Mostly Not Redistribution

The welfare state is debated as a transfer from the rich to the poor. Most of what it spends is something else entirely: money moved between stages of the same person's life.

Look at what the money buys. Education is consumed between five and twenty-two. Health care is consumed disproportionately in the last years of life and the first. Pensions are received after work ends. All of it is paid for during the earning years in between. Nicholas Barr's phrase for this is that the welfare state is a piggy bank as much as a Robin Hood, and lifetime accounting bears it out: the majority of what a typical household receives is money it contributed at a different age. Redistribution between people is real, and it is a smaller share of the total than the political argument assumes.

This matters because a cross-sectional snapshot, the picture nearly all public debate runs on, shows pensioners and students receiving while workers pay, which looks like a transfer between groups. Followed across time, most of those people are the same people at different ages. The snapshot systematically misdescribes who is paying for whom.

Why does the state do this rather than leaving people to save? Individuals could in principle smooth their own consumption, and many do. Private arrangements fail against a specific class of risks: ones that are unknowable in advance or uninsurable in a market. You do not know how long you will live, whether you will develop an expensive chronic condition, or whether your industry will exist in thirty years.

Health insurance is the clearest case of market failure, and it fails through adverse selection. The people most likely to claim are the most likely to buy. Insurers respond by raising premiums or excluding the high-risk, which drives out the healthiest remaining buyers, which raises average risk in the pool, which raises premiums again. The equilibrium is a thin, expensive market that excludes precisely the people who need cover most. Every rich country resolves this the same way, compulsory participation in some form, however different their systems look on the surface.

Against this sits moral hazard: insurance changes the behaviour it insures against. Being covered makes a claim somewhat more likely. The effect is real and usually smaller than assumed, and it is constantly confused with adverse selection in public argument. The two call for different remedies, one about information and the other about incentives.

The best evidence on how far demand-side discipline can go is the RAND health insurance experiment, which randomly assigned families to different levels of co-payment. Higher cost-sharing reduced the use of medical services substantially, and the reduction was roughly indiscriminate between care clinicians judged appropriate and care they did not. Patients are not reliable triage agents for their own symptoms, which is unsurprising and inconvenient: the cure for moral hazard suppresses useful treatment at about the same rate as wasteful treatment.

The upshot is not that moral hazard does not exist. It is that the standard remedy has a cost that is rarely counted, and that a system designed as though the main problem were people over-consuming free medicine will get the trade-off wrong in a predictable direction.