Denseman on the Rattis

Formerly known as the Widmann Blog


Con 276, LD 123, Lab 218

FiveThirtyEight, an American election web site that was the place to visit for predictions about US elections (their coverage of the last presidential election was awesome) have started to make predictions about the UK election.

They are working on a different model than most British forecasters, and in general they’re forecasting a worse result for Labour, and a better one for the LibDems.

However, as far as I can tell, at the moment they’re just playing around with the model, so they’re just showing the outcomes of various scenarios.

In particular, I’m concerned they don’t seem to have any empirical basis for estimating voter movements between the parties – they just seem to make up numbers that add up to the current poll figures.

They’ve promised further blog postings in the near future, though, so hopefully there’ll be some really good forecasts soon.

Update (29/4): They have now updated their prediction to 299/120/199.

6 thoughts on “Con 276, LD 123, Lab 218

  • There was (is) a place that was basically a marketplace for bets on election outcomes for the American election that turned out to be very, very accurate. Unfortunately, I don’t remember what it was called.

  • Thanks! Unfortunately, it’s not very specific about what they mean by winning. I presume they mean ‘most seats’, but it’s by no means guaranteed the party with the most seats will be able to form the government.

  • Interesting. In the case of a coalition, the “winner” is the largest party in the coalition, not necessarily the biggest party in parliament.
    For instance, the Conservatives could get the most votes and the most seats, and Labour could be in third place for votes but second for seats and then be the “winner” if they entered a coalition with the LibDems.

  • Which makes sense, as Labour would then have the biggest word in the resulting government.


Leave a Reply

Your email address will not be published. Required fields are marked *