I'm giving a talk at the National Business Law Scholars Conference today, The CEO Pay “Problem”: It’s All Marjorie Margolies’ Fault, which reviews Michael Dorff's new book on executive compensation.
Here's the text:
The CEO Pay “Problem”: It’s All Marjorie Margolies’ Fault
It is rare that an issue as complex as executive compensation can be traced to a specific decision by a single person at a precise moment in time, but I nevertheless want to offer up a candidate.
The date was May 27, 1993. The place was the floor of the House of Representatives in Washington D.C. The person was Marjorie Margolies.
Newly elected President Bill Clinton’s first budget was in deep trouble. Every Republican in the House had voted against it, as had a number of Blue Dog Democrats (remember them?).
Congresswoman Margolies had opposed the bill, but changed her mind at the last minute after extended personal lobbying by President Clinton himself.
As Margolies walked down the aisle of the House to cast what became the deciding vote on the 1993 budget, numerous Republican Congressmen jeered "Goodbye Marjorie!,” waving handkerchiefs as though saying farewell to a ship sailing out to sea.
It is a moment that does not appear in Michael Dorff’s estimable new book Indispensable and Other Myths (2014), but probably should have.
Dorff does a through and thoroughly persuasive job of showing that the managerial power argument is plagued by internal inconsistencies and, at a great many points, is either unsupported by or even inconsistent with the empirical evidence. Culling through the evidence on both pay and the changing role and composition of boards of directors, he poses a profound question:
If managerial power is the most important cause of rising CEO pay, then why did CEO pay climb most steeply precisely at the same time that managerial power was experiencing its most precipitous decline?<![if !supportFootnotes]><![endif]>
The other principal contending theory—the efficient contracting model—contends that CEO compensation is the product of arms’-length bargaining between managers attempting to get the best possible deal for themselves and boards seeking to get the best possible deal for shareholders. This bargaining process takes place in the shadow of important market constraints, which tend to produce optimal outcomes.
In turn, the undeniably dramatic increase in CEO pay over the last several decades was purportedly driven by systemic changes in metrics such as company size and stock market performance, improved pay for performance techniques, greater reliance on tournament competitions, and so on.
Here, again, Dorff does valuable work in fairly but critically evaluating the arguments. With one exception (about which more later), I found his critique of the justifications offered by proponents of arms-length bargaining persuasive. Like the managerial power story, the optimal contracting story has a lot of holes.
The biggest of those holes is that both models rest on a shared but flawed assumption; namely, that performance-based pay ineluctably leads to superior performance. In fact, however, as Dorff rightly concludes from a lengthy analysis of both empirical evidence and laboratory studies of decision making, “performance pay very likely … does not result in better performance.”<![if !supportFootnotes]><![endif]>
As I likewise observed in a Texas Law Review article reviewing Bebchuk & Fried’s book, there is relatively little evidence that CEOs are motivated by pay, which suggests the possibility that CEOs are motivated principally by other concerns such as ego, reputation, and social effort norms. Put another way, the latter considerations may be the principal mechanisms by which the principal-agent problem is resolved. If so, evidence advanced by either side to show that incentive compensation either does or does not improve performance tells us nothing other than that researchers have mined the data to find a spurious correlation. Worse yet, at least from the perspective of those who wish to use compensation to address the principal-agent problem, the goal of linking pay and performance inevitably will prove an exercise in futility.
Having made that sweeping generalization, it will be helpful at this point to refine matters slightly by disaggregating the problem of incentivizing CEOs and other top managers into three sub-categories. First, it may be necessary to give CEOs an incentive to work hard rather than slacking. Second, it may be necessary to give CEOs an incentive to make shareholder wealth maximizing choices when selecting projects and business ventures for the firm. Finally, it may be necessary to give CEOs an incentive not to engage in self-dealing, whether such self-dealing takes the form of excessive perquisites, conflicted interest transactions with the firm, or outright theft.
