June 1, 2017 by Paul Goldsmith
Polling is a dangerous business. No-one predicted the Conservatives’ majority in 2015 and the EU Referendum result came as a surprise to most people. The polling industry is thus being treated with the type of care taken these days when removing asbestos from an old building. BUT, a poll released yesterday (before the televised debate) by YouGov deserves notice because it is the first one to predict that the Conservatives will NOT win a majority. As well as explaining what they say, it’s worth also looking at their methodology and why they are confident that this rather startling finding could be accurate.
The model is showing that the Conservatives are set to take 310 seats (16 short of a majority), with Labour 257, the SNP 50 and the Liberal Democrats on 10. The other seat counts aren’t particularly notable. The big story is the hung Parliament that could result, from an election designed to increase Theresa May’s majority to help her push Brexit through. In the modern history of calling shock early elections (February 1974 and June 1970 being the best examples, the party that called them has been punished at the polls. But this would be a massive shock.
Yougov admit that the predictions are a midpoint for their model, and the Conservatives could end up with as many as 345 seats (a majority of around 40) and Labour with as few as 223 seats. YouGov also point out that this is just one snapshot based upon data from the past seven days (taking in the social care shambles, Manchester and the Sky News TV ‘debates’. People do change their minds in the closing days of a general election. But they are confident in their model, for a specific reason. It successfully predicted the EU Referendum result, always having Leave ahead, when others weren’t.
Yougov models every constituency, and key voter type in Britain. It uses an analysis of demographics and voting behaviour in the 2015 general election AND last year’s EU referendum. Turnout is based on voter demographics and is based on analysis from the authoritative British Election Study data in 2010 and 2015. This is important because, unlike during the referendum, there is recent data to compare the polling results with to sense check them. Furthermore, adding the EU referendum data takes account of the political shift that happened last year, which is an added layer of information that could push this data towards being realistic.
YouGov is conducting 7,000 interviews with registered voters every day about their voting plans in their constituency. The data is then used to assess how each type of voter will shape the race in each type of constituency.
This results in a model that calculates daily voting intention and seat estimates, based upon the fact that people with similar characteristics tend to vote similarly, but not identically, regardless of where they live. Yet here is a problem, because it cannot account for specific local factors that may shape the vote in some seats (like the nuclear issue did in the Copeland by-election earlier this year when a Tory won a historically Labour seat on the back of concerns about the future of the Sellafield nuclear plant under Labour).
IF the data is accurate, and it is a big IF, we could be about to see one of the biggest election upsets in history. If that happens, it will be a result of the Conservative campaign being one of the worst and most misguided in history, based upon a personality contest relying on a leader with no personality as well as a manifesto which offered so little to sell on the doorstep (which in some ways was sensible if you are sure you will win, but now turns out to be a massive risk). Ducking last night’s debate didn’t reflect well on Theresa May either, although to coin one of her own phrases, I would agree that no Theresa May is better than this, bad Theresa May.
There are many ways to lose an election. I know, because my next book is about the many ways to lose an election. But if YouGov is right, this one will have been a spectacular misjudgement of the electorate.