An evaluation of synthetic household populations for census collection districts created using optimisation techniques

Abstract

Regional policy makers rely on the availability of detailed and current small area data to inform their decision making. The main source of small area socio-demographic data in Australia is the five yearly Census of Population and Housing conducted by the Australian Bureau of Statistics (ABS). Although the Census provides a comprehensive coverage of Australian households, only a relatively limited range of information is collected about these households. Other data sources, such as the ABS Household Expenditure Survey (HES), provide a rich source of household information but are not available for small geographic areas. One solution to this lack of detailed small area data is to combine the information-rich survey data with the geographically disaggregated Census data using synthetic estimation. This paper reports on some initial validation of recent research that uses optimisation techniques to create a set of synthetic households that seek to represent the socio-demographic profiles of each Census Collection District (CCD) in Australia. The ABS Household Expenditure Survey Confidentialised Unit Record File is reweighted for each CCD to match selected variables in the Census Basic Community Profile (BCP). As an initial measure of the success of this technique, the weights generated are used to recreate the socio-demographic profiles of CCDs in the Australian Capital Territory. The socio-demographic profiles created by synthetic household populations are compared with those from the Census BCP to evaluate the degree to which the synthetic households represent the population within each CCD. This research demonstrates the potential for using optimisation techniques to create synthetic data that can be used for small area policy and household level analysis.