Alongside Labour, the Liberal Democrats and the UK Independence Party one other group endured a terrible General Election campaign — the pollsters.
These organisations exist in order to give people a general sense of how Britain is likely to vote in order to help the parties understand the task ahead of them and voters make more informed choices. But for them to be useful, however, they have to be reasonably accurate.
And they weren’t…at all.
Here’s how wrong they got it — the average of the polls going into election day suggested that the Conservatives and Labour were effectively neck-and-neck. But on the day itself they weren’t:
This in accuracy lead to very large spreads between the number of seats that forecasters thought the parties would win and the number they actually did. Such was the confidence with which people were predicting a hung parliament this time around that some sites were effectively saying that it was nigh on a certainty:
And yet as I write the Conservatives have taken 323 seats, already enough to give the party a functional majority in the House of Commons if you exclude seats won by Northern Ireland’s Sinn Féin, which doesn’t take its seats in parliament, and the speaker. By the end of Friday that figure is expected to have risen to as high as 330 — in other words, an outright majority.
So what went wrong? Below are a few theories:
- Shy Tories — There has long been a thesis that supporters of certain parties are embarrassed to admit their voting preference when asked by pollsters. Basically, they lie. If this occurs on a sufficiently large scale it can skew the results as it looks like a whole load of people suddenly changed their mind on polling day.
- The assumptions were wrong — Every polling company weights their sample in order to make the results as representative of the general population as possible. If some of the assumptions they make when they do this are incorrect — for example, over-emphasising the importance of educational levels or the newspaper that people read in people’s voting decisions — then the weighted answer might not be as representative of the country as a whole as they think.
- The methodology was wrong — It could be even worse for pollsters: Their whole model could be wrong. Those that rely on telephone polling could be oversampling a falling number of people who still use land-line phones, leading to an over-sampling of particular groups. Or else online polls could be suffering from the fact that respondents are self-selecting and so could equally be unrepresentative.
- Turnout was significantly different than expected — Voter turnout can impact different parties in different ways and with turnout at 66% this time round (the highest since 1997) it could have had an impact. For example, it has long been considered conventional wisdom that a higher turnout favours Labour because the party enjoyed a structural advantage under the UK’s voting system (it needed to get a smaller share of the national vote per seat won than the Conservatives). However, the SNP surge has shifted that arithmetic and higher turnout in Scotland appears to have overwhelmingly benefited the nationalists while the higher turnout elsewhere does not seem to have helped Labour in the manner that history might suggest.
- People actually changed their minds on the day — It’s not impossible that the fundamental reason that the polls got it so wrong was that people weren’t really sure who they would vote for until they turned up at the polling stations. For example, there were some hints in the polling data that people in England were quite uncomfortable with the idea of Labour doing a deal with the SNP, and when it came to putting a cross in the box it might just have pushed them over to the Conservatives.
Whatever combination of the above is true, we can compare it to the most successful poll of the night — the exit poll. The release of the exit poll after voting closed was the first time people in Britain realised that things might not go as expected with the Conservatives forecast to win 316 seats, way ahead of what the polls had been suggesting, and Labour on 239, way behind what it was being projected to get.
Part of the answer could be that the joint BBC, ITN and Sky exit poll used a far larger sample (roughly 20,000 people from 140 polling stations around the country) and avoids the problem of people changing their mind by asking them how they voted after they had already done so. Again, however, it is also vulnerable to problems of methodological weakness and incorrect assumptions. But on the night it came out top.
There will be lots of soul searching done by pollsters over the next few days and Britons should know in advance of their assessments that the best case scenario for them is that we’re a nation of congenital liars.
NOW WATCH: Briefing videos
Business Insider Emails & Alerts
Site highlights each day to your inbox.