British Election Study research on the shock result of the 2015 General Election led to more accurate polling methods that have been adopted by many leading pollsters.

Impacts

  • Analysis of British Election Study (BES) data provided new weighting procedures to ensure more accurate future polls.
  • BES research underpinned the findings of the independent Inquiry into the 2015 British general election opinion polls, co-written by Professor Green, and commissioned by the British Polling Council and the Market Research Society.
  • BES recommendations to improve polling accuracy caused leading polling experts to change their methodology. Major polling organisations ICM and TNS adopted the advised method (using turnout weighting methods), and YouGov also increased efforts to recruit non-voting respondents.
  • A major 'vote validation' exercise, in partnership with the Electoral Commission, provided a benchmark of the voting population for scholars and non-academic data users, including the polling industry.
  • The BES findings increased understanding of the 2015 'polling miss' through presentations to industry and academic conferences, as well as the media. Their academic paper on the topic is now the third most downloaded survey methods paper of all time (1,144 downloads by June 2017) on SSRN, a major social science repository.
  • During the 2015 General Election and 2016 EU referendum the BES team's analysis on voting behaviours gained extensive coverage across TV, radio, newspapers, blogs and social media, and they were recognised as the first to forecast the Leave vote during ITN's referendum broadcast.
  • After the 2015 election, BES contributed the only external analysts to Labour's official internal review of the election defeat. Their analysis was presented to Margaret Beckett MP and Labour's Chief Executive, Iain McNichol.

“At a time when the whole concept of sample surveys was called into doubt, the ability of the British Election Study to show that well-designed and well-conducted surveys can still represent the population with a high degree of accuracy helped polling organisations – both by helping them understand what might have gone wrong and how to avoid similar problems in future, and by providing support for the whole idea of using surveys to measure voting behaviour and opinions on politics.” (Nick Moon, British Polling Council)

About the research

In the 2015 General Election the opinion polls were wrong – they predicted Labour to be the largest party and a government led by Ed Miliband, whereas the actual election result was an overall Conservative majority.

“Opinion polls are really hard to get right. But when they predict the wrong winner, people start to question how far they can trust not just polls, but data and experts in general. That has important consequences for politics, elections and public faith in surveys,” says Professor Jane Green.

BES data and analysis showed that population samples used in polling were not representative – they didn't capture 'hard to reach' voters, and the pollsters' weighting adjustments didn't correct properly for the skewed distributions of respondents. The BES team implemented procedures to maximise responses from hard-to-reach groups, keeping post-survey weighting to a minimum, and their data correctly reflected the 2015 Conservative lead.

“By surveying 'hard to reach' voters we could counter speculation on why the polls were wrong, with clear reasons that we have shared with the polling industry and the public,” adds Professor Green.

The close working partnership that BES has gained with the media has informed a greater public understanding of election results including the EU referendum and also President Trump's victory, Professor Green suggests. BES data, graphics and analysis now informs millions of viewers on election nights.