![](https://dl6pgk4f88hky.cloudfront.net/2021/06/gettyimages-873171232-scaled.jpg)
Choose your narrative: despite their various difficulties, the Conservatives are only one or two points behind Labour in the polls. The party will surely be back to winning ways once it [insert commentator’s hobby horse conclusion here].
Alternatively, Survation, the only pollster to call the result within the margin of error, shows Labour are eight points ahead. The Conservatives are surely doomed, in a validation for [insert commentator’s preferred politician here].
However, I don’t think that voting intention is particularly useful, certainly not this far out and perhaps not at all.
We know that for a variety of reasons – changing work patterns, different uses of phones, the difficulty of reaching non-political respondents – that it is getting harder and harder to do voting intention correctly.
That’s one reason why Populus has stopped doing voting intention entirely and several other pollsters privately agree that modelling, such as that done privately by Populus for the Scottish Conservatives at the last election, or YouGov’s model which correctly predicted the 2017 election result, is where the action will increasingly be in elections to come.
Although one of the repeated errors that pundits, including me, tend to make at elections is refighting the last war, you might – if nothing else because it’s good to have a system, trust Survation if you really want to trust one, as they got the last one right, or trust the average because that has a better overall success rate than any one pollster.
However, despite the fact they got it right, I don’t think Survation are exempt from the secular problems with voting intention. Indeed, their “correct” polls pre-election all exhibited similar problems to other companies under the hood. Their 5 June poll had too many people down as “certain to vote”, and all three polls understated Labour’s advantage among Remain voters in the election itself.
The industry problem is that they are finding it hard to talk to the right people. Occasionally, talking to the wrong people will get you the “right” result but it doesn’t mean that the methods aren’t flawed. If you polled the offices of the Spectator and the New Statesman, you would get a big Tory vote, a big Labour vote and a Liberal Democrat rump: but that doesn’t mean this is a reliable way of predicting elections.
That Survation doesn’t appear to be exempt from the problem (and polling too many of the politically motivated almost certainly accounts for the underestimation of Labour’s performance among Remainers) means that I am not persuaded that we should expect it to be any more or less accurate than other pollsters at the next election, whenever it may be.
In addition, as I’ve explained before, the trouble pollsters have this far out is it’s a bit like asking if you want red or white wine. The context may change depending on how expensive the wine is, what the food being served is, and so on. Pollsters have the added problem that they are asking not what wine you want tonight or even next week but in five years’ time.
It’s probably better to focus on what polls can reliably tell us, which is how different demographics are moving: we know from the 2017 election that the Conservatives have a problem with the young, social liberals of all types and socially liberal Remainers in particular, while Labour has a problem with the old, the socially conservative, and socially conservative Leavers in particular. We know that to win the next election, both parties need to protect their most vulnerable flank and advance into the others’ territory. And there is plenty of information in polls, modelling, and indeed in local council election results about that in particular, without worrying much about what voting intention alone is saying, particularly this far from an election.