This is the fourth installment of my exploration of the work of Roger Myerson, Nobel Laureate in economics in 2007. It is part of what will be a broader series of articles exploring why economics is unique among the sciences in awarding the Prize to scholars whose predictive work proves profoundly wrong and leads to public policies that cause great harm. The first installment used Myerson’s Prize lecture to explore his paean to plutocracy as the purported unique advantage of capitalism.
The second expanded this point by examining the model he proposed in his December 2012 lecture to prevent bank CEOs from causing harm in response to their “moral hazard.” I explained why Myerson’s model would intensify bank CEOs’ already perverse incentives to engage in control fraud and why his model of compensating bank CEOs was unethical (it rests upon the requirement that shareholders bribe bank CEOs so that they will “imitate” the conduct of ethical bankers) and would fail because it would require shareholders to accept negative returns.
The third installment noted that his September 2012 article argues that:
“In this sense, a tax on poor workers to subsidize rich bankers may actually benefit the workers, as the increase of investment and employment can raise their wages by more than the cost of the tax [p. 25].”
The same September 2012 article by Myerson discusses what he considers the relevant literature on executive compensation for bankers. The context is his unsuccessful effort to develop a model that explains why our supposedly efficient markets are producing recurrent, intensifying financial crises. Here, I discuss only his statements in the article about executive compensation.
A MODEL OF MORAL-HAZARD CREDIT CYCLES (March 2010, revised September 2012)
A brief reprise of Myerson’s December 2012 model of executive compensation
I explained in the first two installments Myerson’s claim that the only reliable means to prevent widespread abusive conduct by bank CEOs is to require that they make massive capital investments in “their” banks. Myerson emphasizes that only exceptionally wealthy individuals would be able to make such investments in a bank. This is the basis for his claim that plutocracy provides capitalism’s unique advantage over communism. Myerson’s December 2012 model calls for a minimum capital investment by CEO of 40% of the bank’s total capital to counter the CEO’s perverse incentives arising from moral hazard. The CEO would own such a huge portion of the stock that he would be the controlling shareholder of any publicly traded bank.
Myerson’s model also calls for extreme compensation for bank CEOs (roughly $1 billion for a mid-sized bank with $20 billion in assets). That compensation consists of two components. A CEO who invests “only” the minimum $400 million in the bank (assuming a total bank capitalization of $1 billion) called for by Myerson’s model would receive an annual $200 million bribe (Myerson euphemistically calls it a “moral-hazard” “rent” paid to the CEO to induce him not to rip off the shareholders by shirking) regardless of whether the bank succeeds. The CEO would also receive what Myerson calls a “wage” of $800 million if the bank’s investments were successful. (The details are set out in my second piece where I scale-up his model to represent my hypothetical bank with $20 billion in assets and $1 billion in capital.) In addition to this compensation the CEO would receive an investment return on his massive capital investment in the bank if the bank were successful.
Myerson’s September 2012 model of executive compensation
Myerson’s September 2012 exposition of the literature on executive compensation and his model of how compensation should be structured differ greatly from his December 2012 model of optimal means of compensating bank managers. Here is how Myerson explained the literature and his model in his September 2012 article:
“Since Becker and Stigler (1974) and Shapiro and Stiglitz (1984), it has been well understood in agency theory that dynamic moral-hazard problems with limited liability are efficiently solved by promising large end-of-career rewards for agents who maintain good performance records. So an efficient solution to moral hazard in banking must involve long-term promises of large late-career rewards for individual bankers. Such back-loading of moral-hazard agency rents requires that bankers must anticipate some kind of long-term relationship with investors [p. 2].
To motivate the agent to supervise appropriately at time t, the agent must be promised a greater reward from success in the next period t+1 than from failure [p. 5].
Fact 1. For any given investment h>0, the agent’s expected reward αv+ (1−α)w at time t+1 is minimized, over all feasible (v,w) satisfying incentive compatibility  and limited liability, by an incentive-compatible contract that promises rewards v=hM/α for success and w=0 for failure [pp. 5-6].
The optimal incentive plan in Fact 2 involves maximal back-loading of rewards (to the agent’s last period t+n) and maximal punishment of failures (termination of service without pay) [p. 7].”
The mathematics can be disregarded. The words convey the key analytical claims. Myerson’s view is that economists have known for nearly 40 years ago that the optimal way for a principal (the shareholders) to compensate an “agent” (the CEO) is to pay the maximum possible portion of the compensation to the CEO only when he leaves the bank. The CEO, of course, needs some current compensation to live on (or vast wealth – the core premise of Myerson’s December 2012 model). Putting together both models, the ability a bank would have to pay zero current compensation to a wealthy CEO is a further reason why Myerson would argue that hiring only wealthy CEOs would make possible the optimal compensation incentives.
If the bank is unsuccessful, the CEO should be “terminat[ed] … without pay.” While Myerson does not mention it, the logic of this analysis is that if the CEO received any compensation from the bank while he was employed it should be “clawed back” if the bank were unsuccessful due to the CEO’s looting, shirking, or other breaches of his fiduciary duties. His model, however, has no claw back provision.
Under Myerson’s September 2012 model the CEO of a bank reporting high profits would receive an extraordinary payment when he retired or resigned. His heirs would receive the payment if he died while the bank was reporting that it was success.
The Four Questions about Myerson’s models of executive compensation
- 1. Why are his models contradictory?
Myerson’s September 2012 remarks about and model of optimal executive compensation should have led him to ask a series of questions. First, why does his December 2012 model call for massive annual payments to bank CEOs? His December 2012 presentation does not call for bank CEOs to be “terminated … without pay” if the bank is unsuccessful (“w [wage] =0 for failure”). Instead, it calls for the bribe to be paid regardless of whether the bank is successful. His December 2012 model calls for a huge, annual “moral hazard” bribe to bank CEOs (roughly $200 million for my hypothetical mid-sized bank). Myerson’s model, if applied to our largest banks, would require the shareholders to pay an annual “moral hazard” bribe to the CEO of over $10 billion.
Myerson’s December 2012 model of how to control the CEO’s perverse incentives arising from moral hazard also calls for a very large “wage” to be paid to the bank’s CEO as soon as what he calls a “project” is purportedly “successful.” The CEO’s “wage” (if the project succeeds) would be roughly $800 million in the case of my hypothetical $20 billion mid-sized bank. (Again, this component of the CEO’s compensation would be over 50 times larger for our largest banks.)
The third form of bank payment to the CEO under Myerson’s December 2012 model is also paid when the “project” is purportedly “successful.” The third form is the return on the CEO’s minimum 40% capital investment in the bank. The CEO would get 40% of the bank’s total profits in addition to his “wage” and his “moral-hazard” bribe.
The first question, again, is why his two models, presented three months apart, are so contradictory. In his December presentation he claimed that only massive stock ownership by the bank’s CEO could successfully constrain their perverse incentives under moral hazard. His December model called for enormous payments to the CEO regardless of his success and staggeringly large payments to the CEO upon the success of “projects” he is managing. In his September presentation, Myerson writes that economists have known for nearly 40 years that the optimal means of constraining a bank CEO’s moral hazard is not to require massive stock ownership but rather to pay the CEO as close to zero a current salary as is feasible, to compensate him richly when he retires (or his heirs, should be die) if the bank is successful and to fire him and pay him nothing should the bank be unsuccessful.
- 2. Why does Myerson implicitly assume control fraud out of existence?
Myerson fails to ask the question of how one evaluates whether a bank is “successful” even though his September and December 2012 models fail if bank CEOs can falsely report that the bank is successful. Investors’ typical evaluation of a bank’s success is based on the bank’s reported income and capital – which is dependent on accounting that can make a deeply insolvent and unprofitable bank report high profitability.
Myerson assumes in both 2012 presentations that bank CEOs act perversely when they experience moral hazard in order to maximize their self-interest even if that is at the direct expense of the shareholders. Myerson assumes that CEOs will readily breach their fiduciary duties if doing so is in their self-interest. Indeed, he defines such abusive conduct by CEOs as “rational.”
Myerson’s contradictory September and December 2012 models have some features in common. They both assume, implicitly, that a bank’s “success” is an objective, readily observable fact. That assumption is false. If Myerson had made the assumption expressly he would have had to think about it and defend the reasonableness of the assumption. His implicit assumption removes any scholarly discipline. No one who understands the most basic facts about banks would ever support Myerson’s assumption. The naïve nature of the assumption is particularly remarkable given its timing – late 2012. We have just gone through a financial crisis in which scores of huge financial institutions reporting that they met all capital requirements (and had passed “stress” tests with flying colors only months or weeks earlier) were actually massively insolvent and collapsed in combined liquidity/asset quality death spirals. Many of these financial institutions (falsely) reported high profitability until shortly before these death spirals.
We know from the last three major financial crises that the fraudulent lenders driving the crises were able to use accounting as their “weapon of choice” to report record earnings until months before their collapses. The National Commission on Financial Institution Reform, Recovery and Enforcement (NCFIRRE) reported on the causes of the savings and loan debacle in 1993. It found that at “the typical large failure” “fraud” “was invariably present” and that accounting fraud was key means to loot the bank. George Akerlof and Paul Romer’s classic 1993 article (“Looting: The Economic Underworld of Bankruptcy for Profit”) confirmed NCFIRRE’s finding and showed how accounting was used to loot in the context of other financial frauds. The CEOs running the Enron-era “control frauds” also used accounting as their primary weapon. As Akerlof and Romer emphasized, control fraud is a “sure thing.” A bank that follows the accounting control fraud “recipe” is guaranteed to report record earnings in the near term. The recipe has four ingredients that will sound immediately familiar to anyone conversant with the current crisis.
- Grow like crazy by
- Making really crappy loans at a premium yield while
- Employing extreme leverage and
- Providing only grossly inadequate allowances for loan and lease losses (ALLL)
Myerson’s September and December models are contradictory, but they share the characteristic that the CEO will make vastly more money if the bank reports that it is “successful.” The CEO can ensure that the bank will report that it is successful. This is obviously possible under Myerson’s December model because it pays the CEO as soon as the bank reports it is “successful.” Under Myerson’s premises about how CEOs act their sole rational strategy would be to engage in control fraud. Indeed, because the bank CEOs and CFOs who followed the fraud recipe would report far greater profits than their honest counterparts the control frauds’ spectacular (reported) results would define “success.” Their honest counterparts would report sharply lower profits than the frauds even in good years and losses in difficult years. The frauds’ far superior (reported) results would greatly increase the risk of honest CEOs and CFOs being “terminat[ed] … without pay.” What I am describing is a “Gresham’s” dynamic (Akerlof 1970) in which bad ethics drives good ethics out of the markets. When cheaters prosper markets become perverse.
While the logic of Myerson’s model would require him to describe control fraud as the “dominant” “game” strategy, he implicitly assumes that fraud is impossible. This makes no sense because economists assume that CEOs and CFOs, particularly those engaged in accounting fraud, have superior information about the true condition of “their” bank compared to outsiders. Accounting fraud has been the dominant driver of recent financial crises, so Myerson’s implicit assumption that such frauds are impossible is bizarre.
The fraudulent CEO and CFO’s superior knowledge about the bank’s true financial condition would allow them to profit even in a system in which they would be “terminat[ed] … without pay” if the bank failed on their watch and even if they received minimal compensation during the time they were employed by the bank. Their asymmetrical information would allow them to retire or resign and take another position while the bank was reporting record profits. The CEO and the CFO would receive extraordinarily large pay upon their retirement or resignation. As I noted, Myerson’s model has no claw back provision. Even if it did have a claw back provision it would often be difficult or impossible for the bank to claw back the payments to the CEO and CFO. If a bank could claw back payments it made to a CEO under whose leadership the bank reported consistent, high profits simply because it began reporting losses years later it would be impossible to get anyone to take a job as CEO of a bank. If the bank could claw back the former CEO’s payments only if it could prove that the former CEO caused the losses then it would be very difficult for the bank to make its case against the former CEO.
Bank CEOs, moreover, can lead forms of control fraud other than accounting control fraud that are impossible to counter under Myerson’s compensation models. Recent examples include deliberately selling customers unsuitable products, overcharging customers, assisting tax fraud, manipulating and fixing LIBOR, and laundering money for drug cartels and terrorists. Because these other forms of control fraud, unlike accounting control fraud, produce real (albeit unlawful and unethical) profits, they will not cause the bank to fail even in the long run. The combination of high profits from fraud without losses from fraud means that these other forms of control fraud would become “dominant” strategies for “rational” bank CEOs under Myerson’s logic even if claw back provisions were universal and could be counted on to make it impossible for CEOs to profit from accounting control fraud and prevented it from creating Gresham’s dynamic. Claw back provisions are actually nearly non-existent and could rarely be successfully enforced against CEOs leading accounting control frauds even if they existed. The fact that accounting control fraud “recipe” is a “sure thing” for a lender’s CEO continues to make accounting the weapon of choice in finance and bad loans the “ammunition of choice.”
To this point I have been discussing the problems caused by bank CEOs leading accounting control frauds. The key thing to understand about accounting control fraud, in the context of determining how to avoid perverse incentives for senior bank officers, is that it produces guaranteed, record (albeit fictional) income in the near to mid-term followed by catastrophic losses.
The fact that accounting control fraud creates only fictional profits means that prior to modern executive compensation accounting control fraud did not create a Gresham’s dynamic. An honest bank could respond to a rival bank engaged in accounting control by gleefully sending its worst prospective borrowers to the rival. The honest bank would remain profitable. The dishonest bank following the accounting fraud recipe deliberately increases adverse selection and fails. Modern executive compensation has produced a Gresham’s dynamic in accounting control fraud that did not previously exist. That, along with the three “de’s” (deregulation, desupervision, and de facto decriminalization) are the principal reasons why we suffer more frequent and severe financial crises in the last few decades. The anti-regulatory policies and the pro-plutocrat policies of economists like Myerson that were purportedly designed to serve as “optimal” bribes to cause immoral CEOs “imitate” (Myerson’s word) honest CEOs by “aligning” the interests of the CEO and the shareholders have actually produced a vastly more criminogenic environment. The logic of Myerson’s model supports this conclusion, but his model implicitly excludes the very frauds his logic predicts will become endemic. As I will develop in a later article on Myerson’s model of government, his pro-plutocrat position also harms society by causing a great increase in inequality and our democracy and economy by fostering crony capitalism.
We should also step back for a moment and ask a question about a concept that theoclassical economists studiously ignore – power. The fundamental theoclassical rationale for modern executive compensation (one that Myerson endorses) has two parts. Because of the separation of ownership and control shareholders typically have only minimal power to control the CEO of a major corporation. The board of directors nominally selects the CEO, but the reality is that the CEO selects the board by choosing its members and voting proxies to elect the members the CEO hand-picked.
The second part of the rationale is that CEOs typically use their dominant power to maximize their self-interest even if doing so comes at the direct expense of the principals (shareholders). The interests of CEOs and shareholders typically conflict. This produces what economists call an “agency” problem. The “unfaithful agent” (CEO) harms the shareholders to make himself better off. Theoclassical economists, however, almost invariably make the implicit assumption that the CEO only harms the shareholders in two manners. CEOs “shirk” (they do not work as hard as they could), they like their perks too much to be fierce cost minimizers, and they avoid taking prudent risks. Theoclassical economists argued that U.S. firms were taking inadequate risks. They argued that this slowed U.S. economic growth. The economics literature explains why shareholders prefer the firms they invest in to take greater risks than do CEOs. The key points are that investors’ “expected value” increases with (prudent) risk in a diversified investment portfolio, it is far easier for shareholders to diversify their investment portfolio than for CEOs to diversify their investment in the firm, and old fashioned CEOs avoid risky investments because they are more likely to fail and spectacular failures pose the only serious risk that the CEO will lose his job.
The bottom line is that CEOs, and the firms they manage, are likely to be too risk averse and too fat and happy. Economists typically propose three fixes for these managerial maladies – modern executive compensation that purportedly “aligns” the interests of the CEO and shareholders through “performance pay” and a far greater willingness to fire the CEO and CFO if they fail to report great financial success, encouraging hostile corporate takeovers, and exceptional corporate leverage (debt).
- 3. Given Myerson’s assumptions, why would CEOs adopt optimal compensation designs?
One fatal problem with these “mechanism” “designs” that economists propose to fix the agency problem posed by the unfaithful, immoral CEO is that they ignore power. Their premises for why we needed to “design” a new executive compensation “mechanism” for CEOs (and other agents) were: (1) the CEOs had the dominant power over the officers, directors, and employees, (2) the CEOs acted to maximize their self-interest even when doing so violated their fiduciary duties and was unethical, and (3) the CEOs could frequently best maximize their self-interest by harming the shareholders.
The obvious question, which those supporting massive pay to plutocrats ignore, is why the CEOs would use their dominant corporate power to adopt the executive compensation “mechanisms” “design[ed]” by the scholars like Myerson. Why would a CEO want to “align” his interest with that of the shareholders? Why would a CEO want to adopt an executive compensation plan that would terminate him without pay – for decades of work – if the bank were unsuccessful? A “rational” CEO would like “his” bank to make him wealthy regardless of whether it succeeded or failed. Short-term reported bank income is far easier to inflate massively through accounting fraud than is long-term (10 years) bank income. Under Myerson’s premises a “rational” banker would structure his executive compensation to have a heavy component of pay for shorter-term reported income.
Recall that economists argue that shareholders prefer the firms they invest in to take considerable risk and that this was contrary to the interests of CEOs even before modern executive compensation. Under Myerson’s plan the shareholders purportedly give the CEO the proper incentive to cause the bank to take far greater risks by providing the CEO with an exceptionally large bonus at the end of his employment with the bank. This mechanism fails to solve the problem that economists identified. The CEO still has no adequate means to diversify his investment in the firm. Myerson’s December 2012 model makes the CEO’s diversification problem vastly more severe by claiming that the minimum safe level of shareholder ownership by a bank’s CEO is 40%. As I explained, Myerson’s minimum investment would require the CEO to own $400 million in bank stock for a bank with $1 billion in reported capital (roughly $20 billion in total assets) and over 50 times that capital investment (over $20 billion) by the CEO’s of the largest banks. Myerson’s December model would require bank CEOs to place all their eggs in a single basket – their wealth, income, and reputation would all depend on the success of the bank. Myerson’s model is designed to compound the bank CEO’s already perverse incentives by pushing the CEO to cause the bank to invest in far riskier investments and activities. All other factors being held constant, the probability of the bank failing increases as it makes far riskier investments.
If the bank fails the CEO loses massively under Myerson’s September and December 2012 models. Under the September model he is terminated and paid nothing for his years of work. Under the December model he receives a large bribe that Myerson (incorrectly) claims prevents the CEO from developing perverse incentives due to “moral hazard” but he loses his entire investment in the bank, an amount at least double the bribe. Theoclassical economists claim that the CEO suffers a terrible loss of reputation if the bank fails. This should be particularly true under Myerson’s December 2012 model because the CEO would be both the controlling officer and the controlling shareholder – 40% minimum stock ownership would cause the CEO to be the controlling shareholder of virtually every publicly traded U.S. bank. A CEO who is also the controlling owner of the bank should be held accountable for any failure and should suffer a great loss of reputation. (Theoclassical economists’ view of reputation is primitive and frequently false, but it is fair to hold them to internal consistency.) Taken together, these factors mean that both variants of Myerson’s compensation model cause bank CEOs to have intense, perverse incentives to engage in accounting control fraud in order to obtain a “sure thing” and report success. When the bank is failing, the CEO’s perverse incentives become even more compelling. Because Myerson’s model is premised on the unethical nature of bank CEOs (though Myerson describes their unethical conduct as “rational” rather than unethical), his logic requires that leading control frauds would be the dominant strategy of bank CEOs and requires the conclusion that his model exacerbates the bank CEOs’ already perverse incentives to lead control frauds. Myerson’s model works best as an explanation for why the designers of modern executive compensation have created a criminogenic “mechanism” that helps generate the epidemics of control fraud that drive our recurrent, intensifying financial crises and the rise of crony capitalism.
- 4. Why doesn’t Myerson note that actual compensation plans are the opposite of optimal?
Recall that Myerson began his explanation of his September 2012 model with the admission that:
“Since Becker and Stigler (1974) and Shapiro and Stiglitz (1984), it has been well understood in agency theory that dynamic moral-hazard problems with limited liability are efficiently solved by promising large end-of-career rewards for agents who maintain good performance records.”
Assume for the purpose of inquiry that Myerson were correct that we have known for nearly 40 years that moral-hazard problems posed by CEOs are severe but “efficiently solved” by compensation plans in which CEOs received virtually all of their compensation only when they ceased to be employed by the bank – and received no pay if the bank is unsuccessful. Only a theoclassical economist could fail to ask two related questions.
Have publicly-held corporations responded to the severe “moral-hazard problems” posed by their CEOs by adopting the executive compensation “mechanism” that “efficiently solved” those problems?
If not, why have they failed to act “rationally?”
Myerson knows that the answer to the first question is a resounding “No!” It is not a case of some firms have failed to adopt the “rational” “design” for executive compensation even after nearly 40 years of knowing how to “efficiently solve” a severe problem – no publicly held firm has adopted the only “rational” mechanism for CEO compensation.
Myerson would hate the answers to the second question. His model is premised on “rational” behavior and the (implicit) claim that the board of directors independently determines the “optimal” design of CEO compensation. His model fails if either of these premises fails – and his exposition falsifies both premises. His model (implicitly) supports the rival premise that at abusive firms where “moral-hazard problems” are most severe the CEOs use their corporate power to design and adopt executive compensation systems that maximize the interests of unethical CEOs at the expense of the shareholders, creditors, the general public, and customers. The actual compensation systems have not “efficiently solved” “moral-hazard problems” – they have made them so much worse that they are often criminogenic.