Fair Pay or Power Play? Pay Equity,Managerial Power,and Compensation Adjustments for CEOs

Compensation Benchmarking for CEOs

Most academic studies of executive compensation focus on the relationships that exist within a firm, such as the relationship between a firm’s top management and board (for reviews, see Devers, Cannella, Reilly, & Yoder, 2007; Finkelstein, Hambrick, & Cannella, 2009). However, there is ample evidence that actual CEO pay setting is significantly shaped by CEOs’ or firms’ standing relative to other CEOs or firms. For instance, Miller (1995) showed that annual changes in CEO salary more accurately reflect performance relative to industry peers than their own firm’s past performance. More direct evidence exists regarding the compensation benchmarking among peers. In a study examining 160 proxy statements filed in 1993, Bannister and Newman (2003) found that 94% used the salaries of a peer group to some extent when setting CEO salaries. Similarly, Bizjak et al. (2008) reported that 96 of the 100 firms in the S&P 500 index used peer groups to determine executive compensation in 1997.

Why do firms engage in compensation benchmarking? Motivated by the criticism that firms use benchmarking to justify excessive pay increases (Ceron, 2004; Crystal, 1991), the existing literature offers two contrasting viewpoints. First, some research findings support the view that firms controlled by entrenched executives use compensation benchmarking in an opportunistic way to justify executive pay independent of performance. Hambrick and Finkelstein (1995) showed that a change in CEO pay at one firm is positively related to changes in CEOs’ pay throughout the industry, which they interpret as evidence of mimetic effects. Hambrick and Finkelstein demonstrated that firms mimic peers (i.e., use them as benchmarks) more actively when the firms are not controlled by strong owners, or in other words when the firms are controlled by entrenched CEOs. Ezzamel and Watson (1998) used data from the United Kingdom to show that firms adjust their executive pay based on the relative standing of their executives’ pay to that of their peers. Ezzamel and Watson also found that the response to peer information is asymmetric in that upward adjustments of pay for previously underpaid executives are stronger than downward adjustments in pay for previously overpaid executives. This finding is consistent with the argument that firms use peer benchmarking data to justify the ratcheting up of CEO compensation. Faulkender and Yang (2010) reported that firms in which CEOs have greater power are more likely to select highly paid peers to justify their levels of CEO compensation, which is also consistent with this view.

A contrasting perspective is that firms use benchmarking to maintain the competitiveness of pay packages and retain managerial talent. According to this perspective, benchmarking stems from concerns about pay equity and fairness in the competitive labor market, rather than a case of entrenched executives extracting rents from their firms. Bizjak et al. (2008) and Bizjak et al. (2011) provide evidence supporting this view. They found that CEOs whose pay is below the median of their peers’ pay typically receive subsequent pay raises that move them above the peer-group median, and that this upward adjustment in pay is not related to poor governance (Bizjak et al., 2008). Bizjak et al. (2011) also demonstrated that corporate governance variables do not explain biases in peer group selection (i.e., choosing peer firms that are bigger, better-performing, and higher-paying than the focal firm). Based on these findings, Bizjak and colleagues argued that benchmarking stems from the efficient behavior of firms in the presence of a competitive labor market for executives.

To understand why firms engage in compensation benchmarking, empirical research on how firms use the benchmarking information of peers to adjust CEO compensation is needed. If firms benchmark CEO pay to maintain pay equity and competitiveness in the executive labor market, then compensation adjustment of their CEOs should be explained by the CEOs’ relative standing among peers with regard to pay distribution. In contrast, if benchmarking is used opportunistically to justify CEO pay raises, the compensation adjustment should reflect the degree to which CEOs have power over the pay-setting process. The following sections present the arguments and testable hypotheses for these views, drawing from theories on pay equity and managerial power.

Pay Equity

According to equity theory (Adams, 1963), individuals evaluate the fairness of their situation by assessing the ratio of their inputs into their work (i.e., ability, effort, and job performance) and the outcomes (i.e., pay, benefits, and working conditions) and comparing their ratio to others’ ratio of inputs and outcomes. When individuals perceive the ratio of their inputs to outcomes to be equal to the ratios of others, equity exists. When the ratios are perceived to be unequal, inequity exists. If an individual’s ratio is lower than the ratios of others, the individual is underpaid; and if it is higher, the individual is overpaid. Adams argued that inequity creates tension or dissonance in the person perceiving it, motivating that person to reduce inequity. Adams described six possible methods of reducing inequity: (a) altering inputs, (b) altering outcomes, (c) psychologically distorting one’s own inputs or outcomes, (d) taking action to change inputs or outcomes of the others used for comparison, (e) changing the comparison other (i.e., the referent), or (f) leaving the position.

CEO compensation provides an excellent context for applying equity theory. First, unlike many other jobs, CEOs are often considered to have the capacity to influence their own pay (Bebchuk & Fried, 2004; van Essen et al., in press), which is one of the most salient “outcomes” discussed in equity theory. Among the six possible ways to reduce inequity, the option of “altering outcomes” has rarely been studied, except for in a case involving employee theft (Greenberg, 1990). Presumably, this is because few other jobs give employees this kind of ability to influence their own pay. Most existing studies have examined other methods of dealing with inequity, such as altering inputs (i.e., job performance; Currall, Towler, Judge, & Kohn, 2005; Werner & Mero, 1999) or leaving the position (i.e., turnover; Fong, Misangyi, & Tosi, 2010; Williams, McDaniel, & Nguyen, 2006).

Second, it is reasonable to assume that CEOs typically compare themselves with the CEOs of peer firms rather than with their subordinates or with individuals in different occupations. O’Reilly, Main, and Crystal (1988) found that the compensation of outside CEOs on compensation committees was positively related to the compensation of the CEO at a focal firm. This suggests that outside CEOs consider themselves to be valid referents for the setting of other CEOs’ pay. The issue of who is chosen as a referent is not clearly addressed in the original formulation of equity theory (Adams, 1963) and has thus been the subject of much research (Brown, 2001; Harris, Anseel, & Lievens, 2008). Having a narrowly defined comparison group, as in the case of CEOs, therefore provides a valuable advantage.

Finally, since publicly traded companies are required to disclose the compensation for their CEOs in proxy statements, information regarding the CEO pay is readily available, thereby enabling CEOs to easily compare their compensation with those of other CEOs. For most other jobs, pay secrecy is the norm. It is thus reasonable to assume that most CEOs are keenly aware of CEO pay at other firms and are particularly aware of the relative standing of their own pay to that of their peers. Moreover, information regarding inputs of CEOs in terms of firm performance is available through the market data. Although firm performance is not the only measure of CEO input, it is certainly one of the most important metrics of CEOs’ contributions to their companies. Changes in firm performance are commonly attributed to a CEO’s competence and leadership.

