Whistleblower laws and corporate fraud: Evidence from the United States

Accounting Forum xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Accounting Forum journal homepage: www.elsevier.com/locate/accfor

Whistleblower laws and corporate fraud: Evidence from the United States☆ ⁎

Adriana S. Cordis , Elizabeth M. Lambert Department of Accounting, Finance, and Economics, Winthrop University, United States

AR TI CLE I NF O

AB S T R A CT

Keywords: Corporate fraud Whistleblower Ethics Qui tam

We use data from the United States to assess whether whistleblower laws that protect private employees from retaliation have an impact on corporate fraud. Currently, eighteen states have whistleblower laws that offer such protection. Our analysis indicates that, in these states, a higher awareness of whistleblower laws is associated with a lower state-level conviction rate for corporate fraud. This finding is consistent with the hypothesis that whistleblower laws that cover private employees have a deterrent effect on corporate fraud, and that awareness of the provisions of whistleblower laws plays a key role in determining their effectiveness as a policy tool.

1. Introduction Whistleblowing is widely believed to play an important role in uncovering government and corporate malfeasance.1 Consequently, many international organizations have become strong advocates for laws that protect whistleblowers from retaliation. Some prominent examples include Transparency International, which has published a set of “recommended draft principles for whistleblowing legislation,” and the Council of Europe, which has called upon member states to establish “a normative, institutional and judicial framework to protect individuals who, in the context of their work-based relationship, report or disclose information on threats or harm to the public interest.”2 In the same vein, a major Australian research project on whistleblowing has recently called for strengthening protections, noting that the preliminary findings of the project “point to a need for further reform and stronger oversight in the public sector, and especially confirm that for the private and not-for-profit sectors, a well-informed legislative overhaul is overdue.”3

☆ This research was conducted while Elizabeth Lambert was a Winthrop McNair scholar. We thank the McNair Scholars Program for generous financial support. We also thank an anonymous referee and participants at the 2017 American Accounting Association Southeast Region Meeting for helpful comments and suggestions. ⁎ Corresponding author at: College of Business Administration, Winthrop University, 209 Thurmond Building, Rock Hill, SC 29733, United States. E-mail address: [email protected] (A.S. Cordis). 1 The term “whistleblower” made its way into the popular lexicon largely due to the actions of U.S. consumer activist Ralph Nader. In January 1971, he and a number of associates gathered in Washington D.C. for a “whistleblower’s conference” that showcased government employees who had publicly exposed government activities that conflicted with the public interest. Although the conference focused on government employees, whistleblowing has taken on a broader meaning in the intervening years. One widely-cited definition, which is due to Near and Miceli (1985), holds that whistleblowing is “the disclosure by organization members (former or current) of illegal, immoral or illegitimate practices under the control of their employers, to persons or organizations that may be able to effect action.” 2 See Transparency International (2009) and Council of Europe (2014) for details. 3 See Brown et al. (2016) for details.

http://dx.doi.org/10.1016/j.accfor.2017.10.003 Received 5 March 2017; Received in revised form 10 June 2017; Accepted 14 October 2017 0155-9982/ © 2017 Elsevier Ltd. All rights reserved.

Please cite this article as: Cordis, A.S., Accounting Forum (2017), http://dx.doi.org/10.1016/j.accfor.2017.10.003

Accounting Forum xxx (xxxx) xxx–xxx

A.S. Cordis, E.M. Lambert

In this paper, we investigate whether whistleblower laws that protect private employees from retaliation are an effective policy tool for deterring corporate fraud. Such laws have garnered considerable attention in the United States in the wake of the numerous accounting scandals of recent years.4 The actions of companies such as Enron, WorldCom, and HealthSouth have caused investors to lose billions of dollars, and have shaken investor confidence in financial statements and auditors (Hobson, Mayew, & Venkatachalam, 2012). Despite the important role of independent auditors in supporting well-functioning capital markets, very few jurisdictions impose a duty on auditors to report irregularities to regulatory authorities.5 Whistleblower laws hold the potential to complement the deterrent effect of external audits by encouraging employees to expose fraud and other forms of corporate malfeasance. Not surprisingly, academic studies of whistleblowing indicate that it is far from a riskless endeavour. Dyck et al. (2010), for example, examine 216 cases of alleged corporate fraud, and report that employees who helped to expose the alleged fraud declined to identify themselves in 45% of the cases to avoid suffering adverse consequences. For the cases in which the employees were named, the majority reported that they were fired or demoted as a result of exposing the alleged fraud. These findings suggest that protection from workplace retaliation should help to incentivize employees to come forward with information about fraudulent business activities.6 We develop empirical evidence on the relation between whistleblower protections and corporate fraud using data from the United States. These data are well suited to such an undertaking due to the multi-jurisdictional aspects of the U.S. legal system. Although there are federal statutes that protect whistleblowers in the public sector, the individual states are free to enact their own whistleblower laws provided that these laws do not conflict with federal law. Most states appear to view whistleblower anti-retaliation measures as a key mechanism for deterring wrongdoing (Callahan and Dworkin, 2000), and all fifty states have some form of whistleblower law in place. But the specifics of these laws, such as type of whistleblower protected, the appropriate disclosure recipient, and remedies for retaliation, display wide variation across states. By exploiting the state-level variation in statutory whistleblower protection, we can assess the impact of whistleblower laws on the prevalence of corporate fraud. Our basic hypothesis is that these laws should help to deter fraud by increasing the likelihood that it will be detected and punished. But we expect the effectiveness of the laws to be a function of several factors. First, laws that offer protection to private employees should be more effective at deterring corporate fraud than those that focus only on public employees. Second, managers and other employees must be aware of the law and its provisions in order for it to be effective. Third, the effectiveness of a law should depend on its overall strength, as determined by the combination of all its features. We use three variables to capture these factors. The first is an indicator variable that identifies the 18 states that protect private employees from retaliation.7 The remaining two variables are designed to measure the strength of whistleblower laws, and the level of awareness with respect to the laws. We use internet searches to construct the awareness variable, as in Goel and Nelson (2014). To measure the strength of whistleblower laws, we use an index that is published by the Public Employees for Environmental Responsibility (PEER) organization. Our measure of corporate fraud is constructed using information compiled from the Executive Office for U.S. Attorneys (EOUSA) by the Transactional Records Access Clearinghouse (TRAC) at Syracuse University. The TRAC database lists the number of corporate fraud prosecutions filed and the number of corporate fraud convictions by state judicial district beginning with 2003. We aggregate the judicial district data to obtain a measure of corporate fraud at the state level, and divide by the number of non-farm employees in the state to obtain a conviction rate. We conduct the analysis using conviction data for the years 2003–2015. The research design is straightforward. We estimate cross-section regressions of the state-level conviction rates for corporate fraud on the three variables discussed above, and include an interaction variable to allow for the possibility that the effect of the antiretaliation provisions of whistleblower laws differs across states with high and low awareness. The regression evidence is consistent with our hypothesis regarding the deterrent effect. Specifically, we find that a higher awareness of whistleblower laws is associated with a lower conviction rate for corporate fraud for states whose whistleblower laws have provisions that protect private employees. This evidence is robust to the use of a variety of control variables. We also consider the impact of state qui tam laws, but find little evidence of a deterrent effect with respect to corporate fraud.8 This finding is not surprising in view of the limited scope of such laws. Our results on the deterrent effect of whistleblower laws complement the recent findings of Wilde (2017), who investigates the effect of whistleblowing by employees of U.S. corporations on the behaviour of these corporations in subsequent time periods. His analysis suggests that allegations raised by whistleblowers have a deterrent effect on both financial misreporting and tax aggressiveness, and that this effect can be observed for at least two years after the year in which the allegation was made. Our analysis is fully consistent with this finding, and suggests that strengthening state-level whistleblower laws is an effective way for U.S.

