Megan McArdle and the Confusing, Logically Dubious, Point-Missing Critique of Elizabeth Warren

Megan McArdle has a pair of posts critiquing the latest Elizabeth Warren (et. al.) bankruptcy study.

Here’s the crux of the critique:

Warren reports that the proportion of bankruptcies attributable to medical causes (medical bankruptcies) has increased.
However, the absolute number of bankruptcies has declined.
And so has the absolute number of medical bankruptcies.
therefore,
Elizabeth Warren is a DIRTY COMMIE LIAR!!!!111!!111!!!111!!

It may surprise you that I’m a little alarmed by what’s going on in that therefore. Here’s McArdle’s account of the therefore:

Yes, but why do they only talk about the proportion? In general, economics papers talk about absolute numbers whenever they can, and use proportions only when things like changes in income and inflation make comparisons between years too difficult. I submit that we want to know, not whether medical bankruptcies are a bigger or smaller proportion of overall bankruptcies, but whether more people are being pushed into bankruptcy by their medical bills. To take the extreme absurd case, if only one person had declared bankruptcy in 2007, but that one person had had huge medical bills, would this be a sign that we need national health care?

But that’s not right. What we really care about is not whether “more people” in absolute terms “are being pushed into bankruptcy by their medical bills.” What we care about is whether the problem of people being financially crippled by medical bills is getting better or worse. The two questions are not the same, and while the first is relevant to the second, it isn’t determinative.

True, ordinarily, an absolute decline in the number of medical bankruptcies would indicate that the problem is getting better, or at least not getting worse. But this isn’t an ordinary situation. As both McArdle and Warren point out, the absolute decline in the number of bankruptcies in general — chopping them in half! — was likely caused by the bankruptcy “reform” legislation enacted in 2005. But that overwhelming exogenous cause makes it impossible to relate a change in the absolute number of medical bankruptcies to anything about the way in which medical bills are affecting people’s finances.

What’s left? The proportion. Surely it’s significant that bankruptcies in general fell by half between 2001 and 2007 while medical bankruptcies only fell by a third — that is, that medical bankruptcies became a higher proportion of bankruptcies. That suggests that there’s something going on — that some variable is causing medical bankruptcies to be more robust against the bankruptcy “reform” bill than other kinds of bankruptcies.

A competent social scientist, looking at that result, would think that there are two possibilities:

a) the effect of the bankruptcy legislation (the only known independent variable for the absolute drop) was different for medical bankruptcies versus other kinds of bankruptcies; or

b) some other independent variable that contributes to medical bankruptcies but not to other bankruptcies was increasing.

a), of course, is the conservative/libertarian story “the bankruptcy bill took out all those nasty fraudsters, and there are fewer fraudsters with real serious problems like medical bills, so of course the proportion of honest medical bankruptcies would increase.” b) is the liberal story “the impact of health care costs is getting worse and worse.”

Picking between a) and b) is difficult, but the point is that for either of them, the thing we’re concerned about — the evidence that prompts the question that we need to answer by choosing between a) and b) — is the changing proportion of medical bankruptcies.

“But,” McArdle might say here, “failing to highlight the fact that we’re comparing declining quantities unreasonably biases the question in favor of b), the liberal interpretation, since a) is only consistent with the real situation — everything declining, but medical bankruptcies declining less — while b) is consistent with both the real situation and a fake situation in which medical bankruptcies actually increase. By implying the fake situation, Warren rigged the game in favor of interpretation b).”

There are two problems with that objection.

First, Warren hardly concealed the decline in the absolute number of bankruptcies. Indeed, the study specifically discusses the effects of the new law (see page 5 of the pdf), and specifically considers and argues against interpretation a).

BAPCPA’s [The bankruptcy "reform" legislation's] effects appear nonselective. 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. It is implausible to ascribe the growing predominance of medical causes of bankruptcy to BAPCPA.

Conversely, there is ample evidence that the financial burden of illness is increasing. The number of under-insured increased from 15.6 million in 2003 to 25.2 million in 2007. Of low- and middle-income households with credit card balances, 29% use credit card borrowing to pay off medical expenses over time.8 Collection agencies contacted 37.2 million Americans about medical bills in 2003. Between 2005 and 2007, the proportion of nonelderly adults reporting medical debts or problems paying medical bills rose from 34% to 41% …

(several footnotes removed)

Yes, that’s right, Warren et. al. acknowledged the difference between interpretations a) and b) and offered evidence for the one they believe. You know, just like good social scientists are supposed to do.

Second, the difference between a) and b) isn’t even a conservative/libertarian vs liberal thing. I tricked you by describing a) as a conservative/libertarian interpretation. In fact, a) is perfectly consistent with medical bills being really bad, because it suggests that the real proportion of bankrupts — that is, the people who were genuinely struggling with their bills — attributable to medical causes was even higher in 2001, once the nasty fraudsters are dropped out. a) is actually more liberal than b), because it suggests that the problem with medical bills is a long-term serious problem.

Share


4 Responses to “Megan McArdle and the Confusing, Logically Dubious, Point-Missing Critique of Elizabeth Warren”

  1. Stephen Bank Says:

    I don’t really have anything to say about this. But it was a good reminder that your blog is worth writing. After all, I find it very much worth reading!

  2. Daniel S. Goldberg Says:

    Lovely. While it is possible, it seems to dubious to this observer — who, you know, kind of thinks about this sort of garbage for a living — that the hyperinflationary growth trajectories of BOTH health care costs and (esp. chronic) disease prevalence have no meaningful relationship in context of bankruptcies.

    To make matters worse, however, is the well-known fact that demand (i.e., costs driven by illness burdens) is far less a driver of health care costs than supply, and more so, that supply is rarely positively correlated with outcomes above a very low level in the developed world. Worse yet, supply is negatively correlated with outcomes in many geographic regions and communities, which implies that rising numbers of people are undergoing medical bankruptcies for care which frequently may be doing them no good and which may well be causing them harm.

    Nice society we’ve got here.

  3. howabout Says:

    how about all the critiques which are up on her blog including some from commentators?

    i think the percentage/raw number critique was fairly flimsy but that study does seem to be riddled with issues and problems—frankly i’m surprised it got through peer review

  4. Paul Gowder Says:

    Howabout, those critiques are more fair. (And, unsurprisingly, the worst accusation they can levy is sloppiness, rather than dishonesty.) I’m not sure they’re conclusive, but they’re worth thinking about.

    (Why am I not sure they’re conclusive? Because there’s a balance to be made between dotting and crossing one’s methodological i-s and t-s and being able to actually study the phenomenon you want to study.)

Leave a Comment