Thursday, October 19, :15-5:45 p.m. Stanford Law School Room 320D

Transcription

1 LAW AND ECONOMICS SEMINAR Autumn Quarter 2017 Professor Polinsky Thursday, October 19, :15-5:45 p.m. Stanford Law School Room 320D Higher Sanctions or Larger Crimes? An Empirical Assessment of DOJ Corporate Monetary Sanctions in the Post-Sarbanes Oxley Era by Mark Cohen (Owen Graduate School of Management, Vanderbilt University) Note: It is expected that you will have reviewed the speaker s paper before the seminar.

2 HIGHER SANCTIONS OR LARGER CRIMES? AN EMPIRICAL ASSESSMENT OF DOJ CORPORATE MONETARY SANCTIONS IN THE POST-SARBANES OXLEY ERA* Cindy R. Alexander** Mark A. Cohen*** October 6, 2017 This draft: Presented at Law & Economics Seminar, Stanford University Law School Preliminary Please do Not Quote, Circulate or Cite without Permission *An earlier version of this paper was presented at the USC Law School. ** Research Fellow, Law & Economics Center, George Mason University. Alexander thanks James Cooper and the staff of the Law & Economics Center for their support for the portions of the DPA study that led to this paper and former colleagues at the DOJ and CEA, Jennifer Arlen, Dan Klerman, Alex Lee and David Reiffen for helpful discussions. Alexander is also affiliated with the SEC; the SEC disclaims responsibility for any private publication or statement by any of its employees. This paper expresses the views of the authors and does not necessarily reflect the views of the Commission of the authors colleagues upon the staff of the Commission. *** Justin Potter Professor of American Competitive Enterprise and Professor of Law, Owen Graduate School of Management, Vanderbilt University. Cohen thanks James Cooper and the staff of the Law & Economics Center for their support for the portions of the DPA study that led to this paper, and to the Owen Graduate School s Dean Fund for Summer Research for research funding.

3 HIGHER SANCTIONS OR LARGER CRIMES? AN EMPIRICAL ASSESSMENT OF DOJ CORPORATE MONETARY SANCTIONS IN THE POST-SARBANES OXLEY ERA Abstract We study the evolution of monetary sanctions corporations following the corporate scandals of amid legal reforms in several areas. First, Congress in 2002 mandated an increase in criminal sanctions for corporations under the Sarbanes Oxley Act (SOX). Second, the Sentencing Commission implemented an Economic Reform Package with an effective date of November Third, Justice Department in 2003 encouraged prosecutors to impose criminal sanctions using nonplea settlements in the form of Deferred Prosecution Agreements (DPAs) and Non-Prosecution Agreements (NPAs) without a guilty plea and thus with lesser risk of collateral sanctions. Finally, the Supreme Court issued a series of decisions that initially limited the ability of judges to make downward departures to sentences negotiated between companies and prosecutors under the Organizational Sentencing Guidelines, and ultimately made the Guidelines nonbinding. This paper uses novel data to test for a link between these reforms and the observable increase in monetary sanctions for companies that occurred between 1997 and We find that monetary sanctions are significantly higher, post- versus pre-2003, on average. Controlling for other factors, we find that the increase in monetary sanctions that might be attributed to the era-specific demand for reform is more readily explained by changes in the seriousness of the offenses and the culpability of offenders entering into settlements after Among the reforms of the period, the use of DPA is associated with higher monetary sanctions than plea agreements, contrary to the view that DPAs are a means for prosecutors to go soft on corporate crime. 1

4 I. Introduction In response to public demand for reform amid corporate scandals, Congress in 2002 required increased criminal sanctions for corporations under the Sarbanes Oxley Act (SOX). In the following year, the Sentencing Commission issued changes to the Guidelines on sentencing of organizations and the Justice Department issued new guidance for prosecutors that encouraged the use of non-plea settlements that would ultimately extend the reach of criminal enforcement into corporations without the risk of collateral sanctions associated with plea agreements. The overall effect was an increase in the number of companies sanctioned for criminal misconduct per year after This was accompanied by a series of unrelated Supreme Court decisions on the exercise of judicial discretion with implications for the sentencing of organizations. These decisions affect the ease with which judges could make adjustments to sentences negotiated between companies and prosecutors at settlement. A prior study (Alexander and Cohen, 2015: 568-9) introduces evidence of a dramatic transformation of corporate criminal enforcement after In particular, within a few years, DPAs and NPAs accounted for nearly half of all criminal settlements for public companies, while at the same time the annual number of plea agreements did not drop. Thus, the reach of corporate criminal enforcement expanded. In addition, monetary penalties were found to be higher, post-sox. The reason for the increase is unclear. It could be attributable to the shift in the form of settlement and also changes in the types of offenses, the severity of offenses, and reforms to the administration of justice that occurred in this period. 1 In this paper, we turn to the question of whether findings of an increase in monetary sanctions after 2003 are attributable to specific changes in the legal environment or whether they arose from a more general increase in the bargaining power of prosecutors at settlement, controlling 1 Changes in the legal environment include changes in the administration of the law that are a matter of public record through statutory reforms or by issuance of guidance to judges and prosecutors by the sentencing commission and the justice department, respectively. 2

