Allocation by lot: a conceptual and empirical analysis

valued because of an atavistic rivalry between the two countries in this respect. The Club Committee decided to have lots drawn for the tickets and to ask the winners to contribute DFL30 to the club funds for the purpose of buying a TV set for the club canteen. The six boys who had won the quiz, and who had already received personal presents, were not given priority in the distribution of the tickets; this decision was later reversed under heavy pressure amounting to, among other things, anonymous telephone threats to the club chairman. Many more illustrations could be given for the strong feelings of aversion that are often evoked by using lottery as an allocative mechanism. On the other hand, rational arguments in favour of lottery have been put forward. Elster (1989a) discusses three types of indeterminacy that might justify random choice between options. One is strict equioptimality, as in choosing between cans of Campbell's tomato soup. A second is equioptimality within the limit of what it pays to find out, that is, the case in which the cost of gathering more information would exceed the marginal utility of the superior option. The third is the incommensurability of options; one might say that, in this case, any investments into the choice procedure are fruitless a priori. The present analysis takes the contrast between aversion and argument as its point of departure. Its scope is allocative problems and particularly those situations in which some public authority distributes scarce indivisible goods among people, rather than problems of choice between goods from a private point of view, whether individual or institutional. In the context of allocation, the question about the feasibility of lottery may be analysed as follows: from a rational point of view, the efficiency of distributive mechanisms is at stake, as it is in other problems of rational choice. However, the interplay between the allocative authority, the target subject, and the general public or common interest introduces two further aspects of feasibility. One is justness, that is, the extent to which the mechanism is compatible with written and