Small area estimates of poverty andinequality statistics, through survey-to-census imputationthat lets consumption be estimated for each and everyhousehold in a census, are useful for at least threereasons. First, they can help improve the effectiveness ofpublic spending, by targeting to prevent the leakage ofbenefits to the non-poor (and prevent the under-coverage ofthe poor). If poor people are concentrated in certain areas,spatial targeting by directing extra development projectsand public services to those areas, may be more feasiblethan trying to individually target the poor. Geographictargeting is highly relevant in countries like Timor Leste,where mountainous topography contributes to high levels ofheterogeneity. In similar environments, such as Papua NewGuinea, the enclave nature of some modern economicdevelopment has created high levels of spatial inequality.The basic details are that household survey data are used toestimate a model of consumption, with explanatory variablesrestricted to those that have overlapping distributions froma census. The coefficients from this model are then combinedwith the variables from the census, and consumption ispredicted for each household in the census. With thesepredictions available for all households, inequality andpoverty statistics can be estimated for small geographicareas (Elbers et al, 2003).2 In the results below, thepoverty statistics that are calculated by using thepredicted consumption data for each census household arereported at the suco level (n=442). For the headcountpoverty rate, the standard errors at the suco level(relative to the poverty index) average one-quarter and sothis is a comparable degree of precision to what the surveyoffered at the municipality level (n=13) for a variable likethe poverty severity index.
PDF
download
878987797
028-85220240
