Unrepresentative samples main cause of polling miss, finds Inquiry
Unrepresentative samples were behind the opinion polls' failure to identify the Conservative's lead over Labour during the 2015 General Election Campaign, according to a major report by the independent Inquiry into the 2015 General Election polls.
The Inquiry was established by the British Polling Council and the Market Research Society and was chaired by Professor Patrick Sturgis, Director of the National Centre for Research Methods.
Professor Jane Green, from The University of Manchester and Co-Director of the British Election Study, was on the nine-strong panel which produced the report.
The panel drew heavily on British Election Study data – and used data provided by pollsters who are members of the British Polling Council - to carry out the research. The conclusion of the report echoed findings of the BES team earlier in the year.
The Inquiry considered eight different potential causes of the polling miss. It concluded that unrepresentative samples was the main cause.
There may have been a modest 'late swing' to the Conservatives between the final polls and Election Day, contributing at most – around one percentage point to the error on the Conservative lead.
The report also concluded that there was evidence consistent with 'herding' among pollsters around the final result.
Professor Green said: "Because the opinion polls substantially underestimated the lead of the Conservatives over Labour, the election campaign convinced many voters that a hung-parliament and coalition government was almost a certainty.
"It is important that the public has confidence in the polls, but also that there is greater understanding about uncertainty in polling estimates, among the public, among politicians, and among journalists."
Among the recommendations, the Inquiry concludes that pollsters should report whether changes in their headline vote intention figures are significant in statistical terms. It is hoped this will reduce the amount of reporting of changes in party support when the underlying data rarely support this.
It recommends pollsters pre-register polls and make their data more available online, and a review of how pollsters use the data for respondents who say they don't know, or won't disclose, who they intend to vote for; and take measures to get more representative samples within the weighting methods they use.
It also recommended that the Economic and Social Research Council should fund a pre- and post-election random probability survey as part of the British Election Study in the 2020 election.