Prediction is Hard, Especially About the Future

Ezra Klein says that he has made testable predictions about the future of health care:  to wit, that in twenty years we will have peer reviewed research showing that health care reform has saved tens of thousands of lives.  I didn't mean to single out Ezra in particular; he is but one of many enthusiastic pundit advocates for this bill from whom I would like to see some criteria for judging success or failure.  But I'm glad he's stepping up with some numbers.

That said, while this is certainly a prediction, it's not a very strong one.  I too, believe that by 2030 we will have at least one peer reviewed research paper showing that health care reform has saved many lives.  There will probably also be at least one peer reviewed research paper showing no effect.  

Given the magnitude of the claims about the uninsured, however, I don't think we need to wait for 2030, or peer review.  If you believe that 45,000 people in this country die from lack of insurance every year--a figure that Ezra, among many, many other commentators, has treated seriously--then conservatively, by 2030, we should have something like 30,000 fewer lives lost every year in the 18-64 age group.  

Yes, I know--you want to tell me that being long-term uninsured may take a toll years later!  But of the panel surveys from which the relative risk estimates are usually taken, the longest follow-up was 16 years.  By 2030, we should be seeing the full effect.  We should be seeing noticeable impacts much faster, particularly if--as many people have argued--insurance status has a huge and measurable impact on trauma outcomes.

That's a 5% drop in mortality--a huge drop against the background rates.  Even if you only think that the correct number of the uninsured is 20,000 a year, you should believe that there will be a drop that is, so to speak, visible to the naked eye.  

If you think that the number will be much, much smaller than that--so small that only tens of thousands of lives will be saved in fifteen years--then it seems to me that you're saying that those estimates were radically off, by a factor of five or more.  Given just how prominent those numbers were in the debate, that's kind of a problem.  It's an even bigger problem because a few thousand deaths a year will not really be distinguishable from statistical noise.

But the biggest problem is how much we'd then be spending per year to get this added benefit. I think it's entirely plausible that we'll be saving 3,000 people a year.  But 3,000 people a year, at a cost of $200 billion, is almost $70 million per life saved.

Of course, there are supposed to be other benefits, like preventing bankruptcy.  But let's say that Himmelstein et. al. are right, and there are about 500,000 "medical bankruptcies" in the United States.  Let us further assume, very implausibly, that these bankruptcies are 100% caused by medical bills. 

Well, the average unsecured debt discharged by debtors in bankruptcy is something along the lines of $50,000. (Yes, this number is deliberately vague, because the only studies I know of were done in the late 1990s--but these numbers do not usually take dramatic fourfold leaps over the course of a decade).  

For $25 billion a year, we could solve the entire problem by just paying off their debts.  Yes, in reality, there would be a moral hazard problem.  I'm just pointing out the relative size of the problem--and in reality, the medical bill problem is almost certainly far smaller than that; more medical bankruptcies appear to be driven by lost incomes than by bill problems.  You do not spend $200 billion to save $12 billion or so for our nation's debtors--and that's really a pretty generous estimate of how big a problem this is.

Thus we quickly end up in immeasurables like morbidity.  In fact, we should be able to measure at least some of this, because a lot of conditions with high morbidity also have high mortality--diabetes springs most readily to mind.    But a lot of morbidity is hard to measure, so if you're a program's advocate, your belief in its efficacy is basically non-falsifiable.

The bottom line is that what I think the public is expecting from the bill is something much more dramatic than a decrease in mortality that won't be large enough to make any noticeable impact on things like US life expectancy or death rates.  And frankly, I think that they are expecting this because a whole lot of commentators--and this is not about Ezra in particular--made it seem as if this was something reasonable to expect.

Does that mean I think the bill will be repealed?  No.  I think entitlements are very hard to repeal or reform.  Ezra regards this as a vindication of the entitlements.  I regard this as a public choice problem that may well push us into a nasty fiscal crisis.  Both may be true, which won't make us any less hosed when the crisis comes.

Ezra's other predictions are not about benefits; they're mostly political.  We are radically in disagreement about the fiscal future, possibly because I regard any future moves to repeal significant revenue and cost control measures as a fairly predictable risk of the bill, while Ezra probably thinks of them as separate events that do more to signal Republican malfeasance than the political sustainability of reforms.

But interestingly, I think we're actually in agreement on the core predictions:  there will be few measurable benefits of this bill, and certainly not any large enough to justify the gross cost of the coverage provisions.  Unless the cost-control provisions are sustained in the face of considerable political pressure, and also do not result in any adverse impact on the elderly, belief in this bill's success or failure rests on outcomes that we won't be able to measure in any reliable way on a national scale.  So it's probably not really falsifiable, especially since some of the benefits that progressives see--like civic togetherness--aren't things that people like me put a high value on, unless that togetherness results in some more tangible improvement in peoples' lives.

But Ezra may vehemently disagree with my interpretation of our areas of agreement--in which case, I fearlessly predict (99.5% certainty) that he will  correct me.