Does financial reporting misconduct pay off even when discovered?

By construction, our subsamples do not overlap. As described in the sample selection table later in the paper, we exclude SEC enforcement cases from the lawsuit subsample and exclude SEC enforcement cases and lawsuits from the restatement subsample.

The following example illustrates the process that we use in each instance. One of the cases in our SEC enforcement subsample involves OCZ Technology, Inc., where Chief Financial Officer Arthur Knapp was charged with accounting fraud. Knapp lost an estimated $3,305,104 of wealth through stockholding, as OCZ’s stock price dropped from a peak value of $10.6 during the violation period to $1.16 after the initial revelation of the misconduct. He also lost an estimated $1186,613 of future earnings, as his employment at OCZ was terminated, although his forgone earnings were partially offset by the earnings from a new job, estimated to be $212,000. Moreover, he had to pay disgorgement and fines totaling $130,000. However, Knapp also derived benefits from his misconduct. His gain from incentive-based compensation during the violation period was $100,000. He earned a profit of $414,786 by trading stock and options before the revelation of the misconduct. He also gained $2,419,308 from his stockholding in OCZ, as the stock price increased during the period of inflated revenues. Overall, netting out all the benefits and costs results in a decrease in wealth of $1,475,623. In some other cases in our sample, the benefits from a misconduct outweigh the costs of getting caught, resulting in an overall increase in wealth.

For instance, to address the concern that a perpetrator would have earned part of the incentive payments even without engaging in misconduct, we estimate the benefits from misconduct excluding gain from performance-based compensation. To address the concern that unrealized gain via stockholding should not be counted as profit from engaging in misconduct, we estimate benefits from misconduct excluding gain in wealth via stockholdings. In subsequent robustness tests, we also vary the estimation method for loss in wealth via stockholdings and change the assumptions on the counterfactual scenario, absent misconduct.

The discussant mentioned that the percentage of perpetrators fired due to the misconduct in our sample (39%) is much lower than the 93% dismissal rate documented by Karpoff et al. (2008). Note that the Karpoff et al. (2008) sample consists of only AAERs, which are, by definition, more serious than the misconduct in our full sample, which includes SEC enforcement actions, class action lawsuits, and restatements.

The 31% detection rate claimed by Dyck et al. (2017) is not without controversy. This inference is based on the ratio of the following two statistics in Table 3 of Dyck et al. 2017: 2.82% of former Arthur Andersen (AA) clients had frauds detected by auditors that took over after AA disappeared, relative to 0.88% of non-AA clients whose frauds were discovered by other auditors (31% = 0.88/2.82). But other data reveal different detection rates. For instance, Krishnan et al. (2007) document that these ratios of detection of “fraud” by auditors who take on AA clients are a function of firm size. In particular, the percentage of clients that are given going concern opinions by auditors that took on former AA clients, relative to those given by auditors to non-AA clients, is (i) 6.2% to 14.5% for clients with assets less than $250 million, (ii) 5.8% to 5.1% for clients with assets between $250 million and $1 billion, and (iii) 5.6% to 2.3% for clients with assets above $1 billion. In other words, the Dyck et al. (2017) result is potentially applicable only to very large firms. Having said that, re-examining the Dyck et al. (2017) paper is beyond the scope of our current project. Hence we take the 31% detection rate as a tentative given in this paper. To provide evidence on the sensitivity of our inferences to the detection rate, Table 7 of reports the results when detection rates vary from 5% to 95%.

However, the stated equilibrium assumes that managers actually know the objective probability that their misreporting would be detected. The literature (e.g., Sah 1991) has long argued that the probability of detection is subjective to the manager, as base rates of detection, especially for the specific kind of manipulation that each manager indulges in, are not widely known. Hence the expected cost-benefit calculation is most likely a subjective assessment conducted by each manager in our sample. The expected costs and expected benefits are unlikely to be same for every manager in a larger sample we present. Therefore, in equilibrium, some managers will misreport as long as they expect their personal costs to be lower than their expected personal benefits.

We do not include cases that allege violations of Section 11 of the 1933 Securities Act. Section 11 provides for liability against issuers who make material misstatements or omissions in a registration statement. Section 11 is not an anti-fraud section and usually appears in lawsuits involving initial public offerings.

We do not include cases that allege violations of Section 12(j) of the 1934 Securities Exchange Act. Firms are usually charged under Section 12(j) when they have a history of delinquent periodic reports and filings. This does not necessarily indicate fraud or intentional misconduct.

An enforcement action is typically announced in both legal proceedings and administrative proceedings if it involves a trial or settlement.

The Department of Justice has the authority to bring criminal charges, while the SEC is only authorized to bring civil enforcement actions. Fraud cases for which the SEC issued final judgments could still be subject to separate investigation by the DOJ.

We have excluded eight firms (19 individuals) from the sample on account of DOJ enforcement. Assuming all these are negative net wealth effect cases, the percentage of positive net wealth effect cases would become 31.7% after including these DOJ cases, instead of the 32.9% reported in the main body of the paper.

A company is required to disclose the compensation package for the top five most highly paid executives in its annual proxy statements. Because some of the perpetrators are not among the highest paid executives in their companies, their compensation is not available in the proxy statements.

