Analysing MORI's 2009 data
Ipsos MORI have put the aggregated results of all their 2009 polling up on their website here (in fact it's been up for a week, but I haven't had chance to look at it properly!). The data is, of course, not particularly up to date (some of it would have been collected over a year ago), but the huge aggregate sample size provides us with some interesting cross breaks.
The overall shares of the vote for the whole of 2009 were CON 42%, LAB 26%, LDEM 19% - the equivalent of a 9.5% swing. If we look at different breakdowns by social class (comparing to MORI's aggregate figures from the 2005 election campaign) we find that ABC1s have a smaller than average swing to the Conservatives: ABs have swung by 6.5%, C1s by 7%. The larger Conservative swings are in C2s (10% swing) and DEs (13% swing). There's a similar pattern if you look at tenure - amongst those who own their homes the swing is 6.5%, amongst those with mortgages it is 10.5% and amongst those in local authority or housing association properties the swing is 13%.
This pattern of swing, with voters in groups that traditionally support Labour swinging the most strongly toward the Conservatives, produces a truly startling pattern when we get to MORI's breaks by type of seat. MORI have the Conservative lead in Lab-v-Con seats with a Labour majority of under 8.7% of 21 points. Depending on exactly what notional figures MORI used and how they treated three way marginals, that represents a swing of around about 12.5%. Looking at Lab-v-Con seats with majorities up to 13.9% the Conservative lead is still 21 points - suggesting an even bigger swing in those seats (somewhere around 14%).
If the Conservative swing is biggest in Lab-v-Con marginals it must be lower elsewhere. It isn't in safe Labour seats, MORI suggest a swing of 13% there. Part of it is Lib Dem seats, where the swing from Labour to the Conservatives is less than 1% (the swing from LD to Con is 7.4%, but I suspect that under-represents how well the Lib Dems would actually do). Where the big swings in Labour seats are really balanced out seems to be in the Tory heartlands - in seats the Conservatives already hold MORI's figures only suggest a swing from Labour of 5%.
Of course the Conservative lead has shrunk considerably since last year, but if a pattern of swing like this happened in reality it could hardly be more perfect for the Tories - tons of extra votes in the seats they need to win, but very few extra votes in the seats they already hold where they don't need them. To be honest though, while I've no reason not to trust the figures, it just doesn't seem believable. Previous elections have never shown differential swings of this degree, we simply don't get swings of 5% in one type of seat and 14% in another. Perhaps this is something different, perhaps this could be a real realignment election, but while other polls of marginal seats have shown bigger swings in marginals, none have shown swings this much bigger. Still, it's intriguing...
