Poverty prediction and targeting over time and space: Evidence from Nigeria

Abstract

Understanding poverty dynamics is crucial to target and tailor economic policies in developing countries like Nigeria—a country at the risk of hosting about a quarter of all people living in poverty worldwide. To facilitate the targeting of poverty-reducing interventions, we build a nationally representative panel dataset spanning 2011–2019 with more than a hundred covariates and apply econometric and machine learning tools to predict and examine factors associated with the static, transient, and persistent poverty status of Nigerian households. Results show that demographic factors, asset holdings, access to infrastructure, and housing indicators can accurately predict poverty in 80% of cases.