Nonprofit Business Advisor, Strategies to Survive and Grow in Tough Times

Separating Chair and CEO Roles: A Flawed Debate

By John Carver December 11, 2013

RESISTANCE to separating the chair and CEO roles may be enjoying a new surge in the United States. Two recent articles illustrate the case: Jeffrey A. Sonnenfeld, one of the most frequently quoted corporate governance voices in the American press, wrote “The Jamie Dimon Witch Hunt” as an op-ed piece in the New York Times, May 8, 2013. (Dimon was being attacked for steadfastly holding on to the combined position at JPMorgan Chase—hence Sonnenfeld’s title.1) Joann S. Lublin’s “Job Split Doesn’t Always Work Magic” was published in the Wall Street Journal on the same date. These articles use research data to prove that whether the roles should be split depends on company circumstances and personalities rather than being determined—as taught in Policy Governance—on governance theory, that is, on a set of preexisting, coherent principles.

Relying on others’ studies as well as their own experience, Sonnenfeld and Lublin both argue that the advisability of splitting the chair and CEO roles is a mixed bag. For example, financially successful companies with combined roles are cited along with troubled companies with split roles, and vice versa. Although there is an undercurrent of disapproval of splitting the roles in both articles, neither author takes an inflexibly dogmatic stand on the matter. Sonnenfeld asks: “If [splitting the roles] was so universally priceless, why did the ‘try it, you’ll like it’ experiments [at IBM, Procter & Gamble, Home Depot, Boeing, Dell, GM] fail so quickly?” He also points out that Enron, WorldCom, Global Crossing, and HealthSouth had split roles “before they collapsed in criminal fraud.” However, both authors leave room for an “it depends” solution.

What it depends on is not entirely clear, but includes a finding that although struggling companies benefited from splitting the functions, companies that were already successfull were damaged by the same action. In other words, splitting the roles might be advantageous in some cases, but there is no overall case to be made for better company performance when roles are split.

Especially for readers outside North America, it should be pointed out that this is not new. There has been long and formidable resistance to separating the chair and CEO roles in the United States and Canada. Sonnenfeld reports, for example, that “only 23% of the S&P 500 firms have a truly independent director as Chairman,” whereas European companies look more favorably on separated roles.

Returning to the basis for resistance, the arguments and data that Sonnenfeld, Lublin, and others cited could lead a person not proficient in Policy Governance to conclude that the Policy Governance insistence on the integrity of role separation is at best arcane and at worst dogmatically pedantic. After all, these authors and most of their colleagues are not casual observers; they are recognized experts in their fields, whose opinion on this matter, unlike the Policy Governance position, is seemingly supported by real-world facts.

The Policy Governance position warns that assigning both chair and CEO jobs to one person under any circumstances introduces a serious role conflict in a setting already fraught with role confusion. More pointedly, it maintains that assigning the chair and CEO functions to the same individual distracts from optimal board job performance and always represents a conflict of interest.

It is safe to assume that if Sonnenfeld and Lublin know about Policy Governance, they do not know it in detail and are therefore unaware of the challenge its provisions pose to their conclusions. Nevertheless, viewing their reasoning in light of Policy Governance concepts lays bare a glaring weakness—a misunderstanding that lies in their concept of governance at the outset.

The Policy Governance concept of governance is that it is the job of a group of equals in ensuring the transformation of owners’ values into organizational behavior and outcomes. A group of equals, even while maintaining its group accountability, can appoint a first among equals to help it fulfill the discipline necessary for successful group performance. In contrast, to appoint a “boss” instead of an official servant-leader weakens and ultimately destroys the group members’ sense and fulfillment of their accountability. Therefore, it is critical for the board to be understood to be superior to the chair, not the reverse, regardless of personalities.

