Republicans, who are modest favorites to take over the House from Democrats, still have a chance to do the same in the United States Senate. But their odds have dropped significantly: from a 26 percent chance last week to 15 percent today, according to the FiveThirtyEight forecasting model.
The main reason for the decline is the outcome of Tuesday’s Republican primary in Delaware, in which the insurgent candidate, Christine O’Donnell, defeated Michael N. Castle. Two recent polls, including one completed after the primary, show her trailing her Democratic opponent, Chris Coons, by margins of 11 percent and 16 percent.
Although Ms. O’Donnell and Mr. Coons remain relatively unknown to some Delaware voters, and a comeback by Ms. O’Donnell is not impossible, the forecasting model gives it only a 6 percent likelihood of happening — and has established Mr. Coons, therefore, as a 94 percent favorite. Had Republican voters selected Mr. Castle instead, the numbers would be exactly the opposite: Mr. Castle would be the 94 percent favorite to win the seat, leaving Mr. Coons with just a 6 percent chance of an upset.
If Ms. O’Donnell were unable to surprise observers again in Delaware, the Republicans could still earn a majority, 51 Senate seats, in one of two ways: either by sweeping the Democratic-held seats that currently appear to be competitive — while holding all of their own — or by putting one or two additional states into play.
The first path — sweeping the Democratic-held seats — remains the clearer of the two. It is not uncommon for a party to win all or almost all “tossup” seats when they are having a strong election night, as the Democrats did to claim the Senate in 2006. The forecasting model accounts for this tendency, in that it assumes that the results of Senate contests in different states will be correlated to some extent.
The first several pickups of Democratic seats should come easily for Republicans. They are almost certain to win in North Dakota, where the Democratic incumbent Byron L. Dorgan has retired, and Arkansas, where a poll today put the Democratic incumbent there, Blanche Lincoln, at a 17-point disadvantage. And they are nearly as likely to do so in Indiana, where Evan Bayh retired.
Pennsylvania also looks good for Republicans: their candidate there, Pat Toomey, has slowly expanded his polling lead over the Democrat, Joe Sestak, and is now more than 90 percent likely to win the seat, formerly held by Arlen Specter, whom Mr. Sestak defeated in the Democratic primary.
Following that are three states in which the election looks more competitive: Colorado, Illinois and Nevada. But the model regards the Republicans as favorites in each, although by a trivial margin in Illinois and Nevada (a reversal from last week in the latter case).
Were Republicans to win these seven states, they would stand at 48 senators, and would need to convert three more Democratic seats to win the chamber. Those gains could potentially come in three blue states — California, Wisconsin and Washington — in which there are vulnerable Democratic incumbents.
The opportunity in California, where voter dissatisfaction is high because of high unemployment and the state’s fiscal crisis, is the clearest. Barbara Boxer, the incumbent, has tepid approval ratings, and a variety of polls show her race against Carly Fiorina polling within the margin of error, although Ms. Boxer remains the slight favorite.
But in Washington state, where Patty Murray is the Democratic incumbent, Republican prospects now appear more tenuous. The polling there has been inconsistent for some time, with some polls showing as large as a 17-point lead for Ms. Murray and others as much as a 10-point advantage for her Republican opponent, Dino Rossi.
Two new polls, however — one by CNN and the other by a local pollster, Elway Research — each show Ms. Murray with a 9-point lead among likely voters. In contrast to most other polls of the state, these were traditional, telephone polls with live interviewers, and did not use automated scripts. This is noteworthy because one of the automated polling firms, SurveyUSA, had underestimated Ms. Murray’s performance in the Aug. 17 primary, in which candidates from both parties appeared on the ballot together.
Whether automated polls, which have performed about on par with traditional ones in recent elections, will do so again this year is one of the most important questions facing electoral forecasters — and an answer is probably impossible until the November election. But in some states, like Washington, they have shown systematic differences, with the traditional polls tending to show more favorable results for the Democrat while the automated polls are more favorable for the Republican. For the meantime, however, the addition of the Elway and the CNN polls has brought Ms. Murray from being a slight underdog to roughly a 70 percent favorite, according to the model.
