IDB's Machine-Learning Forecast of Commune-Level Poverty and Inequality in Haiti

Haiti's national statistics institute (IHSI) has not run a comprehensive household living-conditions survey since 2001-2003, leaving governments and donors with stale, low-resolution poverty data.

IDB researchers Neeti Pokhriyal, Omar Zambrano, Jennifer Linares and Hugo Hernández built a machine-learning framework combining the 2012 household survey with satellite imagery and anonymised mobile-phone call-detail records to estimate income poverty, inequality and standard-of-living deprivation for all 140 communes and 570 communal sections in Haiti, validating the models against observed 2014 indicators before forecasting through 2019 without any new field survey.

The study (IDB, 2020, DOI 10.18235/0002466) found that one in four poor Haitians live in just ten communes concentrated in the Artibonite, Ouest, Nord-Ouest and Nord departments, that Ouest department communes consistently outperform the rest of the country, and that Nord-Ouest communes grew poorer and more deprived between 2014 and 2019 relative to other regions.

The exercise is a research product built to inform policy, not a government-run operational system: no public record ties the estimates to a specific budget or targeting decision, and Haiti's own statistical agency has not published a comparable update since the original 2001-2003 survey.