Our latest forecast shows little change in Republican chances of taking over the Senate after the Nov. 2 election. They now have a 19 percent chance of doing so, according to our forecast model; their chances had been 18 percent in our previous update on Thursday.
There are, however, some important states where Republican prospects have been improving, albeit slightly. One is Illinois, where the last three nonpartisan polls each give the Republican Mark Kirk a small lead; he now has a 1.7-point edge in our adjusted polling average.
Because the number of undecided voters in Illinois remains high — which contributes uncertainty to the forecast — a small lead is less meaningful than it would be ordinarily. Still, our model has Mr. Kirk’s chances of victory up to 64 percent, the highest they have been all year in a race that has been essentially tied for months.
Although there has been no polling in Nevada or West Virginia within the past 48 hours, the forecast had also moved incrementally in the G.O.P.’s direction in those states earlier this week.
The polling in West Virginia has been all over the board, with recent surveys showing everything from a 10-point advantage for the Democrat Joe Manchin to a 7-point lead for the Republican John Raese. The dynamic in the state is unusual — a popular governor, Mr. Manchin, on the ballot in a state where President Obama is exceptionally unpopular — and it is not surprising that we’re seeing somewhat disparate polling. Nor would it be surprising if either Mr. Manchin or Mr. Raese won by a margin in the high single digits on Election Day.
For the time being, however, West Virginia is the very definition of a tossup: our model currently has Mr. Manchin “winning” with 48.749 percent of the vote to 48.748 for Mr. Raese. This, however, is a slight improvement for Republicans, as the model had pegged Mr. Manchin as about a 60 percent favorite a week ago.
Many people are already voting in Nevada, and turnout statistics collected by the state show Democratic and Republican voters having turned out in almost equal numbers so far. That means the race will probably come down to independent voters, a group with whom most public polls show the Republican Sharron Angle having a very small edge; that’s essentially why our model now makes her about a 2-to-1 favorite.
Still, in comments to FiveThirtyEight and to other reporters, Harry Reid’s campaign has been critical of some of the public polling, particularly those polls that do not include any of the six third-party candidates that will appear on the Nevada ballot, nor the state’s unusual option to allow voters to select none-of-the-above. Mr. Reid’s campaign believes these choices will deny Ms. Angle some votes she otherwise might have received, and their internal polls reportedly show their candidate ahead.
It is unclear whether the so-called “full ballot test” would make all that much difference: public polls that have tested the race both with and without the third-party and none-of-the-above options included have found it makes little difference in the margin between Mr. Reid and Ms. Angle.
Still, the public polling was fairly poor in Nevada in 2008, with many surveys having underestimated Barack Obama’s victory margin by several points. And it can be a tricky state for pollsters in many ways, with much of its citizenry keeping unusual hours because they are employed in the gaming and service industries there, and with the state having a lot of new residents that may rely on cellphones rather than landlines or who may otherwise elude pollsters.
Even if Republicans were able to win Nevada, however — as well as West Virginia and Illinois, and other states like Pennsylvania and Colorado where their candidate seems to have a small lead — they would still need to win either Washington or California to get their 51st senator. In those states, we think talk of the races having tightened is a little overdone, and that the Democratic candidates — Patty Murray and Barbara Boxer — still hold a 3- to 4-point lead in each state. This late in the race, a 3- to 4-point lead in the consensus of polls translates to around an 80 percent chance of winning, according to our model.
A similar dynamic is manifest in Pennsylvania: that race clearly has tightened, but most nonpartisan polls still give the Republican Pat Toomey a small lead, including a tracking poll that had previously shown an edge for Joe Sestak. Our model still sees Mr. Toomey as about a 3- or 4-point favorite, which translates into victory chances a hair over 80 percent.
Certainly, Mr. Sestak could win his race in Pennsylvania. But if he does, that might suggest that the enthusiasm gap isn’t quite as large as some pollsters were expecting, or that there were other problems with the polls that had implications in a number of races around the country. In that universe, the Democrats would probably be favored to win at least one or two of the races in the group including Nevada, Colorado, Illinois, and West Virginia — all of which appear slightly closer to us than Pennsylvania — and Republicans would be long shots to win in California or Washington, where Democrats already have an edge.
In other words, Pennsylvania — while fascinating on its own merits — isn’t likely to be the state that prevents Republicans from achieving a majority that they otherwise would have claimed. Instead, Washington and California remain the states that represent the biggest barriers to Republicans winning a 51st seat. What should make Republican poll-watchers happy is if new polls come out in the next week showing their candidates ahead in either of those states, particularly if they come from pollsters whose surveys do not ordinarily show a Republican lean.
There is also one technical change in our latest model. Ordinarily, the model reserves a percentage point or two for third-party candidates. There are four Senate races, however, in which only the Democrat and Republican will appear on the ballot, and there are no third-party candidates at all: these are Pennsylvania, Washington, Kentucky and Alabama. Therefore, our forecasts will now distribute all 100 percent of the vote to the major-party candidates in these states.
We will incorporate this change into our House and governor races’ forecasting beginning with models on Monday.