Recent years have been marked by a worrisome increase in criminal enforcement as a tool of domestic policy. One powerful weapon in this government arsenal is the deferred prosecution agreement (DPA), sometimes called a nonprosecution agreement (NPA), whereby the government agrees to defer or stop prosecution against an individual or firm that complies in full with the conditions of the agreement. These conditions often include requiring the individual or firm to make certain payments to the government or private citizens, to fire senior employees, to restructure key aspects of the business, and to accept continuing government oversight and monitoring of the activities of the firm or specific individuals for a period of years.

These agreements are frequently concluded outside courts of law and thus without judicial oversight. The cumulative cost of compliance with these sanctions imposes a heavy burden on the regulated parties. These include meeting the explicit terms of the DPA or NPA and extend to the disruption of business operations and an evident loss in reputation and business goodwill.

The government’s stated reason for using these agreements is to stem an epidemic of crime that would otherwise allow bad actors within the system to gain at the expense of innocent individuals. Indeed, the rhetoric used to support this position has a crusading quality that misses or understates the underlying difficulties with the practices. This Legal Memorandum examines some of the salient features of these agreements.

• The first section outlines the unwonted zeal that government officials often bring in order to obtain these agreements.
• The second section examines the logic of settlement of both ordinary civil and criminal disputes and explains why the collateral consequences from any unilateral government decision to initiate prosecution are inconsistent with any credible theory of optimal criminal punishment because they enable the government to impose disproportionate punishment on its selected targets through NPAs and DPAs.
• The third section then applies this analysis to one recent DPA that the Department of Justice (DOJ) concluded with great fanfare with Toyota Motors regarding defects in Toyota’s acceleration pedal.

Government Zealotry in a World of Uncertainty

A convenient demonstration of the one-sided view of the topic is found in the rhetoric of Eric Holder, then Deputy Attorney General of the United States under President Bill Clinton and now Attorney General under President Barack Obama. Holder’s 1999 memo started:

General Principle: Corporations should not be treated leniently because of their artificial nature nor should they be subject to harsher treatment. Vigorous enforcement of the criminal laws against corporate wrongdoers, where appropriate, results in great benefits for law enforcement and the public, particularly in the area of white collar crime. Indicting corporations for wrongdoing enables the government to address and be a force for positive change of corporate culture, alter corporate behavior, and prevent, discover, and punish white collar crime.[1]

Holder’s belief, which he still holds today, is that DPAs and NPAs make sure that crime will not pay. The deterrence that the government achieves at low price in these cases pays handsome social dividends, it is claimed, in the form of lower levels of criminal abuse by other businesses, which well understand the unwelcome fate that awaits them if they cross the line into criminal behavior.

Like all grand overgeneralizations, this proposition contains a grain of truth. In a world of certainty, aggressive use of DPAs and NPAs will harm only the guilty while sparing the innocent, thereby giving a huge leg up to private parties that refuse to deviate from the straight and narrow. At the very least, the imposed sanctions will suck all of the gains out of these illicit activities. Extra punishment, often in the form of heavy fines, thus can be routinely imposed without negative side effects. Rational individuals and firms will see the writing on the wall and, so informed, will consciously refrain from taking the actions that provoke this stern government response. The dislocations to the individuals and firms that are caught thus yield large social gains by their indirect effects.

One clear implication of this optimistic theory of perfect deterrence is that the level of criminal activity should trend sharply toward zero. Accordingly, the number of criminal cases will tend to zero as well, which allows the state to pare back its overall enforcement effort.

In a world of certainty, these strong penalties are in principle warranted. The standard theory of deterrence states that punishments should be set high enough to make sure that a wrongdoer is left worse off by committing crimes than by acting in law-abiding ways. The fly in the ointment comes from introducing into the argument certain irreducible facts of criminal enforcement: uncertainty, error, and administrative costs (for government) and compliance costs (for private parties). Now the happy deterrence story turns into a myth.

Each of these three elements forces conscientious thinkers to back off from the gung-ho model of criminal enforcement. The first two elements are linked closely together, as uncertainty pervades the administration of criminal justice. Many cases of criminal liability turn on questions of intent or knowledge, which often require quite specific information as to what was intended in a particular case or why. Other cases depend on the introduction of evidence that is difficult to collect and interpret and that may run afoul of the various constitutional guarantees under the Fourth and Fifth Amendments.

Important errors can thus creep in even if the government agents are scrupulous in their charging behavior, and the dangers become even greater whenever crusading prosecutors take over a high-visibility case. The errors in these cases can go to the issue of whether any crime has been committed at all or whether the prosecutors have ramped up the punishments above and beyond the optimal level, given that excessive enforcement is an inherent risk of criminal enforcement.

It is worth recalling that the once-notorious prosecution of Arthur Andersen, Enron’s auditor, destroyed the entire firm, even though the indictment addressed asserted misconduct in a small portion of the firm concerned with ambiguous instructions on document shredding. To add insult to injury, the United States Supreme Court unanimously found that the government indictment that sank Andersen was defective on elementary grounds.[2] The United States had read the “corruption” requirement out of the relevant statute by claiming that any effort to persuade another person to destroy the documents in question sufficed. That faulty instruction did not distinguish between a program of lawful document destruction and one that was intended to obstruct any government investigation of certain Enron-related transactions.

It is equally worth noting that the recent DPA that hit JPMorgan Chase for $1.7 billion for its alleged role in the Bernard Madoff scandal also rested on a very sketchy theory of criminal liability. Some JP Morgan Chase officers had heard rumors that a “cloud” hung over Madoff’s operation during the course of supplying routine administrative services for the firm. The government’s own exhaustive Investigation of Failure of the SEC to Uncover Bernard Madoff’s Ponzi Scheme—Public Version[3] pointed out that many individuals both inside and outside of the SEC had suspicions that the entire situation was not entirely above board but did little, if anything, about it.

Yet the report did not take any action against any SEC employees, nor did it take any action against any other private parties who might have known this information. Indeed, at no point did the SEC report even mention JPMorgan Chase as a potential target for either criticism or investigation. The risk of selective prosecution by a populist Attorney General against a firm whose head, Jamie Diamond, was a critic of the Obama Administration’s policies on money and banking surely lurks in the shadows of this case.

These cases do not stand alone. In their recent detailed examination of these agreements, James Copland and Isaac Gorodetski of the Manhattan Institute issued a detailed study of DPAs and NPAs. The Shadow Lengthens: The Continuing Threat of Regulation by Prosecution[4] notes that the use of DPAs and NPAs began to increase in 2005 and continues unabated to the present day. Over the past 10 years, they report, public sources reveal about 278 DPAs and NPAs. In the George W. Bush years from 2004 to 2008, there were 115 such settlements, for an average of 23 per year and a peak of 41 in 2007. In the five Obama years from 2009 to 2013, that total rose to 163, for an average of about 33 per year, with a total collection of $25 billion. The list of targeted firms includes Ralph Lauren, Royal Bank of Scotland, GlaxoSmithKline, and HSBC. In February 2014, JPMorgan entered into a DPA over the Bernard Madoff scandal for about $1.7 billion, even without any direct evidence of fraud in the situation.[5]

The threat that drives such settlements is the evident risk that a criminal prosecution can lead to the ultimate sanction against banks: revocation of their charter to do business.[6] Well aware of these consequences, federal prosecutors have exacted a guilty plea from Credit Suisse by persuading both the Federal Reserve Bank of New York and Benjamin M. Lawsky—New York State’s first Superintendent of Financial Services, the most powerful position in New York—not to lift the licenses of Credit Suisse after it agreed to enter a criminal plea. The guilty plea was then coupled with a $2.6 billion fine that did not impose any criminal sanctions on any current employee of the firm, although eight former employees were prosecuted.[7] The market did not respond negatively to the settlement; in fact, its share prices rose, doubtlessly because of the way in which the government pulled its punches.[8]

Settlement Theory

The question then arises whether the outcomes of these settlements should be evaluated on the benign theory that the higher the fine, the better the result, but there are strong theoretical reasons why this optimistic approach leads to a very bad assessment of the overall practice. To see why, it is necessary to give a brief outline of the settlement of disputes, both with and without DPAs or NPAs.

