The Culture of CEO Pay

Response - Online Edition - Volume 96

CEO compensation is a perennial issue, and understandably so, particularly in light of how much CEOs are paid in both absolute and relative terms and given the corporate scandals that make front-page news.[1] How do we explain CEO pay? There are essentially two different narratives. Under the managerial power theory, the CEO has enormous influence and sway over board members who, thanks to legal innovations like the poison pill and the staggered board, are sufficiently insulated from market forces to do the CEO’s bidding, including awarding above-market pay.[2] By contrast, the managerial talent theory, also known as optimal contracting theory, maintains that boards act as independent, neutral pay-setters and that CEO pay reflects compensation for unique managerial talent.[3]

The debate is important because the rightness of any policy recommendation depends on which theory is correct. If the managerial power theory is correct, then we should probably look for ways to hold boards accountable, including by empowering shareholders to apply various carrots and sticks.[4] If, on the other hand, the managerial talent theory is correct, then the system is working the way that it should, and shareholder empowerment would only disrupt the apple cart.[5]

Enter CEO Pay Redux,[6] which is the latest article to wade into this conversation. It is a provocative, interesting piece that takes up an important question—can CEO pay be explained as a reward for talent or the result of CEOs wielding enormous influence over their boards? It seeks to shed light on this question in two different ways. First, it argues that legal scholars have relied on an over-abstraction of reality in describing how CEO pay works and that this over-abstraction (what they refer to as the “static” model) has led to erroneous conclusions.[7] Second, drawing on the authors’ preferred “dynamic” model, it conducts an empirical test in an effort to identify which of these two theories is better supported by the evidence.[8] Despite some methodological and statistical questions and quibbles, I think the article has important points to make on both scores. After discussing these points, I briefly suggest a possible new direction that the legal literature on executive compensation might take in building off of this interesting work.

I. The Dynamic Model

According to the authors, there are two issues that a static model of CEO pay overlooks and a dynamic model picks up. First, there is the reality that if the CEO does a good job, his reservation value increases over time, which creates a retention problem ignored by the static model.[9] Second, a dynamic model illuminates the possibility that CEOs might be too focused on the short-term.[10] The authors point out that compensation solutions that are difficult to explain in the static model as anything other than evidence of managerial power actually make sense in a dynamic model in light of these two concerns.[11] In particular, fixed compensation, long a bugaboo of managerial power theorists who characterize it as the quintessential example of pay without performance, actually might help to mitigate both of these problems in a dynamic model.[12] If the CEO knows that he is going to get a bonus if he does a good job, the result is increasing total compensation commensurate with his increasing reservation value.[13] Additionally, this structure helps address the question of short-termism on the part of the CEO by focusing his attention on long-term compensation.[14] Similar arguments can be made in favor of restricted stock, which is also frowned upon by managerial power theorists.[15]

Of course, the response on the part of the managerial power theorists might be: “Thanks for the clarification, but shareholders still should be empowered.” After all, they might point out, this type of “dynamic” compensation contract only makes sense if the CEO cannot only be rewarded for effort but also be removed for the lack thereof. And, according to the managerial power theory, there are not sufficiently robust mechanisms for removal.[16] The authors’ response is that in a dynamic model, the need for long-term value creation is emphasized, and once you do that shareholder empowerment becomes problematic to the extent that it serves as a vehicle for imposing certain shareholders’ short-term preferences on firms.[17] The authors are of course correct about this, but that does not mean that shareholder empowerment is unnecessary. It just means that it needs to be designed with the concern over shareholder myopia in mind. Nevertheless, the points the authors make about the shortcomings of the model that tends to dominate the legal literature on executive compensation strike me as important and worth paying attention to.

II. An Empirical Investigation of Theories of CEO Pay

Having clarified their model, the authors next turn to an empirical investigation of CEO pay.[18] Their starting point is that there already exists some evidence suggesting that the managerial power theory is not the whole story.[19] For example, the authors point to findings that staggered board provisions are positively correlated with firm value, and they ask rhetorically how this could be true in a world of managerial power.[20] Here is one possibility: staggered boards might help firms get better deals in takeovers by taking the hostile alternative off the table. At the same time, staggered boards might make it easier for CEOs to extract agency rents. If the value-increasing effects of staggered boards outweigh the value-decreasing ones, then one could expect a positive overall correlation between staggered boards and firm value. But that does not mean that they do not also allow for rent extraction in the form of higher CEO pay.

