We are transitioning, a little earlier than I'd planned, into a more or less fully automated version of our Senate race rankings. From now until November, our rankings will be based entirely on polling and other objective variables, with one important exception that I'll explain in a moment. But first, let's get the bad news out of the way for Democrats:
Right now, the program is showing that Democrats will retain an average of 54.7 seats in the 112th Congress. The distribution, however, is slightly asymmetrical, so the median number is 54, and the modal number is 53.
And things could, potentially, get a whole lot worse than that; the program recognizes that the outcome of the different races are correlated based on changes in the national environment. Between the surprise in Massachusetts, and races like California and Indiana which are potentially coming into play, there's about a 6-7 percent chance that Republicans could actually take control of the Senate, and another 6 percent chance or so that they could wind up with a 50-50 split. On the other hand, there's still a 7-8 percent chance that the Democrats could regain their 60th seat if the national environment shifts back in their direction.
I've already incorporated a couple of improvements to the algorithm that we used in 2008. For one thing, as our Presidential model used to do, the program now builds in a 'trendline adjustment' in races where the polling is stale. Although it hasn't been exceptionally dramatic, Democrats have lost a net of 3-4 points over the past 60-90 days, and several races that once looked like toss-ups are now better thought of as leaning Republican.
Another improvement -- also borrowed from the Presidential model -- is that we're now incorporating an adjustment for house effects. For whatever reason, the house effects are fairly strong this cycle -- if Rasmussen and Research 2000 poll the same race, the Research 2000 survey will typically show an outcome about 6 points more favorable to the Democrats than Rasmussen does.
Lastly, as you'll see below, I've modified the program to handle cases in which the identity of the candidates isn't known -- which is the case in most of these races given that all but a handful have a competitive primary in one or both parties. And this is where the subjective part comes in: I'm eyeballing the primary polling and other news items to estimate the chance of each candidate becoming his party's nominee -- the black numbers you'll see in the tables below to the left of the Democratic candidate's name represents the likelihood of each permutation of major party candidates. FWIW, I'm not trying to be 100 percent comprehensive here about things like unexpected retirements; this ought to be a decent, but fairly rough guess about the most salient combinations.
As in the past, the Senate algorithm is based on a combination of polling and regression analysis. In fact, because a lot of these races haven't been polled adequately, the regression component takes on a fairly large weight. The variables in the regression analysis are a state's PVI, an average of the candidate's approval and favorability ratings in races featuring an incumbent, and a variable that runs from 0 to 3 to reflect the highest office that each potential candidate has held, which scales as follows:
3. Current or former governor or senator.
2. Current or former U.S. Rep.; other statewide elected office; mayor of large city.
1. State legislature; mayor of small city; other non-trivial elected office.
0. No substantial experience as elected office-holder.
Eventually, we'll compliment or substitute this variable with fundraising data, but since a lot of potential candidates have not yet filed for office or did so only recently, the FEC data is not terribly informative and won't be for some time.
OK, enough window-dressing ... here are the Senate races ranked in their traditional order from most to least likely to change parties:
Dems have virtually no chance of hanging on in North Dakota. I not sure why people are classifying this as anything other than safe GOP.
This was a tough enough state for Democrats in 2008 and in the 2010 environment ... look out. Lincoln's polling is about where it "should" be, according to our regression analysis. The prospective primary candidate, Lieutenant Governor Bill Haller, would do roughly as well.
This is a very heavy lift for Harry Reid -- if he's the candidate, we have him at less than a 20 percent chance of holding onto his seat. There's plenty of polling in this race and while there are a few incumbents who have come back from the sort of hole he finds himself in, it's awfully rare. Although PPP's polling showed that Shelly Berkley's numbers weren't any better than Reid's, the algorithm begs to differ: she's an experienced office-holder, and it's hard to do worse than an approval rating of about 40 percent. A long-shot candidacy by Vegas mayor Oscar Goodman could actually make Dems the favorites here.
