Social Science Statistics Blog

Two small points. First, as another, in Jim's apt words, "quantitative legal empiricist," I instinctively share his sense of unease with the legal academy's recent (and growing) embrace of all things empirical. That said (and self-interest aside), I have a hunch the current empirical "fad" will likely endure (unlike many other fads that periodically traipse through legal academia). Second, Jim's main substantive point is dead on--precedent matters. Empirical legal scholarship that neglects this point does so at considerable analytic peril.

Just so I understand your concern, Jim, are you talking specifically about modeling decisions as response variables, or is broader than that? And would you be less concerned about Supreme Court decisions, where precedents are kind of like constitutional conventions in the United Kingdom - things to be observed until they aren't?

Michael, Mike, first, apologies. I was testing the blog notification software, and as a result, did not get notified of either comment.

Michael, I hope you're right that you and I are not participating in a fad.

Mike, to answer your questions briefly, I think my worries depend on the modeling that's being done. "Often" (whatever that means), quantitative analysts worry less about the distributions of covariates (or background variables or whatever you call them). If appellate decisions are covariates, and the analyst understands that his/her inferences are limited to the portion of the covariate space covered by the data, then there is less concern. But when appellate decisions are modeled, either as outcomes or as covariates, then my stomach churns.

To answer your second question, in a word, "No." Precendent may matter less to the Supreme Court, particularly on constitutional cases, but my sense is that it still matters. Much of the first year of law school is about beating precedent into law students, and most folks on the high court got through the first year.