Variance Investigation in Agency Settings

Early variance investigation papers (see Kaplan [1975] for a review) were concerned with controlling mechanical production processes. More recently, the question of when to conduct an investigation has been examined from the perspective of agency theory. In an agency setting, the productive outcome is affected by both exogenous factors and the actions of an agent. An incentive problem occurs when the principal cannot observe the agent's action directly. This incentive problem can be reduced if the principal can acquire additional information about the agent's action. Of course, the benefits of acquiring additional information via an investigation must be weighed against the costs of conducting the investigation. For this reason, it is of interest to determine what factors affect the benefits of conducting an investigation. Baiman and Demski [1980a; 1980b] show that the productive outcome that occurs affects the benefits of investigation, and that the decision to conduct an investigation will therefore depend on the production outcome. In deriving their results, Baiman and Demski assume that the distribution of the information signal that is generated about the agent's action is independent of the productive outcome. In this paper, I allow the productive outcome to affect the distribution of the information signal that the investigation generates. This allows for consideration of investigations which provide information about the environment or the