A Little More About Medical Bankruptcy

I want to talk a bit more about why I dislike the Himmelstein et. al. study that found more than half of all bankruptcies were due to medical reasons.  There are a few reasons that I don't find their work very convincing.  First, the rate of respondants attributing the bankruptcy to medical problems was more in line with other studies I've seen, at 30-40%.  The extra folks come from their addition of, for example, people who had medical bills that total 5% of income.  Now, having medical problems is one of the most socially acceptable reasons for bankruptcy (compared to other major causes like divorce, overspending, or a gambling or drug addiction, so we'd normally expect people to overemphasize the medical problems compared to other factors.

They also had a pretty low response rate, which isn't exactly surprising, but of course, when you've got a response rate well under 50%, you have to worry quite a bit about selection bias.

But the biggest problem is the timing.  They conducted the study over a period that roughly corresponds with the first quarter of 2007.  This probably doesn't mean anything to most of you, but in 2005, Congress enacted a bankruptcy reform that went into effect in late autumn.  Partly due to the press hype, which exaggerated how hard it was going to be to file bankruptcy in the future, people freaked out.  Anyone who even thought they might file bankruptcy, rushed to do so before the law took effect.  The result was a giant spike, and then an even more giant drop.

The paper sort of deals with it, but all they say is that there's no difference between the people who filed in 2007, and previous filers.  I have no idea how they can say this, since this is actually the subject of their study--if we had really good data on the average percentage of medical bankruptcies, they wouldn't be studying the question.

Here's their explanation:

BAPCPA's effects appear nonselective(7).  Current filers differ from past ones mainly in having struggled longer with their debts.  New restrictions fall equally on medical and nonmedical bankruptcies, with no preferences for medical debts or sick debtors.

But that's not quite what the paper they cite says; it only studies income and assets, not other characteristics of the debtors.  And there are a number of reasons that the law might have favored sick debtors:  sick debtors might have found it easier to meet the means test; the new law tinkered with the rules surrounding secured debt like auto payments, which made it relatively more attractive for people with unsecured debts like medical bills to file; and people with medical bills might find it harder to time their bankruptcies.

Now, maybe none of these things did affect the sample.  But I don't think you can look at the first quarter of 2007 and say this is a period from which you want to draw broad generalizations about bankruptcies:
You're still in an atypical period after a huge shock to the system.  Now, maybe there was no difference between those who rushed to file, and those who didn't.  But I wouldn't want to bet on that.  And I can tell a plausible story where the people with medical problems can't or won't engage in the kind of strategic behavior that lets people either predict, or stave off bankruptcies . . . which makes me wonder why they chose the period they did.  Everyone I interviewed at the time made it very clear that they thought we were in an atypical period . . . this was the pro-debtor story, because if bankruptcy rates stayed low, it meant that there had indeed been a fair number of strategic bankruptcies, as the banker's associations had been arguing.

They didn't stay low, as you can see; the data runs only to June 2009, but there's little doubt they're still mounting.

Now look at the "clinical findings" section of their paper:

  • 62% of bankruptcies have medical causes
  • Most filers are well-educated and middle class; 3/4 have health insurance
  • The number of medical bankruptcies has increased 50% since 2007

Their press releases and the interviews made it even clearer that they were making a case for national health care, based on the fact that the current system was pushing everything into bankruptcy.

Now, given the quality of the data they had, this is way, way, way too strong a statement of what they found.  This was also true of their earlier study, which defined everyone with medical bills over $1,000 as having a "medical bankruptcy".  Somehow, the parameters they choose always give rise to the conclusion that the medical system is wreaking catastrophe on the maximum number of people.

I find it very hard to trust the studies they do because of this.  In general, I'm really skeptical of papers from activist groups, which is why you rarely see me do more than pluck a very basic data set based on government or similar figures from any advocacy group or think tank that agrees with my ideological priors.  So the way this paper was pushed by PNHP (Physicians for a National Health Plan, of which two of the authors were founders) was already enough to make me worry about it.  The choice of years and methodology was another.  The highlighting of their most eye-popping claims with few caveats is a third.

Yes, they disagree with "my side" (though I was against BAPCPA, and assume there are a significant number of medical bankruptcies).  But I don't think you'll find me citing many papers like this from my own side--and if you do, I'm wrong, and you should call me on it.  I just don't trust this kind of work.