5 for factors such as offense type and seriousness that are known to influence sanctions generally. We analyze novel data on 486 publicly-traded firms that settled charges of criminal wrongdoing pre- and post-2003, over the period. We find that the increase in monetary sanctions is attributable to an increase in crime seriousness and offender culpability as measured under Guidelines and to the introduction of non-plea forms of settlement. That is, firms with DPA settlements receive significantly higher monetary sanctions than those with plea agreements, controlling for other factors that include crime type, seriousness, and culpability. Finally, we find evidence that when the scope for judicial discretion was increased, monetary sanctions decreased, again controlling for other factors. Section 2 provides background on the legal history surrounding the sentencing of organizations in the federal system, and related legal, economic and criminological theories of punishment. We introduce several hypotheses about the determinants of the rise of monetary sanctions around the reform period. Section 3 describes the data and empirical method, followed by a presentation of findings in Section 4. Finally, a concluding section explores alternative hypotheses and some topics for future research on organizational sanctions. II. Hypothesis Development: The Sentencing of Organizations in the U.S. Federal System We study the evolution of monetary sanctions in response to a series of corporate scandals marked by the passage of the Sarbanes-Oxley Act of The reforms of the SOX period were the first major reforms to federal sentencing of corporations since the issuance of sentencing guidelines for organizations in Under the Comprehensive Crime Control Act of 1984, the Congress created the U.S. Sentencing Commission and authored the promulgation of guidelines for prosecutors and judges to follow in the sentencing of individuals and organizations for federal crimes. Prior to 1991, sentences for organizations convicted of federal crimes were constrained only by statutory minimum or maximum penalties and the terms and sanctions imposed were generally 3

6 negotiated between the prosecutor and company subject to federal judicial approval. Few if any criminal charges are settled at trial; instead over 95% of organizations charged with federal crimes plead guilty. 2 When Congress passed the Comprehensive Crime Control Act of 1984, it voiced a belief that the monetary penalties that judges approved for corporations were systematically too low and called for increased severity of corporate sentences (Alexander et al., 1999: 395). Accordingly, in 1991, the U.S. Sentencing Commission ( Commission ) with ratification by Congress adopted guidelines for judges to follow in the sentencing of organizations ( Guidelines ). These Guidelines provided for a detailed calculation of offense seriousness and offender culpability, leading to a prescribed range of monetary sanctions as well as other provisions such as mandatory restitution. While mandatory, judges did retain some authority to depart (upwards or downwards) with explanation although either side could then challenge the validity of the departure on appeal. Criminal fines and total monetary sanctions for organizations rose after the Sentencing Guidelines for Organizations first went into effect in November 1991 (see Alexander et al., 1999). Total sanctions and criminal fines were higher for corporate crimes sentenced after that date than previously, consistent with the Guidelines leading to more stringent sanctions. The increase in fines was greater in cases where the guidelines were legally binding, however, while the increase in total sanctions was about the same. Thus, factors other than the causal impact of the Guidelines on judicial discretion appear to explain the increase in corporate criminal sanctions that occurred over the time period (see Alexander et al., 1999). 3 More than a decade after the first issuance of sentencing guidelines for organizations, changes to corporate criminal sentencing began with the expedited passage of the Sarbanes-Oxley Act of 2 For example, in FY 2016, the Commission reported three organizations convicted at trial (2.3%) out of 132 organizations sentenced under the Guidelines. 3 Among the candidate explanations for the finding that sentences rose in cases that were not guidelineconstrained, a shift in judicial norms or judicial information about best sentencing practice may have coincided with the November 1991 effective-date of the first guidelines for sentencing organizations. Alternatively, judges may have had strategic reasons for cooperating with a perceived Congressional intent to corporate criminal sanctions across the board, not just for crimes committed after November See AAC Section IB (Alternative theories of judges and legal rules). 4

7 2002. Under the Act, Congress instructed the Commission to expeditiously consider the promulgation of new sentencing guidelines or amendments to existing sentencing guidelines to provide an enhancement for officers or directors of publicly traded corporations who commit fraud and related offenses. 4 The Commission issued a report and proposed amendments in January 2003, with formal amendments submitted to Congress in May 2003 and an effective date of November 1, While most of the attention of Sarbanes-Oxley was on corporate governance reforms and increased criminal liability and sentencing enhancements for officers and directors of corporations, some attention was also focused on organizational sanctions. Of particular relevance to organizational sanctions, the Commission under the Act proposed changes in sentencing that included an increase in the guideline fine range ( enhancement ) for fraud offenses involving 250 or more victims, and to offenses that endanger the solvency or financial security of (1) a publicly traded corporation, (2) an organization that employs 11,000 or more employees, or (3) 100 or more individual victims. In either of these cases, the Guideline fine range could increase substantially by as much as 25% to 50%. Finally, the amendments also increased the base fine level for crimes that involved between $200 and $400 million, and for those that involved more than $400 million in monetary loss. 6 In addition to these structural changes in the Guidelines, there is anecdotal evidence of a decline in the general public s confidence in large corporations after An annual Gallup survey asking the public whether they are satisfied or not with the size and power of major corporations, peaked at 50% who responded they were very or somewhat satisfied in January 2002, with this 4 Sarbanes-Oxley Act of Public Law , July 30, 2002, (Section 1104 (a)(2), Amendment to the Federal Sentencing Guidelines). 5 Note that formally, after the Commission proposes any amendments, they go into effect unless Congress modifies disapproves of the amendments within 180 days. 28 U.S.C. 994(p). 6 USSC (2003c), pp. i-ii. 7 For a review and appraisal of the procedures leading to expedited passage of the Sarbanes-Oxley Act, see Romano (2004). 5

