A short one this.
There have been three sets of constituency polls lately which I believe, taken as a whole, look bad for UKIP. The first is the ComRes polling across UKIP target seats, including both Thurrock and South Thanet. It showed the party third overall in those seats, second to Labour, on 21% to the Tories’ 39%. UKIP complain that some seats polled are not their targets, but it also included Grimsby, where Labour have a significant lead. The overall numbers for the Conservatives therefore, are going to be a lot better.
This matched up with the ‘buried’ Thanet poll showing Farage losing to Craig MacKinlay, the Conservative candidate. (No poll has ever shown Labour leading in Thanet).
UKIP’s bacon seemed to be saved in Thanet when Survation, the pollster that previously gave them an outlier lead there, did so again with a poll giving them 39%. Its raw numbers had them on 35% before weighting.
Lord Ashcroft came in with a third poll that included two UKIP marginals. It showed Mark Reckless would lose Rochester and Strood but gave a narrow lead to UKIP in Thurrock.
How can we explain such wild discrepancies?
I asked his Lordship a question, having looked at the data-set, and he said he would dig into it, so I’m hopeful of maybe getting a response. It seems to me that the difference in the numbers comes down to how you count people who have not voted in the past election.
It is axiomatic in polling that those who have not voted before do not vote in the future – or are far less likely to do so.
Survation had come out with another outlier poll that showed a UKIP lead in Thanet. It would appear that the donor commissioned ComRes to try to confirm this unlikely number. ComRes put the Tories in the lead instead. Survation then got the same result next time they polled (from another donor).
Was this a big change due to dislike of Farage’s lack of work in the seat – for example, he missed every single meeting and rally held by the local cross-party Save Manston campaign. Or was it merely the way the pollsters count?
I have heard that Survation do not penalize on the basis of non-voting. This seems to me vitally important if you want to get an accurate number. Survation have a bad record overstating UKIP support, probably because of this. In the Rochester by-election, as one tweeter pointed out to me:
@LouiseMensch Rochester Survation polls: Oct 9% Ukip lead. Nov 12% Ukip lead. By-election result 7% winning margin.
I looked up other polls and results. Survation also over-stated UKIP and under-stated the Tories in Clacton (8% differential to result in UKIP’s favour) and in Newark. In one case in October they found a not-very believable 25% for UKIP nationwide leading to predictions of, er, 128 UKIP MPs. Clearly this was always utter rubbish.
What worries me is that Survation had a long time after Newark to get their polling right; it was UKIP-favouring in Newark, in Clacton, and again in Rochester. Did not the consistency of the pro-UKIP over-estimation bother the firm?
ComRes, under attack by Kippers for showing the Conservatives leading, published their tables. As Mike Smithson of Political Betting said:
Big worry for UKIP in the ComRes South Thanet poll. 31% of its support coming from GE10 non-voters.
This idea, that #ukip are heavily dependent on people who have not voted before, was confirmed to the Spectator by no less a person than Nigel Farage:
Other UKIP supporters on Twitter freely admit that their support is chock-full of people who have never voted before (not just not in 2010)
If we take a look at the last-but-one set of Ashcroft polling (his polls out today show a fall in UKIP support to 11%, -2 and down below even ICM), Lord Ashcroft’s raw numbers show the Conservative ex-MP Jackie Doyle-Price ahead in the seat of Thurrock on every table. It’s only when a certainty to vote weight is applied that UKIP move from below the Tories. However, although Lord Ashcroft’s poll does say how many voters are non-voters from 2010, it doesn’t appear to correlate them to present party support as ComRes does.
This is ComRes non-voter data from Thanet South
As we can see, 13% of Tory support comes from non-voters (the lowest) and 31% of UKIP support from non-voters (the highest).
Therefore, I think since UKIP themselves report the pledged support of large numbers of former non-voters, it would be good to see if Lord Ashcroft can dig out those figures, which might change the weighting in seats like Thurrock. Certainly in a tight race, if Jackie-Doyle price is ahead on raw numbers, a ‘certainty to vote’ would have to be weighed against ‘never actually voted before’.
This is not to say that non-voters will not vote – they may break the habit. But Survation got all the by-elections wrong in UKIP’s favor – every time. As Lord Ashcroft has noted, polls influence results. Hopefully Thanet won’t be fooled by what I think are very wrong numbers from Survation, which will of course, encourage ukip and depress Tories and Labour (even though I allege no political bias there will be a political effect).
ComRes which polled Thurrock has overall Con-UKIP numbers in Con favour, and they are counting non-voters. With Nigel Farage himself calling – a bit late in the day – for people to vote Conservative in Con-Labour marginals like Cannock Chase, being accurate on those margins is important. I hope that if his Lordship can correlate non-voters to parties supported he might consider releasing that data in seats where the polling is tight.
My prediction is that Con will hold Thurrock and South Thanet, UKIP will hold Clacton, and UKIP will lose Rochester. Perhaps UKIP will gain Heywood and Middleton – I say this because it was far tighter for them at the by-election than any pollster predicted, so Heywood may be a special case.