This has been the first UK election in which we have had results from a considerable number of constituency polls, giving us an indication of where the gains are likely to be. However, there are not constituency polls everywhere and sometimes it is worth taking a detailed look at what we can get out of the national data.
A new Guardian analysis of ICM’s campaign polls suggests Labour is performing no better in the marginal seats of the English battleground than it is doing nationwide.
ICM telephone polls have come closest to predicting the final result in three of the past four general elections, and during the campaign so far they have produced worse numbers for Labour than many other firms. Where other pollsters have often shown the race as neck and neck, or have recorded Labour slightly ahead, ICM has chalked up three successive leads for the Conservatives of 6, 3 and 2 percentage points.
Averaging across these results suggests a nationwide swing of nearly 2% – that’s the decline in the Tory share and the rise in Labour’s score since 2010 added together, then divided by two.
In the English and Welsh battleground seats – defined here as seats the Conservatives won by no more than 15 percentage points last time, or Labour won by no more than 10 points – the swing is somewhat smaller. Where the Tories bested Labour by 38% to 36% in 2010, the pooled poll data suggests that the two parties are now tied on 38% apiece, which implies a swing of only 1%.
On first glance, this suggests relatively few gains for Labour next week. Drilling down into the data reinforces the impression that few seats may change hands, because Labour appears to have made even less progress in the Conservative-held marginals than in those it already holds.
The numbers in these sub-samples is small, even with pooled data, so the precise swing should not be taken too seriously – there is a margin of error of about 7 points. However, it shows there is unlikely to be a Labour landslide (although we knew that already).
Yet the distribution of the votes is key, as well as the margin of error. The latest Lord Ashcroft constituency polls showed that there are likely to be Conservative seats going to Labour – even in Peterborough, where the Tories had an 11-point lead in 2010.
Both the Conservative and the Labour heartlands appear as safe as ever – the Tories are sitting pretty on 51% in the seats they won by more than 15 points last time, which is exactly what they averaged in 2010.
Labour has piled on 4 points in those seats it won by more than 10 points, to rise from 47% to 51%. Both these results suggest some resilience of the two main parties on their home turf to the challenge of Ukip, and indeed the Greens.
Within the context of their nationwide collapse in support since 2010, many Liberal Democrat MPs are pinning their hopes on the vote holding up better in their own back yard. In one sense, the pooled polling suggests some support for this: the proportional drop in Lib Dem support in these seats is rather less than a third, compared with more than half in most other categories of seats, where the party is now running in single figures.
In another sense, however, the results ought not to reassure the party. In Lib Dem seats there were obviously many more Lib Dem votes in the first place, and the absolute number of votes they have lost is actually bigger than elsewhere. The party declines by an average of 18 points, from 47% to 29%, in its English and Welsh seats.
The many Scotland-only polls give a much more authoritative impression than the small number of respondents here. But it is striking that in the pooled Scottish sub-sample, which contains more than 150 respondents, the Scottish National party is running at over 50%, and the collapse in Labour support is so great that it has actually dipped behind the Scottish Conservatives.
This analysis was possible using a novel technique. The pollster uses postcodes to identify which constituency its respondents will be voting in. It then groups the seats into different political categories – safe Labour, marginal Labour, Liberal Democrat-held and so on.
In any one poll, the number of respondents in each of these categories can be too small to produce reliable results, but insights can be gained by pooling the figures from all three of the recent campaign polls, which the Guardian has been reporting on a weekly basis since 13 April.
