Adaptive Safety Nets for Rural Africa: Drought-Sensitive Targeting with Sparse Data

Abstract

The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.

Predicting poverty and malnutrition for targeting, mapping, monitoring, and early warning

“…Kshirsagar et al (2017) also highlight the important role of bootstrap aggregation --parameterizing a model by averaging over many subsamples from the data --within model development to achieve a model that performs consistently well across subpopulations within a country. Baez et al (2019) emphasize the value and importance of using child malnutrition indicators on the left hand side of a targeting model when interested in child well-being, rather than consumption or expenditure measures that often prove noisy indicators of welfare among agricultural households and more predictive of long run health and human capital outcomes. In producing a drought-contingent targeting model using Demographic and Health Survey (DHS) data combined with Normalized Difference Vegetation Index (NDVI), Baez et al (2019) find that simple and easily interpreted methods such as logistic regression and classification trees do just as well as black box methods such as random forest and gradient boosting in the prediction of child level stunting, suggesting that the trade-offs between interpretability and strong prediction may be modest.…”

“…Baez et al (2019) emphasize the value and importance of using child malnutrition indicators on the left hand side of a targeting model when interested in child well-being, rather than consumption or expenditure measures that often prove noisy indicators of welfare among agricultural households and more predictive of long run health and human capital outcomes. In producing a drought-contingent targeting model using Demographic and Health Survey (DHS) data combined with Normalized Difference Vegetation Index (NDVI), Baez et al (2019) find that simple and easily interpreted methods such as logistic regression and classification trees do just as well as black box methods such as random forest and gradient boosting in the prediction of child level stunting, suggesting that the trade-offs between interpretability and strong prediction may be modest. Baez et al (2019) also make clear the value of augmenting relatively sparse data, such as DHS household level surveys, with big data such as NDVI satellite data for targeting purposes.…”

“…Conceptualizing resilience as a measure of future well‐being allows one to estimate the probability of staying above a welfare threshold as a function of assets and exposure to shocks (Cissé & Barrett, 2018; Knippenberg et al, 2019). Along these lines, Baez et al (2019) propose a drought‐contingent early warning system to identify regions where children are at risk of stunting due to drought. Knippenberg et al (2019) estimate a food security early warning system, with the food security coping strategies index as the response variable, using LASSO and random forests for feature selection.…”

“…In producing a drought‐contingent targeting model using Demographic and Health Survey (DHS) data combined with Normalized Difference Vegetation Index (NDVI), Baez et al (2019) find that simple and easily interpreted methods such as logistic regression and classification trees do just as well as black box methods such as random forest and gradient boosting in the prediction of child‐level stunting, suggesting that the trade‐offs between interpretability and strong prediction may be modest. Baez et al (2019) also make clear the value of augmenting relatively sparse data, such as DHS household‐level surveys, with big data such as NDVI satellite data for targeting purposes.…”

“…A significant challenge in predicting malnutrition outcomes is that they tend to be measured with noise, and some workhorse measures, such as severe wasting, are, thankfully, less prevalent than poverty (Baez et al, 2019; Head et al, 2017). To address this challenge, Zhou et al (2021) use oversampling combined with random forest and gradient boosting models; they demonstrate that oversampling rare and noisy food insecurity outcomes in model development can significantly improve predictive accuracy in the targeting and forecasting of food insecurity at the cluster level.…”