Furthermore, in the Policy Governance concept of governance, the board proactively establishes the definition of desired CEO performance. It does this in terms of defining desired results, recipients of results, and the priority or cost-effectiveness (“Ends”), along with placing limitations on executive authority to make all other (that is, non-Ends) decisions that could expose the organization to risk. The board is then able to delegate vast power safely to a CEO while simultaneously avoiding the time-consuming bottlenecks of management decisions. The line of authority goes from board to CEO, not board to chair to CEO (else the chair becomes de facto CEO regardless of title). Clearly the board role and thus the chair role are fundamentally different from a CEO role. The board, not the chair, interacts with management. Any blurring of these differentiations—for example, if the chair becomes the board’s boss, the chair becomes the CEO’s boss, or the CEO becomes the board’s boss (which can happen if the chair and CEO roles are combined)—and the system will have lost some or all of its capacity to discharge its enormous accountability while avoiding impeding managerial speed and vigor.

In other words, the critical matter is not whether one person fills the chair and CEO roles. The total design of these critical roles—board, chair, CEO—is the central issue; the combining of two roles in one person is but one element. Just as poor overall design cannot be cured by splitting the roles, combining the positions does not automatically preclude Policy Governance as long as the roles are still distinct, which is why Policy Governance strongly recommends but has never required splitting.

As to the kind of research question to which Sonnenfeld and Lublin addressed themselves, varying the combined versus split roles by itself in the absence of any information about the rest of the governance pattern cannot yield a meaningful conclusion.

Within a coherently designed system, one can vary single elements and gather meaningful data. For example, studying a board functioning in accordance with as lucid a job design as Policy Governance, one might observe the differential effects of, say, having split versus combined roles or employing one set of committees versus another. However, in a system without the theory-based coherence of Policy Governance, varying one element or another cannot lead to so clear an understanding.

This is because in poor systems, what could otherwise be a sensible improvement in one element can lead to worse performance in another and, conversely, an otherwise unpalatable change can lead to better performance. For example, in a poorly designed aircraft, the most ingenious new engine design might spell disaster; a less-than-optimal wing design might do less harm in a more advanced aircraft than in an older one.

Coming back to the board context, consider, for example, a completely unengaged board whose thoughtless inaction leaves its organization in a dangerous situation. Who would fault a board chair who illegitimately assumes board prerogatives in order to take action that saves the company? But in designing a board process, we would never plan for chairs to exercise undelegated authority. The fact that he or she has felt obliged to exercise authority without legitimate authorization by no means indicates that carefully limited chair authority doesn’t matter.

As Miriam Carver and I have specifically explained in a raft of publications, today’s typical governance pattern has grown like Topsy rather than being derived from logical theory. Unfortunately, it is from the current conceptually muddled governance that these authors drew their data, unaware that they had any other choice and, as several comments in their articles suggest, unaware of the deficiency it causes in their argument.

Sonnenfeld, for example, criticizes the two-position solution as “hydraheaded,” which makes sense only if there is a destructive ambiguity in the chain of command. But there is no such ambiguity in split roles (chair and CEO) in Policy Governance. He warns that “the separation of roles escalates palace intrigue,” again a picture of role ambiguity only if the Policy Governance treatment of roles is not followed. He notes that at General Motors, the former independent chairman “seized the chief executive job” from the existing CEO, not aware that if a chair has authority to do that, the board has not followed the careful limiting of chair authority as in Policy Governance. Furthermore, if splitting the roles does, as he claims, “[slow] decision making” and imperil sufficient “prudent risk taking” (I assume caused by a board’s hampering management), then clearly Policy Governance is not in effect.

Lublin notes there is “a mixed impact on shareholder returns when independent chairmen are named to keep a better eye on CEOs.” Conceiving the independent chair as a monitor of CEO behavior is bound to confound the issue of chair-CEO separation; that is why it is not allowed in Policy Governance. She says that those who want separated positions do so in order to put “a more formal check on CEO power.” If her representation of separatists’ intention is accurate—and I’ve no reason to think it isn’t—they themselves have pictured the chair role in a way that repeats the role combiners’ error: to conceive of the CEO as subordinate to the chair, a devastating blow to the board’s ability to hold the CEO accountable.

The direct, unfiltered board-to-CEO relationship is one of the key components in overall good governance design. When a chair acquires a spot on the chain of command, he or she simultaneously inherits some amount of CEO-ness, thereby institutionalizing ambiguity (hence, “palace intrigue,” “hydra-headedness”). In other words, you can split the roles, yet if each role does not carry with it an accountability-enhancing clarity, then it is to some extent as if the roles have not really been split. Moreover, the more the chair is drawn into personal supervision of the CEO, thereby becoming more managerially entangled than other board members, the more he or she is no longer independent after all.

