Estimating and Forecasting Income Poverty and Inequality in Haiti using Satellite Imagery and Mobile Phone Data

The 44-page study estimates the distribution of income poverty and inequality in Haiti's 140 communes and 570 communal sections from features of anonymized mobile phone data and satellite imagery. It works in two stages: small-area estimates built from the 2003 census and the 2012 household survey, then a machine learning model trained on those estimates to produce indicators for 2014 and forecasts for 2019, with 2014 commune estimates checked with the first-stage estimates. It uses the 2012 national poverty line. Three trends emerge: better-performing communes were mostly in Ouest, some Sud communes became more deprived between 2014 and 2019, likely after Hurricane Matthew, and Nord-Ouest communes grew poorer. One in four poor Haitians lived in 10 communes in Artibonite, Ouest, Nord-Ouest and Nord, and only three of 140 communes show low inequality. The authors say the method complements surveys and cannot replace them.