A similar problem is manifest in Wisconsin: among the nine polls of the matchup between Russ Feingold, the Democratic incumbent, and his Republican challenger, Ron Johnson, all but one is an automated poll, and 6 of them are from a single firm, Rasmussen Reports. Although each of the automated polls have shown a roughly tied race, the lone traditional poll — from the University of Wisconsin — gave Mr. Feingold a 5-point edge among likely voters. The forecasting model — partly because it accounts for some non-polling factors like Mr. Feingold’s approval ratings, which are decent — regards him as about a 2-to-1 favorite, but the state badly needs a greater diversity of pollsters.
Therefore, there are eight states in which the Republican gains seem unambiguous, and two more — Washington and Wisconsin — in which it is clear based on the automated polls, but specious based on the traditional ones. Were Republicans to win all 10 of these races, they would control the Senate — provided, that is, that they also hold all of their own seats.
And here is some good news for G.O.P.: they are doing quite well in all of the states currently held by a Republican, and appear increasingly likely to sweep them. Democratic takeover chances have dwindled below 10 percent in Ohio, Kentucky, Missouri, and North Carolina. Democrats have better chances in New Hampshire, but the Republican Kelly Ayotte, who narrowly won her primary on Tuesday night, is the clear favorite there. Meanwhile, Republicans no longer appear to be at as much risk of losing Florida to the independent, Gov. Charlie Crist. Instead, the Republican Marco Rubio is surging as Mr. Crist loses votes to the Democrat there, Kendrick B. Meek. Although three-way races are dynamic and highly uncertain, Mr. Rubio is now almost an 80 percent favorite.
But even if they were to hold all of their own seats, and win tossup races like Illinois and California, Republicans might need some luck — or some further political momentum — to completely sweep them. Thus, they may be looking to put additional races into play, particularly if their opportunity in Washington or Wisconsin is less strong than some polls suggest.
One possibility is West Virginia, where the Democratic governor, Joe Manchin III, is seeking the seat formerly held by Robert Byrd. West Virginia is another state in which there aren’t many polls to pick from: apart from Rasmussen Reports, which has surveyed the race three times and now gives Mr. Manchin only a 5-point lead, the only alternative is from R.L. Repass & Partners, which gave Mr. Manchin a large lead in a poll conducted by Internet.
West Virginia’s is an odd race: Mr. Manchin is very popular, and a plurality of the state’s voters still identify themselves as Democrats. But President Obama lost badly there in 2008 and has become no more popular. It requires skilled polling, which the one-size-fits-all approach of a firm like Rasmussen probably cannot provide. But with the available evidence, the forecast model gives the Republican, John Raese, about a 10 percent chance of a win — better, for instance, than Ms. O’Donnell in Delaware, and probably making him relatively more worthy of Republican attention.
Another opportunity that Republicans may be considering is Connecticut, where two recent polls of likely voters give the Democrat, Richard Blumenthal, a lead of 6 and 9 points, respectively. Here, the model is more skeptical of their chances: although a lead like Mr. Blumenthal’s is surmountable under most circumstances, the new polls, from Quinnipiac and Rasmussen Reports, show there are almost no undecided voters in the race (and have Mr. Blumenthal polling above 50 percent). This is perhaps to be expected, since Mr. Blumenthal, the state’s Attorney General, and his opponent, Linda McMahon, are familiar figures to voters, with Mr. Blumenthal keeping a high profile and Ms. McMahon having put millions of dollars of her own money into advertisements.
While the race in Connecticut has tightened considerably since the spring, there is no evidence that it is doing so further, and with the supply of undecided voters nearly exhausted, Ms. McMahon may be unable to make up much further ground. Although the forecast model’s call — it makes Mr. Blumenthal a 98 percent favorite — is too assertive for my tastes, and although the race is certainly close, its outcome is perhaps not all that uncertain.
Accounting as best as it can for all of these contingencies, the model estimates a 15 percent likelihood of a Republican takeover of the Senate. The party also has an additional 11 percent chance of winning exactly nine Senate races on Nov. 2., in which case they could potentially control the body by persuading Joseph I. Lieberman, an independent, to caucus with them.
Republicans remain strongly positioned to make large gains in the Senate: the model, after 100,000 simulations, has them finishing with an average of 47.1 seats, only slightly down from 47.5 last week. But their winning the relatively specific set of races that they would need to win in order to control the Senate now seems less likely, and the Democrats’ position is as secure as it has been in several months.