8 figure declining to 43% in January 2003; and 38% in January Each of the reforms to the legal environment that we study could have occurred independently of any scandals or heightened public demand for reform. Yet it is natural to consider that they might not have occurred, or might not have occurred at around the same time, without such intense demand from the public. This leads us to our first hypothesis, which is that prosecutors were able to negotiate more effectively for higher monetary sanctions in response to the public demand for reform after 2003, independently of the legal reforms that arose during the period, other things equal: H1: Post-2003, monetary sanctions for corporations settling charges of federal criminal wrongdoing are higher on average than in the prior period, controlling for other factors that generally affect monetary sanctions for corporations under the Guidelines. Other factors include offense seriousness, culpability and offense type. The hypothesis is that monetary sanctions increased simultaneously with, and independently of, the specific legal and policy reforms, e.g., the structural/mandated increase in sanctions under the Guidelines. Changes in the influence of judges over sentencing in this period warrant special attention, given the changes in the ability of judges to adjust sanctions, post-settlement, that also occurred in this period. The tension between the judiciary and the other branches of the government over appropriate sentences for criminals for both street crime and white-collar/corporate criminal offenders has been the subject of extensive study in the prior literature. 9 This tension could reflect differences in the constraints that the different actors face or may arise from institutional differences. For example, federal judges are appointed for life, while Congress and federal 8 Frank Newport, Americans Similarly Dissatisfied with Corporation, Government, Gallop, January 17, This figure continued its decline to a low of 29% in January 2012, increasing to 36% in January See e.g., Reinganum (2000) and Miceli (2008) for formal models. Anderson et al. (1999) provide documentation that the federal sentencing guidelines both increased sentences for individual offenders and reduced inter-judge disparity. 6

9 prosecutors are more accountable to political cycles. 10 Depending on the alignment between the preferences of judges and those of the general public, the difference in term of appointment could lead to differences in the responsiveness of legislators, prosecutors and judges to changes in public demand for aggressive enforcement and thus higher sanctions. Generally, judges have objected to constraints on their exercise of discretion at sentencing, as occurs under the sentencing guidelines, 11 and tend to exercise their discretion by imposing more lenient sanctions than the sanctions that are expressly prescribed by prosecutors and the legislature, on average. 12 Consistent with an effort to obtain higher expected sanctions for corporations, Congress in 2004 passed legislation to limit the ability of judges to make downward departures from the Guidelines. That is, effective May 1, 2004, Congress passed the Feeney Amendment to the PROTECT Act which among other things tightened standards for downward departures making it more difficult for judges to lower monetary sanctions below minimum guideline fines. 13 This action 10 For theory and evidence on how judges make decisions based on their preferences and constraints, see Cohen (1991a, 1992a), Posner (1993, 2008), Epstein and Knight (2013) and Epstein et al. (2013). These results extend beyond criminal sentencing, see e.g., Choi et al (2011) explaining judicial opinion writing using a similar framework. Boylan (2005) provides evidence that the career path of prosecutors is positively related to longer prison sentences. 11 See Cohen (1991a) and Sisk et al. (1998) documenting the adverse reaction of federal district court judges to the loss of discretion under sentencing guidelines. Boylan (2004) finds an increase in judicial retirements following the imposition of the federal sentencing guidelines. Also see Epstein et al. (2013). 12 Empirical evidence on sentencing practice of judges under the sentencing guidelines is consistent with this view. In particular, when given the opportunity to exercise discretion by departing from the guideline sentencing upward or downward, judges are found more frequently to depart downward. For example, the Commission (2003a, Table 26) reported 19,377 downward departures in FY 2003 (29.7% of cases), compared to only 541 upward departures (0.8% of cases). In a report to Congress (following concerns about downward departures, the Commission (2003b, iv) noted that downward departures had increased from 5.8% in fiscal year 1991 to 18.1% in fiscal year Also, see Epstein et al. (2013), cited below, for evidence that when judges were given more discretion following Booker, they were more likely to depart downward from the Guidelines. In addition, Reinganum (2000) demonstrates that even if their underlying preferences are identical, judges might prefer less harsh sentences in a game theoretic model that accounts for negotiating between the offender and prosecutor in the face of plea bargaining. 13 Prosecutorial Remedies and Other Tools to End the Exploitation of Children Today Act of 2003, Public Law , Section 401. Among other things, the act (1) tightened the standard of appellate review for non-guideline sentences, replacing the abuse of discretion standard with de novo review (Section 401(d)(2)), directed the Commission to amend the Guidelines to ensure that the incidence of downward departures are substantially reduced, (Section 401(m)(2)A)), and directed the Department of Justice to resist downward departures not supported by the facts and the law. (Section 401(l)(1)(A)) 7