4 The interest in such laws appears to be growing in other countries as well. Whistleblower laws in Australia, for example, have historically focused on public employees. However, in the Draft Open Government National Action Plan for 2016–2018, the Prime Minister and Cabinet commit to improving “whistle-blower protections for people who disclose information about tax misconduct to the Australian Taxation Office,” and state that they will “consult on other reform options to strengthen and harmonise whistle-blower protections in the corporate sector with those in the public sector.” See https://www.pmc.gov.au/resource-centre/publicdata/australias-first-open-government-national-action-plan-2016-18-fact-sheet. 5 South Africa is perhaps the best example of a jurisdiction in which external auditors are duty-bound to blow the whistle on wrongdoing by their clients. Marouna and Solomon (2014) argue that this reporting duty enhances the perceived legitimacy of the audit profession. 6 Prior research suggests that the likelihood of whistleblowing depends on a number of factors, including organizational policies and procedures (Seifert, Sweeney, Joireman, & Thornton, 2010), the presence of monetary and reputational incentives (Dyck, Morse, & Zingales, 2010), the monitoring role of the press (Miller 2006), and the regulatory environment (Maroun and Solomon 2014). For a detailed review of fraud-related literature, see Trompeter et al. (2013) and Trompeter et al. (2014). 7 State whistleblower laws typically do not provide for pecuniary compensation. 8 These laws, which are specific to the United States, allow whistleblowers (private citizens) to file lawsuits against companies that are defrauding the state government. If the state recovers funds as a result of the lawsuit, the whistleblower is entitled to receive a share of the funds.

2

Accounting Forum xxx (xxxx) xxx–xxx

A.S. Cordis, E.M. Lambert

policymakers to combat corporate fraud. More broadly, we believe that the United States’ experience with whistleblower laws provides valuable policy lessons for other countries. We began the discussion by noting that many international organizations, such as the Council of Europe, have become strong advocates for laws that protect whistleblowers from retaliation. Our findings highlight the benefits of extending such protections to private employees. We hope that these findings will serve to inform the legislative process in both the United States and other countries with respect to the expansion of whistleblower protections in the private sector, and the efficacy of such protections in combating corporate malfeasance. The remainder of the paper is organized as follows. Section 2 provides some background on U.S. whistleblower laws and presents our hypotheses. Section 3 describes the data and empirical methodology. Section 4 presents the main empirical results along with those obtained from several robustness tests. Section 5 contains a few concluding remarks. 2. Background and hypotheses development The landscape of whistleblower laws in the United States is quite diverse. Prior to the U.S. Civil Service Reform Act of 1978, which explicitly recognized the legitimacy of whistleblowing by federal government employees, very few states offered any kind of statutory protection for whistleblowers. Over the next two decades, whistleblowing gradually came to be seen in a more favourable light, and many states enacted their own whistleblower laws. Because there was no consensus among the states as to specific provisions that should be included in such laws, each state’s law reflects its own legislative priorities. In the state of Connecticut, for example, the process began with a 1980 court case in which the State Supreme Court recognized an exception to the employment-at-will doctrine, thereby allowing employees (public or private) to bring “a cause of action for wrongful discharge where the discharge contravenes a clear mandate of public policy.”9 Two years later the state legislature enacted a law that protects employees who disclose illegal activities or unethical practices by their employers. Michigan also enacted its law in 1980, protecting employees (public or private) who report violations of the law, or who take part in court actions. However, a number of other states, such as Illinois, did not enact their whistleblower laws until much later. Most state whistleblower laws protect employees who report wrongdoing from retaliation. But a few states also have provisions in their laws that allow whistleblowers to receive compensation.10 Of particular note, the majority of state laws protect public employees only. Callahan and Dworkin (2000) attribute this fact to “the prevailing view that society has a more significant interest, and a greater stake, in reports of misconduct in government agencies than in the private sector.” In total, 18 states enacted new or amended existing laws between 1980 and 2003 to offer general protection to private employees who expose misconduct. This aspect of the variation in whistleblower laws across states is central to our research design, because our analysis focuses on the question of whether expanding whistleblower protections to private employees helps to deter corporate fraud. 2.1. Hypotheses development The well-known theory of crime deterrence developed by Becker (1968) predicts that fraud can be deterred by increasing the probability that it is detected. Under this theory, the decision to engage in criminal activity, such as fraud, entails a cost-benefit analysis. If the expected benefit from the crime is less than the expected cost (i.e., the sanction multiplied by the probability of getting caught), then a rational individual will choose not to engage in crime. In general, whistleblower laws should increase the probability of detecting fraud because individuals who have knowledge of wrongdoing should be more likely to report it if they are afforded legal protection. But assessing the impact of whistleblower laws on corporate fraud is less straightforward than it might seem. It is tempting to argue that states that afford strong whistleblower protection to private employees should have less corporate fraud than those that afford weak or no protection to these employees. However, this argument ignores the potential for reverse causality. To draw an analogy, suppose we believe that a strong police presence deters crime, but we observe that municipalities with a high number of police officers per capita have higher crime rates than those with a low number of police officers per capita. This is not necessarily inconsistent with a deterrent effect, because municipalities with high levels of criminal activity may be driven to hire more police officers than other municipalities on a per capita basis. Similarly, the provisions of state whistleblower laws that afford strong protection to private employees may have been put in place in response to high levels of fraud in the private sector. Callahan and Dworkin (1994) argue, for example, that legislators enact whistleblower laws because they view whistleblowing as a means to combat fraud, misuse of funds, and other wrongdoing. Because of the potential for reverse causality, we frame our analysis of the deterrent effect of whistleblower laws in terms of the awareness of the law among state residents. Suppose a state affords strong whistleblower protection to private employees, but these employees are largely unaware of the provisions of the law. It stands to reason that the deterrent effect of the law should be relatively weak under these circumstances. We argue, therefore, that the extent to which a whistleblower law helps to deter corporate fraud varies with both the specific provisions of the law and the awareness of these provisions among state residents. This leads us to our first hypothesis. Hypothesis 1. Among states that afford whistleblower protection to private employees, the level of corporate fraud is a declining function of the level of awareness of the law among state residents. 9

