Patrick Sturgis et al, “Report of the Inquiry into the 2015 British general election opinion polls”, London: Market Research Society and British Polling Council.
The opinion polls in the weeks and months leading up to the 2015 General Election substantially underestimated the lead of the Conservatives over Labour in the national vote share. This resulted in a strong belief amongst the public and key stakeholders that the election would be a dead heat and that a hung-parliament and coalition government would ensue.
In historical terms, the 2015 polls were some of the most inaccurate since election polling first began in the UK in 1945. However, the polls have been nearly as inaccurate in other elections but have not attracted as much attention because they correctly indicated the winning party.
The Inquiry considered eight different potential causes of the polling miss and assessed the evidence in support of each of them.
Our conclusion is that the primary cause of the polling miss in 2015 was unrepresentative samples. The methods the pollsters used to collect samples of voters systematically over-represented Labour supporters and under-represented Conservative supporters. The statistical adjustment procedures applied to the raw data did not mitigate this basic problem to any notable degree. The other putative causes can have made, at most, only a small contribution to the total error.
We were able to replicate all published estimates for the final polls using raw micro- data, so we can exclude the possibility that flawed analysis, or use of inaccurate weighting targets on the part of the pollsters, contributed to the polling miss.
The procedures used by the pollsters to handle postal voters, overseas voters, and un- registered voters made no detectable contribution to the polling errors.
There may have been a very modest ‘late swing’ to the Conservatives between the final polls and Election Day, although this can have contributed – at most – around one percentage point to the error on the Conservative lead.
We reject deliberate misreporting as a contributory factor in the polling miss on the grounds that it cannot easily be reconciled with the results of the re-contact surveys carried out by the pollsters and with two random surveys undertaken after the election. Evidence from several different sources does not support differential turnout misreporting making anything but, at most, a very small contribution to the polling errors.
There was no difference between online and phone modes in the accuracy of the final polls. However, over the 2010-2015 parliament and in much of the election campaign, phone polls produced somewhat higher estimates of the Conservative vote share (1 to 2 percentage points). It is not possible to say what caused this effect, given the many confounded differences between the two modes. Neither is it possible to say which was the more accurate mode on the basis of this evidence.
The decrease in the variance on the estimate of the Conservative lead in the final week of the campaign is consistent with herding - where pollsters make design and reporting decisions that cause published estimates to vary less than expected, given their sample sizes. Our interpretation of the evidence is that this convergence was unlikely to have been the result of deliberate collusion, or other forms of malpractice by the pollsters. On the basis of these findings and conclusions, we make the following twelve recommendations.
BPC members should:
The Economic and Social Research Council (ESRC) should:
I am an Associate Professor of Social Research Methods at the London School of Economics and Political Science. From 2016-2019, I am an Associate Editor of the American Political Science Review. My research is focused on the measurement of political preferences from survey, voting, network and text data. Applications of these methods have included citizens, legislators and judges in the US, UK and EU.
To receive updates from this site, you can subscribe to the RSS feed of all updates to the site in an RSS feed reader