We have entered information from around 50 new surveys into our database. But most of them were either in less-than-critical states (two pollsters, for instance, decided to survey all three House races in Utah), or tended to confirm our existing impressions of the races.
Polling junkies out celebrating Halloween this weekend are advised to keep plenty of juice in their cellphone batteries. The Sunday night before Election Day usually features some newspapers issuing their final polls of the cycle in their states. These surveys, which may begin to trickle in during the wee hours of the morning and which often feature pollsters with intimate knowledge about their states, are often quite strong.
At the same time, except in a case like the Maine race for governor, where the numbers appear to be exceptionally volatile, or one like the Illinois Senate race, where there are an abnormally high number of undecided voters, you might want to avoid becoming overly obsessed with which way the races seem to be moving at the last minute.
Most voters —85 or 90 or 95 percent — make up their mind well in advance of the election. While it can be helpful to get a sense for which way the few remaining undecideds in a race are breaking, the last weekend before the election can sometimes produce misleading polling. Weekends, particularly Friday and Saturday nights, are tough to poll in general, and that’s especially so when the weekend coincides with Halloween (and also with get-out-the-vote efforts). In primaries, where a lot of voters do make up their minds at the last minute, you have no choice but to keep an eye on the polls until as late as possible. In general elections, however, polls conducted a week or two before the election often prove to be at least as accurate as polls still in the field on the final weekend.
Here is what our models saw based on the data they had to chew on tonight.
House: I’m going to be especially brief here because we have a very detailed overview of the House landscape due to publish on Sunday morning.
Our forecast is unchanged for the third consecutive evening: Republicans won an average of 53 seats in our simulations, although they were knocking on the door of 54. We have Democrats’ chances of holding the House at 16 percent, which is unchanged from Friday.
Two polls of House districts probably deserve some extra attention, however. A SurveyUSA poll of Minnesota’s 8th congressional district shows the longtime Democratic incumbent Jim Oberstar — the chairman of the House Transportation and Infrastructure Committee — just 1 point ahead of his Republican opponent, Chip Cravaack. SurveyUSA polls have had a rather strong Republican lean this cycle, and have overstated Republican performance in Minnesota in the past. Still, this is clearly a competitive race, and not the sort of numbers that Democrats want to be seeing on the final weekend.
A Mason-Dixon poll in Nevada’s 3rd congressional district, meanwhile, has outstanding trend lines for the Republican Joe Heck, who is trying to knock off the Democratic incumbent Dana Titus. The poll now shows Mr. Heck ahead by 10 points; he had trailed Ms. Titus by 4 points in Mason-Dixon’s previous survey, which had been completed just after Labor Day.
The slight counterpoint for Democrats is that they had very good numbers in the final day of early voting in Nevada, which they might need to save both Ms. Titus and Harry Reid. But if they are rallying, it is probably from behind in these races.
Senate: Republican chances of taking over the Senate are up a tick to 11 percent Saturday from 10 percent Friday. The reason is a Marist College poll showing Patty Murray just 1 point ahead of Dino Rossi in Washington. That’s not a terrible result for Ms. Murray, obviously — she holds the lead, however nominally so, and Marist’s results are unchanged from its poll 10 days ago. Still, it increases Ms. Rossi’s chances in our forecast from 17 percent to 19 percent; our model is very sensitive to any Washington polling.
It’s important to remember that what we’re seeing in Washington is some polls showing a tie, and others showing Ms. Murray with a small lead. That is different than the situation, for instance, in the Florida race for governor, where different polls show each candidate ahead. This is why our model still sees Ms. Murray as a favorite, rather than considering the race to be a tossup. Still, it would be no great surprise if Mr. Rossi won instead.
What would be a surprise is if Democrats won any of the Senate seats currently held by Republicans. The latest one to see his chances drop from slim to virtually none is Paul Hodes in New Hampshire, whom two new polls have trailing the Republican Kelly Ayotte by 15 points.
The Democrat whom we actually have as being the most likely to pick up a G.O.P.-held seat is Scott McAdams of Alaska: our model gives him a 7 percent chance against Lisa Murkowski and Joe Miller amid sparse and hard-to-read polling.
But let’s not lose perspective here. The fact that Republicans are poised to win open-seat Senate races in key swing states like New Hampshire and Ohio by something like 15 points (!!) speaks to how well they’re doing this year, whether or not they ultimately wind up winning the House or the Senate.
Governors: There is no movement of any real consequence in our forecasts. The closest thing to pass for it is the standing of Tom Foley, the Republican nominee for governor in Connecticut, whose chances improved to 16 percent from 14 percent on the basis of a YouGov poll showing him just 4 points behind Dan Malloy.
YouGov’s poll are conducted by Internet, but unlike some online polling companies, it has had decent results in the past, and its surveys show little discernible bias this year. YouGov also has a process that is reasonably well thought out and scientific.
It’s not that I don’t have any concerns about online polls — I do. But they are probably bound to get better rather than worse as Internet penetration increases; the opposite is probably true of the many polls that don’t include cellphones in their samples (landline penetration, which is declining, is now about the same as Internet penetration in American households). In any event, our policy is to weight surveys based on their past performance. YouGov’s polls get a slightly below average weight, but I don’t think there is any basis for ignoring them, and I wouldn’t be surprised if they performed fairly strongly this year.