Archive for the 'Pollsters/polling' Category


Pro-Brexit anti-Lords poll splashed by the Mail comes under fire for “loaded” questions

Thursday, May 24th, 2018

Leo Barasi of the PB/Polling Matters podcast sets out his reasons

Leo Barasi, a regular with Keiran Pedley on the PB/Polling Matters Podcasts is no stranger to PBers and his is always worth listening to.

He’s posted on his “Noise of the Crowd” blog observations about the ComRes poll that featured in yesterday’s Daily Mail. He writes:

The fundamental problem is that the questions were nearly all one-sided agree/disagree questions, with each one loaded against the Lords and Remainers. A couple of examples:

It would be wrong for the House of Lords to try and thwart Brexit [“thwart”!]”

“It is wrong that the House of Lords has already voted against the government on Brexit 14 times”

“There are currently 780 members of the Lords compared to 650 MPs in the Commons. This is too many”

If you really want to measure public opinion you ask a question that presents both sides of an argument equally, then allow respondents to choose which they are closer to. Or if you really have to ask agree/disagree questions, the collection of the questions should be balanced so you’re not pushing a particular argument and you can compare the skewed questions against each other.

A good guide of a fair poll is that you shouldn’t be able to guess the view of the organisation commissioning the poll from the questions. This clearly fails that.

The poll was done for a new pro-Brexit campaign called “We, the People”.

I have invited ComRes to respond.

Mike Smithson


The London election polling test finds that LAB was overstated by 4 points in the final polls

Friday, May 18th, 2018

What does this say about current national polling?

It is not often we get a real election against which we can compare final polls and this month’s London elections provided one such opportunity.

The LAB/CON/LD shares in the final polls from YouGov and Survation are shown in the chart and compared with the overall result.

As can be seen both pollsters had LAB at 51% which compared the 47% that actually happened. Survation got the Tories spot on while YouGov understated the party by 2 points.

Both pollsters understated the LDs – YouGov by 2 points and Survation by a point.

London Borough councillors are elected by block voting. This means that voters could cast as many votes as there were seats in their wards. In most cases this meant three. The vote totals in the chart totals reflect the total number of votes cast for that party in London on May 3rd.

It will be recalled that Survation did best at GE2017 last June.

Mike Smithson


Be wary of YouGov’s finding that Britain’s voting intentions is classless

Saturday, May 12th, 2018

It isn’t new, and it isn’t really backed up by other pollsters or elections

Class is supposed to define British politics. Perhaps that explains the flurry of surprise yesterday when YouGov’s latest poll gave the Conservatives a 3% lead among the C2DE group (43-40-7 among the main parties), and showed Labour doing worse with them than with the ABC1s (43-37-11).

We shouldn’t be overly surprised. For one thing, the sub-groups aren’t statistically balanced and the figures are close enough that it’s quite possible that Labour is actually ahead with the C2DEs; for another, YouGov consistently produce results more favourable to the Tories than, say Survation – which reported a 1-point overall Con lead yesterday, against YouGov’s 5-point margin; but also, and most importantly, that finding is nothing particularly new.

Before going there, a big word of caution: not all pollsters are finding the same thing. ICM’s last poll gave Labour a lead of about 5% with the C2DE group, despite an overall 3-point Con lead. Survation, though they don’t use the standard social group definitions, found a more traditional – and marked – split among income groups, and a very even set of results when voters are grouped by educational level. Mori, which does use the usual groups, reported even more distinct findings: a 6% Con lead among the middle-class and no less than a 17% lead among C2DEs (though the weighting changed a 4-point Labour lead into a 1% Con one overall, so we should perhaps apply a blue shift to those subsamples too). The same is true of ComRes. YouGov seems to be very much the exception.

All the same, while the company might be the exception, its poll isn’t. Go back before the last general election, to when the Tories were racking up big double-digit leads and again, they were clearly ahead with both middle and working-class voters. To take one example, the first poll YouGov conducted in 2017, which produced overall figures of 39-26 to Con (plus UKIP 14, LD 10), had an ABC1 split of 41-26 and a C2DE one of 37-27. Those findings are not unusual for the time.