The stripped-down account of settlements in ordinary civil cases runs as follows. The plaintiff files a complaint that asks for a certain amount of money. The plaintiff knows that there is some chance that she will not win the case and that she will be required to make certain expenditures in order to bring the lawsuit to completion. The expected rate of return from litigation is equal to the anticipated amount of recovery multiplied by the chances of winning, less the costs of bringing the suit, which are incurred in all cases.

In looking at the first term, dealing with expected return, it matters whether the plaintiff is risk-neutral or risk-averse. (There are few cases in which parties prefer to take on litigation risk.) The standard assumption is that there is some level of risk aversion. At this point, the settlement pressure on the plaintiff arises from her ability to eliminate both the uncertainty and the expense of suit. The defendant faces reciprocal incentives. His costs equal the expected amount of damages, which takes into account both the magnitude and the probability of losses, plus the certainty of incurring costs in defending the case.

The reason why settlement is possible in these situations is that both sides gain from eliminating legal uncertainty and litigation costs. So long as the plaintiff’s lowest demand is below the defendant’s highest offer, the parties could well strike a bargain from which both benefit by reducing uncertainty and saving out-of-pocket costs.

No one claims that this process is easy. The parties could differ in their estimation of the relevant variables in the equation. If the bargaining range (between the plaintiff’s minimum ask and the defendant’s maximum offer) is large enough, no result leaves both parties better off. Both parties will be tempted to posture and bluff before reaching the final settlement. Thus, if the plaintiff is willing to settle for $100 and the defendant is willing to pay $200, any settlement between $101 and $199 will leave both sides better off. But there is no reason why the settlement will be for $150, as the bargaining skills of the two parties could push that figure one way or the other. Each round of bargaining reduces the gains from settlement, which places an additional burden to conclude the deal quickly.

That said, even with these inevitable complications, on average this rocky process works tolerably well. The upshot is that the settlement figure tends to reflect the anticipated costs of defendant’s conduct as measured under the underlying substantive rule. The plaintiff gets less through a settlement than through successful litigation. The defendant pays more than is required under successful litigation. The incentives for settlement are properly aligned for both parties.

One key assumption about this simple model is that it introduces no fundamental asymmetries to either the litigation or settlement process. In principle, criminal settlements, often called plea bargains, should be amenable to resolution by the same rules where either fines or years in prison are substituted for the damages that are routinely awarded in civil litigation. To be sure, the criminal system relies on a very different system of discovery[9] and imposes a higher burden of proof on the prosecution, but these changes alter only the probabilities of success, which in turn influences the size of both the expected penalty and the cost of litigation.

Accordingly, the government takes a cut in collecting a criminal fine or imposing a sentence, both based on the perceived weakness of its case. The ultimate sanctions thus fall midway between those of a failed prosecution and those of a successful prosecution, which is as it should be.

Unfortunately, the dynamics of criminal settlement are upset because of the major collateral consequences of any criminal prosecution. I have called this switch the “Grand Inversion”[10] to reflect the brutal fact that the danger of the government’s decision to prosecute for any reason is more deadly than conviction of the same offense on proof beyond a reasonable doubt. In cases against corporations, imprisonment is not an option, but fines are. In this context, the simple inversion means that DPAs and NPAs made before trial could result in larger sums than fines imposed after conviction because both DPAs and NPAs deviate systematically from the proper progression of litigation.

Ideally, the government in a criminal case should get its greatest return after it proves its case in open court beyond a reasonable doubt. If that case is settled prior to the final verdict, the government’s need to show less should mean that it gets less in return.

