A0339
Title: On the use of small area estimation with geospatial data
Authors: Luciano Perfetti Villa - University of Southampton (United Kingdom) [presenting]
Nikos Tzavidis - University of Southampton (United Kingdom)
Angela Luna Hernandez - University of Southampton (United Kingdom)
Abstract: Small-area estimation methods are used for poverty mapping, most commonly with the aid of census data. In the most developed countries, censuses are updated around every ten years but much less frequently than in many countries in the Global South. Geospatial data provide an alternative data source for use in small-area models in off-census years and can improve the frequency of estimates. However, how geospatial data is processed and used in model-based small-area estimation requires careful attention. The purpose is to study theoretically and empirically the properties of small area estimators based on models with geospatial zonal statistics as predictors. The estimators are used to estimate headcount poverty rates for districts in Mozambique. Estimates using geospatial data are compared against estimates produced with the most recent 2017 census (industry standard estimates) and the old 2007 census in Mozambique. The geospatial-based estimates track the industry standard estimates well, but this is not the case for the estimates based on the 2007 census data. The application in Mozambique illustrates the importance of model building/selection when using geospatial data and the potential pitfalls when using old census data.
