Friday, May 01, 2015

Election prediction using my model

The BBC's "Poll of Polls" currently looks like this:

Con: 34%
Lab: 33%
UKIP: 14%
LibDem: 8%
Green: 6%
Other: 5%

I fiddled around a little with my model, and came up with the following:

Defection Con-UKIP - 0.16

Defection Lab-UKIP - 0.11

Loss of votes for Lib when not incumbent - 0.75

Loss of votes for Lib when incumbent - 0.3

Lib transfer votes to green - 0.25

Lib transfer votes to UKIP - 0.1

Lib transfer votes to Labour - 0.3

Lib transfer votes to nationalist - 0.25

BNP transfer votes to UKIP - 0.8

Lib transfer to nationalist that go to labour - 0.5
if no nationalist candidate

Swing Lab/Con/Lib-SNP - 0.35

Swing Lab/Con/Lib-PC - 0.1

Greens standing in 90% - assume all
Defection Lab-Green for new candidate - 0.06

This resulted in the following outcome in terms of percentages and seats:


Con Lab LibDem SNP PC Green BNP UKIP
Percentage of vote 33.4% 33.7% 7.7% 4.6% 1.3% 6.2% 0.4% 12.7%
No of seats obtained 274 270 21 59 7 1 0 0
There are flaws in the model, I've spotted errors in the calculations. However, the numbers are satisfyingly close to the percentages that the opinion polls are returning with. I would like to redo the model more accurately, as I said before, but this is an interesting estimate to go with for now.

No comments: