Big news on the polling front as Newspoll unveils its first set of results based on what The Australian describes as “an improved methodology following an investigation into the failure of the major published polls”. The old series had been limping on post-election with results appearing every three weeks, but this latest result emerges only a fortnight after the last, presumably portending a return to the traditional fortnightly schedule.
The poll credits the Coalition with a two-party lead of 51-49, compared with 50-50 in the result a fortnight ago, from primary votes of Coalition 41% (up one), Labor 33% (down two), Greens 12% (steady) and One Nation 5% (down two). Interestingly, both leaders’ personal ratings are a lot worse than they were in the old series: Scott Morrison’s approval rating is at 43% (down three) with disapproval at 52% (up nine), while Anthony Albanese is at 38% approval (down four, though he was up five last time) and 42% disapproval (up five, though he was down seven last time). No news yet on preferred prime minister, which is presumably still a thing (UPDATE: Morrison’s lead narrows from 46-32 to 46-35).
On the methodological front, the poll has dropped robopolling and is now conducted entirely online. The sample size of 1519 is similar to before (slightly lower in fact), but the field work dates are now Thursday to Saturday rather than Thursday to Sunday. In a column for the newspaper, Campbell White of YouGov Asia-Pacific, which conducts the poll, offers the following on why robopolling has been abandoned:
A decade or so ago, most people had landlines and they tended to answer them. There was very little call screening. This meant getting a representative sample was easier and pollsters did not need to be so skilled in modelling and scaling their data. The truth is, the old days are never coming back. In order to do better, we need to consider what we can do differently. We’ve seen a consistent pattern overseas where telephone polling has become less accurate and online polling more so as fewer people answer phone calls and more and more people are online.
White further notes that “annoying and invasive” robopolling is “answered largely by older people or those who are very interested in politics”, while “busy people who are less interested in politics either don’t answer or hang up”. He also reveals that the new series will “weight the data by age interlocked with education and have precise quotas for different types of electorate throughout Australia”, consistent with YouGov’s methodology internationally.
Hopefully the restated commitment to “greater transparency” means we will shortly see comprehensive details of demographic breakdowns and weightings, a commonplace feature of British and American polling that Australian poll watchers could only envy. Stay tuned.