Challenge Accepted

Jonathan Bernstein has a piece in which he argues that Democrats should be, so to speak, irrationally exuberant about 2012, because doing so will encourage stronger Democratic candidates, both in the form of fewer retirements and in better challengers. I agree with his logic, and his example of how 1994-1996 played out, with Clinton winning big in '96 but Democrats failing to capitalize in the House, possible because of premature Eeyorism, is a good one. Toward the end:

So, whatever they actually believe, Democrats should pound the 1996-2012 analogies, remind themselves that reversals are not only possible but happen all the time (I bet there are all sorts of other historical examples easy to dig up: liberal bloggers, consider that a challenge!), and convince themselves and anyone who will listen to them that good times for Dems are, once again, right around the corner.

I like historical elections numbers, so here goes.
In 1994, Bill Clinton's Democrats lost 54 seats, and the majority, in the House of Representatives; two years later, Bill Clinton won a 9-point popular vote margin and 379 EVs. In 1982, Ronald Reagan's Republicans lost 26 seats of a 192-seat majority; in 1984, well, Reagan kind of cleaned up. In 1954, Dwight Eisenhower's Republicans lost 18 seats, and their narrow majority; in 1956, Dwight Eisenhower won a 16-point victory and 457 EVs. In 1946, Harry Truman's Democrats lost 54 seats and the majority, and in 1948, Dewey defeated... oh wait, Truman won a 4-point victory and 303 EVs. In 1938, President Franklin Delano Roosevelt's Democrats lost seventy-two seats to the Republicans, though not the majority (which tells you something); in 1940, FDR won a 10-point victory and 449 EVs. In fact, in 1942 FDR's Democrats lost another 45 seats, and he won 432 EVs in 1944. In 1922, Warren Harding's Republicans lost 77 seats, and in 1924 his successor, Calvin Coolidge, won 382 EVs and one of the biggest popular-vote margins in modern history. (Not a great example, since Harding had died and been replaced...) In 1914, Woodrow Wilson's Democrats lost 61 seats; in 1916, Wilson won re-election by a narrow margin. There are plenty of times when Presidents have lost big in midterms in the last 100 years and have then won, often by big margins, in the subsequent Presidential election. I ran the numbers, in fact. Using all midterms going back to 1906, the equation relating number of seats the President's party loses in the midterm (x) to the percentage of electoral votes they receive in the next Presidential election (y) is:

y = 0.0021x + 0.6358

That is to say, each extra seat your party doesn't lose nets you, on average, an extra 0.21% of the electoral college vote. With 538 electors, that's an extra 1 elector. So yeah, that sounds bad: it suggests that the expectation for Obama's re-election EV vote is about 65 electors lower than if the Democrats had lost nothing in the House. But the R-squared value for this linear approximation of this trend, the number that tells us how well this trend fits the data, is 0.0347. In rough terms, that means that midterm performance accounts for about 3% of the variance in re-election performance. Practically none. There are plenty of Presidents who way overperformed their midterm results, and plenty who way underperformed them. In fact, the two Presidents whose parties lost more seats than the Democrats did just now in their midterms, Harding and Roosevelt '38, both saw their party win over 70% of EVs in the next Presidential election. And sometimes a good midterm showing doesn't save you: Carter's 1978 Democrats only lost 15 seats, but Carter didn't do so well in 1980. George H.W. Bush's 1990 Republicans only lost 9 seats; Bush lost badly to Clinton. Things change. Approximately 97% of the 2012 election has nothing to do with last Tuesday. For every Hoover or Taft, who got pounded in their midterms and pounded at re-election, there's a Roosevelt, a Truman, a Clinton, an Eisenhower. Do you think Obama is more like Herbert Hoover or Harry Truman? William H. Taft, or Bill Clinton?

Actually, that's a further interesting point. If I split my sample by partisanship, we see something interesting. The relationship between midterm performance (x) and re-election performance (y) for Republican Presidents:

y = 0.0047x + 0.7166

So Republican Presidents lose an average of 2.5 electors for each Congressional seat they lose. And Democrats, then?

y = -0.0002x + 0.55

Yes, that's right: Democrats, on average, win a higher percentage of electoral votes in Presidential elections after their midterms in which they get shellacked. And that's despite Roosevelt's '34-'36 performance, picking up 9 seats in 1934 and then winning 98% of EVs in 1936. Take that away, and we get y = -0.0034x + 0.3951, so Democratic Presidents other than FDR in 1936 win an extra 1.8 electors for each additional House seat they do lose. The R-squared values for these three equations, for what it's worth, are 0.1497 for Republicans, 0.0004 for Democrats, 0.0917 for non-FDR'36 Democrats. Midterm elections make no difference to Democrats. Obama 2012!

UPDATE: I decided to see what happens if I account for "overexposure;" that is, calculate midterm losses in terms of percent of caucus lost rather than absolute seats lost. Now I get, using x = net gains over seats previously held and y = EV% in the next Presidential election:
For all Presidents:

y = 0.7007x + 0.6562    R-squared = 0.0527

For Republican Presidents:

y = 1.233x + 0.7381     R-squared = 0.1729

For Democratic Presidents:

y = 0.0544x + 0.564     R-squared = 0.0003

So, to recap. 17% of Republican Presidents' performance at re-election time comes from midterm performance, adjusted for exposure, and 15% even if you don't account for exposure, but 0.03% of Democratic President's performance comes from the same source. So, Democrats: don't worry, this election literally told you nothing about how Obama will do in 2012.