This big MRP model helps sets the stage for the election

YouGov enters the MRP arena for the first time this election cycle

Feb 15, 2025

Update May 8: This post seems to have got a bit of traffic since the election actually happened. This is a write-up of the first MRP model YouGov published ahead of the election. YouGov’s MRP published a few days before the election estimated Labor would win majority government, with a central estimate of 84 seats.

The swing back to Labor was underway by the time of YouGov’s March MRP which I wrote about for ABC News.

I didn’t write anything on YouGov’s final model because I was far too busy with preparations and rehearsals for election night. My colleague Jacob Greber wrote about it here.

Original post below:

I’ve got a story running today on ABC News about YouGov’s first MRP poll of this electoral cycle, which has a median projection that the Coalition would win 73 seats in an election held today, Labor would win 66 seats, there’d be one Green and 10 independents (including Bob Katter and Rebekha Sharkie in that total as they’re effectively parties of one at the federal level)

If that central projection is the result on election night, it’s hard to see any government being able to form other than a Dutton government.

But there’s plenty of time for things to change, and there are lots of very competitive seats in this model.

It’s easy to fixate on the individual seat estimates from an MRP model, but I’d caution against taking it as gospel. This model will not predict the result of every single seat, and it’s most useful where it identifies broad patterns in voter behaviour.

This poll has clearly done that, and shown that the vast majority of the seats most likely to change hands are in the urban fringes of capital cities, especially Sydney.

As I’ll talk about below, I’m slightly dubious about some of the individual seat estimates, but that doesn’t mean there aren’t valuable insights in this MRP.

If you haven’t already, read my story for ABC News here or watch my segment on Insiders here, and then come back for some more thoughts that didn’t fit in the story. And here’s YouGov’s website with full details.

MRP models aren’t new, but they’re still a relatively novel thing in Australia and there aren’t that many pollsters producing them. Redbridge and Accent Research published three of them last year (Accent Research principal Shaun Ratcliff worked for YouGov on their MRP in 2022), and there were a couple of them published by other groups just before the Voice referendum.

So we don’t have many other recent MRP models to compare this one too.

At least not publicly. Political parties are certainly commissioning their own MRP models for internal use to help them make campaign decisions and direct resources, as campaigners did during the Voice to Parliament referendum.

This time around, YouGov has conducted around 40,000 interviews, and the sample gets trimmed down to an effective sample size of around 9,000, which gets run through the MRP model to produce vote share estimates in every seat. You can read more about YouGov’s methodology here.

The national vote shares and seat projections are broadly consistent with other traditional polls and the RedBridge/Accent MRP published in December (it projected the Coalition would win between 64 and 78 seats, and Labor would win between 59 to 71 seats).

The 2PP has shifted roughly 0.5-1 percentage points toward the Coalition since that MRP was fielded, so it’s no real surprise that this one sees the Coalition projected to win roughly two more seats than RedBridge/Accent found.

The swing appears to be very efficient for Peter Dutton, and the model has the Coalition winning more seats than you might expect from a national two party preferred swing of 3.2 per cent. That’s because the model is picking up (significantly) larger swings in Labor-held seats than Coalition-held ones.

That’s why the Coalition is found to be leading (in some cases very narrowly) in 15 Labor-held seats: Bennelong, Macquarie, Werriwa, Paterson, Hunter, Robertson, Shortland, Eden-Monaro, Gilmore in New South Wales, Chisholm and McEwen in Victoria, Bullwinkel and Tangney in WA, Lyons in Tasmania and Boothby in SA.

A few stray observations and cautions on the cross bench

The model has all the independents retaining their seats except for Dai Le in Fowler, but I think we should be most cautious of the figures in the seats with competitive independent challenges. But this MRP does at least give us a sense of which candidates appear to be at least in the ballpark of being competitive.

YouGov’s MRP underestimated the level of support for independents in 2022, and while they now have more data on independent voters from that election, it’s possible that could happen again. An MRP model can struggle to pick up the weird local factors unique to a single seat, by virtue of the model that seeks to compare similar voters across multiple seats.

Respondents were asked to pick between parties on a ‘generic ballot’, where individual candidates were not named. That’s ordinary practice this far from election, because we don’t know who all the candidates are yet.

In some seats that will cause the independent support to be overstated, but in others, like for example in the case of Dai Le, where local profile and name recognition is so important, it risks understating their support for the time being. It’ll be interesting to see if there are any significant changes once YouGov starts including candidate names closer to the election.

On the flip side, the MRP could also miss dynamics that were relevant in 2022 but are no longer present in 2025. For example, this model suggests that an independent is coming second in Nicholls. Rob Priestly mounted a strong challenge there in 2022 but he’s not running again and I’m not aware of any serious independent challenge in Nicholls this election. It could be that respondents there have said they’d vote for a generic independent where one is not running this time, and also that the model is drawing a comparison between independent voters in Nicholls with independent voters in the nearby seat of Indi.

So tread cautiously on the independents especially!

This MRP also suggests Labor could make some gains in Brisbane, taking up to three seats held by the Greens. They’re narrowly behind Labor in this model in Brisbane, Ryan, and Griffith. And I mean narrowly. The primary votes estimated by the model are:

In Griffith: Liberal 29.3%, Labor 31.9%, Greens 30.7%
In Ryan: Liberal 37.1%, Labor 27.1%, Greens 26.1%
In Brisbane: Liberal 35.7%, Labor 30.0%, Greens 23.7%

Stephen Bates in Brisbane won from third place in 2022 after overtaking Labor during the preference distribution. He wouldn’t be able to do that on those numbers, but in the other two Brisbane seats, Labor and the Greens are so close that it will all depend on who comes in third.

And it’s worth noting the seats where independent candidates seem to be running in second place. They are:

Cowper, where Independent Caz Heise is trying for a second time to win the seat, rated by YouGov as ‘Lean Coalition’ (2CP estimate L/NP 50.9% vs IND 49.1%)
Bradfield, where Nicolette Boele is also running for a second time, which is rated by YouGov as ‘likely Coalition’ (2CP estimate L/NP 51.9% vs IND 48.1%)
Wannon, where Alex Dyson is trying to make it third time lucky, which is rated by YouGov as Likely Coalition (2CP estimate L/NP 52.1% vs IND 47.9%)
Calare, where former Nationals member Andrew Gee is running for re-election as an independent. This model suggests he is well behind, (2CP estimate L/NP 59.7% vs IND 40.3%) but this estimate is complicated by the fact that Gee was elected as a National three years ago and much will come down to his own name recognition as a local member.

If any of them win, that would be at the expense of the Coalition, making it harder for Peter Dutton to find a majority in parliament.

The MRP also has One Nation coming a distant second in Maranoa, as well as a generic independent running second in Nicholls, as discussed above.

The closest the Greens are coming to gaining a new seat, according to the MRP, is Wills. The model has Labor ahead there 53.8% to 46.2%. In this model Macnamara remains a close three-way contest.