peterbirks: (Default)
peterbirks ([personal profile] peterbirks) wrote2017-06-07 10:26 pm

Thoughts on an election, week seven

The latest opinion polls have led to a strengthening in the pound. With two polls showing a double-digit lead again for the Conservatives, those who should know better are talking of a 100+ majority.
But let's look at the trends within particular polling methods.

ComRes's last three polls have been:

ICM's have been:

SurveyMonkey's have been:

Opinium's have been:

Meanwhile, YouGov's have been rather constant at 42:38

So, even over the past 10 days, all the pollsters seem to have the Conservative vote as fairly solid, and the Labour vote either flat or increasing slightly. The situation remains: which pollsters do we believe? Or, alternatively, how many 18 to 34 year olds will turn out tomorrow? 50% (2015) or 64% (referendum) or even more?
With about 70% of this demographic favouring Labour and only 16% voting Conservative, a 15pp increase in turnout would make something like a 3pp difference in the result (which, coincidentally, roughly reflects the difference between the predictions of ICM and YouGov).
I'm going to nail my colours to the mast here and say that my personal feeling is that YouGov has got it right.

Now, Nate Silver wrote an interesting article earlier this week entitled "Are the polls skewed?"
Silver debunks the myth that the polls always underestimate the result because of "shy Tories". He doesn't deny that there are shy Tories; what he points out is that they are only one factor.
He also points out that, although the Conservatives have outperformed the polls 6 times out of the last 7 elections (but only 12 of the last 19), the pollsters might have overcompensated for this. Now, when you work out that the Conservatives have not really outperformed the polls since polls came into being, and that in the past 10 elections when the Conservatives were in front leading up to the election, they *underperformed* the polls in six of them, any assumption that the Conservatives will outperform the polls again is based on very dodgy foundations.
Silver also points out that the pollsters have come up with different reasons for poll misses at different elections. In 2010 they blamed a late swing away from the LibDems. In 1997, 2001 and 2005 they blamed their overestimation of the Labour victory on a low turnout.
As Silver rather pointedly asks "Could it really be a coincidence that all these different errors in all these different elections just so happened to underestimate Conservatives?"
The important point is, the pollsters have "learnt" from their mistakes in 2015, but they have not all learnt the same thing. It could be an overcompensation, it might be spot on.

Anyhoo, for good or evil, I'm going with the tendency to overcompensate and the fact that the Conservatives are ahead to conclude that most of the pollsters have overestimated the Conservative vote, perhaps by 2pp, and underestimated the Labour vote by the same amount. That puts us in the YouGov ballpark of 42:38.

What result does that give us?
Well, it gives my spreadsheet the following:
Cons 327
Lab 242
LibDems 12
SNP 46
PC 3
Green 1
Speaker 1
NI 18
... for an overall majority of a thumping 4.

I always find it funny when politicos "like" anything I post when it supports their party. This is nothing to do with the validity of the analysis -- they just like the result and don't really care how I came by it. if the same methodology had come up with a different result, they would like the methodology less. Go figure.

Now this comes with an important caveat. I'm inserting a big differential in the LibDem vote, not as I originally planned because of Remain vs Leave (that is still there, but muted) but in terms of seats which were LibDem in 2010 and can be won back again, or seats that were won by LibDems in 2015 anyway. If this does not transpire, LibDems could shrink to 6 seats and the Conservative majority would climb to 16.

Now, I am perfectly aware that when I get this wrong, there will be no shortage of people telling me why I got it wrong. That's what always happens in the FX markets and the stockmarkets. It makes one amazed that there aren't more millionaires out there. Everything is obvious after the event. But if you didn't make money on it, I'd have to ask why you waited until afterwards to say why it was obvious.

So I've gone even madder. here are some predictions. These are not the same as my bets. My bets are all about value and I don't think I have had a single bet at shorter than 4/7 (two bets, Lab to win Vauxhall and Lab to win Rhondda).

In Scotland Con to gain:
Aberdeen South
Berwickshire Roxburgh & Selkirk
Dumfries & Galloway
Ochil and South Perthshire
West Aberdeenshire

Lab to gain from Con
Bolton West
Brighton Kemptown
Bury North
Croydon Central
Derby North
Morley and Outwood
Plymouth, Sutton and Devonport
Weaver Vale

Con gain from LibDem
Possibly Carshalton -- very close to call

Lab gain from LibDem
Possibly Sheffield Hallem -- very close to call

My own bets (when not mentioned above) and which I haven't written off as doomed (i.e., I still stand by them)
Cons to win Birmingham Northfield
Lab to win Halifax
Lab to win Wakefield
Lab over 218.5
Lab over 177.5
Con under 399.5
Con under 370.5
Con under 337.5
Turnout 60-65

Written off bets
Cons to win Tooting
Cons to win Ynys Mon
Cons to win Cardiff South
Cons to win Hove
LibDems to win Bradford East
LibDems to win Birmingham Yardley
LibDems to win Bermondsey (still holding out an irrational hope that Simon Hughes might buck the odds).

Final seat predictions again:
Cons 327
Lab 242
LibDems 12
SNP 46
PC 3
Green 1
Speaker 1
NI 18
Overall majority - 4

Post a comment in response:

Anonymous( )Anonymous This account has disabled anonymous posting.
OpenID( )OpenID You can comment on this post while signed in with an account from many other sites, once you have confirmed your email address. Sign in using OpenID.
Account name:
If you don't have an account you can create one now.
HTML doesn't work in the subject.


Notice: This account is set to log the IP addresses of everyone who comments.
Links will be displayed as unclickable URLs to help prevent spam.