Archive for the 'Pollsters/polling' Category


As well as a unique hair-style Trump has another thing in common with Boris – being overstated in the polls

Tuesday, February 2nd, 2016

Iowa polls
Real Clear Politics

All but one of the Iowa pollsters had inflated Trump numbers: All had deflated ones for Cruz and Rubio

One of the things that I’ve written about before here is that Boris gets over-stated by the polls when tested against real election results. It happened to quite a degree at the last London mayoral election and we saw a similar pattern overnight in the results from Iowa.

The table above from Real Clear Politics sets the data out nicely and this is a key thing to take into account in the coming contests over the next five months.

One of the great elements about an election is that you have real data against which you can compare what the pollsters said was going to happen.

I know that the caucuses are not like normal elections such as the one we’ll see in New Hampshire a week tonight. Working out who is actually going to go to a caucus meeting is a massive challenge. We should expect the proper primaries to be better. In previous White House races most pollsters have kept away from the caucus states because of the complications.

Iowa only gets the intention attention that it does because it is first.

Mike Smithson


Corbyn drops to his lowest level yet amongst those who voted LAB last May

Friday, January 29th, 2016

More numbers, no doubt, for Corbynistas to remain in denial about

Every month for more than 40 years the pollster Ipsos MORI has carried out leader ratings. The question format has been the same simply asking people whether they are satisfied or dissatisfied with the performance of the leaders.

As I’ve argued and shown these ratings have proved to be a better indicator of General Election outcomes than standard voting intentions. If we’d have followed these at GE1992 and GE2015 rather than the party vote shares we would have got both elections right.

So to me the most significant feature of the latest Ipsos-MORI phone poll was not the Tories extending their lead but how Corbyn and Cameron were being rated. The particular numbers that I highlight are how those who voted for the main parties at the last general election now rate the leader. My focus is on those proportion who give a positive rating.

The trend is in the chart. Now fewer that half of those who voted for Miliband’s LAB last May are ready to give a positive rating to Corbyn and the trend is moving away from him.

Clearly we are some way away from the general election and we don’t know yet who the Tory leader will be. But even so the reluctance of 2015 LAB voters to be positive about their man doesn’t bode well for the party.

It is not party members who matter, as the Corbynistas assert, but past Labour voters.

Mike Smithson


If all goes to Dave’s plan British voters will be filling in one of these in just 144 days

Wednesday, January 27th, 2016

The online-phone EURef poll gap remains

A picture of the actual referendum ballot paper was featured on the BBC Daily Politics this morning. The pic above is a screen shot.

There’s something about seeing a ballot paper that makes an election very real and all the signs are that Cameron & Co are using their best endeavours to get it out of the way as soon as possible.

The longer the wait the more that Harold Macmillan’s famous “events” can intervene and the government’s strategy seems sensible.

Cameron only agreed to the referendum in the first place as a means of impeding the UKIP surge during the last parliament and it was only because he won a majority that this is now happening.

It is not something I’ve yet bet upon. The divide between the online polls and the phone ones is enormous and cannot easily be explained away. Last month ComRes commissioned two referendum polls at about the same time and as can be seen from the table got two very different outcomes. Online the split was 41% to 41%. The phone poll had it at 56% remain to 35% leave.

My view remains. Whatever Dave recommends will win as has been seen in poll after poll.

Mike Smithson


The turnout problem facing Bernie Sanders as seen by the NYT’s Nate Cohn

Tuesday, January 26th, 2016

The full article is here


A week to go until Iowa: the Great American Gamble – to Trump or No Trump

Monday, January 25th, 2016


Leader ratings side by side: How JC’s doing against DC generally & with party supporters

Monday, January 25th, 2016

The next general election, of course is unlikely to be between Corbyn’s LAB and Cameron’s CON. The latter has made his exit intentions partially clear though we don’t know whether it’ll be before the election or afterwards.

There’s doubt on the Labour side as well. Interestingly in recent days PB’s two LAB post writers, Henry G and Donald Brind, have both suggested that they don’t thing Corbyn will be there at the election.

Whatever the chart above can only be interpreted as bad news for Corbyn. He’s doing significantly worse than Miliband and we all know what happened there.

Of real concern should be the ratings from those who voted for Miliband’s LAB last May.

Mike Smithson


The last time Hillary Clinton fought New Hampshire the pollsters did worse than the British ones at GE2015

Monday, January 25th, 2016

NH Polls 2008

Many punters lost a lot of money that night

One of the great nights on PB was the New Hampshire primary in January 2008. It came a week after the Iowa caucuses where Obama pulled off a significant victory.

Interest was intense as one of the two main parties look set to be the first in US history not to select a white man as nominee.

Following Iowa the media narrative was all on Obama and the final polls for the New Hampshire primary are shown in the chart above. The Real Clear Politics average close with Obama on 38.% and Clinton on just 30%. The Clinton campaign looked doomed and all the money piled against her.

Those who backed Obama at very tight odds on prices lost a lot if money. One of the site’s biggest gamblers told me it was his worst political betting ever.

    For as it turned out Hillary got 39% to Obama’s 36.4% which was a bigger miss than what we saw in the UK on May 7th last year.

Close analysis of what happened found that women were significantly more likely to have chosen Clinton over Obama and, more importantly, were more likely to vote.

What this says about the coming contest against Bernie Sanders I don’t know. He has the benefit of being from the neighbouring state and there’s a long history of New Hampshire favouring such contenders.

Mike Smithson


The pollsters got the big picture at GE2015 absolutely right provided you ignore the rubbish voting intention numbers

Saturday, January 23rd, 2016

The polling inquiry should have considered alternatives to standard voting intentions

While everybody else has been obsessing about voting intention numbers I’ve been looking at how the pollsters did with their lead rating last May and the big picture is in the chart above. This shows the percentage in each of the samples that gave positive ratings to Mr. Cameron and Mr. Miliband.

In the final few days five pollsters asked leader rating questions and the results are featured above. Although their question formats were different the overall trend from each was very similar. Amongst all sampled Cameron had leads of 7% to 20%.

The final two column clusters look at the proportion of declared LAB and CON voters in the specific polls were prepared to give their man positive ratings. Here the LAB voter responses to Miliband are in a fairly narrow range from 68% to 76%. Cameron’s, meanwhile ranged from 81% to 96%.

    That Miliband was trailing across the board amongst those saying they would vote for him is very telling. Clearly the Labour vote was less committed to their leader and, I’d argue, less likely to turn out and vote.

This mismatch between the voting numbers and leader ratings has happened before and when it has the leader ratings have proved the best prediction. In a third of the six general elections in the past 24 years the voting numbers have been wrong. In all elections since 1979 the lead numbers have got it right.

My mistake was not to have produced this chart last May. I’ve learned.

Mike Smithson