Search

One major limitation to addressing child undernutrition is a lack of the information required to target resources. This article extends the small-area estimation technique of Elbers, Lanjouw, and Lanjouw (2002, 2003) to jointly estimate multiple equations while allowing for individual-specific random errors across equations (in addition to cluster- and household-specific random errors). Estimates of the prevalence of stunting and underweight for children under age 5 in Cambodia from 17 Demographic and Health Survey strata are disaggregated into 1,594 communes by combining the Demographic and Health Survey data. The estimates are consistent with the survey-only estimates at the aggregate and primary sampling unit levels. The accuracy of the commune-level estimates is comparable to the survey-only estimates at the stratum level. The results are robust, and the estimates are useful for policy analysis and formulation. The small-area estimates can be presented in various ways. The strengths of each representation are also discussed. [ABSTRACT FROM PUBLISHER]

Summary: We compare commune-level poverty rankings in Cambodia based on three different methods: small-area estimation, principal component analysis using aggregate data, and interviews with local leaders. While they provide reasonably consistent rankings, the choice of the ranking method matters. In order to assess the potential losses from moving away from census-based poverty mapping, we used the concentration curve. Our calculation shows that about three-quarters of the potential gains from geographic targeting may be lost by using aggregate data. The usefulness of aggregate data in general would depend on the cost of data collection. [Copyright &y& Elsevier]