How is CEO pay set? Theorists study this question through building models that are later tested by empiricists. While a great deal has been learned from theoretical models (see the surveys of Edmans and Gabaix 2016 and Edmans et al. 2017), they make several critical assumptions. For example:
- The standard objective function is firm value minus CEO pay, so the only downside of high CEO pay is to directly reduce shareholder value. However, it may also demotivate employees or harm the firm’s customer reputation. Moreover, the board’s objective function may not be shareholder value but to avoid losing a say-on-pay vote.
- Many models feature a binding participation constraint, which implies the CEO would leave if they were paid $1 less. However, it is not clear that the participation constraint binds in practice, given the magnitude of CEO salaries.
- Almost all models also feature a binding incentive constraint. However, the incentive constraint may not bind if intrinsic motivation or reputational concerns are sufficient to motivate the CEO.
- The CEO’s utility is a function of effort and consumption only. However, real-world CEOs likely also care about their reputation, fairness, and being appreciated by directors and investors.
- Most models feature no constraints other than participation, incentive compatibility, and perhaps limited liability. However, boards and CEOs do not negotiate pay in a vacuum. They may be constrained by norms on what investors and stakeholders deem acceptable, proxy advisory guidelines, history, and tax and accounting considerations.
Theories are typically evaluated not by the realism of their assumptions but by the empirical consistency of their predictions. While empirical research has also taught us a great deal, it too has limitations. First, data only documents the outcome of an optimisation problem and not the underlying program that led to it. Even if data is consistent with a model, a very different model may have generated it. Second, many key ingredients of compensation models are difficult to measure and thus test.
In Edmans et al. (2021), we survey over 200 non-executive directors of FTSE All-Share companies and over 150 institutional investors in UK equities on how they set or influence pay. Most questions allowed respondents to add free-text comments, and we conducted 14 post-survey interviews. The answers reveal several interesting results which challenge existing theories. We organise them into four groups:
Objective and constraints
We first ask respondents to rank the importance of three goals when setting CEO pay. Sixty-five per cent of directors view attracting the right CEO as most critical, while 34% prioritise designing a structure that motivates the CEO. For investors, these figures are 44% and 51%, respectively. This reversal reflects a theme that recurs throughout our survey – directors view labour market forces, and thus the participation constraint, as more important than investors, who prioritise the incentive constraint.
Only 1% of directors and 5% of investors view keeping the level of pay down as their primary goal. This is consistent with CEO pay being a small percentage of firm value, whereas the effects of CEO ability (Gabaix and Landier 2008) and effort (Edmans and Gabaix 2011) increase with firm size.
However, boards feel restricted by far more than the participation and incentive constraints standard models focus on. Sixty-seven per cent of directors admit that they are willing to sacrifice shareholder value to avoid controversy on CEO pay – from parties such as proxy advisers, employees, and customers.
Surprisingly, the strongest constraint is the need to obtain investor support, even though this should be automatic if boards are setting pay optimally. Instead, directors believe that shareholder guidelines, paradoxically, harm shareholder value: 77% report that such constraints have forced them to offer a lower level of pay, and 72% an inferior structure.
Most models of CEO pay take the ‘shareholder value’ view that pay is set by a single principal, a shareholder-aligned board. The main alternative is the ‘rent extraction’ view, whereby boards are captured by CEOs and thus do not seek to maximise value (e.g. Bebchuk and Fried 2004).
However, our free-text fields and interviews suggest a third perspective – directors and investors share the same objective (shareholder value) but view the world differently. One possibility is that directors better understand the CEO labour market, whereas shareholders push for changes that would violate the CEO’s participation constraint or demotivate the CEO. Another is that boards overestimate the value of their CEO or underestimate their latitude to improve pay.
To help disentangle these interpretations, we ask the 77% of directors who were forced to offer lower pay about the consequences. While 7% report that the CEO left, and 13% that they hired a less expensive CEO, 41% admit that there were no adverse effects. This result is meaningful since any self-serving bias would discourage this response. Thus, at least in some cases, boards overestimated the negative consequences of tough decisions on CEO pay.
However, 42% reported that the CEO was less motivated, suggesting that the level of pay affects incentives. This contrasts standard theories, where motivation depends only on pay-performance sensitivity. Instead, it is consistent with fairness being an important motivator. In the ‘gift exchange’ model of Akerlof (1982), if a CEO is given a gift of fair pay, they will reciprocate by providing discretionary effort. Similarly, Hertzberg (1959) suggests that fair pay is a ‘hygiene factor’ that demotivates if not provided.
Incentives and variable pay
There is greater agreement on the second set of questions – the role of financial incentives in motivating CEOs. Both boards and shareholders believe they are relevant but of secondary importance. The CEO’s intrinsic motivation and personal reputation are seen as most important, yet are absent from nearly all theories.
However, financial incentives still matter because they reinforce intrinsic and reputational incentives. CEOs believe it is fair to be rewarded financially for good performance; perceived unfairness would erode their intrinsic motivation. As one respondent stressed, “the retrospective acknowledgement of exceptional performance is important”. Separately, an increase in realised pay signals the CEO’s performance to outsiders, boosting the CEO’s reputation.
These responses suggest that incentive pay may work through different channels to standard models. In these models, the CEO only improves firm value if their utility from consuming the resulting pay increment exceeds the effort required to do so – the contract offers sufficient consumption incentives. Our respondents instead suggest that variable pay provides ex-post recognition. A CEO does not need the extra pay to finance consumption but believes it is fair to be recognised for a job well done.
