That consensus, throughout our history, has shifted and evolved as our nation has grown more enlightened and our understanding of liberty has broadened. For instance, today every adult can vote, no person can own another, and our children are in integrated schools rather than sweatshops. These changes are the direct result of the appropriate exercise of each branch’s authority, working not only in harmony but in competition as well.

It seems you’re characterizing reversal rates as something more than they actually represent. They’re not a reflection of “activism” at all, but merely a cataloguing of outcomes during the exercise of judicial authority as it’s evolved over the centuries. The trouble with the “activism” framing is that it makes the judicial process seem sinister and negative, and as an overstepping of boundaries, when in fact, it’s not. The reason we have multiple layers of judicial scrutiny is because our democracy works hard to be just. The unfortunate label “judicial activism” should be rejected rather than “researched.”

I missed replying to this comment the first time around. The overwhelming majority of standards of review used are de novo and clear error. The other standards are used less frequently. I am coding for different areas of law although I am not making particularly fine distinctions. Criminal cases, for example, have lower reversal rates regardless of the standard of review used so I have tried to make sure the mix of criminal cases is at least reasonably close among judges.

So far, I haven’t done much to deal with selection effects for a few reasons. First, my sample sizes for each judge are large enough such that the case mix isn’t much different (at least for the areas I have coded). Second, so far my data hasn’t indicated that the activism differential varies substantially based upon area of law. Third, since I’m still gathering data, I haven’t done much to normalize small differences simply because I can’t be sure if the observed differences will continue in other circuits. Fourth, even though I have a lot of data in one sense, when I use judges as a unit of measure, I only have 52 individuals with adequate sample sizes. So, I haven’t been able to really test for statistical significance in a meaningful way to decide if I need to make some refinements.

At some point, I might have a circuit that has a very different reversal rate overall or a different case mix. In such cases, I have a couple of strategies in mind to deal with those situations. Probably the toughest case for me will be when I include the D.C. Circuit since that court has a very different docket (and I’m not even bothering to look at the Federal Circuit).

If you have any more ideas, let me know – I’d love to hear them. Thanks.

Corey

I have been working my way backwards through this very interesting serious of posts. Could you elaborate a little on how, if at all, the case mix differs between the cases under a deferential standard v. a non-deferential standard. What strategy did you adopt to deal with the different selection effects that might be underlying your results?

Cheers

Matt