Some existing studies have found evidence supporting equity theory within the context of CEO compensation. Using data from 199 firms in the United Kingdom, Ezzamel and Watson (1998) found that when executives are underpaid or overpaid relative to those at other firms, their firms typically adjust their compensation in the subsequent year. Although this reaction to inequity is consistent with the predictions made by equity theory, it is unclear whether such adjustments are driven by executives’ reactions to inequity or to boards’ attempts to maintain competitiveness in order to retain and motivate executives. More directly addressing CEOs’ behavioral reactions to inequity, Fong et al. (2010) examined voluntary turnover and changes in firm size and profitability following pay inequity among 932 U.S.-based CEOs. Consistent with the predictions of equity theory, Fong et al. found that underpaid CEOs are more likely to increase the size of their firms or withdraw from their firms than are overpaid CEOs. Fong et al. also reported evidence that overpaid CEOs increase their efforts to increase firm profitability, presumably as a means of reducing pay inequity.

An important question in equity theory is who is chosen as a relevant other or referent (Brown, 2001; Harris et al., 2008). Adams’s (1963) equity theory argues that individuals compare themselves with others who are similar to them; however, it does not provide specific arguments about how referents are chosen. For CEOs, it is reasonable to assume that other CEOs are chosen as referents. More specifically, CEOs are likely to focus on other CEOs whose characteristics are similar to their own, particularly in terms of firm size, performance, and industry, because these are widely known to be the key determinants of CEO compensation. By cognitively controlling for the determinants (or inputs) of pay, CEOs would then be able to evaluate how their own pay (or outcomes) measures up against those of other CEOs. However, the referents chosen cognitively by CEOs may differ from those chosen officially by the board of directors for the purpose of compensation benchmarking, which are not necessarily selected based on the principle of similarity and may serve other strategic purposes of the firms. Recent research has demonstrated that boards have a tendency to choose peer firms that are larger and higher paying than their own (Bizjak et al., 2011; Faulkender & Yang, 2010).

Reactions to inequity may differ in cases of underpayment and overpayment. Adams’s (1963) original theory suggested that for a given magnitude of inequity in absolute terms, the intensity of the reaction is stronger for underpayment than for overpayment. According to Adams (1963: 426), “the thresholds for inequity are different (in absolute terms from a base of equity) in cases of undercompensation and overcompensation. The threshold would be greater presumably in cases of overcompensation, for a certain amount of incongruity in these cases can be acceptably rationalized as ‘good fortune.’” Similarly, Mowday (1996: 55) argued that “individuals are somewhat more willing to accept overpayment in an exchange relationship than they are to accept underpayment.” Consistent with this logic, Brown (2001) found that, in a survey of public sector research workers in Australia, underpayment had a greater impact on employee satisfaction with pay level than did overpayment. In a study of executives from the U.K., Ezzamel and Watson (1998) showed that upward adjustments of executive pay following underpayment were greater than downward adjustments following overpayment, resulting in a trend of upward ratcheting. These findings are consistent with the argument that underpayment triggers stronger reactions than does overpayment.

Ezzamel and Watson (1998) reported an asymmetry in the magnitude of compensation adjustment following upward versus downward payment, but the intensity or strength of the compensation adjustment may also differ in the cases of underpayment and overpayment. In other words, annual changes in a CEO’s pay may reflect annual changes in peer CEOs’ pay more or less strongly depending on whether the CEO in question is underpaid or overpaid. Underpaid CEOs may feel a strong urge to convince the board of directors to raise their pay, and may justify such raises by relying on the logic of external fairness. Members of a board’s compensation committee may also raise concerns about the competitiveness of executive compensation packages and avoid underpaying CEOs relative to the going rate in the executive labor market. Crystal (1991: 43), a former compensation consultant, describes a typical scenario: “If . . . you and your top managers are underpaid, you start the meeting with your newly retained compensation consultant by suggesting that he (or she) perform a survey of what other companies pay.”

A CEO’s motivation to restore equity may be weaker in the case of overpayment than underpayment, and for overpaid CEOs, the demand for peer benchmarking may not be as strong as it is for underpaid CEOs. Peer benchmarking may be unnecessary, or even undesirable, for an overpaid CEO, because external comparison would highlight the fact that the CEO is overpaid, and potentially prompt the board to adjust the compensation downwards. An overpaid CEO would rather rely on the logic of internal evaluation to justify his or her pay, for example, by citing above-average firm performance and the CEO’s contribution to it. Similarly, members of a board’s compensation committee may be less motivated to restore pay equity when the CEO is overpaid than when the CEO is underpaid. The financial cost of overpaying the CEO relative to the market is not directly borne by the directors, who typically do not have a substantial ownership stake in their company (Deutsch, Keil, & Laamanen, 2011; Linn & Park, 2005). Directors can justify CEO overpayment by citing the need to maintain the competitiveness of the compensation package and to motivate the CEO to improve future performance.

Combining the aforementioned arguments, the following hypothesis states that the strength of compensation benchmarking depends on whether a CEO is underpaid or overpaid. The strength or intensity of compensation benchmarking is studied by examining the relationship between annual changes in a CEO’s pay at a focal firm and annual changes in CEO pay at peer firms. If compensation benchmarking is vigorously used to establish a CEO’s compensation, then that CEO’s pay will be adjusted largely based on how much the compensation for the CEO’s peers is adjusted. Therefore, adjustments of CEO compensation will closely reflect the adjustment of peer CEOs’ compensation. The hypothesis predicts that the strength of the relationship between a CEO’s compensation adjustment and peer CEOs’ compensation adjustments differs in the cases of overpayment and underpayment, because CEO underpayment more strongly motivates the CEO to restore pay equity than CEO overpayment does.

Moderating Effects of Power

One limitation of equity theory is that it ignores the issue of power (Cook & Emerson, 1978; Hegtvedt, Thompson, & Cook, 1993; Kabanoff, 1991), which is particularly important in addressing situations in which individuals react to inequity by altering pay outcomes. The authority to determine employee compensation is given only to those with legitimate power in an organization; therefore, to influence compensation, one must possess a considerable degree of power over the pay-setting process. Other types of reactions to inequity that have been heavily studied in existing literature (i.e., altering inputs, adapting psychologically, or leaving the field) may require little or no power over employment relationships.

