Jonathan Chait of TNR makes some good points about establishing the baseline for the upcoming Dem debacle:
It's certainly legitimate to question the policies or the tactics of the Obama administration. But in order to have that conversation, you need to begin with a baseline expectation. What sort of performance should we expect normally? Clearly, in the current environment, it's not rational to expect the majority party to escape any losses whatsoever. If you want to blame the Democrats' loss on bad messaging or wimpy policies or rampaging socialism, then you need to establish how you'd expect them to do given normal messaging and policies.
Political scientist Douglass Hibbs has a model of the election. It takes account of three factors:
1. The presence of a midterm election, which generally results in losses for the president's party.
2. The incumbent party's "exposure" -- the more seats you hold, the deeper into hostile territory you're stretched, and the easier it is to lose seats.
3. Personal income growth, which heavily influences out-party behavior.
The model does not include presidential approval rating. Indeed, it doesn't include anything other than structural factors. That model predicts the Democrats will lose 45 seats in the House.
If those predictions pan out Obama and Team Progressive will have underperformed by only a few seats. Now taking a step back, one might wonder whether there was really nothing the Dems could have done to improve the economy (which would have improved personal income growth). Saying that the Democrats will lose seats because they will take the blame for a weak economy sidesteps the question of whether that blame is fairly placed. No one thinks that the President has total control over the economy, but would Mr. Chait really argue that the current state of the economy is utterly divorced from anything Obama might have done about it?
But that is not what struck me about this modeling! What struck me about the data fitted to years past (in the Hibbs paper), is that there was *no* Watergate effect in 1974 and *no* impeachment effect in 1998.
By way of contrast, Republicans did out-perform the model in 2002 and under-perform in 2006; both results could fairly be attributed to terror and Iraq (but see below).
Republicans also underperformed by about 20 seats in 1970, which I would attribute to Cambodia and Kent State. In 1994, Republicans outperformed by (approximately) a dozen seats, which may be a useful guide to the likely direction and magnitude of mis-estimation this time around.
Still, I remain deeply suspicious of a model that is telling me that the resignation of a President in 1974, or the impeachment of one in 1998, had no impact on the subsequent midterm elections.
Oh well - Prof. Hibbs has been doing this for a long time.
SELF-QUIBBLE: Matching under-performance and over-performance with non-economic news (as I did above for 1970, 2002 and 2006) makes a certain sense, since this model does not specifically incorporate other variables (Hibbs did make a Presidential "Bread and Peace" model).
On the other hand, every day the financial news includes some good news, some bad news, and the reported market movements. Sometimes there is an obvious connection; other days, what is obvious is that the reporter simply said, hmm, the market went up so I guess the good news was the driver. And maybe it was, or maybe the driver was something overlooked by reporters but not investors.
BONUS QUIBBLE: Just to stick with Watergate, the 1974 data point was used to estimate the model parameters, so in effect, 1974 dragged the model towards itself. Maybe the model result would be quite different and there would be an obvious 'Watergate effect' if 1974 had been held out of the data used to estimate the model. Or maybe the weak economy in 1974 created an environment in which Nixon's opponents could force him to resign and in 1999 Clinton survived due to a strong economy bouying his popularity.