Distribution of likely number of Democratic Senate seats
Since we started doing the Poll Explorer model, there's been only one really game-changing event: the decision by underwhelming Democratic candidate Chad Taylor to drop out and give well-funded independent Greg Orman a clear shot at Republican incumbent Pat Roberts in the Kansas Senate race. Unfortunately, we're still trying to figure out how that changed the game; he's been deliberately cagey about saying which party he'd caucus with. One thing seemed pretty clear last week, though: Orman looked very likely to win. There was only poll of a two-way Orman vs. Roberts election, from Public Policy Polling, and it gave Orman a 10-point lead. That's outside the margin of error, so that necessarily gave Orman a greater than 95 percent chance of winning.
Now, however, we have some additional polling that makes it clear that Orman isn't as likely to win. Ordinarily, you'd think a second poll with Orman in the lead would only solidify his standing ... but the poll, from SurveyUSA (a trustworthy and usually accurate pollster), gave Orman only a 1-point lead, 37-36. The SurveyUSA poll also had an important difference from that previous PPP poll. It included Taylor's name as well, necessitated because Kansas Secretary of State Kris Kobach ordered Taylor's name to stay on the ballot.
SurveyUSA informed its respondents that Taylor's name was on the ballot but that he wasn't running any more. Despite that piece of information, 10 percent of the respondents still said they were voting for Taylor. (The poll didn't delve further into whether those 10 percent were that hardcore in their devotion to the Democrats, or were just particularly obtuse.)
Unfortunately, Kobach's meddling, and the subsequent structure of the Survey USA poll, caused us to have to redo the Kansas portion of our model one more time. We could no longer use that PPP poll, which didn't include Taylor at all, but we also couldn't include the pre-dropout polls which included Taylor's name, since they were taken under a totally different set of assumptions (i.e. that Taylor was actually campaigning).
So, we're back to having only one Kansas poll in our Senate model (the SurveyUSA poll), and now, instead of a mortal lock for Orman, it's a total coinflip. (That may be more realistic, though; Orman's appeal to disaffected Republicans may be diminished now that he's become something of a de facto Democrat.)
We'll discuss the rest of this week's events over the fold:
If you click through to the permanent Senate page for the Poll Explorer, you'll notice we've added a new little widget to further explain the Kansas situation, below the rest of the Senate totem pole. Thanks to the 1-point lead in this new poll, Orman has only a 57 percent chance of winning the race. From there, he'll caucus with the Democrats only if they have a clear majority, or, according to our expectations, 50 percent of the times when Orman's caucus decision is determinative (i.e. if the Senate splits 49 D/50 R before Orman decides). Since only about half of the Senate scenarios get the Democrats that far, only 36 percent of our simulations end with Orman a) winning and then b) joining the Democrats.
The 49 D/50 R scenarios just aren't that common, though; the final number that we've added to our Kansas widget is the odds of that 49 D/50 R split and Orman's caucus vote being the decisive one. This happens in only 8 percent of all simulations, so the odds of that unusual drama playing out are pretty slim. Most simulations either end with the Democrats getting to 50 or more, or Republicans getting to 51 or more, without Orman's help, reducing Orman to just a bonus member for the winning side anyway in the other 92 percent of simulations.
Despite the decline in the odds of a Democratic "win" in Kansas from 54 percent a week ago to 46 percent on Monday and now just 36 percent today, the median number of Democratic seats held steady at 49. In fact, the Democrats' odds of retaining control of the Senate ticked up very slightly, from 46 percent overall on Monday to 47 percent overall on Thursday. There are a lot more moving parts in the model than just Kansas, and a few other things broke the Democrats' way this week.
The biggest move was in Michigan, where a new poll from the Glengariff Group gave Democratic candidate Gary Peters his first double-digit lead of the race, 47-37. That boosted odds of a Democratic win in Michigan from 67 percent up to 81 percent. However, a smaller move in North Carolina may have had a bigger impact on the Democrats' overall odds, considering how close to 50-50, and how close to the model's pivot point, the NC-Sen race is. Kay Hagan's odds of a win in North Carolina increased from 47 percent to 55 percent. Oddly, that may be thanks to a Republican internal poll that the NRSC leaked, showing a tied race; once you've factored in the penalty for partisan polls, that's effectively a poll with Hagan leading by 3.
Turning to the gubernatorial races, you'll notice that the median number of Democratic gubernatorial seats ticked back down to 22 from 23 previously; 22 would represent a net gain of one seat. The model has fluctuated back and forth between these two numbers for several weeks.
The drop seems largely caused by the shift in Rhode Island, which fell from 99 percent Democratic odds to 56 percent odds. That's not because of a new poll, but rather because of previous polls being subtracted from the database. There's been precious little polling of the general election in the Rhode Island race this cycle, and those polls tried a number of different candidate configurations that never came to be.
With the primary election finally over this week, we know the Democratic and Republican nominees are Gina Raimondo and Allan Fung, respectively. Unfortunately, though, there's been a grand total of one poll of this matchup all cycle; it's from all the way back in October 2013 (check out the size of the error bars on the Rhode Island trendlines on the Poll Explorer permanent page, which are huge, given how long ago the poll was). That poll from Brown University gives Raimondo only a 38-36 lead over Fung. We expect future polls of the race to give a much larger lead to Raimondo, but, for now, that's all we can base the model on.
That's offset somewhat by improved Democratic odds in Kansas, where that same SurveyUSA poll that we mentioned above gave Paul Davis a 7-point lead, pushing Dem odds in the gubernatorial race there from 74 percent to 85 percent, meaning it's now just as likely to flip as Maine is! On the downside, Democratic candidate Ed FitzGerald's odds in Ohio fell from 8 percent all the way down to 1 percent, thanks to two new polls finding him down by 19 and 20 points.