A proof-of-concept model of Australia’s overseas migration system: projecting temporary and permanent populations, visa switching, and visa-specific migration flows

Forecasting net overseas migration (NOM) is very challenging, as demonstrated by the fact that most NOM forecasts in Australia prove wide of the mark not just in the long-run but even within the first one or two years of the forecast horizon. The forecasting challenges include: (1) a complex migration system in which the Australian Government sets the number of permanent residence visas it will make available each year and makes regular changes to migration rules and processes, (2) NOM data which undergoes non-trivial revisions between preliminary and revised versions, (3) hard-to-predict global economic and political trends which affect international migration patterns, and (4) limited understanding and data on moves between visa and citizenship statuses within Australia. This paper attempts to address the last of these challenges. It presents a proof-of-concept model of the Australian migration system, the first to incorporate immigration and emigration flows by visa/citizenship category and, importantly, shifts between these categories within Australia. We used data from ABS overseas migration statistics, Estimated Resident Populations, the Australian Census of Population and Housing, and the Person-Level Integrated Data Asset (PLIDA). The population groups consist of Australian citizens, permanent residents, international student temporary visa holders, other temporary visa holders, and New Zealand citizens. We then created a population projection model to produce illustrative population scenarios for the future, including (1) retention of current migration policy settings, (2) a smaller Migration Program and a cap on international student immigration, and (3) a hypothetical global pandemic. The model permits policy settings to be included in projection assumptions, incorporates several feedback effects, and generates non-linear projections of NOM in many cases. The main limitation currently is the quality of some of the input data, but our proof-of-concept model has important implications for forecasting and policy-based scenario analysis in other high-immigration countries.