Surprised that the computer liked this one as much as it did, but Pat Toomey's lead is only growing in the polls and Specter's approval ratings are fairly wretched. Joe Sestak might be a marginally better choice, although he hasn't really clicked in the general election polling. Exit question: would the Democrats have been better off had Arlen Specter never switched parties?
I'm actually a touch more optimistic than the computer on this one since what little I've seen of Jane Norton hasn't impressed me, but the cycle has turned sour on the Democrats while they were busy getting their act together in Colorado. Bennet's approval ratings are middling enough that it's a toss-up as to whether he or Andrew Romanoff would be the better general election candidate.
Delaware's ranking has moved down a bit in relative terms as other races have lapped it, but not so much in absolute ones. Still, this is one of the more salvageable races for the Democrats if Beau Biden is the candidate.
Although Alexi Giannoulias isn't a terrific candidate, Illinois is blue enough that he's the favorite if he wins the primary next week. The other Democrats poll several points behind him and lack the experience of having run a statewide campaign.
Finally, our first Republican-held seat! It's nothing that Robin Carnahan is doing wrong, but her polling has started to have a little bit more trouble weathering the national storm.
Here's a surprise -- although not really to the Republicans, who keyed into the opportunity almost immediately after Scott Brown's victory. Indiana, last year aside, is still a fairly red state, and the political climate over the past year hasn't permitted conservative Democrats like Bayh to stay smilingly above the fray. One caveat: the computer stupidly doesn't distinguish between Mike Pence and John Hostettler, because neither candidate has been polled and the only variable it uses to distinguish them is their highest level of office achieved (both have been U.S. Representatives). But Hostettler is a notoriously weak candidate and Pence, I'm sure, would be a strong one.
Undoubtedly the most boring race of the cycle -- both Democrats poll about equally against Rob Portman.
Paul Hodes might be saved if the Republicans nominate a wingnut rather than Kelly Ayotte.
Speaking of which, the Republicans are potentially taking a risk in Kentucky if they nominate someone who's never run a campaign before rather than their Secretary of State.
This is an interesting one. Boxer, whose approval ratings have been tepid, was probably going to get a bit of a mulligan if the politically daft Carly Fiorina had been the nominee, but Tom Campbell, who just entered the Senate race, would be a more formidable threat. Boxer is the favorite nevertheless, but any scenario in which the Republicans hope to pare the Democrats down to 49 or 50 seats will probably have to run through California.
It's probably generous to Charlie Crist to suggest that he's even-money to win the primary at this point. The Meek-Rubio polling has been all over the place as neither candidate is terribly well known outside of his base.
Tommy Thompson could be threatening, but Feingold's approval ratings are robust enough that he'll only have trouble against more the generic GOPers if the national climate goes from bad to nightmarish.
Winnable with a tip-top campaign for the Democrats. Elaine Marshall looks like the best candidate.
A lot of scenarios that are hard to game out, made more complicated by the fact that Kristin Gillibrand's approval ratings are all over the board depending on which poll you look at. But the algorithm's basic intuition seems sound: a straight toss-up if George Pataki is the nominee, and otherwise Dems will probably need to screw up to lose the seat.
A non-generic Republican would have a fighting chance, but Patty Murray's approval ratings have been reasonably sound.
Not the right cycle to win a race like this.
Johnny Isakson's approval ratings are fairly mediocre, but Dems don't really have a candidate.
Linda Lingle's probably not quite competitive enough to warrant the gamble here.
Wyden shouldn't have any problems.
Democrats need to have a contingency plan here in case J.D. Hayworth is the nominee; a non-generic candidate like Phoenix Mayor Phil Gordon would likely be competitive.
Democrats likely need a scandal or gaffe to have much of a shot here.
How bad would this cycle be for Democrats if Chris Dodd were still the candidate?
It's something of a moot point now, but it's disappointing that Democrats never mustered a candidate here back when the cycle was looking more competitive.
As you might imagine, this is quite a bit of work, so the Senate race rankings will continue to be updated only on a monthly-or-so basis until we get a little deeper into the cycle.