Last year's general election was an embarrassment for the polling industry © Bloomberg
Experimental feature

Listen to this article

00:00
00:00
Experimental feature
or

The polling industry is facing another furore over its work on the EU referendum, an influential pollster has warned, just days before the final inquest is published into what went wrong with last year’s general election predictions.

Polling companies were left embarrassed last year after a surprise Conservative election win, which none of them had accurately predicted. On average, the final published polls underestimated Tory support relative to that of Labour by 5.5 percentage points.

Recent surveys about people’s voting intentions in the forthcoming EU referendum campaign have raised fresh concerns about pollsters’ accuracy, with a substantial divide between the outcomes predicted by different polls.

Surveys conducted by telephone in the past year have consistently shown leads for the Remain campaign that are 15 to 20 points higher than those carried out online.

Until now it has not been clear why this is — or which of the methods is more accurate. But research published on Tuesday by one of the few people to predict the outcome of last year’s election accurately has suggested that, for the referendum, the phone polls are more likely to be right.

The difference in outcome is caused by two factors, conclude James Kanagasooriam, from polling company Populus, and Matt Singh, a polling analyst. These are: how the question is phrased, and the failure to take into account the wider social attitudes of the people surveyed.

Mr Singh, a former Barclays trader, correctly predicted the general election result and was vindicated after arguing that all the polls were wrong.

The findings by Mr Singh and Mr Kanagasooriam suggest that the campaign for Britain to remain in the EU has a larger lead in public opinion than many people believe.

The report also suggests that: “traditional [polling] modelling may no longer be sufficient to draw an accurate and representative sample of the voting public”.

Britain rarely holds a referendum, making it hard for pollsters to calibrate their methodology to focus on the target question.

On the EU referendum question, online polls almost always present an explicit “don’t know” option while telephone polls generally do not, meaning voters are “forced to choose”, Mr Singh said.

Meanwhile, pollsters have failed to take into account differences in social attitudes, particularly in relation to issues of equality and national identity, the research found. Having broadly liberal or conservative attitudes is a better indicator of an individual’s referendum vote than any other factor, the research found.

This means “people who seem otherwise alike in terms of who they are, where they live or how they’ve voted previously can turn out to be different on the question of the EU referendum”, according to Mr Singh.

People who commission polling work — such as politicians and the media — demand quick turnround times, which means, owing to the current polling methodology, they struggle to connect with hard-to-reach sections of society, such as those who are more socially liberal, the research found.

To improve their sampling, pollsters should pay more attention to social attitudes, the research said.

Populus — which publishes both internet and phone polls — and Mr Singh looked at 80 EU referendum polls that have been published since last year’s general election.

Copyright The Financial Times Limited 2017. All rights reserved.
myFT

Follow the topics mentioned in this article