Updating small area estimates during intercensal years using geospatial data and SPREE-type methods
Conference
65th ISI World Statistics Congress 2025
Format: IPS Abstract - WSC 2025
Keywords: geospatial, poverty, small area estimation, spree
Session: IPS 961 - Use of Geospatial Methods for Small Area Population Estimation
Tuesday 7 October 10:50 a.m. - 12:30 p.m. (Europe/Amsterdam)
Abstract
The estimation of poverty measures in small areas is typically implemented using population census data. However, lack of access to updated census data is a common issue in many countries, particularly in the global south. In addition, reliable estimation during the intercensal period, especially as we move from the census date, remains an issue. One approach to avoiding the need to access census microdata is to use auxiliary information from alternative data sources. Using geospatial data in model-based estimation has gained renewed interest due to the frequency of data availability and advances in data processing. Geospatial data have been used in countries with infrequent census data collection to map small-area poverty, and despite only serving as proxies for household characteristics, the results are, so far, promising. An alternative approach to address the possible lack of census data and estimation in the intercensal period is to use Structural Preserving Estimation (SPREE) methods and extensions.
In collaborative work with the World Bank in Mozambique, we produce poverty estimates for districts in the country using a ‘gold standard’ method with the most recent, 2017, census micro-data and an empirical best predictor. We further estimate poverty measures using the older 2007 census and geospatial data as two alternative sources of auxiliary information and different model specifications. Estimates from the various methods are compared to estimates obtained by applying SPREE-type methods, to design-unbiased direct estimates, and are assessed against estimates using the ‘gold standard’ method .
The application in Mozambique highlights the common challenge that analysts face when they lack access to updated census data. Our findings are not relevant only for countries in the global south. Developed economies can benefit from this research too. In collaboration with the UK Office for National Statistics, we transfer our experiences from working with data in Mozambique’s to the UK context. We estimate income deprivation for middle super output areas and local authority districts by analysing data from the UK Family Resources Survey (FRS) alongside geospatial data. Through a critical assessment of the results from the two case studies, we gain valuable insights about the advantages of utilising alternative data sources to produce official statistics.