Shooting the pollsters

It was bound to happen sooner or later. As Americans and their leaders prepare for the most important and divisive presidential election in decades, pollsters, who like to think of themselves as scientists outside the arena of political contention, have become targets of partisan venom from both the Kerry and Bush camps.

It isn’t surprising. Why should pollsters be above the fray in this titanic struggle between red and blue America, when nothing else—Senator Kerry’s war wounds, the still-wrenching images of the collapsing twin towers, the president’s family name, lewdly derided by Whoopi Goldberg—has been off limits?
Gallup was accused of Republican bias as its results in September tended to show President Bush with an especially big lead. An activist group even took out a full-page ad in the New York Times reminding readers that George Gallup Jr., the son of the company’s founder, is an evangelical Christian whose faith might be tainting his figures. But Gallup has vehemently denied these charges, and it isn’t the only group being accused of publishing skewed results.

Many major polling organizations have complained of receiving angry letters and phone calls in response to their published polls. Suspicion of pollsters and their methods has reached such a height that some news organizations have begun backing away from public pulse-taking. MSNBC, which had planned to air focus groups led by pollster Frank Luntz following televised presidential debates, axed the idea. The Washington Post reported that a network spokesperson had explained, “We think our viewers should be able to make up their own minds without ‘scientific’ help.”

Canadian pollsters were also on trial not so long ago. North of the border, though, skepticism of pollsters in the media and among the public was inspired by fears not of partisanship, but of mere ineffectiveness. The perceived problem in the recent Canadian federal election was not that various polling organizations differed from one another suspiciously, but rather that the pollsters were unanimous—and wrong. Such was the claim of many pundits, anyway.
As voting day approached the polls appeared in rapid succession, each more or less affirming that the Conservatives were closing the gap on the Liberals. A minority government was almost certain, but what kind of minority government was anyone’s guess.

The election did indeed yield a minority government, but the results were a far cry from what pollsters had predicted on the last Friday before the vote. The Liberals won a minority government by a comfortable margin: 135 seats to 99 for the Conservatives. As expected, the Bloc did well in Quebec, winning 54 seats. The New Democrats won 19 seats, more than the last time but far fewer than polling during the campaign suggested. The Greens did not win a seat but they did capture four per cent of the popular vote, enough to allow them to cash in on federal campaign funding and—who knows?—maybe an invite to the televised leaders’ shouting match in the next campaign.

Did the pollsters blow it as so many journalists, in giddy fits of schadenfreude, declared? Were pollsters, as the Globe’s Jeffrey Simpson in his post mortem column stated, the “night’s biggest losers”?

In fact, there is no evidence that the pollsters were wrong when they declared a dead heat in the week before election day. All were within their advertised margin of error of each other. True, the seat projections they and others offered based on their data took liberties that science might have suggested imprudent, but there is no evidence that they gauged the popularity of the parties wrongly at the time they came out of the field with their last published polls on Thursday, June 24. Rather, there is considerable evidence that there was a shift in voter preferences in the last day or two of the campaign, away from indecision and other parties toward the Liberals.

Ipsos-Reid claimed that its polling over the final weekend of the campaign showed the Liberals picking up votes on the Sunday evening. Compass pollster Conrad Winn claimed that fully one quarter of the voters said they made up their minds on election day or the day before. Voters making their decision in the final days of the campaign opted for the Liberals by a margin of nearly two to one over the Conservatives.

What the published polling data made clear to voters was that a Conservative minority government, dependent on the Bloc Quebecois for the balance of power, was more than plausible. A significant number of voters, after reflecting on this information, changed their votes to avoid that outcome. Those who remained undecided throughout the campaign, particularly Ontarians, said “Better the party I dislike than the party I fear; I’m holding my nose and voting Liberal.” A significant number of New Democrats switched from their first choice to their second, the Liberals. Even some who had intended to vote Conservative decided in the end they did not think their party was ready to govern, particularly if their fledgling government would have to do business with Gilles Duceppe’s separatists. That the Conservative vote this year was eight points less than the combined Alliance/Progressive Conservative vote in 2000 says that Joe Clark was not the only former Tory to abandon the new party.

In the infamous 1948 U.S. election, pollsters found that Republican governor Thomas Dewey was way ahead of Democratic president Harry Truman and concluded it was not necessary to do any more polls since it was clear that Dewey would win. Back then, pollsters and their media clients learned that a lot could happen in the last ten days of an election campaign. In the 1980, American voters didn’t decide until the final weekend before the election that Republican candidate Ronald Reagan’s values were, after all, within the American mainstream.

In our most recent campaign, enough Canadian voters decided the opposite about the new Conservative Party—that its values were not in the Canadian mainstream—to deny Stephen Harper the minority government so many had predicted. Pollsters and voters alike learned once again that much can happen in the final days and hours of a campaign, especially when the electorate is conflicted. Far from a failure for the polling industry, this year’s federal election was evidence of the crucial role polls can play in helping voters exercise their franchise in full awareness of the context and the stakes of their decision.