Within the 100 lawsuits, 39 cases are settled, 14 are ongoing, and 47 are dismissed. We include dismissed cases in our sample, because the extremely negative stock returns on the announcement date suggest that these lawsuits are likely to involve nonfrivolous allegations. However, the dismissed status also suggests a lack of factual support for the allegation. If we restrict our lawsuit subsample to settled or ongoing lawsuits, the percentage of perpetrators who experience positive net wealth effect becomes 33.6%, as compared to 32.8% using the entire lawsuit subsample.

We use Exposure_end_date from Audit Analytics Corporate and Legal database to define both the end of violation and initial revelation date for firms in our class action lawsuit subsample. The data manual states that Exposure_end_date is the end of the violation period and typically the date when the misconduct first becomes public knowledge.

We ignore compensation clawbacks because they are rarely enforced. In our experience, directors and officers’ insurance (D&O) covers legal fees. However, whether our accused managers are covered by D&O insurance is unknown, as the SEC does not require disclosure of D&O insurance for U.S. firms. Our understanding is that D&O insurance in public companies typically covers legal fees unless the manager is convicted of fraud in a court of law. We rarely observe such convictions in practice.

We vary these assumptions in robustness tests. For example, we assume that perpetrators would have received bonus and stock and option awards without engaging in misconduct and that their firms’ stock prices would have changed at the rate of value-weighted market return. The benefits due to misconduct are computed as the actual benefits minus the benefits perpetrators would have received, absent misconduct.

If the number of shares sold is smaller than the number of shares obtained through option exercises in the same Form-4 filing, we assume that the perpetrator sold some of the shares obtained through option exercises. If the number of shares sold is greater than the number of shares obtained from option exercises, we treat the excess number of shares sold as stock sales.

A skeptic may object to our assumption that managers would keep unrealized gains on their stockholdings if a fraud is not discovered. If the delayed bad news eventually comes out, even without the discovery of fraud, there would still be an associated loss, even in the absence of fraud detection. To address this issue, we conduct a sensitivity test, reported in section 5.5, where we group the unrealized gain and unrealized loss together and only consider the net change in the value of the managers’ stockholdings. Moreover, we could not find evidence in the literature for the conjecture that “the bad news eventually comes out.” If anything, legal experts that we have consulted suggest that managers misreport or commit fraud to delay bad news until something good happens, such that the bad news is offset against the good news. Consistent with such conversations, Graham et al. (2005) find that CFOs of underperforming firms state that they are more likely to delay release of bad news.

In Section 3.1.2, we identify the number of shares out of each stock acquisition that have later been sold. The number of unsold shares is simply the remaining number of shares from each stock acquisition. We use the number of unsold shares to compute gain in wealth via stockholding because this avoids double-counting the trading gain from selling stocks and options.

We determine the acquisition price of each share using the same method we describe in Section 3.1.2. For shares that perpetrators already own on the violation beginning date, we assume the purchase price to be equal to the firm’s stock price on the violation beginning date. For shares acquired through open market purchases during the violation period, the purchase price is equal to the stock price on the day of acquisition. For shares acquired through restricted stock awards during the violation period, the purchase price is zero. For shares acquired through option exercises, the purchase price is the option exercise price.

Some perpetrators (195 out of 499) did not report any trading during the violation period. We hand-collect the number of shares they own from the latest proxy statement filed before the initial revelation date and assume their stockholding to stay constant between the violation beginning date and initial revelation date. For these perpetrators, the gain in wealth via stockholding is computed as the number of shares held times the difference between peak-date price and the price on the violation beginning date.

Some firms (4 out of 227) were delisted prior to the initial revelation date, we substitute the price one day after initial revelation date by the delisting price.

Studies have used different time windows around the event date to identify forced turnover. Hennes et al. (2008) find that the termination of executives starts half a year before the restatement date and the number of executives fired becomes constant around one year afterward. Karpoff et al. (2008) document that the termination of executives happens through the SEC enforcement process, which lasts for 57 months on average. From Figure 2 in their paper, most of the executives are terminated within (−1, +2) years surrounding the announcement date.

Parrino (1997) and Jenter and Kanaan (2015) assume that turnover is voluntary if the executive was above the age of 60. We use age 65 as the cutoff because many companies have mandatory retirement policies that require CEOs to retire at the age of 65.

We classify an executive as fired in the case of bankruptcy because the executive would have lost his or her future earnings. Only six firms in our sample that went bankrupt after the misconduct was revealed.

Conference participants suggested that we ought to account for the possibility that (i) executives’ salaries grow much faster than the average inflation rate of 2% per year and that (ii) executives may move to higher positions in other companies as they become more senior, instead of staying in the same company and position, as assumed here. Therefore forgone earnings would be higher than the current estimation. We acknowledge this perspective but prefer to stay away from speculating what the exact rate of salary increase, future employment prospects, or both might be.

When estimating the new average annual compensation, we omit the first year the perpetrator starts working at the new public company, because there is often an unusually high amount of bonus or option awards that will result in an overestimation of the new annual compensation.

We have not used explanatory variables that indicate the strength of a firm’s corporate governance because the firms in our sample are too small to be included in standard corporate governance databases. ISS Riskmetric, for example, only covers S&P 1500 firms. As a result, we do not have corporate governance data for majority of the firms in our sample.

Following Becker (1968), to derive the breakeven probability of getting caught, we compare the expected benefit accruing to a manager from misconduct with the expected sanction. The expected sanction, in turn, is a function of the probability of getting caught and the magnitude of sanction imposed once the manager is caught.