Sunday, 2 August 2009
I heard somewhere that what people do on blogs is write comments about what other people said on blogs. I guess I do a bit of that, but I'm sure my rate is below average. Therefore time to catch up.
In this post I'll look at a couple of interesting points on the American healthcare debate:
Megan McArdle looks at rates of insurance among unhealthy people and finds that the adverse selection theory is wrong. But I think there's a serious flaw in her argument. Her finding is that since uninsured people are about as healthy as insured people, adverse selection can't be happening. But she relies on the following assertion: "we would expect the uninsured to be sicker than the general population". I don't think that's correct at all. Surely sick people have far more incentive to get insured than healthy, and so we'd expect the uninsured to be substantially healthier than the average? After all, we would assume that people who don't own cars probably have less need of cars than average; and those who don't own swimsuits probably do not want to swim very much. Therefore, if a significant proportion of uninsured people are sick, and therefore need healthcare very much, there does appear to be a big failure in the market.
Megan also says that we don't want to gut the healthcare system in order to solve a problem affecting only 0.3% of the population. But actually the healthcare system's job is mainly to help people who are sick. Thus it's sick people, not the whole population, who are the relevant population here. On the basis that 15% of the population don't have insurance, and that percentage is slightly higher among those who are ill (in "fair" or "poor" health), the healthcare system is failing about 18% - more than one in six - of the relevant population. Perhaps that still doesn't justify a complete restructuring; but the argument is very different.
Next, Greg Mankiw makes a very good point: a lot of the healthcare debate - and many other economic debates - reveal something about the kind of institutions that people trust. Greg explains his reasons for trusting market-based solutions and private companies more than governments.
(An aside: Greg quotes Paul Krugman saying he doesn't trust private healthcare because "your treatment is their cost" and thus they have an incentive not to treat you. But that's true of all private companies: providing any service to you is a cost to them. Single-play game theory indicates that all companies should try as hard as they can to rip you off. The sandwich shop should give you the smallest, cheapest sandwich they possibly can. Multiple-play theory, however, shows why they don't - because they want you to like them and to come back again. There is far less repeat purchasing in healthcare than in sandwiches, so that argument is less strong. However the principle still works with cars, which we buy only occasionally, because there is a wide popular conversation about cars which shares consumer information about them and boosts the value of a good reputation. Maybe that would work for healthcare too)
This - and something I was asked by a potential customer a few weeks ago - got me wondering about the best ways to model trust. My theory of how people make economic decisions is broadly based on the idea that they have a mental model of the world, with many different components, each of which has some kind of expected utility function attached to it. For instance I expect that another cup of coffee will bring me a certain level of happiness; making lunch similarly; and going out for a beer something else again (at which point I may want to take into account the difference between short and long term utility).
These expected utility functions each have some degree of credibility associated. I can predict better how the coffee is going to make me feel than the lunch, partly because I have just had a coffee and can remember it; and partly because my lunch today will be something new that I haven't had before. I might really enjoy the lunch or it might leave me indifferent. The lunch utility function has much less credibility than that of the coffee (even though my actual enjoyment of the lunch could well be greater).
I believe that this credibility value is correlated with trust. If we buy a private health insurance policy - or decide to rely on the public system - my uncertainty about the level of utility it will provide is mainly mitigated by my trust in the provider. I have a reasonable idea about the level of care I'll get from the NHS. But for me, there is much less certainty (through lack of experience) about the service level and outcomes of private health treatment. I suspect it would be better, because after all BUPA has to compete with the NHS and still get people to pay money, but I don't (yet) trust the company enough to be confident it would be much better.
Of course this doesn't just apply to healthcare: a big challenge for my software company, for instance, is to get people to trust us enough to try us out. We know we'll do a good job, but the prospective customer doesn't.
There are other factors involved in credibility apart from trust - uncertainty about the general present state of the world, unpredictability of future circumstances, and in some cases straightforward randomness; but trust has a big influence and is perhaps the easiest one to control.
One way that trust can be enhanced is by providing more information. In software we do that through references and testimonials, detailed proposals, and by offering low-risk low-cost trials of our services which will give people an understanding of what we do - and attach more credibility to our utility function. Another is through personal affinity - we tend to trust those that we feel something in common with - and so those customers that we personally get on with are more likely to believe that we can provide a good service.
Healthcare providers might do well to consider this. Transparency could be good for everyone - that is, if it were enforced across the board. Shouldn't you be able to find out both the death rates for certain operations at your local hospital and also the proportion of claims that your insurance company refused or policies they rescinded last year? Assuming that the refusals are rare, releasing these figures should enhance overall trust in the sector; and if they are not, then consumers are entitled to know it, and competition should drive them down. Of course no company would release these figures alone - it would take legislation or some other means of enforced coordination.
This leaves open Greg's question of how much you should trust the federal government. No doubt some incentives could be designed for that situation too, but I'll leave that for another time.