10 strengthened the Guidelines and reduced judicial discretion. The effects of the Feeney Amendment were limited, however, through a series of court rulings that led eventually to the elimination of the mandatory nature of the Guidelines as a constraint on judges in sentencing. The Supreme Court in Blakely on June 25, 2004 held that judges cannot base the sentence on factors not proven in court. 14 This signaled the end of the mandatory federal Guidelines for judges, even while the ruling was based on sentencing guidelines in the State of Washington, and the Court specifically noted that it was not ruling on the federal Guidelines as such (see Scott, 2012). Indeed, several months later (January 12, 2005), the Supreme Court made it clear in Booker that the federal Guidelines were no longer to be mandatory and judges were free to sentence offenders within the confines of any statutory minimum or maximum penalties. 15 Two further court rulings (Kimbrough and Gall) on December 10, 2007 clarified and expanded the scope of judicial discretion and the non-mandatory nature of the Guidelines. 16 While the Guidelines were no longer a constraint on judges, they must still calculate the Guideline sentence and consider them in the sentence they impose, 17 and the Guidelines also remained a consideration of prosecutors in negotiating settlements. Combining this historical narrative with theory and previous evidence that on average judges prefer more lenient sentences than the Congress or executive branch leads to our second hypothesis: H2: More judicial discretion lowers monetary sanctions for organizations settling charges of criminal wrongdoing, controlling offense seriousness, culpability and other sentencing factors under the Guidelines. The Justice Department s guidance for prosecutors on criminal settlement practices for 14 Blakely v. Washington 542 U.S. 296 (2004). 15 United States v. Booker, 543 U.S. 220 (2005). 16 Kimbrough v. United States, 128 U.S. 558 (2007) and Gall v. United States 128 U.S. 586 (2007). 17 According to the Commission (2006: v), the circuit courts now have uniformly agreed that all post- Booker sentencing must begin with calculation of the applicable guideline range. 8

11 corporations also evolved over the period. In January 2003, the office of the Deputy Attorney General issued the first in a series of memoranda on best practices for prosecutors in reaching criminal settlements with corporations. The first of these, the Thompson Memo, 18 called for the use of deferred prosecution agreements ( DPAs ) a nd non-prosecution agreements ( NPAs ). Subsequent policy memos by Deputy Attorney Generals McNulty (2006), 19 Filip (2008), 20 and Morford (2008) 21 clarified DOJ policies related to various aspects of settlements. Some commentators have expressed the concern that the practice of using non-plea settlements (DPAs and NPAs) to resolve corporate criminal investigations promotes underdeterrence of corporate crime. 22 The concern is in part that non-plea settlements, by eliminating the guilty plea cause the offending company to face a lower expected cost of the collateral effects of having a criminal conviction, which include the expected current and future costs of any debarment or delicensing that a company might face. There is also a concern that the absence of a guilty plea 18 Memorandum from Larry D. Thompson, Deputy Att y Gen., U.S. Dep t of Justice, to Heads of Dep t Components & United States Attorneys, Principles of Federal Prosecution of Business Organizations (Jan. 20, 2003) available at aiv_dojthomp.authcheckdam.pdf. 19 Memorandum from Paul J. McNulty, Deputy Attorney General, U.S. Dept. of Justice, to Heads of Dept. Components & United States Attorneys, Principles of Federal Prosecution of Business Organizations (Dec. 12, 2006) [hereinafter McNulty Memo ], available at The McNulty Memo clarified when prosecutors may seek waiver of legal privileges (Alexander and Cohen, 2015: 551). 20 Memorandum from Mark R. Filip, Deputy Attorney General, U.S. Dept. of Justice, to Heads of Dept. Components & United States Attorneys, Principles of Federal Prosecution of Business Organizations (Aug. 28, 2008) [hereinafter Filip Memo ], available at pdf. The Filip memo clarified how cooperation should be defined as a mitigating factor and can be seen as an effort to alleviate prosecutorial pressure on firms to cooperate. (Alexander and Cohen, 2015: 550) 21 Memorandum from Craig S. Morford, Acting Deputy Attorney General, U.S. Dept. of Justice, to Heads of Department Components, Selection and Use of Monitors in Deferred Prosecution Agreements and Non- Prosecution Agreements with Corporations United States Attorneys (March 7, 2008) [hereinafter Morford Memo ], available at The Morford Memo distinguished between NPA and DPA as distinct forms of non-plea settlement and provided guidance on the drafting of settlement provisions that call for a corporate monitor, and the role of the monitor relative to that of the Justice Department.(Alexander and Cohen, 2015: 552). 22 See e.g., Mokhiber (2005) and Uhlmann (2013). 9

