In my previous blog post, I looked at the four constituency polls conducted in Wales by Lord Ashcroft during 2014. One slightly puzzling feature of them that I pointed to was the relatively strong performance of Plaid Cymru in all four seats. Given that none of the four constituencies polled is conceivably a Plaid target seat for 2015, and given also Plaid’s rather limited progress in the national polls, it was rather surprising to observe Plaid’s vote share in all four constituency polls up by several points on Plaid’s actual 2010 election performance in those places.
To what might we attribute this slightly puzzling result? It seems very unlikely, as I said, that Plaid can have been targeting substantial resources at any of these seats. It is possible, perhaps even likely, that Plaid’s vote in these seats will get squeezed as election time approaches, and voters in those respective constituencies face up to the question of who can conceivably win in their locality. (Although, to be fair, Lord Ashcroft has sought to address this issue by asking people first a ‘standard’ voting intention question and then one related specifically to the particular constituency they are in). We don’t have equivalent constituency polls from this point in the electoral cycle prior to the May 2010 election: perhaps Plaid at that point would have been in a promising position in these seats, only to see votes drain away later.
(Of course, there are plenty of keyboard warriors/fantasists, lurking on Twitter and in the below-the-line comments sections of many publications, who will insist that most poll results are just arranged to suit the interests of those paying for them. To which all I can say is that, if Lord Ashcroft is a closet Welsh nationalist, he has been keeping it very well hidden…).
More seriously, there is one other factor that may at least partially explain why Plaid are showing up better in Lord Ashcroft’s polls than one would expect from the overall picture of the national polls. Below is a table based on dividing the ten publicly-reported national polls on general election voting intention in Wales in 2014 into two groups: those seven conducted on-line (all by YouGov) and the three conducted by telephone. The table then lists the average general election vote intention figure for each party in the two types of poll:
As we can see, for most parties here the differences between the two types of poll are small. They are also quite explicable. Labour does a bit better in the phone polls than the online ones because Labour’s support ebbed during 2014 and two of the three phone polls were conducted in the first five months of the year, when Labour was doing a bit better. Very similar factors can account for the small differences between the two types of poll for Lib-Dem voting intentions. Meanwhile, the Tories – whose poll rating was pretty stable throughout 2014 – see little difference between the two types of poll. UKIP were doing better by the end of the year than at the beginning, and thus score better in the online poll average.
The puzzle remains Plaid Cymru, for whom there is the largest gap between phone and online. There was little or no obvious trend in their general election voting intention throughout 2014 within YouGov’s online polls. Moreover, all three of the telephone polls put Plaid at a higher level of support than did any of the online ones. In short, there seems to be a genuine difference between the two methods in the support levels they are showing for Plaid Cymru. (Albeit this is a difference of relatively modest size: it’s not like we are talking about phone polls showing Plaid’s support levels double those of the online polls).
These differences may well, in turn, go at least some way to explaining why Plaid have been performing quite well in Lord Ashcroft’s constituency polls. These polls are all conducted by phone – they have to be, as no online company (not even one as well-established as YouGov) have sufficient respondents in their panel to produce 1000-sample polls in individual constituencies. If phone polls are, in general, producing systematically better figures for Plaid than on-line ones, then we shouldn’t perhaps be quite so surprised at Lord Ashcroft’s constituency results for the party.
At least two further questions remain, though I’m not sure I have adequate answers to either of them. First, why do we see these differences in the two types of poll? There may be some social factors behind it: one colleague with whom I raised this issue pointed to the often poor broadband connections in much of rural west Wales as being potentially a factor leading online polls to potentially under-state Plaid support slightly. Or there may be some more directly political factors at play (though I’m not quite sure what there would be); in this regard, it is perhaps worthy of note that the best poll result for the SNP since the September independence referendum (that by Ipsos MORI in October) was also a phone poll.
Second, the other obvious question is which one is right?! It is very difficult to say. ICM and YouGov are both highly (and quite deservedly) respected companies internationally with an impressive record for measuring voting preferences accurately in elections and other events. Perhaps we will see these slight differences erode as we get nearer the election – just as the pollsters tended to converge as the Scottish independence referendum approached. But the possibility that one polling method is tending to systematically under- or over-state the support of one significant party in Wales is something that is at least worth keeping an eye on, I think.