Since Tuesday, headlines all over the city have trumpeted the obvious fact that Freddy Ferrer would need no small miracle to win this election. Their source was none other than Quinnipiac’s most recent election poll, which found Bloomberg leading Ferrer over 2 to 1-a stunning statistic in a city where Democrats outnumber Republicans 5 to 1.
In its story “BLOOMY UP 61 – 30: POLL,” (October 25, 2005) the New York Post does not mention that these numbers refer to likely and not registered voters. When reporting on polls, the press often seems to take for granted that their audience understands the caveats and limitations inherent in the numbers. “I think that the public polls have served to guide campaign coverage in a way that is very typical,” says Jeffrey Pollock, Ferrer’s pollster. “We see the press corps following the numbers and mentioning them ad nauseam.”
Quinnipiac itself left out several interesting numbers from its press release. The 31 percent gap is among likely voters and “leaners,” or, anyone currently undecided but leaning towards a particular candidate. Among registered voters, however, this margin is only 22 percent, Bloomberg leading Ferrer 51 to 29. This 9 percent variance between registered voters and likely voters is significantly higher than what polls showed in 2001, when on October 24, Green led Bloomberg 51 to 35 percent among likely voters, and 49 to 35 percent among registered voters—only a 2 per cent difference.
Quinnipiac’s pollster, Doug Schwartz, claims that this disparity can simply be attributed to the character of a given race. “It varies from election to election,” he says. “In a presidential election you’ll have greater turnout than you will in an off year, election, for example.”
However, the New York City electorate has been relatively consistent over the past several mayoral and presidential elections. The turnout between 1997 and 2001 for the mayoral elections varied only by 100,000 voters, and the presidential elections of 2000 and 2004 within 200,000 voters—small potatoes in a city of roughly 8 million.
So maybe this gap can be chalked up to something other than a capricious electorate. Both Quinnipiac’s and Marist Institute’s polls employ random digit dialing, wherein numbers are called randomly and respondents are subsequently asked a series of questions-a screen-to determine whether or not they are likely to vote. While the details of these screens are not generally shared with the public, they are used to determine a potential voter’s interest in and knowledge of a particular race.
In a May 2001 study on likely voter screens, the Pew Research Center evaluated its own poll of the 2001 Philadelphia mayoral election. It concluded that, among all voters that could be confirmed, 17 percent were evaluated as unlikely voters and subsequently cast a vote, and 10 percent said they were going to vote and didn’t. Ideally, these two numbers would be equal. The significant discrepancy suggests that the screen held voters to pretty high standards, asking them about their previous voting history and the amount of “thought” they’d given to the election.
The Pew Center has since removed a particular question from its likely voter scenario: “Do you know where you go to vote?” Pew’s Michael Dimock, who conducted the 2001 study and strongly advocates the screens as a method, claims that this question is “no longer as valid a measure in an era when people are voting by mail.”
Quinnipiac, whose near perfect poll in 2001 did not include this question, has since added it. In New York, however, just under 2 percent of all votes cast in the 2001 mayoral election were via absentee ballot.
Surely, evaluating these polls is a delicate balance, and when Bloomberg is 22 percentiles ahead of Ferrer, one asks the point of tossing around all these numbers. However, sometimes, small changes in poll numbers can make a big difference in analyzing the race. In the September 12 poll, Bloomberg was ahead of Ferrer 52 to 38 percent among likely voters with leaners. Among registered voters, his lead was much thinner: 46 to 36. As Pollock points out, the 50 percent threshold is an important one for incumbents, who rarely gain undecided votes over the course of an election. “I was surprised that more emphasis wasn’t put on both numbers,” says Pollock. “The 46 percent was buried in the coverage, barely mentioned.”