But it does not. That benign outcome is frustrated by the complex interaction between the settlement process used in criminal cases and the heavy collateral consequences that can be brought to bear against many individuals and firms. In many instances, the simple fact of criminal prosecution by the federal government triggers the loss of licenses to do business under independent state or federal laws even if no conviction follows. For a large corporation, or even for a successful professional, the consequences of that loss can easily dwarf the criminal penalty.

Likewise, both firms and individuals fear criminal prosecution because of the effects on their business reputation even if their licenses are not suspended. The maxim of most people is “steer clear of trouble,” especially when one must explain one’s actions to restive shareholders or business partners. Who wants to deal with any individual or firm that is under indictment, lest they expose themselves to derivative lawsuits or other sanctions from dealing with a person or firm under a cloud? It is better to be safe than sorry, so in a competitive market, the prudent party finds some new trading partner that does not labor under these evident disabilities.

Critically, the stakes are higher because most complex forms of dubious conduct may generate potential liability under multiple theories, which makes negotiating a comprehensive settlement difficult. The expected values of fines and punishment therefore can exceed the values that satisfy any requirement of optimal deterrence—that is, under any theory that weighs excessive deterrence on the same scale as inadequate deterrence.

These bargaining dynamics tie into a system that features high levels of prosecutorial discretion and that imposes precious few constraints in charging defendants. The relentless expansion of the criminal law adds further avenues for potential prosecution, which makes it even more difficult to cabin prosecutorial discretion.

Indeed, once the prosecution is brought, the criminal defendant knows that the probability of a loss of licenses, reputation, or goodwill approaches certainty, dwarfing any possible fine on conviction. It constitutes the entire loss of business. In the Arthur Andersen prosecution, the indictment carried a maximum penalty of $500,000, but the fact of prosecution reduced a firm worth $9 billion to bankruptcy, dislocating hundreds of partners and causing some 28,000 employees to lose their jobs—none of whom was guilty of any wrongdoing.

No legislature would pass a law stipulating that Arthur Andersen’s supposed corruption offense in shredding documents should carry anything like a $9 billion penalty. Why, then, arm the government with a weapon that in effect can impose such penalties even before anything resembling due process has run its course? No firm that faces these collateral costs will do anything other than capitulate, given that the burdens under a DPA or NPA leave it better off capitulating than winning in court after the prosecution has taken place.

Overdeterrence leads to misallocation of social resources by forcing defendants to take excessive precautions against losses, thereby diverting resources better spent elsewhere to reducing the risk of criminal behavior. Yet those better services and better products could also reduce the risk of loss from product failure or financial irregularity. When the criminal law forces a real diversion of resources to compliance work, overall levels of innovation could slow down with attendant social losses in other areas, including safety issues. This point is not entirely hypothetical, as is made clear by looking at a recent DPA introduced with great pomp and circumstance.

The DOJ Toyota DPA with a $1.2 Billion Financial Penalty

The concerns mentioned above are not just theoretical abstractions. They set the dynamics for many government deferred prosecution agreements. On Wednesday, March 19, 2014, the Department of Justice triumphantly announced that it had exacted a $1.2 billion DPA from Toyota because of the company’s serious mishandling of two serious defects that led its accelerator to stick with potentially fatal consequences. The DPA also called for “an independent monitor to review and assess policies, practices and procedures relating to TOYOTA’s safety-related public statements and reporting obligations.”

It should be noted that the investigation that led to the deferred prosecution agreement was not for the safety losses associated with the use of the defective accelerator, but rather was separate punishment for “a criminal wire fraud charge” brought against Toyota, which had “defrauded consumers in the fall of 2009 and in early 2010 by issuing misleading statements about safety issues in Toyota and Lexus vehicles.”

The agreement lasts for three years, after which the government will seek to have the charges dismissed against Toyota if it “abides by all terms of the agreement.”[11] Nothing is said about what happens if Toyota slips up on any matter, large or small, within that time period.

