2. Pennsylvania (R-Specter)
There's been some contradictory polling in Pennsylvania, with Quinnipiac placing conservative challenger Pat Toomey, who would be a heavy underdog in the general election, 14 points ahead of Arlen Specter. Franklin and Marshall, meanwhile, gives Specter a 15-point lead. I don't know that there's any a priori way to say which poll is correct. Polling primaries is intrinsically pretty difficult, and polling races more than a year in advance is intrinsically pretty difficult, which means that polling primaries a year in advance is really difficult. If you simply average the two numbers, you get the Republican primary being a toss-up. Each of these polls, it should be noted, were taken before Specter's intention to vote against the Employee Free Choice Act became known to voters, a position which will presumably help him the primaries while probably harming him with Pennsylvania's fairly union-friendly general electorate.
Let's do some fuzzy math here. Assume that there is a 80 percent chance that Specter's health is such that he chooses to run for re-election. If Specter runs for re-election, there is a50 percent chance that he survives the primary. If he survives the primary, there is a 80 percent chance that he wins the general election, where the Democratic opposition has been a bit disorganized. If he loses the primary, however, or chooses not to run, the Republican nominee (presumably Toomey) will probably only have about a 10 percent chance of retaining the seat for the Republicans. Run all the numbers, and that yields a 62 percent chance that the Democrats pick up the seat. Granted, it's fairly dubious to imply any sort of precision when we're dealing with guesstimates like these. But my impression is that the Democrats' odds in New Hampshire, with Hodes now having the primary field to himself, are a just a touch better than that. Also, Specter's nay vote on EFCA probably eliminates the chances of the Democrats picking this seat up through a party switch.