At 10pm on 7th May the exit polls were met with gasps as they showed an outright victory for the Conservatives. But how could this be? For months the polls had shown a dead heat between Labour and the Conservative Party; a country with a First Past the Post electoral system, which has more often than not returned a strong government in the face of a minority vote, was surely heading for a second coalition government. The only question was the colour of the lead party and which of the smaller parties would support them. Suggestions of a Labour government supported by the SNP were dismissed out of hand with senior spokespersons declaring ‘we’re going out to win’. Even the Independent came out in support of the continued coalition between the Conservative Party and the Liberal Democrats. So sure was everyone that no-one would have bet against another coalition. The outcome left the pollsters dumbfounded, irrespective of the agency or the methodology used, support for the Conservative Party was consistently under-estimated.
Early explanations for the failure of the polls include low Labour turnout, last minute switching between parties, ‘Don’t knows’ making up their minds. My concern with this is that we are missing out on some fundamental issues. Warning – I’m about to raise an issue that is not sexy. It is not as interesting as which party might’ get into bed with another’, ‘what are the red lines – the deal makers and the breakers’ but it needs to be tackled it is a concept called margin of error. The statistics around survey research can get complicated, but I will keep this simple. Assume, an opinion poll finds that 30% of respondents intend to vote Labour. That 30% is an estimate and all survey estimates have some margin of error associated with it (error can occur for a number of reasons which I don’t have space to explore here but accept that it is natural, the job of the survey researcher is to minimise). If 1,000 people were surveyed there is a margin of error of approximately plus or minus 3%. This means that the likelihood of the population voting Labour would be somewhere between 27% and 33%. Let’s assume the same survey finds 30% of the population are voting Conservative. Same margin of error and logic applies. To use a sample to estimate the outcome of a larger population is perfectly legitimate but the sample needs to be representative of that population. Assuming that there was some inherent bias which means that the poll under-estimates support for the Conservatives and over-estimates support for Labour, such that the level of support for the Conservatives is really 33% and support for Labour is down at 27% suddenly we have a difference of 6% and predictions of a coalition are looking a whole lot more shaky!
Opinion polls are important and they’re not going away. That is why everyone, particularly the pollsters and the media, have a responsibility to ensure that the polls are reported in an open and transparent way that people understand. Unless such issues are addressed the pollsters are at risk of seeing the same failures repeated in the future. It could be argued that we’re already seeing history repeating itself, with the phenomenon of ‘the shy Tory’ given as the explanation in 1992 as the reason for under-estimating Conservative Party support by 5%.
This week YouGov released a poll about the Labour leadership election which suggests that Jeremy Corbyn could win on the first round with 53% of the vote. There have been reports of halls packed out with people who have turned up to hear Corbyn speak and not having been able to get in and shouts of ‘Jez we can’. So it’s a done deal then? Corbyn’s got this all sewn-up? Not only should Liz Kendall withdraw from the race but the rest of the field might as well pack up and go home too? How quickly we forget the lessons of our past. Less than 3 months out from an election where the polls were somewhat off I would urge caution. Not only for the reasons I’ve laid out about opinion polls, but let’s look at the other evidence. In terms of CLP nominations, 152 constituencies nominated Corbyn. That means that 228 didn’t. Analysis has shown that CLP nominations don’t always convert into votes and that constituencies who have nominated a left-wing candidate haven’t gone on to vote in the same way. CLP nominations will come down to who you can get out to a meeting on a particular night and if the campaign came down to mobilisation Corbyn would surely be the outright winner. But winning the ground-war isn’t enough. A ground-war can win over hearts but not minds; the Labour Party demonstrated this in May. By every account, Labour won the ground war in mobilising activists over the Conservatives, but it wasn’t able to convince enough voters that they had the right proposition to govern. Don’t get me wrong, I’m not saying Corbyn can’t win – there is a very good chance he could. What I am saying is that no-one should think this is a done deal. With ballot papers starting to drop through people’s doors and four weeks before voting closes there is – to end on a cliché – all to play for.