See https://h2o.law.harvard.edu/cases/3555 for more information. Illinois, Florida, Oregon, South Carolina, and Wisconsin have such provisions. Only Illinois and Florida provide substantial compensation.

10

3

Accounting Forum xxx (xxxx) xxx–xxx

A.S. Cordis, E.M. Lambert

Note that this hypothesis focuses on the interaction between provisions of the law that offer protection to corporate whistleblowers and the extent to which potential whistleblowers are aware of these provisions. It posits that the marginal effect of protecting private employees on corporate fraud varies with the awareness of the law because deterrence requires knowledge of the provisions of the law. Testing this hypothesis is the main focus of our empirical analysis. But we also consider whether state qui tam laws have an impact on corporate fraud. These laws mirror the provisions of the Federal False Claims Act, which allows whistleblowers to file lawsuits on behalf of the federal government. Specifically, state qui tam laws allow whistleblowers to file lawsuits against companies that are defrauding the state. If the state recovers funds as a result of the lawsuit, the whistleblower is entitled to receive a share of the funds. Medicare fraud is the most common type of fraud addressed by state qui tam laws. However, a number of states have laws that apply to a broad range of state-funded programs. The possibility of receiving large monetary awards is designed to act as a powerful incentive for potential whistleblowers. This suggests that qui tam laws might be effective in deterring corporate fraud. On the other hand, the scope of the laws is quite narrow. They only apply to corporations that do business directly with the state government. This suggests that their impact on the overall level of corporate fraud at the state level might be fairly small. But it is clearly possible that they have a non-negligible deterrent effect. This leads us to our second hypothesis. Hypothesis 2. Among states that have qui tam laws, the level of corporate fraud is a declining function of the level of awareness of the law among state residents. 3. Data and methodology Many different acts are commonly classified as corporate fraud. Some common examples include misuse of accounts, financial accounting misstatements, and fraudulent expense claims. But measuring corporate fraud is fraught with challenges because the true underlying level of fraud is inherently unobservable. Our approach to measuring fraud mirrors that employed in studies of public corruption. Although the true underlying level of public corruption is unobservable, researchers have documented the effect of a wide range of variables on corruption using data on convictions for corruption (see, e.g., Brown, Smith, White, & Zutter, 2015; Campante and Do 2014; Glaeser and Saks 2006; Smith 2016). We follow suit by using data on convictions for corporate fraud to conduct our analysis of the deterrent effect of state whistleblower laws. 3.1. Corporate fraud data We construct our measure of corporate fraud using the TRACFed database maintained by the Transactional Records Access Clearinghouse (TRAC), a data gathering, data research, and data distribution organization at Syracuse University.11 This database includes detailed information on white collar crimes obtained from the EOUSA. White collar crime is defined as criminal prosecutions of nonviolent crimes involving deceit, concealment, subterfuge, and other fraudulent activity. TRAC data have been used in a number of studies in the accounting, finance, and economics literature (e.g., Cordis and Warren 2014; Guedhami and Pittman 2008; Hanlon, Hoopes, & Shroff, 2014; Hoopes, Mescall, & Pittman, 2012), by government bodies (e.g., Ginsberg 2014; Grassley 2013), and by high-profile media outlets such as The New York Times (e.g., Tillman and Pontell, 2016),12 The Washington Post (e.g., Rampell, 2016),13 or The Wall Street Journal (e.g., Rubenfeld, 2015).14 TRAC employees follow a number of strategies to ensure data quality, including comparisons of aggregate data counts, comparisons of records across multiple databases, and record-by-record matching procedures.15 We focus on the subset of white collar crimes that are classified by prosecutors as corporate fraud. The data consist of the number of corporate fraud prosecutions filed and the number of corporate fraud convictions. These data are reported by state judicial district beginning in 2003. We aggregate the judicial district data to obtain a measure of corporate fraud at the state level. Acts that are classified as corporate fraud include false statements or entries, concealment of assets, bank fraud, fraudulent claims, bankruptcy fraud, securities violations, and so forth. Our measure of corporate fraud for a given state is the total number of corporate fraud convictions over the years 2003–2015 divided by the non-farm employment (in millions) for the state. As a robustness check, we also use the total number of corporate fraud prosecutions filed over these years divided by the non-farm employment. The number of corporate fraud convictions per million employees ranges from zero to 2.575. The ten states that have the highest number and the lowest number of corporate fraud convictions per million employees are listed in Table 1. New York leads the states with the highest number of convictions per million employees. Eight states have no corporate fraud convictions over the sample period: Alaska, Hawaii, Idaho, Montana, North Dakota,

