We study intertemporal incentive properties of conditional accounting conservatism. Conservatism has detrimental and beneficial properties. In our first model, conservatism introduces downward bias in the first period; any understatement of first-period performance is reversed in the second period. A conservative bias is not costly in the first period but instead is costly in the second period when a new manager may be rewarded for the performance of his predecessor. In an extension on learning, we illustrate a beneficial role of conservatism in fine-tuning incentives. In the second model, conservatism is modeled as recognizing effort-independent bad news early and good news late. Recognizing bad news early can be optimal because of intertemporal rent shifting, which improves incentives via an “incentive spillback.” We also study overlapping projects (a multi-task setting) in which an interior accounting system can be optimal to avoid making one of the overlapping projects an incentive bottleneck.

We examine the design of asset revaluation policies in settings where a regulator can mandate fair value disclosure in order to mitigate a lemons problem in the asset resale market. The welfare-maximizing policy generally prescribes fair value certification for the lower asset values and (less costly) historical cost reporting for the higher asset values. The potential for voluntary certification can reduce welfare by increasing equilibrium certification costs and promoting underinvestment in socially valuable projects. Thus, a single regulated source of information (mandated disclosure) can be preferable to two sources of information (mandated and voluntary disclosure).