It's hard to believe that there were some complaints this weekend about how few polls there were. Once the work week started, the steady downpour of polls began in earnest, with just two weeks left until the election; there were 33 polls on Wednesday alone, for instance.
And yet, none of those dozens of polls really reshaped the overall model, or even game-changed any one individual race. There was only one race, out of all the Senate and gubernatorial races that we cover, where the odds shifted in one candidate's favor by 10 percent or more: it was the North Carolina Senate race, where Kay Hagan's odds of winning fell from 73 percent to 60 percent. (And that was largely because of a poll from the awful Gravis Marketing, who claimed that Hagan was trailing Thom Tillis 48-43. That's the only poll in over a month to give Tillis a lead; the other two polls in North Carolina that came out this week, from PPP and SurveyUSA, both gave Hagan a three-point lead. That Gravis poll aside, this is really the most consistent Senate race in the country; almost invariably, for several months, pollsters have found this race to have a 2-to-4 point lead for Hagan. But the Gravis poll was a big enough deviation to put a small dent in the trendlines.)
That stability — as I've said in just about every post about the Outlook model this month — is bad for the Democrats in the Senate, but good for them in the gubernatorial races. In the Senate, they really need to make a move in the Colorado and Iowa races; in August and September when the Democratic candidates seemed to have small leads in these races, that was what was keeping Democratic odds of holding the Senate around the 50-50 mark. Once Bruce Braley and Mark Udall started losing ground in mid-September, that's when the model's predicted Democratic seat count got really locked in at 48.
And in the gubernatorial races, the last month has seen a couple key races, ones that were looking like they were going to get away from us for much of 2014, start to crystalize with small Democratic advantages. In particular, I'm talking about the Florida and Illinois races, where the Republican candidates' financial advantages let them dominate the conversation over the summer, but where the Democratic candidates have started to eke out small but consistent leads in the last month ... partly thanks to having saved their money for financial parity in the closing months, but also thanks to a drumbeat of unflattering news stories about Bruce Rauner, and debate debacles for Rick Scott.
One encouraging thing about the increasing stability, though, is that it means that the model, and the very concept behind poll aggregation, is doing exactly what it's supposed to. When you have only a few polls of a race, and a new poll comes along that's dramatically different, it totally changes (or at least appears to change, regardless of whether it reflects reality or not) the direction of the race. But when you have several polls a day of a race, and most of them show the same thing, the races tend to settle down into a rut (and even a glaring outlier, like the Gravis poll of North Carolina, gets drowned out by all the other polls around it).
Think of poll aggregation in terms of the fable about the blind men describing the elephant. Each pollster, by way of the hundreds of people they sample, is each describing only a very small portion of the elephant that is the electorate. When you have only a couple pollsters describing a race, you can't be sure whether they're fumbling around with a tail or a trunk or a tusk. But when you have dozens of pollsters at work talking to thousands of people, and somebody aggregating all their responses, then you can, piece by piece, start to rough out the shape of the actual elephant.
We'll talk a little more about the week's polls over the fold:
On the Senate front, the big drop, as mentioned before, was in North Carolina, where Kay Hagan's chances fell from 73 percent on Monday to 60 percent today; even when downweighting the Gravis poll for being a partisan pollster, it still registers as a Tillis lead. Mark Udall also lost more ground in Colorado, mostly because of Wednesday's poll from Suffolk that gave Cory Gardner a 7-point lead; Udall's odds fell from 30 percent to 23 percent. That means that challenger Michelle Nunn now has better odds in the Georgia race than Udall does in Colorado. Together, those were enough to push Democratic odds of controlling the Senate down from 34 percent to 30 percent today.
On the gubernatorial side, most races were pretty stationary as well, but two Democrats saw their chances improve significantly. One was in Kansas, where Paul Davis is back above water, thanks to a 5-point lead over Sam Brownback in Monday's poll from Monmouth; that pushed his odds from 46 percent back to 55 percent, though that's still a far cry from the near-lock he seemed to have on the race in August. Davis's decline seemed to move in parallel to Greg Orman's decline in the Senate race, and it seems likely that some of the boost that Pat Roberts has been getting from national Republican groups to save the Senate race has rubbed off on Brownback as well.
The other is Connecticut, where a Quinnipiac poll on Tuesday gave Democratic incumbent Dan Malloy a one-point lead; that moved Malloy's odds up from 28 percent on Monday to 39 percent now. Malloy had been above water for a few weeks after a good run of polls (including an eight-point lead in a PPP poll), but got pushed back down by an outlier-ish Rasmussen poll over the weekend that found Malloy trailing Tom Foley by 7.
That contrasts a bit with Wisconsin, where some similar polling news — a one-point lead for Mary Burke in the most recent Rasmussen Reports poll on Wednesday — didn't help Burke to move back above water. Burke's odds actually fell from 46 percent to 43 percent, as the two other polls this week (from PPP and St. Norbert) both had Burke trailing Scott Walker by one point. If you're wondering what race in the country comes the closest to being a true tossup, if it isn't the Florida governor's race, it's probably this one: every single Wisconsin poll in the last couple weeks has been a one-point lead for one or the other candidate or an outright tie.