Poor areas, or only poor people?

Anti-poverty programs often target poor areas even when there is seemingly free migration. Should such programs focus instead on households with perso nal attributes that foster poverty, no matter where they live? Possibly not ; there may be "hidden" constraints on mobility, or location may reveal oth erwise hidden household attributes. Using survey data for Bangladesh we fin d significant and sizable geographic effects on living standards after cont rolling for a wide range of nongeographic characteristics of households, as would typically be observable to policy makers. The geographic structure o f living standards is reasonably stable over time, consistent with observed migration patterns, and robust to testable sources of bias.