Micro-Level Estimation of Child Undernutrition Indicators in Cambodia

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.