Friday, May 07, 2010

Conservatives Vs Labour Vs Liberal Democrats ?

At the moment I am more interested in Nate Silver Vs the BBC

For decades the BBC has used a uniform swing model which predicts that vote shares are a function of a fixed constituency effect and a nationwide election specific effect. Alarmingly this works very well in the UK where, it seems, most voters don't change their minds.

Silver adopted a similar method to predict democratic primaries. His downweighting of old polls reflects the assumption of much greater variance of opinion shifts than the BBC but also much much lower variance than standard horse race coverage which stressed the campaign narrative.

Silver won the gold prize in 2008. Now he is playing in the international league.

So far, it looks like the BBC won. Their simple model predicted a hung parliament and, with most seats decided they are sticking with the prediction. Silver predicted a Conservative majority given the results of exit polls.

At the moment (or until recently) Silver still hopes to eak out a narrow win.

5:44 AM [Nate]. So, here's where we stand right now. The national vote is tracking to Conservatives 37.1, Labour 29.1, Liberal Democrats 23.7.

On those figures, a uniform swing model would predict Conservatives 296, Labour 258, Liberal Democrats 64; our model would predict Conservatives 333, Labour 224, LibDems 64. It appears that the actual results will be somewhere in between the two figures, with Conservatives short of a majority but north of 300 seats, perhaps somewhere in the neighborhood of 310-315.


He seems to be using GMT so that was 2 hours 27 minutes ago (EDT it would be 2 hours 33 minutes from now).

If the Conservatives were to win 315 seats that would be 18 less than predicted by Silver and 19 more than predicted by the BBC, so Silver still claimed a chance to win -- by one seat.

No comments: