The Game Theoretical CEO: An Inexplicable Lawful Agent

By William K. Black
(Cross posted at


This is the sixth (and final) of my series of articles on the work of Roger Myerson, a 2007 Laureate in Economics.  Myerson’s work on CEOs is typical of the game theoretical approach to explaining the behavior of CEOs and firms, so I am discussing an exemplar rather than an outlier.  This installment discusses some of the fatal flaws that I argue characterize the game theoretical work on CEOs by the Laureates.  I will urge that they are weakest where they believe they are strongest – their models.  The article explains why the models are specified incorrectly because the models have no coherent theory (or understanding) of fraud or ethics.  The game theoretical Laureates (Laureates) make unsupportable implicit assumptions that are belied by the data and internally inconsistent with their explicit assumptions.

Replacing the Laureates’ unsupportable assumptions would be fatal to their models.  Instead of predicting efficient, incentive-compatible behavior by CEOs and firms, their models, if properly specified in a manner that is logically consistent with their explicit assumptions, would predict endemic control fraud.  The Laureates’ policy recommendations are so criminogenic that they have greatly increased the harm done by control fraud.  Their policies have made the incentives CEOs face far more perverse, producing the epidemics of accounting control fraud that drive our recurrent, intensifying financial crises.  The irony is that the scholars who purport to have made the study of incentives and their design their area of expertise have failed to understand CEO incentives and have designed mechanisms that they thought improved those incentives but actually made them far more perverse.

The good news is that their explicit assumption of (and definition of) “rational” behavior is (depending on how one wishes to characterize the flaws) either wrong or so immoral that well-functioning societies reject it.  Classical economists generally had a superior understanding of fraud by CEOs than do the game theoretical Laureates.

The Laureates’ explicit assumptions about CEO behavior and goals

The Laureates three explicit assumptions about CEOs are that they are rational wealth maximizers who differ in skills.  It is unclear why they assume wealth maximization v. utility maximization for they do not explain the basis for the assumption.  Putting aside the enormous problems that economists have identified with assuming either utility or wealth maximization I note for these purposes that when considering the importance of fraud and morality the differences between assuming wealth v. utility maximization can be large.

The assumption of differential skills is realistic.  Myerson makes the simplifying assumption that CEOs are either “bad” or “good.”  Good and bad, however, have no moral content as Myerson uses the terms.  The difference is simply skill.

Myerson’s Contradictory Models of CEO behavior

Myerson takes the moral out of Moral Hazard

I explained in prior articles that Myerson presented two inconsistent models of bank CEO behavior only three months apart in 2012 without any discussion of why he did so.  Both articles purported to counter the CEO’s “moral hazard.”  As I explained in a prior article, moral hazard theory predicts that it will produce either (or both) excessive risk by a CEO or that the CEO will lead a “control fraud.”  That term, from criminology, refers to leaders of seemingly legitimate entities that use them as a “weapon” to defraud.  In finance, accounting is the “weapon of choice.”  Bank CEOs can lead multiple kinds of control frauds because they are not mutually exclusive.  Myerson, however, implicitly assumes control fraud out of existence.  More radically, he assumes all immoral behavior out of existence – in models that purport to model “moral hazard.”  Myerson’s approach is not only false and fatal to his model and policy designs – it is also bizarre.

It is a particularly odd implicit assumption for Myerson to make even from a traditional economics perspective for his work is part of a body of work on CEO incentives that stressed two other asserted problems with CEOs – “shirking” and unwillingness to have the firm undertake some prudent risks.  Shirking is a form of fraud and is immoral.

The Laureates do not explain why they implicitly assume shirking and risk avoidance by CEOs out of existence.  The implicit assumption that CEOs do not seek to avoid prudent risks (the original claim of neoclassical scholars like Michael Jensen who developed the rationale for what the Laureates would later call “incentive-compatible” executive compensation) is deeply troubling because Myerson’s September 2012 model predicts the opposite.  Myerson’s model predicts that CEOs will cause their firms to take on even highly imprudent risks because they are subject to moral hazard.  Myerson’s September 2012 argument is that CEOs have asymmetrical payoffs.  If they cause the firm to take great risks that succeed the CEOs receive extraordinary compensation.  If the great risks lead to the firm’s failure (and in an efficient market the only way to secure a much higher expected return is to take a much greater risk of failure) the CEO will not be liable to the shareholders for the loss (and would not have the net worth to make them whole even if he were not protected by limited liability and the business judgment rule).

The Laureates should, of course, be troubled by the contradictory predictions that neo-classical economists have made about CEOs’ appetite for risk.  They should be troubled enough to make explicit assumptions about CEO behavior and incentives supported by theory and data.  It is disturbing that it did not lead to these minimum responses.

Moral Hazard predicts increased control fraud under Myerson’s model

Understanding the different modes of control fraud and their implications is essential to modeling moral hazard, so Myerson’s implicit decision to assume control fraud out of existence is fatal to his model.  There are several means of classifying control frauds, e.g., by type of entity (for profit, non-profit, or governmental) or whether the fraud is “opportunistic” or “reactive” (a failing firm creates even more intense perverse incentives for the CEO).  For purposes of discussing Myerson’s work classification based on the identity of the primary intended victim is most useful.

  1. Anti-consumer control fraud (failure to pay full compensation due, violating safety and health requirements protecting workers)
  2. Anti-purchaser control fraud (the deliberate sale of “lemons” (inferior quality and/or dangerous goods that maim and kill)
  3. Anti-public control fraud (tax fraud, environmental crimes, bribery)
  4. Accounting control fraud (“looting” – principal victims are creditors and shareholders)

Control fraud creates a fatal flaw in Myerson’s September and December 2012 models.  Under the September model the great bulk of the CEO’s compensation is not paid until the CEO leaves the firm or dies.  At that point he receives an enormous payment if the firm reports that it is profitable or none of his deferred income if the firm reports it has been unprofitable.  This represents the “mechanism” that Myerson has “designed” to ensure that the CEO’s compensation is “incentive compatible” and will not succumb to “moral hazard.”  Myerson’s implicit (counterfactual) assumption are that shareholders have power over corporations, recognize the “agency” problem created by the CEO’s moral hazard, act successfully to counter it by preventing certain CEO abuses, and that the CEO lacks the power and the guile to counter his principals’ efforts to restrain his abuses.  Each of Myerson’s implicit assumptions frequently proves false, but this article will focus on the specific problems posed by control fraud.

Control fraud poses three fatal problems for Myerson’s assertion that he has designed a compensation mechanism that will prevent CEOs from engaging in abusive behavior due to moral hazard.  First, the Laureates’ models assume that it is costly for shareholders to monitor and constrain “unfaithful” agents.  Indeed, their theory is often referred to as “agency cost.”  Shareholders, therefore, would not under Myerson’s model (which assumes they are wealth maximizers) undergo the cost of preventing the CEO from causing the firm to engage in abuses that increase the firm’s reported income and increase the shareholders’ wealth.  If anything, the shareholders would encourage control fraud under Myerson’s model.  This is particularly true if the bank were failing because control fraud could allow existing shareholders to maintain or even increase the value of their stock and perhaps to sell their stock and achieve a gain rather than suffer a loss.

Second, Myerson’s mechanism does not simply fail to prevent control fraud, it greatly encourages it.  The September version of the compensation plan for CEOs that Myerson designed vastly increases the CEO’s incentive to engage in fraud because it will produce far greater income, with much greater certainty, and allow him to avoid the draconian denial of compensation that Myerson’s mechanism would provide to CEOs who failed to report firm profitability.  George Akerlof and Paul Romer explained why control fraud is a “sure thing” in their classic 1993 article (“Looting: The Economic Underworld of Bankruptcy for Profit”).

“[M]any economists still seem not to understand that a combination of circumstances in the 1980s made it very easy to loot a financial institution with little risk of prosecution. Once this is clear, it becomes obvious that high-risk strategies that would pay off only in some states of the world were only for the timid. Why abuse the system to pursue a gamble that might pay off when you can exploit a sure thing with little risk of prosecution?” (Akerlof & Romer 1993: 4-5).

Third, control fraud is dynamic and criminogenic – control fraud begets control fraud.  George Akerlof explained this process in his most famous article on markets for “lemons” in 1970 where he described anti-purchaser control frauds that deceived the consumer about the quality of the goods being sold.  Akerlof coined the economic metaphor we know as a “Gresham’s” dynamic.

“[D]ishonest dealings tend to drive honest dealings out of the market. The cost of dishonesty, therefore, lies not only in the amount by which the purchaser is cheated; the cost also must include the loss incurred from driving legitimate business out of existence” George Akerlof (1970).

When firms that engage in control fraud gain a competitive advantage market forces become perverse.  Control fraud creates perverse incentives that can make fraud common in particular trades or even entire nations with weaker law enforcement and regulatory systems.  Akerlof used two examples of anti-purchaser control frauds in his “lemons” article.  He discussed only the financial aspect of the losses to consumers, but both of his examples posed health risks to the consumers (selling defective cars and adding small stones to bags of rice).  More vicious forms of anti-purchaser control frauds are common.  They maim and kill tens of thousands of people, e.g., the sale of fake infant formula and anti-malaria drugs.

One of the defining elements of most crimes is that they often inflict severe negative externalities and reduce societal wealth in addition to the harm they cause the direct victim.  The reduction in the scope of women’s freedom inflicted by fear of sexual assault is one example most people understand.  In the case of accounting control fraud the negative externalities can include the loss of trust.  Fraud consists of the perpetrator getting you to trust him – followed by him betraying your trust to take something of value from you.  As a result, elite accounting control fraud is the most destructive acid in destroying trust in the markets.

Accounting control frauds also have the ability and incentive to suborn “independent professionals” (auditors, appraisers, attorneys, loan brokers, and credit rating agencies) and pervert them into the fraud’s most valuable allies.  CEOs and their senior officers are ideally placed to be able to use the ability to hire, fire, promote, and compensate to suborn these professionals.  This is why accounting control frauds that are massively insolvent and unprofitable are frequently able to report record profits for many years.  It also explains why we observe recurrent “echo” epidemics of control fraud among supposedly independent professionals.

Markets and executive compensation mechanisms cannot be “efficient” or optimal if they induce control fraud.  Markets beset by substantial control fraud are perverse and dangerous, even murderous.  Regulatory “cops on the beat” supported by a vigorous, competent criminal justice system that is willing to prosecute elites are essential to prevent the frauds from gaining a competitive advantage and creating a Gresham’s dynamic.

Myerson’s model does not simply lack any mechanism from countering control frauds that make the firm more profitable – it requires that such fraud become the “dominant” strategy.  This is inherent in a model that eliminates ethics and sanctions and conflates “rationality” with wealth maximization.  A CEO that fails to engage in control fraud is not “moral” under Myerson’s model – he’s meshugenah.  Gregory Mankiw (unintentionally) captured the full insanity of conflating “rational” and “wealth-maximizing” when he was the discussant for Akerlof and Romer’s presentation of their paper on “looting.”  Mankiw said ““it would be irrational for savings and loans [CEOs] not to loot.”  I call this “Mankiw morality” and it is characteristic of the Laureates’ models.

