Measuring Intentional Manipulation: A Structural Approach

Abstract

Based on a sample of about 1,400 CEOs, I estimate the extent of undetected GAAP violations and managers' manipulation costs using a dynamic finite-horizon structural model. The model features a risk-averse manager, who receives cash and equity compensation and maximizes his terminal wealth. I find that the expected cost of manipulation is low. The probability of detection is 6%, and the average misstatement results in a 24% decrease in the manager's wealth if the manipulation is detected and the manger is terminated. Based on the estimated parameters, the implied fraction of CEOs who manipulate at least once during their tenure is 73%; the value-weighted bias in the stock price across manipulating CEOs is 6.97%, and the value-weighted bias in the stock price across all CEOs is 2.82%. Finally, I find that the model-implied measure performs at least six times better in terms of the root mean squared error out-of-sample than any of the five discretionary accruals measures that, in addition to GAAP violations, capture manipulations that do not violate the GAAP.