This is where the assertion, made in the light of YouGov’s poll yesterday (and without reference to other pollsters), that “Corbyn is losing the working class” is a bit of a fallacy. He has already, within three years, lost it, regained it, and lost part of it again. Even if we accept YouGov’s subsamples, the 37% share they give Labour among the working class is still well above what it was in the first four months of last year.

In fact, whether or not the gap between how the classes vote has closed (or even reversed), what is clear is that it has been reducing for a long time. This tweet helpfully summarises Labour’s lead among the working class over the last 40 years. What’s notable is how that lead has shrunk when measured against the overall figures.

In 1992, Labour did 18% better with the working class than they did overall. Blair’s landslide changed nothing in that respect: it was still 18% in 1997. Come 2005 and it had dropped to 12%, in 2010 it was just 8%, in 2015 it was back up to 13% but in 2017 it fell right down to 4%.

We can explain the spike in 2015 as a consequence of working-class Tory voters defecting to UKIP; a phenomenon which lasted just one election. So while Labour certainly has a problem in retaining its traditional vote under Corbyn, that problem started well before him. (It is of course true that the reverse also holds: Labour might be losing working-class votes but it’s picking up middle-class ones to compensate).

In theory, this should mean that safe seats on both sides should be becoming less so, and indeed there is evidence of that, both from the general election and this recent round of local elections. We should, however, beware of being blinded by the exceptions. For all that Kensington or Canterbury were spectacular in the 2017 general election, or that London and the South East saw a relatively poor Tory performance while the Midlands and North generally saw a strong one for a party eight years in office, the great majority of seats did not change hands.

My gut feeling is that YouGov are out on this and the other pollsters are closer to being right (which also has to raise a question about YouGov’s top-line figures). However, while the trend it’s picking up on might be exaggerated, it is still there. That raises very interesting questions about all the parties’ strategies for the next election. It also implies that with both middle- and working-class voters leaving their traditional political homes, firewall seats won’t be anything like as safe as they once were – on all sides.

David Herdson


Polling analysis: Corbyn is a liability to Labour while TMay has returned to being an asset to the Tories

Wednesday, April 25th, 2018

The YouGov favourability trackers are just about the only polling where we can compare leaders with their parties on the same basis. The same question is asked in exactly the same form to the same sample whether people have a favourable or unfavourable opinion of CON/LAB/TMay/Corbyn.

It is also a tracker which is asked in the same form at regular intervals which means there are enough data points to examine trends.

The movement in the leader and party ratings since the general election was called just over a year ago is shown in the chart above. As can be seen Theresa May was doing better than her party but went behind after the election. Only in the past few weeks is she doing better again

    The Corbyn and Labour party figures have been consistent over the time period. Labour Party has always been viewed more favourably than its leader a situation that remains.

In the chart the unfavourable figure is subtracted from the favourable figure to give a net number in each case for each data point.

All this rather undermines the notion that is often heard by his supporters that it was Mr Corbyn rather than the party that gave the red team a better than expected result last June.

Mike Smithson


Exactly a year ago this weekend ComRes had TMay’s Tories 25% ahead

Sunday, April 22nd, 2018

How things have changed since

It is just a year since Theresa May made her fateful and what will be her career defining announcement about calling a general election to secure a bigger majority.

On the weekend after the news we had the initial round of voting intention polls of the campaign and those are shown in the chart above.

As can be seen the one that stands out is ComRes, which had been the most accurate pollster two years earlier at GE2015. This had the biggest Conservative lead – a whopping margin of 25 points over LAB.

    Although the final lead on election day was just 2.5% it is too easy to conclude that those late April polls were wrong.

Only a couple of weeks after the general election announcement there were the local elections where the Tories made big gains doing substantially better than had been predicted.

It was those real elections that seemed to validate the polling and reinforce the view that Mrs May’s gamble was going to pay off. The big question was not whether there would be a Tory overall majority but would it be a landslide.

My guess is that it might well have done so but for the length of the general election campaign and for the over-confidence it engendered in the Tory camp that led to the manifesto debacle and Mrs May believing that she didn’t have to face Corbyn in a leaders’ TV debate.

