Files
Abstract
For effective decision-making, policymakers and program managers often need
detailed information about the welfare of the population, including knowledge about
which specific areas are most affected by poverty and undernutrition. Household sample
surveys are an important source of information, yet because the typical sample size is
only a few thousand observations, the information is only useful for inferences at high
levels of aggregation, such as the nation or large regional units. In contrast, data sources
with wider coverage, such as national censuses, rarely capture detailed information on
welfare levels. Recently small-area estimation techniques have been applied to the study
of poverty to produce estimates of poverty, or poverty maps, for small geographic units.
This paper uses household survey and unit record census data from Tanzania to
explore the possibility of applying small-area estimation methods to the study of
children’s nutritional status as measured by anthropometry. Overall, undernutrition
models have had lower explanatory power than poverty models, which has important
implications for the precision of the small-area estimates. The analysis finds that
applying small-area estimation techniques to anthropometric data is feasible, although the
relatively low explanatory power of the regressions does limit both the degree of
disaggregation possible and the power to detect significant differences in undernutrition
prevalence between districts and subdistricts. In the case of Tanzania, the nutrition
mapping approach reveals considerable heterogeneity in nutritional status within regions
and within districts. The most striking finding is the much lower levels of undernutrition
in areas classified as urban, including relatively small district centers.