relationships for their maize marketing, hence helping to maximize their household economic welfare.
The number of adult male and female family members influences the choice of maize marketing channel. The addition of an adult female in the household decreases the probability by 2.8% that the household would sell maize to retailers in the main market and increases the probability of selling it to consumers in the local market by 2% (cf. Aregu et al. 2011). The addition of an adult male in the household increases the probability by 2.6% that they sell maize to collectors at the farm gate; however, they would decrease the probability of selling to consumers in the local market by 1.7% (cf. Amani 2014 for Burkina Faso and Rwanda).
Growing improved maize varieties increases the probability that producers sell maize to retailers in the main market and to collectors at the farm gate by 5.5% each; however, it decreases the probability by 6.3% (as compared to growing traditional maize varieties) that they sell them to wholesalers. These results are linked to the quality of improved maize seeds used by farmers and the storage capacity of traders involved. According to the Ethiopian Seed Association (2014), a lack of quality seed is one of the critical constraints to increasing production and productivity in Ethiopia (see also Gebre et al. 2019). On the other hand, the FAO (2015) and World Bank (2018) note that maize traders in Ethiopia face constraints in the capacity of their storage facilities. Maize traders in our study lack capital to invest in large modern maize storage. Compared to other traders, wholesalers are able to store maize for much longer, up to 2 ~ 3 months depending on the market price. Our results indicate that producers are more likely to sell improved maize to retailers or collectors than wholesalers as retailers and collectors sell products immediately after they purchase them. In the study area there are no quality standards nor grades in the maize market, and collectors mix improved maize with traditional maize varieties. Most wholesalers receive maize from collectors, often in such a mixture.
The area of farmland allocated to maize production increases the probability that the farm household sells maize to retailers in the main market rather while it decreases the probability of selling directly to consumers in the local market (cf. Amaya and Alwayng 2011).
The amount of maize households sell influences their choice of marketing channel. Our results indicate that a one-quintal increase in the amount of maize sold increases the probability by 12% that the household sells maize to collectors at the farm gate (with ‘sale to consumers in the local market’ being the base market for comparison). This could be related both to the relatively small amounts required in the local market and poor access to roads and trucks to transport their produce to market. However, it decreases the probability by 1.9% that the household sells maize to wholesalers. An increase in the price of maize increases the probability that farm households sell maize to consumers in the local market while it decreases the probability that they sell maize to wholesalers or collectors.
Since that rational farmers would prefer to sell produce in the market where they can reap the most benefit (Mmbando et al. 2016), an increase in the cost of marketing increases the probability that producers sell maize to wholesalers or collectors rather than directly to consumers in the local market (cf. Masuku et al. 2001).
The age of the household head increases the probability that female and joint decision-making households sell maize to collectors at the farm gate (Table 5). Older farmers (who are most likely the household head) sell farm produce to a closer market (cf. Amaya and Alwayng 2011; Mmbando et al. (2016). As age increases, they lose interest in traveling (even to the local market) and shift to focus on selling produce at the farm gate.
Number of adult females in the male decision-making household is negatively associated with selling to retailers in the main market while positively associated with selling to consumers in the local market. In female decision-making households, number of adult females is positively associated with selling directly to consumers in the local market but negatively associated with selling to wholesalers. In joint decision-making households, an increase in the number of adult females increases the probability by 5.5% that joint decision-makers sell directly to consumers in the local market. The finding might be related to household production capacity. For some agricultural activities such as plowing with oxen and planting, male and female labour is not interchangeable. Plowing with oxen is culturally considered as a male task in the study area (Gebre et al. 2020). Thus, given the gendered division of labour for agricultural production, a higher number of adult females in the household (with the number of working age adults in the household held constant) may lead to diminished household farm output. This in turn leads to less marketable produce. Women who prefer to sell smaller quantities are more likely to sell in the local market.
Given the gender division of labour in agriculture, a higher number of adult male family members could provide more household production. In all the three decision-making types of households, an increased number of adult men lead to more sales to collectors. In contrast, an increase in the number of adult males in the male decision-making household decreases the probability of household maize sales to consumers by 3.7% since local market exchanges are dominated by women.
In our sample, the majority of female decision-makers are household heads. The adult males in these households are usually their adult sons may prefer to sell maize to wholesalers and claim the income. Our results show that an adult male added to a female decision-making household increases the probability of household maize sale to consumers or collectors by 4.3% and 5.7%, respectively, whereas the probability of selling maize to wholesalers would decrease by 10.1%.
Similarly, an increase in the number of adult males in joint decision-making households increase the possibility of conflict between men and women selling maize and controlling income. A woman who jointly makes decisions with a man in the household would be unwilling for him to sell maize in distant markets. Our results indicate that in fact they generally agree to sell to collectors at the farm gate and share control over the income from the sale.
Number of livestock owned by male decision-making households increases the probability that they would sell maize to wholesalers in town (cf. Aregu et al. 2011).
Planting improved maize varieties decreases the probability that male decision-making households sell to wholesalers in town and increases the probability they sell to collectors at their farm gate. For female and joint decision-making households, growing improved maize variety increases the probability that they sell maize to collectors at the farm gate and retailers in the main market, respectively. Our results also indicate that growing improved varieties results in a decrease in the probability that female and joint decision-makers sell maize to consumers, by 12.8% and 11.5%, respectively but rather sell to collectors at the farm gate. However, collectors in the study area are aware of the farmers’ lack of storage facilities and set a lower price than the market in order to take advantage of the farmers’ need to sell directly after harvest. They also set prices according to their social relationship with the farmer. Male decision-makers with good connections to maize collectors may receive a relatively higher price than female decision-makers. Joint decision-makers are more likely to sell to retailers in the main market where prices are higher.
The area of farmland allocated to maize increases the probability that female and joint decision-making households sell maize to collectors at the farm gate (cf. Amaya and Alwayng 2011) since they usually lack trucks to transport their produce to main/distant markets.
.Access to credit services increases the probability that male and joint decision-making households sell maize to collectors at the farm gate. They may receive credit from collectors in advance of maize sales as nationwide evidence suggests (Rashid et al. 2010; Abate et al. 2015; World Bank 2018). Female decision-makers may fear risks of debt default associated with receiving advance credit from collectors and hence rely on selling directly to consumers in the local market.