In this article, I have tried only to avoid diluting the focus on the futility of conclusions like those expressed by Sonnenfeld and Lublin. However, as an additional matter, I should also mention that board performance is not the unequivocal determinate of organizational performance anyway, as I illustrated with school boards (“Is Student Achievement the Test of School Board Governance?” Board Leadership, no. 104, July–August 2009). The opinions and data in the two articles cited here assume such a direct link or, rather, make the even riskier assumption that the merits of a single structural element in governance can be assessed by company performance. (It is consequential that Sonnenfeld and Lublin’s focus is on structure. Assessing a piece of structure apart from the process surrounding it leads to unreliable conclusions.) Lublin, at least, does recognize that problem by quoting Gary Wilson, former independent chairman of Northwest Airlines, who said, “There are a million things that can affect shareholder returns and any number of things that are more important than a separate Chairman.”

Given the primitive conceptual state of governance, it is no wonder that Sonnenfeld can say that in his thirty-five-year career studying corporate governance, he has “come across no evidence to suggest that anything would be gained by separating [CEO and chair] roles.” I have no reason to doubt that his impression is exactly right, but not for the reasons he would give.

If I’ve left the impression that I think voices on the split role side of the debate are negligible or weak, let me correct that. For instance, following the Sonnenfeld op-ed, no less a luminary than Ira M. Millstein, chairman of the Center for Global Markets and Corporate Ownership, wrote a letter to the editor of the New York Times. Millstein maintained that for a board to accomplish its task “requires a leader who is not the chief executive” and if a “lead director [who is neither chair nor CEO] is strong enough to truly lead the board—it may well serve the same purpose,” although, he continues, “I would still assert that the separate Chairman is clearer in terms of leadership.”2 Former attorney general and New York governor Eliot Spitzer believes splitting the roles is a matter of principle. He has argued, “Shareholder democracy doesn’t function in terms of a meaningful check on the CEO. Therefore, the only check that remains is the board. And if a board is, as we’ve seen far too often, dominated by the CEO himself or herself, that check also disappears.”3 Former chairman of the Federal Deposit Insurance Corporation, William A. Isaac, made a case “that the functions of Chairman and CEO of financial institutions should be separated” and that “the common thread among [failed] institutions is that in nearly every case, the banks had a strong and dominant CEO and a weak and compliant board of directors.”4

This entire debate demonstrates how resolving secondary issues can be futile when the primary issue is unacknowledged. The primary issue here is that the basic design and purpose of governance cry out for a coherent framework. Not only have the world’s many corporate governance codes not offered a resolution of that omission, they have not even noticed its absence. Secondary issues, just like separate components in a machine, include what should characterize separate roles, how those roles should interact, and what insulation or protection some roles need from others. There is nothing esoteric about the proper order of those considerations, just a careful progression from broad decisions to smaller ones. Still, much of the governance debate revolves around various secondary issues made virtually unsolvable by a conceptually scrambled foundation. We chase about with one secondary issue after another, while underpinning concepts of governance go largely unexamined, like a person frantically swatting flies, oblivious to having left a window open.


1. As it turned out, the majority of JPMorgan Chase shareholders voted against a proposal that would split the chair and CEO roles. However, there were other issues afoot, largely matters of personality and the fear that Dimon might not stay without having both roles. As a former JPMorgan Chase executive, William Daley, said, “The vote and the debate right now about splitting the CEO and Chairman have nothing to do with splitting the CEO and Chairman.” Moore, M. J., Harper, C., and Kopeki, D. “The Only Man for the Job . . . and That’s the Problem,” Bloomberg Business Week, May 20–26, 2013, 62.

2. Millstein, I. M. “A Separate Chairman?” New York Times, May 17, 2013. Retrieved from

3. Quoted in Moore, “The Only Man for the Job,” 63.

4. Isaac, W. M. “Split Chairman and CEO Roles? It Depends,” American Banker, May 15, 2013.