11

We obtained the data under license from TRACFed (http://tracfed.syr.edu/). “Corporate Fraud Demands Criminal Time,” available at http://www.nytimes.com/2016/06/29/opinion/corporate-fraud-demands-criminal-time.html?emc= eta1&_r=1. 13 “Fear (or lack thereof) of getting caught by the tax man,” available at https://www.washingtonpost.com/news/rampage/wp/2016/08/31/fear-or-lack-thereofof-getting-caught-by-the-tax-man/?utm_term=0.34bfe4963fa6. 14 “Corruption Currents: White-Collar Crime Prosecutions Reach 20-Year Low,” available at http://blogs.wsj.com/riskandcompliance/2015/07/31/corruptioncurrents-white-collar-crime-prosecutions-reach-20-year-low/. 15 See, for example, Long et al. (2004) for a detailed description of strategies used by TRAC employees to ensure data quality. 12

4

Accounting Forum xxx (xxxx) xxx–xxx

A.S. Cordis, E.M. Lambert

Table 1 States with Most and Least Corporate Fraud Convictions per Million Non-Farm Employment (2003–2015). Most Convictions

Least Convictions

State

Convictions per Million Non-Farm Employment

State

Convictions per Million Non-Farm Employment

New York Alabama Connecticut Vermont Rhode Island Minnesota Ohio Nebraska Utah Florida

2.575 2.299 2.136 2.033 1.680 1.625 1.552 1.427 1.302 1.293

Nevada Arizona Alaska Hawaii Idaho Montana North Dakota New Mexico South Dakota Wyoming

0.054 0.026 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

New Mexico, South Dakota, and Wyoming.16 3.2. Data on state whistleblower laws We use four variables to capture key aspects of the variation in whistleblower laws across states. Two of these variables, Whistleblower Law (WB Law) and Qui Tam Law reflect the presence or absence of specific provisions in the state laws. The third variable, Whistleblower Law Awareness (WBL Awareness), is a measure of awareness about whistleblower laws, and the fourth variable, Whistleblower Law Strength (WBL Strength), is a measure of strength of the whistleblower laws. The definitions of these variables are as follows. The Whistleblower Law variable is simply an indicator that equals one for the states with laws that protect private employees who blow the whistle from retaliation. Currently, eighteen states have whistleblower statutes that afford such protection.17 Anti-retaliation provisions should encourage private employees to expose misconduct, thereby deterring corporate fraud. As discussed earlier, however, we do not necessarily expect a negative correlation between Whistleblower Law and corporate fraud convictions in view of potential reverse-causality considerations. Similarly, the Qui Tam Law variable is an indicator that equals one for the states that have such laws. Qui Tam laws are also known as False Claims Acts. The federal False Claim Act was originally enacted in 1863 and substantially strengthened in 1986 and then again in 2009. It provides considerable monetary incentives to whistleblowers that bring lawsuits in the name of the government against individuals and companies engaged in fraud (Callahan and Dworkin, 2000). Whistleblowers are awarded up to 30% of all the money recovered (plus reasonable expenses and attorney’s fees) if they prosecute the claim themselves, and up to 25% of the money recovered if the government joins in the lawsuit. Twenty-nine states have adopted False Claim Acts similar to the federal law.18 The Whistleblower Law Awareness variable is designed to capture the general level of whistleblower law awareness within each state. We use the methodology pioneered by Goel and Nelson (2014) in their study of whistleblower laws to construct this measure. Specifically, we use two Internet search engines — Google and Yahoo — to record the number of hits obtained for each state when searching for whistleblower laws. Because the search results include “official publications and web postings about whistleblower laws, newspaper articles, advertisements and opinions by law firms on the subject, academic references, media reports of related court cases, etc.,” Goel and Nelson (2014) argue that the resulting measure “can be interpreted as capturing the diffusion of awareness about whistleblower provisions.” Similar hit-based awareness measures have been used in other contexts (see, e.g., Goel, Nelson, & Naretta, 2012). We construct our awareness measure in two steps. First, we search for the state name in conjunction with “whistleblower law,” e.g., “Alabama whistleblower law.” Second, we use a modified set of keywords to narrow the results returned for a set of 12 states. This step is necessary because the basic search results for these states pulls in extraneous information. For example, some of the results obtained by searching for “Virginia whistleblower law” are actually for West Virginia. Because we follow the Goel and Nelson (2014) procedure as closely as possible, we refer the interested reader to their paper for a very detailed description of the full process. All searches were conducted on the same day in May 2016. The Whistleblower Law Awareness variable is the average of the Google and Yahoo search hits on a per capita basis. It has an average value of 0.27 across all states. Vermont has the highest level of awareness 16 It is highly unlikely, of course, that these states have no corporate fraud. Clearly our convictions-based measure is a noisy proxy for the true underlying level of corporate fraud, because some fraud is never detected (or prosecuted). Despite this shortcoming of convictions-based measures of crime, they have been successfully used in a host of studies in the literature (see, e.g., Alt and Lassen, 2008; Brown et al., 2015; Campante and Do, 2014; Cordis and Milyo, 2016; Glaeser and Saks, 2006; Smith, 2016). 17 These states are Arizona, California, Connecticut, Florida, Hawaii, Illinois, Louisiana, Maine, Michigan, Minnesota, Montana, New Hampshire, New Jersey, New York, Ohio, Oregon, Rhode Island, and Tennessee (source: www.ncsl.org and www.uslegal.com, last Accessed May 2016). 18 These states are California, Colorado, Connecticut, Delaware, Florida, Georgia, Hawaii, Iowa, Illinois, Indiana, Louisiana, Massachusetts, Maryland, Michigan, Minnesota, Montana, North Carolina, New Hampshire, New Jersey, New Mexico, Nevada, New York, Oklahoma, Rhode Island, Tennessee, Texas, Virginia, Washington, and Wisconsin.

