An Algorithmic Approach to Contracting in an Infinite Agency Model

The paper describes the numerical results obtained when solving an infinitely repeated agency model with strategic income reporting. The model distinguishes itself from previous models of infinitely repeated agency problems on two accounts. First, due to the accrual nature of income reporting, the uncertainty cannot be modeled as independent, identically distributed noise terms. Second, the 'state' is not fully observed by the principal. A consequence of the correlation over time is that the problem has to be solved as a non-linear dynamic programming problem. Using a modified version of Gauss-Seidel value iteration, the computations show that the stock price of a firm fluctuates around a constant. The fluctuations depend on the company's retained earnings and a summary statistic of previous signals. The firm's dividends depend on retained earnings, the firm's accounting reporting strategy, and a summary statistic for previous signals