Management scholars have argued that in the context of executive compensation, the organizational and political power of top executives often plays a significant role in the pay-setting process (Bebchuk & Fried, 2004; van Essen et al., in press). Top managers’ power may stem from various sources such as the CEO’s formal position (as in the case of a CEO who is also the board chair), composition of the board (executives may sit on the board or outsider directors may have professional or personal connections to management), and psychological biases and tendencies of the directors to favor management (e.g., in-group bias and the norms of reciprocity). In addition, CEOs often have a significant influence over director nominations, affecting the composition of the board and the psychological tendencies of its members (Bebchuk & Fried, 2004; Westphal & Zajac, 1995).

Because managers generally prefer a higher level of compensation that is less risky (i.e., less contingent upon future performance), managerial power theory predicts that executives with a greater degree of power over the board will receive higher levels of pay and forms of pay that are less sensitive to performance. For example, studies have documented that CEO pay is higher when the CEO also serves as the chair of the board (Core, Holthausen, & Larcker, 1999) and when a greater proportion of directors serving on the compensation committee have been appointed during the current CEO’s tenure (Main, O’Reilly, & Wade, 1995). A meta-analysis of 219 studies showed that CEOs with greater power—measured by CEO-chair duality, the size of the board, and the ownership structure—tend to receive higher levels of pay (van Essen et al., in press). In a paper related to the current study, Faulkender and Yang (2010) found that firms in which the CEOs have relatively greater power (i.e., CEOs who are the chair of the board and who have a longer tenure) are more likely to choose benchmarking peers that are larger in firm size and higher paying than are firms that have relatively less powerful CEOs.

The motivation-ability framework (Anderson & Butzin, 1974; Lawler, 1966; Reinholt et al., 2011) provides a useful tool for integrating equity theory and the managerial power perspective. Originally developed to explain variations in job performance among individuals, a motivation-ability framework posits that ability moderates the relationship between motivation and performance. Although the original theory focuses on the role of functional ability rather than power, the concept of ability can be modified to address the role of power in organizations. Similarly, the concept of motivation can specifically refer to an individual’s political motivation to influence organizational processes and alter their outcomes. Therefore, a modified version of motivation-ability framework posits that the organizational process and outcomes such as executive compensation are shaped jointly by the political motivations and political abilities of the individuals involved in the process.

Equity conditions create motivation for action: “The presence of inequity will motivate Person to achieve equity or reduce inequity” (Adams, 1963: 427), but individuals may vary in terms of their ability to actually perform mitigating actions, particularly those that alter outcomes. In organizational settings, such ability typically stems from an individual’s organizational and political power. Mintzberg (1983) argues that the exercise of power in an organization requires both motivation and skill/ability. Using the motivation-ability framework and Minzberg’s concept of power, Liu, Liu, and Wu (2010) showed that political motivations and political skill (i.e., ability) interact to explain individuals’ career growth potential, as rated by individuals’ direct supervisors.

If power relationships shape the process of executive compensation, CEO power over the board may moderate the relationship between pay inequity and peer benchmarking. The motivation-ability framework can integrate pay equity theory and the managerial power perspective to provide more fine-grained predictions. CEOs who are relatively underpaid, and thus motivated to mitigate pay inequity, are likely to push boards to perform compensation benchmarking and to ascribe greater importance to external comparisons with peer CEOs in setting future compensation. To convince the directors of their companies and justify upward compensation adjustments, these CEOs may rely on the logic of pay equity and fairness. When underpaid CEOs have sufficient power over the board, directors may feel obliged to respect the CEOs’ concerns about pay equity. As a result, compensation committees may become more willing to incorporate peer benchmarking information into the pay-setting process and to match a CEO’s compensation with the compensation of peers more closely. On the other hand, if underpaid CEOs lack power to influence the board, their preference for pay equity and peer benchmarking may not be acted upon to justify upward adjustments in pay. Without significant pressure from CEOs, directors have greater autonomy in implementing internal, rather than external, evaluations, based on the principle of pay for performance. This suggests that the effects of CEO pay inequity (i.e., underpayment of the CEO) on the strength of peer benchmarking depend on the level of power that CEOs have over the board. In other words, the relationship between pay inequity and peer benchmarking is moderated by CEO power. As for the previous hypothesis, the strength or intensity of peer benchmarking is measured by the relationship between changes in a CEO’s pay at a focal firm and changes in peer CEOs’ pay.

The situation is different when CEOs are overpaid relative to their peers. As argued above, overpaid CEOs may want to minimize the compensation committee’s use of benchmarking information in setting their pay. Whether they are able to influence the compensation committee in this way, however, depends on their power over the board. Overpaid CEOs with sufficient power may be able to influence the board to avoid the extensive use of peer benchmarking and to justify the current level of pay by citing excellent firm performance under their leadership. On the other hand, if overpaid CEOs lack power, they may be subject to a more intensive external evaluation, and to a subsequent downward adjustment of compensation. The following hypothesis states the moderating effect of power in the case of CEO overpayment.

Data and Methods

Control Variables

The regression models include several control variables that may affect changes in CEO compensation. Firm size was measured in total assets in logarithm, and percentage changes in assets from year t – 2 to year t – 1 were entered into the model. For accounting measures of firm performance, the change in ROE from year t –2 to year t –1 was used. When return on assets (ROA) was used instead of ROE, the results were similar to those reported in this article. Total return to shareholders was used as a market measure of performance, calculated as the change in stock price from year t – 2 to year t – 1, adding dividends and dividing by the price at year t – 2. Levels of corporate diversification may also affect the intensity of peer benchmarking, as firms that are more diversified may find it difficult to focus on a homogeneous set of peer firms. The models in this study include a control for changes (from year t – 2 to year t – 1) in the number of business segments that a firm has in the same three-digit industry as the firm’s core industry, which is then divided by the firm’s total number of business segments. The data on business segments were obtained from Compustat Segment files.

A potentially confounding factor is the tendency for mimicry among firms. As proposed in institutional theory (DiMaggio & Powell, 1983), organizations may be subject to institutional forces that push them to seek the appearance of greater legitimacy by imitating each other’s practices without necessarily increasing substantive or economic effectiveness. Although such institutional forces can function at the organizational-field level across all firms, it is necessary to control for some firm-level variations in the mimetic forces that drive companies to engage in peer benchmarking. This study uses a proxy measure of interorganizational ties that is often utilized by researchers to measure mimetic processes: board interlocks. ² Ties among firms created by interlocking directorates facilitate the exchange of information and diffusion of novel practices, thereby serving as conduits for organizational mimicry (Haunschild, 1993; Mizruchi, 1996; Westphal & Zajac, 2001). Using board data from RiskMetrics and the information found in proxy statements, this study controls for changes in the number of board interlock ties to peer firms from year t – 2 to year t – 1, divided by the total number of directors per firm.

