A fortnight after delivering Malcolm Turnbull the embarrassment of his 30th successive poll defeat, Newspoll’s outsized influence on Australian political discourse is being demonstrated once again.

Oddly enough, the latest kerfuffle arises from a result that seemed to suggest Turnbull’s dreaded milestone of a fortnight earlier hadn’t done him the least bit of harm.

Conducted two weekends ago from an unusually large sample of over 2000, the latest result had Labor’s lead down to 51-49 — lower than it had been at any stage since the start of the Coalition’s losing streak.

There's more to Crikey than you think.

Get more Crikey for just $199 $99.

Subscribe now

This came as an unpleasant surprise to Labor’s social media army, who had been gorging themselves on a week of bad headlines for the government arising from the banking royal commission.

When unhappy partisans and unexpected poll results combine, conspiracies are soon theorised — and an apparent tweak to Newspoll’s preference methodology gave them at least a kernel of fact to work from.

Underneath the two-party result lay the same major party primary votes as a fortnight ago: 38% for the Coalition, putting them two points above where they were stuck through 2017, and 37% for Labor, who have held steady.

The Coalition may have benefited from softening support for One Nation, who were tracking near 10% before the Queensland election campaign but have lately been around two points lower.

Presumably these are voters disposed to favour the Coalition over Labor in any case, so the change may not have done much to disturb the “true” picture after preferences.

However, the latest two-party result forms part of a bounce in the Coalition’s favour since late last year, part of which seems to have arisen from Newspoll no longer allocating preferences strictly according to how they flowed at the 2016 election.

A very sound justification for this would involve One Nation, who has been on 7% to 10% over the past year, and for whom 2016 election preference flows offer an unreliable guide, as they only contested a 10th of the 150 lower house seats.

As psephologist Kevin Bonham observes, exceptional circumstances prevailed in some cases — notably in Longman, where the party saw fit to campaign vigorously against Wyatt Roy.

The overall result was a 50-50 preference split that is surely misleading for a party whose current support base is engorged with Coalition defectors.

Certainly it was not reflected at One Nation’s two big state elections in Queensland and Western Australia, when the party’s preferences broke at least 60-40 to the Coalition.

A switch to this ratio shortly after the Queensland election would neatly explain Newspoll’s recent two-party form — a change usually sufficient to sway the rounded total a point in the Coalition’s favour if One Nation support is approaching double figures.

A report on the matter in The Australian reveals that the new formula reflects “the observed preference flows from both elections”, but the exact nature of the change remains undisclosed. The company did something similar before the Queensland election, when a preference flow was derived from the three previous election results to avoid replicating the unusual preference backlash against Campbell Newman’s government in 2015.

This approach looked pretty good after the election, and it’s likely its present approach will too.

However, a prerogative for pollsters to keep their mix of herbs and spices secret is problematic, particularly as the long-term escalation of the minor party vote makes preference allocations ever more significant.

The lack of detail stands in contrast with the polling published in Britain by Galaxy’s parent company, YouGov, which includes detailed accounting of demographic breakdowns and weightings, and would assuredly be as forthcoming on preferences if that were a factor in the country’s electoral system.

However, the very reason the British polling industry has felt compelled to observe higher standards of transparency is that it would invite ridicule if it sought to claim, as Galaxy did yesterday, that its “track record speaks for itself”.

If ever the sorts of failures seen in Britain at the 2015 general election and 2016 Brexit referendum are replicated here, a day of reckoning may arrive that will shine light on the dark corners of Australian opinion polling.

But for the time being, followers of the political horse race will have to keep taking Newspoll’s word for it.

Correction: This article originally stated that YouGov Galaxy had released little more than a tweet regarding its polling methodology. At the time of publication, this was untrue. In fact, two articles about the Newspoll methodology were published in The Australian on Monday.

There's more to Crikey than you think.

It’s more than a newsletter. It’s where readers expect more – fearless journalism from a truly independent perspective. We don’t pander to anyone’s party biases. We question everything, explore the uncomfortable and dig deeper.

And now you get more from your membership than ever before.

Peter Fray
Peter Fray
Editor-in-chief of Crikey
Get more and save 50%