We’re gearing up for what is certain to be a very exciting and — given that there are a number of crucial races on the West Coast, particularly for the Senate — very long Election Night.
Among other things, we’re hoping to be able to update our House and Senate takeover projections as the night progresses. While we almost certainly won’t be updating our forecasts for individual seats, and we definitely won’t be “calling” any races until The New York Times does, we do hope to provide some forecast of the overall number of seats that Republicans are most likely to win in each chamber, and their probability of taking over both the House and the Senate. These top-level projections would be updated a couple of times an hour as the returns roll in.
In the process of preparing our model to do this, I’ve started to hone in on the seats that are likely to tell us the most about the disposition of the House as Election Night progresses.
In particular, what I’ve done is to take all 435 House seats and sort them in order of the margin we project in each one — from the Republican Ron Paul’s district, the Texas 14th, which we expect him to win by about 65 points, to the Democrat José E. Serrano’s New York 16th in the Bronx, where he should be re-elected by 70 points or so.
The outcomes of lopsided races like Mr. Paul’s and Mr. Serrano’s aren’t likely to be terribly exciting to many people other than Mr. Paul, Mr. Serrano and their immediate families. But a tremendous number of House seats are competitive this year.
In a series of charts below, I’m going to list what our forecasting model considers to be the roughly 150 most competitive House races, as of Saturday afternoon. This obviously casts a very wide net; it’s basically every race in which we expect the two candidates to finish within 20 points of each other.
The charts are organized by the time that we should expect to see returns starting to trickle in from each state. There are few judgment calls involved here, based on the past experiences of The Times’s data team, as some states have multiple poll closing times. As veteran election-watchers know, Indiana and Kentucky, where most polls close at 6 p.m. Eastern time, should be the first states to begin reporting results. Alaska, meanwhile, won’t start releasing vote totals until 1 a.m. Eastern, after people in the Aleutian Islands have had their chance to vote.
The House seats are further divided into three columns. The seats that reflect the G.O.P.’s path of least resistance to taking over the House are in the leftmost column. If the Republicans won exactly the seats in the left-hand column, but no others, they would gain a net of 39 seats from Democrats, and control the House, 218-217. This basically involves those seats where the Republican candidate is favored by 3 or more points by our model.
Seats where the G.O.P. could begin to build on their majority are in the middle column. These are seats that Republicans they are favored to win by our model, but by fewer than 3 points. If Republicans won each of these seats, their gains would total a net of 59.
(A technical note: our simulations have the Republicans picking up an average of 53-54 seats, rather than 59. This is because the forecasts are somewhat asymmetric: there are 14 seats in which we have the Republican favored by 0 to 2 points, versus 6 seats like this for Democrats. Our official forecast looks at these races probabilistically — that is, if our model has the Republican projected to win by 0.01 points, his winning chances are 50 percent and some very small fraction. We don’t “call” the race for him. But if we do allocate all of the toss-up seats to one or another party, no matter how trivial its lead, we’re showing Republicans as favorites to win a net of 59 seats from Democrats.)
Finally, the seats in the far-right column reflect the tsunami possibility: those that Republicans would need to win to achieve a gain of 60 seats or more.
Each competitive race is represented by a little box that contains several key pieces of information. For the most part, this should be pretty self-explanatory:
The one thing that I’d like to draw your attention to is the statistic in parenthesis in the lower left-hand corner of the box: this is what we call the magic number. What this statistic indicates is how many seats we’d expect the Republicans to gain on the Democrats over all if they won this particular seat and all seats in which we have them favored by a larger margin.
In this particular example, for instance — the New York 19th congressional district, in which we have the Republican Nan Hayworth favored by 3 points — the magic number is 43. That means that if Republicans won this seat — and all other seats in which they were favored by more than 3 points, but none of the seats in which we had them favored by fewer than 3 points — they would finish with a gain of 43 seats on Democrats over all on the night. Another way to look at the magic number is that it’s the number of seats we’d expect Republicans to win nationwide if they won this particular district by exactly 1 vote (and we had no information about what had taken place in any other district).
Occasionally, the magic number will be negative; these are seats that, if the G.O.P. lost them, would imply that it were actually going to lose seats in the House overall.
What you should be looking for is whether Republicans are consistently winning seats with magic numbers in the 60s, 70s, 80s or higher. If so, they could be in for a very big night. Conversely, if Democrats are holding onto seats with magic numbers in the teens, 20s, or 30s, that means they are overperforming their forecasts and could hold the House.
By consistently, by the way, I do mean consistently: individual districts are fairly hard to forecast, and so Republicans almost certainly will pick off a few seats with very high magic numbers, and Democrats will almost certainly hold on to some with very low ones, regardless of what is happening elsewhere in the country.
You should also be watching the margin of victory in each district, particularly for districts in which enough of the vote has been counted that The Times has called the race. Our forecasts are calibrated to an overall Republican gain of between 50 and 60 seats. If they’re consistently winning their races by a larger margin than our model expects, that means their gains are likely to be somewhere beyond 60 seats. If they’re underachieving their margins, on the other hand, it may be below 50 seats, and Democrats might hold the House.
One reason we do try to be so precise with our forecasts — projecting a margin of victory in each race rather than putting them into broad categories like “toss-up” and “lean Republican” — is exactly so that it can serve this benchmarking function. It’s not that we know exactly what is going to happen on Tuesday; quite to the contrary, we think other forecasters are being incautious in not acknowledging the degree of uncertainty inherent to forecasting this House election.
But we can show you the blood, guts and entrails of roughly what a Republican gain of about 55 seats would look like, if that turns out to be the number: it would look something like this.
All right, that’s enough buildup. Let’s show you what to look for beginning at 6 p.m., when we’ll begin to see the first results from Indiana and Kentucky.
Baron Hill’s seat, the Indiana 9th, has long been one of the most competitive in the country. I don’t think you should get too swept up in the results of any one particular congressional district — not when there are 435 of them in every corner of the country. But Mr. Hill, a middle-of-the-road Democrat who ordinarily performs strongly in his fairly rural, somewhat Republican-leaning district, but who voted for the health care bill and the stimulus, is in a position that is fairly typical for Democratic incumbents around the country this year. Also, the district has a magic number of 41, which means that it’s right at the cusp of what Republicans would need to take over the House. If they fail to win it, that could be the first sign that they’re liable to do a hair worse than expected. If they win it by a margin in the high single digits or the double digits, however, it could suggest that a lot of Democratic incumbents, many of whom are less skilled than Mr. Hill at understanding how to run a strong campaign in their districts, are going to be in trouble.
Joe Donnelly, in the Indiana 2nd district, is one Democrat whose polls have held up fairly well in spite of the Republican wave. Our model has him favored by just 2 points, however, and if he were to lose, that would be a good early sign for Republicans.
Indiana’s 8th district, vacated by Brad Ellsworth, is very likely to be a Republican pickup. If they’re having trouble winning it, that’s a reasonably bad sign for them.
Indiana’s 7th and 3rd congressional districts are not likely to be especially competitive. If these races wind up within the single digits, something really weird might be afoot.
I’d be a little bit more cautious about reading too much into the two Kentucky districts on our chart, the 6th and the 3rd, just because Kentucky is a fairly idiosyncratic state to begin with, and both the polling and the Senate race have been strange there. Still, John Yarmuth’s 3rd district, which encompasses Louisville, reflects a strong potential upside case for the G.O.P. if they were to win it.