To test the hypotheses drawn from equity theory, it is important to control for variations in CEOs’ inputs such as skills, effort, and performance. Many studies have demonstrated that changes in CEO compensation typically depend on a CEO’s tenure. Annual increments in CEO pay may decrease with tenure; or on the contrary, CEO tenure may represent a CEO’s social and political influence over the board, which may positively affect annual changes in pay. To control for either effect, the models include CEO tenure in years. Similarly, the models also control for CEOs’ ages. Older executives may have more industry experience and a stronger social network in the business community than younger ones, but may possess knowledge and expertise that are not as up-to-date as those of younger CEOs. Information about CEO tenure and age has been taken from the ExecuComp database and proxy statements.

Firms vary in their timing of the fiscal year end. Benchmarking requires information about multiple firms’ performance and compensation levels, which means that it is important to match the timing of information releases across firms. Because the majority (64%) of firms in the sample had a December fiscal year end, the regression models include a dummy variable for a December fiscal year end. The substantive findings of the study remain unchanged when the analysis excludes firms with a fiscal year end other than December. Table 1 presents the means, standard deviations, and correlations for the variables. No variance inflation factor (VIF) score exceeds 2.00 and the mean score is 1.17, which is well below the threshold of 10 that is typically considered to represent potential multicollinearity. All dollar values have been adjusted for inflation.

OPEN IN VIEWER

Table 1

Means, Standard Deviations, and Correlations for Variables

Variables	Mean	SD	1	2	3	4	5	6	7	8	9	10	11	12
1. ΔCEO pay	0.00	0.60
2. ΔAssets	0.08	0.19	.08**
3. ΔROE	−0.12	20.13	.12**	.04**
4. ΔShareholder returns	−0.03	0.72	.13**	−.03**	.08**
5. ΔDiversification	−0.01	0.13	.02	−.02	−.01	.01
6. ΔBoard interlocks	0.00	0.03	.00	−.01	−.01	.01	−.01
7. CEO tenure	10.81	7.88	−.01	.03**	−.01	.00	−.01	.00
8. CEO age	57.11	6.85	.00	−.03**	.00	.01	−.03	.00	.46**
9. Decenber fiscal year end	0.64	0.48	−.02	−.01	.02	.01	−.01	−.01	−.05 **	.11**
10. ΔShareholder returns (peers)	0.00	0.43	.08**	.05**	.07**	−.26 **	.00	.00	.00	−.01	.01
11. ΔCEO pay (peers)	0.00	0.29	.27**	.05**	.05**	−.01	.01	−.01	.00	.01	.02	.16**
12. CEO underpayment	−0.11	0.26	−.51**	.07**	−.01	−.12**	−.04**	.02	−.03	.02	.01	.02	−.01
13. CEO overpayment	0.11	0.22	−.29**	.07**	.00	−.15**	−.03	.00	.04**	.01	−.01	.00	−.04**	.22**

**

p < .05.

Results

Table 2 presents the fixed-effects regression results predicting annual changes in CEO compensation. Model 1 presents the baseline model with control variables only. Model 2 includes changes in CEO compensation among peer firms, which are defined as other companies in the same two-digit industry and in the same size group (i.e., top or bottom half) as the focal firm. The results for Model 2 show that the coefficient for changes in peer CEO pay is positive and significant, which is consistent with the observation that firms use benchmarking information about peers to adjust their own CEO’s pay. In Model 2, the coefficient estimates for the control variables are not significantly different from those in Model 1.

OPEN IN VIEWER

Table 2

Fixed-Effects Regression Models Predicting Annual Changes in CEO Compensation

	Model 1	Model 2	Model 3	Model 4	Model 5
ΔAssets	.04	.04	.29***	.37***	.05
	(.04)	(.04)	(.04)	(.04)	(.04)
ΔROE	.00***	.00***	.00***	.00***	.00***
	(.00)	(.00)	(.00)	(.00)	(.00)
ΔShareholder returns	.10***	.10***	.07***	.05***	.03**
	(.02)	(.02)	(.02)	(.02)	(.01)
ΔDiversification	.12	.11	.14	.03	.09
	(.10)	(.10)	(.12)	(.06)	(.10)
ΔBoard interlocks	−.14	−.14	−.01	.05	−.04
	(.21)	(.21)	(.18)	(.19)	(.19)
CEO tenure	−.01	−.01	−.05***	−.06***	.00
	(.01)	(.01)	(.01)	(.01)	(.01)
CEO age	.00	.00	.00	.01	.00
	(.01)	(.01)	(.01)	(.01)	(.01)
December fiscal year end (dummy)	.13	.13	.03	.03	.12
	(.13)	(.12)	(.12)	(.14)	(.14)
ΔShareholder returns (peers)	−.05**	−.06**	−.07***	−.06***	−.04*
	(.02)	(.02)	(.02)	(.02)	(.02)
ΔCEO pay (peers)		.20***	.24***	.08	.19***
		(.05)	(.05)	(.06)	(.05)
CEO underpayment (dummy)			.37***
			(.02)
ΔCEO pay (peers) × CEO underpayment (dummy)			−.16**
			(.07)
CEO underpayment (continuous)				−.98***	.36***
				(.17)	(.03)
CEO overpayment (continuous)				−.48***	−.26***
				(.16)	(.02)
ΔCEO pay (peers) × CEO underpayment (continuous)				−.62	−.06
				(.69)	(.10)
ΔCEO pay (peers) × CEO overpayment (continuous)				.02	.14
				(.16)	(.09)
Year 1998	.28	.31	.04	.01	.23
	(.20)	(.20)	(.03)	(.03)	(.20)
Year 1999	.30	.32	.14***	.11***	.28
	(.20)	(.20)	(.04)	(.03)	(.20)
Year 2000	.31	.33	.13***	.14***	.28
	(.20)	(.21)	(.03)	(.03)	(.20)
Year 2001	.19	.24	.06	.07**	.16
	(.21)	(.21)	(.03)	(.03)	(.20)
Year 2002	.36*	.38*	.28***	.25***	.28
	(.20)	(.20)	(.04)	(.04)	(.20)
Year 2003	.33	.36*	.33***	.33***	.32
	(.20)	(.20)	(.04)	(.03)	(.20)
Year 2004	.41**	.41**	.43***	.43***	.34*
	(.20)	(.20)	(.04)	(.03)	(.20)
Year 2005	.30	.32	.35***	.37***	.25
	(.20)	(.20)	(.04)	(.04)	(.20)
Constant	−.17	−.19	.14	−.13	−.17
	(.41)	(.41)	(.33)	(.35)	(.40)
N	5,239	5,237	4,363	4,363	5,237
R 2	.10	.11	.23	.40	.22

Note: Robust standard errors are in parentheses.

