Yesterday, I wrote the following:
Nine out of 10 ‘SHOCKING’ poll results aren’t shocking if you’re paying attention.
What I meant is that a lot of the polling results that people find surprising can be attributed to random variation, or to a race being hard to poll for some reason and in which there is inherently a lot of uncertainty, or a race that hasn’t been polled very much and in which there is little basis for comparison, or some combination of those factors.
But every now and then, a poll comes along that really forces us to reconsider what we thought we knew about a particular race.
There is one such poll out this morning from Quinnipiac University, which finds the Republican nominee of governor in New York, Carl Paladino, only 6 points behind the Democrat, Andrew Cuomo. Mr. Cuomo has 49 percent of the vote in the poll, and Mr. Paladino 43 percent.
Polls taken before the primary on Sept. 14 showed Mr. Cuomo leading Mr. Paladino by much wider margins, ranging from 23 points to 45 points. This includes three prior polls by Quinnipiac University, which had shown leads of 36, 30 and 37 points for Mr. Cuomo.
It’s just not very often in a general election campaign that you see a candidate’s lead shrink from about 30 points to 6 in just a month or so. It would not be so unusual in a primary campaign, when voters often make their choices late. But you could spend all day looking at general election polling results and still have trouble coming up with more than a handful of similar examples in recent elections.
Several qualifiers are in order. First, Quinnipiac did not include the Conservative Party’s nominee, Rick Lazio, in their poll. (Mr. Lazio was beaten by Mr. Paladino in the Republican primary.) I criticized Rasmussen Reports for making the same decision to omit Mr. Lazio.
I asked Doug Schwartz, director of the Quinnipiac University Poll, why Mr. Lazio was not included.
“Lazio has not announced his intentions,” Mr. Schwartz wrote in an e-mail. “So we didn’t include him. If he decides to stay in the race, we will include him in the next survey.”
Quinnipiac’s polling center is run by really smart people who are trying to get this sort of thing right, and usually do. But I don’t agree with their decision any more than I did Rasmussen’s. Mr. Lazio’s name will appear on the ballot in November unless he drops out — which he does not sound inclined to do so, at least so far — so the default decision should probably have been to include him in the polling. This is particularly so since a previous poll of the race as a three-way contest showed Mr. Lazio running ahead of Mr. Paladino (although well behind Mr. Cuomo). At the very least, Quinnipiac could have tested results both with and without Mr. Lazio.
Of course, the fact that the race now appears winnable for Mr. Paladino may change its dynamics, and increase the pressure on Mr. Lazio to truncate his bid.
The other caveat is that Quinnipiac switched from a registered voter model to a likely voter model. It is hard to know exactly how much difference this makes — but on average, over the course of this election cycle, likely voter polls have favored Republican candidates by about 4 points compared with registered voter polls (perhaps reflecting greater enthusiasm among Republican voters in most regions of the country).
Sometimes, the difference has been smaller, while other times it has been as large as 10 points or so. Pollsters use different types of likely voter models, and their impact on the results is correspondingly different.
Still, it is not probable that the switch to a likely voter model accounts for all, or even most, of the gain for Mr. Paladino that Quinnipiac has shown here. Instead, the voter model is probably one of several factors that has contributed to the narrowed margin.
I asked Mr. Schwartz about the partisan composition of the electorate in his likely voter sample; he said it was 22 percent Republicans, 36 percent Democrats, 31 percent independents and 11 percent voters who identified with some other party or who didn’t disclose their partisan affiliation.
For comparison, according to exit polling in the 2004 election, voters in New York State were composed of 46 percent Democrats, 28 percent Republicans and 25 percent independents.
That Quinnipiac is showing a much larger fraction of independents and “other” voters is not necessarily revealing, since there are a lot of different ways of asking about partisan identification. Meanwhile, Quinnipiac shows a 14-point gap between Democratic and Republican voters, whereas there was an 18-point gap in 2004 — not all that extreme a difference. So I don’t know that Quinnipiac’s likely voter model is necessarily to “blame” here.
There are also some hints in the poll that Mr. Cuomo’s fundamentals are sound, so to speak. Some 51 percent of voters have a favorable impression of him, as compared to 34 percent unfavorable. And 67 percent approve of his performance as New York’s attorney general. That perhaps suggests that there are plenty of votes available to Mr. Cuomo, provided he doesn’t take them for granted — which arguably he is doing by having ignored Mr. Paladino’s invitation for a debate, for instance.
The close result in the poll probably reflects some combination of the failure to have included Mr. Lazio, and some support for Mr. Paladino that is fairly “soft.” More likely than not, it also reflects some statistical noise (as Rasmussen Reports, for instance, showed a 16-point margin for Mr. Cuomo).
But, let’s not bury the lede too much: This race has gotten a lot closer. Probably not as tight as 6 points — but much closer, nevertheless. As I noted previously, there are several reasons to think that this race was, in fact, liable to tighten. But I doubt that many observers expected it to tighten this much this quickly. I certainly did not.
UPDATE: Mr. Schwartz has written me with the partisan makeup of Quinnipiac’s previous poll of New York State, which was conducted among registered voters. It was 21 percent Republican, 38 percent Democrat, 33 percent independent, and 7 percent other or don’t know. That’s only a couple of points different, across the board, from the partisan identification in Quinnipiac’s poll of likely voters this morning.
So, at least based on these numbers, Quinnipiac does not look to have applied an especially unusual likely voter screen, and the movement toward Mr. Paladino is mostly explained by other factors.