Daniel Fischel infamously claimed in his book decrying the criminal prosecution of Michael Milken that it was “the government’s” fault that so many CEOs looted “their” S&Ls because they had financial incentives to loot.  This is moral reasoning that one would be embarrassed to hear coming from one’s three-year old.  It is typically impossible to remove the financial incentive to engage in fraud.  Both of Myerson’s 2012 models of CEO compensation that are supposed to transform the CEO’s moral hazard into “incentive compatible” (and the models of all the Laureates that implicitly assume control fraud out of existence) greatly increase the incentive to engage in control fraud.  A different historical example and form of white-collar crime may help make the point clear.  I have drawn on the British Library’s description of The East India Company (Company).  Its first voyage to India and Sumatra was on the verge of commercial failure when one of its entrepreneurs saved the project.

“The problem was that Sumatrans were not very interested in trading their precious spices for these goods – given the climate in Sumatra, wool seemed particularly useless to the natives. So Lancaster acted quickly: he decided to capture a Portuguese carrack (a large ship) and steal its rich cargo of gold, silver and Indian textiles. These goods allowed him to buy pepper at Aceh.”    

A company that first prospered through theft was not destined to be a beacon of moral integrity so it is not surprising that the Company’s strategic plan was to become a de facto military force with its own (not very) diplomatic policy:  “The Company started to protect its trade with its own armies and navies….”  These actions of the Company were only prudent given that the Company’s definition of “protect[ing] its trade” included the seizure of large vessels and their “rich cargo.”

The Company then proceeded to get Britain into a war with China, but at least the Company’s motives were pure and the British got Hong Kong.

“From China, the Company bought tea, silk and porcelain. The Chinese wanted silver in return. Over the next 100 years tea became a very popular drink in England, and there was a fear that too much silver was leaving the country to pay for it. To stop this happening, the Company became involved in a triangular trade by smuggling opium (a highly addictive and illegal drug) from India into China.

The Company grew opium in India. They were looking for something that the Chinese would accept instead of silver, to pay for the goods they bought at Canton. Opium was a valued medicine which could deaden pain, assist sleep and reduce stress. But it was also seriously addictive and millions Chinese became dependent on the drug.

Although opium smoking was a subject of fascinated horror for Europeans, the Company actually encouraged people to use the drug in China – sales of opium were extremely lucrative. As a result, millions of Chinese would die from opium addiction, and the very fabric of Chinese society was threatened.”

When the humiliation of China produced the Opium War it was fought by the Brits on grounds the WTO would have loved:  the British demanded the right to engage in the “free trade” – of opium.  Small wonder that when Gandhi was asked what he thought about “English civilization” he replied:  “I think it would be a good idea.”

The Company serves as an example of several of the key points I have made.  The fact that a firm’s CEO tries to increase his wealth provides no assurance that he will run the firm in an honest, ethical, efficient, or socially productive fashion.  The Austrian concept of “spontaneous order” can lead to “entrepreneurial” solutions involving slavery, theft, terror, mass murder, and piracy.  The markets can destroy for wholly uncreative reasons and results that harm society.  The Company had a contemporary counterpart, The South Sea Company that produced the famous bubble.  The principal value of the South Sea Company was that:

“[A]fter the conclusion of the Treaty of Utrecht (1712), the South Sea Company was given sole rights to carry on British trade with Spanish America – the so-called South Seas. The South Sea Company had also obtained for 30 years the Asiento de Negros, a contract to be the sole supplier of slaves to the South Seas. Britain already had colonies in the Caribbean and, as a result, had a large share of the slave trading market in the Western Hemisphere. The Company certainly looked to be well positioned in this fruitful new market.”

Modern economic historians who specialize in statistical approaches have an impulse I have never been able to comprehend to write in these callous terms.  I will not attempt to defend their practices.  For my purposes I will emphasize that slaves are certainly sold at things called “markets” but this is in not a “market” – it is a crime site.  Slavery is kidnapping, conducted through force, maintained by physical and psychological torture during and after the mass murder of the middle passage, inherently tied to the theft of labor, and heavily associated with rape and racism.  It is not “fruitful.”  It was murderous.  It caused, and continues to cause incalculable harm.  Slavery harmed society.    