5

Accounting Forum xxx (xxxx) xxx–xxx

A.S. Cordis, E.M. Lambert

(1.57) and Ohio has the lowest level of awareness (0.03). The Whistleblower Law Strength variable is an index developed by the Public Employees for Environmental Responsibility (PEER) organization. PEER rates the coverage, usability, and strength of each state’s law on a 100-point scale, and publishes a “whistleblower report card” based on the ratings.19 Strong laws have broad coverage of disclosures, well-defined protections for whistleblowers, and clear remedies for retaliation. California, Tennessee, and Nevada have the highest ratings on the strength dimension; Georgia, Indiana, and South Dakota have the lowest ratings. 3.3. Controls We recognize that the number of corporate fraud convictions per capita is likely to be influenced by factors other than our main variables of interest. Accordingly, we use a couple of key control variables in our baseline empirical specifications in order to mitigate omitted variables bias. In addition, we assess the impact of including more controls as part of our robustness checks. Our first control variable, Income Inequality, is a measure of income inequality developed by Frank et al. (2015). Numerous studies link income inequality to crime levels (see, e.g., Brush 2007; Choe, 2008; Chiu and Madden, 1998; Fajnzylber, Lederman, & Loayza, 2002; Kelly, 2000). Braithwaite (1991) argues that inequality explains both street crimes, which are motivated by need, and whitecollar crimes, which are motivated by greed. Our income inequality variable is the income share of the top 1% of the population in the state. The average top 1% income share across all states is 15.35%. Connecticut, New York, and Nevada, have the highest levels of inequality. Iowa, North Dakota, and Alaska have the lowest levels. Our second control variable, EOUSA employment per capita, is designed to capture the variation in law enforcement effectiveness across states. We examine this variable because increasing law enforcement resources could deter employees from engaging in corporate fraud. Indeed, Baer (2008) argues that one way for government agencies to deter corporate fraud is to increase the probability of detection by hiring more law enforcement agents. Our proxy for law enforcement effectiveness is the number of federal full time employees in the EOUSA for the state divided by the state population. Table 2 reports the data sources for the whistleblower law and control variables used in the analysis. It also provides summary statistics for these variables. 3.4. Regression specifications We use linear regression methods to assess the impact of state whistleblower laws on corporate fraud. Our baseline analysis is conducted as follows. First, we test Hypothesis 1 by estimating the model Fraud Convictionsi = α + β1WB Lawi + β2 WB Lawi × WBL Awarenessi + β3 WBL Awarenessi + β4WBL Strengthi + γ' Controlsi + εi,

(1)

where i is a state index and Controlsi is a column vector that contains our controls. The main coefficient of interest is β2. Hypothesis 1 predicts that this coefficient is negative. Next, we test Hypothesis 2 by estimating the model Fraud Convictionsi = α + β1 QuiTam Lawi + β2 QuiTam Lawi × WBL Awarenessi + β3 WBL Awarenessi + γ' Controlsi + ϵi.

(2)

Again, the main coefficient of interest is β2. Note that this specification does not include our measure of the strength of whistleblower laws because this measure does not focus on state qui tam laws. 4. Empirical results We present the results of our baseline regression specifications in Tables 3 and 4. The former is for the model in Eq. (1) and the latter is for the model in Eq. (2). The results reported in the first column of each table use corporate fraud conviction rates as the dependent variable (as shown in Eqs. (1) and (2)). We recognize, however, that the bar for convicting someone of fraud is high (proof beyond a reasonable doubt), so we also fit the model using corporate fraud filing rates (i.e., the number of fraud cases filed by prosecutors per million employees) as the dependent variable. This measure is broader than the conviction rate because it includes fraud cases that are ultimately dropped and those that result in acquittals. The results for the alternative model are reported in the second column of the tables. Because it turns out that the estimates look very similar across models, we focus our discussion on the results presented in the first column of each table. First consider the results in Table 3. The estimated coefficient on the Whistleblower Law variable (an indicator variable) is 0.96 with a t-statistic of 3.43. Thus the intercept in the regression specification displays a highly statistically significant upward shift for the 18 states whose whistleblower laws cover private employees. There are at least two potential explanations for this finding. One possibility is that the laws in these states encourage whistleblowing by private employees, leading to more corporate fraud being uncovered and hence higher conviction rates. Another possibility is that states that suffer from a lot of corporate fraud are simply more likely to afford whistleblower protections to private employees. Of course both of these could be true simultaneously. The estimated coefficient on the Whistleblower Law Awareness variable is also positive and statistically significant at the 5% level 19

Detailed information on how the ratings were developed is available at www.peer.org.

6

Total number of corporate fraud convictions divided by total nonfarm employment in millions, 2003–2015. Dummy variable that equals 1 if the state has a whistleblower law that protects private employees, 2016. Dummy variable that equals 1 if the state has a Qui Tam law, 2016. Internet-based measure of whistleblower law awareness (see text for details), 2016.

Index of coverage, usability, and strength of whistleblower laws. A higher index value implies a stronger law; the index ranges from 0 to 100, 2015. Interaction term, the product of the Whistleblower Law and Whistleblower Law Awareness variables. Interaction term, the product of the Qui Tam Law and Whistleblower Law Awareness variables. Income share of the top 1% of the population, 2003. Full time federal employees in the EOUSA per population (in millions), 2003.

Corporate Fraud Convictions per Employment Whistleblower Law

Whistleblower Law Strength

7 15.35 37.13

0.11

0.07

56.60

3.31 13.88

0.13

0.13

11.11

0.50 0.25

0.48

0.36 0.58 0.27

0.66

Standard Deviation

0.69

Mean

11.14 17.02

0.00

0.00

22.00

0.00 0.03

0.00

0.00

Min

25.51 75.48

0.52

0.52

78.00

1.00 1.57

1.00

2.58

Max

Frank et al. (2015) TRAC

Whistleblowerlaws.com and authors

USlegal.com and authors

Whistleblowerlaws.com Authors, based on Goel and Nelson (2014) PEER.org

USlegal.com

TRAC & BEA

Source

Notes: TRAC is the Transactional Records Access Clearinghouse; BEA is the Bureau of Economic Analysis; PEER is Public Employees for Environmental Responsibility; EOUSA is the Executive Office for U.S. Attorneys. N = 50.

