Does this new machinery of dirty tricks have a major effect on election results? Don’t get too riled up. A local professional pollster, Lee Miringoff of the Marist Institute for Public Opinion, doubts it has a significant effect. “Polls are not self-fulfilling,” he said.
SUNY New Paltz professor Jerry Benjamin, an expert in New York government and a former Ulster County Legislature chairman, says that candidates are always releasing poll results, and that the creation of momentum, true or false, has long been a staple of politics. “Reporters using candidate polls identify them as such, or should,” said Benjamin. “Nobody pays attention.”
It’s certainly easy to find examples of manufactured momentum. Take the New York Post front page the day before this year’s election. “Dead heat,” screamed the huge headline type. “Poll shock: GOP surge ties two top NY races.” Albany might not be immune to the throw-the-bums-out fever sweeping the country after all, the Post declared. The latest Siena College poll had showed Republican Dan Donovan tied with Democrat Eric Schneiderman at 44 percent each in the attorney general’s race. The same percentage was also found for both candidates in the contest for state comptroller between Republican Harry Wilson and Democratic incumbent Tom DiNapoli.
The Post found a political science professor from Iona College willing to accept the presumptive validity of the Siena numbers. “The sudden gains of Donovan and Wilson is part of a larger theme that’s being played out nationally,” Jeanne Zaino was quoted as saying. “When you’re tied like this, the independent vote is going to decide. In these cases, you wouldn’t want to be an incumbent like DiNapoli or an insider like Schneiderman.”
Schneiderman had been ahead by seven points a few weeks previously, so his opponent, Donovan, naturally saw the importance of momentum as a political game-changer. “The polls show we are now tied,” the Post quoted him as saying. “People know that he is part of the problem in Albany and he’s tied in with all the people that are now getting investigated.”
The unofficial vote the morning after the election showed a different result. Schneiderman had 55 percent of the total statewide vote. Donovan had 44 percent — a landslide. No polling margin of error could explain the result. In the last 24 hours, it seems a miracle had taken place.
Either that, or the Siena poll should not have been given the credibility that it had been given. The Schneiderman-Donovan election proved the Siena poll an “outlier,” the statistical term in the trade referring to an observation that appears to deviate markedly from other members of the sample in which it occurs. The difficulty in the polling trade, of course, is that it takes other observations to make an outlier an outlier. The “sudden gains” reported could have been accurate, though it turned out not to be.
To be fair, the other New York statewide race being discussed was much closer, as the Siena poll had correctly predicted. Harry Wilson, the Republican candidate for comptroller, was a political novice with considerable entrepreneurial experience on Wall Street. He gave Tom DiNapoli a run for his money in competing for the job that manages the state’s money. In the unofficial statewide vote count, DiNapoli ended up with 50 percent of the statewide vote to Wilson’s 47 percent. That contest was at least within the margin of error.
Neither major party has a monopoly on cooking the numbers.
Some pollsters affiliated with each have better reputations than others. The Republicans have their favorite pollsters, the Democrats have theirs. Charlie Cook, himself of the conservative inclination, is widely thought of as the country’s most accurate and knowledgeable election prognosticator. Cook’s website charges for his services.
Legitimate polls, said Benjamin, report their methods, timing and questions. Does the general public know that statistically one out of 20 polls is a stinker, he asked? A prominent example of such a stinker was the poll that predicted Carl Paladino only 6 percent behind Andrew Cuomo. It proved an outlier.
Rasmussen “quite biased”
Two days after last week’s election, Nate Silver, who writes a blog called FiveThirtyEight for The New York Times, asked how accurate eight major pollsters had been, and whether they were consistently biased (the eight, in order of accuracy, were Quinnipiac, SurveyUSA, YouGov, Public Policy Polling, Mason-Dixon, Marist, CNN/Opinion Research and Rasmussen Reports. The best, Quinnipiac, missed the final result, the margin between the candidates, by 3.3 percent and showed little overall bias. The worst performer, Rasmussen, was in a class by itself, with a 5.8 percent error rate. Rasmussen was consistently biased toward the Republicans by an average 3.9 percent. Silver calls Rasmussen’s 2010 polls “quite biased.”
Rasmussen conducted 105 of the 287 polls, suggesting that its bias was unlikely to have been a statistical anomaly. According to Silver, Rasmussen and its subsidiary Pulse Opinion Research have been under heavy criticism this election cycle for methodological shortcuts and unusual statistical assumptions. Silver noted that its president, Scott Rasmussen, was himself a conservative, “the same direction in which his polls have generally leaned.” Many Rasmussen and Pulse Opinion Research polls are commissioned by Fox News.
Rasmussen, Silver said, does not call cell phones and does not call back respondents it initially misses. These cost-saving measures are alone enough to bias a sample. Cell-phone users tend to be younger and slightly more urban than land-line users; the better educated and minorities also are heavier cell users.
Bias is likely to be interpreted as suggesting a deliberate skewing of results rather than poor procedure. According to Silver, it isn’t necessarily that. It’s simply a statistical term conveying the accuracy — or lack of it — of predictions.
In a post-election blog, Miringoff, longtime director of the Marist surveying operation, noted that the McClatchy-Marist national poll numbers, like many other polls, thought the gap in voting for Congress would hover around 6 percent in favor of the GOP among those most likely to vote. Team Obama’s efforts to rally the Democratic base to close the so-called “enthusiasm gap” clearly came up short, said Miringoff. Although the McClatchy-Marist national poll found a narrowing gap in the closing days of the campaign from a plus 23 percent in favor of the GOP over the Democrats to a plus 14 percent, this, he said, was clearly a case of too little, too late.
Silver said the Marist polls ended up this year with an error rate worse than five of the eight pollsters he surveyed.
Miringoff said he does not do surveying for political candidates. But he does collaborate with media companies. The 10 races he did this year, all of which he noted correctly predicted the result, were on Marist’s tab. In 1978, said Miringoff, Marist became the first college to use undergraduate students for its survey research (“Their experiences,” its website explains, “allow them to weave political science, computing, communications, marketing and psychology into an interdisciplinary learning experience.”)
“You don’t vote differently if you see a poll,” argued Miringoff. “Most voters aren’t moved by reports of who’s winning.” Fewer young people voted this year because they had lost some interest in politics since 2008, he said, not because people were telling them what to do.
The turnout recipe for this midterm election had a decidedly different flavor than 2006 or 2008, concluded Miringoff. “This year’s electorate was older and more conservative. Not a good sign for the Democrats,” he wrote. “This data all identifies the strong headwinds that the Democrats had to navigate. With the energy of the Tea Party, the sleepiness of the president’s electoral base, and the shifting tides of independent voters, the White House was truly in for a shellacking.”
Campaign experts use polls to make strategic decisions on resource allocation and policy emphasis, said Miringoff — though they might tell journalists another story. Sometimes campaigns spend money just to make their opponents spend it, too.