12 could enable the company to escape some of the negative publicity that might contribute to the costs of settlement and are associated with costs of reputational damage from getting caught. For the purposes of this paper, the implication of these concerns to the extent they are valid is that the monetary sanction should be higher for a given offender and offense if there is a non-plea settlement. That is, assuming that prosecutors generally prefer higher sanctions, they should be able to extract a higher monetary sanction in exchange for allowing the company to have a non-plea settlement and thus a lesser risk of costly collateral sanctions. The hypothesis that monetary sanctions should be higher under non-plea than plea settlements is consistent with the fact that the former are a subject of lesser judicial review. 23 That is, the chance of a judge lowering a sentence after it has been negotiated by the prosecutor and company is lower with a non-plea than with a plea agreement. Independently of whether the total sanction is higher for companies under nonplea than plea settlements, we thus have the hypothesis that the monetary sanctions should be higher: H3: Deferred Prosecution and Non-Prosecution Agreements raise monetary sanctions for organizations settling charges of criminal wrongdoing, controlling for offense seriousness, culpability and other sentencing factors under the Guidelines. In testing these three hypotheses, we consider differences in marginal sanctions controlling for other factors that are known to influence the magnitude of the sanction, such as the mix of offenses that receive sanctions. 24 III. Data and Empirical Method Our initial data consist of a comprehensive compilation of all known federal settlements of plea agreements, DPAs and NPAs with a public company (or its subsidiary) and a prosecutor of the Justice Department, including its Divisions and Offices of US Attorneys, between 1997 and The 23 Consistent with this, some commenters have suggested that NPAs and DPAs will lead to overreach by prosecutors who will be able to extract higher payments or other concessions from firms at settlement than would occur with a guilty plea and greater judicial oversight. See e.g., Janis (2005) and Epstein (2006). 24 These issues are discussed further in Alexander and Cohen (2015). See also []. 10

13 construction of the sample of 486 settlement agreements is explained in Alexander and Cohen (2015: ). About 2/3 of cases (329) are plea agreements, followed by NPAs (91) and DPAs (66) (Alexander and Cohen, 2015: Table 3). However, the composition of cases has changed considerably over time. While 98% of cases prior to 2003 were pleas, that percentage decreased to 63% between 2003 and 2006, and 56% between 2007 and Thus, within less than 10 years following the introduction of DPAs and NPAs, they accounted for nearly half of corporate criminal settlements by publicly traded companies. Dependent Variable The key dependent variable for this paper is the total monetary sanction including any criminal penalty and other civil settlements with government or state agencies that are noted in the settlement agreement or DOJ press release. As explained in Alexander and Cohen (2015: 582-3), this definition is necessitated by differences in the approaches taken in NPAs and DPAs compared to pleas where a criminal fine might not be specified. Our goal is to place all settlements on an equal footing in terms of the total monetary sanction paid by the corporation. Thus, this definition would include restitution that is specifically mentioned in the settlement documents or DOJ press release, but would exclude civil settlements not part of the agreement such as class action lawsuits. In that sense, one can view our dependent variable as being the global settlement with DOJ of the alleged criminal wrongdoing. The Legal Environment As discussed in Section 2, during the time frame that our study covers, several important legal and policy changes occurred. Table 1 presents the timeline of the changes regarding corporate criminal sentencing that have occurred since the mid-1980s that are the focus of this study. Similar timelines appear in studies by the Sentencing Commission and others of post-booker sentencing behavior (USSC, 2012: 56). 25 The third column of Table 1 identifies the direction of the effect of each 25 Scott (2012) uses a similar time period in his empirical analysis of sentencing criminal offenders in 11

14 expected change on the scope for judicial influence over the monetary sanction ( judicial discretion ), while the last column provides the implied sign of the hypothesized effect of the change on the monetary sanctions. Recall first that one of the express purposes of the Guidelines was to increase the severity of sanctions imposed on corporate offenders in part by reducing the influence of judicial discretion at the time of sentencing. That is, on average, judges were assumed to prefer less severe sanctions than those imposed by the Guidelines. Accordingly, policies that relax constraints on judges should result in decreased criminal sanctions as influence shifts away from judges. Conversely, measures that make it more difficult for judges to influence settlement outcomes should lead to more severe sanctions. The Thompson Memo encouraged prosecutors to substitute the use of DPAs and NPAs for traditional plea agreements in settling criminal investigations of corporations. There are several reasons why corporate executives might agree, and indeed prefer, to settle with a DPA or NPA, holding the monetary sanction constant. They might believe that DPAs and NPAs reduce their risk of collateral sanctions, such as debarment from government contracts, relative to plea agreements. 26 Prosecutors could then bargain for higher monetary sanctions in exchange for agreeing to offer a DPA or NPA. On the other hand, to the extent that DPAs and NPAs have extended the reach of the criminal law by allowing prosecutors to charge and settle cases with companies for offenses that would have been difficult to obtain a criminal conviction, we might expect less severe monetary sanctions. When these two effects are combined, while the Thompson Memo may have limited the exercise of judicial discretion in principle, its overall impact on sanctions is an empirical question. Massachusetts. See also Yang (2014), who studies the period prior to and after Koon v. United States 518 U.S. 81 (1996). 26 Alexander and Arlen (2017) compare the differences in the abilities of prosecutors and government officials to waive collateral sanctions, such as debarment and delicensing under DPA versus plea agreements, and conclude that the risk to the company of a costly collateral sanction, triggered by settlement, from a plea agreement is greater than from a DPA, even after considering the potential for waivers to occur. 12

