The website of Nate Silver, the American polling expert, surveyed all of Britain’s public-opinion surveys on Election day in Britain and declared that the chance that David Cameron’s Conservatives would win a majority of seats “was vanishingly small when the polls closed — around 1 in 500.”
But that is precisely what happened, leading Nate Silver to write a piece titled “The World May Have a Polling Problem.” He listed the errors that overtook “probably the four highest-profile elections of the past year, at least from the standpoint of the U.S. and U.K. media”:
1) The final polls in the Scottish-independence referendum showed the “no” side winning by just 2 to 3 percentage points. It won by nearly 11 percentage points.
2) Polls “significantly underestimated the GOP’s performance” in the 2014 midterm elections last November. In the Senate, GOP margins were about 4 points better than the polls in the average Senate race. The 2014 election was, on average, tied with 2002 as the second-worst polling year for Senate races in a quarter century (the worst was 1998).
3) In Israel, polls badly underestimated the performance of the right-wing Likud Party, projecting it would win about 22 seats in the Knesset; in fact, it won 30.
Silver came up with various explanations for the errors, noting first of all that voters are becoming harder to contact, so pollsters rely less on direct contact and more on online questionnaires. Some of those online polls abandon probability sampling, the bedrock of polling methodology. In addition, he also observed that “some pollsters have been caught withholding results when they differ from other surveys, ‘herding’ toward a false consensus about a race instead of behaving independently.”
Two examples of “herding” in the 2014 election appeared in the Kansas and Virginia Senate races. Pollsters there didn’t release their numbers because they deviated too much from what others showed. In that way, Virginia voters were badly served because no late published poll caught Republican Ed Gillespie’s last-minute surge that almost defeated Democratic incumbent Mark Warner.
In Britain’s election last week, we see another troubling instance of “herding.” The polling firm Survation admitted that its final poll showed the Conservatives with a lead of 37 percent to 31 percent over the Labor Party — almost the exact final result. The company’s CEO explained why he failed to publish the poll: “The results seemed so ‘out of line’ with all the polling conducted by ourselves and our peers — what poll commentators would term an ‘outlier’ — that I ‘chickened out’ of publishing the figures — something I’m sure I’ll always regret.”
This is playing with dynamite. Pollsters in Britain have long realized the potential polling problem created by the “shy Tory” vote — referring to those voters who don’t want to admit to pollsters that they are going against the grain of media coverage and might cast a politically incorrect vote. But this is the third time in 50 years that the shy Tories have swung an election (it also happened in 1970 and 1992). Apparently, polling companies haven’t been able to solve the problem. Chris Hanretty, the academic who advised the BBC on polling, told Britain’s Guardian newspaper that he feels “a little bit foolish,” in the wake of the election. “We categorically ruled out a [Conservative] majority. . . . We should have expected far more ‘shy Tories.’”
“There is a pro-Labor and anti-Conservative bias in polls.”
Pollster Stephen Fisher also told the Guardian that the problem is more complex than that. Noting that polling companies have consistently exaggerated the Labor vote since the 1970s, he concluded: “Polling companies have done a lot of work to try to counteract this, but it keeps happening. There is a pro-Labor and anti-Conservative bias in polls.”
Rem Korteeweg, a senior research fellow at the Center for European Research in London told the New York Times: “People say who they are voting for with their heart and then vote with their wallets.”
#related#I’m not sure that’s the case, but even if it is, the problem of people giving politically correct responses to pollsters isn’t confined to Britain. As Nate Silver concluded, “Polls, in the U.K. and in other places around the world, appear to be getting worse as it becomes more challenging to contact a representative sample of voters.”
The “science” of polling has been through rough patches before, and experts have conducted reviews and done postmortems. But whatever the reason — the increased use of cell phones, the inability of pollsters to reach people, and a tendency to avoid giving politically incorrect answers to strangers on the phone — the problem isn’t close to being fixed.
Leighton Vaughan Williams, director of the Political Forecasting Unit at Nottingham Business School, sums it up: “If you really want to know a likely election result, ignore the polls and look at the betting markets. It’s a mystery why pollsters are taken more seriously.” From now on, if pollsters are to regain any credibility, they must be more transparent, spend the money to conduct real surveys, resist the temptation to withhold polls they don’t like, and realize that more and more of the public is starting to pick up a whiff of the same bias in pollsters that they detect in the media as a whole.
— John Fund is national-affairs correspondent for National Review Online.