During the federal election count, I identified a number of issues with the data being published by the AEC. Two main issues emerged, where the AEC’s estimates did not seem to be correct.
I have been publishing a data repository for some years now, and it mostly focuses on publishing state and local election results, which are often published in a form that is not particularly useful for data analysis. The AEC’s data is published in a very clean and consistent format and I’ve never previously had the need to produce my own files. But this time there are a few issues that I thought made it worthwhile to produce my own spreadsheets. They are designed to perfectly match the original AEC files (with the removal of the introductory row) but with correct datapoints.
The two main issues that motivated this work were:
- The AEC’s estimated two-candidate-preferred margins in a series of non-classic seats were quite strange. This means the 2CP swings published by the AEC are also off. I’ve explained how I calculated these corrected swings here.
- The AEC has not conducted a 2PP count for the seat of Bradfield, and I don’t agree with the AEC’s method of estimating that figure. I originally explained my logic here, and then updated the figures when 3CP data was published.
There is a third, less significant issue. In the seats of Bendigo, Brisbane and Nicholls, the AEC has acted as if there is no 2CP history that can be used to calculate a valid swing. This is correct in a seat like Bean, where there is no Labor vs Independent precedent. In seats like those, the AEC publishes the percentages for both candidates, and thus the swings add up to 100 (whereas in most seats they add up to 0). In practice you just need to exclude those swings entirely.
But this doesn’t make sense for those three seats, because in all three cases they are now a classic race between Labor and Coalition. Since 2PP is calculated everywhere, it’s possible to calculate a 2PP swing between 2022 and 2025 in those three seats. In Brisbane and Nicholls, the seat was non-classic in 2022, but the leading Coalition party was the same in both cases and thus the 2022 2PP is perfectly comparable to the 2025 2CP. In the case of Bendigo, it went from being Labor vs Liberal to being Labor vs Nationals. Reasonable people could disagree but I think it’s fair to treat that as a comparable result.
There have been examples in the past of similar issues – Whitlam in 2016 comes to mind – but since I was fixing the other issues I thought I’d fix this one too.
So you can now access this data from my data repository. I will leave these files in the free public section of the repository. Generally the most recent election for each jurisdiction is free for everyone to access, with the full repository available for those who donate $8 or more per month via Patreon.
