Lib Dem poll ratings – why I’m putting my trust in ICM

by Stephen Tall on March 3, 2015

Occasionally, when I see the latest poll showing the Lib Dems’ poll rating dribbling along at 6-8%, I wonder if I was stupidly over-optimistic when I nailed my colours to the mast with my prediction in January that the party would win 12% and 32 seats in May’s general election.

Quite possibly so. But my house of straw/brick was at least built on some solid foundations, primarily my trust in ICM’s polling credentials.

ICM is rightly, in my view, regarded as the ‘gold standard’ pollster. Its final polls proved to be most accurate in each of the 1997, 2005 and 2010 elections (in 2001, NOP took the prize but ICM were still very close).

Its methodology is controversial because it adjusts its reported figures by re-allocating a proportion of those voters who say they are undecided back to the party they say they voted for at the last election. (Lord Ashcroft and Survation also do this, but no other pollsters.) This means ICM polls are part-snapshot of how people say they will vote now; and part-forecast of how people will actually vote in May.

As Anthony Wells has noted‘When ICM first pioneered it in the 1990s it helped the Tories (and was known as the “shy Tory adjustment”), these days it helps the Lib Dems, and goes a long way to explain why ICM tend to show the highest level of support for the Lib Dems.’

My hypothesis is this: the Lib Dem vote is notably non-tribal (Lib Dem voters are much more likely than Labour/Tory voters to say they may yet change their mind). The closer we get to an election, the more the Lib Dem vote tends to firm up. This is often then reported as a “Lib Dem surge” and attributed to the burst of publicity the party tends to get during an election campaign when broadcasters have no choice but to acknowledge the party’s existence. What ICM in effect does is anticipate that eventuality. That’s why, by the time we get to eve-of-poll, the different polling companies’ final polls have converged — but why, at this point in time, they haven’t.

The difference the methodologies make to forecasts of how the Lib Dems will do are illustrated by the graph below, comparing ICM and YouGov’s poll ratings for the Lib Dems over the past year:

icm v yougov

As you can see, over the last year the companies have diverged a lot. ICM consistently reports the Lib Dems at c.10-12%; YouGov consistently at c.6-8%. This makes a huge difference to the likelihood of the Lib Dems surviving the May election with at least 30 MPs.

What we don’t know, of course, is whether ICM’s methodology will work in the changed circumstances of Coalition politics. My bet is that it will, and that’s what my predictions are based on.

However, it’s only fair to note that ICM is an outlier among the polling companies, which means that even the ‘polls of polls’, which aggregate data in order to try and smooth out the random noise of day-to-day polling fluctuations, tend to show the Lib Dems at or around the figures shown by YouGov.

I reckon that, by polling day, YouGov (and the other pollsters) will have moved towards ICM’s current figure rather than the other way around. In 64 days, we’ll know.


It seems to me that ICM has also best matched the local election results.

I’d also argue there may well be an issue with false recall. Even directly after the 2010 GE, about 30% of the population recalled voting LD (see here, for example, Now, they may have considered it at the height of Cleggmania, but they didn’t do it. Only about 24% did, with a significant swingback to Labour.

Now, usually the effects of false recall are small. But here they could be significant. These 5% are probably answering voting intention questions as saying they will vote Labour – in many ways, they were the softest supporters the Lib Dems had, switching to Labour before the 2010 GE. But, in claiming to pollsters that they voted LD in 2010, they are adding a significant number to the LD->Lab switchers. This could easily mean that the LD vote share given by many opinion polls is low by 2-3 percentage points. Not much for Lab or the Tories, but a big deal when you’re slugging it out and polling around 10%.

I really wonder how much of an effect the pollsters weightings for people mis-remembering how they voted are having.

by Liberal Will on March 3, 2015 at 3:35 pm. Reply #

I agree with you that in normal circumstances, we’d expect to see an uptick due to more publicity – but we’ve had plenty of publicity as a result of being in coalition, so I think that effect will be less this year.

by Zoe O'Connell on March 3, 2015 at 6:17 pm. Reply #

I'd agree with Zoe – expecting an uptick In the campaign seems over-optimistic to me. Didn't we actually drop in the polls during the European campaign?

by Nick Barlow on March 3, 2015 at 8:08 pm. Reply #

But, Nick, my point is I’m not expecting an uptick in the polls – I’m saying I trust ICM’s polling, which consistently shows the LDs on c.10-12%. (Maybe rightly, maybe wrongly: we’ll see.)

by Stephen Tall on March 3, 2015 at 10:41 pm. Reply #

But the reason you give for believing ICM’s polling is because they’re predicting what will happen in April.
“The closer we get to an election, the more the Lib Dem vote tends to firm up. This is often then reported as a “Lib Dem surge” and attributed to the burst of publicity the party tends to get during an election campaign when broadcasters have no choice but to acknowledge the party’s existence. What ICM in effect does is anticipate that eventuality. ” The question is not about their polling, but the assumptions behind the methodology they use to mare that prediction.

by Nick on March 3, 2015 at 10:55 pm. Reply #

[…] Lib Dem poll ratings: why I’m putting my trust in ICM by Stephen Tall on Stephen Tall. They are the gold standard, and more accurately reflect the Lib […]

by Top of the Blogs: The Lib Dem Golden Dozen #414 on March 8, 2015 at 7:00 pm. Reply #

Leave your comment


Required. Not published.

If you have one.