As the DOJ’s announcement relates, the tragic accident that precipitated this DPA occurred in 2009 in San Diego, California, when a family of four was killed after “a Lexus dealer had improperly installed an incompatible all-weather floor mat into the Lexus ES350 in which the family was traveling and that mat entrapped the accelerator at full throttle.” It seems quite clear that this unfortunate set of events creates serious liabilities for Toyota and its dealer even under the narrowest substantive theories of product liability, which restricts recovery to those cases in which the defendants are responsible for a latent defect in a vehicle that remains in its original condition and causes serious harm when the vehicle is used in its normal and proper fashion.

By the same token, however, this simple fact pattern also makes it clear that in the individual case cited, the main culprit was the dealer who made the improper installation, not Toyota, whose mats caused no danger in the overwhelming number of cases when they were properly installed. It is equally clear that any measure of the severity of the defect is influenced by the likelihood that some future mishap will occur.

Yet nothing in the DOJ account suggests that this defect caused any other fatal accident or that the probability of further loss was high. Nor is there any recognition in the announcement that Toyota had already been subject to heavy tort liability that already is calibrated to supply the appropriate level of deterrence for the particular deaths.

The situation is still more complex because this particular defect related to a matter brought properly before the National Highway Traffic Safety Administration (NHTSA), one of whose major functions is to see that automobile companies correct defects in automotive design before they can cause further harm to individuals. The DOJ report makes a powerful case that Toyota wrongfully concealed from NHTSA vital information about the dangers to the accelerator pad, not only regarding the mat issue, but also regarding the risk that it would stick for other reasons. With respect to the floor mats, Toyota managed to limit the recall to the mats without having to make any changes in the vehicle, “a result that Toyota employees touted internally as a major victory: ‘had the agency…pushed for recall of the throttle pedal assembly (for instance), we would be looking at upwards of $100 + million in unnecessary costs.”[12]

The DOJ press release uses this quotation to establish the bad faith of Toyota. In fact, it suggests a narrative that is at least as critical of NHTSA as it is of Toyota. Toyota position’s is that fixing the mat is a cheaper way to solve this problem if it secures clearance for the accelerator pedal, even if the mat is improperly installed. That solution could be $100 million cheaper than a more expensive product redesign. If it is, Toyota’s position is consistent with advancing overall social welfare, while NHTSA, by contrast, has not followed a sensible cost-benefit analysis.

Again, there is a serious social loss if NHTSA uses the occasion of a potential product defect to impose on firms wasteful expenditures that offer little or no additional safety benefit to their customers. If that charge is true, a top-to-bottom examination of how NHTSA conducts its safety investigations is needed.

The DOJ release quoted Transportation Secretary Anthony Foxx thanking NHTSA investigators who worked “tirelessly” in the investigation and “succeeded in this effort in spite of extraordinary challenges.”[13] But the overzealous enforcement of its safety mandate carries with it serious risks for consumers, and the government’s conduct puts all automobile companies in a difficult position. If they do not report defects that have an exceedingly low probability of causing further harm, they risk having the book thrown at them. If they do report those low risks of harm, they could be saddled with costly requirements to make unnecessary or even counterproductive design changes. An overzealous NHTSA thus creates safety risks of its own.

These observations, of course, do not excuse Toyota’s conscious decision to conceal safety risks from NHTSA. Nor is it my point here to contest the “record civil penalties of more than $66 million”[14] levied against Toyota for its serious and deliberate breach of statutory duties. That fine sets an instructive baseline against which to measure the $1.2 billion fine that Holder announced for the criminal wire fraud case.

By any standard of criminal justice, that fine is absurdly out of proportion to Toyota’s injury to its consumers. The wire fraud charge is marginal to the serious issues raised by this case, which are the deaths and the breach of statutory duty, both of which were independently addressed previously. The supposed consumer deception in this case cannot therefore include any cases of physical harm because the civil suits brought against Toyota already fully accounted for them—perhaps with punitive damages. Nor can it include the punishment for seriously misleading NHTSA.