Income Inequality EOUSA Employment per Capita

Qui Tam Law X Law Awareness

Whistleblower Law X Law Awareness

Qui Tam Law Whistleblower Law Awareness

Description

Variable

Table 2 Descriptive Statistics.

A.S. Cordis, E.M. Lambert

Accounting Forum xxx (xxxx) xxx–xxx

Accounting Forum xxx (xxxx) xxx–xxx

A.S. Cordis, E.M. Lambert

Table 3 OLS Model of Whistleblower Laws and Corporate Fraud in the United States. Dependent Variable

Whistleblower Law Whistleblower Law X Law Awareness Law Awareness Whistleblower Law Strength Income Inequality EOUSA Employment per Capita Constant N F-value R2

Corporate Fraud Convictions per Employment (1)

Corporate Fraud Filings per Employment (2)

0.96*** (3.43) −2.79*** (−3.66) 1.18** (2.49) −0.01 (−1.31) 0.05 (1.42) −0.01* (−1.99) 0.57 (0.81) 50 4.84*** 0.38

0.87*** (3.19) −2.36** (−2.63) 1.07** (2.04) −0.00 (−0.07) 0.02 (0.70) −0.01** (−2.20) 0.42 (0.67) 50 4.35*** 0.30

Notes: The dependent variable is the total number of corporate fraud convictions and the total number of corporate fraud filings, respectively, per total non-farm employment in millions over 2003–2015. Both models present the results of OLS regressions. ***, **, and * denote statistical significance at the 1%, 5%, and 10% level, respectively; robust t-statistics are reported in parentheses. All variables are defined in Table 2.

(1.18 with a t-statistic of 2.49). Note that this coefficient is the estimated marginal effect of awareness for states that do not afford whistleblower protection to private employees. It is therefore indirectly consistent with the hypothesized deterrent effect. To see why, consider a scenario in which private employees have no protection from retaliation if they blow the whistle on corporate fraud. If they know that this is the case, then they may be less likely to expose corporate fraud. For states whose whistleblower laws do not cover private employees, a higher awareness of the law means that private employees are more likely to know that they have no protection under the law. Thus high awareness may actually lower the likelihood that corporate fraud is detected, thereby undermining the deterrent effect of the law. The estimated coefficient on the interaction between the Whistleblower Law and Whistleblower Law Awareness variables is −2.79 with a t-statistic of −3.66. To see how to interpret this coefficient, note that it represents the shift in the estimated marginal effect of Whistleblower Law Awareness for the 18 states whose whistleblower laws cover private employees. In these 18 states, a higher awareness of whistleblower laws is associated with a lower conviction rate for corporate fraud. Specifically, the estimated coefficient on awareness for these states is 1.18–2.79 = −1.61, which is statistically significant at the 5% level. This finding lends support to Hypothesis 1. In other words, it is consistent with the view that whistleblower laws that cover private employees help to deter corporate fraud, and that awareness of the laws plays a crucial role in determining their effectiveness in this regard. The estimated coefficient on the Whistleblower Law Strength variable is −0.01 with a t-statistic of −1.31. It is therefore statistically indistinguishable from zero. As for the two controls, the estimated coefficient on Income Inequality has the anticipated sign, but is not significant at standard levels. The estimated coefficient on the EOUSA Employment per Capita variable also has the anticipated sign, and it is statistically significant at the 10% level. Now consider the results in Table 4. The estimated coefficient on the Qui Tam Law variable is 0.16 with a t-statistic of 0.55. So the shift in the intercept for the 29 states with qui tam laws is not statistically significant. Similarly, the estimated coefficient on the Whistleblower Law Awareness variable is statistically indistinguishable from zero (0.84 with a t-statistic of 1.55). Although these results contrast sharply with those obtained for whistleblower laws that cover private employees, they are not particularly surprising given the limited scope of qui tam laws. The estimated coefficient on the interaction between the Whistleblower Law and Whistleblower Law Awareness variables, which is the relevant quantity with respect to Hypothesis 2, is −1.35 with a t-statistic of −1.66. Thus the sign is consistent with that predicted by Hypothesis 2. But the magnitude is such that the estimate falls slightly short of being statistically significant at the 10% level. Note that using prosecutorial filings in place of convictions produces an estimated coefficient that is statistically significant at 10%. Nonetheless, the support for Hypothesis 2 is clearly very weak based on the available evidence. Overall the analysis suggests that the deterrent effect of state qui tam laws is much less pronounced than that of whistleblower laws that cover private employees. This is probably because most corporations are not engaged in business activities that fall within the scope of the qui tam laws.

4.1. Robustness checks One potential drawback of our research design is that we investigate the impact of state whistleblower laws that cover private employees separately from that of state qui tam laws (i.e., using separate regression models). We adopt this approach because of data limitations. With only 50 observations on state-level conviction rates for corporate fraud, it would be imprudent to include a large 8

Accounting Forum xxx (xxxx) xxx–xxx

A.S. Cordis, E.M. Lambert

Table 4 OLS Model of Qui Tam Laws and Corporate Fraud in the United States. Dependent Variable

Qui Tam Law Qui Tam Law X Law Awareness Law Awareness Income Inequality EOUSA Employment per Capita Constant N F-value R2

Corporate Fraud Convictions per Employment (1)

Corporate Fraud Filings per Employment (2)

0.16 (0.55) −1.35 (−1.66) 0.84 (1.55) 0.07 (1.65) −0.01* (−1.71) −0.11 (−0.14) 50 4.20*** 0.23

0.28 (0.96) −1.52* (−1.70) 0.95 (1.66) 0.04 (1.24) −0.01** (−2.29) 0.30 (0.55) 50 3.45** 0.20

Notes: The dependent variable is the total number of corporate fraud convictions and the total number of corporate fraud filings, respectively, per total non-farm employment in millions over 2003–2015. Both models present the results of OLS regressions. ***, **, and * denote statistical significance at the 1%, 5%, and 10% level, respectively; robust t-statistics are reported in parentheses. All variables are defined in Table 2.

