The Only Way is Ethics: When it comes to reporting polls, journalists need to stop being smug and wise up

Remember when political polls were infallible? Prior to the general election in May, there were plenty of people – pollsters, naturally, but also journalists and politicians, too – who took it for granted that polls of voter intention were bang-on. A Tory majority in the forthcoming parliament was far from most minds.

There was so much faith in the results, in fact, that many were prepared to dismiss the election night exit poll as erroneous, despite it being – in theory – more likely to be a better predictor. After all, an exit poll asks people how they voted, not what their intention is. Paddy Ashdown may never quite live down his reaction to the exit poll’s announcement, memorably offering to “eat my hat” if it was proved right.

Since May, polling companies have been scratching their heads to work out what went wrong – with few clear answers. Journalists, meanwhile, chortled at the pollsters’ misery, largely ignored their own complicity, and pledged not to be fooled again – at least until the next time. But following Jeremy Corbyn’s remarkable elevation to the leadership of the Labour Party, polls are suddenly right back in fashion.

Results from a recent example made the front page of The Independent a week or so ago, with the headline reading: “Corbyn ‘loses fifth of Labour voters’”. This, said a couple of readers, didn’t make sense. Well, maybe it did; and maybe it didn’t.

The poll, conducted by ORB, had asked respondents whether they agreed with the following statement: “With Jeremy Corbyn as leader of the Labour Party, I am MORE likely to vote Conservative at the next general election.” Among those who voted Labour in May, a fifth agreed or strongly agreed, hence our headline.

But, of course, the interpretation of the results of the survey might depend on how respondents understood the question. Did they think, for instance, that they were indicating a preference for the Conservatives over Labour? Or was the comparator actually time? In other words, did respondents believe they were being asked whether they were more likely to vote Conservative than they had been at the last election? That allows for a much more ambiguous conclusion, that people might be incrementally more likely to vote Tory now, but that isn’t to say they wouldn’t vote Labour.

After all, the poll also asked respondents separately whether they agreed with this statement: “With Jeremy Corbyn as leader of the Labour Party, I am MORE likely to vote Labour at the next general election.” Especially for those who voted for Miliband and co in May, this must have seemed much more obviously a comparison between discrete periods. Of Labour voters, 37 per cent disagreed with the statement, which means they must feel less likely to vote for a party led by Corbyn; but it doesn’t necessarily follow that they would vote for another party or, indeed, abstain. Similarly, can it be concluded that the 8 per cent of those who voted Tory last time but who agreed they were “more likely” to vote Labour with Corbyn in charge would actually do so?

Political polling is a vast and lucrative industry. And, in some ways, the recent reminder of polls’ imperfections makes them even more fascinating. Results will continue to make headlines. But if we have learnt anything from the last election, it’s that polls must be closely scrutinised – and we should consider what those being surveyed may have thought they were being asked before we make assumptions about the answers they gave.