In this paper, I shall investigate a role for trade unions in increasing economic efficiency. This role is enabling an improvement in the allocation of riskbearing between a firm and its employees. The essential idea is straightforward. If employers and employees are both risk averse, the optimal wage contract will not leave one party bearing all the risk if state-contingent contracts are possible. Grossman and Hart (1981) give several reasons for firms being risk averse. That firms are indeed risk averse would seem to follow from the fact that employees bear substantial risks. Few firms offer a job for life at a guaranteed real wage. Employees bear the risks of layoff, redundancy and changes in the real value of their earnings. With transactions costs in finding and moving to another job (which may be substantial in periods of high unemployment), these are real risks. Yet the states of the world in which these eventualities are to occur are rarely specified in any legally enforceable form. Indeed, it is hard to imagine that they ever could be, both because of the problem of envisaging all the relevant states of the world (the bounded rationality of Simon (1979)), and because of the problem of defining those states in such a way that the courts could, in the event of a dispute, decide which state had occurred. In the absence of legal enforceability of state-contingent contracts, the analysis depends on whether or not decisions about production must be made before a firm knows which state has occurred. If they do not have to be, then it is not essential to negotiate any contract before the firm knows the state but it may still be desirable from a risk-sharing point of view to have a long-term contract covering different possible states. For that case, Hall and Lilien (1979) and Grossman and Hart (1981) have argued that some degree of state-contingency may be achieved by negotiating a contract which allows the firm, in the light of the state which occurs, to choose between limited combinations of employment and wages. The firm may, for example, be permitted to reduce wages under the contract but only at the expense of reducing employment so that it is only in its interest to do so if marginal productivity is low. This can overcome some of the moral hazard problems arising when states are observed only by firms but, as Grossman and Hart show, only at the cost of Pareto inefficiency in the employment contract. At an empirical level it may explain provisions for compulsory overtime at premium rates. This type of contract will work only in the case in which the firm observes the state of the world before making employment decisions but, given that