number of explanatory variables in the regression models. This is why we fit separate regression models for the two types of whistleblower laws. That said, it is possible that our approach misses something important because it fails to capture the correlation between the two types of whistleblower laws. To investigate whether this is a legitimate concern, we estimate a comprehensive regression model that nests the models in Eqs. (1) and (2) as special cases (the detailed estimation results are omitted in the interests of space). This does not produce any major changes in our findings. The analysis still indicates a higher awareness of whistleblower laws is associated with a lower conviction rate for corporate fraud in the 18 states that provide protection to private employees, lending strong support to Hypothesis 1. Another, presumably larger, concern is that our estimates of the impact of whistleblower laws might be adversely affected by omitted variables bias. The most common approach for addressing this concern is to add more controls to the regression model. Although we are limited in our ability to expand the number of explanatory variables by the small number of observations in our dataset, it is reasonable to consider several additional controls. We therefore investigate the impact of three additional controls on our results: daily newspaper paid circulation per capita, the percentage of population aged 25 and up with a high school diploma or higher, and the gross state product per capita.20 Each of these variables has been linked to state-level variation in public corruption in the economics literature, suggesting that they may help to explain variation in corporate fraud across states. Table 5 presents the results of adding the new controls. The first column is for the model in Eq. (1), and the second column is for the model in Eq. (2). None of the additional controls has a statistically significant effect on the corporate fraud rate. The estimated coefficients on newspaper circulation have the anticipated sign, but both t-statistics are a little too small to be statistically significant at the 10% level. Note, however, that adding the new controls causes the R2 statistic to increase from 38% to 44% for Eq. (1) and from 23% to 32% for Eq. (2). Overall the addition of the new controls has little impact on our inferences regarding the impact of the two types of whistleblower laws. Of particular note, the estimated coefficient on the interaction between the Whistleblower Law and Whistleblower Law Awareness variables is −2.61 with a t-statistic of −3.36, and the estimated marginal effect of Whistleblower Law Awareness for the 18 states whose whistleblower laws cover private employees is 0.91–2.61 = −1.70. Hence the findings continue to lend support to Hypothesis 1. Although omitted variables bias can never be completely ruled out in most empirical settings, we see no reason to argue that it is a pressing concern with respect to our results.

5. Conclusions Recently, Ireland put in place one of the strongest and most sweeping whistleblower laws in Europe: the Protected Disclosures Act, 2014. A notable feature of the law is that it extends protection to employees in all sectors of the Irish economy. The lessons from our analysis of U.S. data suggest that this law draws on best practices in this regard. In particular, the evidence indicates that whistleblower laws that protect private employees reduce the prevalence of corporate fraud by increasing the probability that corporate malfeasance is detected and punished. This deterrent effect of whistleblower laws is broadly consistent with standard arguments from the economic theory of crime. 20

All of these variables are taken from the Statistical Abstract of the United States.

9

Accounting Forum xxx (xxxx) xxx–xxx

A.S. Cordis, E.M. Lambert

Table 5 Robustness Checks: Additional Controls. Dependent Variable: Corporate Fraud Convictions per Employment OLS (1) Whistleblower Law Whistleblower Law X Awareness

OLS (2)

0.91*** (3.43) −2.61*** (−3.36) −0.01 (−0.02) −0.84 (−1.00) 1.11** (2.20)

Qui Tam Law Qui Tam Law X Law Awareness Law Awareness Whistleblower Law Strength Income Inequality EOUSA Employment per Capita Newspaper Circulation per Capita High School Diploma Percentage Gross State Product per Capita Constant N F-value R2

1.44*** (3.12) −0.01 (−1.18) 0.05 (1.62) −0.02** (−2.60) 2.71 (1.61) −0.04 (−1.32) −0.39 (−0.32) 3.35 (1.31) 50 4.36*** 0.44

0.07* (1.78) −0.02** (−2.66) 3.27 (1.49) −0.04 (−1.36) 0.50 (0.42) 3.01 (1.15) 50 2.90** 0.32

Notes: The dependent variable is the total number of corporate fraud convictions per total non-farm employment in millions over 2003–2015. ***, **, and * denote statistical significance at the 1%, 5%, and 10% level, respectively.

We believe that there are at least two key policy implications that emerge from our results. First, more states should consider extending whistleblower protections to private employees. The reduction in corporate fraud that could potentially be achieved by adopting a more uniform standard of whistleblower protection across states is a worthy policy objective. Because our analysis highlights the crucial role of awareness of whistleblower laws on their deterrent effect, it also suggests that states could potentially benefit from public service messaging that increases public awareness of the protections that are afforded to whistleblowers in both the public and private sectors. Second, and more broadly, more countries should consider adopting whistleblower laws that provide broad protections to both public and private employees. There seems to be a recent trend on the part of well-respected international organizations, such as Transparency International and the Council of Europe, to favour such laws. Perhaps our findings, and those of other researchers, will bolster the legislative prospects for meaningful reforms in this area. References Alt, J. E., & Lassen, D. D. (2008). Political and judicial checks on corruption: Evidence from American State Governments. Economics & Politics, 20(1), 33–61. Baer, M. H. (2008). Linkage and the deterrence of corporate fraud. Virginia Law Review, 94(6), 1295–1365. Becker, G. S. (1968). Crime and punishment: An economic approach. Journal of Political Economy, 76(2), 169–217. Braithwaite, J. (1991). Poverty, power, white-collar crime and the paradoxes of criminological theory. Australian and New Zealand Journal of Criminology, 24(1), 40–58. Brown, N. C., Smith, J. D., White, R. M., & Zutter, C. J. (2015). Greased wheels or just greased palms: political corruption and firm value. [Available at SSRN: https://ssrn. com/abstract=1882011 or https://doi.org/10.2139/ssrn.1882011]. Brown, A., Dozo, N., & Roberts, P. (2016). Whistleblowing processes & procedures: An Australian & New Zealand snapshot. Preliminary results of the Whistling While They Work 2 Project. Brisbane: Griffith University [Retrieved from http://www.whistlingwhiletheywork.edu.au/wp-content/uploads/2016/11/WhistleblowingProcesses-Procedures-Snapshot-Prelim-Results-Griffith-University-8-Nov-2016-FINAL2.pdf]. Brush, J. (2007). Does income inequality lead to more crime? A comparison of cross-sectional and time-series analyses of United States counties. Economics Letters, 96(2), 264–268. Callahan, E. S., & Dworkin, T. M. (1994). Who blows the whistle to the media and why: Organizational characteristics of media Whistleblowers. American Business Law Journal, 32(2), 151–184. Callahan, E. S., & Dworkin, T. M. (2000). The state of state whistleblower protection. American Business Law Journal, 38(1), 99–175. Campante, F. R., & Do, Q. (2014). Isolated capital cities, accountability and corruption: Evidence from US states. American Economic Review, 104(8), 2456–2481. Chiu, W. H., & Madden, P. (1998). Burglary and income inequality. Journal of Public Economics, 69(1), 123–141. Choe, J. (2008). Income inequality and crime in the United States. Economics Letters, 101(1), 31–33. Cordis, A. S., & Milyo, J. (2016). Measuring public corruption in the United States: Evidence from administrative records of federal prosecutions. Public Integrity, 18(2), 127–148. Cordis, A. S., & Warren, P. L. (2014). Sunshine as disinfectant: The effect of state Freedom of Information Act laws on public corruption. Journal of Public Economics,