Nevertheless, the authors are surely correct that the managerial power theory is not obviously right. After all, there are market forces at work that surely cabin managerial power. These forces are blunted to a certain extent by antitakeover devices like staggered boards and poison pills. But there are still markets, and there are fiduciary obligations that, while weak, effectively require CEOs to pursue shareholder value even if the law itself is not always willing to enforce that norm.[21]

If it is not obvious that one of these two theories is right as an a priori matter, then the question is: how should we go about testing them? Let us first talk about the ideal way of answering that question and then consider how the authors go about it.

A. The Ideal Regression

In a perfect world, a test of the two theories would consist of two steps. First, you would consider the determinants of CEO pay. Your dependent variable would be CEO pay, and your independent variables of interest would be measures of managerial talent and managerial power. And then, of course, there would be a number of control variables.

What do we mean exactly by managerial power? There are two aspects: what I will call the board independence component and the market insulation component. The board independence component is the degree to which the board is truly capable of making objective decisions free of the CEO’s influence. Independence in this context might mean one of two things. There is, on the one hand, formal independence: at a high level of generality, are there any relationships, for example familial, that might prevent an independent judgment?[22] But then there is a more searching form of independence, which goes beyond formal relationships and considers cultural ecosystems.[23] While it will be difficult to come up with proxies for either type of independence, the second presents more of a challenge.

The second aspect of managerial power, what I am calling the market insulation component, is the degree to which the board is insulated from market forces. Even if the CEO can exert significant influence over the board, that might not translate into higher CEO pay if the board is worried that such over payment will attract market scrutiny from shareholders, consumers, or, most likely, potential acquirers eager to realize efficiencies. Here, unlike with the independence variable, there are several proxies in use in the literature to measure board entrenchment—in other words the board’s degree of insulation from market forces.[24]

That is managerial power. What about managerial talent? How could one measure that? Presumably talent is more strongly rewarded the more competitive the market tends to be. So, one could look at proxies for competitiveness, including industry concentration indices and levels of M&A activity.

All of that constitutes only the first step of the empirical test. The second step would be to then test the effect that CEO pay has on firm value. After all, who cares if CEO pay is determined by managerial power if it ultimately has no effect on firm value?

It is at this point that I think there is an important clarification to make, and one that is not often emphasized in the literature. CEO pay is presumably not as relevant to all firms for the simple reason that good CEOs are not equally valuable at all firms. At firms that enjoy durable competitive advantages, it is probably not as valuable. At firms that do not, where execution matters a lot, it is probably more valuable. There is the old Peter Lynch quip about wanting to find a business that could be run by an idiot because at some point it will be.[25] Mr. Lynch is essentially saying that there are certain businesses that are so insulated from competition that CEO execution is much less important. So, the ideal regression would probably control for different types of competitive advantage. How might one do this? Return on capital—earnings divided by book value—is a traditional metric for measuring a firm’s competitive advantage.[26] Firms that have high returns on capital over long periods of time are usually thought to benefit from a durable competitive advantage.[27] But any company could have a high, one-period return on capital. To really control for this, you would need to average the return on capital for some extended period of time, maybe 10 years.[28]

B. The Regression in Reality

The authors run these types of regressions in their paper, and they find some evidence for the managerial talent theory. However, as is often the case for anything short of real-world experiments, there are some limitations to their study.

As a preliminary matter, the authors do not really test the board independence component of managerial power. They only test the market insulation component.[29] This is a problem for their project because it does not really allow one to measure the effect of managerial power on CEO pay. For example, as discussed in more depth below, the authors find that the presence of various antitakeover devices, like poison pills and staggered board provisions, is not correlated with CEO pay, and they view this result as inconsistent with the managerial power theory.[30] But even if these results are reliable (and they are not without issues), does this conclusion really follow? After all, it could be the case that CEOs are paid more than they would be in an arm’s-length transaction and yet still not so much that it would justify a hostile bid, in which case, antitakeover devices would not affect CEO pay. For this reason, I am not particularly convinced by the authors’ interpretation of their finding that proxies for competition (in product markets, labor markets, and the market for corporate control) are positively correlated with CEO pay. It is certainly possible that this result suggests that what is really being rewarded is talent. But because of the lack of strong proxies for board independence, this result is not inconsistent with some (weak) version of the managerial power theory.

