‘Unrepresentative samples’ led to inaccurate electric predictions, report finds

‘Unrepresentative samples’ led to a failure to accurately predict the results of May 2015’s general election, according to a report from the British Polling Council and the Market Research Society.

Pollsters mistakenly predicted a hung parliament rather than a Conservative majority and the report suggests that this was down to pollsters using methods that ‘systematically over-represented Labour supporters and under-represented Conservative supporters’.

On the basis of this, the report makes a number of recommendations on how to improve polling accuracy. It advises that polls should review existing methods for determining turnout probabilities, as too much reliance is currently places on self-report questions which require respondents to rate how likely they are to vote.

The also suggests that pollsters should review current allocation methods for respondents who say they do not know or do not want to disclose which party they intend to vote for. Additionally, it recommends that pollsters should take measures to obtain more representative samples, as well as investigate new quota and weighting variables.

The reports reads: “On average the final estimates of the polling companies put the Conservatives on 34 per cent and Labour on 34 per cent. No individual poll put David Cameron’s party more than a point ahead. Yet in the event the Conservatives won 38 per cent of the vote in Great Britain, Labour 31 per cent. As soon as this discrepancy became apparent, the British Polling Council (BPC) and the Market Research Society (MRS) immediately agreed that they should jointly sponsor the establishment of an independent inquiry into the performance of the polls at the election.

“As an immediate result of this report, MRS will be working with the Royal Statistical Society (RSS) to update their joint guidance on the use of statistics in communications; issuing new guidance on research and older people; producing a simple guide for the public on how to read polls; and reminding accredited company partners of the elements of the MRS Code of Conduct which are particularly relevant to the issues raised in the report.”