Budget Makers as Agents : A Preliminary Investigation of Discretionary Behavior Under State-Contingent Rewards

This analysis reflects the presumption that the principal-agent relationship is central to an understanding of government spending. Our model's structure is adapted from Jack Hirshleifer's (1970) use of the state-preference theory of decision-making under uncertainty to delimit the information asymmetry problem that lies at the heart of the principal-agent relationship. Our principal conclusions are quite straightforward and can be summarized as follows. If elected officials reward budget makers solely on the basis of budget accuracy, budget forecasts will not be biased, i.e., they will always equal expected revenue. Since in our model budget makers are assumed to be risk averse with respect to rewards, we conclude that risk aversion is not sufficient to explain biased forecasts. Alternatively, if rewards are mixed - based partly on budget accuracy and partly on other criteria - budgets will equal, exceed, or fall below expected revenue, according to the characteristics of the probability distribution of tax revenues. Because these distributions are likely skewed to produce budgets that fall below expected revenue, we conclude that mixed rewards are both necessary and sufficient to explain the budget makers' observed bias. Finally, insofar as reward structures are determined by the budget makers' political masters, we would not discount the possibility that this bias ultimately reflects the interests and preferences of their masters, although the more conventional view is that the preferences of elected officials run in precisely the opposite direction