It seems unlikely that performance-based compensation schemes deter an executive bent on self-dealing. Because the impact of a self-dealing transaction is spread across all shareholders, the negative effect on the executive’s equity-based compensation and existing equity holdings likely will be a small fraction of the positive benefit to be obtained from the self-dealing transaction. Accordingly, rather than counting on compensation or other market-based forces to constrain self-dealing, corporate law seeks to deter it through harsh penalties and strong norm-setting rhetoric.
It also seems unlikely that performance-based compensation schemes do much to affect the degree of managerial slack. In the first place, slackers will rarely climb to the top of the greased pole. By the time a CEO prevails in the repeated promotion tournaments that must be won in order to reach the top of the corporate hierarchy, the CEO likely will have fully internalized a slate of social effort norms. Finally, once the CEO reaches the top of the corporate ladder, he remains subject to monitoring and competition by his board, his immediate subordinates, the media, and activist shareholders. As such, we would expect the CEO’s performance to be more sensitive to self-esteem and reputation than to variable incentive compensation. Accordingly, monetary compensation probably serves less as a direct incentive than as a status counter for top CEOs.
In contrast, performance-based compensation plausibly should affect CEO incentives when making capital budgeting decisions. Suppose the CEO is presented with two mutually exclusive projects requiring roughly the same capital investment, but with significantly different net present values (NPVs). Because risk and return are positively correlated, the higher NPV project typically will be the more risky one. All else being equal, that is the project stockholders will prefer. Because managers are inherently more risk averse than diversified shareholders, however, managers may prefer the less risky project. Here then is where performance-based compensation theoretically comes into play, by giving managers an incentive to choose the more risky project. On the other hand, however, if CEO behavior is motivated mainly by the non-pecuniary incentives discussed earlier, performance-based compensation will not be effective even in this context.
To be sure, it seems implausible that there is no motivational link between pay and performance. In many settings where incentive-based compensation seems appropriate, we observe a mix of fixed and variable pay. In major motion pictures, for example, we observe leading actors being paid a fixed salary plus a percentage of the gate. Star athlete contracts also often include performance-based bonuses. This pattern suggests that pay and performance are linked to some extent, which comports with common sense. Dorff’s analysis, however, confirms at length my intuition that the link between pay and performance is far weaker than is commonly supposed.
And so we come back to that critical moment when Marjorie Margolies cast the deciding vote in favor of the Omnibus Budget Reconciliation Act of 1993. Among other things, that law amended § 162(m) of the Internal Revenue Code. The amendment “limits public companies’ ability to deduct the compensation for the CEO and certain other senior executives to the extent a covered executive’s pay exceeds $1 million per year. The first million paid to each executive may be deducted, but every dollar after that must be counted in the company’s income (even though the company is really paying that money as salary to the executive).”<![if !supportFootnotes]><![endif]>
As Dorff points out, although CEO pay had been growing fast relative to other metrics during the 1980s, it was in the 1990s that CEO pay really started to explode as a percentage of corporate profits and total wages.<![if !supportFootnotes]><![endif]> In large measure, this occurred because of a huge shift towards options and other forms of incentive pay that were tax favored under the 1993 amendments.
Despite this temporal correlation, Dorff downplays the role that regulation in general and § 162(m) play in our story.
In contrast, I think it was a pivotal moment. As former SEC Chairman Christopher Cox noted, it deserves a “place in the museum of unintended consequences.”
It’s not just that § 162(m) created tax incentives for using performance pay schemes. I believe § 162(m) is a prime example of how law, by expressing social values, can change social norms or even create new norms.
Dorff recognizes that norms can be powerful, citing the oft-remarked example that people “leave tips in restaurants even when traveling in a strange city they never intend to visit again, and even when eating alone,” although doing so makes no sense from a rational actor model.<![if !supportFootnotes]><![endif]>
What I think happened in the 1970s-1990s was the emergence of a changed boardroom norm. The change began in the 1970s when financial economists (in a development Dorff ably traces) began pushing the idea of using pay for performance. Boards then gradually began responding by increasing the amount of incentive pay they used.
