uture consumption would be purchased some of whose expected benefit exceeded expected cost at the margin(where the premium exceede actuarially predicted cost)and some of whose expected benefit fell short of expected cost at the margin (where the premium was less than the actuarially predicted cost)(Arrow [1]). The second derives from a zero price implying excess demand and a failure to satisfy the necessary condi tion that expected marginal valuation should equal marginal cost. Neithe of these two objections, however, can be regarded as sufficient to show the relative inefficiency of the NHS. First, the ma a dis crimination ment is a condition relating to a hypothetical world of zero transaction contract, and information costs. All real world institutions have degrees of pooling built into them as a means of economizing on the costs of collecting enough information for thecorrect'premiuims to be found. The comparison is properly one between different conceivable real world institutions where second-best solutions must be discovered. At least, such solutions may be second-best by comparison with a hypothetical ideal, but they may be first-best in terms of what can actually be done. This is clearly not a matter to be settled by a priori reasoning since the correct degree of pooling- depends upon the costs of administering various pooling/discrimination. gan empirical matter. There is also an alternative argument which asserts that pooling as under the NHS is actually more efficient than risk discrimination. Risk discrimination requires that individuals in groups. having a higher incidence of sickness should pay higher premiums since they impose higher social costs. With pooled risks in the NHS, however, he premium takes the form of a tax- price which, to the extent that the NHS is financed out of general funds and the tax-system is progressive, divorces the premium from the individuals risk and relates it instead to- come. Since the poor would tend to be less discriminated against under his system than one in which risk discrimination was the rule, there may be some social benefit from such a form of organization relative to one in was less prominent. This argument is an important one which we return later. At this point, however, it is observed that whatever merit the argument may have, it is not related to the matter of optimizing cost uncertainty. Instead it concerns the incidence of absolute costs. It is therefore not relevant to the discussion of this section The second objection to the NHS system is also derived from comparing the real world with a hypothetical ideal instead of a realistic, or conceivable alternative. In any known health care system, however, there is some response to the problem of the uncertainty of costs which incorporates insurance or pooling elements by which all fail the test of comparison with a hypothetical ideal. The reason for this is that so long as the tax-price (or premium) does not exceed the expected value of consumer's surplusReproduced with permission of the copyright owner. Further reproduction prohibited without permission