A commanding Coalition lead with a record One Nation vote is a call for change, or is it a narrow parliamentary advantage for Labor and the Greens?
Victoria’s latest polling offers enough ammunition for competing political narratives. What it does not offer is certainty about who will govern after November 28.
That distinction is being buried beneath the numbers.
Two assessments of Victoria’s election prospects have produced sharply different headlines: one suggests the Coalition could win majority government, while the other places Labor and the Greens within reach of a combined parliamentary majority.
The gap demands scrutiny. But the central problem is that the releases measure different things over different periods, making a direct comparison less straightforward than their headlines suggest.
The Australian Financial Review’s Redbridge/Accent Research poll is a recent statewide voting-intention survey. Roy Morgan’s release combines months of survey responses with electorate profiles to assess parliamentary outcomes and One Nation’s electoral opportunities.
Both offer useful information. Neither provides enough evidence to treat its preferred outcome as settled.
The Redbridge/Accent survey, conducted between 2 and 8 October among 1,008 respondents, recorded the Coalition on 30 per cent, Labor and One Nation on 24 per cent each, the Greens on 16 per cent and other candidates on six per cent.
Respondent preferences produced a Coalition lead over Labor of 56–44. Compared with the previous survey, Labor was unchanged, the Coalition and Greens each rose one point, and One Nation and others each fell one point.
Those figures support a clear conclusion: this particular survey found Labor in a weak position and the Coalition substantially ahead after preferences.
They support a less confident conclusion about momentum.
A one-percentage-point movement is too small to establish a change in public opinion without examining the uncertainty surrounding both surveys. Differences in respondents, weighting and ordinary survey variation can produce movements of that size.
The Coalition’s reported rise should therefore be distinguished from its broader lead. Likewise, the Greens reaching a polling high does not necessarily demonstrate a sudden breakthrough.
Labor remaining on 24 per cent also means the latest survey does not show a fresh decline since Redbridge’s previous reading. It shows persistent weakness.
For Premier Ben Carroll, that is still a serious political problem. A government seeking a fourth term cannot take reassurance from being steady at a low level. But accurate criticism requires distinguishing a poor position from a newly worsening one.
Roy Morgan’s assessment tells a different story. Published on 9 October, it projects 40 seats for Labor, 31 for the Coalition, 11 for One Nation and six for the Greens.
That would leave neither Labor nor the Coalition with the 45 seats needed for a majority in the 88-member Legislative Assembly. Labor and the Greens would hold 46 seats between them, compared with 42 for the Coalition and One Nation.
The release draws on responses from 7,708 Victorian electors collected between March and September. Its “Winnability Score” combines voting intention with the degree to which an electorate’s population resembles a party’s support base.
The distinction between those two products is fundamental.
| Feature | Redbridge/Accent October survey | Roy Morgan October release |
|---|---|---|
| Underlying period | 2–8 October | March–September |
| Reported sample | 1,008 respondents | 7,708 electors |
| Main output | Statewide voting intentions and preference estimates | Seat projection and One Nation opportunity rankings |
| Headline implication | Coalition could win majority government | Neither major party grouping wins a majority alone |
| Main scrutiny required | Sampling, weighting and preference estimates | Timing, electorate modelling and uncertainty around seat outcomes |
Roy Morgan’s larger overall sample is a strength, but it does not resolve the question of how accurately the release measures the electorate’s present mood.
Pooling interviews over several months provides more observations and can help identify patterns among demographic groups. It also creates a potential tension with describing the result as an election held “today”.
An interview conducted in March captures a different moment from one conducted in September. The importance of that difference depends on how the analysis weights older responses, accounts for changes in voting intention and incorporates political developments during the collection period.
The public release does not explain those adjustments in sufficient detail to independently reconstruct the headline seat forecast.
That is a transparency limitation, rather than proof the forecast is incorrect.
The same caution applies to the relationship between Roy Morgan’s opportunity rankings and its projected victories.
An electorate whose residents resemble One Nation supporters may offer favourable conditions for the party. That does not establish that its candidate will win.
Candidate recognition, local campaigning, incumbency, competing candidates and preference flows can affect the result. A demographic opportunity score needs a clearly explained connection to electoral outcomes before readers can judge how it supports a forecast of 11 seats.
