Final Wrap of the Dunkley By-election

The result of the Dunkley by-election was declared last month, but as usual it has taken a couple of weeks for the the Australian Electoral Commission (AEC) to finalise and release its full suite of election statistics on the by-election.

This post wraps up the final figures, looking at turnout, preference flows by candidate and result by vote type.

In short, the by-election saw a normal change in turnout, the pattern of vote by type was normal, and the pattern of preferences flows was normal. The by-election produced an overall swing of 3.5% which is on par with an average by-election swing.

Turnout

Before the Dunkley by-election there were several stories mentioning concerns that there would be a low turnout After the election I was asked numerous questions on twitter on whether the turnout in Dunkley was unusually low.

The turnout in Dunkley was not low. It was lower than at the last general election, but by-election turnout is always down on the previous election.

The table below shows turnout at all by-elections since 2014 compared to preceding election turnout. As all electoral divisions have a characteristic pattern of high or low turnout, the relevant statistic to examine is change in turnout, the figure shown in the final column.

Turnout at Federal By-elections 2014-2024
Turnout
Division Year PM (Party) Prev Elec By-elec Change
Griffith Feb 2014 Abbott (LIB) 93.1 82.1 -11.0
Canning Sep 2015 Turnbull (LIB) 92.0 79.5 -12.5
North Sydney (No ALP) Dec 2015 Turnbull (LIB) 92.3 78.4 -13.9
New England Dec 2017 Turnbull (LIB) 93.4 87.1 -6.3
Bennelong Dec 2017 Turnbull (LIB) 91.7 86.0 -5.7
Batman (No LIB) Mar 2018 Turnbull (LIB) 89.7 81.5 -8.2
Longman Jul 2018 Turnbull (LIB) 91.7 84.3 -7.4
Braddon Jul 2018 Turnbull (LIB) 94.1 90.4 -3.7
Mayo Jul 2018 Turnbull (LIB) 92.5 85.5 -7.0
Fremantle (no LIB) Jul 2018 Turnbull (LIB) 88.8 66.1 -22.7
Perth (no LIB) Jul 2018 Turnbull (LIB) 88.0 64.1 -23.9
Wentworth Oct 2018 Morrison (LIB) 86.2 78.1 -8.1
Eden-Monaro Jul 2020 Morrison (LIB) 93.3 89.1 -4.2
Groom Nov 2020 Morrison (LIB) 93.1 81.7 -11.4
Aston Apr 2023 Albanese (ALP) 92.5 85.6 -6.9
Fadden Jul 2023 Albanese (ALP) 86.5 72.5 -14.0
Dunkley Mar 2024 Albanese (ALP) 90.1 83.8 -6.3

Across all 17 by-elections, the average decline in turnout was 10.2%. If you exclude the four by-elections that were skipped by one of the major parties (Batman, Fremantle, North Sydney, Perth), the average falls to 8.0%.

In Dunkley the decline in turnout was 6.3%, lower than the drop at the other two by-elections in this term, Aston and Fadden. In short there was nothing unusual about the turnout at the Dunkley by-election.

Vote by Type

There was nothing exceptional in the results based on the type of vote cast. The proportion of votes cast on polling day rose from 43.8% to 49.2%, perhaps explained by the absence of absent voting at by-elections. In Dunkley at the 2022 election, Absent voting had been 3.3% of all votes and Pre-poll Declaration Votes, essentially absent pre-polls, had been 3.6%.

There was a slight rise in pre-poll ordinary votes from 30.8% to 31.5%, and a slight rise in Postals from 18.3% to 18.9%. Lower turnout will explain some of these shifts but there is nothing unusual in the vote by type figures.

As is usually the case, Labor polled better on polling day than than with pre-poll votes. Once all postal votes were counted, Labor even recorded a narrow majority amongst postal voters. The chart below shows the two-party preferred result by vote type.

The swing by vote type was 2.9% against Labor on polling day, 4.4% with pre-poll and 3.8% on postal votes for a final 3.6% swing.

Looking at how the matched polling place swing behaved, on election night with all polling places having reported the projected Labor two-party preferred vote was 53.4%. At the end of election night with pre-polls added the prediction was 52.8% and the final result was 52.7%.

