Friday, June 05, 2009

Elizabeth Warren's Bankrupt Study

Megan McCardle does a nice job of noting that TARP Overseer Elizabeth Warren's latest addition to Harvard level research screams advocacy. Would such lame methodological rigor be acceptable by a straight white male studying clean coal at Harvard? Warren and her colleagues (spokesmen for Physicians for a National Health Program--no axe to grind there) note the increase in the percentage of bankruptcies attributed to medical expenses, while neglecting the overall number of bankruptcies (down 50%).

Her process is to ask people who went into bankruptcy to what degree it wasn't their fault. They asked a select group of 2314 bankrupt people motivated to respond to their 5251 randomly selected debtors. These responders had 'the same' assets and liabilities as non-responders, implying to the authors no selection bias, ignoring the different propensity of such responders to confabulate tendentiously. They looked for the following, all of which count as 'causing' their bankruptcy:

• Specifically identified medical problem of the debtor or spouse or another family member as a reason for filing bankruptcy.
• Specifically said medical bills were a reason for bankruptcy.
• Lost two or more weeks of wages because of lost time from work to deal with a medical problem for themselves or a family member.
• Mortgaged their homes to pay medical bills.
• Spent more than $5,000 or 10% of annual household income in out-of-pocket medical bills

Hey, if they say so! Why would they lie? I have seen people fail in business many times. I don't remember anyone saying: it was my fault. They always have an excuse, this is human nature. Perhaps, like overconfidence, a benign bias that would otherwise make us too depressed. The reason a company or project fails is never a bad strategy or management, but rather, someone else screwed them over, or some act of God.

Bankruptcy is still somewhat shameful, not so much for reneging on obligations but rather admitting one is currently poor. I can't believe a study adressing some personal failure is done in this way. I presume most criminality is due to police brutality and biased judges and prosecutors under this approach.

But if saying 10% of your income went to medical expenses caused bankruptcy, and America spends 15% of its GDP on healthcare, I'm surprised it is not 100% caused by healthcare. Of course using this definition, it is also probably 100% caused by food. And housing, another 100% cause. What major sector of expenditures doesn't cause bankruptcy?

I would think a convincing study would be look at the medical debts outstanding at bankruptcy and compare them to debt outstanding. As the authors note, medical debt sometimes go to credit cards or collection agencies, but that's the kind of dirty empirical work that should be de rigueur for America's premier university, and there are ways to get at the problem. To merely ask pathetic people why they are pathetic is so manifestly biased, it generates very uninteresting results. A neat senior thesis perhaps, for a high schooler, but such are the standards at the American Journal of Medicine. The fact that they used a logistic regression and Chi-squared at 2-tailed t tests does not make this rigorous (it's a standard procedure in any econometric software).

Harvard. These are the people who don't know what causes seasons (it's the tilt, not the distance). Our best and brightest.


Anonymous said...

You are on fire lately Falken! Keep up the good work!

Anonymous said...

This is a giant turd sandwich here. Just one visit to the emergency room can cost someone 10% of their income ... even if there's nothing wrong with them. If a family needs two incomes to pay the mortage on their $150,000 home, and one of them falls sick, even with health insurance they are SOL.

Anonymous said...

the statistics observation in your first paragraph is exactly what i was thinking when i was paging through malcom gladwell's last book (i just had to see what he was on about). he started going on and on about some statistics for some isolated variable without ever mentioning the overall totals.
it's troubling that the general reading public probably never asks
this question themselves.
it is so basic. and yet this guy just glossed over it like no one would care about the overall totals. i had to put the book down.
this person you target here. she has a J.D. but does she have an undergraduate science degree? if not, i'd bet that's the problem. without ever having to do this grudge work, she probably had to get someone to help her do the minimum statistical analysis that Am J Med requires. and i would doubt she would understand it. heck, many faculty do no even understand it. they just do it because it's required.

it's neither here nor there: the data is garbage, like you said. questionnaires are not the best way to figure out why people do what they do, and anyone in tune with current methods should know this. it's cheap though. and so it will never disppear.