Contracting on Aggregated Accounting Estimates

Using a principal-agent setting in which the agent has a rich action space, we provide a novel framework for studying accounting measurement and aggregation. We show that the optimal contracting process can be decomposed into three stages: constructing unbiased estimates of items that the principal values, aggregating those estimates using the weights in the principal's objective (as opposed to weighting by sensitivity or precision), and compensating the agent on the aggregated estimate. This decomposition provides a theoretical justification for contracting on highly-aggregated accounting metrics. Moreover, our results reconcile the conflict between the stewardship and valuation uses of information; when the agent has flexible control over firm performance, evaluating the manager and valuing the firm are one and the same. In a highly tractable specification of our model in which normal distributions arise endogenously, we show that optimal measurement rules are conservative yet produce unbiased estimates. Furthermore, we show that a weaker link between investment and future returns warrants more conservative treatment of expected future benefits, providing a rationale for the immediate expensing of R&D, the capitalization of PP&E and the accrual of credit sales