A Bonus YouGov Poll!

Well, there went my quiet Saturday… After posting this morning my reflections on last night’s BBC Leaders’ Debate, I thought I could have a few hours off and enjoy the Newmarket Guineas meeting on TV. (That’s horse racing, in case you were wondering).

However, Plaid Cymru have decided to interrupt my few, well-earned hours leisure by releasing a new Welsh poll. Like our Welsh Political Barometer polls it was conducted by YouGov, with the fieldwork done on Tuesday to Thursday of this week. The detailed tables are here.

First things first – can we trust this poll? After all, it was commissioned by a political party competing in this election. Well, I’ve looked through the details of the poll, and it seems quite ‘legit’ to me. It has been done to YouGov’s normal professional standards, with a similar sample size to the Barometer polls; the questions seem to me both sensibly ordered and reasonably worded. The data has also been weighted and reported in YouGov’s normal manner. It might well, be, of course, that had the results been unfavourable to Plaid’s narrative then the poll findings would not have been published. But while I would dispute some aspects of how Plaid seem to be spinning the results (see below), the results themselves seem as reliable as any poll to be conducted by YouGov.

So what do the results show? Well, here are the figures on the standard general election voting intention question (weighted, as YouGov have been doing with all their polls in the final few weeks prior to polling day) by likelihood to vote. (You can find the results without likelihood to vote weighting in the detailed tables):

Labour: 39% (-1 on the most recent Barometer poll)

Conservative: 26% (unchanged)

Plaid Cymru: 13% (+1)

UKIP: 12% (-1)

Liberal Democrats: 6% (unchanged)

Greens: 3% (-1)

Others: 2% (+2)


So these figures show only very small changes on the last Barometer poll, all well within the ‘margin of error’. Still, you’d always rather be going up rather than down, even within the margin of error, and Plaid will be pleased to have edged up a bit further – thereby relegating UKIP narrowly into fourth place.

An interesting feature of this poll, however is that it followed up the standard ‘generic’ voting intention question with one that asked respondents to focus on the specific dynamics of their constituency. (See the tables for the exact wording on this). This is very similar to the approach that Lord Ashcroft has followed in his constituency polls. It is also something that Survation have been doing in some of their recent GB-wide polling. What difference does this make? Well, here are the figures for voting intention once respondents have been primed to focus on their constituency:

Labour: 37%

Conservative: 25%

Plaid Cymru: 15%

UKIP: 12%

Liberal Democrats: 7%

Greens: 2%

Others: 2%


So overall this constituency-specific question doesn’t make a great deal of difference. But it does seem to help Plaid a little bit (and also the Lib-Dems to an ever smaller extent) – it pushes them up to 15%, which would be, if they achieved it, Plaid’s best-ever general election vote share in Wales.

On Plaid’s website, my old friend and former colleague Dr Dafydd Trystan (Plaid’s National Chair) has compared the results of this second question with those from the generic voting intention question in the first YouGov poll of the campaign, that for The Sun newspaper at the very end of March. I think this is a tad naughty – we are comparing the results of two different questions here. There is also a more general question regarding the use of these questions that ask people to think about their specific constituency: once the election is over we may be able to tell whether they, or the traditional generic questions, have proven better at ‘predicting’ voter behaviour.

I’ll up-date this post later today with UNS and Ratio Swing projections of these findings. I’ll also post some time (tomorrow, I expect) about the other findings of the poll.

Postscript: OK, so I’ve worked through the seat projection numbers for both the generic and the constituency-specific voting intention questions from this poll. And I’ve done them for both Uniform National Swing and Ratio Swing (boy do I know how to have a fun time on a Saturday afternoon…).

So, first, the standard voting intention question. Here are the UNS seat projection numbers:

Labour: 28 seats (holding all 26 from 2010, and gaining Cardiff Central and Cardiff North)

Conservatives: 8 seats (losing Cardiff North to Labour, but gaining Brecon & Radnor from the Liberal Democrats)

Plaid Cymru: 3 seats (holding the 3 seats won in 2010)

Liberal Democrats: 1 seat (holding Ceredigion, but losing both Cardiff Central and Brecon & Radnor)

And here are the Ratio Swing projections:

Labour: 28 seats (holding all 26 from 2010, and gaining Cardiff Central and Cardiff North)

Conservatives: 8 seats (losing Cardiff North to Labour, but gaining Brecon & Radnor from the Liberal Democrats)

Plaid Cymru: 4 seats (holding the 3 seats won in 2010 and gaining Ceredigion from the Liberal Democrats)

