Taxation of capital gains at realization may distort individuals' decisions regarding holding or selling during an asset's lifetime. This creates the problem of designing a tax structure for capital gains so as to induce efficient patterns of holding an selling. Several tax structures are explored in this paper. Linear taxation, at rates which rise with the holding period, can achieve the first best, even under the conditions of limited information that we postulate. The form of the optimal tax is independent of the stochastic structure of rates of return. We also derive the optimal nonlinear tax under the constraint that it be independent of the holding period. Second-best tax rules are examined. Results in a two-period model are contrasted with those in a continuous time framework. Also treated is the case in which the returns to the asset under consideration depend on the aggregate quantity invested

Taxation of capital gains at realization may distort individuals' decisions regarding holding or selling during an asset's lifetime. This creates the problem of designing a tax structure for capital gains so as to induce efficient patterns of holding an selling. Several tax structures are explored in this paper. Linear taxation, at rates which rise with the holding period, can achieve the first best, even under the conditions of limited information that we postulate. The form of the optimal tax is independent of the stochastic structure of rates of return. We also derive the optimal nonlinear tax under the constraint that it be independent of the holding period. Second-best tax rules are examined. Results in a two-period model are contrasted with those in a continuous time framework. Also treated is the case in which the returns to the asset under consideration depend on the aggregate quantity invested

Tournaments, reward structures based on rank order, are compared with individual contracts in a model with one risk-neutral principal and many risk-averse agents. Each agent's output is a stochastic function of his effort level plus an additive shock term that is common to all the agents. The principal observes only the output levels of the agents. It is shown that, in the absence of a common shock, using optimal independent contracts dominates using the optimal tournament. Conversely, if the distribution of the common shock is sufficiently diffuse, using the optimal tournament dominates using optimal independent contracts. Finally, it is shown that for a sufficiently large number of agents, a principal who cannot observe the common shock but uses the optimal tournament does as well as one who can observe the shock and uses independent contracts.

Tournaments, reward structures based on rank order, are compared with individual contracts in a model with one risk-neutral principal and many risk-averse agents. Each agent's output is a stochastic function of his effort level plus an additive shock term that is common to all the agents. The principal observes only the output levels of the agents. It is shown that, in the absence of a common shock, using optimal independent contracts dominates using the optimal tournament. Conversely, if the distribution of the common shock is sufficiently diffuse, using the optimal tournament dominates using optimal independent contracts. Finally, it is shown that for a sufficiently large number of agents, a principal who cannot observe the common shock but uses the optimal tournament does as well as one who can observe the shock and uses independent contracts.