Earning Management in Two-Period Principal-Agent Models

Abstract

Theory suggests that in a multiperiod principal-agent setting optimal cash flow based contracts should exhibit memory of wages (Lambert, Bell Journal of Economics, p. 447); that is, wages should depend on both present and past outcomes. There is only weak empirical evidence such as raises/pay cuts and promotions/demotions for this theoretical result. Thus the conclusion can be made that real management compensation is in fact rather memoryless. In addition, these memoryless contracts are usually based on reported earnings, not on cash flows. This paper offers an explanation for this phenomenon.

Although cash flow based contracts with memory c. p. strictly Pareto-dominate memoryless ones, the welfare improvement may be outweighed by an increment of transaction costs, i. e., costs of writing and specifying the memory contract. This paper focuses on the second question: Why does practice prefer earnings-based to cash flow-based memoryless contracts even if the agent has discretion to manage earnings to maximize her own utility? It will be proven that at least in a setting without the agent's access to capital markets, the former contracts strictly Paretodominate the latter. To present a strictly formal proof, a model modifying Lambert's (1983) scenario is introduced in which the agent is allowed to manage earnings in period 1; in period 2 there is the reverse effect resulting from the 'clean surplus' or 'tidiness' property of accrual accounting. The interpretation for the result is that earnings management enables the agent to smooth consumption over time. Thus the (risk-neutral) principal can lower the expected wage payment to implement the optimal action pair without violating the agent's reservation utility. Consequently the agency is strictly better off.

A brief analysis of earnings management in a dynamic principal-agent setting with free access to capital markets concludes the paper.

JEL Classification: C72, D82, M41, M43, M49