What is Quinnipiac doing to Eliot Spitzer's lead?
Last week, Quinnipiac University released its third poll of the comptroller's race among likely Democratic voters, which showed Eliot Spitzer and Scott Stringer tied at 46 percent.
It was a noteworthy result, since it seemed to have reversed the gains Spitzer made in a Quinnipiac poll just two weeks earlier, which itself represented a change from a first Quinnipiac poll on July 25, in which Spitzer led Stringer, but within the margin of error.
The result, a tie, was also sharply at odds with a Siena poll that came out shortly afterward, showing Spitzer with a solid lead.
So what's up with Quinnipiac?
Spitzer had a commanding lead in that first poll among likely Democratic voters who are black, 63-33 percent, while Stringer led among likely Democratic voters who are white, by 56-38 percent. In attempting to "weight" the respondent-groups, Quinnipiac projected black voters to be only 29 percent of the primary electorate, and white voters to make up 44 percent. You can see all that in the crosstabs of that first poll.
Then Quinnipiac changed its assumptions.
In the second Quinnipiac poll, on August 14, when Spitzer went out to a big lead, he increased his margin among black voters, but that advantage was magnified by the poll's black-turnout projection, which went up in the second poll, it was 35 percent.
Two weeks later, Spitzer and Stringer were in a tie, according to Quinnipiac's third poll, with 46 percent each.
Again, Quinnipiac had adjusted its weights. In that poll, on August 29, Spitzer's lead among likely Democratic voters who are black was greatly reduced, down to 52-40, within the 8.31 margin of error for that category. Also, the turnout projection for black voters went from 35 percent to 27.
There was other readjusting, too: Likely Democratic voters who are women—another category where Spitzer once led—were also downgraded as a projected share of the electorate, going from 59 percent of the primary in the second poll to 55 percent in the third, after starting off in the first Quinnipiac poll at 61 percent.
Quinnipiac's adjustments were bigger than in other polls released around that time from Marist College (July 25 and August 16) and Siena College (August 30). All three of those polls had the exact same weight given to likely Democratic voters who are women: 57 percent (Marist, Marist and Siena). Both Marist polls projected likely Democratic voters who are black would make up 28 percent of the primary; Marist's poll was one percent different, at 27 percent.
The weights in the Marist and Siena polls were more steady, and partially as a result, their measurements of the candidates' overall support were, too. Marist polls had Stringer at 36 percent, while Siena had him at 35 percent.
Spitzer's support among likely Democratic voters shifted among the Marist and Siena polls, but those changes were smaller than Quinnipiac's.
In the first two Marist polls (July 25 to August 16), Spitzer went from 48 to 54 percent. At nearly the exact same time, the first two Quinnipiac polls (July 25 to August 14) showed Spitzer gaining seven points, from 49 to 56 percent.
Then, he dropped. Spitzer's support among likely Democratic voters dipped, from 54 percent to 50 percent, from Marist's August 16 poll to Siena's August 30 poll. That's a four-percent dip. The change Quinnipiac's polls detected, from August 14 to August 29, was more significant: 10 percent.