Marginal Utility

by Tom Bozzo

Earlier today, I threatened to take on a post at Mahalanobis by HedgeFundGuy, "Bankrupt Survey Data," which purports to debunk a recent survey of personal bankruptcies*, cited in Paul Krugman's Tuesday New York Times column among other places. Regular commenter Ken Houghton egged me on, so here goes.

1. HedgeFundGuy suggests, "Survey data are always very suspect, which is why economists generally do not use them." This is simply ludicrous. It took me about two seconds to draw up a short list of survey-based data widely, if not universally, used by U.S. (macro)economists:

• Payroll employment, based on the Current Employment Statistics survey
• National unemployment rate, based on the Current Population Survey
• Consumer price and producer price inflation, based on surveys
  
• Personal consumption expenditures and other GDP components, go ahead and see how many surveys are mentioned in the NIPA methodology summary (PDF) — it's Several.
  

BTW, some of the above survey data are "tendentious." For instance, the CPS determines self-employment through self-reporting by survey respondents, and under certain circumstances actually being paid is not necessary to be considered "employed" for CPS analysis purposes.


2. HedgeFundGuy's argument regarding the bankruptcy study is... lacking. Here is a summary:

• Survey data are always suspect.
• Different surveys of bankruptcy causes yield different results on the importance of medical causes. Those surveys! (HedgeFundGuy does not establish that the surveys in question were defining medical causes of bankruptcy identically.)
• HedgeFundGuy doesn't believe the results of other surveys measuring a completely different thing (hunger and 'food insecurity'). This occupies about 1/3 of the post.
  
• Therefore, it follows that something is wrong when the bankruptcy study shows a limited causal role for "personal vices" in bankruptcy filings. HedgeFundGuy complains that "vices" that have DSM classifications are considered medical contributors to bankruptcy. (Substance and gambling addictions are cited by 2.5% and 1.2% of survey participants, respectively.) HedgeFundGuy suspects self-reporting bias.
• HedgeFundGuy faults the authors for failing to cite an industry study and a 1999 Journal of Finance paper, and for citing a co-author's relatively well-known book. HedgeFundGuy offers no credit for citing a more recent American Economic Review paper that discusses determinants of personal bankruptcy filings. (The AER paper's data were derived from — wait for it — the Panel Study of Income Dynamics, another of those pesky surveys!)
• HedgeFundGuy complains that the study, which is clearly not intended for the purpose, fails to address causes of the increase in bankruptcy filings during the 1990s.
• HedgeFundGuy goes off the reservation in suggesting that credit card lending practices might actually contribute to the increase in personal bankruptcy filings. (Hat tip: Ken.)
  
• HedgeFundGuy suggests that doctors always do bad sociological research. Maybe, but the listed coauthors are two doctors, a lawprof, and a sociology prof.

I've previously expressed sympathy with the argument that surveys results in medical journals (not to mention most statistical results published in social science journals) should be better documented. However, the Health Affairs paper is not, properly construed, making an extraordinary claim requiring extraordinary justification.

In short, far from tearing the study up, as Alex Tabarrok suggests, HedgeFundGuy has offered nearly nothing to actually discredit the study. Five demerits for the Marginal Revolutionaries: they won't make the blogroll this quarter.

HedgeFundGuy might have made some hay out of the AER paper (cite) had he been so inclined. That paper, which I skimmed for the purposes of this post, provides mixed evidence for exogenous events as drivers of bankruptcy filings. The authors suggest that filers' strategic behavior is of primary importance, though their analysis also appears to show that factors such as divorce and income shocks have very large effects on the marginal probability of bankruptcy filing. (They also offer repeated caveats about the difficulty of identifying behavioral hypotheses from their analysis.) It should be common knowledge that health shocks can readily cause large adverse income shocks.

---------------

Todd Zywicki is a little more infomative at Volokh Conspiracy, but still relies too much on the Straw Man. For instance, he complains:

Nor does the requirement of $1,000 in unpaid medical bills within 2 years of bankruptcy seem like a very plausible measure of serious financial problems.

The abstract of the Health Affairs paper states that

Among those whose illnesses led to bankruptcy, out-of-pocket costs averaged $11,854 since the start of illness... [emphasis added]

and, Earth to GMU, if you're not rich, nearly $12,000 in out-of-pocket medical expenses is a lot of money.

But it's getting late. I'll just add a parting shot that Zywicki has a law review article making, among other things, an Olsonian collective-action case for why the role of creditors in pushing bankruptcy reform might be overstated. I don't think Zywicki's argument holds water, but I'll save that one for another day.

-------------

* "MarketWatch: Illness And Injury As Contributors To Bankruptcy," by David U. Himmelstein, Elizabeth Warren, Deborah Thorne, and Steffie Woolhandler.