Media commentators seem to have caught up with the fact that UKIP are challenging for third place in the polls… except that it isn’t quite a fact. As ever, the actual picture is more complicated and whether UKIP are competing for third place, or have secured third place, or are trailing in a very distant fourth place depends on the polling company and their methodology.
There are currently nine polling companies conducting regular or semi-regular voting intention polls, four of them have shown UKIP ahead of the Lib Dems – in some cases like Survation almost always, in some cases like YouGov very rarely. The other five have never done so, and in the case of ICM or ComRes’s phone polls have never shown UKIP even vaguely close to third place.
The perception that they are consistently competing for third place comes from a couple of factors. One is publication bias – polling companies that show UKIP in third or challenging for third tend to include them in their topline figures. Polling companies that have them in a distant fourth don’t really bother. This is understandable enough, but means the good figures get far more prominence than the bad ones (it is further exacerbated by the reporting of polls – UKIP in third place is a news story worth reporting. UKIP back in the crowd with 3% or 4%, as ICM, MORI and ComRes have sometimes shown them in recent months isn’t). The other factor is that UKIP tend to do better in online polls, and online polls tend to be more frequent. Between the start of June and the 25th October when I collected up the data for this post there were 153 voting intention polls, 135 were online. Even putting aside YouGov, who conducted over two-thirds of all the polls in that period, there were 33 online polls compared to 18 phone polls. The polls that are best for UKIP are also the most frequent.
The graph below shows the average score for the Liberal Democrats and UKIP for each polling company between June and October, ordered from those showing the biggest Lib Dem leads over UKIP on the left, to those showing the biggest UKIP lead over the Lib Dems on the right.
At the left ICM and ComRes (phone) are showing the Lib Dem very solidly in third place, with UKIP not far above their 2010 level of support. Populus and MORI have a big gap between the two parties. TNS and YouGov both, on average, have the Lib Dems in third place, but have both on occasion shown UKIP in third place or equal with the Lib Dems. Only Opinion and Survation give UKIP a higher average level of support than the Lib Dems, and only Survation significantly so.
As usual there are different reasons for this pattern. For example, Survation give UKIP by far the highest level of support because they prompt from them in their main question (something I’ve written about at more length here). ComRes’s phone polls probably show the lowest level of UKIP support because they use a harsher likelihood to vote filter for minor parties than for the main three (their topline figures include those who say they are 5/10 or more likely to vote for Con, Lab and LD, but only those 10/10 certain to vote for minor parties).
The main cause of the difference however, as I’ve marked on the chart, seems to be between online and telephone polling. For whatever reason, online polls seem to show a higher level of support for UKIP than telephone polls do. This doesn’t seem to be a “minor party” factor, if you compare “others” in general online polls show more support for minor parties, but take out UKIP from the equation and it vanishes, as I’ve shown in the graph below. For some reason, the consistent difference only shows up with UKIP.
There are various possible explanations for this difference. One is interviewer effect, perhaps people are more willing to admit to an anonymous web-interface that they are going to vote for UKIP than they would be to a live human being on the end of a telephone. The other possibility is that there some sort of sampling difference in one mode or the other. Are the sort of people who would support UKIP systemically less likely to agree to take part in a cold-called telephone interview than other people with the same demographic profile? Or are the sort of people who would support UKIP more likely to join an internet panel than other people with the same demographic profile? We cannot tell.
What do we have to judge by? At the last election all the main pollsters got close to UKIP’s actual level of support, so that doesn’t give us any real clue. Since UKIP’s support started to really increase in late 2011 we have had very few big electoral tests to check by, and very few where both telephone and online pollsters have conducted comparable polling.
There were two polls for the Feltham & Heston by-election, one
online one from Survation and one commissioned by Lord Ashcroft and presumably carried out by Populus over the phone. Populus showed UKIP at 5%, Survation showed them on 7%. UKIP ended up getting 5.5%, so hardly a conclusive answer (UPDATE – plus, as I should have spotted, the Survation poll was telephone anyway, so the comparison couldn’t have told us anything about online vs phone polls anyway!)
In the London elections this year all pollsters overestimated support for UKIP slightly, but the only telephone poll (by Populus) showed UKIP candidate Lawrence Webb on 3% the same as online polls by ComRes and Opinium. Besides, by describing themselves as “Fresh Choice for London” UKIP did not put their party name on the ballot paper, which may have depressed their actual vote – we don’t know if pollsters slightly overestimated their support, or if UKIP underperformed by failing to mention UKIP in their party description on ballot papers.
Even the worse polling for UKIP shows them increasing their support from the last election, mostly quite significantly, so I think we can be confident they have gained support since 2010. Just how much we can’t really tell – the pollsters range from the relatively modest all the way up to trebling their support or more.