The published appendix ranks electorates by One Nation’s “winnability”. It does not provide a complete set of projected candidate vote shares, winning margins and uncertainty ranges for every district.
Without that information, readers cannot readily assess whether the projected seats are comfortable victories or narrow contests vulnerable to small changes.
Redbridge faces a different limitation: moving from a statewide lead to a parliamentary majority.
Victoria’s lower-house elections use full preferential voting. Where no candidate receives an absolute majority initially, preferences are distributed through the elimination process until a winner emerges. The decisive contest therefore depends on which candidates remain in contention in each electorate.
A statewide Coalition–Labor two-party estimate is informative, but it does not reproduce every local contest.
Some electorates may finish as Labor versus the Greens, the Coalition versus One Nation, or another pairing. Where several parties are competitive, relatively small changes in first preferences can alter the elimination order and redirect the path to victory.
The Coalition’s six-point statewide primary-vote advantage over both Labor and One Nation cannot establish that it leads those parties everywhere.
In the Financial Review report, Redbridge director Tony Barry and Accent Research principal Shaun Ratcliff make their majority-government assessment conditional on the polling result being replicated at the election.
That condition matters. It preserves the distinction between a plausible consequence of the survey and a prediction guaranteed to occur.
Preference estimates deserve particular attention in this contest.
Asking respondents how they would preference parties can capture changing attitudes that historical election flows might miss. But answers given before the final candidate field and campaign are known remain statements of intention.
How-to-vote recommendations may influence voters; they do not control voters’ choices. A statewide preference pattern may also conceal substantial differences between electorates.
Roy Morgan acknowledges that elimination order and preference flows can change its assessment of a seat. Redbridge’s reported Coalition lead likewise depends on how respondents’ preferences are interpreted.
The relevant question is how sensitive each conclusion is to different plausible preference outcomes.
Would modest changes leave the Coalition comfortably ahead? Would they move enough marginal electorates to overturn Roy Morgan’s projected balance?
There is also a distinction between sampling uncertainty and forecast uncertainty.
A conventional margin of error describes uncertainty around a survey percentage under specified assumptions. It does not automatically account for recruitment bias, inaccurate recall, changes after interviewing or errors introduced by converting votes into seats.
Nor does a statewide sample of 7,708 mean every electorate has been directly surveyed with comparable precision. Modelled local estimates borrow information from wider patterns; their reliability depends on whether those patterns accurately describe the district.
Redbridge and Accent have previously published useful methodological detail. Their September Victorian release identified online-panel recruitment, demographic quotas, weighting efficiency and the exclusion of respondents who remained undecided after a follow-up question. It also warned that subgroup findings carried greater uncertainty. Those disclosures concern September’s survey and should not automatically be assumed to describe October’s unchanged.
Equivalent detail for the latest survey is needed to assess recruitment, effective sample size, undecided voters and the preference questions behind its headline figures.
One argument can, however, be rejected without needing further methodological disclosure: that Labor cannot lose support while the Greens gain.
There is no electoral rule requiring those parties to move together. Dissatisfied Labor voters may move left, right or towards independents. Different groups can shift in different directions simultaneously.
Headline percentages alone cannot identify those movements. Evidence about previous votes and current intentions would be needed to establish where the Greens’ support originated.
Finally, Roy Morgan’s combined parliamentary totals should not be confused with governing agreements.
Forty-six Labor and Greens members could provide the numbers for a government, but those numbers would not settle the terms of support. The same applies to combining Coalition and One Nation seats.
The most defensible reading is that Redbridge’s recent survey provides evidence of a substantial Coalition advantage under its preference assumptions, while Roy Morgan’s longer-period analysis identifies a possible fragmented parliament.
Their disagreement does not prove that both are wrong. It demonstrates why voters need more than the headline: recent fieldwork, clear methods, electorate-level evidence and an explanation of how readily the forecast changes when its assumptions change.
Those are the tests both assessments should face before their numbers become campaign certainties.
Support our Journalism
No-nonsense journalism. No paywalls. Whether you’re in Australia, the UK, Canada, the USA, or India, you can support The Australia Today by taking a paid subscription via Patreon or donating via PayPal — and help keep honest, fearless journalism alive.
