A few days ago, I suggested that we could invert the usual forecast-the-election-from-the-economy rule and instead use historical election returns to make inferences about past economic trends.
Bob Erikson is skeptical. He writes:
It is an interesting idea but I don’t think the economics-vote connection is strong enough to make it work. At best econoims explains no more than “half the variance” and often less. Like I [Bob] am on record as saying the economy has little to do with midterm elections (AJPS 1990) unlike prez elections.
Damn. It’s such a cute idea, though, I still want to give it a try.
Andrew,
Economics has used this idea of transposing the treatment and response — I am unable to comment on the success or failure thereof. It's called "reverse regression" and is often applied in discrimination cases. The seminal article is by Goldberger, "Reverse Regression and Salary Discrimination", J Human Resources, 1984.
Regards,
Bruce
Bruce: Yes, calibration and reverse regression are well known in statistics as well. I actually wrote a paper on the topic in 2004, in the context of laboratory assays. What I was calling "cute" here was not the general idea of reverse regression but the specific suggestion of applying it to elections and historical economic trends.