The only purpose served by this huge fine was to correct the public misimpression created by Toyota’s false reassurances that it had gotten to the “root” of the problem when it had not in fact done so. No evidence shows that Toyota’s misleading statements had any influence on consumer behavior in buying Toyota vehicles that even approaches the impact of either the deaths on the one hand or the misrepresentations and concealments from NHTSA on the other. Why, then, should DOJ impose a penalty that is nearly 20 times larger than the one that NHTSA imposed for a far more serious offense? And why did DOJ worry about oversight of public releases when the serious issues are in the ongoing relationship that Toyota has with NHTSA?

We have, therefore, a DPA the severity of which is utterly inconsistent with the severity of the underlying offense. It is critical to recall the government rhetoric that led to this outsized penalty. Secretary Foxx insists that the full brunt of these penalties is that “they send a powerful message to all manufactures to follow our recall requirements or they will face serious consequences.”[15] But there is nary a word about whether these processes are sensible or not, nor is there an awareness that the heavier the penalties, the longer it will take to bring mistakes in the NHTSA processes to light.

U.S. Attorney for the Southern District of New York Preet Bharara toes the same line when he insists that “tens of millions of drivers in America have an absolute right to expect that the companies manufacturing their cars are not lying about serious safety issues; are not slow-walking safety fixes; and are not playing games with their lives.”[16] Again, however, there is not a single word about why the NHTSA penalties were insufficient to address this issue, nor is there any explanation regarding what the optimal penalties should be, or how fast the corporate corrections should be made, or even whether the fines should be against the corporation and their innocent shareholders instead of the individual officers in the company who hatched the inappropriate plan. This soaring rhetoric of absolute rights leads the analysis in the direction of counterproductive overdeterrence.

The one good piece of news from this public announcement is that it is subject to judicial review. Ideally, that review should have taken place prior to announcement of the agreement so that Toyota was not presented with a fait accompli in the hearings that follow. As matters now stand, it looks as though the judge in question will be asked to rubber stamp the agreement. Toyota, which remained mum when the DPA was announced, cannot be expected to raise a word in protest, lest the process ratchet up even further for noncooperative behavior.

Yet in light of all of the difficulties associated with DPAs in general and with this DPA in particular, the judge who hears this case should take the initiative and ask the DOJ to explain why the penalty on a single count of criminal wire fraud charge should generate liabilities that are far greater than those imposed by either the tort system or NHTSA itself. It would be a sensible move for that judge to appoint an independent party to challenge the settlement, given that Toyota cannot be expected to do so. This proposal is not inconsistent with recent British efforts to impose extensive oversight on DPAs and NPAs.[17]

Conclusion

The hard question that remains is whether any other effective steps can constrain the use of DPAs and NPAs. The problem is a difficult one. Right now, the most blatant abuses look to be a thing of the past. It is no longer permissible to do what Christopher Christie did when he was U.S. Attorney for the District of New Jersey, when his DPA with Bristol–Myers Squibb over violations of the securities laws led to the endowment of an ethics program at Seton Hall Law School, Christie’s alma mater.[18] Nor is it possible today to include in a DPA a provision that requires a company such as KPMG LLP to renege on its promise of legal assistance for its employees who are facing government prosecution.[19]

But stopping these evident abuses does not get to the core of the problem with DPAs and NPAs: their inverted incentive structure, which gives potential defendants far more to fear from prosecution than from conviction. The modest proposal for third-party review offers the best shot at some kind of external control of prosecutorial behavior. The system of external review might also be expanded to include periodic reports to the public over any significant patterns in the administration of these agreements.

It is far too soon to say whether these measures could undercut the destructive dynamic that now drives these settlements. It may seem trite to say, but perhaps the best protection against abuse would come from prosecutors internalizing a set of norms that helps to guard against excessive practices. Judging from the record of the past decade, that level of self-awareness has been sorely missing at the federal level in recent years.