Liberal Democrats: 0 seats


What about the constituency-specific question? Well, for Uniform National Swing the projected seat results are identical to that for the generic question, with the two Cardiff seats plus Brecon & Radnor being the only seats to change hands. But when we use Ratio Swing, we get a slightly different outcome:

Labour: 27 seats (gaining Cardiff Central and Cardiff North, but losing Ynys Mon)

Conservatives: 8 seats (losing Cardiff North to Labour, but gaining Brecon & Radnor from the Liberal Democrats)

Plaid Cymru: 5 seats (holding the 3 seats won in 2010, and gaining Ceredigion from the Liberal Democrats and Ynys Mon from Labour)

Liberal Democrats: 0 seats


Those are the seat projections I get from applying these formulae to the polling numbers. Could I remind you all to please not refer to them as my predictions as to what will happen next week.


  • J.Jones

    I may be dumb but…What’s the question in the first part, the one captioned “Headline voting intention”?
    What does “likelihood to vote taken into account” mean? Are Yougov only counting “certain to vote” or have they a sliding weighting; something like 90% of all certain, 60% of 9 on the scale, 40% of 8 etc.?

    • Roger Scully

      I think it’s a sliding scale, whereby those choosing 9 on the 0-10 scale count as 0.9 of a full respondent, those choosing 8 count as 0.8 etc.

      Headline voting intention is just the response to their standard question on how you would vote in a general election.

      Sorry that these replies are slightly vague, Jon, but I’m at home today and the full details are somewhere in the office.

      • J.Jones

        How polls might misrepresent has always fascinated me. 73% are absolutely certain to vote but in the last election 64.75% actually did vote although it’s reasonable to suppose that that 64.75% didn’t come entirely from a group of people who describe themselves as “absolutely certain”.

        What we can say is that someone who is over 60 years of age, educated middle class background, and living in North Wales is very likely to vote.

        Someone who is 18 years old and with a poor education, living in the S. West of Wales is very unlikely to vote.

        So the most predictably “true” polling is in one section of the population and the most unpredictable and volatile polling is amongst a different section.

        We have Labour, Tories and UKIP drawing their support amongst the most stable voting population and Plaid and Greens drawing a substantial amount amongst those least likely to vote. And by least likely to vote I mean not what they say they will do but what the post election analysis suggests that they actually do.

    • Roger Scully

      Jon – sorry about not getting back to you on this sooner. Has been quite a busy week.

      Anyway, I checked. The standard YouGov general election voting intention question (which is very similar to that used by many companies) is: “If there were a general election held tomorrow, which party would you vote for?”

  • J.Jones

    The other question is…is this all that the poll asked; voting intention? Plaid didn’t ask Yougov to test some of their policies for popularity at the same time did they?

      • J.Jones

        The reason that I asked is because, obviously coincidentally, there was an article in the Daily Post saying that Plaid had polled through Yougov on questions like “Who do you think best represents Wales in the General election” (Leanne Wood got the majority vote.) There was also a question on Wales’ finance something like “Do you think that Wales should be allocated the same funding as Scotland or less?”… a question that references one of Plaid’s headline claims ( I think 78% said that we should).

        Obviously Plaid is rolling in cash and carried out separate polls.

        From my own experience with online opinion polls they are rarely limited to just voting intention. I quite accept that which mobile provider I use or would choose doesn’t influence my voting intention but some lists of questions do focus my attention on a set of general viewpoints…only to find at the end of the poll a quick question on voting intention.

  • David Jones

    I don’t understand why someone would change their voting intention for “thinking about your constituency”. Where else will they be voting?

    Why are the electorate so stupid?

    • Roger Scully

      Because most people – whether stupid or not – don’t think about electoral and party politics most of the time. Many of them do need to be reminded of the specific context of their own constituency – and parties often put quite a lot of effort into doing so.

  • Welshguy

    Difficult to know what to make of Plaid Cymru these polls as they’re coming from a relatively low base, so the margin of error could obscure any meaningful trend. The difference between Plaid polling 10% and 15% is huge (the equivalent, relatively speaking, of a 15 point swing for a party polling around 30% like the big 2 do): 10% would be a grave disappointment for Plaid even if they hold onto all their current seats; but to get 15% would be an impressive result and bode very well indeed for next year’s Assembly election.

    electionforecast.co.uk’s figures seem to intriguingly suggest that Plaid will do especially well in some S Wales seats, possibly earning a higher share of the vote in places like Llanelli, Rhondda and Neath than in Ceredigion or Ynys Môn (but still being more likely to win the latter two due to divided opposition). God knows how accurate their figures are though, and anything still seems possible!

Leave a Reply

Your email address will not be published. Required fields are marked *