Friday, October 25, 2013

When the Value of Information is Negative

It's 2023, and data scientists develop a DNA test that perfectly predicts each person's future disease and thus each person's exact lifetime medical expenses.
  • In one country, insurers price all "insurance policies" to fully reflect these future costs. Everyone pays exactly what they will get back.
  • In a second country, insurers are required to provide insurance at standardized rate which ignores the DNA test and everyone is required to buy it.
Question 1: Ceteris paribus, which country has better insurance? Which has higher overall welfare? If you haven't taken the DNA test yet, which would you rather live in?

Question 2: Would any of your answers change if the test was imperfect and only predicted some of your medical expenses?


  1. Dear Professor Brynjolfsson,
    I'll think about those questions. Meanwhile, here are my answers to the two questions you asked me on my blog:

    1. Terrific.

      I look forward to reading your thoughts on the questions above.

  2. Mike Davis ( 25, 2013 at 2:42 PM

    I actually use a very similar hypothetical in class when I talk about adverse selection. The real risk here is not that you will get the disease, the real risk is that you will have the bad DNA. This might seem like a distinction without a difference but it matters a lot. Risk averse people want to smooth risks and insurance is one way they can do that. People might not care whether the risks are reduced through DNA insurance or disease insurance, all they care about is price. But insurers would have to recognize that these are two different products with two different kinds of issues.

    So to answer question 1, start by asking what happens if there is no regulation of insurance markets. Would insurance markets look like Country #1, which appears to be a place where there is no DNA insurance available?

    Maybe. You can certainly tell stories about how adverse selection destroys the market for DNA insurance

    Or maybe not. Remember that both insurers and their potential customers have incentives to try and mitigate adverse selection problems. Motivated buyers and sellers acting in free markets can come up with creative solutions.

    But now suppose you’ve decided that the market for DNA insurance is doomed to failure, would you rather live in Country 2?

    Maybe. Certainly if you haven’t yet taken the DNA test, you’re upset by the fact that you can’t buy DNA insurance. Maybe this world of “standardized rates” comes close to approximating what you’d get if you could buy DNA insurance.

    Or maybe not. Do you really think the regulators are going to get the standardized rate right? You don’t have to be a libertarian crank to think that rent-seeking and the usual parade of public choice horribles are going to mess things up. Maybe you’d prefer a different solution to the failure of the market for DNA insurance.

    Question 2 makes the analysis more complex in its particulars (how imperfect does the DNA test have to be?) but I think you’d still approach it the same way.

    1. Thanks for your thoughtful reply. There are clearly potential rent seeking opportunities in the public system, and for that matter, in the private system. It would be good to learn what the "creative solutions" that motivated buyers and sellers would come up with look like. We don't have the perfect DNA testing, but we do have increasingly good prediction of health care costs, so its not a purely hypothetical question.

  3. ANS:
    Question 1
    "Ceteris paribus, which country has better insurance?"
    -- Better for whom? The net givers? The net takers? Providers who will get paid for every patient? Patients who decide they'd rather have the money for their care than the care itself?
    "Which has higher overall welfare?"
    -- The system where people are responsible for the costs of their own healthcare would have higher overall welfare, because so many medical maladies are preventable, being exposed to the full cost of care as a result of their decisions, they'd be more likely to make better decisions regarding their health, and therefore be more healthy, and increased overall health and well being would undoubtedly result in higher welfare relative to the other country (genes aren't destiny, not now, and not in the future, again, highlighting your limited understanding of the delivery of medical care as described on econlib).
    "If you haven't taken the DNA test yet, which would you rather live in?"
    -- I'd rather live in the non-redistributionist society.

    Question 2:
    Would any of your answers change if the test was imperfect and only predicted some of your medical expenses?
    -- No, because this hypothetical is absurd, as no DNA test will ever serve as an even remotely good proxy of a patients lifetime medical expenses. There are certainly many genetic conditions which predispose to any number of disease. However, hundreds of billions of dollars in medical spending are for treatment of ailments brought about by patient choice. Smoking, alcoholism, engaging in high-risk sexual activity, violence and stupidity resulting in trauma, failure to comply with medications and lifestyle modifications needed to manage and minimize the damages of chronic disease, willful self injury -- all of the above are huge drivers of medical expenditures, and all are preventable. 'Bad luck' as you referred to in you econlib post, certainly befalls many, but more often than not, what you presumably attribute to 'bad luck' has nothing to do with luck at all, but is the result of 'bad choices'.

    1. Of course, after you know your test outcome, some will be lucky and some will be unlucky. But what about BEFORE you know your test outcome. Would you like to have the opportunity to buy insurance against the bad outcome?
      I agree that many outcomes are partly the result of choices and partly the result of luck. For the sake of this question, I'd be interested in your response to the situations that are due to luck (I assume you agree such cases exist). Should we try to insure against those cases or is it hopeless?

    2. Such cases certainly exist. Take childhood cancer...that's almost invariably bad luck, and is occasionally bad genes. And, the costs of their medical care is already effectively covered by a host of programs. You won't find children dying in the streets because the boogeyman medical industrial complex turned them away for inability to pay.

      But such cases are not the fiscal black hole relentlessly gobbling up our national resources. A huge percentage of expenditures for medical care in this country go towards providing care for conditions brought about or made worse by willful actions...not genes, not luck.

      I'm all for insurance (in the true sense of paying a certain amount now proportionate to the probability of a future event with contractually stipulated benefits which the insured will receive should such event occur). However, most can't seem to separate insurance in this sense, from that which is erroneously called 'insurance' but is effectively prepayment of medical costs.

      So, Dr. Brynjolfsoon, a question for you: Who should pay for the treatment costs such as frequent hospitalizations, lung biopsy, thoracic surgery for lobectomy, chemotherapy, and pharmaceutical costs for some 60 year old who has a 75 pack-year history of smoking?

