LAUSR

BACKGROUND The diets of millions of poor individuals lack adequate amount of essential nutrients. OBJECTIVE To examine the determinants of household dietary diversity in Burkina Faso and assess whether the choice of a diversity metric matters. METHODS Using survey data from 2014, we construct 3 metrics-Household Dietary Diversity Score (HDDS), Berry Index (BI), and Healthy Food Diversity Index (HFDI). Unlike the oft-used HDDS, the BI captures the quantity distribution of food items while the HFDI captures all 3 aspects of a healthy diet-count, quantity distribution, and health value. We fit linear (for BI and HFDI) and Poisson (for HDDS) models controlling for several socioeconomic and climatic covariates. RESULTS Some parameter estimates are sensitive to the diversity metric with fewer significant covariates observed in the HFDI model. Overall, diets are more diverse for households in urban areas, with female or better educated heads, with higher asset-based wealth and with more diverse on-farm production, while remoteness reduces dietary diversity. Higher precipitation seems to reduce diversity, potentially driven by the spatial heterogeneity in precipitation and on-farm production diversity. CONCLUSIONS The sensitivity of estimates to the metric used underscores potentially more complex interactions that determine the quantity distribution of food items consumed. Policies that enhance on-farm production diversity, market access, and women's empowerment may help improve dietary diversity and subsequent nutritional benefits. Efforts should be made to compile health value data that are relevant to developing countries facing nutrition transition.