[*] among many

Mr. Mouse: remember that this is not a causal model; I’m describing a statistical relation here, which (I think) has the interpretation given above. There are plenty of positive swings to LD which have them gaining seats; the split 33C, 37L, 25LD which I am looking at on my screen right now has them picking up 12 seats. They still lose Sheffield Hallam even in this scenario though, because in my model it is basically a Tory seat.

But the NI parties have close to fuck-all influence on the larger British politic, with the one exception being when Gerry Fitt was stupid enough to vote yes on the no-confidence motion against Jim Callaghan in 1979, thus helping to usher in Thatcher (another six months, and Labour might have recovered in polls sufficiently to win an election). And Plaid are competitive in only 2-3 seats.

So, with the exception of the Scottish Nats, I can’t see why nats would improve Daniel’s ARSE.

I think this points up a fundamental problem with your model. I can conceive of circumstances in which an increase in a party’s national vote share actually decreases its support in a particular constituency in some causal way, but they’re pretty bizarre.

Mind you, that sort of intuition has led me up the garden path before so I could be wrong here. (E.g. I still feel strongly that using imaginary numbers to solve real-world differential equations is Bad and Wrong, but the bridges stay up despite my qualms.)

I think what’s actually going on is the well-attested tactical vote/targetted seat campaigning effect, which gave the non-linear result for the LibDems last time round, and which might give entertainingly non-linear results this time.

In the spirit of the LazyWeb, I’d like to see modelling that takes on the tactical vote issue explicitly. I imagine some fiendishly clever feedback mechanism that magnifies the expected results (poll data?) in a constituency in to a propensity for anti-Tory tactical voters to go one way or ‘tother.

I think this time round we can expect anti-Labour tactical voters in a way we’ve not really seen before, and hence about whom we have insufficient data so cannot make robust predictions. That’s never stopped people before, though!

If you ignore the nationalist parties, then NI is in play. It won’t be, however; in all likelihood, NI will be split between DUP and Sinn Fein, with UUP and SDLP being splattered.

(I don’t know enough about Scotland or Wales to comment, so won’t.)

I quite agree on the Nashies, but it really was too much trouble. I may chuck in another adhoc fix.

I don’t know if I buy this anyway – there is no shift from Labor to Lib Dem at all? So the protest vote angle won’t work out at all?