I alluded before to the possibility that these results from the regression involving the governance variables (like poison pills and staggered boards) on CEO pay are not without issues. There are three in particular. First, the authors find that the presence of a poison pill has no effect on CEO pay. This is actually not at all surprising, given the fact that any board can adopt a poison pill literally in a matter of hours, and therefore it is irrelevant for governance purposes whether a board has actually done so.[31] The authors acknowledge this fact but nevertheless seem to draw inferences regarding the lack of correlation.[32] These inferences are simply not justified.

Second, the authors find that there is no correlation between CEO pay and the presence of a staggered board—a governance device commonly thought to insulate the board from the market for corporate control,[33] arguably the most robust of available market forces.[34] The problem though is that the authors do not distinguish among different types of staggered board provisions, and not all provisions are created equal. It has been shown that the only staggered boards that are really effective against hostile takeovers are those that (1) are located in the corporate charter and therefore cannot be eliminated without the board itself calling a meeting, and (2) where the effects of the staggered board cannot be circumvented by simply increasing the size of the board or removing directors and filling the resulting vacancies.[35] Staggered board provisions that satisfy those two requirements are significantly (statistically and practically speaking) more successful in warding off hostile bidders than staggered board provisions that do not meet those requirements.[36] In other words, not all staggered board provisions effectively insulate boards. Therefore, lumping all staggered board types together, as the authors do, does not allow us to really test the effect of staggered boards on CEO pay. Consequently, we also cannot view this result as inconsistent with the managerial power theory.

Third, the authors find that CEO pay is increasing in the percentage of outstanding shares owned by institutional investors.[37] They interpret this finding as inconsistent with the managerial power theory, and there is a clear logic to this conclusion.[38] If CEO pay is set by managerial power rather than talent, the reasoning goes, then one would expect challenges to that power—for example, by sophisticated shareholders with concentrated ownership—to decrease, not increase, CEO pay. However, there is an interpretive problem here that results from the failure to properly test for board independence, as discussed above. There is a weak version of the managerial power theory, alluded to above, that would allow for market forces, including shareholder activism, to establish the range of possible compensation contracts.[39] Within that range, managerial power might then allow the CEO to get more money than he otherwise would have in an arm’s-length negotiation. For this reason, although I think that this result about the effect of institutional investors on CEO pay is very provocative and interesting, I am not sure that it is as damaging to the managerial power theory as the authors suggest.

These are the issues raised by the regressions testing of the effect of various governance variables on CEO pay. But let us assume for a moment that these results are fairly reliable and, as the authors insist, roughly inconsistent with the managerial power theory. What about the second step, the regression of firm value on CEO pay? The authors find a positive correlation between CEO pay and firm value.[40] Thus, one could, in an ideal world, use this result to complete the story being told: “CEO pay is largely determined by managerial talent, not managerial power, and it is important to firm value.”

However, the problem with this conclusion is that there are statistical issues with the valuation regression. CEO pay drives firm value, but firm value also drives CEO pay. The authors try to cure this endogeneity problem with an interesting event study.[41] The trick is to find some event that affects CEO pay but does not affect firm value, which then allows one to isolate the effect that CEO pay has on firm value.[42] The authors seize on a really clever idea: a 2004 accounting rule change that required companies, for the first time, to expense stock options when they are granted.[43] This caused stock options to become less attractive from an accounting perspective, leading to their diminished use, but it should not have affected the cash flows of the firm.[44] After all, stock options represent a cash outflow regardless of how they are accounted for, and the market should realize this. For this reason, the rule change should not affect firm value, even though it likely affects CEO pay.[45] Thus, the accounting change is an exogenous shock that affects CEO pay and allows one to isolate the effect that those changes have on firm value.[46] Here is how the authors describe how the event study should be interpreted: “[T]he association between the changes in compensation that were caused by [the accounting rule change] and the changes in firm value that took place after the rule’s introduction can accordingly be interpreted as providing plausible causal evidence for how changes in compensation affect changes in firm value.”[47]

