Adjusting historic election results for boundary changes

2

If you’ve been reading my federal election guide, you would have noticed a new feature: a chart showing the two-party-preferred vote in that seat over recent decades. It features a dotted green line, which is my estimate of how the modern 2022 electoral boundaries voted at every election since 2004.

In this blog post I will explain how to read these charts, and how I produced this data.

I’m going to use a couple of examples to explain the value of this data. Firstly, my old seat of Macarthur.

The chart has three features. The blue line is the two-party-preferred vote in New South Wales. The red line is the two-party-preferred vote in Macarthur as it was then drawn. The dotted green line is my estimate of that vote for Macarthur of 2022. I use Coalition or Labor two-party-preferred vote depending on who currently holds the seat, but they are simply a mirror image of each other.

Macarthur’s electoral boundaries have bounced around a lot over the decades, due to its position on the main pathway for shifting electoral boundaries between Sydney and rural NSW. If Sydney has a half-quota of population left over, the only way to dissipate that is by shifting the electorates in the south-western corridor: Fowler, Werriwa, Macarthur, Hume.

I show a break in the red line wherever the electoral boundaries changed. For example, there was a major change prior to the 2016 election. The Liberal Party had polled 61.4% after preferences in 2013, but the redistribution cut that margin down to just 53.2%. This is due to Macarthur being pulled more into the Labor-friendly suburbs of Campbelltown and away from Camden.

There was another dramatic change favouring the Liberal Party prior to the 1993 election, and another pro-Labor shift prior to the 2001 election, and minor changes prior to 2007 and 2010. The pre-2010 redistribution made the seat into a notional Labor seat, but a swing to the Liberal Party helped them retain the seat.

Since Macarthur has changed shape, it is hard to detect what are real shifts in vote, and what changes are due to redefining what counts as “Macarthur”.

The dotted green line helps resolve that. Since the current boundaries have not changed since 2016, the dotted green line disappears under the red line since that last redistribution, but before that it shows that the current Macarthur would still have been won by the Liberal Party in 2013, but would have been won by Labor in 2007 and 2010, and only narrowly lost in 2004.

If you didn’t know about the redistributions in Macarthur, you could mistakenly think that the seat has shifted much more strongly towards Labor since 2013, even though about 8% of that shift is due to that pre-2016 redistribution. Macarthur does swing a lot – an 11% swing in 2007, a 7% swing in 2013, a 9% swing in 2016.

Two other quick examples. First Macquarie:

Macquarie doesn’t usually experience much redistribution change, being wedged into the north-west of Sydney, covering the Blue Mountains and Hawkesbury regions. But when it was redrawn, it was redrawn dramatically. The pre-2007 redistribution removed the Hawkesbury region and added much of central west NSW, increasing Labor’s two-party-preferred vote by about 9% and flipping Macquarie into a notional Labor seat.

Labor then gained a 6.5% swing on top of that at the 2007 election. The combination of the redistribution and the swing shifted Macquarie from an 8.7% Liberal margin to a 7.0% Labor margin.

The electoral boundaries were redrawn again prior to the 2010 election, and the Macquarie changes were completely reversed, cutting Labor’s margin by 6.9%, and the Liberal Party regained Macquarie in 2010.

This story is made clear in my chart. The red line shoots up to reflect the 2007 version of Macquarie, but the dotted green line perfectly lines up with the 2004 and 2010 versions of Macquarie.

Finally, let’s look at Gilmore, a seat with much less dramatic redistribution effects.

It demonstrates an example of a seat that has genuinely shifted quite a long way. The electorate was redrawn into a notional Labor seat prior to the 2010 election, but it swung back to the Liberal Party, and some of the areas that swung most strongly were removed prior to the 2016 election. Using my green dotted line, you can see that the Labor two-party-preferred vote has shifted from 41.6% in 2004 to 52.6% in 2019.

That will be the topic of my next blog post: when you remove the effects of redistribution, which seats have shifted the most compared to the country as a whole over 15 years of elections.

Two-party-preferred is a simple statistic and can mask a lot of interesting trends on primary votes. I will also be using this data in other ways to look at primary vote trends, but for the purpose of these charts the two-party-preferred vote does a good job of telling the story of who wins, at least in most seats.

Finally, how did I calculate this dataset? Since 2013, the AEC has published data on how many people from each SA1 (the smallest level of ABS analysis) voted at each polling place, or via various other voting methods. I can use this data to calculate estimates of vote for each SA1. Prior to 2013, the AEC provided this data to the parliamentary library, who did a similar analysis.

I have then taken this data from 2004 onwards, and steadily converted the data to match more recent datasets. This has been complex, since the small areas have changed three times over this time period: 2001 CCDs, 2006 CCDs, 2011 SA1s and 2016 SA1s. It does mean there is some lost information, and some loss of precision, but I think the Macquarie chart (and other examples) shows that my estimates come very close to matching the results in the cases of seats whose boundaries are the same or similar to what they were in 2004.

That’s it for now. I’ll be back tomorrow with some more analysis

Liked it? Take a second to support the Tally Room on Patreon!
Become a patron at Patreon!

2 COMMENTS

  1. Ben, do you have a similar report for the federal seat of Chisholm, Victoria. I live in the seat.

Comments are closed.