Election prediction

by Daniel on May 5, 2005

Quickly quickly, here’s my prediction. I’m using vote shares from the IG betting market, because I think that “The Wisdom of Crowds” probably works quite well for mass estimation problems like this, but maybe not so good at projecting the results of its mass estimation onto a difficult electoral college problem. So I’m working on the assumption of 32.5% Conservative, 37% Labour and 24% LibDem and using my own Allocated Regional Swing model, documented here last week.

That gives me the following seat predictions:

Labour: 388
Conservative: 190
LibDem: 53.

Labour majority 130, and presumably Blair decides that it was a referendum on the war (and the lying about the war) after all.

Note that the LibDems get royally screwed by first-past-the-post; they get a swing of 5% and pick up three seats. I have a really ace triangle plot showing this but I don’t know how to upload images on the new WordPress site (Update: thanks Henry!)
Excel rules ok
I’ll do a proper post after the election explaining how I got it so wrong. Meanwhile, below the fold is my list of possible seat changes; it’s longer than Martin Baxter’s, but this is mainly because the list is drawn from a slightly different model; I wanted one as long as would be possible consistent with my overall predictions to be a bit more interesting.

Now I’m off to vote (LibDem, if anyone cares. Sorry Dobbo, you’re a really nice guy but you’re not standing for the Frank Dobson party. You’re standing for the Labour Party and that means Blair).

Argyll and Bute-LAB-Gain from-LIB
Bexleyheath and Crayford-CON-Gain from-LAB
Braintree-CON-Gain from-LAB
Brecon and Radnorshire-CON-Gain from-LIB
Bristol West-LIB-Gain from-LAB
Cambridge-LIB-Gain from-LAB
Cardiff Central-LIB-Gain from-LAB
Ceredigion-LIB-Gain from-NAT
Cheadle-CON-Gain from-LIB
Clwyd West-CON-Gain from-LAB
Colne Valley-LIB-Gain from-LAB
Cornwall South East-CON-Gain from-LIB
Dorset South-CON-Gain from-LAB
Edinburgh Pentlands-CON-Gain from-LAB
Folkestone and Hythe-LIB-Gain from-CON
Forest of Dean-CON-Gain from-LAB
Gordon-CON-Gain from-LIB
Greenwich and Woolwich-LIB-Gain from-LAB
Hammersmith and Fulham-CON-Gain from-LAB
Harborough-LIB-Gain from-CON
Harrogate and Knaresborough-CON-Gain from-LIB
Hereford-CON-Gain from-LIB
Hornchurch-CON-Gain from-LAB
Inverness East Nairn and Lochaber-LIB-Gain from-LAB
Kettering-CON-Gain from-LAB
Lancaster and Wyre-CON-Gain from-LAB
Leeds North West-LIB-Gain from-LAB
Liverpool Wavertree-LIB-Gain from-LAB
Maidenhead-LIB-Gain from-CON
Milton Keynes North East-CON-Gain from-LAB
Monmouth-CON-Gain from-LAB
Newbury-CON-Gain from-LIB
Norfolk North-CON-Gain from-LIB
Northampton South-CON-Gain from-LAB
Selby-CON-Gain from-LAB
Sheffield Hallam-CON-Gain from-LIB
Shipley-CON-Gain from-LAB
Southport-CON-Gain from-LIB
Tatton-LIB-Gain from-CON
Thanet South-CON-Gain from-LAB
Welwyn Hatfield-CON-Gain from-LAB
Weston-super-Mare-CON-Gain from-LIB
Ynys Mon-NAT-Gain from-LAB



nick 05.05.05 at 7:20 am

Folkestone and Hythe-LIB-Gain from-CON

Michael Howard’s going down?


Alex Fradera 05.05.05 at 7:24 am

Well, I just voted for frank, so from the blog’s eye view it’s neck and neck, with honestly everything to play for. And should my man win, it will be proof of the power of the internets, and I will be on the right side of history!

Today is medication-free day.


Keven Lofty 05.05.05 at 7:57 am

Michael Howard’s going down?

Now that would be amusing.

For what it’s worth I made a bet a couple of days ago for a 132 Labour Majority. Not quite so confident now, but you have to stick with your convictions.