*

p < .10. **p < .05. ***p < .01.

Based on the view that underpaid CEOs are motivated to restore equity more strongly than overpaid CEOs, Hypothesis 1 proposed that the relationship between changes in a CEO’s pay and changes in peer CEOs’ pay is stronger when the CEO is underpaid rather than overpaid. To test this hypothesis, Model 3 includes the interaction term between changes in peer CEOs’ pay and a dummy variable for CEO underpayment, which is set to 1 for underpaid CEOs (i.e., the regression residual has a negative value) and 0 for overpaid CEOs (i.e., the regression residual has a positive value). Hypothesis 1 predicted a positive and significant interaction effect. However, the coefficient estimate for the interaction term is in fact negative and significant (β = –.16). This indicates that the relationship between changes in a focal CEO’s pay and changes in peer CEOs’ pay is weaker for underpaid CEOs than for overpaid CEOs. For overpaid CEOs (underpayment dummy = 0), the slope of the relationship between changes in a focal CEO’s pay and changes in peer CEOs’ pay is .24. For underpaid CEOs (underpayment dummy = 1), the slope is .24 – .16 = .08.

Although Model 3 provides a direct test of Hypothesis 1, it uses a dichotomized variable of overpayment and underpayment, which is constructed from a continuous measure of pay equity variable. To avoid information loss from dichotomizing the continuous measure, Model 4 uses the continuous measures of CEO underpayment and overpayment, as described in the methods section above. The CEO underpayment variable is equal to the residual from the regression predicting CEO compensation in year t – 1 if it is negative, and is zero otherwise. Similarly, the CEO overpayment variable is equal to the residual if it is positive, and is zero otherwise. The main effect of CEO underpayment indicates that the compensation for underpaid CEOs, who have negative values for CEO underpayment variable, tends to be adjusted upward the next year. By contrast, if a CEO was overpaid in the previous year, the current compensation for the CEO tends to be adjusted downward. As is consistent with equity theory, deviations from pay equity (“pay anomalies” [Ezzamel & Watson, 1998]) lead to compensation adjustments that bring pay closer to the average.

To test Hypothesis 1, interaction terms between changes in peer CEOs’ pay and overpayment/underpayment variables are included in Model 4. A significant interaction term would indicate that the effect of changes in peer CEOs’ pay on changes in focal CEO’s pay varies depending on the overpayment and underpayment of the CEO. Hypothesis 1 predicts that the interaction effect between changes in peer CEOs’ pay and CEO underpayment is significantly greater than the interaction effect between changes in peer CEOs’ pay and CEO overpayment. Contrary to Hypothesis 1, the coefficient estimate for the interaction between peer pay and CEO underpayment (β = –.62) in Model 4 is smaller than the coefficient estimate for the interaction between peer pay and CEO overpayment (β = .02). A t-test suggests that the difference between the two coefficients is not statistically significant. Therefore, there is no evidence that the effects of changes in peer pay differ between the cases of underpayment and overpayment.

As a robustness check, the overpayment and underpayment variables are modified to focus on situations where the CEOs are severely underpaid or overpaid. Following Fong et al.’s (2010) approach, CEO overpayment is coded 1 if the residual from the regression predicting CEO compensation in year t – 1 is in the highest quartile in the sample, and 0 otherwise. CEO underpayment is coded 1 if the residual from the regression predicting CEO compensation in year t – 1 is in the lowest quartile in the sample, and 0 otherwise. Model 5 of Table 2 shows the results. Again, the coefficient estimate for the interaction between peer pay and CEO underpayment (β = –.06) is smaller than the coefficient for the interaction between peer pay and CEO overpayment (β = .14). The difference between the two coefficients is not statistically significant. In sum, Hypothesis 1 is not supported. The relationship between changes in a CEO’s pay and changes in peer CEOs’ pay does not significantly vary based on pay inequity.

The lack of support for the predictions based on pay equity theory may be due to the omission of managerial power in the models. CEOs have varying degrees of power, which determine whether they can actually address pay inequity and pressure the board to incorporate benchmarking information into pay settings. As such, CEO power may moderate the relationship between a focal CEO’s pay and peer CEOs’ pay. Hypotheses 2 and 3 predicted the moderating effects of CEO power. Specifically, Hypothesis 2 suggested that for underpaid CEOs, the relationship between changes in a focal CEO’s pay and changes in peer CEOs’ pay would be stronger for CEOs who have greater power over the board than for CEOs with relatively less power over the board. Hypothesis 3 predicted the converse, suggesting that for overpaid CEOs, the relationship between changes in a focal CEO’s pay and changes in peer CEOs’ pay would be weaker for CEOs with greater power over the board than for CEOs with relatively less power over the board.

The models in Table 3 test Hypotheses 2 and 3. The sample is split into underpaid and overpaid CEOs. Underpaid CEOs are those with a negative value of the residual from the regression predicting CEO compensation in year t – 1; overpaid CEOs are those with a positive value of the residual. Hypothesis 2 is tested in Models 1, 3, 5, and 7, which conduct estimations for underpaid CEOs only. Models 2, 4, 6, and 8 test Hypothesis 3 for overpaid CEOs. The hypotheses are tested by using the interaction terms between changes in peer CEOs’ pay and the focal CEO’s power, measured in terms of four aspects: CEO-chair duality, proportion of outside directors, proportion of directors appointed after the CEO, and proportion of outside directors appointed after the CEO. Models 1 and 2 use CEO duality as an indicator of CEO power. For underpaid CEOs (Model 1), the interaction between peer pay and the dummy variable for CEO duality is positive and marginally significant in a two-tailed test (p = .80). This suggests that the strength of the relationship between changes in the focal CEO’s pay and changes in peer CEOs’ pay is greater for CEOs who hold the position of board chair than for CEOs who do not hold the position of board chair. This is consistent with Hypothesis 2. However, Model 2 indicates that among overpaid CEOs, the interaction between peer pay and CEO duality is not significant. Hypothesis 3 is not supported. Figure 1 demonstrates the interaction effect for underpaid CEOs. Following Aiken and West’s (1991) procedure, the values of high and low peer pay are set at 1 standard deviation greater than and less than the mean, respectively. All independent variables are standardized for the graph. The lines represent the relationship between changes in peer CEOs’ pay (x-axis) and changes in a focal CEO’s pay (y-axis). The slope for CEO-chairs is steeper than that for non-CEO-chairs, consistent with Hypothesis 2.

