Friday, August 10, 2007

Electoral Calculus is A Bit Rubbish

One of the things which always happens when polling information comes out is that someone feeds the numbers into Electoral Calculus and quotes its considered opinion to illustrate what this new information would mean for the next election.

So it was today, for instance, over on Iain Dale's blog. He posted a story about some Scottish polling data (mysteriously knocking 2% off the Lib Dems to begin with, now corrected), and within hours a comment appeared saying:
For Westminster seats electoral calculus converts this into SNP 51 (up from 6)seats and labour 8 (down from 41).

Lib dems & the other party are forecast to have nil.

As pointed out in another blog this in itself would eliminate labours UK majority at Westminster.
All very well and good, but when you look at the Electoral Calculus website, you do have to begin to question whether it really knows what it's talking about. Any Lib Dem will instantly have their eye caught by their prediction that we are going to be absolutely slashed at the next election to 19 seats. So click on the link to which seats they see changing hands. Pretty much all the changes are Lib Dem losses. And most of them strike me as being completely based in fantasy land.

From my own background, I am somewhat surprised to see David Howarth apparently on the list to lose to Labour. This being Cambridge, a town which shows no sign of tiring of its status as an island of pretty solid yellow in a sea of blue. I know there are boundary changes, but having looked at them, I don't think they pose a huge threat.

Others faced with the chop include Chris Huhne, Lynne Featherstone, Norman Lamb, Evan Harris, Michael Moore, Jo Swinson, and quite a few who I just can't see losing their seat. So I thought I'd conduct a bit of an experiment. I put in the results of the 2001, 1997 and 1992 elections. Perfect polling data, so would it get it right?

Well, here's a graph:


The white gradient-y ones are the genuine results, the solid ones the predictions from Electoral Calculus. In each case, it underestimates the party in power, and the Lib Dems. Now, I should mention to be fair that when you put the numbers in, it offers you the chance to put in a "Tactical Vote Swing" for each party, or in other words, to put in a fiddle factor to make things work right.

I sympathise with the difficulty of trying to predict elections, but really, is it such a good idea to authoritatively announce that the Lib Dems would lose so many seats on current polling data when, fed with real data from elections, the predictions it makes are wildly off, and always much too low?

2 comments:

Nich Starling said...

Electoral Calculus is absolute shit. It works on simple uniform swings and takes not account of anythin except national opinion polls.

Iain Dale used to quote it in leaflets in North norfolk in the vain hope it would persuade people to vote Tory. It didn't.

James Graham (Quaequam Blog!) said...

Any system that relies on a universal national swing is fairly useless at calculating what the Commons will look like. It isn't that national swing doesn't count for anything, just that it is only one factor.

Yet time and again, pundits and rival party activists get excited about what it suggests is going to happen in the next General Election. I have been confidently informed that the Lib Dems will lose seats from both Labour and Tory activists in 1997, 2001 and 2005. It's why the froth coming from Pravdale et al a couple of months ago about Ming was so misplaced.