Bankruptcy Reform & the Force of Selection

ark.jpgMichael Heise highlights a very vigorous debate in the American Bankruptcy Law Journal on the empirical study of bankruptcy reform.

  • It started with a paper by Bob Lawless, Angela K. Littwin, Katherine M. Porter, John Pottow, Deborah Thorne, and Elizabeth Warren (hereinafter the “Six”), Did Bankruptcy Reform Fail? An Empirical Study of Consumer Debtors, 83 Am. Bankr. L. J. 27 (2009). The authors of that paper found that there was “no change in the income levels of bankruptcy filers after the [bankruptcy reform] amendments [and that] debtors filing for bankruptcy in 2007 have even greater debt loads than their counterparts from 2001.”
  • Rafael Pardo responded in Failing to Answer Whether Bankruptcy Reform Failed: A Critique of the First Report from the 2007 Consumer Bankruptcy Project , 83 Am. Bankr. L. J. 47, in which he made several critiques of the Six’s method and lack of nuance regarding the Code.
  • Pardo’s critique seems to have pushed the Six to a very defensive response (not yet on SSRN, but which you can find on WL), called “Interpreting Data: A Response to Professor Pardo”
  • Pardo has now responded.

You should read the papers, if you are interested in empirical analysis of bankruptcy, or if you just relish a well-articulated methods discussion. I’ll leave the merits largely alone, with one exception. The papers appear to disagree about a relatively fundamental question of predicted selection effects.

Selection effects operate at various stages of litigation to remove certain cases via settlement or unilateral withdraw from the fight. As is well known, these “missing” parties and claims make it quite difficult to analyze the effects of changes in law on datasets of outcomes of litigation. (As an illustration, imagine that Noah had a policy that denied snakes entry into the ark (as some posit). Rational snakes, knowing about the policy, wouldn’t approach the choosing point, and, therefore, we wouldn’t see the policy’s effects in action.) A law might make it easier for plaintiffs to win, but the resulting set of cases will be unaffected because more defendants settle both before and after filing. Selection turns out to play a very important role in this bankruptcy dispute.

In the first paper, the Six compared characteristics of two population of litigants: filers from 2001 and filers from 2007. (That is, filers operating before and after BAPCPA, the Bankruptcy Reform Act of 2005). The research question that the Six asked was whether BAPCPA had a deterrent effect on “high-income abusers” of the bankruptcy system, and they made inferences about a missing “800,000 filers” in 2007 (who would have filed but for BAPCPA’s new means test) to perform an analysis. After finding that income didn’t differ between populations, they conclude that their dataset failed to show that BAPCPA achieved its purported effect.

Obviously, for this assumption to work, the missing (deterred) filers in 2007 plus the remainder of 2007 filers must look like the undeterred 2001 population. The papers appear to agree that the operative question is how would the means test affect the population of filers. The Six rightly point out that it is “virtually impossible to gather data on deterred consumers who did not actually file bankruptcy.” So, who are they?

The Six are focused on the “upstream deterrent effects” of BAPCPA, not its “intra bankruptcy operative effects.” In other words, they believe that people who know they will flunk the means test will not file at all. Like a “Litigation Masculinization Reform Act”, requiring complaints by women to be dismissed, the means test operates to generate a sharp selection. But, Pardo replies, the means test contains “inherent uncertainty and ambiguity”: many attorneys and their clients “will file for Chapter 7 relief with the good faith belief that they do not run afoul of the means test but may nonetheless end up facing a dismissal motion due to a different expectation of outcome by the moving party.” That is: Pardo believes that selection will operate (at least partially) after filing; the Six believe it will operate entirely before filing. More succinctly: Pardo posits selection effects that are, as is traditional, largely ex post; the Six offer a theory of complete ex ante selection.

The Six’s ex ante position is too strong. In my work on veil piercing, I have found that plaintiffs routinely file cases that have essentially zero chance of success: they try to pierce the veil of public corporations; they advance plainly frivolous consolidation motions; they attack the dominion and control of LLC members. Why? Sometimes, they are pro se. Other times, they have bad lawyers. And sometimes, even good lawyers misunderstand complex law, perhaps because clients offer bad information, or because lawyers are subject to cognitive biases like optimism too. And even very sophisticated parties will take litigation positions with small chances of success. Consider defense counsel in federal criminal trials. Given the government’s success rate at trial, why don’t such cases select out of the system? In part, I think it’s a function of demographics, but also I think that there may be a bit of a lottery effect. Parties facing bad outcomes may be risk seeking, even given terrible odds. Moreover, parties that are one-shot players are more likely to file and proceed further in litigation than parties who are repeat and risk-neutral. Thus, I don’t really understand the Six’s confidence in their total ex ante selection story, though I grant there are many things about how the consumer bankruptcy system operates which I don’t understand. Perhaps Pardo is wrong to claim that the means test is ambiguous, and instead it’s more like gender. But the Six don’t make that argument, or if they do, it’s not at all clear. This all suggests that at least on this one point, Pardo’s critique has teeth.

The Six obviously disagree. They suggest that instead of “offering useful ideas of how to explore the available empirical data or build new data sets, he impugns our methodology, our logical assumptions, and our very understanding of BAPCPA’s means test . . . We confess that we would rather have been analyzing more of our data and expanding the collective knowledge about post-BAPCPA debtors . . . ” I too must confess: that’s not the tone of response to a comment on method that I would have expected from this noted group of empirical scholars. It sounds much more like an advocate’s brief. Or, worse, the death penalty debate. I certainly don’t think there is an affirmative obligation to be nice when responding to critics, but I do think that there is a norm of being charitable to others’ arguments, attempting to relay them in the clearest possible light, so that readers can be helped to understand the merits. (This view, incidentally, was forcefully advocated by Dan Markel some time back. I disagreed then, but I’ve been convinced over time, and I think the norm of charity is particularly important in technical pieces, when readers don’t have easy access to the underlying data.) The Six’s response failed that standard.

But that comment aside, the merits of this dispute are obviously of significant interest to the consumer bankruptcy community, and the methods will be interesting to empiricists and lawyers alike. I encourage you to check it out and make up your own mind as to whose story you find persuasive.

[Note: practicing what I preach, I showed this post in a draft form to both Pardo and to Katie Porter, one of the six. Nonetheless, all errors are mine.]

(image Source: Wikicommons, Noah’s Ark by Edward Hicks. Which animals selected out of the Ark? Apparently, none.)

You may also like...

3 Responses

  1. Nate says:

    Excellent post. Thanks.

  2. Clinton says:

    You are a very smart person! I never would have thought about it this way.