RS 05.05.05 at 8:11 am

I was fiddling with the BBC seat calculator http://news.bbc.co.uk/1/shared/vote2005/seatcalculator/html/default.stm and I put in the most recent opinion poll of the time (37/32/24/7), and by fiddling a bit with tactical voting in marginals, I get something like 70-90 majority for Labour, and 60 odd seats for the Lib Dems.


sharon 05.05.05 at 8:54 am

Howard losing is a possibility but not *that* likely. The ones to watch are Oliver Letwin (majority 1414) and David Davis (majority 1903). Neither of whom appear to be on your list?


Simon 05.05.05 at 9:01 am

Labour will probably regain Leicester South from the Lib Dems. It’s natural Labour territory and the Lib Dem MP – who won the seat in a byelection last year – has been virtually invisible.


James 05.05.05 at 9:01 am

Greenwich & Woolwich to switch to Lib Dem? That is a major swing with Lib Dems from 3rd to 1st. I voted for the Lib Dems in this constituency, but I did’t really believe it will make any difference. If you are right, poor old Nick Raynsford. Good guy in a crap position.


Daniel 05.05.05 at 9:39 am

Letwin and Davis keep their seats in my model but only by like 500 votes. Wouldn’t be surprised to see total decapitation; I think that a fiver each on Redwood, Hague and Bojo might be quite a decent investment.


Thomas L. Tompkins 05.05.05 at 9:50 am

According to this Labour will lose 15 seats, but Conservatives & Lib Dems will gain 27. Do you think that the smaller parties will actually do worse, or is a problem with the model?

Also in the list of seats that will swing, you only mention PC losing one seat, and gaining another. Where are the other 12 that the other parties will lose?


Kevin Phillips Bong 05.05.05 at 10:08 am

Now I’m off to vote…

I thought voting was for saps?


john 05.05.05 at 10:38 am

Wow, that BBC calculator is crazy – to get the Liberal Dems to have the most seats, they have to beat Labour by 10 points (and the Tories with almost as much). Jesus. What are the assumptions behind it?


Daniel 05.05.05 at 10:40 am

Scotland mainly, and mainly due to boundary changes. Though it is certainly a problem with the model that I basically assume zero swing in the Nat and minority party vote.


Dominic Murphy 05.05.05 at 10:47 am

What happened to your once-firm avowal that you were gonna suck it in and vote Tory, that being the most effective protest vote, all things considered? (I’m not trying to be obnoxious – I’m actually curious about why you changed your mind.)

And any particular reason for giving my home constituency (Ceredigion) to the LDs, or did the model just happen to decree it? (I haven’t lived there since I came to the US, so I don’t have any intuitive basis for a prediction myself.)


Daniel 05.05.05 at 10:49 am

Ahhhhh I see. I have 10 seats which are recorded in my model as having been Labour in 2001, but which didn’t actually exist, and two ghost Tory seats. Martin Baxter’s pivot table deals with these boundary change effects in some clever way but my model was based on the raw data underlying it. Sorry fans, but if anything this should improve the accuracy of the aggregate projections; I think I would actually have chosen to do it this way if I had realised in time.

John: it’s not the BBC calculator that’s crazy. It is correctly representing a crazy electoral system.


Daniel 05.05.05 at 10:53 am

Dominic: basically because the Tories have no chance in Holborn & St Pancras, where I now live. When I wrote that post I was living in Oxford West.

Ceredigion is likely to be a crappy prediction looking at it; the LibDems are the biggest of the three main parties, and they get a positive swing locally. I assume a zero swing for the Nats, so they suck up enough Labour and Tory votes to win it. I doubt that they will get it, though I reserve the right to boast like billyo if they do.


urizon 05.05.05 at 11:17 am

Blair will win, and then Labour will chuck him overboard. It could take weeks, perhaps months, but it will happen sooner or later.

Blair is done; Labour marches onward.


Daniel 05.05.05 at 11:52 am

Greenwich & Woolwich is a pretty wild prediction but if the model works at all (yes yes yes) it’s going to work in these way-out seats. Remember that this was Rosie Barnes’ seat for the SDP back in the day and they were the second biggest party in 1992; the reason that their support is so low down in 2001 will have a lot to do with tactical support of Labour which can’t be counted on.


RS 05.05.05 at 12:05 pm

“Dominic: basically because the Tories have no chance in Holborn & St Pancras, where I now live. When I wrote that post I was living in Oxford West.”

