Rather than write a daily post on vote updates, I’ve been putting my effort into ensuring results are as accurate as possible on the ABC Queensland Election Results site.
While the ECQ conducted indicative preference counts for all votes counted on the weekend, it has ceased undertaking further preference counts and removed the weekend’s totals from its website.
Before the data was removed, I captured all the available preference flow data and used it to re-build the preference formulas in the ABC’s election computer. The ABC results site is now applying the weekend’s preference flow data to the latest first preference totals being released by the ECQ.
In this post I will provide some analysis of these flows compared to 2020, and also look at the dramatic increase in swing that occurred on election night as pre-poll centre counts began to arrive.
As at the close of counting on Monday, the swing against Labor on polling day votes was 4.9%, but 8.6% for Pre-polls and 10.3% on the initial count of postal votes. That explains why the swing rose through the evening.
There’s more on vote by type and preference flows inside the post.
Vote by Type
In a post I published before the election, I pointed out that the result on election night would not be clear until we started to receive results from Early Vote Centres. Whatever the split of votes between Labor and the LNP early in the evening, it would shift if the swing on Early votes was significantly different to polling day votes. Given the LNP’s lead in opinion polls had narrowed through the campaign, it was entirely possible that Labor would perform less well with Early votes compared to election day, which meant any swing against Labor would increase as the night wore on.
And so it proved to be. Early on election night it appeared the swing was in the range 4-5%, but once pre-poll votes arrived the swing rose above 7%.
The table below breaks down two-party preferred vote by type in 2024 compared to 2020. The totals for Early votes include only within district Early Votes, and the Postal Vote total includes only votes received by and counted on Saturday night.
| 2020 Two-Party Vote | 2024 Two-Party Vote | ||||
|---|---|---|---|---|---|
| Party | Labor % | LNP % | Labor % | LNP % | Swing |
| Polling Day | 55.5 | 44.5 | 50.6 | 49.4 | -4.9 |
| Early Votes | 52.4 | 47.6 | 43.8 | 56.2 | -8.6 |
| Postal | 51.3 | 48.7 | 41.0 | 59.0 | -10.3 |
As I have done for three decades, on election night I was analysing results based on swing rather than two-party preferred vote. If you go back to the original work on Australian election prediction by Dr Ross Cunningham, the reason for using 2-party preferred % swing rather than two-party preferred % vote was because the variance of swing was smaller than the variance of 2PP percent. Unlike 2PP%, swing was also not correlated with booth size.
That swing did not correlate with booth size meant you could use swing to produce an unbiased estimator of the final result, and that estimator would also have smaller variance than you get if you used 2PP%.
The problem emerging with this model is that if pre-poll votes behave differently to on the day votes, then the assumption there is no correlation between booth size and swing is breached.
Early votes come from giant count centres that can be ten time the size of an average polling place. They report late in the evening, and their size means they have greater impact on calculation of swing than the average polling place. So having measured a consistent swing for polling place votes through the evening, different behaviour with early votes shifted the predicted swing value. The early evening swing calculation become a biased estimator of the final swing because swing was correlated with booth size.
The problem for anyone like me trying to analyse on election night is you can’t know how different early vote swing is until an early vote centre reports. You can have a consistent swing over a dozen polling places and then one giant pre-poll centre arrives and shifts the value of the swing. And in many electorates at this year’s Queensland election, there was only a single early voting centre.
When I started covering election in the early 1990s there would be only 10-15% of votes cast as pre-poll, postal and absent. You left some wriggle room in your projection at the end of election night to account for a post-election night shift, almost always against Labor. Now these votes make up a majority of the count, and their arrival late in the evening substantially changes the count, as we saw on Saturday.
We have always known that postal votes favour the conservative side of politics, and that absent votes favour Labor. But there is not a consistent pattern with early votes. Early votes range from little difference (NSW 2019) to a significant difference (NSW 2023) to a massive difference (QLD 2024). What this difference is matters more as the proportion of pre-polls grows.
