This paper investigates methods for producing Australian intercensal population estimates.
Since the 1981 Census, the ABS has been producing estimates of estimated resident population (ERP) on Census year, and updates these with vital statistics for the intercensal periods. Estimates of Aboriginal and Torres Strait Islander ERP on each Census year have also been produced since the 1991 Census. However, the lack of sufficiently reliable births, deaths and migration data for Aboriginal and Torres Strait Islander people, and the intercensal volatility in the Census counts of Aboriginal and Torres Strait Islander people, do not make it possible to use the so-called standard approach to make intercensal estimates. Instead, ABS has produced population projections which have been widely used as a proxy for population estimates for various purposes. For some important users, however, this proxy is not sufficient or appropriate and therefore there has been significant demand from both government agencies and the community for intercensal population estimates. This paper investigates methods for producing intercensal population estimates. Due to the complexity of this task, the paper focuses on methods for estimates at the Australian level only. After highlighting key issues in data quality, and making an assessment of a series of historical ABS projections, the paper examines three key factors driving the growth of Aboriginal and Torres Strait Islander population and which should form part of any estimation method: (1) fertility levels and trends; (2) mixed parentage and births to non-Indigenous mothers; and (3) the extent of changing identification for Indigenous status on Census form. The paper concludes that the identification change plays a very important role and finds promise in using births registered in the last intercensal period to project identification change in the next Census. Building from these analytical results, the paper proposes an enhanced demographic balancing equation method for estimating intercensal population, and an iterative approach for estimating the intercensal population. The limitations of the proposed method and future directions are also discussed.