The importance of ex-post recognition has two implications. First, it suggests that a CEO assesses their pay not only for the consumption utility it provides but also against their expectation of a fair reward. This expectation is believed to be affected by at least two reference points – the CEO’s contribution to the company and the pay of their peers.
Second, flow pay plays a special role not provided by the CEO’s equity stake. In standard theories, only total incentives matter – it is irrelevant whether they stem from changes in flow pay or the value of the CEO’s equity. Empirically, incentives from equity holdings are much greater, so standard measures of CEO incentives ignore flow pay (e.g. Hall and Liebman 1998). However, changes in flow pay provide greater ex-post recognition because they require a discretionary decision by the board and are voted on by shareholders. They are also publicly disclosed, boosting the CEO’s reputation. Thus, they may be important even if the CEO holds significant equity.
One other reason for variable pay supported by both directors and investors is for the CEO to share external risks with investors and stakeholders. This is surprising and contradicts standard theories (e.g. Holmström 1982) since it implies inefficient risk-sharing. It is, however, consistent with a fairness model in which directors and investors also evaluate CEO pay relative to a set of reference points that includes shareholder returns. As one investor explained, the CEO should be a co-owner who is “there for the journey” alongside other shareholders.
Our third set of results concerns the level of pay. We first ask what determines the pay of a new CEO. Both directors and investors view CEO ability as most important, consistent with Gabaix and Landier (2008) and Terviö (2008). Unexpectedly, pay at peer firms is seen as more important than the new CEO’s actual outside options, such as pay at their prior firm and at other firms they could move to.
One explanation, supported by the free-text fields and interviews, is that peer compensation matters not only because it determines the CEO’s alternatives, but because it is a reference point the CEO uses to assess whether their pay is fair.
When asked about increases in expected pay for incumbent CEOs, both directors and investors state that the primary justification is good recent performance. This is surprising given the substantial equity holdings that CEOs have, but is consistent with changes in flow pay providing ex-post recognition.
Directors and investors disagree on whether current pay levels are appropriate: 77% of investors view CEO pay as too high, and many believe that large cuts would have no adverse consequences. Most directors disagree, claiming that large cuts would force them to recruit lower-quality CEOs and adversely affect CEO motivation.
‘Suboptimal’ pay practices
Finally, we study the reasons for apparently suboptimal pay practices. The first is the limited use of relative performance evaluation, in contrast to Holmström (1982). Directors support three explanations, absent from existing theories, for why they do not filter out industry conditions from all performance measures.
One is again fairness – CEOs should benefit from an upswing since investors and stakeholders do. The other two reasons are practical – it can be difficult to find suitable peers for some firms or to observe peer performance for some performance measures. Explanations proposed by existing models, such as keeping pay competitive with peers during upswings (Oyer 2004), receive little support.
A second apparently suboptimal practice is the short-term nature of many pay incentives. Here, directors’ and investors’ views differ sharply. Seventy-eight per cent of investors believes the CEO would make better decisions if incentives were more long-term, as predicted by Edmans et al. (2012). Fewer than 6% agree with each of three potential concerns: that long-term incentives are less effective motivators, would jeopardise CEO retention or recruitment, or would require a costly adjustment in pay level.
In contrast, directors view incentives as already sufficiently long-term, and only 21% believe that further lengthening would improve decisions. Instead, they view all three concerns as important.
In sum, we have identified interesting directions for future academic research by uncovering important determinants of CEO pay, such as fairness, that have hitherto been overlooked.
Our results may also help practitioners by highlighting differences of opinion between directors and investors that cause conflicts about pay packages. Investor engagement often focuses on details of the contract itself, but initiating dialogues on these deeper disagreements may ultimately lead to more fruitful conversations on executive pay.
Akerlof, G A (1982), “Labor contracts as partial gift exchange”, Quarterly Journal of Economics 97: 543–69.
Edmans, A, and X Gabaix (2011), “Tractability in incentive contracting”, Review of Financial Studies 24: 2865–94.
Edmans, A, and X Gabaix (2016), “Executive compensation: A modern primer”, Journal of Economic Literature 54: 1232–87.
Edmans, A, T Gosling and D Jenter (2021), “CEO compensation: Evidence from the field”, CEPR Discussion Paper 16315.
Edmans, A, X Gabaix and D Jenter (2017), “Executive compensation: A survey of theory and evidence”, in B E Hermalin and M S Weisbach (eds.), Handbook of the Economics of Corporate Governance, Amsterdam: North-Holland, 383–539.
Edmans, A, X Gabaix, T Sadzik and Y Sannikov (2012), “Dynamic CEO compensation”, Journal of Finance 67: 1603–47.
Gabaix, X, and A Landier (2008), “Why has CEO pay increased so much?”, Quarterly Journal of Economics 123: 49–100.
Hall, B J, and J B Liebman (1998), “Are CEOs really paid like bureaucrats?”, Quarterly Journal of Economics 112: 653–91.
Herzberg, F (1959), The motivation to work, New York: Wiley.
Holmström, B (1982), “Moral hazard in teams”, Bell Journal of Economics 13: 324–40.
Oyer, P (2004), “Why do firms use incentives that have no incentive effects?”, Journal of Finance 59: 1619–50.
Terviö, M (2008), “The difference that CEOs make: An assignment model approach”, American Economic Review 98: 642–68.