Which explains why all parties and broadcasters were cautious in calling the election result on Saturday. On the day votes pointed to a swing that would produce a hung parliament, but the larger swing that produced an LNP majority government only emerged once Early votes arrived. The 3.1% Labor gap between polling day and early votes in 2020 ballooned to 6.8% in 2024, the election day swing only 4.9% compared to 8.6% for early votes.
There is no guarantee that the same gap in swing would open in 2028, but how are you to judge there is a different gap until the first early vote counts arrive. In the end all you can do is be cautious until you actually see early vote centres report.
Preference Flows
The table below summarise preference flows of minor parties at the 2024 election compared to 2020. For the Greens and One Nation I have included both a total flow and average flow per electorate to account for the higher flow that often occurs in seats where a party gets its highest vote. The table excludes districts that did not finish as two-party preferred contests.
| 2020 Preferences to | 2024 Preferences to | |||
|---|---|---|---|---|
| Party | Labor % | LNP % | Labor % | LNP % |
| Greens – total flow | 80.1 | 19.9 | 80.7 | 19.3 |
| Greens – average flow | 77.4 | 22.6 | 78.1 | 19.3 |
| One Nation – total flow | 33.8 | 66.2 | 26.2 | 73.8 |
| One Nation – average flow | 33.0 | 67.0 | 26.4 | 73.6 |
| Family First | .. | .. | 37.7 | 62.6 |
| Katter’s Australian Party | 42.1 | 57.9 | 31.6 | 68.4 |
| Legalise Cannabis | 58.2 | 41.8 | 58.7 | 41.3 |
| Animal Justice | 61.8 | 38.2 | 56.4 | 43.6 |
The main point to note in the above table is that One Nation flows were consistently higher to the LNP than in 2020, and the LNP also benefited from the return of Family First.
From 1998 to 2016 One Nation preferences tended to be a variable feast. It is only since the 2017 Queensland and 2019 Federal elections that One Nation preferences have begun to consistently favour the LNP.
Katter’s Australian Party preferences also favoured the LNP more in 2024, mainly due to the party moving from open tickets to recommending preferences for the LNP in the three Townsville seats. In Mundingburra KAP preferences to the LNP increased from 57.0% in 2020 to 73.5% in 2024, in Thuringowa from 50.7% to 73.3%, and in Townsville from 56.2% to 64.5%.
The flow of LNP preferences in South Brisbane demonstrate the power of major party how-to-votes. In 2020 the LNP recommended preferences for the Greens in South Brisbane and 67.8% of LNP voters followed the recommendation and gave preferences to the Greens. In 2024 the party reversed its recommendation and 76.0% of LNP voters preferenced Labor ahead of the Greens.
A Further Note on South Brisbane
As mentioned earlier, the ABC election site is now calculating estimated preferences using the flow data from election night. This includes for South Brisbane.
The complication with South Brisbane is that who finishes second and third will determine who wins. If as is currently the case, the LNP finishes third, Labor wins easily on LNP preferences. If the LNP overtakes Labor to finish second, the Greens win easily on Labor preferences.
Currently in South Brisbane, LNP preferences flow 76.0% to Labor and One Nation’s 72.4%. In Maiwar, Labor’s preferences are flowing 75.0% to the Greens and a similar flow would be expected if Labor is excluded in South Brisbane.
Which is why the order of exclusion is so important in South Brisbane. There are two factors that could cause the LNP to overtake Labor.
- (1) – the addition of further absent and postal votes boost the Liberal vote.
- (2) – One Nation has 3.3% of the vote and its preferences could push the LNP ahead of Labor.
As I write (5pm Tuesday 29 October with 74.6% counted), Labor is on 32.0% and the LNP 29.8%. The ABC computer is currently projecting first preference vote shares of Labor 32.2% to LNP 31.4%, close enough for One Nation preferences to overturn Labor’s lead. It is important to note these projected votes shares are based on incomplete Postal, Absent and Absent Early totals so they may bounce around.
I note there are some cries for a three-way preference count to be conducted to resolve who goes out first, but this can only deal with the One Nation preferences. It won’t provide an answer on whether votes yet to be counted will change the vote shares. The ECQ won’t be conducting a three-way preference throw but scrutineers have been free to calculate their own.
The information I have been given is that One Nation preferences are splitting 65% LNP, 20% Labor and 15% Greens.
