Day 19 - 2025 Federal Election

It is unusual for a second term government to outperform its first term 2pp vote share. Today's polls suggest it might be possible for this Labor government.

Day 18 - 2025 Federal Election

Today's Newspoll 52/48 in Labor's favour on the two-party preferred (2pp) metric is the same as the last Newspoll. My Bayesian models assume that if the national voting intention for Labor is 52 per cent, we will see a smattering of 53 and 54 polls, as well as a smattering of 51 and 50 polla. With the exception of Roy Morgan, we are not seeing polls above 52.5. And I start to worry when the polls do not have the variability I would otherwise expect. I am not yet concerned, nor am I alarmed, but I am watching. 

My Bayesian aggregation of Labor's 2pp is 51.1 per cent (based on pollster estimates). It is lower that the most recent polls, because they are clumped together, and it is balancing the absence of polls above the median with the earlier historical polls. For Roy Morgan, I am adjusting its recent polls as a separate series, because the relationship to the trend suddenly changed, suggesting a change in collection practice that impacts on the house-effects part of the model. 

Day 17 - 2025 Federal Election

 Not a lot to report today. The betting market sees it much as it did yesterday.

Day 16 - 2025 Federal Election

Today the individual seats market has Labor ahead (the favourite) in 75 seats.  But the sum of probabilities suggest that Labor most likely will only win 71 seats (after giving Labor and the Coalition one each from the "any other candidate" grouping, which remains overstated even after treating the odds to partially correct for the favourite-longshot bias problem). 

Day 15 - 2025 Federal Election

Today a YouGov poll was released. It had Labor on 52.5 per cent of the two party preferred vote share: a better result for Labor than the 2022 election. This poll is up 1.5 percentage points on the previous YouGov poll from two weeks ago. Of the 10 polls released since the election was called, Labor has been the winner in eight, one was a tie and one had the Coalition ahead. 

Day 14 - 2025 Federal Election

I want to talk about "Others", but first a quick look at the betting market, which now has the probability of a Labor win at 76 per cent. The odds are $4.00 (for the Coalition) and $1.26 (Labor).

Day 13 - 2025 Federal Election

Today, let's look the complete set of charts from the individual seats market. Overall, we can see another net seat with Labor in the lead. The two seats that were tied yesterday (McEwan and Robertson) have bounced to a Labor lead.

Day 12 - 2025 Federal Election

Yesterday a Roy Morgan poll was released that had Labor on 53.5 per cent of the two-party preferred vote. Morgan may have changed its polling approach in February 2025. Historically, their polls have been around the trend of the other polling firms. Now they appear to be consistently above the trend output from other pollsters. Out of an abundance of caution, I have set the post 15 February 2025 polls from Roy Morgan as a new series. Of note, while the earlier Morgan series tracks pretty close to the aggregation, the later Morgan polls track 2.2 percentage points more favourable to Labor on average than the aggregation. Overall, this change moves the aggregation around 0.3 percentage points less favourable to Labor.

Day 11 - 2025 Federal Election

Yesterday brought us a new Newspoll: 52 to 48 in Labor's favour in two party preferred terms. If repeated at an election, this would deliver a similar result to the 2022 election. Adding the latest Newspoll to my Gaussian Random Walk (GRW) Bayesian model to aggregate polls, it looks like Labor leads with around 51 to 49 per cent of the two party preferred vote. If replicated at an election, this is most likely to result in a minority Labor government.

Day 10 - 2025 Federal Election

Twenty-seven days until the election on 3 May. The punters are firming on Labor winning this election.

Day 9 - Federal Election 2025

As we wake up this morning, China has retaliated to the Trump tariffs, the S&P500 is down 17.5 per cent on its February peak (down six per cent overnight) and the Australian dollar is trading at 60 cents. The Trump gloss is starting to peel off and in the process it is hurting the Coalition.

Labor continues to push ahead in the betting market. The odds at 7am this morning (my daily collection time) were Labor on $1.53 and the Coalition on $2.50. As I write now at 8.30am, it is $1.50 and $2.62. The odds have flipped from where they were just three weeks ago. The chart is from 7am this morning.

