Electoral Unpredictability
By Andrew Hawkins, Chairman

Theresa May claimed that her decision to call last week’s snap election was made while on a walking holiday in Wales before Easter.  Well, there is a long list of people who wish Theresa May had found another way of relaxing - including UKIP, the Lib Dems, Greens, the media (facing its third national election in as many years), pollsters and, of course, Theresa May herself.


From our perspective at ComRes, it was the third time our Voter Turnout Model had been used, and the second in a national election.  As I wrote on 30 May, the forecast of a 100+ Tory majority had been predicated on voters behaving ‘in a way they broadly did in 2015’ – which, of course, they didn’t.  We have already embarked on an in-depth look at what needs doing to improve accuracy for next time.  The BBC’s Peter Barnes opined of the pollsters’ performance more generally that ‘it looks like the errors were caused at least in part, by fighting the last war’. Yet inevitably we must rely to a large extent on historical behaviour to predict it in the future.


One pointer to the shock result was contained in the detail of our final, eve of election poll.  We had been tracking reported postal votes for the final month or so of the campaign, but were prohibited by law from publishing them.  By the end of May the Tories had an 11-point lead among the 16% who said they had voted in this way.   This rose a week later to 16 points as the proportion of postal votes climbed to 23%.  But on the eve of polling day we found 26% had voted by post but that gap had closed to just 5 points, fuelled by a surge in postal voting among Labour-supporting under-35s.


The other surprise of that final poll was the near absence of any variance between reactions to a Corbyn-led government and that of a May-led one.  When we asked voters to imagine how they would feel waking up on Friday to a government led by either leader, exactly 50% said they would be “delighted” with each respective outcome.  On almost every other measure there was no significant difference between reactions to a Labour or Tory win.  Having enjoyed eye-watering poll leads just weeks earlier, these were all signs that the wheels were well and truly off the Conservative campaign.


The outcome produced a result that is unique in recent political history: the largest combined vote share for the two main parties since 1970, at 84.5%.  It must feel that the bar has just been raised impossibly high for both Labour and the Conservatives.  Labour’s vote share, at 41%, was its highest since 2005 when Tony Blair won a majority of 66.  The Conservatives’ vote share, at 43.5%, was its highest since 1983 when Margaret Thatcher won a majority of 144.


There are several obvious consequences.  First, will the next election be as volatile?  If a 40%+ vote share no longer wins you a majority, and more importantly if voters actually change their minds (and thus their behaviour) during election campaigns, then electoral uncertainty levels go off the scale.


Second, campaigning needs to change.  Voter sentiment can turn 180 degrees almost overnight.  Nothing can be taken for granted, including pledges of party loyalty collected more than a few days before polling day.  At national level, there is now an expectation that party leaders are accessible, open and willing to debate.  It was clear back in 2010 that the TV debate genie was out of the bottle.  After this election, only a fool would make the mistake of refusing to take part.


Third, we shall be taking a close look at how we can revise our modelling to take account of this new reality.  Given the fluidity of decision making, one obvious need is to conclude our final poll as late as possible to account for last-minute changes of mind.


We got much about GE2017 right: the initial popularity of Theresa May and her Party, the toxicity of the Conservative Manifesto, the surge in support for Labour, the collapse of UKIP.  Having been closest in 2015, we will be working to get once again not only the underlying realities correct but the headline vote shares too.


There has been much speculation about the timing of when the next national electoral test will come.  Whenever that is, ComRes will be ready, but we stand alongside Brenda from Bristol in hoping that it won’t come around too quickly.