The British Election Study seems to focus on 'nationalized' politics, yet even the most cursory look at the data suggests that politics is playing out in deeply different ways across the country. For a start, the Conservatives - the ruling party, albeit in coalition - are barely represented in Scotland. Second, the Conservatives also win barely any seats in any major city outside London. A quick and dirty regression on English data alone tells me that just two variables - region, and the urban-rural classification of the constituency, explains 43% of the Conservative vote share at the constituency level (my guess is that increases a lot once we include Scotland, which I haven't entered the data for yet). Labour are a little less regionalized, but the two variable model still explains 34% of the variation. The Liberal Democrats, perhaps not surprisingly, present a more mixed picture (16% of variation explained).
So, class voting may be dead in the strictest sense, but we still know a lot about a voter by knowing which region they live in, and whether they live in a city, a town or suburb, or in a rural area. I don't know how much of this is an indirect measure of class (some of it must be), but my guess is that this indicates a high degree of endogeneity of the kinds of preferences that the 'valence' models use to predict the vote.
Politically, this is a big deal, since a coalition largely representing the wealthy South and suburban and rural Britain is slashing public spending in Northern urban areas with weak private sectors. I don't think this can end well.