OPEN IN VIEWER

Table 3

Fixed-Effects Regression Models Predicting Annual Changes in CEO Compensation for Underpaid and Overpaid CEOs

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7	Model 8
	Underpaid	Overpaid	Underpaid	Overpaid	Underpaid	Overpaid	Underpaid	Overpaid
	CEOs	CEOs	CEOs	CEOs	CEOs	CEOs	CEOs	CEOs
ΔAssets	.10	.40***	.11	.41***	.10	.41***	.10	.41***
	(.09)	(.06)	(.09)	(.06)	(.09)	(.06)	(.09)	(.06)
ΔROE	.00**	.01***	.00**	.01***	.00**	.01***	.00**	.01***
	(.00)	(.00)	(.00)	(.00)	(.00)	(.00)	(.00)	(.00)
ΔShareholder returns	.06***	.05*	.06***	.05*	.06***	.05*	.06***	.05*
	(.02)	(.03)	(.02)	(.03)	(.02)	(.03)	(.02)	(.03)
ΔDiversification	.31*	−.09	.31*	−.09	.32*	−.09	.32*	−.08
	(.16)	(.09)	(.17)	(.09)	(.17)	(.09)	(.17)	(.09)
ΔBoard interlocks	.48	−.03	.44	−.02	.43	−.05	.42	−.05
	(.59)	(.17)	(.60)	(.17)	(.61)	(.17)	(.61)	(.17)
CEO tenure	−.01	−.32**	−.01	−.32**	−.01	−.32**	−.01	−.33**
	(.01)	(.15)	(.01)	(.15)	(.01)	(.14)	(.01)	(.14)
CEO age	.00	.24*	.00	.24*	.00	.24*	.00	.25*
	(.00)	(.15)	(.00)	(.15)	(.00)	(.15)	(.00)	(.14)
December fiscal year end (dummy)	.09	−.18	.10	−.16	.09	−.17	.09	−.17
	(.11)	(.25)	(.12)	(.25)	(.12)	(.24)	(.12)	(.24)
ΔShareholder returns (peers)	−.05	−.05*	−.05	−.05*	−.05	−.05*	−.05	−.05*
	(.03)	(.03)	(.03)	(.03)	(.03)	(.03)	(.03)	(.03)
ΔCEO pay (peers)	.02	.11	−.02	.19	.16	.39***	.20	.45***
	(.15)	(.09)	(.26)	(.19)	(.14)	(.13)	(.13)	(.14)
CEO duality	−.01	−.10
	(.05)	(.09)
ΔCEO pay (peers) × CEO duality	.27*	.06
	(.15)	(.10)
Proportion of outside directors			−.17	.17
			(.18)	(.13)
ΔCEO pay (peers) × Proportion outside directors			.36	−.06
			(.35)	(.29)
Proportion of directors appointed by CEO					−.03	.12
					(.15)	(.14)
ΔCEO pay (peers) × Proportion directors appointed by CEO					.11	−.42**
					(.24)	(.20)
Proportion of outside directors appointed by CEO							.07	.12
							(.15)	(.10)
ΔCEO pay (peers) × Proportion outside directors appointed by CEO							.02	−.45**
							(.19)	(.18)
Year 1998	.02	.01	.02	.01	.02	.02	.03	.02
	(.04)	(.04)	(.04)	(.04)	(.04)	(.04)	(.04)	(.04)
Year 1999	.16***	.05	.16**	.05	.17***	.05	.17***	.04
	(.06)	(.05)	(.06)	(.05)	(.06)	(.05)	(.06)	(.05)
Year 2000	.10**	.16***	.10*	.16***	.10**	.17***	.10**	.16***
	(.05)	(.04)	(.05)	(.05)	(.05)	(.04)	(.05)	(.04)
Year 2001	.10*	.01	.09*	.02	.10*	.02	.10*	.02
	(.06)	(.05)	(.05)	(.05)	(.06)	(.05)	(.06)	(.05)
Year 2002	.26***	.22***	.24***	.22***	.25***	.23***	.25***	.23***
	(.07)	(.06)	(.07)	(.06)	(.07)	(.06)	(.07)	(.06)
Year 2003	.27***	.35***	.26***	.35***	.26***	.35***	.27***	.35***
	(.05)	(.05)	(.05)	(.05)	(.05)	(.05)	(.05)	(.05)
Year 2004	.32***	.48***	.30***	.48***	.31***	.48***	.31***	.48***
	(.06)	(.06)	(.06)	(.06)	(.06)	(.06)	(.06)	(.06)
Year 2005	.27***	.51***	.25***	.51***	.26***	.51***	.26***	.51***
	(.05)	(.06)	(.05)	(.06)	(.05)	(.06)	(.05)	(.06)
Constant	.02	−10.79	.16	−10.91	.03	−10.70	.03	−11.06*
	(.18)	(6.85)	(.23)	(6.85)	(.18)	(6.76)	(.18)	(6.61)
N	2,122	2,241	2,122	2,241	2,122	2,241	2,121	2,238
R 2	.09	.24	.09	.24	.09	.24	.09	.25

Note: Robust standard errors are in parentheses.

*

p < .10. **p < .05. ***p < .01.

OPEN IN VIEWER

Figure 1

Annual Changes in Focal CEO Pay and Annual Changes in Peer CEO Pay for Underpaid CEOs Who Are Chairs and Nonchairs

Models 3 and 4 use the proportion of outside directors as a measure of CEOs’ potential power. Boards with a higher proportion of outsiders are assumed to be less sensitive to the influence of management, and can therefore constrain CEO power more effectively. For both underpaid (Model 3) and overpaid CEOs (Model 4), the interaction terms between changes in peer CEOs’ pay and the proportion of outside directors are not significant. Therefore, Hypotheses 2 and 3 are not supported when using the proportion of outside directors as a measure of CEO power.

