Introduction

The owner of an enterprise wants to put it in the hands of a manager. In each of a number of successive periods (month, quarter, year) the profit of the enterprise will depend both on the actions of the manager and on the environment in which the enterprise is operating. The owner cannot directly monitor the manager's actions, nor can the owner costlessly observe all of the relevant aspects of the environment. The owner and the manager will have to agree on how the manager is to be compensated, and the owner wants to pick a compensation mechanism that will motivate the manager to provide a good return on the owner's investment, net of the payments to the manager.

This is the well-known principal–agent problem, with the additional feature that the relationship between the owner (principal) and the manager (agent) may last more than one decision period. Some other principal–agent relationships in economic life are client–lawyer, customer–supplier, regulator–public utility, and insurer–insured. The insurer–insured relationship gave rise to the term moral hazard, and the first formal economic analysis of moral hazard was probably given by K. J. Arrow.

Early formal analyses of principal–agent relationships focused on “short-term” or “one-period” situations. The simplest of these can be modeled as a two-move game, in which (1) the principal chooses a compensation function (made known to the agent) and (2) the agent chooses the action. The outcome is then determined as a function of the agent's action and some exogenous stochastic event, and the principal compensates the agent as a function of that outcome, according to the compensation function the principal previously announced in the first move.