U.K. Pollsters Were Talking to the Wrong Voters, Study Findsby
Polling companies struggled to reach young and working class
Larger, slower, face-to-face survey got closer result
The failure of British polling companies to predict David Cameron’s election victory last year may have been because they were talking to the wrong people, according to an analysis by one of the country’s leading political academics.
The result, with Cameron winning a majority that few had predicted, meant a crisis not just for the opposition Labour Party but also for pollsters. The British Polling Council ordered an immediate inquiry, which will present its interim conclusions on Jan. 19.
John Curtice, professor of politics at Strathclyde University in Glasgow and president of the BPC, analyzed the results of the British Social Attitudes survey, an annual look at how the country views a wide range of issues. In a paper published Thursday, he argued it had been more successful than regular opinion polls in finding supporters of Cameron’s Conservative Party.
“The source of the polls’ problem did not lie in how the pollsters’ questions were answered but rather in who was answering them in the first place,” Curtice wrote.
In particular, the pollsters were better at finding middle-class than working-class people and worse at finding the young than older voters. Because this meant that those they did find were given greater weight, there was a danger of error if the interviewees were unrepresentative of their groups. Curtice also found evidence to support the so-called “Lazy Labour” effect, with non-voters more likely to be Labour-supporting.
While most political opinion polls rely on one or two questions asked of around 2,000 people by telephone or online, more than 4,000 people are interviewed face-to-face for the BSA. And whereas pollsters generally keep trying to contact new people until they have enough responses, those conducting the BSA identify target interviewees and make at least six attempts to contact each.
In practice, this method is likely to be too slow and time-consuming to be of practical use in trying to predict election outcomes. It’s also not perfectly accurate. Curtice’s paper shows that, while the BSA did find -- after the event -- a Conservative lead over Labour that eluded other pollsters, it overestimated the share of the vote that both Labour and the Tories would get by about 2 percent.