According to a new poll, Anthony Weiner is trailing City Council Speaker Christine Quinn by nine points in the New York Democratic mayoral primary. But on Monday, he was up three. What gives?
We spoke by phone with Andrew Gelman to find out. Gelman is a professor of statistics and political science at Columbia University who has contributed to The Times' FiveThirtyEight blog in the past.
First, the details on the polls.
|Monday's poll||Today's poll|
|Results||Weiner, up by three (25 to 22%)||Quinn, up by nine (27 to 18%)|
|Margin of error||3.6 percent||4 percent|
|Polling firm||Quinnipiac University||Siena Research Institute|
|Sample size||738 registered Democrats||1,010 registered Democrats|
|Polling dates||Monday, July 8 — Sunday, July 14||Tuesday, July 9 — Monday, July 15|
While he wasn't familiar with these particular polls, Gelman explained a few possibilities. The first: "Public opinions can change." Even if the poll was conducted over largely the same time period, "a primary election campaign for mayor can be very volatile," he explained. "You have many candidates and don't want to waste your vote," meaning that people may be more likely to be willing to change their minds, especially since New York City Democratic primaries don't exactly include a broad range of political philosophies.
Gelman noted that even that switch of one day can have a big influence. Different people are home during the day and in the evenings on any given day of the week and during different time periods. So, while the dream goal of conducting a survey would be to have a unified list from which each pollster draws a random sampling of names, the reality includes a lot more fluctuation. And that fluctuation leads to different results.
Don't blame the margin of error, by the way. The margin of error simply means that "if you do two surveys at the exact same time, from the exact same survey organization, you'd expect to see that level of variation." If Siena College had done the same poll with the same sample size over the same days, in other words, it may have ended up as much as 4 percent from what it reported today.
(Incidentally, neither of the sample sizes in these two polls is suggestive of the discrepancy. After a certain point, having more respondents doesn't yield a significantly more accurate result, as we've noted before.)
Then what's a voter / reader / candidate for mayor to do with these conflicting results? "The natural thing to do would be to average them," Gelman says. In a sense, that's what FiveThirtyEight did over the course of the campaign last year, tracking meta-trends in political polling while also doing its own original research. If the wisdom of the crowd is how we get our poll results, the wisdom of a crowd of polls can do us one better.
So we have our answer: Christine Quinn is leading in the race for mayor by three points, according to the polling average. At least until next week.