Only the dedicated psephologists among us are crass enough to say so out loud, but one of the big questions in politics at the moment is how much of an opinion poll bounce Tony Abbott can expect in the wake of his generally well-received handling of the MH17 tragedy. Notwithstanding three new federal polls that were published earlier this week, the present answer is that it's still too early to say. Check back on this space in a week's time and you will find the question attended to in unseemly detail.
For the time being, the big news in polling is that Fairfax has published the last result we're ever going to get from Nielsen, which has chosen to close its Australian polling operation. This stands to deprive hardcore polling wonks of the intelligence Nielsen provides as to what might be going on beneath the primary vote trends -- recent evidence of which could take some of the shine off whatever poll boost might be coming the government's way.
Along with Roy Morgan, Nielsen has been one of only two pollsters to make known the responses of minor party and independent voters when asked how they propose to allocate their preferences, through what's known in the trade as a respondent-allocated two-party preferred result. Other pollsters derive their two-party preferred results by applying preference flows from the previous election, as indeed Nielsen did in the headline figures as they usually appeared in Fairfax papers.
This practice accommodates the reality that poll respondents often don't conceive vote choice any more deeply than identifying a particular party to support, the realities of compulsory preferential voting notwithstanding. For many, preference allocation is a matter of following a how-to-vote card that won't be available to them until they arrive at the polling booth.
Furthermore, the sample of minor party voters in a given opinion poll is rather small, so there's a big chance preference ratios derived from their responses will prove to be off the mark. Such considerations have grown in importance along with the rise in minor party support over the decades, making primary vote totals ever more deficient as a guide to the likely result. The shortcomings of respondent allocation were illustrated when Newspoll employed it throughout the election year of 2004, causing its two-party preferred results to be overly generous to Labor -- which, as Peter Brent of Mumble
explains in detail, created an exaggerated impression of Mark Latham’s political successes during the early months of his leadership.
In its final poll before the election, Newspoll turned in a set of primary vote numbers that leaned only slightly to Labor, perhaps representing its failure to catch the effect of the infamous Latham handshake on the last day of the campaign. However, the two-party preferred result from the poll was 50-50, which meant the pollster was seen to have misread an election that returned the Howard government with an increased majority and control of the Senate -- the actual two-party preferred result having been 52.7-47.3. Newspoll promptly reverted to using previous election flows, and has since had no cause to regret the fact.
Even so, the previous election method is not without problems of its own, as illustrated last year when Labor's share of Greens preferences rose from 78.8% at the 2010 election to 83%, and that of all other minor parties and independents combined rose from 41.7% to 46.7% -- the latter shift pointing to the problems that can arise when an untested new player such as Palmer United arrives on the scene.
Using preferences from the 2010 election, a poll that correctly recorded the primary vote last September would have sold Labor short by about 1% on two-party preferred. It is thus advisable to take an occasional look at what respondent-allocated preferences might be telling us, and the loss of one of the only two sources of such information seems an opportune moment to do so.
The two charts below show trend readings of Labor’s share of non-major party preferences as recorded by Nielsen and Morgan.
The chart on the left shows Nielsen’s results going back to the 2010 election, together with the share of preferences Labor would have received going off 2010 and 2013 election preference flows -- with slight fluctuations reflecting changes in the Greens’ share of the overall non-major party vote.
This clearly shows that something quite striking happened in June 2013, and it is obviously no coincidence that the effect precisely coincides with Kevin Rudd's return to the prime ministership. Nielsen had hitherto recorded Labor skirting the middle of the 2010 and 2013 election preference flows, but it would soon overshot the latter figure to reach 69% by polling day -- substantially more generous than the 61% they actually received, lest this be seen as an endorsement of the respondent allocation approach. Not only did this effect outlive Labor’s primary vote “sugar hit” after the return of Rudd, it has continued apace after the election to reach its present dizzying height of 75.5%.
Going off the primary vote numbers currently recorded by my BludgerTrack poll aggregate
, and assuming no change in Greens preference flows on last year's election, this implies that Labor's share of preferences from smaller parties and independents has gone from below 50% to nearly 70%.
The chart on the right performs the same exercise on the respondent-allocated figures from Roy Morgan, going back as far as the adoption of Morgan’s current poll methodology last March, with Morgan in blue and Nielsen in red for comparison. While this series fairly consistently has Labor’s preference share around 5% lower, the shape of the trend forcefully confirms the impression conveyed by Nielsen.
It should again be stressed that it's very unlikely to play out the way Nielsen has been reading it at a real election, but it's at least clear that those of a pox-on-both-your-houses persuasion are feeling especially sour towards Tony Abbott's government. All the more reason then for nervous Coalition MPs to await a post-MH17 poll swing with a little more eagerness than they might care to publicly admit.