Day 8 - 2025 Federal Election - Betting Market

What fun we have when the betting market provider introduces multiple independents into seats. I needed a quick recode to capture this situation. The affected seats were Monash in Victoria (Deb Leonard and Russell Broadbent) and Calare in NSW (Andrew Gee and Kate Hook).

More importantly, there was some movement on both the overall winner charts and the individual seat charts. Both movements were in Labor's favour. 

Overall winner

Day 7 - The 2025 Federal Election

Let's look at the steps between the election having been called and the election being all wrapped up.

Date	Event
Parliament prorogued	8.29am 28 March 2025
House of Representatives dissolved	8.30am 28 March 2025
Issue of writs for the election	31 March 2025
Close of electoral rolls	8pm 7 April 2025
Close of nominations	12 noon 10 April 2025
Declaration of nominations	11 April 2025
Start of early voting	22 April 2025
Close of postal voting applications	6pm 30 April 2025
Election day	8am to 6pm 3 May 2025
The return of writs	No later than 9 July 2025
The new Parliament will first sit	Within 30 days of the return of writs

The next critical day on this schedule is to ensure you are enrolled to vote before the electoral rolls close on Monday 7 April 2025. Early voting starts on 22 April. Election day is Saturday 3 May 2025. 

Day 6 - 2025 Federal Election

The betting market has Labor slightly ahead to form government, with odds of $1.72 for Labor to the Coalition's $2.10.  And I should note that I have updated my daily collection time for these odds from lunch time to breakfast. 

Day 5 - 2025 Federal Election

Yesterday's polls were plentiful. Roy Morgan had Labor on 53 per cent of the two party preferred vote (unchanged on its previous print a fortnight ago). Newspoll had Labor at 51 per cent (up 2 on three weeks ago). Resolve Strategic had Labor on 50 per cent (up 5 from mid February) and Freshwater Strategy had Labor on 49 per cent  (unchanged from a fortnight ago).

As a consequence, all of my polling aggregation models have moved further in Labor's favour. But importantly, they suggest that the rate of movement towards Labor may be slowing or even coming to a stop. 

Day 4 - 2025 Federal Election

Today I am looking at the individual seat betting markets. These markets are tricky to analyse because of the favourite-longshot bias problem. To reduce the impact of this problem, I multiply the odds from each seat by its square root. This has the effect of increasing the odds for longshot outcomes, and thus reducing their probability. To calculate probabilities I take the reciprocal of the odds and then adjust for the bookmaker's overround.

From the individual betting markets we can deduce which parties might win the election. For example, we can count the party that is ahead in each seat. This suggests that the punters think the most likely outcome is one in which no party secures 76 of the 150 seats to form government in their own right.

Day 3 - 2025 Federal Election

Around lunchtime every day I take a snapshot of the odds for the winner of the next Federal Election at sportsbet. I use it as a barometer of public intuition for who might win the next election. At the moment it is very close. Today the win probabilities are effectively 50/50 for both Labor and the Coalition (ignoring the long shot outcome of any other party). 

Day 2 - Federal Election 2025

Today I spent some time looking at the published two-party preferred (2pp) opinion poll estimates for each of the Australian states and territories. [Note: I have excluded Tasmania, the Northern Territory and the Australian Capital Territory because there was not enough polls for these jurisdictions]. 

We will start with the headline chart that shows the current state of polling compared with the outcome of the previous election. As you can see, with the exception of Queensland, Labor state polling is below what each State achieved at the 2022 Federal Election. In the three largest states, it is below 50 per cent. 

Friday 28 March - we have an Election date

It has happened. This morning a Federal election was called for Saturday 3 May. The Parliament has been prorogued (a fancy word for closed). The House of Representatives has been dissolved. The Government is now a caretaker government, and (by convention) it only takes major decisions in consultation with the Opposition.

Over the next six weeks I will be blogging most days on electoral and polling matters. Today we will look at the betting market and primary vote polling in the states.

The betting market

The betting market at sportsbet has moved around a bit over the past few days. Around noon today, it gave Labor a 52.6 per cent probability of winning the 2025 election. That is very close to even money, with Labor as the slight favourite going into this race.