15 The PROTECT Act constrained the exercise of judicial discretion by limiting downward departures in Guidelines cases; hence we anticipate an increase in sanctions for plea agreements negotiated under the Guidelines. Blakely applies to state sentencing guidelines and has an ambiguous effect on sanctions in principle. 27 As discussed in the previous section, Booker and Kimbrough/Gall had the effect of relaxing the constraints that the Guidelines had placed on individual judges at criminal settlement. Assuming judges prefer lower sanctions than are prescribed by the guidelines, we should find lower sanctions, post-booker and Post-Kimbrough/Gall, than previously, cet. par. While we are unaware of prior studies examining the impact of changes in the legal environment on corporate sanctions, several recent studies have examined the impact of some of these changes on federal criminal sentences for individual offenders. Freeborn and Hartmann (2010) found that the PROTECT Act reduced downward departures from the Guidelines sentence, resulting in an average increase in prison sentence of two months. Scott (2012) studied the impact of Booker on sentences for individual street crime offenders in Massachusetts and found that average sentence length increased following Booker and again following Kimbrough and Gall. 28 However, some judges displayed what Scott termed a free at last pattern by significantly reducing their average sentences below Guidelines levels. Thus, following the increase in judicial discretion brought about by these two court cases, the impact of the judge on sentence length grew considerably. Excluding sentences subject to mandatory minimums, Scott finds that pre-booker, judges explained 1.4% of the variance in sentences. This increased to 3.1% post-booker, and 8.0% post-kimbrough/gall (Scott, 2012: Table 2). Similar findings are reported in Yang (2014). Epstein et al. (2013: 243) confirmed this increased variability in sentencing by examining 27 Commentators at the time anticipated that the Supreme Court would ultimately rule in a similar manner and limit downward departures, hence increasing sanctions. 28 The analysis in Scott (2012) is for street crime, so it appears that judges on average in Massachusetts felt constrained by the Guidelines and would have preferred longer sentences, on average. In our study on corporate crime, we anticipate the opposite effect as judges have previously been shown to prefer less severe sanctions than Congress or prosecutors. 13

16 departures from the Guideline sentence range. They found that following Booker, 36.8% of all sentences were below the Guideline range (compared to 28.4% just prior to Booker), while 1.6% were above the range post-booker (compared to 0.7% just prior to Booker). In a regression analysis explaining the fraction of cases above or below the Guidelines by District, these differences were statistically significant (Epstein et al, 2013: 245). 29 Control Variables In addition to the key legal and policy variables identified above, we include a number of control variables in order to test our three hypotheses: Offense Magnitude Our ability to determine whether or not monetary sanctions increased following any of these indicators of change in the legal environment critically depends upon comparing offenses that are similar with respect to other factors that might influence sanctions. Otherwise, we might attribute an increase in sanctions to a tendency for either more serious crimes (and/or more culpable offenders) being committed and/or prosecuted. In identifying measures of offense magnitude, we turn to the related academic literature. The economics literature on optimum sanctions characterizes the sanction that minimizes the social cost of crime as based on the social harm and probability of detection. This was first formalized in the context of individual wrongdoing by Becker (1968), and extended to corporate criminal sanctions by Shavell (1987), Cohen (1992b), Polinsky and Shavell (1993) and others. 30 Criminological theories of punishment include enforcement objectives of deterrence, incapacitation, rehabilitation, 29 They also report that in addition to the increase that occurred following Booker, an additional statistically significant increase beyond this initial effect in 2009 and 2010 several years after Booker. Although they attribute this to delayed adjustment to Booker. 30 Additional issues relative to optimal penalties that are not examined in this paper include whether or not sanctions should be monetary versus non-monetary (e.g., Shavell, 1987), and the potential trade-off between individual and corporate criminal sanctions for the same underlying offense (e.g., Polinsky and Shavell, 1993). Mullin and Snyder (2009) contains a useful summary of this theoretical literature. 14

17 restitution, and retribution (sometimes referred to as just punishment or just deserts ). 31 When Congress passed the Crime Control Act of 1984, which established the U.S. Sentencing Commission, it enumerated similar goals of sentencing. 32 These goals were formalized by the Commission in the context of corporate sanctions with the 1991 Guidelines. While the policy recommendations might vary depending upon which goal(s) of sentencing is adopted, virtually all of these normative theories lead to a straightforward policy recommendation that ceteris paribus, monetary sanctions should be increasing with the monetary loss caused by crime. 33 Previous empirical evidence also supports this finding. For example, Cohen (1996) found that monetary sanctions were increasing in monetary loss 34 in 285 private and public corporate offenders sentenced between 1984 and However, data on monetary loss has proven to be particularly difficult to obtain in corporate criminal cases. For example, while Cohen (1996) was able to obtain monetary loss data in about 30% of cases (285 out of 961 corporate convictions), Alexander and Cohen (2015) were only able to estimate monetary loss in 13% (62 out of 486) of publicly traded companies that settled allegations of criminal wrongdoing with a plea agreement, DPA, or NPA. 35 Thus, operationalizing a control variable solely on monetary loss is particularly 31 These four purposes of punishment date as far back as the early 1800s with Jeremy Bentham s writings (see Crimmins, 2017). As Parker (1989: 563-4) explains, the traditional view of retribution has been generally modified to focus on just desert or proportionality. 32 Guidelines, Chapter 8 Introductory Comments. 33 As explained by Parker (1989: 555), a legal theory of deterrence would base the monetary sanction on the gain to the offender in order to provide an incentive not to commit any illegal acts whereas the alternative Becker approach focuses on harm to the victim (and society), allowing for some crimes to be committed because the cost of deterrence is too high relative to the social gain. 34 Although Cohen (1996) used the term harm, to avoid confusion we use instead the term loss. Cohen (1996) was not estimating social harm, but instead included harm to victims which might be considered a transfer and not a social loss. But see Becker (1968, 171, note 3.) suggesting that these transfers might be considered a social harm in the case of theft. 35 See Alexander and Cohen (2015), note 165. Several earlier studies were cited in that note where loss estimates were available in a significantly higher percentage of cases (e.g., Cohen (1989: 617) estimates monetary loss in 62% of cases (178 out of 288); and Cohen (1991b: 258-9) estimates monetary loss in 52% (262 out of 505). However, as noted, those earlier studies were based partly on detailed analysis of presentence investigation reports that are not publicly available to researchers; hence, studies relying upon publicly available information post-1990 are likely to have significantly fewer cases where monetary loss can be estimated. 15