The shareholders of the East India Company and the South Sea Company, under the Laureates’ theory of wealth maximization would have no incentive to perceive the theft of a Portuguese vessel and its rich cargo, the huge export of opium to China, or sparking the Opium War as presenting any “agency” problem requiring any discipline.  The East India Company’s actions may have killed millions of Chinese, but they added to the shareholders’ wealth.  Similarly, shareholders purchased stock in the South Sea Company precisely because it had a monopoly right to commit one of the greatest crimes in history.

When the Laureates conflate rationality with wealth-maximization and ignore ethics they create models and policies that encourage horrific crimes by CEOs that we have spent centuries trying to eradicate.

I have explained in prior articles why Myerson’s December 2012 paean to plutocracy is even more criminogenic than his inconsistent September model.  The December 2012 raises the stakes greatly in favor of control fraud because it produces a “sure thing” and because Myerson’s model would require CEOs of very large firms to be billionaires who put a huge portion of their wealth at risk of suffering a total loss.  Myerson’s December 2012 model would maximize adverse selection of CEOs for the ones most willing to serve would be those most likely to lead control frauds because they had used Myerson’s model of politics to ensure that crony capitalism would protect them accountability.

The Laureates’ Models Contradict their Claims about Model Specification 

The supposed strength of the Laureates’ models is actually their greatest failure.

“Appearances to the contrary, the principal feature of the cliometric [quantitative historical] approach is neither its preference for hard, quantitative data nor its reliance on computer technology and mathematics. Rather, it is the effort to specify explicitly the usually implicit assumptions about causation that underlie and make possible any explanation of human behavior. When formally stated and systematized, these theoretical assumptions constitute the cliometrician’s “model,” and in the ideal case the model should be clear enough to express in the perfectly unambiguous form of an algebraic equation. The mathematical aspect of the method derives from this effort to specify assumptions rigorously. [1]

Conventional historians tend to dismiss all this painstaking specification of assumptions as misplaced precision, or, worse, a futile aping of scientific method. If a cliometrician were to write the history of the crucifixion, according to a current historical joke, he would begin by counting the nails. But the cliometricians are on strong ground when they reply that conventional historians often get away with fuzzy thinking by leaving their theoretical assumptions implicit, or, as is often the case, simply unexamined.”

[1] R. W. Fogel, “The Specification Problem in Economic History,” Journal of Economic History, Vol. 27 (Sept. 1967), pp. 283-308.  [Robert Fogel was a Laureate in economics known for his work (Time on the Cross) arguing that slavery was far more benign than (modern) historians have admitted.  He died very recently.  In future columns, I will address his work.]

“The True & Tragical History of ‘Time on the Cross’” – The New York Review of Books

My criticism is that it is the Laureates’ whose models of CEO behavior have consistently and repeatedly failed to make explicit their assumptions, examine their appropriateness, and test their logical consistency and validity.  This makes their models and recommended policies so badly flawed that they are criminogenic.  It means that the patina of rigor in their work is actually rust arising from methodological errors that produce erroneous predictions and criminogenic policies convenient to the Laureates’ ideological dogmas.  When their models are specified correctly they falsify the Laureates’ predictions and warn that their policy advice is perverse.


Bill Black is the author of The Best Way to Rob a Bank is to Own One and an associate professor of economics and law at the University of Missouri-Kansas City. He spent years working on regulatory policy and fraud prevention as Executive Director of the Institute for Fraud Prevention, Litigation Director of the Federal Home Loan Bank Board and Deputy Director of the National Commission on Financial Institution Reform, Recovery and Enforcement, among other positions.

Bill writes a column for Benzinga every Monday. His other academic articles, congressional testimony, and musings about the financial crisis can be found at his Social Science Research Network author page and at the blog New Economic Perspectives.

Follow him on Twitter: @williamkblack


One response to “The Game Theoretical CEO: An Inexplicable Lawful Agent