As another measure of CEO power, Models 5 and 6 use the proportion of directors appointed after the CEO’s appointment. A high proportion of directors appointed during the CEO’s tenure indicates the CEO’s greater influence over the board. For underpaid CEOs (Model 5), the interaction between changes in peer CEOs’ pay and the proportion of directors appointed after the CEO is not significant. Hypothesis 2 is not supported. However, Model 6 shows that for overpaid CEOs, the interaction is negative and significant, suggesting that the relationship between changes in a focal CEO’s pay and changes in peer CEOs’ pay is weaker when a higher proportion of directors were appointed during the CEO’s tenure. This supports Hypothesis 3. Figure 2 illustrates the moderating effect, using the overpaid CEO sample. As in the previous graph, the high and low values are set at 1 standard deviation greater than and less than the mean, respectively. All independent variables are standardized for the graph. For overpaid CEOs, the relationship between changes in peer CEOs’ pay (x-axis) and changes in a focal CEO’s pay (y-axis) is weaker and the slope is less steep when a higher proportion of directors were appointed during the CEO’s tenure.

OPEN IN VIEWER

Figure 2

Annual Changes in Focal CEO Pay and Annual Changes in Peer CEO Pay for Overpaid CEOs at High Versus Low Proportion of Directors Appointed by CEO

In Models 7 and 8, the variable for the proportion of directors appointed after the CEO is modified to measure the proportion of outside directors appointed after the CEO. The results are similar to those of Models 5 and 6. The interaction between changes in peer CEOs’ pay and the proportion of outside directors appointed after the CEO is not significant for underpaid CEOs (Model 7) but negative and significant for overpaid CEOs (Model 8). This supports Hypothesis 3. Figure 3 illustrates the interaction effect for overpaid CEOs. The high and low values are set at 1 standard deviation above and below the mean, respectively. All independent variables are standardized for the graph. For overpaid CEOs, the relationship between changes in the focal CEO’s pay and changes in peer CEOs’ pay is weaker when a higher proportion of outside directors were appointed after the CEO’s appointment. Therefore, when the proportion of directors (or outside directors) appointed after the CEO is used to measure CEO power, Hypothesis 3 is supported.

OPEN IN VIEWER

Figure 3

Annual Changes in Focal CEO Pay and Annual Changes in Peer CEO Pay for Overpaid CEOs at High Versus Low Proportion of Outside Directors Appointed by CEO

As a robustness check, all models in Tables 2 and 3 were reestimated in several alternative ways (results are available upon request). The findings were substantively similar across all variations. First, the peer group definition was modified in the following ways: (1) using three-digit industry only, (2) using three-digit industry and firm size (above and below average), and (3) using two-digit industry and firm size quartiles. Second, the findings were robust when either base or total compensation was used as the dependent variable. Third, instead of using the log difference in pay as the dependent variable, the percentage change in pay from year t – 1 to year t was used. Fourth, for a smaller sample, in which information on CEOs’ full career history was available, the insider/outsider status of the CEO was controlled for. Fifth, the models were reestimated excluding the 42 CEO-firm-year observations for cases in which co-CEOs existed. Finally, the substantive findings were not sensitive to different estimation techniques and specifications, including (1) random-effects models, (2) random-effects models with dummy variables for one-digit SIC industries, and (3) OLS regression models without fixed effects.

Discussion and Conclusion

Drawing from psychological and political perspectives on executive compensation, this study examined the use of benchmarking information about peers in setting CEO compensation. The analysis of data from 1,555 CEOs indicates that firms closely link their own CEO’s pay to the pay of CEOs at peer firms. Consistent with the motivation-ability framework, the results suggest that the relationship between a focal CEO’s pay and peer CEOs’ pay is moderated by the level of a CEO’s power over the board. CEOs who are underpaid relative to their peers may desire to mitigate pay inequity by urging their boards to implement peer benchmarking and incorporate benchmarking information into the process of setting their compensation. Overpaid CEOs, on the other hand, may be motivated to avoid an extensive use of peer benchmarking and prevent downward adjustments of compensation. Without the political and interpersonal ability to influence directors effectively, however, CEOs are unable to act upon their preference for peer benchmarking. Predictions based solely on equity theory, without considering the moderating effect of power, were not supported in empirical testing. Contrary to equity theory’s prediction that CEOs react more strongly to pay inequity when they are underpaid than overpaid relative to their peers, the results suggested that the effect of pay inequity on compensation adjustment was not significantly greater for underpaid CEOs than for overpaid CEOs. The motivation to react to underpayment is not enough to cause adjustments in compensation. To remedy the conditions of underpayment successfully, CEOs also need to be in a position that provides sufficient power and influence over the process of determining executive compensation.

When underpaid CEOs are in a position to influence the board, a subsequent increase occurs in the use of peer benchmarking information and the upward adjustment of compensation. This is the case for CEOs who also hold the board chair position. When these CEOs are motivated to resolve underpayment by emphasizing the logic of market fairness and peer comparison in their communication with the board, the directors are more likely to pay attention to the CEOs’ demands and to decide to grant them pay raises, because the CEO-chairs are in a position with authority and influence over the directors.

For overpaid CEOs, compensation benchmarking is rather undesirable. Overpaid CEOs are motivated to minimize the use of benchmarking information in setting their compensation. The findings suggest that when overpaid CEOs possess power over the board, indicated by a greater proportion of handpicked directors, they may be able to influence the board to avoid the extensive use of peer benchmarking and a downward adjustment of compensation.

The measures of CEO power used in this study do not uniformly support the hypotheses. Hypothesis 2 is supported only when CEO duality is used as a measure of CEO power. Justification of upward adjustments in pay may require a clear demonstration of power and authority. CEO duality signals formal authority based on structural position of the CEO (Main et al., 1995; Pollock et al., 2002). The other measures used in this study—the proportion of outside directors and the proportion of (outside) directors appointed after the CEO—may be too subtle to capture the CEO’s formal authority. On the other hand, Hypothesis 3 is supported only when the proportion of (outside) directors appointed after the CEO is used as a measure of CEO power. The proportion of outside directors, a popular measure in the literature, does not offer support for either of the hypotheses. Researchers have raised doubts whether the traditional inside/outside director distinction captures boards’ independence (Finkelstein & Mooney, 2003; Johnson et al., 1996). Owing to the social and psychological relationships that exist between directors and the managers, directors who are formally categorized as outsiders may not be truly independent of management. Compared to the inside/outside distinction, the proportion of (outside) directors appointed during the current CEO’s tenure more accurately captures the presence of loyal directors who are likely to succumb to a CEO’s preference for peer benchmarking, particularly when it comes to the avoidance of a downward adjustment of compensation.

