The projected demographic impact of COVID-19 in Australia

I've been meaning to post about the demographic impact of COVID-19 for some time, based on some work I've been doing for local councils in New Zealand. However, before I get to my work (perhaps in my next post), I thought I should post about this new working paper by Elin Charles-Edwards (University of Queensland), Tom Wilson (University of Melbourne), Aude Bernard and Pia Wohland (both University of Queensland), that presents some scenarios on the impact of COVID-19 on the Australian population (see also this article in The Conversation about the research).

Projecting the future population in a sensible way requires understanding that there are essentially only three things that happen to the population: (1) people are born; (2) people move from place to place; and (3) people die. So, if you can project fertility, migration (international and internal), and mortality, then you can project the population - this is the basis of the cohort-component method, which involves projecting each component of population change.

The biggest problem with estimating the demographic impact of COVID-19 is uncertainty - we don't know what the impact will be on any of the components of population change. New Zealand has been lucky - it appears the impact on mortality is minimal if anything. Australia has been similar (they have more COVID-19 deaths, but also have a larger population, and deaths in Australia are still not high compared to other countries). Despite media articles suggesting that there could be a 'COVID baby boom', that seems unlikely. In fact, fertility tends to decrease in times of economic recession. However, the biggest uncertainty comes from migration. How long international borders will remain closed is a big unknown, reduced international migration as a result of COVID-19 will have the largest impact on the future population.

Charles-Edwards et al. tackle this uncertainty by presenting three scenarios for the future population of Australia:

1. A light scenario, which assumes a 28% decrease in net international migration for 2020/21, with a quick return to normal after one year;
2. A medium scenario, which assumes a nearly two-thirds decrease in net international migration for 2020/21, with a gradual return to normal after four years; and 
3. A severe scenario, which assumes zero net international migration for 2020/21, and a gradual return to normal only after eight years.

Their scenarios also include some differences in fertility and net internal (inter-State) migration (but no differences in mortality). They also run a business-as-usual No pandemic scenario. They find that:

The reference No pandemic scenario has Australia’s population reaching 27.6 million by 2025, 29.5 million by 2030 and 33.2 million by 2040... Based on the modelled scenarios, COVID-19 is expected to have a measurable and persistent impact on Australia’s population. Under the Severe scenario, Australia’s population will reach 26.6 million by 2025, 29 million by 2030 and 31.8 million by 2040; 1.4 million or four per cent fewer than the No pandemic scenario. The impact is less under the Light and Moderate scenarios, with Australia’s population reaching 0.18 million fewer and 0.50 million fewer by 2040 respectively... The impacts of COVID-19 are felt most strongly in the short term with annual population growth dropping [from] 1.40 per cent in 2020-21 in the no pandemic scenario to 1.14 per cent in the Light scenario, 0.78 per cent in the Moderate scenario and just 0.41 per cent in the Severe scenario. For historical context, Australia’s annual population growth last dropped below 0.78 per cent in 1935, and last dropped below 0.41 per cent in 1916...

Those impacts are quite large (and much larger in relative terms than the impacts I have estimated for some districts in New Zealand, as I will discuss in my next post). At the State level, they find that:

In relative terms, the largest impact out to 2040, based on the severe scenario, will be in Victoria followed by Western Australia, both of which will have a population 5 per cent smaller than in the no pandemic scenario.

However, as I noted above, there is a huge amount of uncertainty here. The international migration scenarios in the Charles-Edwards et al. research are based on a survey of just six demographic experts. It's hard to see how you could do much better than this though. As they point out in the conclusion to their working paper, identifying leading indicators of demographic change (or even contemporary indicators of change) is incredibly difficult. Yet, that is what you would need in order to have demographic projections that adjust dynamically to real-time events. However, as I will demonstrate in my next post, it would be easy to overstate the impact of these events, especially when you look at them alongside the impacts of large events in the past. I think that is a real risk with this Australian research.