If you believe today’s Newspoll and the write-up from Simon Benson, the government suffered a slump in its fortunes over the last three weeks, with its 53-47% two-party-preferred (2pp) dominance over Labor reduced to a 51-49%lead.
Benson was at a loss to explain this slump. “The poll was conducted during a period in which the political contest had cooled, with the federal parliament on a six-week autumn-winter break,” he acknowledged. But he offered some ideas: it included a period when “Angus Taylor came under pressure over his alleged failure to declare family business interests as required”. There were also concerns about the economy and the trade war and Australia’s resistance to acknowledging climate change at the Pacific Islands Forum.
The broader problem is Newspoll itself: as INQ explored in its polling series, not merely was Australia’s most influential poll — the poll that decides the fates of prime ministers — wrong before the election, but it has the least transparent methodology of the major polls. This week’s result should no more be believed than its election-eve poll predicting a Labor victory — and until YouGov allows greater transparency about Newspoll, that scepticism should remain firmly in place.
But Benson’s efforts to explain a shift since the last poll points to another problem (one I too was guilty of myself back in the Essential poll days, even though we tried to downplay the 2PP result). Benson is an experienced enough political journalist to know that pretty much nothing has happened that will have caught the attention of voters in the last three weeks. But to make a yarn out of it he can’t say that either the shift is within the poll’s margin of error (so doesn’t likely mean anything), or that the entire Newspoll methodology has been demonstrated to be flawed.
Polling must always reflect some change in the electorate in response to what’s happened in Canberra — otherwise why bother reporting them?
This is the weird conspiracy of the governing class about polls — political and media elites jointly insist that polls reflect shifts in electoral sentiment from week to week. Except, they’re the main victims of this conspiracy too. The electorate doesn’t care. The average voter doesn’t even know who Angus Taylor is, or that there was a Pacific Islands Forum. A large swathe of them completely ignore the media’s political coverage. Any change in sentiment take months to happen, unless there’s a high-profile event like a change of leader.
But politicians and political journalists tell themselves and each other that the electorate is obsessing over politics as much as they do. It’s a complete delusion.
Without necessarily defending the methodology of polls, the point that seems to be overlooked by most commentators is that there is a “margin of error” of about 3% in polls of about 1500 people. Simply, that means that, all other things being equal, a poll of 51-49 would mean that a poll of the whole electorate (15 million) could be between 54-46 and 48-52. So a “drop” of 2% from 53-47 could mean that in reality nothing has changed- or it could actually be as high as 54-46 or as low as 48-52.
The weird thing about the pre-election polls was that they were so consistent. The very meaning of the concept of “margin of error” is that there will be a variation. Conceivably, the LNP was always ahead in the polls- statistically, every poll might have been randomly showing the LNP being behind- but this is “very unlikely”.
Until everyone is schooled in the meaning of “margin of error”, then we will continue to be victims of this statistical silliness.
Voters by and large, voted against labor, not for the coalition at the last election mainly because of the franking credits issue, they did not endorse Scomo but simply rejected Shorten and Bowen, but now that Scomo`s recession is hitting home they will turn away from his government more and more, if there are any by elections called over this next term we will probably see the return of a hung parliament unless Fryenbrain lifts newstart and starts spending on stimulating the dying economy.
Except there is nothing for them to turn to. Albanese has gone missing. He has the smell of old Bomber Beazley about him. Shorten might have been a party hack but at least he was visible…even if you didn’t quite like what you were looking at.
I haven’t seen any analysis, yet, that shows that the pre-election polls were incorrect. Indeed, as the INQ articles point out, they were the same, to within the 3% margin of error mentioned by Jedimaster, even though they used different polling methods. So the safest bet is to assume that they were right.
There is no conflict between that conclusion and the outcome of the election, because the two events measure different things. A whole-country-population poll will only be reflected in corresponding numbers of won seats if the distribution of voting intention is uniform across the country, and across seats. Which of course it isn’t. Until pollsters actively attempt to estimate won seats, rather than individual voting intention, we won’t learn anything interesting from them.
The point I’m making is that there is no precise “right”- that “margin of error” should be called a “margin of uncertainty”.
Of course the elections are won on an electorate-by-electorate basis – and that is where the LNP focused its attention- on the Queensland and WA seats that they had to keep- which they did with Palmer and Hanson second preferences. But they also won the overall vote by a margin that was within the “margin of uncertainty”.
As I tried to say before, given the large number of polls, some of them should have put the LNP ahead, simply on the supposed random errors. The fact that none did points to systemic faults in the methodology.
The point I’m making is that there is no precise “right”- that “margin of error” should be called a “margin of uncertainty”.
Of course the elections are won on an electorate-by-electorate basis – and that is where the LNP focused its attention- on the Queensland and WA seats that they had to keep- which they did with Palmer and Hanson second preferences. But they also won the overall vote by a margin that was within the “margin of uncertainty”.
As I tried to say before, given the large number of polls, some of them should have put the LNP ahead, simply on the supposed random errors. The fact that none did points to systemic faults in the methodology.
“If you believe ……” why go on?
I’m really wondering – why do we (non-politicians) care at all about polls?
Even if the methodology wasn’t so obviously constrained by a basic problem (you can’t know whether you are getting genuine responses from a sufficiently representative sample) you have unavoidable statistical noise that creates uncertainty close to the margins being measured.
Do we really think that Dorothy and Kevin’s thoughts two years out from an election mean anything, given most people are increasingly disengaged from political reporting altogether, other than a few soundbites on 2GB or a factious headline in the Daily Terror?
And after all this, there’s a clear sense that polling actually affects voting. A significant wedge of people voted Leave as an anti-EU protest, thinking that Remain would win the day. Voters in swing states opposed to Trump staying at home because they thought Hillary’s 3% lead and media declarations assured her of victory.
It’s not reliable, it creates shallow political intrigue and instability, and it can affect voting intentions. Given political parties conduct their own polling, is there any need for the meeja to keep this pointless exercise going at all? Just say no.