So you would have voted against a sitting Lib Dem MP in favour of the Tories?


Daniel 05.05.05 at 12:08 pm

Yes, mainly to annoy my brother-in-law.


des von bladet 05.05.05 at 1:00 pm

Yes, mainly to annoy my brother-in-law.

Now _that_’s expressive voting!


Simstim 05.05.05 at 2:12 pm

Perhaps the *real* post-left/right split in politics will be between the “annoy your in-laws” and “annoy your blood family” tendencies?


pedro 05.05.05 at 2:27 pm

I can’t wait to be able to annoy my in-laws. (Still not a U.S. citizen.)


nikolai 05.05.05 at 3:58 pm

Wow, that BBC calculator is crazy… What are the assumptions behind it?

The assumption’s just a uniform swing.

I voted Labour. But I really hope we don’t get one of the insane majorities that are being floated around here. A majority in the 100s will let then do all sorts of stupid things, whereas if they’re kept to below 40 at least it’ll put the handbrake on some of their crazier ideas. A large majority – and with the Tory crippled and offering no opposition – will be an absolute curse.


RS 05.05.05 at 5:41 pm

Exit polls suggest no insane majority for Labour (66 on the Beeb), and the Lib Dems not doing too great (as always seems to be the case with our barmy electoral system).


urizon 05.05.05 at 5:45 pm

I happen to be a SUNY student. In my experience — at SUNY New Paltz, anyway — Horowitz’ work has already been done for him. More than fifty percent of the New Paltz faculty are untenured adjuncts who make approximately two thousand dollar for a 16 week course. The number of tenured professors is shrinking every year, and the way things are going, there will be a couple of tenured professors in each department with dozens of adjuncts doing all the heavy lifting for a pittance.

Now we’ve got that mendacious, fascist fuck Horowitz going after tenure within the SUNY system. What tenure, David, you pea-brained douche-bag?


Nice when two different agendas dove-tail so seamlessly into one another, huh?


urizon 05.05.05 at 5:45 pm

Oops. Wrong thread. Hehe.


Lee Bryant 05.05.05 at 6:10 pm

388 for Labour is, I think, an overestimate. The ‘anti’ effect and a slight increase in turnout might yet combine to reduce their majority more than your guess implied.


g 05.05.05 at 8:29 pm

Well, the Lib Dems have indeed taken Ceredigion…


dominic murphy 05.05.05 at 8:41 pm

Indeed. Boast away, boyo.


John Quiggin 05.05.05 at 9:31 pm

I just saw a report that George Galloway had beaten Oona King. This seems a pretty dreadful result, considering what I’ve read of the relative merits of the candidates. Still, it will presumably cause Blair more discomfort than it causes me.


ab 05.05.05 at 9:48 pm

It’s now 4am British time, and it looks like Labour holds on to a majority anywhere between 30 and 80 seats.

Galloway beating Oona King would be a disaster; I hope it’s not true.


Daniel 05.05.05 at 10:15 pm

Well it’s 4am and I’m going to bed. My predictions look for shit. But Martin Baxter (who I may have referred to on IRC as “the hated Baxter” which was silly; he’s a good guy and his book on martingale theory is excellent, so sorry) looks just as bad on the overall and I think slightly worse on inividual seat changes. It looks like LDs have done about as bad as I thought in two-way battles with Tories and much, much better in battles with Labour. See ya in the morning. (btw, I still think Labour hold BG&B)


ab 05.05.05 at 10:32 pm

4.25 am — 324 seats — Labour WON — hurray!


ab 05.05.05 at 10:39 pm

F&CK! — 4.35 am — Galloway just won Bethnal against Oona King — What a tragedy!


nick 05.05.05 at 10:55 pm

Well, Galloway’s cockheaded exchange with Paxman (who was almost equally cockheaded, I must say) is one for the records.


KCinDC 05.05.05 at 11:26 pm

I just heard that interview, Nick. Yikes! I would like our American journalists to be less obsequious, but I don’t think that’s the model to follow. Where can an interview go with an initial question like “Are you happy you knocked a black woman out of Parliament?”


dsquared 05.06.05 at 3:58 am

Btw, this looks pretty good for the betting markets. They missed Bethnal Green pretty badly, but they got the size of the majority and the LibDem seats pretty much spot on when all simple modelling approaches were a mile out.

Comments on this entry are closed.