In June 2015, Professor John Curtice, professor of politics at the University of Strathclyde, came to share his thoughts on the performance of the polls in anticipating the outcome of the 2015 UK General Election. Curtice delivered this talk in his personal capacity, and not in his role as president of the British Polling Council.
When we first invited Professor Curtice to give a seminar on the performance of the polls we had no idea that the 2015 election was set to join the 1970 and 1992 elections as an occasion when the polls were perceived to have got it ‘wrong’. On average the pre-election polls predicted a dead heat between Labour and the Conservatives; in the event the Conservatives were seven points ahead and unexpectedly won an outright majority.
Curtice started by suggesting that the reason there is so much criticism of the performance of the polls in the 2015 election is that they got the story wrong. The opinion polls were widely regarded as indicating that there would be a hung parliament when the Conservatives won an overall majority. In 2001 there was in fact just as big a difference between the average level of Conservative and Labour support in the final polls and the eventual result, but as in the polls still pointed to a Labour overall majority the error did not receive much attention. The 2001 polls simply pointed to Labour winning by a larger majority than proved to be the case.
He described how the 2015 opinion polls were conducted and assessed some of the possible reasons why they overestimated Labour and underestimated Conservative support. There were two methods used for data collection: volunteer online panels and partially random and partially quota telephone samples. Both methods are potentially problematic. For example, the response rates are completely unknown. Equally, in both cases the samples often have to be quite heavily weighted in order to make them look demographically and politically representative.
Curtice outlined some possible explanations for the discrepancy between the polls and the results, including the ‘lazy Labour’ and ‘shy Tory’ theories. The ‘lazy Labour’ theory hypothesises that there was differential turnout between Labour voters and Conservative voters, and that the attempts the polling companies made to weight their data by reported propensity to vote did not adequately capture the relative reluctance of Labour supporters to go to the polls. The ‘shy Tory’ theory hypothesises that some Conservative voters do not like to tell polling companies that they support the Conservatives. However, it is not clear that either theory provides an adequate account of what went wrong. Those polling companies that have gone back to the people they interviewed in their final polls and asked them how they actually voted have still failed to find anything like a seven point lead for Conservative over Labour. That suggests the samples may not be adequately representative.
It was noted that the opinion polls were quite accurate for the smaller parties; the reason for the debate about the polls was the error in their estimate of Conservative and Labour support. There was discussion about whether the weighting strategies were appropriate as the adjustments may be serving to overestimate Labour support. This is because they typically involve both upweighting younger (Labour) voters and downweighting middle class (Conservative) voters.
Curtice finished by reminding the audience that he was speaking in a personal capacity, and that the British Polling Council has set up a wholly independent inquiry into the performance of the polls and people can submit evidence to the inquiry at www.ncrm.ac.uk/polling/.