Consistent poverty comparisons and inference

Building upon the cost of basic needs (CBN) approach, an integrated approach to making consumption-based poverty comparisons is presented. This approach contains two principal modifications to the standard CBN approach. The first permits the development of multiple poverty bundles that are utility consistent. The second recognizes that the poverty line itself is a random variable whose variation influences the degree of confidence in poverty measures. We illustrate the empirical importance of these two methodological changes for the case of Mozambique. With utility consistency imposed, estimated poverty rates tend to be systematically higher in rural areas and lower in urban areas. We also find that the true confidence intervals on the poverty estimates—those incorporating poverty line variance—tend to be considerably larger than those that ignore poverty line variance. Finally, we show that these two methodological changes interact. Specifically, we find that imposing utility consistency on poverty bundles tends to tighten confidence intervals, sometimes dramatically, on provincial poverty estimates. We conclude that this revised approach represents an important advance in poverty analysis. The revised approach is straightforward and directly applicable in empirical work.