How Do Investors Respond to Restatements? Repairing Trust Through Managerial Reputation and the Announcement of Corrective Actions

In fact, Johnson (2008) reports that over 10% of public companies issued restatements in 2006.

In the attributions literature, this type of self-serving explanation or account is common when negative events occur because it reduces management’s personal responsibility for the event (Schlenker et al. 2001). In a restatement context, top management has “a natural tendency to offer excuses in response to actual or anticipated questions” as a way to distance itself from the misstatement (Reuber and Fischer 2009; Elliott et al. 2012, p. 517).

SOX holds top managers responsible for creating and maintaining an internal control system over financial reporting. In addition to signing a statement taking responsibility for the financial statements, SOX requires CEOs to sign a statement taking responsibility for the effectiveness of internal controls. By establishing a system of effective internal controls, fraud within the firm should be prevented and/or detected. Thus, even though top managers had no direct involvement in the fraud that led to the restatement, one could argue that top managers are indirectly contributing to the fraud. For example, perhaps because top managers failed to adequately create and maintain a strong control environment, other employees were not prevented from engaging in fraud. Further, top managers are in positions of authority, meaning that they are expected to anticipate negative outcomes. Thus, even if top managers were not directly involved in the circumstances leading to negative outcomes, they are likely to be held accountable (Tennan and Affleck 1990).

Mercer (2004) delineates two dimensions of managerial reputation: perceptions of competence and perceptions of trustworthiness. Our operationalization directly affects perceptions of competence (Goodman et al. 2014; Trueman 1986), and may affect perceptions of trustworthiness. That is, we are silent as to whether the inaccurate forecasts are biased in a certain direction, meaning that the forecasts could be inaccurate due to either incompetence and/or to a desire to mislead investors. Thus, we view this as a strong manipulation of managerial reputation.

Appointment of a Chief Ethics Officer or a Chief Compliance Officer became increasingly common following SOX (Clark 2006). Sometimes, this position reports directly to the Board of Directors, but just as frequently, this position may report to another senior-level position within the company (Fox 2010). In our experiment, we indicated that the Chief Ethics officer would report to both the CEO and the Audit Committee. To the extent that this choice weakens the new position, it biases against us observing treatment effects for our manipulation. Thus, we view this as a conservative design choice.

The Institutional Review Board at the university where the data were collected approved the use of human subjects in this experiment.

To provide comfort that responses to this question were not due to random chance, we also asked participants to confirm that the CEO did not make additional promises beyond the corrective actions and stated, “The company’s press release ____ that the CEO personally promised to do whatever he could to ensure that the company’s future financial reports are of the highest quality and that statement fraud never occurs.” Participants responded by circling “announced” or “did not announce.” Eighty-three percent of participants correctly indicated that the press release did not contain such a promise.

Excluding the 18 participants that failed one or both of our manipulation check questions does not change our inferences. All results reported in Fig. 1 continue to achieve conventional significance levels even when based on the 76 participants that passed both manipulation checks. Thus, the subsequent analyses include all participants.

Two items, which address blame and responsibility, did not load on either factor. Thus, we conducted the PCA without these items.

We rely on mean responses to the relevant questions for each factor, rather than on factor scores. Both approaches are used in practice (O’Rourke and Hatcher 2013). However, using mean responses makes it easier to interpret our factors vis-à-vis our response scale.

As an additional test, we examined the standardized residual errors for all omitted paths in Fig. 1 to determine whether inclusion of any of these paths would meaningfully improve model fit. None of the omitted paths excluded from the model demonstrated standardized residual covariances above 2.0, which is the generally accepted cutoff for statistically meaningful modifications (Kline 2015). Accordingly, we did not include any of these omitted paths in Fig. 1.