Saturday, April 24, 2010

Not Got a Clue... And Nor Has Anyone Else

It's been a great day so far. Lovely sunshine, a fantastic morning's campaigning in Dagenham & Rainham, West Ham winning at home and more or less saving themselves from relegation, now a couple of hours at the City Inn catching up, before doing the BBC News Channel paper review at 11.20pm. Every day should be like this!

Of course, the first thing I have done is look at the polls. And boy are there a lot of them. Here are how the three main parties fare...

CONSERVATIVES
BPIX (Mail on S) 34
YouGov (S Times) 35
ICM (Sun Telele) 35
OnePoll (People) 32
IPSOS MORI (Now) 36
ComRes (Indy oS) 34

LABOUR
BPIX (Mail on S) 26
YouGov (S Times) 27
ICM (Sun Telele) 26
OnePoll (People) 23
IPSOS MORI (Now) 30
ComRes (Indy oS) 28

LIBERAL DEMOCRATS
BPIX (Mail on S) 30
YouGov (S Times) 28
ICM (Sun Telele) 31
OnePoll (People) 32
IPSOS MORI (Now) 23
ComRes (Indy oS) 29

Now, if you average those out you get

Conservative 34.8
Labour 27.4
Liberal Democrat 28.0

So with only ten days to go the LibDems seem to be slightly dropping but not by much. But the one more or less consistent thing in these polls is that Labour is in third place in the popular vote (in 5 out of six of the polls).

The other consistent thing in these polls is that they virtually all point to a hung parliament with the Conservatives the largest party. However, and this is what should spur on every Tory activist, the party is only two or three per centage points from where it needs to be to get a majority. And if it is true that much of the LibDem surge is coming in the North and among the youth vote in university towns, the swing needed for the Conservatives to get an overall majority may be less than some people think.

But having said that, I may as well have not written that paragraph, because like any other political commentator at the moment I am floundering around, grasping at straw, because the fact is that none of us have any idea what is going on. And if we make a prediction which comes true, it is certainly more luck than judgement.

15 comments:

Red Rag said...

Jesus,
a post of yours I agree with....I need to go and lie down.

Senn the Cartoonist said...

10 exciting days...

Irene said...

The answer is Prime Minister Cameron!

Alan Douglas said...

"And if we make a prediction which comes true, it is certainly more luck than judgement."

Or you are Vince Cable, predicting equal and opposite results within days of each other, so can claim to be "always" right. Like the doc who would predict "It's a boy" and write in his diary "girl".

Alan Douglas

Cantstandcant said...

Iain, unless I'm missing something here, or you are basing this on sample sizes that don't seem to be in the public domain yet, you need a new calculator.

I make the respective averages:

Con 34.3 (not your 34.8)
LD 28.8 (not 28.0)
Lab 26.7 (not 27.4)

So the Con/LD gap is 5.5pc, not 6.8pc. And the LD/Lab gap is 2.1, not 0.6. Small but undeniably significant differences, unless my Casio calculator and mental maths is in a bad way.

And if you remove the possible rogue/outlier from the NoW, you get Con 34, LD 30, Lab 26.

Type even the first into the BBC seat calculator and you find DC about 50 short of a majority, so perhaps not quite as rosy a picture as you suggest.

The King of Wrong said...

@Alan Douglas:
That's a bit harsh. After all, Vince Cable predicted all 17 of the last 2 recessions...

(w/v "fled cat" - Macavity?)

richard.blogger said...

Iain you are right that no one quite knows what the polls mean. However I would caution you from simply taking an average, that is a simple mistake to make from someone not used to looking at trends. I would recommend that you plot the polls and look at median values (possibly removing clear outliers). But more important is the change in the polls *the trend*. ukelectiontrend/ do this, and give a fascinating picture.

1) Since October Conservative support has dropped at a rate of one point per month.
2) Since mid-October Labour support has risen at a rate of 0.75 point per month
3) Lib Dem support was flat from Nov to March

This is quite interesting, the polls usually tighten before the polls but why was the LD support flat for so long? The changes are important, but more important are the median values. Nov 09 these were Con 41, Lab 29, LD 19. Given the rates given above the May figures should be Con 34 and Lab 34. Hmmm.

So what has happened?

1) During March LD support has risen by 3 pts per month
2) During March Lab support has dropped by 2 pts a month.
3) Over March Cons support dropped at the same rate as before

Clearly there was a switch in March from Lab to LD

April is the interesting month.

1) Cons support started to go up from the beginning of the month to a peak at the point of the manifesto launch. Then it dropped 2 pts in two days. Then there was the debate.
2) Lab support flattened off between the manifesto launch and the debate.
3) LD still rising at the same rate as before, with a slight dip at the manifesto launch. Then a rise after that.

