How an excess of more than 8,000 deaths can turn into ‘3,000 fewer than expected’
Australia had 3,200 fewer deaths than expected in 2024 and 3,800 fewer in 2025, according to a new excess mortality report 1 report from the Australian Bureau of Statistics (ABS). The Bureau’s previous release 2 had shown 8,700 more deaths than expected in 2023. The apparent turnaround is not evidence that Australia’s excess deaths problem has resolved, rather, it is caused by a redefinition of ‘expected’ deaths.
Excess mortality is calculated by subtracting expected deaths from actual deaths. Actual deaths are counted, but expected deaths are projected from deaths in a historical period (known as the ‘baseline’). Change the baseline or projection method, and the result can change dramatically.
The ABS previously used a ‘Serfling model’ based on the seven years before the 2020 pandemic, in other words, based on death rates in the ‘old normal’ pre-covid world. Despite criticism 3 from independent analysts like Dr Andrew Madry that it underestimated excess deaths, the model 2 still showed an 11.7% peak in 2022, reducing to a still historically high 5.1% excess deaths in 2023.
Normally, when a respiratory epidemic passes unhindered through a population, the spike in deaths is followed in the next year by a deficit of deaths, because frail (mostly elderly) people have died a year early. Surprisingly, despite the 2022 peak being the highest since World War II 4, no mortality deficit appeared in 2023. That’s concerning: it could mean that the bulk of the excess mortality was due to something other than deaths from the respiratory infection itself (covid).
It is thus important to know whether a mortality deficit occurred in 2024 or 2025. The Burnett Institute reported 5 7,600 excess deaths for the first 11 months of 2024 adding that the level was similar to 2023. We have waited in vain for more than two years for the ABS to publish its 2024–25 Serfling results.
Instead, the Bureau now presents a new ‘Generalised Additive Model’ (GAM) using a rolling seven-year baseline that includes covid years (except 2022). Because mortality was elevated during this period, the model projects more expected deaths. A higher expectation makes continuing elevated mortality look smaller and can even produce a deficit of deaths. This is exactly what has happened.
The ABS says the change is justified because the pandemic was officially declared to have ended in May 2023. Thus, we no longer need to compare deaths with pre-pandemic levels, the Bureau argues. But where is the mortality deficit that normally marks the end of a respiratory epidemic?
The ABS appears to have accepted a ‘new normal’ in which covid is “a persistent contributor to mortality”. Enshrining high death rate years in the GAM inevitably extinguishes the useful excess mortality numbers (relative to pre-pandemic levels) and removes the impetus for the Department of Health, Disability and Aging to investigate the actual ongoing excess.
What numbers would the ABS get if they ran the Serfling model for 2024 and 2025? Independent analyst Colin Millen estimates 8,800 excess deaths in 2024 and 7,600 in 2025, using the Serfling baseline. His 2024 estimate broadly aligns with the Burnett’s figure. Far from showing an epidemic-signature deficit, the covid era appears to have a long tail of excess deaths.
Getting the official ABS Serfling numbers would, I believe, confirm that Australia is experiencing a highly unusual ‘long-tailed pandemic’ – which may not be a respiratory pandemic at all. If so, all hypotheses for factors capable of causing long-term harm need to be rigorously examined.
Of course one hypothesis must be the one believed by the ABS: that after covid infection (even mild infection?) people become damaged and more likely to die. Another hypothesis that must be investigated is that mass immunisation with novel genetic vaccines (that had no long-term safety data when approved) damages people and makes them more likely to die. There will be other hypotheses too.
The ABS’s new model is not a tool for getting to the truth: unexplained excess deaths occur, high-mortality years are added to the baseline, expected deaths rise, and a deaths deficit is reported. As we move forward with the rolling baseline, the unexplained episode will gradually disappear from the statistics. Changing the baseline does not explain the deaths. It redefines them as normal. It extinguishes the very evidence of excess deaths.
In the interests of transparency, and to increase understanding with the hope of preventing further loss of life, the ABS should publish the Serfling results for 2024 and 2025 without delay.
P.S. Thank you to Philippa Farmer for drawing the rabbit in a magician’s hat.
1. ABS release on Excess Mortality 3rd July 2026. https://www.abs.gov.au/statistics/health/causes-death/excess-mortality/latest-release#weekly-all-cause-mortality-australia
2. ABS release on Excess Mortality to December 2023 https://www.abs.gov.au/articles/measuring-australias-excess-mortality-during-covid-19-pandemic-until-december-2023
3. Quadrant article on Excess Deaths and mRNA vaccines https://quadrant.org.au/magazine/health/excess-deaths-and-mrna-vaccines/ and Madry, A. (2025) ‘The Statistician’, in R.Clancy and M McCann (eds) Covid through our eyes. Halstead Press, p134.
4. Sydney Morning Herald article on 2022 excess mortality https://www.smh.com.au/politics/federal/almost-20-000-more-people-died-in-australia-last-year-than-anticipated-20230303-p5cp64.html
5. Burnet Institute report. https://www.burnet.edu.au/media/myjhpzax/mortality-during-the-covid-19-pandemic-in-australia_20250311.pdf
