Adam Liptak often writes about empirical studies of the Supreme Court, and his latest (with Alicia Parlapiano) reports on a new study relevant to the upcoming Supreme Court confirmation process. The study, by Lee Epstein, Andrew Martin and Kevin Quinn, measures (among other things) the relative ideology of a handful of possible nominees to fill the current Supreme Court vacancy.

After eyeballing the results, though, they struck me as a bit off. For example, the study suggests that Judge Bill Pryor of the 11th Circuit is more liberal than Judge Raymond Kethledge of the 6th Circuit. Also, it reports that Judge Merrick Garland of the D.C. Circuit is slightly more liberal than Judge Jane Kelly of the 8th Circuit. These results struck me as counterintuitive, so I took a closer look at the study.

Here’s how the study measures the ideology of the potential nominees. Because the potential nominees are circuit court judges, it uses the following proxy for their ideology:

• If a judge is appointed from a state where the President and at least one home-state senator are of the same party, the judge is assigned the ideology of the home-state senator.

• If both senators are from the President’s party, the judge is assigned the average ideology of the two senators.

• If neither home-state senator is from the President’s party, the judge receives the ideological score of the appointing President.

The study defends this methodology as “the tried-and-true approach” used in an earlier study, by Micheal W. Giles, Virginia Hettinger and Todd Peppers, “Picking Federal Judges: A Note on Policy and Partisan Selection Agendas,” 54 Political Research Quarterly 623 (2001). Liptak’s article further comments that the new study “used a common and reliable political science measurement to make predictions about the potential nominees.”

The approach seems puzzling in this particular context, though. Instead of actually looking at how the judge has ruled, it looks only at the measured ideology of the politicians apparently responsible for the potential nominee’s prior appointment. As I understand it, when the study says that Judge Garland is slightly more liberal than Judge Kelly, it is really saying that President Clinton was slightly more liberal than President Obama — ideologies that are then imputed to the respective nominees. Under this approach, Judge Garland and (say) Judge David Tatel are deemed to have the same ideology because they were both Clinton appointees to the D.C. Circuit. And if there were two vacancies from a single state, and a compromise was reached that filled them with one liberal nominee and one conservative nominee, the study would say the two nominees have the same ideology.

I understand using this methodology when we don’t know anything about the ideology of nominees. If you know zero about them, this approach is better than nothing. It probably gets you in the ballpark most of the time. And I understand using the methodology if we are making make generalizations about lots of judges, which is what the 2001 Giles study did. The Giles study studied the impact of ideology for every circuit judge, so it sensibly used a proxy to measure that to generate a large data set.

With that said, using the same proxy seems of questionable value when measuring the differences in ideology among a handful of possible nominees who are all known and have long records of judicial service. The chances that this methodology generates a more accurate signal than the alternative of reading the judges’ opinions and talking to people who know them seem remote. When a judge has been on a court for 20 years, it seems to me, you know the ballpark already. To continue the baseball analogy, at that point you want to know exactly where on the field the player is standing. The judge’s record of judging would seem to provide a more promising measure of that than the ballpark proxy used here. It’s a subjective measure, true, but I would think it’s a more accurate one.

Or so it seems to me. The authors of the study are all prominent empiricists, and perhaps I just misunderstand the study. If the authors think I am missing the point, I’d be happy to print a response from them.