What the authors find is that changes in compensation levels due to the introduction of the new accounting rule have no effect on firm value. One might conclude from this, consistent with the quoted language above, that the practical significance of CEO pay is overblown and that on average it does not affect firm value. However, the authors have a different interpretation: “that the overall level of compensation before the introduction of [the new accounting rule] was set ‘efficiently’ . . . .”[48] I must admit to being somewhat puzzled by this interpretation. If the coefficient on the variable in question—a variable measuring compensation levels at those firms that had been users of stock options prior to the accounting rule change[49]—were statistically significant and practically insignificant, then I can see how one could draw a conclusion about the efficiency of pay practices prior to the rule change. For example, if that coefficient had been statistically significant and negative, then presumably that would mean that CEOs were being paid too much prior to the rule change and that firms could have increased firm value by paying them less. In fact, the authors offer precisely this interpretation with respect to a slightly different variable, which measures the use of equity-based compensation by companies that used stock options prior to the accounting rule change.[50] Accordingly, the authors understand the negative, statistically significant coefficient attached to this variable to mean that “the more firms affected by the introduction of [the new accounting rule] reduced the use of equity-based pay in the two years after the rule change, the greater the increase in firm value.”[51] But that interpretation survives scrutiny because the coefficient is statistically significant.[52] That is not the case with respect to the earlier interpretation regarding the level of CEO pay more generally.

To be clear, however, this finding about the structure of CEO pay prior to the accounting change is a really interesting result, and it suggests precisely what the authors say it does: “that option-based pay seems to have been overused in recent years.”[53] But notice that all we are talking about here is the structure of CEO pay. As I interpret the data, we cannot make similar observations about efficient compensation with respect to the level of CEO pay. The authors want to say that the reason for this market distortion is an over-reliance on market forces, and they want to do this because it allows them to then criticize the managerial power theorists’ prescription of shareholder empowerment.[54] But while I am frankly not sure of the wisdom of the shareholder empowerment prescription, I am also not sure that that this data supports such skepticism. An alternative reading of the authors’ event study is that it shows that increased exposure to market forces is going to cause boards to focus on whatever Wall Street cares about, which, illogical though it may seem, is not always just cash flows but accounting earnings as well. So if you have ill-conceived accounting rules that are going to artificially boost accounting earnings, like the accounting rule in question prior to the change, then yes, increased exposure to market forces will make that ill-conceived accounting rule more relevant at those firms that are so exposed. But, that does not mean that we should not expose boards to greater market forces. It just means that we should not adopt ill-conceived accounting rules.

III. Going Forward

CEO Pay Redux is an insightful, interesting addition to the literature on executive compensation. The question it raises is this: What is the next logical step in this literature? I have no doubt that there are many additional empirical pieces that could build fruitfully off of this one, and I hope that that happens. But this article has left me thinking of a slightly different research agenda, one having to do with in-depth studies of the “culture” of firms. Companies with similar governance features are nevertheless different in a number of unobservable ways that generally fall under the umbrella of “culture.” For example, on the surface, companies like Berkshire Hathaway and Markel look very similar to companies like Sears and Aeropostale.[55] They all have non-staggered, independent boards.[56] They all operate in competitive product markets (insurance for the first two and retail for the latter two).[57] Their executive compensation practices are generally similar, with a mix of options and restricted stock.[58] And yet, in other ways these companies could not be more different. The latter two are not particularly well known as bastions of shareholder-friendliness, whereas the first two are.

But one only gets a sense of these cultural differences from a close institutional analysis. Berkshire and Markel both issue SEC filings that are similar to each other but that are radically different from those of many other companies. These disclosures are remarkably clear. They sound like they are talking to the shareholders themselves, because they are.[59] They often include a bracingly frank discussion of how management thinks about valuation of the firm.[60] They talk about the business, pointing out how they think about it, what really matters, why it is good, and where the weaknesses are.[61] These cultural factors can be seen not just in SEC filings. For example, Markel has a long internal document that defines the “Markel Style.”[62] The company is very discriminating about who they hire, saying things like “not everyone is cut out for Markel.”[63] They say the same things about shareholders.[64] They are trying to cultivate a particular culture.[65]

I think that legal scholars would do well to explore these cultural questions. In empirical research, when we miss the culture, the results often end up being unreliable because cultural factors are omitted, and yet they affect both firm value and CEO pay, for example.[66] Moreover, I think that law professors are well-suited to take up this task. There is a rich literature in the law on the diffusion and perpetuation of norms and a tradition of close institutional analyses.[67] I think such studies would be welcome compliments to the fine empirical work done in papers like this one.