What was clear here is that the public switched support away from the Cons when the manifesto was launched. In short, the manifesto was a big turn off: the drop in Cons support is very marked. Lab support flattening means that the response to the manifesto was neutral. Not good, but at least it is not as bad as the decline the Cons suffered. Basically the Lib Dem rise is now coming from Cons. Now for the debate.

1) Cons support is still dropping, at a rate of 7 pts in 14 days.
2) Lab support is dropping at the same rate, but has 4 days less drop than Cons
3) LD rise is 7 pts in 14 days.
4) The LD polls are much more scattered than the other two, which implies that there is a lot more inaccuracy.

What is worrying is that neither Cons nor Lab show a tailing off of the decline. Both are declining at the same rate. The difference between the two is 5 pts.

Question - what is the Labour core vote? I have seen it suggested to be 25%, but I have no idea if this is the case. If that is the case then Lab support decline can continue for four more days and won't go any lower.

Question - what is the Cons core vote? I have heard it to be 30%, in which case Cons trend can decline for four more days and won't go any lower.

The trends since the Cons manifesto launch show falling support for Cons and Lab at the same rate, and consistently. If they continue at these rates then at the election they could be Cons 24 Lab 19. I think there is little chance of that.

That's as far as my analysis of the trends goes.

wild said...

The conventional wisdom that because Cameron has an (at times almost embarrassingly) higher IQ than Nick Clegg or Gordon Brown, he should pull his punches, in order to avoid looking too slick and arrogant in the eyes of potential Lib Dem voters, who instinctively root for the less able and chippy younger brother.

I think this advice is wrong.

Cameron is never going to win over the chippy vote (or for that matter the airhead vote). He should concentrate on those who want their Prime Minister to be the most able man in the room.

I sense that while Cameron (rightly) despises Brown he has to fight off the desire to be polite (or if you like "patronise") Nick Clegg.

He needs to remember that Clegg's politics may be adolescent, but they token a narcissism that is actually quite repellant in sombody in his 40's.

Cameron needs to view Clegg not as a youthful idealist, but as yet another middle class "legend in his own mind" Euro Federalist who will keep the Leftist elite in power for another five years, indeed if he can get away with it (by changing the electoral system) in perpetuity.

A totalitarian social democracy in which an elite lecture and feed off the decline of a country about which they neither know nor care.

Unknown said...

You're absolutely right, and it's quite extraordinary that, 11 and a half days out from a general election, you can find pretty much every single election result being predicted by someone - Tory win, Labour win, even Lib Dem win, and everything inbetween. The British media, who are obsessed with trying to tell people what will happen before it's happened, are having huge trouble accepting that they just don't know. I'm loving it.

By the way, it'll obviously be a hung parliment with the Tories as the largest party...hoho

Tom said...

It all depends on turnout doesn't it?

Brian said...

I'm dreaming of a Tory/UKIP coalition government with Nigel Farage as Europe Minister and Sir Ming Campbell as Speaker.

Unknown said...

Why should anyone believe the polls? I've been a YouGov member for a long time yet have only once been "selected" to do a political poll and not in the current election period. So consider the following facts:
1. Peter Kellner is a Labour supporter.
2. His wife, Baroness Ashton, was appointed by Brown to the post of Europe's High Minister (Foreign Secretary in English). A post for which she had no mobvious qualifications or relevant experience.
3. I am a registered Tory voter in my YouGov profile.
How many others are there like me?

Sean said...

If there were to be a hung parliament, I expect the most likely coalition would be Tory/LD. In 1997 the LD put their faith in Labour, only to be betrayed: no reform of the voting system followed. Furthermore it makes little sense to ally with a failing party whose support is crumbling day by day.

I'd rather the Conservatives won outright. I still believe that when it comes to the moment of putting a "X" on the ballot paper swing voters will go Tory as a lower risk choice.

However, a Tory/LD coalition would have a mandate for change and a healthy majority no matter which of the current polls you choose.

The Remittance Man said...

Sunday morning pedantry.

Even the polling companies admit their polls are accurate only to +/-3 percent. Quoting average percentages to one decimal place is a little superfluous - round it all up and make it look tidy please.

Calum said...

Iain,

Statistics hat on. I'm afraid you just can't take averages of polls: if you do, what you are doing is effectively creating a larger sample and asking what the cumulative result is. The problem is, mathematically, going from a poll of (say) 3000 people to a poll of 9000, or even 30000, does not make any real difference to the accuracy. In fact, because each of these polls is done using different methods, averaging them is probably comparing apples and oranges.