      You've already astutely noted the seemingly irreconcileable flaw in our system of health finance in one of your econlib posts, that being that even those who don't have the foresight to buy insurance or the means to pay for their own care still receive treatment. Until society is comfortable letting those without ability to pay (be they sympathy compelling cases like some 60 year old lady who is a pleasant grandmother of 5, but never worked a day in her life, accumulated no assets, and never bought insurance who now has cancer and needs expensive care, or irritating cases like the 20-year old gang-banger shot several times during a drug deal gone bad who needs life saving surgery and several hundred thousands of dollars in care so he can get back out on the streets) die at the ER doors, healthcare finance will always be a standoff between heartstrings and pocketbooks. Modern medical care has become so advanced that it can better the lives of many; however, those successes come as the result of hard work by untold millions (from medical providers actually delivering that care, to the janitors who keep the facilities reasonably clean, to the engineers who design the new tools, to the IT guys who keep it all connected such that information is available and can be used when needed, to the factory workers making the various precision tools) that many people can quickly incur treatment costs (just barebones costs, not including the profit needed to incentivize these millions of workers to do their work) that far exceed their lifetime economic productivity.

      As such, debates about healthcare finance quickly become charged because the issue is fundamentally one of allocation of scarce resources, wrapped with all the emotion that surrounds our own mortality, coupled with our fear of disease, and colored by the psychological impact of seeing others suffer which we innately want to alleviate.

      In an (almost) perfect world where all disease was merely the result of bad 'luck' in the form of genetic defects, I could probably support a system where everyone paid into one big insurance pool, payouts from which covered medical care from those who happened to lose the genetic lottery. However, in the real world, where a large portion of disease is the result of poor choices, such a system effectively serves as a large transfer payment from the prudent to the imprudent and creates tremendous moral hazard.

    3. I think your last paragraph summarizes the issue well.

      Ideally, we'd like to design a system that provides as much insurance as possible with as little moral hazard as possible. The extent of moral hazard caused by health care payments softening the blow of getting sick is amenable to empirical research. I don't think it's as severe as you suggest.

  4. This comment has been removed by a blog administrator.

  5. I think this example misses some key points. Brynjolfsson has applied the assumptions that insurance is a 1:1 payment system for all individuals (risk is not spread across pools) and that individuals have no bearing on their health care consumption (DNA and bad luck).

    I wonder if these misunderstandings are common among others with similar leanings.

    1. Yes, in my hypothetical and extreme example, the company can perfectly predict your future costs so there is no longer any pooling. To some, this might seem like a "perfect" market. But my point is that in this extreme case, there is no longer any opportunity for risk pooling so people lose the chance to buy insurance. Ex ante, they are worse off.

      My second example invites you to consider whether there is still a loss of risk pooling when the ability predict future costs is imperfect. What do you think?

    2. If it were possible to know in advance all future medical consumption (and to assume that chance or choice could not affect medical expenditures), insurance as we currently know it would not exist -- there would be no demand for such a thing. Medical services then become the commodity that other living expenses are.

      Let's assume that you must consume a particular number of calories each day in order to live and that you know in advance how many calories you need. Are you worse off for not having a pool over which to spread that cost? This may seem trivial now, but calories weren't always as cheaply available as they are today. In your hypothetical, insurance isn't needed and a more clear market develops for medical services.

      Seeingthat Q2 introduces a sometimes unreliable test, it may increase costs by bringing insurance back into the market, which inherently drives up patient cost when it is used as a payment clearing house, as it is today.

    3. This comment has been removed by the author.

  6. Q1: Using the DNA test to estimate one's medical expense in order to price health insurance premium accordingly is not much different from the existing practice in the private health insurance market. Insurance companies gather and use information about their customers such as body weight (obesity), smoking or drinking habit (cancer), and gender (pregnancy). What is different in this case is that the DNA aims for perfect price discrimination at individual level while the other customer information was used at group level.

    If the goal of having this health insurance plan in Country #1 is to achieve the highest market efficiency, then this is a great news. However, for a topic like health insurance, we must consider its impact on equity. The people with bad DNA (high medical expense) are not necessarily the ones with high wealth (ability to afford the cost). Thus, the government needs to step in (or don't let the market be highly efficient) to make sure everyone gets the medical treatment they need whether they can afford it or not.

    I would like to live in Country #2 than Country #1.

    Q2: In a way, we are already living in this environment.

  7. I wrote a paper about that issue published in 1994 (!).

    Tabarrok, A. 1994. Genetic Testing: An Economic and Contractarian Analysis. Journal of Health Economics 13:75-91.

    Ah, how time flies. Basically I proposed genetic insurance. If the test turned out bad you would be paid enough to be able to afford actuarially sound health premiums at the now higher rate, ie. an attempt to bring Rawls back in!

    Cochrane's time consistent health insurance later used the same idea in a more expanded context

  8. I think if you are wealthy you would want to stay in Country 1 where you know how much you need to pay and you don't want to pay for others. If you are not wealthy you can not take the chance and have to be in Country 2.
    This is similar to the social health care debate in the US. Above a certain wealth level many don't want to pay extra taxes for others (who cannot afford, perceived to have more health risk as well) because they can control their own risk/cost and they do not want to dilute the pool with more risk. In your example the country 1 risk is 0 because you exactly what will happen. In the current case a higher risk pool and cost vs. lower risk pool and cost.

  9. there is an implied assumption that data privacy and HIPAA concerns - are taken care off and patients don't bear the adverse affects ( discrimination?) of their health related information being freely disseminated or discussed by unrelated stakeholders. Hence one would be Ok to be in Country 1. Else Country 2.