Is the race tightening? Decoding the most significant parts of YouGov’s new election MRP survey

Not surprisingly, the second big YouGov MRP survey is attracting as much attention as the first, two weeks ago. It has reduced its headline projection of the Conservative majority from a comfortable 68 to a nervy 28. Why? In some ways the answer is less dramatic than it would seem at first sight—but in one way, of potentially huge significance.

The big picture—Conservative support nationally unchanged on 43 per cent, Labour up two points to 34 per cent, Liberal Democrats down two to 12 per cent—matches the way the polling average has moved over the same period. There was a modest shift from Lib Dem to Labour around two weeks ago; since then, the figures for each party have been stable.

One clear reason for the decline in the projected Conservative majority flows from that overall shrinking of the Tory lead from 11 to 9 per cent. If the latest national figures are plugged into a simplistic, old-fashioned uniform swing model—in which the movement in party support since 2017 is assumed to be identical in every constituency, regardless of the differences between north and south, Leave areas and Remain areas, prosperous and struggling constituencies and so on—the projected result would be: Conservative 345 MPs, Labour 224, SNP 41, Lib Dem 18. This compares with YouGov’s detailed seat-by-seat projection: Con 339, Lab 231, Lib Dem 15, SNP 41. The small differences are well within the margin of error of any system of converting votes to seats.

It was a similar story last time. YouGov’s MRP projections in 2017 captured attention because they projected a hung parliament, while almost all the conventional polls pointed to a comfortable Conservative victory. But the main reason was that YouGov’s MRP surveys reported a national Conservative lead of only three per cent. Any votes-to-seats calculation would have shown a hung parliament. The big numbers were driven mainly by the overall structure of YouGov’s sample, not its MRP algorithm.

However, where MRP scored a palpable hit was the way it showed how massively different seats were behaving. A uniform swing calculation would have got the overall result right, while getting dozens of individual constituencies hopelessly wrong. YouGov’s MRP did far batter at identifying different trends below national level—correctly showing Labour ahead in Canterbury and Kensington, seats it had never won before, even in Tony Blair’s landslide years, while losing Mansfield, which it had never done before.

Likewise, the real interest in YouGov’s latest survey is what it shows is happening below the national surface. Let’s start with the overall seat numbers. A conventional uniform swing calculation of a two point reduction in the Tory lead from 11 to 9 per cent would point to an 18-seat reduction in the Tory majority, from 58 to 40. YouGov’s MRP projection indicates a 40-seat reduction in the majority. On these figures, Tories are slipping from victory to defeat in more than twice as many seats as we would normally expect: they have moved from outperforming to underperforming their uniform-swing projected majority.

This brings us to the really significant part of YouGov’s latest research. In the seats where the party was most narrowly ahead two weeks ago, it has lost most ground since then. To be more precise, Labour has gained most ground in the seats where it is breathing down the Tories’ neck, while Lib Dems have gained ground in some of their target seats, even while slipping back nationally. It seems that some voters in some areas are beginning to shift to the candidate best placed to defeat the Conservatives.

These movements are uneven and usually small. They are not happening everywhere. But they seem to be happening in enough seats to cost the Tories around 10-12 seats they looked likely to win a fortnight ago. This also helps to explain why YouGov’s projection for Lib Dem seats is up from 12 to 15, while the party’s nationwide vote share is down from 14 to 12 per cent.

The point about these trends is that the analysis holds good even if a number of YouGov’s individual seat projections turn out to be wide of the mark, as they were last time. If the third-place anti-Tory vote is being squeezed—and YouGov’s evidence is certainly plausible—then this will cost the Tories seats this week, even if they are not always the particular seats identified by YouGov.

This raises the intriguing possibility that tactical voting decisions, which many voters take only at the very end of an election campaign, could chip away further at the Tory majority, and possibly wipe it out completely, even if the final polls show little or no change in national vote shares. And there is of course a chance that there will be late national movement and/or polling error.

In these final hours, the possibilities are numerous—and they could move the dial in either direction. I am sure there will be surprises on election night; but what particular surprises they will be, I am not at all sure.

 

Peter Kellner is a political analyst and former president of YouGov. He did not work on the MRP or have access to internal methods