18 difficult. The criminological approach of just deserts calls for two broad factors that are used to calculate a penalty seriousness and culpability. 36 This conceptual approach was adopted by the Commission when they promulgated their Guidelines generally, and specifically when they adopted a monetary penalty range for convicted organizational offenders. 37 Seriousness may be thought of as a proxy for loss even when loss is not monetized. The Guidelines operationalize seriousness by calculating an offense score which then gets translated to a base fine that is the maximum of monetary loss, gain, or a number derived from a default fine table that is based on offense-specific characteristics. This base fine is then adjusted using a culpability score that provides a final fine range depending upon various factors such as whether or not there was high level involvement, cooperation, obstruction of justice, etc. The culpability of the offender, an essential part of criminological theory of sanctions, may be a suitable proxy for the probability of detection which is a critical element of the optimal penalty literature. 38 For example, one key component of the Guideline s culpability score is whether or not the organization cooperated with investigators and/or attempted to obstruct justice. To control for seriousness and culpability, we have adopted the methodology of the Guidelines that arrives at a fine range from which the offending organization is supposed to be sentenced. This fine range is based on a combination of the seriousness of the offense and the culpability of the offender. Details on the methodology are in Alexander and Cohen (2015: ). 36 See Packer (1968). 37 the fine range should be based on the seriousness of the offense and the culpability of the organization. The seriousness of the offense generally will be reflected by the greatest of the pecuniary gain, the pecuniary loss, or the amount in a guideline offense level fine table. Culpability generally will be determined by six factors that the sentencing court must consider. The four factors that increase the ultimate punishment of an organization are: (i) the involvement in or tolerance of criminal activity; (ii) the prior history of the organization; (iii) the violation of an order; and (iv) the obstruction of justice. The two factors that mitigate the ultimate punishment of an organization are: (i) the existence of an effective compliance and ethics program; and (ii) self-reporting, cooperation, or acceptance of responsibility. (Guidelines, Chapter 8 Introductory Commentary ). 38 CITE 16

19 Our approach of relying on Guidelines-based indicators of severity in this study is a departure from earlier literature that relies upon monetary loss or gain to measure offense seriousness. 39 An advantage of the current approach is that it permits comparisons to be made across offenses even where no monetary loss or gain is involved (or when monetary loss and gain are unknown), or when the monetary loss or gain do not adequately account for the seriousness of the offense (e.g., risk of long-term health impacts or national security-related violations), thereby expanding the size of the usable sample beyond what would arise if the study were limited to those offenses involving monetary transfers that are readily measurable. While the Guideline fine range should be calculated for all corporate offenders that plead guilty, this information is not always available publicly. Moreover, it is only sporadically calculated in DPA and NPA documents. Thus, as described in Alexander and Cohen (2015), the Guideline fine was recorded from publicly-available documents when available, and estimated using the Guideline Manual and public documents when the actual calculations were not available. We use the midpoint of the Guideline fine range as our measure of offense magnitude. Because they serve somewhat different purposes, and we have varying sample sizes, we use all three measures of offense magnitude. Our largest sample is for the culpability score (n=420), since this does not require any information on the magnitude of loss. Second, we have data on the offense score in 353 cases. The smallest sample is for the midpoint of the Guideline fine (n=317), because it requires that we have both an offense score and a culpability score. Other Control Variables We have controlled for several additional factors that might affect the level of sanction ultimately imposed. First, in order to control for any offense-specific harms or characteristics not otherwise captured in the Guidelines calculations, we include indicator variables for broad crime type categories: non-healthcare fraud; domestic bribery/kickbacks; Foreign Corrupt Practices Act; 39 See e.g., Cohen (1991b). 17

