An extensive look at the debut entry for what promises to be a monthly federal polling series from the Age/Herald.
The Age/Herald have published their first poll of federal voting intention since the 2019 election, dispensing with the services of Ipsos (who happened to be the least wrong pollster at the election) and enlisting Resolve Strategic, which is run by Jim Reed, who once worked for Coalition pollsters Crosby Textor. As national political editor Tory Maguire explains, the polling failure of the last election has inspired the pollster and its publisher to cast around for a fresh approach, the salient feature of which is not telling us straight what the two-party preferred is.
We do get primary votes though, and they are quite a bit different from Newspoll’s, with the Coalition on 38% rather than 40% and Labor on 33% rather than 38%. This means higher scores for minor parties, which happens to replicate a peculiarity of Ipsos. The Greens are on 12% compared with Newspoll’s 11%, but most striking is a 6% reading for One Nation compared with 2% from Newspoll. The latter result is, hopefully, a teething problem: it approximates the party’s 5.4% Senate vote in 2019, but most assuredly would not be matched in the House of Representatives since the party contests few seats there. It also seems highly unlikely that One Nation would be bearing up so well given its recent performance at state elections, with its share of the upper house vote in Western Australia having crashed from 8.2% to 1.5%.
Applying preference flows from 2019, this lands pretty much bang on 50-50 nationally, with the Coalition leading 53.4-46.6 in New South Wales (a 1.6% swing to the Coalition) and 54.3-45.7 in Queensland (a 4.1% swing to Labor), but trailing 53.8-46.2 in Victoria (a 0.7% swing to Labor) and 52.4-47.6 in Western Australia (an 8.0% swing to Labor). The New South Wales result is more favourable for the Coalition than the recent quarterly breakdown in Newspoll, which had it at 50-50, but the results for the other three states are about the same. Distinctions by gender are slight in the case of voting intention, except that the Greens are three points higher among women and One Nation are two points lower, and confounding in the case of personal ratings: Scott Morrison’s net approval is four points stronger among women than men, while Anthony Albanese’s is five points weaker.
Personal ratings are measured on a four-point scale of very good, good, poor and very poor, which is similar to Essential Research but different from Newspoll’s straightforward satisfied and unsatisfied responses. Morrison registers a combined good rating of 50% and a poor rating of 38%, a net rating of plus 12% that compares with plus 17% from a recent Essential poll and plus 15% from a not-so-recent Newspoll. Anthony Albanese scores 35% good and 41% poor, for a minus 6% rating that compares with plus 5% from Essential and plus 2% from Newspoll. Morrison is credited with a 47-25 lead as preferred prime minister, compared with 47-28 in Essential and 52-32 in Newspoll.
The poll wins points for transparency, at least by Australian standards, in providing breakdowns by state (or at least, the four biggest states), gender and age cohort. If I’m reading the small print correctly, the New South Wales and Victoria breakdowns will be published as a two-month rolling average, combining the current and previous poll. The idea seems to be that these states will have results with sample sizes robust enough to allow the Age/Herald to analyse them with a straight face: readers who choose to probe deeper into the breakdowns will be advised to exercise their own caution. (UPDATE: It seems I’ve read this wrong, and that there will actually be state voting intention results published every two months, based on the combination of two monthly polling samples). However, the results as published still leave a fair bit missing, as the poll is also weighted for education and income (and the survey includes a question on religion), for which breakdowns are not provided. There is also the rather glaring absence of any detail on field work dates: we are told only that the survey was conducted “in April”.
One of two accompanying reports by David Crowe relates the following detail absent from the published numbers:
Support for the Coalition in primary vote terms has fallen since the last election among voters who described themselves as Christian, dropping from 56 to 49 per cent. While support for Labor among this group rose from 28 to 29 per cent, the change was within the margin of error. In a more significant shift, the same cohort increased its support for independents and minor parties from 15 to 22 per cent. The Coalition has lost ground among voters across the board since the last election, with its primary vote slipping from 41 to 38 per cent, but the shift was strongest among immigrants and people from “non-Anglo-Saxon” backgrounds. Immigrants have reduced their support for the Coalition from 48 to 40 per cent since the election, while increasing their primary vote for Labor from 30 to 35 per cent. Those from “non-Anglo” backgrounds reduced their support for the Coalition from 44 to 35 per cent, while increasing their support for Labor from 31 to 36 per cent.
The other tells us the following:
The swing against the Coalition was spread evenly across most demographic groups but was more pronounced among those on higher incomes, with support falling from 49 to 43 per cent among those earning more than $100,000 a year. Labor gained support from the same workers, with its primary vote rising from 29 to 33 per cent.
There are further attitudinal results available in a nicely laid out results display page, including the finding that 44% expect the Coalition to win the next election compared with 28% for Labor. The display includes, under “comments”, sampling of qualitative responses that aim for an impressionistic view of why the ratings for each question are what they are. The poll was conducted by phone and online from a sample of 2000, compared with the Newspoll norm of around 1500. However, the phone sample of 400 appears to be a one-off of this “baseline survey”: it seems that in future the series will be a monthly online poll from a sample of 1600.