Late March Update

An Australian Federal election is likely to called in the next few weeks, and could be called as early as this weekend. Yesterday, the betting market has saw Labor pull ahead as the favourite to form government after the election. 

Mid-March 2025 Update

Since Donald Trump was sworn in as the President of the United States of America, the Australian polls have been trending away from the Coalition. Now Labor is is ahead on all of my aggregation models of the two-party preferred vote. If an election was held now, the sizeable "third party" vote would see either Labor or the Coalition form minority government. 

There was movement at the station ...

With apologies to Banjo Patterson for the headline, we are seeing a significant change in the polls. Since the start of 2025 we have seen a significant drift in primary voting intention away from the mainstream parties (Labor and the Coalition) to the Greens and Others. One consequence of this drift has been a weakening in the Coalition's lead in the two-party preferred polls, especially if we allocate preference flows based on the flows at the previous 2022 election. We are also seeing Albanese firming in the attitudinal polls. 

Let's start with the first preference, primary vote polls. Both Labor and the Coalition have declined since the start of 2025. 

Late February 2025 Polling Update

Two-party preferred

The localised regression has Labor on 49 per cent, and the Coalition on 51 per cent of the two party preferred vote (2pp/TPP). There is not much change since the start of 2025.

Betting and polling update: mid-February

With the Victorian by-elections on 8 February for two seats, we have seen some movement on the betting market for the next Federal election. Currently, punters are giving the Coalition a 58 per cent probability of forming government after the next Federal election. Note: the gaps in the following charts are those days where I did not collects odds from SportsBet.

Code clean-up: Bayesian Aggregation models

I recently spent some time cleaning up the Python code I use to produce the Bayesian Aggregation charts.  When I originally wrote the code some two-plus years ago, it was all in the global name-space. For short analytical notebooks, that's generally not a problem, But for larger notebooks (such as this one) it can hide name clashes, which can result in subtle errors. 

About six months ago, one of my models - the Gaussian Process model - just randomly stopped working. The error message was obscure (something about the linker not working). It was not immediately obvious what the problem was. And I did not want to spend the time diagnosing the actual problem. 

In the past week, I decided to bite the bullet, and encapsulate the Python code into functions, so that the code was mostly not in the global name-space. I also decided I would make the Python code lint compliant (using black, ruff and pylint) and type safe (using mypy). I have now completed that process (well almost, I still have one function with too many local variables). 

The unexpected benefit: the Gaussian Process model works again (even though I did not fiddle with that bit of the code). The lesson learnt (again): doing too much work in the global namespace can result in subtle and hard to detect errors. The other lesson learnt: keep the code clean, and regularly check it with linting tools such as mypy and pylint. 

If you want to see the rewritten notebook, it is here. The main supporting functions (including the Bayesian models) are here.

Anyways, the most recent endpoints for the three models I run are as follows (the values in the table are percentages). 

	2pp vote ALP	2pp vote L/NP	Primary vote ALP	Primary vote GRN	Primary vote L/NP	Primary vote Other
Gaussian Random Walk - Normal likelihood - fixed priors	48.82	51.16	30.67	12.40	39.39	17.53
Gaussian Process - Normal likelihood - fixed priors	49.18	50.82	31.02	12.46	39.10	17.41
150-day local regression	48.42	51.58	30.18	12.65	39.81	16.52

My preferred model remains the Gaussian Random Walk. As I have noted before, the Gaussian Process model does not perform well when the polling information is relative rare (as it was in 2022 and into 2023. The Gaussian Process model also tends to revert back to the mean at the ends of the series. The local regression model can be overly influenced by the last few data points on the right hand side.

The latest charts follow.

Polling update mid January

January is normally a quite month for polling in Australia, and 2025 is no exception. Nonetheless there have been a couple of polls since Christmas 2024. These polls suggest little movement since Christmas in two-party preferred terms. You will note that I now indicate on the first house-effects chart below, those polling houses which I have constrained so that their house effects sum to zero. I exclude pollsters with fewer than 5 polls, and the prior methodologies from pollsters which have changed methodologies.