10

Accounting Forum xxx (xxxx) xxx–xxx

A.S. Cordis, E.M. Lambert

115, 18–36. Council of Europe (2014). Recommendation CM/Rec(2014)7 on the protection of whistleblowers and explanatory memorandum. [Retrieved from https://rm.coe.int/ 16807096c7]. Dyck, A., Morse, A., & Zingales, L. (2010). Who blows the whistle on corporate fraud? The Journal of Finance, 65(6), 2213–2253. Fajnzylber, P., Lederman, D., & Loayza, N. (2002). Inequality and violent crime. Journal of Law and Economics, 45(1), 1–39. Frank, M., Sommeiller, E., Price, M., & Saez, E. (2015). Frank-Sommeiller-Price series for top income shares by US states since 1917. [Available at http://www.shsu.edu/ eco_mwf/inequality.html]. Ginsberg, W. (2014). The freedom of information act (FOIA): Background, legislation, and policy issues, Congressional Research Service Report. Glaeser, E., & Saks, R. (2006). Corruption in America. Journal of Public Economics, 90, 1053–1072. Goel, R. K., & Nelson, M. A. (2014). Whistleblower laws and exposed corruption in the United States. Applied Economics, 46(20), 2331–2341. Goel, R. K., Nelson, M. A., & Naretta, M. A. (2012). The internet as an indicator of corruption awareness. European Journal of Political Economy, 28(1), 64–75. Grassley, C. (2013). Time change: We the People: Fulfilling the promise of open government five years after the Open Government Act, Statement by Senator Grassley, Senate Judicial Committee Hearing. Guedhami, O., & Pittman, J. (2008). The importance of IRS monitoring to debt pricing in private firms. Journal of Financial Economics, 90(1), 38–58. Hanlon, M., Hoopes, J. L., & Shroff, N. (2014). The effect of tax authority monitoring and enforcement on financial reporting quality. The Journal of the American Taxation Association, 36(2), 137–170. Hobson, J. L., Mayew, W. J., & Venkatachalam, M. (2012). Analyzing speech to detect financial misreporting. Journal of Accounting Research, 50(2), 349–392. Hoopes, J. L., Mescall, D., & Pittman, J. A. (2012). Do IRS audits deter corporate tax avoidance? The Accounting Review, 87(5), 1603–1639. Kelly, M. (2000). Inequality and crime. Review of Economics and Statistics, 82(4), 530–539. Long, S. B., Roberge, L., Long, S. B., & Lamicela, J. T. (2004). Balancing data quality against time and money constraints: Tricks from the trenches. Proceedings of the 29th annual SUGI conference [Retrieved from http://www2.sas.com/proceedings/sugi29/094-29.pdf]. Maroun, W., & Solomon, J. (2014). Whistle-blowing by external auditors: Seeking legitimacy for the South African audit profession? Accounting Forum, 38(2), 109–121. Miller, G. S. (2006). The press as a watchdog for accounting fraud. Journal of Accounting Research, 44(5), 1001–1033. Near, J., & Miceli, M. (1985). Organizational dissidence: The case of whistle-blowing. Journal of Business Ethics, 4(1), 1–16. Rampell, C. (2016). Fear (or lack thereof) of getting caught by the tax man. The Washington Post [Available at https://www.washingtonpost.com/news/rampage/wp/ 2016/08/31/fear-or-lack-thereof-of-getting-caught-by-the-tax-man/?utm_term=.34bfe4963fa6]. Rubenfeld, S. (2015). Corruption currents: White-collar crime prosecutions reach 20-year low. The Wall Street Journal [Available at http://blogs.wsj.com/ riskandcompliance/2015/07/31/corruption-currents-white-collar-crime-prosecutions-reach-20-year-low/]. Seifert, D. L., Sweeney, J. T., Joireman, J., & Thornton, J. M. (2010). The influence of organizational justice on accountant whistleblowing. Accounting, Organizations and Society, 35(7), 707–717. Smith, J. D. (2016). US political corruption and firm financial policies. Journal of Financial Economics, 121(2), 350–367. Tillman, R. H., & Pontell, H. N. (2016). Corporate fraud demands criminal time. The New York Times [Available at https://www.nytimes.com/2016/06/29/opinion/ corporate-fraud-demands-criminal-time.html?emc=eta1&_r=1]. Transparency International (2009). Recommended draft principles for whistleblowing legislation. [Retrieved from https://www.transparency.org/files/content/activity/ 2009_PrinciplesForWhistleblowingLegislation_EN.pdf]. Trompeter, G. M., Carpenter, T. D., Desai, N., Jones, K. L., & Riley, R. A., Jr. (2013). A synthesis of fraud-related research. Auditing: A Journal of Practice and Theory, 32(1), 287–321. Trompeter, G. M., Carpenter, T. D., Jones, K. L., & Riley, R. A., Jr. (2014). Insights for research and practice: What we learn about fraud from other disciplines. Accounting Horizons, 28(4), 769–804. Wilde, J. H. (2017). The deterrent effect of employee whistleblowing on firms' financial misreporting and tax aggressiveness. The Accounting Review, 92(5), 247–280.

11