Although local elections are not a straightforward guide to national elections, they do give us some indication of how voters are feeling. And they also tell us something about how opinion polling is working. This time, it looks as if the pollsters will be doing some hard thinking.
Traditional opinion polls work at national level. They take a controlled sample of people spread across the country, and ask about their voting intentions. If they get their sampling right, they can then predict how the country as a whole might vote in a general election on that day.
But this cannot tell us how people might vote in a general election, some months later. More importantly, under our first past the post electoral system, 45% of the national vote does not mean 45% of the seats in Parliament, which is what matters for government.
So, to predict a general election result, we need data at the level of individual constituencies. But nobody does conventional opinion polling at constituency level, because it is extremely expensive: the numbers involved would be huge, and getting the sample balanced in every constituency would be very difficult.
Local election results
So we depend on two other sources. One is local elections, and last week’s provide us with some insight. But local factors may play very differently in local elections from national ones. Local elections are also complicated, because different councils elect at different times. There were no local elections last week in London, Birmingham and some other large areas of England, and none in Wales or Scotland. In the areas where there were elections, some elected the whole council, but others only elected a half or a third of councillors. So, translating this evidence into a Westminster pattern is very difficult.
With those caveats, comparing last week’s results with the last ten national opinion polls suggests that the national polls are broadly right on the Conservative vote share, but they overstate the Labour one, and heavily underestimate the LibDems and Greens. This seems to be why Electoral Calculus, who have a good track record in general elections, was wildly out in its predictions of seat changes in these elections. It underestimated Conservative losses by 700, Labour gains by 140, LibDems gains by 400 and Green gains by 200.
MRP polling
The second source of evidence at constituency level is MRP polling. This takes national data on how particular groups of people intend to vote, and applies those numbers to the known profile of the population in each constituency. So, for example, we know that graduates are more likely to vote Labour than Conservative, which increases the chances of electing a Labour MP in a constituency with a high proportion of graduates. A host of factors like this are fed into an equation, to produce a prediction for every constituency.
This has a good track record, and YouGov’s MRP poll in April, which sampled 18 Councils across England, demonstrated that MRP polling can be relatively accurate. They rightly made the (unremarkable) prediction of an overall swing from the Conservatives. However, they underestimated the scale of the swing. Between them Labour, LibDems and Greens took control of 6 more councils than predicted. This may reflect a time delay between their poll and the actual election, with people swinging more strongly against the Conservatives in the final days. It probably also reflects the difficulty of adjusting for tactical voting. If the sort of people who have traditionally voted Labour vote switch locally to LibDems or Greens, the MRP models may struggle.
Flaws in polling
Tactical voting clearly played a major part in East Suffolk, where the predictions were of a Conservative hold, or perhaps a swing to Labour. But the Greens, LibDems and Independent had a formal alliance. Between them they stood a candidate in every ward, but they did not compete with each other, thus maximising the impact of the combined vote. As a result, the Conservatives lost 24 seats, and the alliance now has a majority. This is not what the polls predicted.
There are two other constituencies where the Electoral Calculus prediction of Labour wins seem improbable. In Chelmsford, the LibDems, who were already in control, took two seats from Independents, and there are no Labour councillors. Without any councillors to provide a backbone for the local party it is difficult to see how Labour, who came third in 2019, could overtake the LibDems and overturn a 20,000 Conservative majority.
Watford is a similar case. Again, the polls suggest a Labour win, but the LibDems, who already control the council, kept all their 27 existing seats (of 36 on the council), and there are no Labour councillors. With three of their four neighbouring districts already in LibDem control, it seems unlikely that Watford will see a sudden surge of Labour support.
An end to two-party politics?
So, it looks as if the MRP polling can be relatively accurate, but other methods can be challenged by the current public mood to vote for “anyone but the Tories”. All those who share that view are likely to congregate around whichever party seem most likely to defeat them, and in these constituencies last week’s results are fairly clear. It is plausible to suggest that, where there is already a clear LibDem or Green lead, that will increase at general election time. That is good news for LibDems and Greens, and less so for Labour.
Of course, this was a local election, and voters may return to their old allegiances next year. But it begins to look like a bit of a gamble.
