What is clear from the Survation poll of the Eastleigh constituency, commissioned by UKIP’s Alan Bown, is that local factors are at play in Eastleigh that are seeing UKIP maintain strong support compared to nationwide polling.
Several outlets report on the poll, including the UKIP website, declaring that UKIP is on course to win the seat at the General Election because Survation has them attracting an incredible 32% support.
However, here we are seeing the same problem with the figures that are being reported as we saw in Survation polls for UKIP at the end of last year in Great Grimsby and Thanet South, where selective use of the polling data is giving a distorted view of the electoral calculus in the constituency.
What is being reported on and played up by the media is Table 5 in the Survation poll.
The data presented concerns this:
Q3 Voting Intention Tables – Normal Weighted Table and Likelihood Weighting
Q3. If that general election were to be held tomorrow, which party do you think you would be most likely to vote for in your Eastleigh constituency?
The results given, and being used as a sensational story, are:
UKIP – 31.7%
Conservative – 28.8%
Lib Dem – 25.7%
Labour – 12.0%
Green – 1.1%
Party not listed – 0.7%
But the figures do not give the true picture because they only show the preference of 316 people and crucially ignore the responses of a significant proportion of those polled:
Base : Respondents likely to vote in May 2015 election and Excluding Undecided/Refused
It is the exclusion of undecided voters and those who refused to say which party they would support which makes these headline figures basically worthless.
For the meaningful figures one must look at Table 4, for that is where we find the polling data for respondents who are likely to vote in the May 2015 election, including those who are yet to decide who to vote for, or who refused to tell the pollsters who they will support. The numbers based on a sample of 472 people in the constituency are:
Undecided – 25.7%
UKIP – 21.2%
Conservative – 19.2%
Lib Dem – 17.1%
Labour – 8%
Refused – 7.6%
Green – 0.7%
Party not listed – 0.5%
No less than 33.3% – or exactly one in three people polled – are yet to make up their mind how to vote or have refused to say how they will vote. If that is representative of the electorate across Eastleigh then of the possible 53,760 votes that might be polled (the 69% turnout in 2010) up to 17,900 votes are still up for grabs and are currently going to an unknown destination. That is a huge number – larger than the winner of the 2013 by-election received – and ultimately it is those votes that will decide which party’s candidate wins the election.
This is not to do UKIP down. Of course, it could be good news for UKIP if most of those people choose to vote for the party. They would win a Parliamentary election for the first time. But it could still be bad news for UKIP because those undecideds could break for the Conservatives or Lib Dems, putting either of those candidates into Westminster. There is an ’embarrassment factor’ in polling that first emerged clearly in 1992, which particularly affects the Conservatives, where voters who are leaning towards voting Tory decline to say so because of the party’s unpopularity in the media. That is one of many factors which underlines why there is a need for realism and perspective when looking at these polling figures.
In Eastleigh there is still everything to play for. It remains a three-way marginal and UKIP is polling over 21% – which is way above their nationwide average. But things are not necessarily as rosy for UKIP as the sensationalist headline suggests and events between now and May could change things in Eastleigh dramatically for or against UKIP.
Autonomous Mind 
AM, I hope you’re as critical when poor poll results are announced for UKIP.
Absolutely, but the issue here is the cherrypicking of information to give the impression that support is greater than it actually is. One cannot pretend that the undecided voters do not exist.
We do not “pretend that the undecided voters do not exist”. Do look at the data tables, undecided and refused voters are not discarded there is an adjustment for them based on past vote and they are included in the headline figures. This whole article is then wholly inaccurate.
A fascinating Twitter exchange with you Damian. I’ve offered you a £100 bet if you stand by your figures and believe UKIP will get 32% in Eastleigh.
But most importantly I’ve asked you for your original figures and +/- error margin, but as yet you have ignored my requests.
You can’t escape from the fact your are applying forumlae for undecideds rather than accepting their vote could go anywhere. You say it is accepted practice, but it does not mean your filtered result mirrors reality.
Everything in your report has been put through a voter intention filter that applies an immediate bias in favour of UKIP. This is why so many people look at polls and say ‘nonsense’. I didn’t criticise Survation or the methodology, rather the selective use of figures. But the more you have explained your actions on Twitter, the more there is to criticise.
** You say that in Table 6 the undecided voters have not been ignored. You’re right, your methodology replaces them…
What is missing, though, is the raw data on voting intentions, so that we can’t see the base figures without the filters. Then, each time a filter is applied, there is an associated error rate/uncertainty, which probably means that the cumulative errors exceed by a significant amount, the margins between the parties.
This is, of course, why these heavily filtered results are of very dubious value, especially in the context of the inherent dishonest in not prominently stating the margins of error.
1) You don’t need Survation to provide you with a margin of error or wait. It’s a simple calculation – here’s a calculator you can use for any poll: http://www.raosoft.com/samplesize.html – you will discover that the MOE is wider but not much wider than if we had used a larger sample.
2) The unweighted data is in the first row of table 1 – it’s not hidden – and by the way, not all polling firms show the level of detail we do.
3) You are correct – we do not accept that people’s vote could “go anywhere” – we make a judgement as other polling companies do. This judgement is indicated in the tables – it’s not a secret or a departure from commonly accepted practice.
4) The methodology for undecided voters does not “replace” their views. In this case it says that half of the undecided voters will vote for the party they voted for just over a year ago in the Eastleigh Byelection.
It takes their stated previous choice and then discounts that 50%. That’s not a “bias” anymore than ICM/Guardian’s monthly allocation of DK 2010 Lib Dem voters to the LDs. We do not accept as you say that “their vote could go anywhere” – we say that voters are more likely to vote for a party they recently voted for than a party that they have not recently voted for -do you disagree? This adjustment makes the effective sample size larger and the margin of error lower – something you say you are concerned about.
Please do not say any more that we are not giving you information “we have ignored your requests”. I have given you my personal email and offered to ask any question you have.
I have no interest in betting money with you. If you are interested in whether our polling is accurate you only need to wait a few weeks as we have also provided an Eastleigh Borough Council voting intention (a snapshot, not a prediction) that you can compare to the actual figures on May 22nd.