What I think the 1993 budget did was to validate that emergent norm. As Richard McAdams has argued, law does not simply affect social norms by creating incentives, but also by “signaling the underlying attitudes of a community or society.” He argues that “legislative outcomes are positively correlated with popular attitudes and therefore provide a signal of those attitudes. Independent of the sanction, the legislative signal [causes] people to update their prior beliefs about what others approve and disapprove.” In turn, because individuals are “motivated to gain approval and avoid disapproval, the information signaled by legislation affects their behavior.”<![if !supportFootnotes]><![endif]>
If that’s right, we may have an explanation for what Dorff claims is a puzzle about § 162(m). He asks: “Why … when increasing options led to rapid growth in pay, did boards refuse to reverse course? Why did they go along with spiraling pay? They certainly had the tools to cut back, even though encumbered by admittedly clumsy federal laws such as §162(m).”<![if !supportFootnotes]><![endif]>
The answer lies in the same behavioral economics to which Dorff pays so much attention elsewhere. The 1990s saw a paradigm shift that created a new social norm validated by Congressional imprimatur and backstopped by tax sanctions and incentives.
CEO pay thus is an artifact of neither managerial power nor arms-length bargaining. Instead, its current structure and size is a product of a paradigm shift that has become deeply embedded in the social norms that regulate boardroom decision making. Which explains why boards refused to “reverse course.”
Regulation played a critical—albeit unintended—role in this process by validating the emergent norm and giving it a seal of approval not just from leading theorists but also the Congress and President of the United States.
Does this suggest that regulation is also the solution? My co-panelist, Professor Kevin Murphy, has done great work in showing how regulation pervasively distorts executive compensation. Our author, Professor Dorff, acknowledges Murphy’s concerns, concluding that “government should … tread carefully in this area as new legal rules are likely to have unforeseen (and undesirable) consequences.”<![if !supportFootnotes]><![endif]>
I concur. Dorff persuasively argues that directors are subject to all sorts of systematic behavioral biases and errors. But so are regulators. As Jennifer Arlen observes:
The very power of the behavioralist critique—that even educated people exhibit certain biases—… undercuts efforts to redress such biases [because regulators will have their own biases]. In addition, the decisions of government actors also may be adversely influenced by political concerns—specifically interest group politics. Thus interventions to “cure” bias-induced inefficiency may ultimately produce outcomes that are worse than the problem itself.<![if !supportFootnotes]><![endif]>
What then should we do? I think the solution is a massive deregulation of CEO pay. Eliminate all tax rules that penalize some pay formats and all rules that subsidize others. Eliminate say on pay, the Section 16(b) exemptions for incentive pay, and so on. Leave in place only simple, clear disclosure rules (with no nonsense about CEO-worker pay ratios or what have you), and let sunlight be your disinfectant and electric light your policeman.
<![if !supportFootnotes]><![endif]> Lucian Bebchuk Jesse Fried, Pay Without Performance: The Unfulfilled Promise of Executive Compensation 6 (2004).
<![if !supportFootnotes]><![endif]> Id. at 5.
<![if !supportFootnotes]><![endif]> Dorff at 120.
<![if !supportFootnotes]><![endif]> Dorff at 146.
<![if !supportFootnotes]><![endif]> Dorff at 82.
<![if !supportFootnotes]><![endif]> Dorff at 19-20.
<![if !supportFootnotes]><![endif]> Dorff at 55.
<![if !supportFootnotes]><![endif]> Richard H. McAdams, An Attitudinal Theory of Expressive Law, 79 Or. L. Rev. 1 (2001).
<![if !supportFootnotes]><![endif]> Dorff at 93.
<![if !supportFootnotes]><![endif]> Dorff at 231.
<![if !supportFootnotes]><![endif]> Jennifer Arlen, Comment: The Future of Behavioral Economic Analysis of Law, 51 Vand. L. Rev. 1765 (1998).