Caperton v Massey I

Though my goal here is to look at the Supreme Court's use of science in the cases that are argued before it this year, I actually want to start two terms ago, with Caperton v Massey. Caperton, a case involving campaign financing and judicial recusal, does not immediately seem like a major "empirical evidence" case. But Kennedy's majority opinion is built around a wholly unsupported empirical statement, and Roberts wrote a lenghty dissent that accurately highlights the challenges empirical evidence poses for the Court--but then ends his opinion with a potentially major empirical gaffe of his own. So Caperton not only provides a good summary of the troubles empirical evidence creates for judges, but it also serves as a good case study for the ways that unacknowledged empirical assertions can ambush judges, even when the judges are thinking carefully about that very issue. This post will look at Kennedy's opinion, my next one at Robert's.

The facts in Caperton are relatively straight-forward. In 2002, Massey Coal lost a major tort case ($50 million), and in 2004, while the case was working its way on appeal up to the state supreme court, Massey's CEO, Don Blankenship, spent more than $3 million to successful help a challenge, Brent Benajmin, unseat an incumbent. In 2005, Brent Benjamin repeatedly refused to recuse himself from Massey's case when it reach the supreme court, and he he voted with the majority in two 3-2 decisions that reversed the judgment against Massey. The plaintiffs then argued to the Supreme Court that Benjamin's refusal to recuse himself violated the Due Process clause.

Justice Kennedy, and a majority of his colleagues, agreed. The standard they adopt appears to be grounded in psychology: citing Withrow v Larkin, they state that a conflict of interest requires recusal if "under a realistic appraisal of psychological tendencies and human weakness, the interest poses such a risk of acutal bias or prejudgment that the practice must be forbidden if the guaranteeof due process is to be adequately implemented." But Withrow cites no scientific support for its "realistic appraisal," nor do any of the cases it cites to bolster its "pyschological" approach. Nor does Kennedy's majority.* Not a promising start.

In fact, at the heart of Kennedy's argument is a bald, potentially incorrect empirical claim. Kennedy simply asserts that "Blankenship's campaign contributions--in comparison to the total amount contributed to the campaign, as well as the total amount spent in the election--had a significant and disproportionate influence on the electoral outcome. And the risk that Blankenship's influence engendered actual bias is sufficiently substantial" to require recusal.

Really? First, as is always the case in lobbying expenditures, causality is tough to disentangle. Blankenship didn't just randomly choose Benjamin, but likely selected him because of his political views. So did the money cause Benjamin to vote how he voted, or did Benjamin's proclivity to vote a certain way cause the money to flow to him? The former is troubling, the latter is just the nature of judicial politics. Second, the incumbent wasn't without campaign money of his own, he made numerous missteps during the election, and only one of West Virginia's major papers endorsed him over Benjamin. And third, the marginal effect of campaign spending is ambiguous, but it appears to exhibits steep decline after a certain point; it is thus unclear whether $3 million is all that much more effective than, say, $500,000. Kennedy says there was a "serious, objective risk" that Blankenship effectively chose his own judge via his campaign contributions, but he marshals no real evidence to support this claim, and it is not a clear-cut assertion like "parachutes prevent deaths."

Furthermore, how do we know when we cross the "sufficiently substantial" risk of bias line? First, as a normative matter, what is the number that goes with "substantial"? But second, and more empirically, how do we estimate the effect of money on that risk? Kennedy cites no psychological studies of any sort about how people's decision-making processes are consciously or subconsciously altered by, say, the way other people have supported them. Perhaps this isn't Kennedy's fault: it could be that such evidence simply doesn't exist (see p. 565 here, n. 17 in particular--and note that the Rotunda piece was published in 2003, so it was available to the Caperton Court). But the fact that the evidence doesn't exist doesn't mean that Kennedy is simply free to state that the bias is present.

That is not to say that Kennedy's opinion is necessarily wrong. But it is unacceptable to say "if condition X holds, recusal is necessary" and to then assert that X holds when X is a profoundly difficult empirical question. If Lee Epstein is right that there is little evidence on this topic, then Kennedy's opinion must reflect that. An empirically honest opinion would be framed along lines such as this: 

 

1. Campaign contributions run the risk of influencing candidates in ways we find unacceptable. [This is clearly true--that risk is present, even if we cannot quantify its size.]
2. In this particular case, we lack evidence that the expenditures explicitly induced Benjamin to vote a certain way
3. We also lack more generalizable evidence that expenditures of this sort induce recipients to change their minds at least x% of the time [where x is the point of "substantial" influence].
4. In this case, we cannot take absence of evidence to necessarily mean evidence of absence.
5. Therefore.... 

Well, that last step is the tricky part. The court could either say that in the absence of evidence judges are presumed to be unbiased or that they must recuse themselves. But in either case the court must make it clear that it is not making an empirical claim about judicial behavior but just a naked--and ultimately rebuttable--legal presumption required to reach a decision that day. 

Such an approach could** yield several benefits. If nothing else, it would clearly be much more transparent. But it could also ensure that the Court's opinions respond more quickly to developments in empirical evidence. It could also force the justices to try to gather more rigorous empirical evidence, and where such evidence clearly doesn't exist to think more carefully about how to make error-cost tradeoffs in the presence of genuine uncertainty. As I said in my first post, one of my primary goals with this project is to see if a case-by-case survey of the Court's use of empirical evidence might shed some useful light on exactly what could work and what it could do. 



* It appears that the briefs do try to cite some evidence. But to the extent that this evidence shaped the Court's opinion, it should be cited in the opinion, if only to (1) make transparent how the justices are reaching their conclusions and (2) allow other researchers to check the validity of the cases on which the Court relies.

** Yes, yes, I know. I lack empirical evidence to support my own claims, and I am using that ultimate weasel-word, "could," that I have and will criticize the Court for using as well. And while part of me wants to argue "what could be the downside of the Court being more honest about how it reaches its conclusions?!" I realize that the issue isn't that easy. The Court lacks an enforcement arm, so its opinions are followed because people feel somehow compelled to do so. If the Court started saying "we're just guessing here," perhaps it would lose some of its moral suasion. Of course, perhaps we should be troubled that the Court is effective only because it lies at some level, but that is an issue for another day.