Small Area Poverty Estimation by Conditional Monte Carlo

Small area poverty estimates are important for social and economic policy, however the required data are often unavailable. This paper presents a Small Area Estimation (SAE) technique called Conditional Monte Carlo (CMC). CMC provides robust estimates of small area poverty rates, subject to fewer restrictive assumptions than existing methods. We present a theoretical derivation followed by a numerical validation. Using Mexican data, CMC replicates small area poverty rates with precision, successfully controlling for unobserved heterogeneity in the relationship between predictor and outcome variables through discriminate microdata sampling. CMC produces spatially-referenced microdata, providing a platform for agent-based modelling and microsimulation analysis.