Distribution of likely number of Democratic Senate seats
If you'd stepped away for two weeks and only just now were coming back to look at the Daily Kos Elections Poll Explorer, you'd assume nothing has changed at all. However, if you checked in with us last Thursday, you'd get a bit of a whiplash effect. A week and a half ago, the Democrats had a 46 percent chance of retaining control of the Senate, and they
still have a 46 percent chance of retaining control of the Senate. And yet,
last Thursday they had a
56 percent chance of retaining control of the Senate. What happened?
Well, what happened last week was the dropout of Democrat Chad Taylor from the Kansas Senate race. That left us with only a race between Republican Pat Roberts and well-funded independent Greg Orman, and the only poll of a head-to-head matchup between the two of them gave Orman a 10-point lead. However, it's not a sure thing that Orman would caucus with the Democrats. He plans to caucus with whoever is in the majority. He didn't give any indication what he'd do if there was a 49 D/50 R situation; our solution to that was to assume a 50-50 chance of him caucusing with either party in such a situation, which is why so few scenarios end with either 49 or 50 Democrats anymore, as you can see in the histogram above.
What happened over the weekend, though, was the release of a huge slew of polls from YouGov on behalf of the New York Times, covering every state with a Senate race (even the snoozers like Hawaii and Wyoming). YouGov is a well-respected pollster (though they're based in the UK, where they have some tangential links to the Conservative Party there), turning in a middle-of-the-pack performance in 2012.
However, this year, they've often put up results more Republican-friendly than traditionally suspect pollsters like Rasmussen. For instance, YouGov is the only pollster since April to give a lead to Terri Lynn Land in Michigan, and now they've done it twice.
Similarly, their new Alaska poll, where Republican Dan Sullivan leads by six among likely voters, is considerably more bearish on Mark Begich than any other non-Republican internal pollster has ever seen (which is even odder when you consider that their previous Alaska poll was an outlier in the opposite direction, giving Begich an 11-point lead). The NYT's
explanation is that the Alaska Republican primary finally rattled to a conclusion, giving Sullivan a potential post-primary bounce as supporters of his rivals grudgingly get on board.
The net result of the onslaught of YouGov polls was that Democratic odds in two competitive races, where the incumbents have been keeping their heads above water pretty well, now fell below 50 percent: in Alaska, Begich's chances fell from 56 to 46 percent, while Kay Hagan's chances in North Carolina fell from 58 to 47 percent. Those aren't big moves, in the grand scheme of things (they remain coin-flips, in other words), but 50 is an important psychological barrier. Even bigger declines were in Arkansas, where Marist, in addition to YouGov, also released a poll giving Republican Tom Cotton a lead over Dem incumbent Mark Pryor, and in Louisiana, where Mary Landrieu was temporarily benefiting from an oddball lead in a Republican internal poll but trailed by 2 in the new YouGov poll. Pryor and Landrieu's odds decreased from 47 and 76 percent to 28 and 51 percent, respectively.
That was offset somewhat by better-than-average results from YouGov in New Hampshire, Colorado, and Iowa, which pushed Democratic odds up there, though. Odds in New Hampshire rose from 78 to 87 percent, from 62 to 75 percent in Colorado, and from 46 to 57 percent in Iowa.
You may be wondering "Well, what about Kansas?!? That's where we really need the data," considering that we only have only poll (from PPP) testing only Roberts vs. Orman. Unfortunately, that's the one YouGov poll, out of dozens of states, that we didn't get. They were in the field when Taylor dropped out (they have unusually long field dates, from Aug. 18 to Sep. 2), so they had to discard their Kansas data. You might notice, though, that Democratic odds of a "win" in Kansas still fell, from 54 percent to 46 percent. That's because, again, Orman plans to caucus with the majority party. Notice how those percentages closely mirror the overall Democratic odds; that's because a "win" in Kansas is largely dependent on a Democratic win overall.
This sudden decline in Democratic Senate fortunes isn't limited to our model; it had a similar effect on the other polls-only models; the Huffington Post Pollster model went from a 40 percent chance of Republican takeover to a 54 percent chance (exactly the same as us), and the Princeton Election Consortium, while still more bullish on Democratic odds than other models, still went from a 25 percent chance to a 30 percent chance. Interestingly, the models that are more fundamentals-reliant barely moved, suggesting that the Democratic decline moved their odds more in line with what the broader fundamentals were suggesting (I'd observed several weeks ago that Senate Democrats were seeming, in the polls, to overperform those same fundamentals). The NYT model moved from 62 to 61 percent chance of a GOP takeover, while 538's model moved from 64 percent to 65 percent. (As of this writing, the Washington Post hadn't updated its model.)
The gubernatorial news, on the other hand, has been getting better for the Democrats in the past few weeks, and just continues to get better in today's installment. We'll discuss that over the fold:
Overall odds of the Democrats gaining gubernatorial seats is up to 66 percent, from 60 percent last week. The median number of
gubernatorial seats that the Democrats are likely to control after the election has ticked back up to 23, which is a two-seat improvement over where they currently are.
The modal number is still 22; to understand why the median is higher, take a look at the histogram. The bell curve has a much bigger tail on the Democratic side, meaning they could plausibly wind up with a result as high as 26 or 27 seats without the model or the underlying polls being "wrong." Even though 22 is the number that comes up in the most Monte Carlo simulations, half of all results are at or below 23, while half of the results are at or above 23. To oversimplify it with a sports metaphor, the Democrats have all the upside in the gubernatorial races, unlike the Senate races.
The big mover in the gubernatorial model, stunningly, is Wisconsin. This race, pitting Republican incumbent Scott Walker against Democrat Mary Burke, seemed like a long-shot even several months ago, but lately it's moved into a position where Walker is slightly, but unambiguously, an underdog. Since last week, Democratic odds in this race shot up from 56 percent to 81 percent. That's thanks entirely to a poll from We Ask America, a little-known pollster who work mostly in Illinois but who found a 48-44 lead for Burke late last week. We Ask America is affiliated with the conservative Illinois Manufacturers' Association; in our database, we regard them as a Republican internal pollster, meaning that Walker is further penalized, on top of already being down 4 points.
Marist also released several gubernatorial polls over the weekend; that was good news in Colorado, where Democratic odds moved up from 53 to 73 percent, but bad news in the open seat in Arkansas, where Democratic odds fell from 40 percent to 17 percent.