In total there were seven weeks between the initial call and parliamentary vote to authorise it and the June 8 election.

So we cannot conclude that the polls weren’t wrong in late April last year. What they do show is that there was a dramatic change in views of the incumbent Conservative government and particularly the Prime Minister as a result of the campaign itself.

It is very hard to envisage the circumstances in which there will be the next Conservative 25% lead.

Mike Smithson


Latest PB/Polling Matters podcast: Are you racist? Syrian airstrikes & the Lords report on polling

Wednesday, April 18th, 2018

On this week’s PB/Polling Matters podcast Keiran Pedley is joined by Matt Singh (Number Cruncher) and Adam Drummond (Opinium) to discuss:

1) Why voting intention polls and perceptions of party leaders seem to be moving in different directions

2) Reactions to the Windrush scandal and how pollsters deal with sensitive questions around immigration

3) An exclusive survey from Opinium for PB that shows 1 in 10 Brits believe the Russian military accusation that Britain staged the Douma chemical attack in Syria

4) What the Lords has to say about the future of polling and how pollsters should react

You can find data tables for the Opinium poll on Syria here

Follow this week’s guests:





There’s the potential for Labour to get a long term polling boost because of their anti-semitism issues

Sunday, April 15th, 2018

Chart from polling conducted by YouGov for the Campaign against Antisemitism

Older voters are more likely to endorse at least one anti-semitic statement, and remember older voters are more likely turn out to vote.

Conventional wisdom suggests that Labour will take a long term hit in the polls because of the recent coverage of their anti-semitism issues but recently we’ve seen conventional wisdom proven to be very wrong, this might be another example.

Before anyone accuses me saying older voters, Tories, and Leavers are all anti-semites it should be noted that a majority of those demographics didn’t agree with a single anti-semitic statement but a significant minority did.

That significant minority could determine who wins the next general election, if a couple of thousand voters had voted differently last June Jeremy Corbyn would have become Prime Minister.

Coupled with his other popular policies such as inter alia renationalising the railways, ending austerity, not loading students up with debt, and opposition to bombing Syria, this might might Corbyn even more appealing to some voters.

I don’t think it is any accident that Nick Griffin, a man who in the past has said ‘that there was a systematic and deliberate policy whereby six million Jews were gassed to death is for a variety of forensic and common sense reasons, [is] utter nonsense’  declared his support from Jeremy Corbyn this week on not bombing Syria and talked about voting Labour for the first time in his life.

The political kaleidoscope in Britain has well and truly been shaken, the old assumptions don’t always hold true. Labour have a proud tradition of opposing racism and bigotry, this week they have won the support of someone who in 1997 published booklet entitled Who are the Mind Benders? which outlined a Jewish conspiracy to brainwash the British people in their own homeland.



Theresa May now level-pegging with “Don’t know” as to who would make the best PM

Saturday, April 14th, 2018

Corbyn, as almost ever, trails behind

This polling, from the latest YouGov, rather sums up British politics at the moment. When respondents were asked who would make the best prime minister 37% said Mrs May which is exactly the same number who said they didn’t know

Neither of the main party leaders is able to show widespread support. On the party splits just 70% of current Labour voters opted for Mr Corbyn. Mrs May gets 90% of Tory ones.

Of course it is likely that at least one of these leaders may not be the party’s flag bearer at the next general election which, if it follows the time table in the Fixed Term Parliament Act, will not happen for at least four years.

Corbyn, as has been observed, is struggling in the standard leader ratings when voters are asked to rate who they find the most favourable or are satisfied or give their approval to. But, as has been observed many times because Corbyn is so overwhelmingly popular amongst LAB’s membership his position within the party is totally secure.

The same cannot be said for Mrs May who has defied gravity, apparently, and is still there after her general election loss of the Conservative majority 10 months ago. Her great strength is that not one of the alternatives commands anything like the support that would be necessary.

The voting intention numbers in this latest poll see both the Tories and LAB on 40% with the LDs up 2 to 9%.

Mike Smithson