With virtually no one any longer contending that the Democrats will reclaim a majority in the House--or even pick up, at best, more than 3 or 4 seats--the only possibility of a shift in congressional control from the 2004 elections comes in the Senate. Here, the race seems very much like the current one in the Electoral College--slightly favoring the Republicans, and with as much chance as a significant GOP gain as a Democratic victory.

Counting Jim Jeffords as a Democrat, the Dems start the race behind 49-51. And, making the (not unreasonable) assumption that if Kerry is elected, Massachusetts voters will replace him with a Democrat, the Dems need to pick up one net seat to take the majority if Kerry is elected and two seats if Bush is re-elected. Quite beyond the issue of coattails, then, a Democratic majority probably depends on a Kerry victory.

Three seats seem all but certain to change hands: Illinois, where everyone's favorite GOP candidate, Alan Keyes, trails by 51 points in the latest poll (68-17); Georgia, where the Dems have had no chance for months; and South Carolina, where the Dems probably lost their only chance at victory when former GOP governor David Beasley was defeated in the Republican runoff. So, not counting any of the close races, Republicans start the contest with a one-seat gain.

Five seats seem to have shifted in the direction of the party that currently occupies them over the past several weeks--Washington, North Carolina, and South Dakota for the Democrats; Missouri and Pennsylvania for the Republicans. That leaves five more open seats (Florida, LA, Oklahoma, Colorado, and Alaska) that will decide, with the Dems needing to win all five to reclaim the Senate if Bush wins, and four of the five if Kerry prevails.

The Florida race has been slow to develop, but it's been a decade since a Republican has won a Senate race there, and the Democrats certainly nominated a solid candidate in Betty Castor. In Oklahoma and Alaska, the Dems nominated their strongest conceivable candidate in both races (Brad Carson and Tony Knowles) and the Republicans nominated two deeply flawed candidates in Tom Coburn and Lisa Murkowski. Nonetheless, although it's certainly possible, it's hard to imagine the Dems winning both of these states, especially since Bush seems likely to carry both by 25 points or more: no Democrat has won a statewide federal race of any kind in Alaska since 1974, and David Boren is the only Democrat to win an Oklahoma Senate race since 1966.

If Oklahoma and Alaska split, that leaves Louisiana and Colorado to decide the outcome. The only Democrat to win a Senate race in Colorado in the last 18 years was Ben Nighthorse Campbell, but he quickly defected to the GOP, and Democratic nominee Ken Salazar recently lost his lead to GOP candidate Pete Coors. In Louisiana, Dems have increasingly become reliant on the state's peculiar election system, in which candidates from all parties appear on the ballot with a runoff the first Saturday in December if no candidate receives a majority. They've erased large Republican leads in the open primary in the 1996 and 2002 Senate races and the 2001 gubernatorial contest. At some point, though, it would seem as if their luck will run out. If I had to guess at this stage, I would say that Salazar will win Colorado and GOP nominee David Vitter will capture Louisiana, which would produce a 50-50 Senate.

There is, however, one other historical trend worth considering. In each of the last four elections, there has been one notable Senate upset: 1996 in Nebraska, with Chuck Hagel over Ben Nelson; 1998 in North Carolina, with John Edwards over Lauch Faircloth; 2000 in Washington, with Maria Cantwell over Slade Gorton (courtesy of the final absentee ballots from Seattle, counted days after the election); and 2002, with Saxby Chambliss over Max Cleland. Alarmingly for the Dems, the only possible candidates for an upset at this stage seem to be Democrats--Tom Daschle in South Dakota, Patty Murray in Washington, and, perhaps most likely, Wisconsin's Russ Feingold, who has demonstrated a tendency to fade late in both 1992 and 1998. This is one historical pattern that Democrats will be hoping to avoid this election day.