Preference Flows

The chart below is the final preference flow data by candidate for the by-election. Labor led on first preferences so the distribution of preferences merely confirmed the first preference winner.

Overall preferences flowed 59.1% to Labor compared to 58.8% in Dunkley at the 2022 election. Vote for third parties declined from 27.2% in 2022 to 24.5% at the by-election.

Greens preferences were 84.3% at the by-election compared to 89.2% in 2022. Libertarian preferences flowed 73.5% to the Liberal Party compared to 76.6% at the 2022 election.

In 2022 One Nation preferences flowed 63.1 to the Liberal Party and UAP preferences 70.0%. Neither party contested the by-election, the parties polling 7.9% between them in 2022 and their absence for the by-election played a part in the 6.7% rise in Liberal primary vote.

The absence of One Nation and UAP candidates was made up as a preference source by Independent Darren Bergwerf. In 2022 he polled 3.9% and his preferences split evenly. At the by-election he polled 5.9% and his preferences flowed 66.0% to the Liberal Party. In 2022 there was a donkey vote factor in Labor’s favour with Bergwerf’s vote but not at the by-election.

Donkey Vote

The term ‘donkey vote’ often gets confused with an informal vote, but a donkey vote is not informal.

A ‘classic’ donkey vote is a formal vote that is completed by numbering the ballot paper top to bottom. In some cases this is a deliberate vote by the voter. In the Dunkley by-election the Liberal candidate Nathan Conroy was top of the ballot paper. A Liberal voter voting 1 Conroy then numbering straight down the ballot paper was making a perfectly rational choice. Liberal preferences were never going to be distributed so there was little point in fussing over who to number 2 to 8.

I highlighted the word ‘classic’ in the above paragraph as there is another type of donkey vote which can be important on large ballot papers. That is a preferential donkey vote where a voter who gave a first preference to a candidate elsewhere on the ballot paper, and maybe a second or third preference, then revertsing to numbering straight down the ballot paper. This can create a preference drift to a candidate appearing higher on the ballot paper.

In the 1980s, David Peetz published a paper examining the donkey vote based on looking at after preferences swing. He found that if the positional order of the final major party pairing was reversed, there was a consistent shift in swing of about 1% produced simply by the position change. The comparison measured not just the advantage of being top of the ballot paper, but a value that included both classic and preferential donkey voting.

In Dunkley there was a switch in major party positions. At the 2022 election Labor was in spot 2 and the Liberal candidate in slot six. At the by-election the Liberal candidate was top of the ballot paper and the Labor candidate Jodie Belyea 8th and last. As already mentioned, this may help explain the different preference flow from Independent Darren Bergwerf.

This suggests that maybe 1% of the swing at the Dunkley by-election could be down to just the change of positions on the ballot paper. That may be so, but the two-party preferred swing remains at 3.5% and all other discussions of the true swing are estimates as against the figure that determined the result.

And on the issue of two-party preferred swing, some people have dismissed its usefulness. Labor partisans have argued the the Labor primary vote being up 0.8% was a more important measure. Liberal partisans have argued that the Liberal primary vote rise of 6.7% is more important.

In the end, primary vote support for both major parties was impacted by changes in minor party nominations and changes in minor party support. The lack of One Nation and UAP candidates played a party in the rise in Liberal primary vote. At the 2022 election Labor received as preferences around a third of votes cast for One Nation and UAP. Labor’s small primary vote rise may reflect some of those voters switching to Labor, as well as a decline in Greens support.

Arguments over primary vote just ignore the fact that Australian elections are decided by the result after preferences. In a contest at two elections between the same two final parties (this definition is important), what matters is the two-party preferred swing, a single measure of shifts between the two major parties. First preference shifts between the various candidates on the ballot paper, and the addition or absence of candidates since the previous election, end up resolving into a single measure of after preferences swing.

To say there was a 3.5% swing to the Liberal Party in Dunkley is to say that there were 3.5% more voters that listed the Liberal Party with a preference before the Labor candidate. That’s the way you measure swing. Primary vote shifts play a party, but in a contest involving the same final two parties, how many voters listed one party before the other, and what percentage changed since the previous election, is the measure that matters.

Dunkley By-election Polling Place Results


(Click on polling place for results)

Leave a Reply