Officials count ballot papers in WitneyA couple of months ago, I wrote:

It seems that UKIP are very good at getting out the vote, but they’ll need to broaden the number of people willing to vote for them to have a serious chance of winning a Westminster seat.

As they’ve now won a Westminster seat, let’s look at the evidence from the by-elections to see what happened.

As I said in that last post, the interesting thing to look at in the UKIP vote isn’t the share of the vote, but the share of the electorate. We spend an inordinate amount of time comparing the share of the vote in elections without pointing to the fact that we’re often looking at vastly different numbers of total votes being cast. For instance, Labour’s share of the vote went up slightly (40.1% at the 2010 general election to 40.9% on Thursday) in Heywood and Middleton, but that masks the fact that the number of votes cast for them dropped by almost 7,000 (18,499 to 11,633). In 2010, they had 23% of the electorate voting for them, last Thursday just 14.7% did.

UKIP’s share of the electorate in Heywood and Middleton was 13.9% – slightly ahead of how they did in Newark, but behind their previous high-water mark of 14.7% in Eastleigh. It’s now at third place overall for them though, because of the Clacton result. Douglas Carswell not only got well over 50% of the vote, he got the support of 30.4% of the electorate. In contrast to Labour in Heywood and Middleton, he was only around 1,700 votes short of the total he received in 2010 (22,867 to 21,113). In percentage terms, his vote on Thursday was 92.3% of his 2010 total, Labour’s in Heywood and Middleton was 62.9%.

So, Heywood and Middleton was around the top of the range we’re currently seeing for UKIP votes, but Clacton was well off the scale. Carswell’s 30.4% was more than double the highest share of the electorate UKIP have received before (14.7% in Eastleigh) and if he can retain that share, he likely will retain the seat next May.

The question is whether UKIP can repeat this feat in other constituencies. Thanks to Ford and Goodwin’s research (if you haven’t read Revolt on the Right yet, you really should) we know that Clacton is the most demographically favourable seat for UKIP. So, we would expect the UKIP share of the electorate to be higher in Clacton than anywhere else, but the question is whether that alone would explain it.

I don’t have all Ford and Goodwin’s data to see what the difference in demographics is, and how much that might explain the change in vote. As I see it, that’s one of three factors likely to predict possible UKIP success, along with the general level of support for UKIP and local support for the candidate. The ideal way to test this would be through an experiment where we ran an election in another constituency with different demographics but with the other two variables either the same, or easily measurable. Incredibly, Mark Reckless’s defection gives us exactly the chance to do that. If UKIP support in opinion polls is about the same at the time of Rochester and Strood as it is now, and the UKIP candidate is also a sitting Conservative MP who’s defected, then the different demographics of the seats should play an important part.

Of course, this could be completely wrong, and the Clacton result might be more easily explainable because of the level of local support for Carswell rather than the local demographics. Other results still seem to be holding up the idea that UKIP have a ceiling of support amongst the electorate, though that segment of the electorate – Ford and Goodwin’s ‘left behind voters’ is heavily concentrated in Clacton. At the moment, we only have the one data point of Clacton to suggest UKIP can win a seat when the turnout gets higher, but Rochester and Strood will give us some useful extra information.