  1. .The authors summarize these factors in the opening paragraphs of the article. See K.J. Martijn Cremers, Saura Masconale, & Simone M. Sepe, CEO Pay Redux, 96 Texas L. Rev. 205 (2017).
  2. .Lucian Bebchuk & Jess Fried, Pay without Performance: The Unfulfilled Promise of Executive Compensation 61–64 (2004).
  3. .See, e.g., Steven M. Bainbridge, Executive Compensation: Who Decides?, 82 Texas L. Rev. 1615, 1629–32 (2005) (describing the optimal contracting theory).
  4. .See, e.g., Lucian Arye Bebchuk, The Case for Increasing Shareholder Power, 118 Harv. L. Rev. 836 (2005) (describing the need for greater shareholder control over the board).
  5. .See, e.g., Bainbridge, supra note 3, at 1643–61 (outlining several reasons that increased shareholder power over boards would undermine corporate functions).
  6. .Cremers et al., supra note 1.
  7. .See id. at 209.
  8. .See id.
  9. .Id. at 216, 229, 231.
  10. .Id. at 211, 230.
  11. .Id. at 231.
  12. .Id.
  13. .Id. at 232.
  14. .Id.
  15. .Id.
  16. .Id. at 231.
  17. .Id. at 243, 256, 266.
  18. .Id. at 236.
  19. .Id. at 246.
  20. .Id. at 213, 227, 246.
  21. .See, e.g., Stephen M. Bainbridge, Unocal at 20: Director Primacy in Corporate Takeovers, 31 Del. J. Corp. L. 769, 778 (2006) (explaining that failures of the duty to maximize shareholder wealth are often shielded by the business judgment rule).
  22. .See, e.g., Auerbach v. Bennett, 393 N.E.2d 994, 1001 (1979) (applying a formalistic test for independence based on board members’ direct connections with the corporation).
  23. .See In re Oracle Corp. Derivative Litig., 824 A.2d 917, 928, 930 (Del. Ch. 2003) (examining complex ties between board members, the corporation, and the university for which the board members were tenured facility with administrative roles to determine whether the board members were truly independent).
  24. .See, e.g., Cremers et al., supra note 1, at 126–27; Marcel Kahn & Edward B. Rock, Corporate Constitutionalism: Antitakeover Charter Provisions as Precommitment, 152 U. Pa. L. Rev. 473 (2003).
  25. .Peter Lynch, One Up On Wall Street 130 (2012) (“When somebody says, ‘Any idiot could run this joint,’ that’s a plus as far as I’m concerned, because sooner or later any idiot probably is going to be running it.”).
  26. .See, e.g., Bruce Greenwald et al., Value Investment: From Graham to Buffett and Beyond 3–16 (2004).
  27. .See, e.g., id.
  28. .This represents a potential weakness in the authors’ control variables. In their event study, discussed in greater detail below, the authors include a profitability variable in their controls, but the profitability variable is just a one-year lagged variable. To really control for durable competitive advantages, you probably need the average over the prior ten years or something similar.
  29. .See Cremers, et al., supra note 1, at 263–50.
  30. .See id. at 244, 246–47.
  31. .See, e.g., John C. Coates IV, Takeover Defenses in the Shadow of the Pill: A Critique of the Scientific Evidence, 79 Texas L. Rev. 271, 326 (2000).
  32. .See Cremers et al., supra note 6, at 246 (“Taken as a whole, the results on Staggered Board and Poison Pill . . . seem incompatible with managerial power theory, as there is no evidence that having these defenses in place results in higher levels of executive compensation or changes in pay structure.”).
  33. .See Lucian Arye Bebchuk et al., The Powerful Antitakeover Force of Staggered Boards: Theory, Evidence, and Policy, 54 Stan. L. Rev. 887 (2002).
  34. .See, e.g., Frank H. Easterbrook & Daniel R. Fischel, Corporate Control Transactions, 91 Yale L. J. 698, 711–14 (1982).
  35. .See Bebchuk et al., supra note 33, at 894.
