Evaluating food transfers in Botswana using multiple matching methods

Data from a household survey in Botswana were analysed to assess the effectiveness of food transfers. Food transfers have been contentious, and much assessment has focused on market participation while overlooking direct impacts on hunger. This paper quantitatively assesses food transfers while controlling for selection bias. Responses to food insecurity questions and Body Mass Index of the head of household were tested for association with socioeconomic variables using logistic and ordinal regression. Respondents who were female (p < .0001), lived outside major cities and towns (p = .0002), lived in larger households (p < .0001), or had never received formal education (p < .0001) were more likely to report hunger and be underweight. After controlling for selection bias using several Propensity Score and Exact Matching methods, we showed that food transfers reduced the proportion of respondents reporting hunger (p = .017 with Optimal 1:2 matching). This demonstrates that food transfers targeted the hungry and alleviated household food insecurity.