Combining Census and Survey Data to Study Spatial Dimensions of Poverty

Abstract

Poverty maps, providing information on the spatial distribution of living standards, are an important tool for policy making and economic research. Policymakers can use such maps to allocate transfers and inform policy design. The maps can also be used to investigate the relationship between growth and distribution inside a country, thereby complementing research using cross-county regressions. The development of detailed poverty maps is difficult because of data constraints. Household surveys contain data on income or consumption but are typically small. Census data cover a large sample but do not generally contain the right information. Poverty maps based on census data but constructed in an ad-hoc manner can be unreliable. The authors demonstrate how sample survey data and census data can be combined to yield predicted poverty rates for all households covered by the census. This represents an improvement over ad hoc poverty maps. However, standard errors on the estimated poverty rates are not negligible, so additional efforts to cross-check results are warranted.