I have built the above One Nation split into the ABC’s election computer as the first step of calculating preferences. The computer will then decided which of Labor or the LNP is in third place and apply the relevant preferences on exclusion.
This could result in a flip-flopping between the Greens and Labor being ahead as the count progresses. The computer has been locked to leave the seat as in doubt with whoever is ahead as leading as each vote update is received.
I am bemused by the 15% preference flow from One Nation to the Greens. i would have thought ON and GRs have zero in common and indeed ON voters would feel nothing but antipathy towards the GRs.
COMMENT: There are always some voters who will vote for every other candidate on the ballot paper before reaching one of the major parties.
Antony, do we see a variance in flows depending on ticket placement?
eg if there were 3 candidates and the middle got less votes do we see most voters naturally vote down the ticket before returning to the top?
so do they vote 3,1,2 or 2,1,3 and would be interesting to note if this impacted the preference flows you have written about above.
COMMENT: I can’t say I’m aware of any research on the subject. The only obvious impact of ballot paper order is the donkey vote, numbering straight down the ballot paper, and this also can be seen on long ballot papers when people have to complete a sequence beyond the candidates they choose to vote for.
Is Maiwar actually assured as Retain for the Greens? Currently 75% of the vote is counted. Let’s say in time formal votes will be received for 90% of the seat. If say 58% of those as yet uncounted votes go to the LNP on a two-party preferred basis, I think that gives LNP the edge. Is that in the realms of possibility? For reference you show 59% of postal votes going to LNP(2 party vs ALP) state-wide.
COMMENT: There has been an error in the Liberal total corrected by the check count and Maiwar is safely won by the Greens.
Be interested to see your thoughts on both Mulgrave and Mirani.
In Mulgrave is it still possible for the KAP to bump up to second based on preferences from those below them in the totals?
In Mirani – I see the lead KAP had in 2pp is shrinking – is there a trend in postals? Based on your data – do you think it’s likely to close further and even for the LNP to jump ahead?
COMMENT: The LNP have won Mirani. Andrews did very badly with Absent votes, most of which would have been cast in Mackay and Rockhampton where there was no KAP candidate to assist with pre-poll campaigning.
I have the suspicion the LNP will win Mulgrave but have been unable to get any scrutineer information on whether the KAP candidate can catch Labor in the race for second.
antony what is the two party preffered vote in araluen, johnston, goyder, blain, katherine, nelson and etc in 2024 nt election
COMMENT: I’m not aware if one exists. I must ask the Electoral Commission if they undertook a separate 2PP count as they have done in the past. If they haven’t then everyone is free to calculate their own.
If they do win South Brisbane it’d twice this past two weeks that the Greens have made an unlikely comeback. It’s nice to see them benefiting from the electoral system and preferences at least sometimes
Hi Antony,
Thanks for your commentary.
Regarding South Brisbane, do you have the 3 way split of PHON votes?
It looks like 72.4% flows to LNP, but is this on a 2CP basis rather than 3CP?
COMMENT: All last week I was using the split 65% LNP, 20% ALP, 15% GRN. It has proved to be slightly higher for the LNP but still not enough to close the gap between Labor and the LNP.
Is South Brisbane evidence of ‘centre squeeze’? A problem with our version of preferential voting. Here the problem is that the more voters turnout for LNP it makes things worse for them. i.e. Greens get elected instead of Labor (Labor being LNP’s preferred winner in that head to head). If LNP voters didn’t vote then they’d get a better outcome. Their voting has a negative effect
I think OPV would be a better option because for Labor v Greens contests, Liberal/LNP voters can simply choose to exhaust their preferences if they have no desire to see either party win.
At least, Labor and Greens would be required to compete head on in a proper two-party matchup rather than rely on a favourable preference recommendation from the Liberal Party/LNP.
I really think pre-polling should be restricted. I know the horse has bolted on this issue, but there were definitely a difference between prepoll voting vs “on the day” voting outcomes.
Political parties are know this and are taking advantage. It’s also pretty obvious that fixed terms are making parties lazy until the last few months of a term.
