Agency

The Economic Modeling of Agency

In the economic tradition, the investigation of agency is linked to the growing awareness of market and hierarchy failures, a major trend of research during the 20th century. Within mainstream economics, the behavior of an agent is modeled as the unique result of a cognitive evaluation of his or her own interests, choices, or preferences. Failures only emerge from imperfections in the information available to the agent. They may be caused by the cognitive limits or bounded rationality of the agent as formulated by Simon and others since 1947, or, result from information asymmetries, i.e., when some agents have valuable information unknown to others, and when this is common knowledge to all. For instance, we all know that physicians know things that we are unaware of. A central issue for economists is to overcome such imperfections with standard economic instruments, i.e., by designing contracts with the “best” monetary incentives or reward schemes.

Thus, standard agency theory models a situation where one agent, called the principal, has contractually offered wages or a reward scheme to another agent who commits himself or herself in return to perform a task that brings some value to the principal. In simple models, this agent can hide the real level of an effort in doing the job. Hence, if the agent receives a fixed wage for the job, the agent will “rationally” choose the lowest level of effort whenever the principal cannot observe this level of effort. In such a case, and in a large variety of variants (see Levinthal's 1988 writing), economists have searched for the best reward system that would induce the agent to chose the level of effort that also optimizes the principal's profit. Results are rather intuitive and synthesized by Milgrom and Roberts. They indicate that the principal should design incentives that take into account all the direct or indirect information about agent effort, for instance, how effort is linked to performance output. The principal may also accept to pay for monitoring, i.e., take actions that give the principal additional knowledge [Page 44]about the agent's effort. Yet, if incentive schemes are built only on partial aspects of the task, they may induce the agent to select the type of efforts that maximizes the performance indicators, while endangering other, nonmeasured dimensions of the job.

...