Integrating Survey and Geospatial Data for Geographical Targeting of the Poor and Vulnerable: Evidence from Malawi

Summary

Introduces a cost-effective strategy combining household consumption surveys, geospatial data, and a simulated partial registry (450 villages across 10 impoverished Malawi districts) using XGBoost to produce village-level poverty estimates. Partial registry model achieves rank correlation of 0.75 with actual welfare, outperforming PMT scores, Meta Relative Wealth Index, and survey-plus-geospatial predictions alone (rank correlations -0.02 to 0.2). Results robust to Gaussian noise in proxy indicators.