administrative costs are absent. The existence of self-insurance is not a fficient condition for sub-optimality. The relevant comparison is, as has been emphasized above, between conceivable real world institutions In practice, one suspects that a pooling system with compulsory member ship will provide lower unit administrative costs than a voluntary risk discriminatory insurance system, though the matter is an empirical or and one' s suspicions are not always well founded. The problem is more complicated than this, however, since both compulsion and pooling imply. social losses to be offset against any cost reductions, so the conceptual argument is less unambiguous than might appear and the empirical exercise would involve some quite heroic cost-benefit efforts 7. The problem of moral hazard, which has previously been alluded to, rises with some forms of insurance because the consumer of medical care is confronted with a marginal cost at the point of receiving care that is less than the true marginal social cost of provision, hence leading to some loss of welfare. If an individual would consume X'units'of medical care if he contracted a particular sickness(and had to pay the full cost)his behaviour changes when he insures against the costs, whatever they may amount to, of that contingency. Ignoring transaction, etc, costs, with unit costs of providing care at c, he might expect to pay an actuarially fair premium of p(cx), where p is the probability of falling sick. Being insured, however, the marginal cost of further treatment(e.g. more days convalesce- ing in hospital, more physican visits, hiring more eminent physician) is ro to him, leading him to consume X units of care where x>x,at a cost to the insurance company of p(ex), assuming constant costs per unit of care. The actuarially fair price therefore rises, leading to a reduction in the number ofinsured persons, i. e. an increase in the number of uninsured risk-averse persons. In addition, each insured person incurs a dead-weight loss on all extra-marginal units of care consumed beyond the point at which marginal valuation equals marginal social cost. The problem is miliar to that faced by the NHS-type organization, which has zero user prices and which satisfies demands in excess of the optimal amount. There are, however, differences in the way the two types of system distribute their excess burdens. In the market, the excess burden can be avoided by lf-insurance, but the self-insured also lose, of course, by virtue of being confronted with a range of premiums which, although they may reflect the ill sooial costs incurred by the mpany, do not reflect the lower premia that would exist if moral hazard were absent. In the NHS he excess burden is incurred by all. 1 It does not follow that because the market permits individuals to escape the excess burden(though at a cost 1 And some individuala'dead weight utility loss may exceed, with compulsory the utility gain from risks avoided. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.Reproduced with permission of the copyright owner. Further reproduction prohibited without permission