Last week, Professor Paige Marta Skiba of Vanderbilt Law School presented her paper, titled, “Race, Gender, and Political Ideology in Personal Bankruptcy Outcomes,” to the Law and Economics Workshop here at Chicago (the most recent draft is available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1548473).
What It's About
The study collected 9,526 bankruptcy petitions across the United States, comparing dismissal rates based on gender, race, and the probable political allegiances of the judge and county. Although the dismissal rate differences between male and female petitioners were negligible, the study indicated that African American petitioners had their petitions dismissed more than White petitioners (24% compared with 7%), with this general trend holding true across all judge demographic types. Yet, African American petitioners’ repayment plans, once approved, did not require Black petitioners to pay significantly more than their White counterparts. Results also suggested that gender and political affiliation of judges may affect petitioners’ Chapter 13 approval rates (White male Democrats in a Republican county and White female Republicans in a Republican county tend to dismiss at a higher rate). The study also investigated the effects of race and gender on decisions to file under Chapter 7 versus Chapter 13, the amount of attorney fees charged, but most of the discussion in the workshop centered on the issue of dismissal.
What Was Discussed
From the outset, the workshop participants seemed most interested in discussing methodology, with critiques being about evenly divided between 1. concerns about the accuracy of the measurements and 2. the problem of omitted variables:
On the accuracy issue, several participants raised concerns about the way the study determined the race of petitioners. Since petitions do not contain information about race, the study had extrapolated probabilities of race based on zip codes and surnames (matching surnames to census data, and only using those names where probability of race membership passed a certain threshold). One concern was that this method of determining race, by using zip codes, could inadvertently introduce confounding variables based on neighborhood locations (which may relate to real estate values). The study also collected data on race and gender of bankruptcy judges by looking at pictures available to the public.
The point was made that, at least for judges, race data could be readily available by contacting the appropriate person in each circuit. The presenter accepted the suggestion, although it seemed like a minor critique, since realistically the two methods were unlikely to create a big difference in error rate. The more interesting point was the issue of coding the politics of the judge. The study classified judges as Democrat or Republican based on the affiliation of the person who appointed the judge, and the study classified county political leanings based on the location of the courthouse (which is usually urban) rather than the political makeup of the judicial districts where the petitions were filed (which may be rural or have different political sympathies).
Then, on a more general note, several commentators raised the possibility that the sample of petitions in the data set needed to be more randomized, since a disproportionate number of petitions seemed to be from certain states, even when taking population into account (otherwise, the discrepancies may be skewed by regional differences).
The more interesting debate (at least to me) came in the discussion about omitted variables. Although the study controlled for some of the non-racial factors, such as asset value, wages, and number of creditors, several workshop participants suggested that there are other non-racial factors that could account for the discrepancies in dismissal rates. For example, living in a predominantly Black neighborhood may affect the rate of increase in home value, which may legitimately affect a judge’s assessment of the petitioner’s repayment plan. (The depressed home value increase rate may reflect racial prejudices from society-at-large, but one could argue that these larger social inequities aren’t what bankruptcy law is structured to address—affirmative action hasn’t really caught on in bankruptcy.) There may also be racially correlated differences in why people file bankruptcies. One commentator suggested that in order to isolate the discrepancies that the judge is directly responsible for, the study would need to filter out voluntary dismissals (e.g., when a petitioner filed simply to temporarily stay creditors), but as it turns out, the study already filtered such results out of the sample.
Now, at this point, one might stop to ask, why the fuss over methodology? Well, for one, a lot is at stake. If the paper successfully proves what it sets out to prove—that bankruptcy judges are using their discretion to deny debt relief to Blacks because of their race—the implications are huge. At the least, it gives us fuel for our cynicism. It may imply that judges can’t be trusted to be fair, and if that is the case, then perhaps the legislature ought to move to reduce discretion. And it gives credibility to those who argue that there is something sinister about legal system as a whole, where even facially neutral law has been deliberately designed with gaps that permit groups in power to maintain the racial status quo. Maybe a little resistance is understandable when the possible disillusionment is high; we want to make sure this thing is for real.
But, on a more practical level, as well, methodology matters: it tells us whether we need to be concerned about the discrepancy after all, and if so, at what bottleneck in the process of justice we need to apply the remedy. Suppose, as a few commentators in the workshop suggested, any discrepancy that we see in bankruptcy petition dismissal rates isn’t really about judges’ perceptions of race, but about actual cultural differences in consumption patterns between racial groups. If one race invests more in homes and another in cars, there may be non-racist reasons for higher dismissals: i.e., the way the bankruptcy law defines which assets courts must consider when approving a Chapter 13 repayment plan. If it turns out that the race’s own consumption preferences account for the differences in dismissal rates, then we can expect that public reactions will no longer be so uniformly negative as when the judge’s racial attitudes account for the differences. Some people will respond by saying, “Who cares? People are allowed to invest in cars, homes, or whatever as they prefer, and let them live with the consequences. It’s not the law’s purpose to distribute consequences equitably.” For such people, the discrepancy exists, but it isn’t really a problem for the law. But even for those others who still find the outcome unjust, such people will now know where to direct their efforts at reform: the problem isn’t with judicial discretion, but with the structure of the law itself (perhaps the law should be revised to count assets differently). At what stage is the discretion coming in? Can we be sure that it’s the judge’s bias showing through, or could it be, as one participant suggested, the bias in the trustee’s recommendation, or some other player’s influence?
The study, as it stands, can’t make any recommendations about what to do about the racial discrepancy problem, even if it proves it exists, because it isn’t clear yet what’s causing the discrepancy. I think the general assumption the paper makes is that the discrepancy must come from the judge’s discretion (e.g., because of the judge’s race, gender, political affiliation, or generally racist attitudes) or the lawyer’s strategic choices (e.g., because of the lawyer’s race or the lawyer’s Chapter 7 and 13 experience). But this is making the big assumption that the law itself is neutral; after all, a perfectly unbiased judge, applying a biased law, will still produce biased effects.
It seems to me that one simple way, moving forward, to improve on the clarity of the paper’s important message—that there may be impermissible race factors tainting bankruptcy petition dismissal decisions—is to set some parameters defining what should and shouldn’t be permissible for judges. What is the normative world the paper exists in? Although this is an empirical paper, answering this normative question seems important in determining the right framework of controls in this study to answer a legally meaningful question. Is it permissible for a judge to consider race-correlated factors, even if a judge cannot directly use race as a proxy and let race itself influence decisions? Of course, the answer to this must be yes to some extent, since there are certain systemic inequalities and insofar as race is correlated with financial patterns, the bankruptcy laws will have disproportionate effects among races. Although the paper controls for asset value and other financial factors in a broad sort of way, it may be appropriate to look more specifically at what factors the legal regime builds discretion in for bankruptcy judges. Are there instances where the judge dismissed a petition even though it fit all of the typical legal considerations for approval? Controlling for everything the law says is relevant, do discrepancies still occur (using the consumption differences example, if the law treats different types of assets differently, can the study point to whites and blacks who had chosen to hold a similar mix of assets and yet had different petition results)?
The discrepancy that the paper raises is an important one, but we need to find out what’s causing it—both to prove the discrepancy exists for an impermissible reason and to identify where to target the reform.