To Boldly Go Where No Welsh Poll Has Gone Before

A few people have asked me recently about whether there is any further polling scheduled in Wales before the Assembly election. Obviously I can’t speak for others – there may be someone else planning to run a poll that I am unaware of. Perhaps Lord Ashcroft will suddenly develop a burning desire to know about the likely outcome of the Welsh Assembly election? Who knows? However, I can tell you about what the team behind the Welsh Political Barometer – ITV Cymru-Wales, The Wales Governance Centre at Cardiff University, and YouGov – are planning.

First of all, there will be one more of our regular Barometer polls before the election. As is normal for the ‘final call’ poll prior to an election, we will be timing this poll for as close as possible to the Assembly election. So sampling will continue until well into Wednesday 4th May. The results will then be published just after 6pm that evening – on Wales at Six, the ITV website and also this blog. The poll should run most of the normal questions that we include; however, given the proximity of the Assembly election, we will doubtless focus most of our attention on the vote intention numbers for the devolved election. As per normal, we will also publish Uniform National Swing seat projections (and I’ll put Ratio Swing seat projections as well on a post on this blog).

That would normally be that. But in addition, we are planning something extra for you this year. On election day itself our friends at YouGov will be conducting an ‘On-the-Day’ poll in Wales: where they approach a sample of people online and ask whether or not they voted, and – for those that say they have voted – which way their cast their ballots.

This on-the-day model is something that YouGov have tried twice before. In the 2014 Scottish independence referendum their on-the-day poll was the closest to the actual result of any poll published in the final week. Their on-the-day poll at the 2015 general election was somewhat less successful; like the final polls conducted before the election it somewhat understated Conservative support and over-stated Labour. In short, the on-the-day poll was ‘wrong’ for the same reasons that the pre-election polls were ‘wrong’. YouGov have adjusted their methodology significantly since the general election (as discussed here), and we must hope that these adjustments will allow us to identify patterns of voting on election day with a greater degree of accuracy.

One thing I must try to make clear is that an on-the-day poll is a distinctly different creature from the Exit Polls that are conducted at the general election. The Exit Poll conducted last year asked voters at a large number of polling stations across Britain to submit mock ballot papers, immediately after they had cast their actual votes. It had a very large sample (approximately 22,000 voters) gathered from more than 140 polling stations within 133 constituencies. (Mainly for that reason the Exit Poll was a very expensive exercise, with its costs shared between the BBC, ITV and Sky News.) The large sample, and the substantial variety of places where mock ballots were gathered, allowed the team analysing the figures to examine patterns of electoral support in various different types of seat. Those patterns could then be compared with very detailed information about behaviour in the previous election: because we know the exact election result in each constituency, and have a very good idea – from the history of both general and local elections – of the behaviour in particular wards within those constituencies. From all this information the precise seat projections could then be derived by Prof John Curtice and his team.

Our Welsh on-the-day poll will necessarily be a rather less elaborate (and very much less expensive!) operation. YouGov will seek to gather information on the behaviour of a representative sample of more than 1,000 voters across the whole of Wales. That information will be generated via YouGov’s normal methodology of internet sampling. Specifically, YouGov will re-contact respondents who participated in their final-call poll, and in the final week of sampling for the Welsh Election Study, to ask them how they voted. By doing this we aim to see if there has been any change in voting, and estimates of the national vote share for each of the main parties will be produced.

What we will not be able to do is produce detailed figures by region, or by type of seat. We will, though, use the estimates of national vote share from YouGov to produce an initial seat projection of the election result. Those figures will be released by ITV on their live Welsh show just after 10.30pm on election night.

How close will our on-the-day poll be to the final result? Obviously, I very much hope that it will be very close. The poll will give us our first indication of which way the election result appears to be heading. But of course we have never done an exercise like this before in a devolved election, so there must be some risk that the poll will prove to be less accurate that we would all hope for. And even if the overall national vote share figures turn out to be broadly correct, one thing that will need looking out for on the night will be differential swings in different places. In 2011, for instance, there was a notably stronger swing to Labour in south Wales than in the Mid & West Wales and North Wales regions.

All that said, for the first time ever in a National Assembly for Wales election – indeed, I believe for the first time ever in any devolved election – we will have some indication of which way things might be heading very soon after the polls close at 10pm. I think both ITV and YouGov should be warmly applauded for this initiative, which I hope will add an interesting new dimension to Welsh Assembly election night.


  • Harry Hayfield

    And don’t forget you will have a real election in a real ballot box (Mynyddbach on Swansea) to help with your calculations in Swansea East constituency as well!

  • J.Jones

    How will you select your sample Roger? I mean, to maximise your understanding of how people have voted and what the likely outcome will be you would have to get 1000 people who have said that they were 10/10 likely to vote.These will be the people that YouGov approaches. The problem may be that people who were 9/10 and 8/10 likely to vote actually have voted but are less committed, less knowledgeable voters voting with a different agenda.
    The other thing that I have noticed was a recent piece from the WM:-


    Look at Plaid’s social media presence. It dwarfs that of other parties and I do believe that online pollsters fail to recognise how Plaid supporters “game” pollsters.

    • Roger Scully

      I won’t select the sample, JJ – I leave that up to YouGov. They will be going back to people who they interview in the immediate pre-election poll, and in the last week or so of our (separate) Welsh Election Study campaign wave. They will ask people if they have actually voted, and if so in what direction.

      On your second point: yes, I am aware that Plaid supporters tend to be particularly active on social media. Same is true of SNP supporters. Mostly down to the age and social profile of their support base, I think. In any case, there is no record of YG polls in Wales tending to over-state Plaid support. The fairly consistent over-statement has been of Labour.

  • Brian Hughes

    Dr Scully, J. Jones is one Jacque Protic, a notorious anti-Welsh bigot. How do I know? I overheard him telling someone, while obviously inebriated, in a pub in LlanfairPG last year.

  • J.Jones

    Brian. I don’t know about Jacques Protic but I haven’t been in a pub in Llanfair PG for years and I rather object to being called an anti-Welsh bigot. You keep popping up spreading a malicious falsehood and it happens to such a degree that it constitutes harassment. Take a break.

Leave a Reply

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