Previous research attributes earnings management to economic incentives to improve the appearance of company performance or to reach important benchmarks, such as targeting to meet (or simply beat) analyst forecasts. My colleagues, Tien-Shih Hsieh of the University of Massachusetts–Dartmouth and Jean C. Bedard of Bentley University, and I investigated whether earnings management is associated with CEO “overconfidence”—a tendency to overestimate one’s own knowledge and abilities.
Prior literature shows that overconfident CEOs have unrealistically high expectations of their company’s future performance and a belief that they can ensure that high performance is achieved. Research has demonstrated that companies with overconfident CEOs issue more optimistic earnings forecasts and are more likely to have financial misstatements that are later the subject of Securities and Exchange Commission (SEC) enforcement actions.
A 2014 study by Banerjee et al. found differing implications of overconfident CEOs for companies before, versus after, the Sarbanes-Oxley Act of 2002 (SOX). Before SOX, results were consistent with the negative implications of the other studies previously cited (companies with overconfident CEOs make decisions that distort investments and load excessive risk into the company). However, after the passage of SOX, overconfident and non-overconfident CEOs behaved similarly, and the negative impacts of CEO overconfidence on firm value and industry-adjusted accounting returns decreased. This implies that the regulatory constraints imposed by SOX were effective in reducing overconfident CEOs’ opportunistic behavior.
Since earnings management affects accounting returns, and firm value could be improved (at least in the short term) by better accounting returns, the missing piece in the picture of overconfident CEOs is whether their companies have systematically greater upward earnings management. Recent research reports a general shift away from accruals-based earnings management toward real activities-based earnings management after SOX and a related decrease in the frequency of targeting to meet or just beat analyst forecasts. However, overconfident CEOs might not follow this pattern if their greater expectations lead them to “push the envelope” of earnings management, regardless of the regulatory constraints of SOX.
Thus, we investigated whether the improved post-SOX performance of overconfident CEOs’ companies could be at least partially explained by earnings management. We examined several mechanisms through which overconfident CEOs might “make their numbers” in the pre- and post-SOX time periods, including accrual-based and real activities-based earnings management.
We tested our hypotheses in a sample of observations from 1991 to 2009, in which all CEOs had the opportunity to exercise in-the-money options (i.e., have a positive estimated option value). We classified a CEO as overconfident if s/he held stock options when those options were more than 67 percent in the money at least two times during our sample period.
Our results showed a positive and significant interaction of CEO overconfidence and SOX on companies’ signed discretionary accruals, which implies that companies with overconfident CEOs are more likely to use accruals to manage earnings upward than non-overconfident CEOs in the post-SOX period. We also found that CEO overconfidence was associated with abnormal levels of income-increasing real activities-based earnings management. Specifically, we found CEO overconfidence to be associated with lower abnormal discretionary expenses and higher abnormal cash flow from operations. We also observed a reduction in abnormal production costs post-SOX, regardless of CEO overconfidence. Lastly, we found that CEO overconfidence was positively associated with just meeting or beating analysts’ consensus forecast in the pre-SOX period, but this tendency was eliminated in the post-SOX period.
Overall, we found evidence of greater earnings management in the pre-SOX period among companies whose CEOs were classified as overconfident—and of continuing earnings management in such companies through accruals, operating cash flow, and discretionary expenses in the post-SOX period. Our results imply that at least some of the improved operating performance demonstrated by Banerjee et al. results from actions taken by overconfident corporate executives that do not build long-term value for their companies. In the post-SOX environment, we found no evidence of a greater tendency by overconfident CEOs to manage analyst forecasts. Our results illustrate SOX’s success at curbing some earnings management behaviors and the important role that those charged with governance play in assessing CEO overconfidence and acting to constrain inappropriate, short-term oriented behaviors of some CEOs and top executives.