  36. .See id. at 930, 937 (finding that companies with “effective” staggered boards—in other words, those that cannot be easily circumvented—are 26% more likely to remain independent in the face of a hostile bid than companies with other types of staggered board provisions).
  37. .See Cremers, et al., supra note 1, at 212.
  38. .See id. at 213.
  39. .See id. at 228.
  40. .See id. at 210.
  41. .See id. at 251.
  42. .Id. at 259.
  43. .Id. at 258.
  44. .Id.
  45. .Id. at 264.
  46. .Id.
  47. .See id. at 259.
  48. .Id. at 264.
  49. .Id.
  50. .Id. at 264.
  51. .Id.
  52. .Id.
  53. .Id.
  54. .Id. at 234–35.
  55. .Charles Sizemore, Is Sears the Next Berkshire Hathaway?, Forbes, [].
  56. . Berkshire Hathaway, Proxy Statement (Schedule 14A) 2–4 (Mar. 17, 2017); Markel, Proxy Statement (Schedule 14A) 2, 9 (March 23, 2017); Sears, Proxy Statement (Schedule 14A) 9–10 (Mar. 31, 2017); Aeropostale, Proxy Statement (Schedule 14A) 10 (May 8, 2015).
  57. . See, Aeropostale, Proxy Statement (Schedule 14A) (May 8, 2015) (“As many of you know, Aeropostale Inc. has continued to operate in a difficult teen retail environment for the last several seasons and we have generated a significant amount of change over the last year in response.”); Erik Holmes, What is Berkshire Hathaway?, Wall Street Journal, []; Insurance and reinsurance, Markel, []; Is Sears Dying a Slow Death?, Fox News, [].
  58. . Markel, Proxy Statement (Schedule 14A) 18 (March 23, 2017); Sears, Proxy Statement (Schedule 14A) 20–26 (Mar. 31, 2017); Aeropostale, Proxy Statement (Schedule 14A) 23 (May 8, 2015).
  59. .See Berkshire Hathaway, Letter to Shareholders, Berkshire Hathaway Inc., [] (“To the Shareholders of Berkshire Hathaway Inc.:”); Markel Corporation, Letter to Shareholders, Markel,
    pdf?la=en&hash=708EAE40927243FB60FDC30FAF23C292BF0CFC06 [] (“To Our Business Partners”).
  60. .See Berkshire Hathaway, Letter to Shareholders, Berkshire Hathaway Inc., [] (“Nevertheless, I very much like our own prospects. Berkshire’s unrivaled financial strength allows us far more flexibility in investing than that generally available to P/C companies. The many alternatives available to us are always an advantage; occasionally, they offer us major opportunities. When others are constrained, our choices expand.”).
  61. .See, e.g., Markel Corporation, Letter to Shareholders, Markel, http://www.markelcorp.
    708EAE40927243FB60FDC30FAF23C292BF0CFC06 [] (“[T]he results from Markel Ventures are both better and worse than what the raw numbers present.”).
  62. .See The Markel Style, Markel Corporation, [] (“The Markel way is to seek to be a market leader in each of our pursuits.”). This is only a summary of the internal document, which is not available to the public.
  63. .See Markel Corporation Annual Report and Form 10-K (2012), [] (“We are not for everyone.”).
  64. .See Markel 2015 Shareholder Meeting Notes, ValueWalk
    2015/05/markel-2015-shareholder-meeting-notes/ [] (quoting the Markel Co-CEO, Tom Gayner, as saying that not all investors will understand what Markel is trying to do). 
  65. . See The Markel Style, Markel Corporation, [] (explaining the elements that Markel wishes to cultivate in order to create a particular culture within their company).
  66. .See generally, Yair Listokin, Interpreting Empirical Estimates of the Effect of Corporate Governance, 10 Am. L. Econ. Rev. 90 (2008).
  67. .See, e.g., Tiffani N. Darden, The Law Firm Caste System: Constructing a Bridge Between Workplace Equity Theory & the Institutional Analyses of Bias in Corporate Law Firms, 30 Berkeley J. Emp. & Lab. L. 85, 85 (explaining that the article’s purpose is to study the intersection between the “workplace equity theory and the institutional analyses of law firm diversity”).