20 antitrust; import/export and immigration violations; money laundering/tax violations; environmental and safety violations; healthcare fraud and FDA violations; and other (see Alexander and Cohen, 2015). In some cases, the record indicated that the sanction was influenced by the firm s reported ability to pay. 40 These cases have been identified with an indicator variable to control for these non-relevant reductions in sanctions. Finally, we include indicator variables to identify when the company settling the case was a subsidiary of a parent company that was also settling a case at the same time for the same underlying offense. 41 The empirical analysis proceeds in two parts. First, we present evidence from univariate comparisons of monetary sanctions and other variables across subsamples, distinguished according to whether the settlement date is pre- or post Then we turn to the multivariate analysis of the data, which permits tests for robustness of findings of change in monetary sanctions, pre- versus post-2003 to controls for other factors that also changed over the period. In the multivariate analysis, we use the mid-point of the Guideline fine range as the baseline control variable, which permits us to test for evidence that changes in the era, form of sanction, and/or legal environment led to increases in the monetary sanction relative to the mid-point, controlling for other factors Further, we regress the monetary sanction on an era indicator variable with control variables that include proxies for the harm caused by the offense, primarily the Guideline s definition of seriousness as measured by the offense score under the Guidelines. We also use the midpoint of the estimated Guideline sanction as a control variable although this is a composite of seriousness and culpability. We have also included in separate specifications the two relevant components of the Guidelines offense score and culpability score. 40 This occurred in 21 out of 486 cases; see Alexander and Cohen (2015: note 193 and accompanying text). 41 This occurred in 65 out of 486 cases where 28 parents settled with 27 subsidiaries; see Alexander and Cohen (2015: Table 7). 18

21 IV. Evidence Table 2 presents summary statistics for the full sample of 486 settlements, by era. As shown in Table 2, the mean total monetary sanction was $85 million, with the median being $11.6 million % of settlements (368) occurred in the post-sox/dpa era beginning Additional legal environment variables identify the Protect Act (5.6%), Blakely (4.7%), Booker (23%), Kimbrough/Gall (42.2%), Filip (36.6%), McNulty (53.7%), and Morford (40.9%). Each of these time periods has been defined as shown in Table 1. In addition, we identify which of these settlements are NPAs (18.7%), DPAs (13.6%), and plea agreements (67.7%). Table 2 also reports several subsamples of interest environmental and safety offenses, and antitrust offenses. The mean sanction was $6.9 million (median $1.3 million) for environmental and safety offenses and $63.6 million (median $25.5 million) for antitrust. Excluding environmental, safety and antitrust offenses, the mean total sanction was $120.3 million and the median was $18.1 million. As shown in Table 2, we were able to estimate a culpability score in 420 cases, an offense score in 353 cases, and determine a Guideline fine range in 317 cases. 43 For these 317 cases, the mean of the fine range was $182 million, with the median being $9.8 million. Table 3 compares the relevant variables in the pre- versus post-sox (2003) era. As shown, the average monetary sanction nearly doubled from $57 million to $94 million (p <.10), while the median increased from $9.2 million to $13.6 million. However, for the subsample where a USSC Guideline fine could be estimated, the average guideline fine increased from $64 million to $210 million (p <.05), with the median increasing from $4.4 to $13.1 million. Similarly, the estimated offense score increased from an average of 19.6 to 25.0 (p <.01), while the median increased from 17 to 26. Thus, while it appears that the severity of monetary sanctions increased substantially in the post-sox era, so did the severity of offenses. There is little difference in the culpability scores (mean 42 Dollar values are expressed constant year 2003 dollars using the consumer price index. 43 Note that in cases where there is no Guideline fine, the sanction may be as high as the statutory maximum. In 69 out of the 317 cases (21.7%), the range was based on statutory maximum penalties because no other Guideline fine is otherwise mandated. 19

22 5.0 pre-sox versus 4.9 post-sox; medians both 4.0), however. Table 4 reports on the same comparison on a subsample that excludes environmental, safety and antitrust offenses. The findings are similar, with the total sanction increasing from $89.4 million to $126.3 million, although this difference is not statistically significant. Once again, the midpoint of the USSC guideline fine increased from $77 million to $262 million (p <.10), while the offense score increased from 24.7 to 30.1 (p <.01). The culpability scores remain similar (mean 5.06 pre- versus 4.73 post; median 5 pre- versus 4 post), with differences statistically insignificant. In the next section, we explore this simple comparison further by controlling for the legal environment and other corporate and crime-specific factors to better understand the source of these increases. The findings from multivariate analysis of the monetary sanction appear in table 5. The evidence is that monetary sanctions were higher for corporations, post-2003, at p<.01 after controlling for crime type and the legal/policy environment (consistent with H1). See columns 1 and 2 of table 5. Column 2 adds indicator variables for NPAs and DPAs. Controlling for these forms of settlement, we still find higher monetary sanctions post However, the magnitude and significance level of the post-2003 indicator variable is now lower at p <.10. The post-2003 era indicator loses its explanatory power when indicators of offense seriousness and culpability are added to the model as explanatory variables, however. In particular, column 3 reports on a model where the midpoint of the Guideline fine range is an explanatory variable. In that model, we find that increased seriousness of the prosecuted offenses and culpability of the offender (as proxied by the Guidelines fine range), rather than the exercise of discretion to obtain higher sanctions for a given level of severity, best explains the increase in monetary sanctions in the data. Thus, the evidence does not support H1, as post-sox, the higher observed sanctions appear to be fully explained by the severity of offenses being punished. Model 4 replaces the midpoint of the Guidelines fine range with its two components, seriousness (proxied by the Guidelines offense score) and culpability (proxied by the Guidelines 20