Monday 10th May 2010 | 12:30

How did the Poll Centre seat projection perform?

The Poll Centre projection was the only global model to exactly predict the number of Conservative seats, though the Labour and Lib Dem seat totals were further from the mark.

The general election results were the ultimate test for seat prediction models – how did the Poll Centre projection perform?

The projection was exactly accurate on the number of Tory seats.  If the Conservatives win Thirsk and Malton, then the PoliticsHome Poll Centre projection will have correctly forecast the number of seats won by the party (307). Our headline result therefore accurately forecast that Tories would fall 19 seats short of a majority. 

However, the projection (which uses a statistical analysis of recent voting intention data) underestimated Labour performance and overestimated the number of seats that the Lib Dems would win. 

The projected Tory share of the vote, at 36.9%, was slightly higher than the actual share (36.1%).  The projection underestimated Labour’s share of the vote (projected 27.6%, actual 29.0%) while significantly overestimating the Lib Dem share (projected 27.2%, actual 23.0%).  The model had a lower projection for  the share of the vote gained by other parties (projected 8.3%, actual 11.9%). At least in part, this is because it did not include Northern Irish seats. 

The Poll Centre’s projected outcomes in individual constituencies were correct in 86% of seats.

Rob Ford, co-author of the Poll Centre methodology, comments:

The model did perform well, although there was a large slice of luck involved. We were in fact wrong to assume that the Tories would outperform in the marginals, but this was balanced by Lib Dem underperformance everywhere to deliver roughly the right result.

We did see very clear patterns of differential swing in Scotland, as we predicted, although the differences were even larger than the polls had suggested. There were also differential patterns in Wales and in seats with large ethnic minority populations. These would both have been near the top of my list of expected differential effects, but we had no polling evidence on them so did not incorporate them in our model.

Of course, the big failure of the polls was the massive over-estimate in Liberal Democrat support. This had less effect on our model as our uniform swing based prediction gave them less chance of winning over a large number of seats even on a high swing.

By giving extra weight to the last day's polls, we also reduced the error on our forecast by successfully detecting the late swing to the Conservatives and Labour evident in these polls - our final forecast upgraded the Conservatives' share by 2 percentage points and Labour by 1 percentage point.

The big story, though, with regards the UNS vs differential swing debate is that the pattern of swing was remarkably uniform:

The change in Conservative vote varied by less than two percentage points moving from their weakest to their strongest areas, and they actually underperformed somewhat in their weakest areas relative to the average

The change in Labour vote varied somwehat more, but there was no systematic relationship with prior strength - if anything the party performed worse in areas where it started off somewhat weaker.

The change in Liberal Democrat vote showed more evidence of proportionality, falling back three points in the strongest areas while rising in the weaker areas. But even here the evidence of proportional swing was weak and patchy at best.

Given the lack of any clear relationship between prior strength and outcomes, we would expect proportional swing based models to perform quite poorly, and so it has proved.

By contrast, the PoliticsHome Poll Centre model performed very well given the poor performance of the polls, and the exit poll analysis run by the BBC, which built in demographic and political predictors of deviation from uniform swing, also performed extremely well. At 10pm on election night, it issued predictions for each of the main parties which look to come very close to the outcome, although at the time of publication they were regarded as very suprising and implausible.

Visit the Poll Centre. 

The final Poll Centre projection (not including Northern Irish seats):