It's been about two months since we last issued a Senate forecast and now I remember why: these things take a whole day of work. So let's not waste any further time; here are some emerging themes that I see.
Nationally, the trends are very flat. We are now using generic ballot polling, rather than the polling from individual Senate races, to create our trendline adjustment, a feature that was imported from our Presidential model. (Trust me, it's better this way.) However, we may as well not have bothered; we show essentially zero change in the national environment over the past several months, and only a net gain of one or two points for Republicans since the start of 2010. In contrast, Democrats lost about 12 points on the generic ballot over the course of 2009. They are not really climbing out of the hole the dug themselves, but on the other hand, it does not appear to be getting worse.
Locally, Democrats helped themselves in the primaries. Democratic fortunes were improved by the primaries in Nevada and Pennsylvania, California, North Carolina, and Kentucky, and worsened probably only in Arkansas (and South Carolina, which they had almost no chance of winning anyway.) This accounts for most of the movement in the rankings. Whereas, as of our last update, or simulations were projecting an average of 54.0 Democratic and 46.0 Republican seats, we now show 55.2 Democrats, 44.2 Republicans, and 0.6 Charlie Crists.
Republicans will need a lot of luck to take over the Senate. There are eleven Democratic-held seats that we show Republicans with a nontrivial chance of winning. In four of them, they are heavy favorites: North Dakota, Arkansas, Indiana, and Delaware. Four more are toss-ups: Pennsylvania, Nevada, Illinois, and Colorado. Finally, there are three where they are underdogs: Washington, California, and Wisconsin.
Republicans would need to win 10 of these 11 races to take over the Senate; even if they gained further momentum nationally (our model does not assume that the races behave independently), this is somewhat unlikely, given the idiosyncrasies involved in many of the contests. Meanwhile, they would need to hold Ohio, which is a toss-up with a slight Democratic tilt, and Missouri, which is a toss-up with a slight Republican tilt, as well as retain Kentucky, North Carolina and New Hampshire, either have Marco Rubio win in Florida or persuade Charlie Crist to caucus with them, and avoid a wildcard somewhere like Arizona or Louisiana.
Therefore, even though states like Washington and Wisconsin are now in play (more debatably so in the latter case), this is counteracted by the fact that they are now engaged in competitive contests in places like Nevada, Kentucky, and Pennsylvania -- which looked like clear edges for them before. In April, or model showed a 6 percent chance of a Republican takeover; this month, it shows an identical 6 percent chance, provided that Charlie Crist is counted as a Republican, and roughly a 4 percent chance if he isn't. If Republicans could hypothetically persuade both Crist and Joe Lieberman to join them, however, their chances would improve to 12 percent.
Alternatively, they can hope that Rasmussen is right and everybody else is wrong. Right now, we are identifying about a 4.5-point Republican-leaning house effect in Rasmussen's polls, relative to a robust average, weighted based on pollster quality. But suppose for a minute that Rasmussen is right and everyone else is wrong, and we calibrate to Rasmussen's average instead. In that case, the projected number of Republican seats goes from 44.2 to 47.4, and their chance of taking over the Senate more than triples to 22 percent, counting Charlie Crist as a Republican.
We are using the new version of the pollster ratings for this update, by the way, which still rates Rasmussen as a somewhat above-average pollster. The problem is not necessarily that Rasmussen is bad but rather that they're so dominant; something like 40 percent of our state-level polls were put out by Rasmussen, and they are just about the only pollster at all in some states. A guiding principal of our house effects adjustment is that nobody should be advantaged (or disadvantaged) simply because they poll more often, so it effectively mitigates the impact of Rasmussen flooding the zone, as it would for any other pollster which did something similar. With that said, Rasmussen certainly could be right, so it's useful to check sometimes on what what the overall Senate picture would look like through their lens.
If Democrats somehow got a wind at their backs, they have enough offensive opportunities to take advantage of it. Suppose on the other hand that the Democrats got a 3-point boost nationally (or the current average of polls is biased 3 points against them, which is effectively the same thing). In that case, they would have about a 27 percent chance of actually regaining a 60-seat majority, and closer to a 40 percent chance if they could persuade Charlie Crist to caucus with them. There's no particular reason to think that this will happen, however, particularly with economic momentum being rather tepid.
Some comments on individual races. For the most part, I'll let the numbers speak for themselves, but some brief comments on a few contests:
Arkansas. Our model now shows Blanche Lincoln's chances to be close to zero (technically, about 0.3 percent, which rounds down to zero). Bypassing an esoteric debate about whether the probability distribution should be more fat-tailed than we have it (if I could get 300:1 on Blanche Lincoln, I'd definitely take it), it's not hard to see why she's in trouble. She's way behind, her approval ratings are terrible, the polling is robust, the GOP nominated a competent challenger, Arkansas is a red state, and the numbers in races with incumbents tend to be less amenable to dramatic last-minute comebacks than those with two newbies. If Democrats spend much money on her, they are potentially costing themselves a win somewhere like Missouri or North Carolina, or a hold in an Illinois or a Washington.
Delaware. The numbers have reverted to the mean here a bit because there has been very little polling. And the dynamics are the reverse of what they are in other "sure things" on the Republicans' target list like Arkansas, North Dakota, and Indiana; Delaware is a blue state, and the Republican candidate is the de facto incumbent. Mike Castle will probably be fine, but it's worth keeping an eye on.
Florida. We're somewhat crudely trying to model a 3-way race by running three 2-way races and then amalgamating them together. It's not an ideal solution by any means, but it will probably change, so I'm not going to go into any great amount of detail telling you how ugly it is. With that said, somewhat to my surprise, the polling has pretty clearly established Charlie Crist as a favorite.
Nevada. That this race has become competitive again is mostly about Sharon Angle, but not completely so: Harry Reid's favorability/approval numbers have also improved some, and are now merely godawful rather than mind-bogglingly wretched.
New Hampshire. Although Paul Hodes has been underwhelming, the Republican primary is not until September 14th here, and the model thinks that Kelly Ayotte's numbers are running a little ahead of where they should be even relative to the very pernicious national environment for Democrats. There's also the chance that Republicans could nominate someone other than Ayoette, like Bill Binnie, who would turn the race into a toss-up.
North Carolina. While Democrats have a shot here, don't get too excited about that Rasmussen poll. Yes, ordinarily a Rasmussen poll showing a dead heat is good news for Democrats, but their polls also have this weird tendency to give very large bounces to candidates who just won primaries, as describes Elaine Marshall in this case.
Arizona. Although Democrats would prefer to run against J.D. Hayworth, John McCain's approval numbers are quite poor, and it's not completely out of the question that they could make the general election competitive against him.
Wisconsin. The model is not really willing to bank very much money on Russ Feingold being beaten by some no-name based on Rasmussen polling alone, especially with Feingold's approval numbers still being okay-ish. If other pollsters start to show the same thing, as PPP has hinted that they might, it could start to feel differently.
Numbers follow below. And yes, I know this particular graphic is ugly to the point of being unreadable, but we'll have much, much prettier ways to present this information once we get to the New York Times.