Comparing Delivery Channels to Promote Nutrition-Sensitive Agriculture: A Cluster-Randomized Controlled Trial in Bangladesh

"[E]mbedding nutrition content within agricultural extension workers' services could be a plausible alternative to co-locating specialized nutrition workers with agricultural extension workers."

Making agriculture "nutrition-sensitive" assumes collaboration between the agriculture and nutrition sectors, yet little evidence-based guidance exists on how the sectors should collaborate. This paper describes a randomised controlled trial (RCT) that compared two models of delivering nutrition behaviour change communication (BCC) messages jointly to husbands and wives in Bangladesh: (i) deploying female nutrition workers versus (ii) training mostly male agriculture extension workers - also referred to as sub-assistant agricultural officers (SAAOs) - to deliver nutrition BCC messages. Such staffing patterns are based on traditional perceptions that agriculture is a male domain, and nutrition female; they also assume that female nutrition workers are better able to interact with mothers in delivering nutrition BCC. The RCT further examines whether a co-resident mother-in-law (MIL) affects the effectiveness of each type of extension worker.

The analysis is based on the Agriculture, Nutrition, and Gender Linkages (ANGeL) project in rural Bangladesh, implemented by the Ministry of Agriculture of the Government of Bangladesh. ANGeL was designed by the International Food Policy Research Institute (IFPRI) to inform scalable approaches for gender- and nutrition-sensitive agriculture in rural Bangladesh through agricultural training, nutrition BCC, and gender sensitisation trainings. ANGeL's nutrition training focused on increasing knowledge of the importance of diverse diets, including micronutrient-rich foods and animal source foods, for young children and for women of reproductive age. The goal was not only to affect individual diets but also to increase women's empowerment within the household, motivated by the documented links between empowerment status, agricultural production diversity, and nutritional outcomes in Bangladesh.

To assess whether male extension agents deliver nutrition training as effectively as female nutrition workers, ANGeL included both a treatment arm with nutrition training delivered by SAAOs (T(SAAO)) and a treatment arm with nutrition training delivered by trained female nutrition workers who lived locally and were recruited specifically for the programme (T(APK), which stands for "ANGeL Pushti Kormi" - Pushti Kormi means "nutrition worker"). In both treatment arms, the trainer (either the SAAO or the APK) provided 19 nutrition training sessions over a 17-month period  (July 2016 to December 2017) to husbands and wives in participant farm households. Training sessions were interactive, including lectures as well as discussions, practical demonstrations, and question-and-answer sessions. Both husbands and wives were expected to attend each session, and active participation from both men and women was encouraged.

Because training would be conducted by SAAOs and APKs each assigned to a "block", cluster-randomisation was conducted at the block level. One village from each block was randomly selected. Within each of these villages, 25 farm households with at least one child under 24 months were randomly selected to participate. This yielded 625 households in each treatment arm (1,250 households in total) and 875 households in the control group, for a total sample of 2,125 households. Baseline data were collected between November 2015 and January 2016, prior to the start of the nutrition BCC sessions. Endline data were collected between January and March 2018, after BCC sessions had ended.

Fidelity of implementation - whether the programme was implemented as designed - was high in both treatment arms. Participants reported valuing the training they received in terms of both information learned and improved confidence, relationships, and social ties. There were no meaningful differences in the assessment of trainings provided by SAAOs or APKs.

Both approaches increased nutrition knowledge of men and women, household and individual diet quality, and women's empowerment. For example, the APK treatment resulted in a statistically significant increase in consumption of eggs and dairy, but not fruits, vegetables, or fish. The magnitudes of these effect sizes were large when expressed as percent increases - 47% for eggs and 36% for dairy - but, given the low baseline mean levels of consumption of these foods, the absolute level of the change is modest. The coefficients of the SAAO treatment arm on the consumption of eggs and dairy were positive, but not statistically significant. Also, for women's empowerment outcomes, there were significant positive impacts from both treatment arms relative to the control group. The women's empowerment score increased by 0.03, and the prevalence of empowered women increased by 5-6 percentage points.

The only evidence of same-sex homophily comes from an attitudes score, which increases more for men if they were trained by SAAOs, and more for women if trained by APKs.

Co-resident MILs change the empowerment environment within the household. They significantly weaken programme impacts of both SAAO and APKs on the women's empowerment score, with a weak or undetectable impacts on whether women are empowered or the total attitudes score. While the MIL's presence has no impact on the effectiveness of the SAAO treatment on men's empowerment scores, it significantly weakens the impact of the APK treatment on men's empowerment. Notably, coresident MILs may not support changes in gender norms that challenge traditional patterns (men helping in the domestic sphere, contributing to the family) but go along with those that are consistent with cultural stereotypes (men's contribution to the community). It is possible that including MILs in nutrition trainings might help to bring them on board, rather than have them offset the positive impact of nutrition BCC. Doing so, however, requires attention to ensuring that the content of the training resonates with MILs.

In reflecting on the findings, the researchers ask, Can training men and women jointly overcome the usual barriers faced in training those of a different gender? Although they cannot answer this question definitively in the absence of a treatment arm where men or women were trained alone,  qualitative work on ANGeL reveals that men and women beneficiaries in the T(SAAO) and T(APK) arms valued the joint training of husbands and wives. For example, a woman beneficiary in the APK treatment arm said, "If I attend the training sessions alone, I (have) to explain in detail to my husband. It can be tough for me to convince him. Now, since we go together, he knows all the things. We discuss and take decision easily."

Thus, the results suggest that opposite-sex agents may not necessarily be less effective in providing training. In South Asia, where agricultural extension systems are male-dominated, training men to deliver nutrition messages may offer a temporary solution to the shortage of female extension workers and offer opportunities to scale and promote nutrition-sensitive agriculture. However, the finding that the presence of MILs within households modifies the programme's effectiveness on some nutrition, empowerment, and attitude measures suggests that accounting for other influential household members is a potential area for future programming.