Six out of ten people don't care about polls

Six out of ten people don’t pay any attention to opinion polls, Ipsos Mori’s Ben Page told the Today programme this morning.

Page was speaking alongside Tim Harford of Radio 4’s statistics programme, More or Less, on the question of whether we really need opinion polls.

In a noble effort to improve the public’s understanding, Page and Harford spent some time discussing how polls work and what they really mean. But the difficulty is that however upfront pollsters are with the caveats and disclaimers, what the media and the public essentially want from an opinion poll is the answer to the question Who’s going to win?

Most newspaper editors and their readers skip past the boring small print about margins of error and snapshots in time in the same way that they click ‘Agree’ to the terms and conditions of some new piece of software. They don’t care – and pollsters know it. There’s an unspoken agreement that they just want to know Who’s going to win?

Page pointed out this morning that the overall performance of voting intention polls over time is actually very good (on average, less than 1% out in the last general election and only 1.5% out in every election since 1950, he claims). But it’s the cock-ups that people remember. The UK’s pollsters famously failed to predict the Conservative win in 1992, and in New Hampshire in 2008, nearly all the pollsters predicted a clear Obama win in the democratic primary, only to watch Hillary Clinton coasted to victory. What happened? Was it the female voters? The late deciders? Was it Clinton’s appearance crying on TV? Who cares? All that matters to the public is that the polls failed to tell us who was going to win.

Asked whether he thinks the polls will get it right this year, Harford said: “I think the polling companies are doing their best but in a year when everything seems to be up in the air, I think [it’s] more likely that the opinion polls will be misleading.” That seems to be a common view – differences between the parties are barely bigger than the margins of error in a lot of the recent polls, and there are all sorts of weird factors making the outcome tricky to predict.

Even if the polls get it almost right, that will be enough to provide a basis for plenty of outraged articles on 7th May. Those of us who work in the industry know that polls are useful even though they have their limitations. But expecting the public to understand or care about those limitations might be asking too much.