COMMENT: Shortening the period of pre-polling seems to be regularly suggested, but if it just leads to an increase in postal voting it is a bad outcome. Postal voting is more expensive than pre-polling, sees more ballot papers rejected, is a problem with the decline on postal service frequency, and is the form of voting where political parties have most inserted themselves between the Electoral Commission and voters. If the pre-poll voting period is to be shortened further, there should also restrictions on parties trawling for postal votes and collecting postal vote applications.
All postal vote applications should be returnable to the Electoral Commission, not to parties and candidates. Anyone wanting a postal vote should always apply through the Electoral Commission’s application portal as this speeds up the dispatch of the postal vote pack.
Barron River, yet again the Barron-meter!
In 2020, state-wide 2PP 53.2% Labor, Barron River 2PP 53.1% Labor.
In 2024, state-wide 2PP 53.9% LNP, Barron River 2PP 54.0% LNP.
The standard has been raised for bellwether seats!
There are two things that could cause a difference between early votes & election day votes:
a) Different types of people choosing to vote early or on the day. I get the impression that, on the whole, early voters vote more to the Right than voters on the day. This could be checked through the last 4 years of Federal & State elections.
b) People changing their minds during the campaign. This can be seen from opinion polls published as the campaign progresses. Big swings during a campaign are rare, but not unknown. One certainly happened in Qld this time.
COMMENT: We know that pre-polls are more conservative leaning than on the day votes. But the percentage they lean by varies from election to election. As pre-polls make up a huge proportion of the vote, and they come in late, a difference in their variation from on the day votes is what makes it hard to make predictions as early as we used to. Early on the night in Queensland, based on the pre-poll to on the day gap in 2020, the LNP were only going to win narrowly, but the gap was much larger and the swing grew and the LNP won comfortably.
At both elections the pre-poll vote was more conservative than on the day votes, but the gap was bigger in in 2024. Or to put it another way, the swing against Labor was bigger with pre-polls the it was with on the day votes. Until the first pre-polls arrive, you can’t really asses the difference. On election night 2024, it was immediately obvious in our state-wide projections that bigger swinging pre-poll centres were starting to report.
Hi Antony, In ref to the difference between early and on day voting is also explained by the degree in which peoples minds are already made up and set. I know from personal experience in my voting life when I have a strong enthusiasm for one party or change I cannot wait to get out and vote. No last minute Champaign will change it, the last 4 years was enough! I think we are seeing evidence of this determination and enthusiasm in USA right now.
Antony, is it possible to check the preference flows in the seat of Redcliffe for the 2024 election. Where did the Green preferences actually flow to?
COMMENT: The flows from election night being used by the ABC are Greens 79.9% to ALP, 20.1% to LNP, One Nation 30.0%/70.0%, Independent 48.6%/51.4%.
Is there a preference tendency for independents generally per se? i.e. do they tend to be closer to 50-50 or eventually favour the LNP or Labor? (from my (own experience I’d doubt they’d favour Labor).
COMMENT: It depends entirely on what the Independent recommends with their how-to-vote material, if they choose to recommend preferences and whether they bother to campaign or distribute how-to-vote material.
Is there a place where I can go to understand what the different vote types are on the ECQ website? What are Returning Officer votes, what’s the difference between absent-early and absent-election day votes, and so on? I’ve searched ECQ and can find no description of what all these mean.
COMMENT: Returning Officer votes are an early vote centre, but one conducted on the premises of a Returning Officer rather than as a separate declared Early Vote Centre. Not all Returning Office offices took early votes. Absent Early votes are early votes taken at Early Vote Centres outside of the voter’s home district. Absent Election Day votes are votes taken on election day at polling places outside of a voter’s home district.
Note that the Brisbane CBD and Brisbane CBD Early centres are outside the boundaries of 92 districts but are declared as centres able to take ordinary rather than declaration votes for each district.
Antony.
Your comparison please re informal votes this year compared to 2020.
Des
COMMENT: Informal voting was up from 3.4% to 3.9%. I can’t offer any sensible comment on why. The ECQ always undertakes and publish research on informal voting after the election. The research examines informal votes and categorises them by type. The two largest categories, as is always the case, will be blank ballots and ballots with numbering errors.