TY - BOOK ID - 134130464 TI - Brazil Within Brazil : Testing the Poverty Map Methodology in Minas Gerais AU - Ebers, Chris AU - Lanjouw, Peter AU - Leite, Phillippe George PY - 2008 PB - Washington, D.C., The World Bank, DB - UniCat KW - Confidence intervals KW - Descriptive statistics KW - Education KW - Enumeration KW - Geographical Information Systems KW - Precision KW - Predictions KW - Reliability KW - Sample design KW - Sample surveys KW - Science and Technology Development KW - Science Education KW - Scientific Research and Science Parks KW - Small Area Estimation Poverty Mapping KW - Standard errors KW - Statistical and Mathematical Sciences KW - Validity UR - https://www.unicat.be/uniCat?func=search&query=sysid:134130464 AB - The small-area estimation technique developed for producing poverty maps has been applied in a large number of developing countries. Opportunities to formally test the validity of this approach remain rare due to lack of appropriately detailed data. This paper compares a set of predicted welfare estimates based on this methodology against their true values, in a setting where these true values are known. A recent study draws on Monte Carlo evidence to warn that the small-area estimation methodology could significantly over-state the precision of local-level estimates of poverty, if underlying assumptions of spatial homogeneity do not hold. Despite these concerns, the findings in this paper for the state of Minas Gerais, Brazil, indicate that the small-area estimation approach is able to produce estimates of welfare that line up quite closely to their true values. Although the setting considered here would seem, a priori, unlikely to meet the homogeneity conditions that have been argued to be essential for the method, confidence intervals for the poverty estimates also appear to be appropriate. However, this latter conclusion holds only after carefully controlling for community-level factors that are correlated with household level welfare. ER -