2012 Senate Election Forecasts

I know I will regret this. I'm not drunk enough to do this yet (8:54 PM)
Now I am (9:30 PM)

If not specified I predict that the same caucus will keep the Senate seat (so for Connecticut the Democratic Caucus not the Connecticut for Lieberman party)


Senate

Maine              King (I) who will caucus with Democrats
Massachusetts Warren (D)
Virginia            Kaine (D)
Florida             Nelson (D)
Indiana             Donnelly (D)
Missouri           McCaskill (D)
Wisconsin        Baldwin (D)
Nebraska        Fischer (R)
N Dakota        Berg (R)
Montana          Rehberg (R)
Arizona            Carmona (D)
Nevada            Heller (R)

Mostly I just predict the candidate who is ahead in the average poll will win.  For 6 races this margin is tiny and I think.

Virginia:  Some averages show this closer than close.  An important reason is a Roanoke College poll which shows Allen (R) up by 5.  Roanoke doesn't poll outside of Virginia and doesn't have much of a track record. Their numbers bounce around a lot.  I guess that they don't do much demographic weighting so the sampling error is equal to the stated sampling error (most pollsters weight to make their sample of adults match census data on some dimensions reducing sampling error then ignore this when making cautious estimates of standard errors).  So I weight them less nudging Kaine into a lead.

In Wisconsin again the averages are very close with Baldwin very slightly ahead.   There are some pollsters which I suspect of overestimating Republican vote (Wenzel Strategies and Rasmussen).  This is enough to make me guess that she will slightly outperform the average poll which implies Senator Baldwin.

N Dakota is very close.  Without looking at polls I would guess a Republican is more likely to win.  The basically tied polling doesn't move my posterior appreciably from this prior.

Montana is very close.  In spite of my gross error in 2004, I still have some sense that undecideds break against the incumbent (Tester (D)).  Also many polls name the libertarian.  I think slightly more people say they will vote for a no hope of winning third party candidate than actually throw their votes away and I assume that their second choice is Rehberg.

Carmona is slightly behind in Arizona.  I am counting on Hispanic citizens who are even more reluctant to talk to pollsters in English than to vote.

Nevada is close too, but not close enough for me to bet on the hard to poll Hispanics effect and the strength of trade unions and Democratic turn out the vote efforts.

New Senate guess same as the old Senate with 51 Democrats 2 independents and 47 Republicans.

Update: ooops that's embarrassing.  New Senate guess Democrats gain one with 52 Democrats 2 independents and 46 Republicans.  I see I agree with Nate Silver except on Indiana and Arizona.  I am quite confident about Indiana.  Silver's model doesn't consider whether there has been a big event like a candidate saying rape babies are gifts from God.  So the Howie DePauwe poll showing Donnelly ahead by 11 is just tossed in the smoothed average.  I am not at all confident about Arizona (or for that matter Montana where Silver gives Tester (D) a 33% chance).
/update

Sanders  will remain a more reliable Democrat than any Democrat.  King will cause the caucus about as much heart ache as Lieberman.    If my guesses are right, the Democratic caucus will move left with conservadem Ben Nelson gone, centrists Conrad and Tester gone, liberal Warren added, and mainstream Democrats added (2 replacing outgoing mainstream Democrats).  The Republican caucus would move to the right with moderate Snowe sometimes pseudo moderate Brown and not as extreme as Cruz anyway Hutchison gone.  So I guess the Senate will be less harmonious and more partisan hard as that is to imagine.

Warning.  My predictions are almost always wrong (my 2010 Senate predictions were correct).