A Good Egg: An Evaluation of a Social Behavior Change Campaign to Increase Egg Consumption Among Children in Rwanda

Abstract

Childhood malnutrition, which is endemic in rural areas of low-income countries, leads to a host of deleterious outcomes such as poor cognitive development, worse educational attainment and lower lifetime wages and productivity. However, promoting the consumption of eggs among young children has emerged as a promising strategy to combat childhood malnutrition. Eggs are a convenient and readily available source of animal protein and contain nutrients that promote cell growth and brain development.

Rwanda has high rates of malnutrition, particularly in its rural areas, and despite the relatively common practice of chicken rearing, one of the lowest rates of egg consumption in the world. In this dissertation, I evaluate the impact of two interventions which aim to increase the consumption of eggs among young children, using a difference-in-difference estimation strategy with propensity score matching. In the first intervention (poultry-only), participants of a large agricultural social enterprise purchased Sasso chickens, which are bred to be hearty to rural conditions, on credit, thereby increasing access to eggs. The second intervention (poultry + SBCC) arm layered on a social behavior change campaign (SBCC) of radio messages, limited in person training, text message reminders and posters, all of which aimed to encourage families to feed eggs to children. This campaign was designed based in-depth formative work that identified key motivational constraints to egg feeding. I also included a comparison area, which was supposed to have excluded any similar interventions. However, mid-way through the data collection, I discovered that this area had begun a chicken and egg distribution campaign along with an egg promotion campaign, and thus no longer served as an appropriate counterfactual to the two other study arms. Therefore, the primary analysis of interest estimated the impact of the SBCC campaign by comparing the poultry + SBCC to the poultry-only interventions.

I found a relatively modest but statistically significant increase in the number of times per week respondents in the poultry + SBCC arm reported feeding eggs to children of .275 (p=.02) compared to the poultry-only arm. This low impact was possibly due to the rather limited exposure of many respondents to the campaign, as under half of respondents reported having heard the radio campaign at all. This low listenership was a somewhat anticipated artifact of using a community radio rather than the more popular national radio station, which was necessary in order to have unexposed comparison areas. In addition, in-person training and poster distribution was curtailed due to COVID-19 related restrictions. So, while the impact was rather low and unlikely to make a significant difference in nutritional outcomes, it suggests that a higher saturation campaign might achieve even greater impacts.

There were important trends in the types of respondents for whom the behavior change campaign had a greater impact. Respondents who were feeding boys had a much larger impact (.417 increase in egg feeding per week, p=.002) than those who were feeding girls (.138 increase in egg feeding per week, p=.255). In addition, the campaign had a larger impact on those who were already feeding eggs to children (.629 increase in egg feeding per week, p=.004) than those who were not engaging in those practices at baseline (.256 increase in egg feeding per week, p=.002). Future campaigns should ensure higher saturation of messaging, include specific messaging around the importance of feeding girls as well as boys and could potentially target the easier to reach segment of those who were already occasionally feeding eggs to children. However, the harder to reach non-egg feeders may have greater needs and may also need to be prioritized.

Because the primary outcome of interest in this study (egg consumption among children) was self-reported and prone to social desirability bias, I included a social desirability scale, the Marlowe-Crowne Social Desirability Scale (M-C SDS) Short Form – C, in this study. This study thus also provided an opportunity to validate this scale, which has increased in use in the past decade in Africa, using a large sample of rural East African respondents over two survey rounds. I found that this scale had a low internal reliability across both rounds (alpha<.4), and exploratory factor analysis found that there were five, rather than the commonly found two, underlying constructs. These findings suggest that the M-C SDS Short Form C is not a valid scale for measuring social desirability in rural Rwanda, and possibly other rural areas in Sub-Saharan Africa. Researchers seeking to assess and control for social desirability in these contexts should develop a more appropriate instrument or find other ways to mitigate social desirability bias.