14:30, Tuesday, 17 November, 2015
Location: University of Wollongong, Building 8, Room G25
Speaker: Mr. Sumonkanti Das PhD Research Student of Statistics National Institute for Applied Statistics Research Australia (NIASRA) School of Mathematics and Applied Statistics University of Wollongong
Title: Robust Poverty Mapping Inference
Abstract: The ELL (Elbers, Lanjouw and Lanjouw, 2003) methodology for poverty mapping was developed by the World Bank and is widely used in developing countries. The method combines household survey income and expenditure data with population census data but does not require any linkage between the datasets at the household level. However, the ELL method has also been criticized because of its assumption of negligible between area variability in calculation of small area poverty estimates and the mean squared error (MSE) of these estimates. In particular, these MSEs are significantly underestimated when the between area variability in the income and expenditure data cannot be adequately explained by the explanatory and contextual variables in the regression model. In this presentation a method of MSE estimation for ELL-type estimates is proposed which is robust to the presence of significant unexplained between area variability in the variables underpinning the poverty estimates. The proposed idea is to estimate a correction factor for the ELL based MSE estimation based on the relationship between estimated variance components when data are generated by a three-level model but are subsequently fitted via a two-level model. Simulation results provide evidence that this correction to ELL based MSE estimates performs well in a realistic poverty estimation scenario. An application to a poverty mapping study of Bangladesh is described. Significant between-area variability is observed in the considered expenditure survey data particularly among the urban small areas. This variability is captured via the proposed correction to the standard ELL methodology to estimate the MSEs of poverty estimates. The proposed methodology is also applicable to MSE estimation of ELL-type estimates of small area linear and non-linear parameters.