The Sunday Long Read
In a startling result, the recent YouGov poll found that in his own, usually safe, Conservative constituency, Rishi Sunak was beaten into third place by ‘don’t know’. When so many people are disillusioned about politics, could this be the election when the ‘don’t knows’ decide the result?
Polls are not predictions
We read opinion polls because we want to know what would happen in the next election. But in reality they can only tell us how people say they would vote if there were an election today. And there isn’t an election today. When the election comes, people will think harder about the question, and they may change their minds.
At present, the polls tell us two important things.
Labour will win
First, the 500 polls Since Rishi Sunak became Prime Minister have all produced similar results. Despite different samples and methodologies they cluster steadily around a Labour lead of 18 points. If anything, the lead is rising very slowly. None has ever shown that lead fall below 10 points.
The current figures are similar to Labour’s poll position a year before Tony Blair’s landslide victory in 1997. So, it will be very surprising if Labour does not win.
There are a lot of ‘don’t knows’
Second, an unusually large number of people say they don’t know how they would vote (16% in YouGov’s latest poll). This reflects a widespread disillusion with politics. But the ‘don’t knows’ are not evenly distributed across parties. While Labour and Liberal Democrats are holding on to most of their 2019 voters, the Conservatives are losing theirs. In the latest poll from Redfield and Wilton, less than half of those who voted Conservative in 2019 would do so again. They report that over a third are switching: evenly divided between Labour and Reform. But it the remainder decide to go back to the Conservatives they eat into Labour’s lead.
Have the Tories hit a tipping point?
A traditional opinion poll can tell us about how the votes will divide up at national level. But we vote in constituencies, and party support is not evenly spread across the country. Labour voters are concentrated in a small number of mainly urban constituencies, while Conservative voters are spread more evenly. So, under our ‘first past the post’ electoral system, when the national vote divides evenly between Conservative and Labour, the Conservatives win most seats in Parliament. But they are more vulnerable to a large swing.
So there is a tipping point, where a large number of Conservatives are suddenly at risk. This is why some commentators now expect bigger Conservative losses than the overall national figures might suggest.
And tactical voting could increase the shift. It seems likely that the 14% of 2019 Liberal Democrat voters who told YouGov they ‘don’t know’ are waiting to see how to most effectively defeat a Conservative in their particular constituency. And on the latest YouGov polling, there are only two seats where the Conservative vote share is now larger than the combined Labour/LibDem/Green share.
How do polls deal with the ‘don’t knows’?
Whether we have reached such a tipping point depends critically on how, and whether, the ‘don’t knows’ vote. There has been some debate about whether they could upset the predictions.
Different pollsters take different approaches to the ‘don’t knows’. YouGov, Redfield & Wilton and Deltapoll all disregard them. They argue that a poll is not a prediction, but a measure of what people say now. If someone says they don’t know, a pollster should not make a decision for them. However, in the past, people who say ‘don’t know’ tend to return to their previous party when the election happens. If they do this this time, these polls will have underestimated the Conservative vote.
Other pollsters accept that we want predictions, and try to estimate how the ‘don’t knows’ would actually vote. Ipsos and BMG both ask a follow up ‘squeeze’ question, along the lines of “if you had to choose which party do you think it would be” and count that answer. Other firms look at past behaviour for a guide to what proportion of the ‘don’t knows’ will return to the party they voted for last time.
Constituency level polling
A traditional way of estimating the likely result of the next election is to assume that the swing from one party to another is the same across the country. Historically, this ‘uniform swing’ has proved fairly accurate, and on this basis, an 18 point Labour lead could give Labour a working majority, of perhaps 30 seats.
But people vote in constituencies, and constituencies are not all alike. In 2019 in the North, the swing from Labour to Conservative was much greater in the Northern ‘Red Wall’ than In the rest of the country. The Liberal Democrats traditionally focus their efforts heavily in key constituencies, where their performance often exceeds the national swing. So, we need to understand what is happening at constituency level.
MRP polls try to estimate results at constituency level, by matching the profile of the population in each constituency with the national voting intention of those groups. For example, national polling tells us how white graduate women in their 50s say they will vote. Local data tells us how many white graduate women in their 50s there are in each constituency. These two are combined, alongside hundreds of such results, into a complicated formula, to arrive at a polling result for each constituency.
YouGov and Survation have both done this recently. Both predict a very comfortable Labour majority, but they differ over the scale. Survation, who ignore the ‘don’t knows’, suggests a Labour majority in Parliament of 284 seats (higher that Tony Blair’s peak of 247 in 2001). But if those 2019 Conservatives who say ‘don’t know’ now actually turn out for the Tories, that will be an overestimate. YouGov takes a different approach. They allocate the ‘don’t knows’ to the groups they are most like, and expect more of the don’t knows to return to the Conservatives, which produces a lower majority of 120, but with more Liberal Democrat MPs (48).
The more cautious YouGov poll still suggests that In the East of England the Conservatives would lose 14 seats to Labour and 3 to Liberal Democrats.
Conclusion
Barring an extremely unusual event, Labour will win the next election. A lot of those who voted Conservative in 2019 have already switched to Labour or Reform. That leaves only 13% who say they don’t know how they would vote. This could shrink the Labour majority in Parliament, but not enough to change the result.
And for them all to go back to voting Conservative would require the Conservatives do something remarkable; for Labour to make a big mistake; or for Reform to stand down (as they did in key seats in 2019).
The experience of the last 15 months suggests that none of these is likely to happen. We are unlikely to see Survation’s 284 seat majority, and we may not see YouGov’s 120, but it is highly unlikely that we will have another Conservative government in a year’s time.
