The polls might still be overstating Labour
Disillusionment and disengagement rather than defection is the danger
The Ipsos-Mori poll this week contained a paradox. On the one hand, Labour’s headline voting intention share was 34%, some way up on their General Election performance. On the other, Jeremy Corbyn’s approval ratings were awful. His overall score of -34 was bad enough but his net rating with Labour’s own voters, at -7%, was considerably lower than Theresa May’s approval rating of some +16% with those same voters. In fact his true overall rating may be even worse: 13% of Conservatives stated they were satisfied with how Corbyn was doing his job, which is not necessarily an endorsement of his effectiveness in leading Labour. What’s going on?
The simple answer to that is that Mori report Labour to be picking up support from the Lib Dems and UKIP faster than they’re shedding it. The increased Tory lead is the consequence of a better retention rate of 2015 voters (though both are high: Con leads with 94% to Labour’s 90% among the sample that generates the headline figure), and the Tories gaining former UKIP and Lib Dem voters even faster than Labour.
The Conservative figure I can understand. A new leader is in place and against the divisions or irrelevance of her opponents and the tarnished reputation of her predecessor, she is being bathed in a very favourable light. That won’t last but for now she can enjoy her honeymoon with the public.
The Labour score makes a lot less sense and we ought to interrogate it far more closely.
Mori do apply a turnout filter – only those who say they are 9/10 or 10/10 to vote are counted – but I’d question whether even that is tough enough. The public invariably overestimates their willingness to cast a vote. There are some technical reasons for why a 100% turnout is impossible such as double-registration of students studying away from home but these aren’t sufficient to account for the difference between the actual turnout and those the polls suggest would happen.
Mori report almost 70% as ‘certain’ to vote, 75% as 9+ out of 10 (the base they use for their headline figures), and 80% as 8+. By contrast, the last general election achieved only a 66% turnout and that was the best this century. It is true that the EURef generated a 72% turnout but it would be foolhardy to read that across to a general election, where different factors are in play and where the result that each vote contributes to is in many cases much less in doubt than the referendum was.
And Labour voters above all have a history of not turning out. The ten lowest turnouts in the 2015 election outside of N Ireland were all in seats won comfortably by Labour. If they were the only place that a discrepancy between anticipated and actual voting took place, it wouldn’t matter. They’re not.
For all the attempts to rework methodology over the years, polls seem to retain an enduringly stubborn bias to Labour when it matters. To answer why that is is to seek the holy grail of polling but one factor I suspect is at play is that those with a broad inclination to Labour are disproportionately more likely to say that they’ll vote and then not follow up on that claim than their Tory equivalents.
Were it only in safe seats that the phenomenon displayed itself then it wouldn’t matter for the outcome. A seat won on a 40% turnout is worth the same as one won on double that. However, that’s probably not the case. Seats are not homogenous throughout and the key marginals will contain strong Labour and strong Tory areas; areas which in local elections exhibit a similar trend of differential turnout are likely to carry their habits of voting or not voting into a General Election.
Similarly, we know that Labour’s support is skewed to the young and the Tories’ to the elderly, and we also know which group is far more likely to actually cast their ballot papers. The polls should be correcting for this but if the polls are wrong about intended turnout – and they invariably are – then they may not be correcting enough.
To some extent we shouldn’t make a general case out of the Mori poll. The 34% they reported was well above the level that other pollsters have found (generally, a couple of points either side of 30% for the last two months; ICM recorded a 28% Labour share this week), but the general question still applies: is it really credible that a party with a leader that is viewed so poorly across the board, and particularly by those who say they’d vote for it, would really poll at or above the level they achieved in the 2015 election?
The evidence from real elections is mixed. The results from both May and from local by-elections point to churn rather than any consistent movement to or from Labour – which might suggest that they should be at least at their 2015GE share. However, William Hague’s Conservatives also recorded good interim election results in 1998-2001 which flattered to deceive when it mattered (although Hague’s Tories were so far behind in the national polls that they weren’t so much out of sight as lapped).
Successful prediction is the art of sifting useful evidence away from that which misleads. Which of today’s evidence is misleading us? My guess is that it’s the Labour retention figures; that were they asked to, far fewer of those inclined to Labour but who aren’t satisfied with Corbyn would turn out than say they would at a time when the government of the country was at stake. And unless Labour can sort out its leadership problem or the pollsters can sort out their problems with their intention to vote figures, that structural error is likely to remain.