How accurate is a poverty map based on remote sensing data? An application to Malawi

This paper assesses the reliability of poverty maps derived from off-the-shelf remote-sensing data. Employing data for Malawi, it first obtains small area estimates of poverty by combining household expenditure survey data with population census data. It then ignores the population census and obtains a second poverty map by combining the survey with predictors of poverty derived from remote sensing data. The two approaches reveal the same patterns in the geography of poverty. However, there are instances where the two approaches obtain markedly different estimates of poverty. Poverty maps obtained using remote sensing data may do well when the decision maker is interested in comparisons of poverty between assemblies of areas yet may be less reliable when the focus is on estimates for specific small areas. •Remote-sensing-based and census-based poverty maps show close correspondence.•For select districts, however, the two approaches obtain markedly different estimates of poverty.•The random offsets embedded in public use coordinates do not meaningfully alter the estimated poverty map.•Yet random offsets may, however, lead to an over-estimation of statistical precision.•The random offsets partially destroy the spatial correlation structure in the data and error term.