The ups and downs of opinion polls

Polls are a central feature of modern democratic politics, but wayward electoral predictions have caused many to lose faith in them

Where did opinion polling begin?

Like most modern political phenomena, in America. Between 1916 and 1932, a popular news magazine, the Literary Digest, attracted much publicity by correctly predicting the results of presidential elections by mailing millions of mock ballots to voters and counting the returns. But in 1936, George Gallup, a statistician turned market researcher, became convinced the magazine was heading for a fall: the mailing lists it used came from telephone directories and car registrations, and thus excluded poorer voters. So Gallup conducted his own poll. It was far smaller but it used “quota sampling”: it targeted a proportionate mini-electorate, reflecting the race, age, sex and income ratios of the real one. Gallup correctly predicted both that FDR would win, and that the Literary Digest would wrongly call his rival, Alfred Landon, as the victor at 56%. The Literary Digest went out of business, and the polling industry was born.

Was it greeted with enthusiasm?

From the beginning there were doubts: the word pollster was initially an insult, like “huckster”. But polls were enthusiastically adopted by the media, and by interest groups eager to commission pollsters to ask leading questions and gather “scientific” proof that the public agreed with their concerns. Pollsters were soon claiming to be able to assess matters as elusive as whether we believe in God, whether we’re happy, and who we trust. Their surveys have come to influence everything from policy on such key questions as immigration, gay marriage and drug legislation, to what kind of consumer products we get offered. For Gallup, the new technique offered the best way to “take the pulse of democracy”. But many remained sceptical. To French philosopher Jean Baudrillard, polls were a sorry reflection of a mass media age in which “there is more and more information, and less and less meaning”. They fail to distinguish between strong and flimsy preferences, say the critics, and they tend to trivialise politics.

In what way are polls said to cheapen politics?

It’s often suggested they encourage politicians to pander to the latest popular sentiment – usually based on ignorance or short-term selfinterest – rather than making longterm decisions for the general good. Polls are also said to divert attention from the issues: their dominant role in the news narrows the media focus to the performance of the leaders, turning elections into horse races. But the most telling criticism is that polls undermine the democratic process by influencing the election result itself.

And do they influence elections?

Almost certainly. For a start, they often affect the timing of elections, because politicians only call elections when they think they can win. Research over the past 30 years suggests they also influence the way people vote, particularly in the late stages of a close-fought campaign. They can encourage more tactical voting, as voters bear in mind the likely balance of power based on polling predictions. They can also induce a “bandwagon effect” – voters plumping for the party that the polls deem the likely winner. A detailed study of the 2003 Netherlands election found some 20% of voters were affected by polls, with many swinging behind the surging Pvda (Labour) party. Such distortions of “the people’s will” are even more striking when predictions are wrong.

And have they often been wrong?

In recent times spectacularly so, both in the US (where almost all polls predicted a win for Hillary Clinton), and in Britain, where the pollsters have called three UK elections wrongly in three years. In the 2015 general election, they said Labour and the Tories were neck and neck; in the EU referendum, they predicted a win for Remain; in this year’s election, they foretold a clear Tory win, wildly underestimating Labour’s vote share. This distorted media coverage: in 2015, the media fixated on the prospect of a hung parliament and the SNP holding the balance of power. As a result, many voters probably cast their vote on a false prospectus. “If the polls had reflected reality,” said an aide of then-labour leader Ed Miliband of the 2015 election, “it would have been a totally different campaign.” It has also been suggested that many voted for Brexit as a protest, thinking Leave would lose, and then suffered pangs of “Regrexit”.

What can be done to correct such distortions?

Some 40 nations now ban opinion polls in the run-up to elections, to avoid affecting the result. In Israel, it’s illegal to publicise an election-related poll in the four days before the vote. In Norway, there’s a 24-hour ban; South Korea has a blackout lasting seven days. However, in the US such a ban would be unconstitutional on freedom of speech grounds; it would probably be unacceptable in this country too. In any case, for all their shortcomings, polls are still arguably preferable to the alternative. “I think a society is likely to operate more effectively if it understands itself better,” says the psephologist John Curtice. Without polls “all you get is a bunch of politicians saying everybody thinks that, or everybody thinks this, we’re winning votes, no, we’re winning votes. How the hell do we know who’s right?” The best solution is probably for the media to give less credence to polls (the BBC now won’t lead a bulletin on a poll result) and for the polling companies to improve their accuracy.

How can the polls get better?

An in-depth study commissioned by the British Polling Council after the 2015 result has identified a series of failings, including giving insufficient weight to postal votes and to the low turnout among certain social groups. Pollsters were also considered to be guilty of “herding” – adjusting their results to bring them closer to other companies’ results. But the basic problem was “unrepresentative sampling”: their samples of voters under-represented some types of voter and over-represented others – a problem that goes to the very heart of polling (see box).