The findings suggest that the processes of compensation adjustment differ in the cases of underpayment and overpayment, shedding light on the underlying behavioral mechanism that has led to the rise in executive compensation over the past two decades. Ezzamel and Watson (1998) report on one aspect of this asymmetry by demonstrating that adjustments of executive compensation following underpayment in firms in the U.K. are greater than downward adjustments following overpayment. Using simulation methods, DiPrete et al. (2010) showed that the continuing practice of such asymmetric responses to pay inequity would lead to the dramatic increase of executive pay across the board over time, a phenomenon which is often called the “Lake Wobegon” effect (Morgenson, 2006). It is implicit in all existing discussions on this topic that all CEOs can potentially benefit from these increases. The present study, however, shows that CEOs may vary in their political ability to respond to pay inequity, and that this variation leads to differences in compensation adjustments over time. Only powerful CEOs are able to successfully react to pay inequity by promoting upward adjustments when they are underpaid and by avoiding downward adjustments when they are overpaid. Eventually they seem to benefit from the asymmetric ratcheting-up process leading to potentially dramatic compensation increases.

The main contribution of this study is that it incorporates both equity and power perspectives to explain compensation adjustments. It has been argued that the concepts of equity, fairness, and distributive justice are inherently intertwined with power (Cook & Emerson, 1978; Hegtvedt et al., 1993; Kabanoff, 1991). Pay inequity motivates individuals to reduce inequity, but political ability or power is necessary to achieve the desired outcomes. This study adopts a motivation-ability framework to integrate the motivational side of pay equity theory and the political side of the managerial power perspective. Existing studies of peer benchmarking have addressed either a pay-equity (Ezzamel & Watson, 1998) or managerial power perspective (Bizjak et al., 2008, 2011; Faulkender & Yang, 2010) alone, and no study has thus far addressed both perspectives simultaneously. As Porac et al. (1999) and Wade et al. (2006) show by using different types of analyses, executive compensation provides a context in which the mechanisms of equity and power relations function simultaneously. Several organizational conditions facilitate the joint functioning of equity and power relationships: uncertainties and ambiguities about the evaluation of executive talent, availability of peer information, considerable autonomy and authority of top managers, and limitations of board vigilance.

This study makes specific contributions to pay equity theory. Commentators have noted at least two major limitations of the theory (Gerhart & Rynes, 2003; Mowday, 1996). First, only selected consequences of pay inequity—most notably pay dissatisfaction (Williams et al., 2006), performance (Werner & Mero, 1999), and turnover (Fong et al., 2010)—have been heavily studied, although the original theory of Adams (1963) proposed several different types of consequences, including the effort to alter job rewards and the outcome. The current study examines how executives react to pay inequity by influencing the pay-setting process, which is a consequence that has rarely been studied. Executive compensation provides a unique setting in which executives have a considerable degree of influence over their own pay.

Another limitation of equity theory is that empirical researchers have reported equivocal findings about the difference between underpayment and overpayment situations. More specifically, research findings on overpayment have been less consistent than those on underpayment (Ambrose & Kulik, 1999; Mowday, 1996). This suggests that the nature of reactions to overpayment may be specific to their contexts. This study provides evidence on this matter from the context of CEO compensation. The findings of this study suggest that the moderating effect of power differs in the cases of underpayment and overpayment. Underpayment prompts the CEOs to exercise formal authority over the board to justify an upward adjustment of pay through a more extensive use of external benchmarking. Overpayment, by contrast, motivates CEOs to rely on a more subtle form of interpersonal influence in order to avoid benchmarking and a downward adjustment of pay.

The contributions of the present study to the literature on executive compensation are twofold. First, its findings extend existing knowledge about the peer benchmarking of CEO compensation. Although several studies have documented the prevalence of competitive benchmarking and its impact on the ratcheting up of executive pay (Bizjak et al., 2008, 2011; DiPrete et al., 2010; Faulkender & Yang, 2010), the question of what firms do with peer benchmarking information has not been thoroughly investigated. Existing literature is unclear as to whether firms benchmark to accommodate concerns about pay equity or because organizational conditions enable executives to exercise their power over the pay-setting process. For instance, Ezzamel and Watson (1998) found that the amount of upward adjustment of pay for underpaid CEOs is greater than the amount of downward adjustment for overpaid CEOs; however, they neither provide explanations for the asymmetric adjustment of compensation, nor explore the organizational conditions that moderate the relationship between pay inequity and peer benchmarking. The present article introduces power relationships as the moderating variable that explains the asymmetry discovered by Ezzamel and Watson. Hambrick and Finkelstein (1995) argued that the close association between an individual CEO’s pay and average CEO pay in the industry is the result of a mimetic effect across firms, but do not observe or measure the actual actions of interorganizational imitation. The present study controls for board interlocks as interorganizational ties that are known to facilitate mimicry, ruling out mimetic effects as an alternative explanation. Finally, Bizjak et al. (2008) argued that the upward adjustment of pay for underpaid CEOs stems from firms’ efforts to retain executive talent rather than from poor governance. Although this may appear inconsistent with the findings of the present study, it may not necessarily be the case: The dependent variable in Bizjak et al. (2008) is the upward adjustment of a CEO’s pay from below the median to above the median for peer CEOs’ pay, whereas the present study analyzes the strength of the association between a focal CEO’s pay and peer CEOs’ pay, which is a more proximate (albeit indirect) measure of peer benchmarking.

This study also contributes to existing literature on the role of managerial power in the executive compensation context. Although numerous studies have directly and indirectly addressed the impact of managerial power on the level of executive pay and the relationship between pay and performance (van Essen et al., in press), relatively little attention has been paid to the issue of how managerial power affects the pay-setting process. Managerial power should affect the pay-setting process before it is reflected in the compensation awarded to executives. This study explores an important aspect of the pay-setting process—compensation benchmarking—and how it is shaped by managerial power. Its findings suggest that CEOs’ power over the board enables them to influence the pay-setting process. With sufficient power and control over the board, CEOs can strategically use compensation benchmarking to their own benefit, promoting benchmarking when they are underpaid and avoiding benchmarking when they are overpaid. In the end, this practice of compensation benchmarking, shaped by the CEO’s political influence, is likely to result in an increase in CEO pay, as has been vividly described by compensation consultants and journalists (Ceron, 2004; Crystal, 1991). This may explain the close statistical association between CEO power and pay that has been reported by many researchers. The present study is an attempt to specify the relationship between power and pay, by arguing that a CEO’s power leads to compensation benchmarking that is implemented in a manner that is more favorable to the CEO and is likely to result in increases in pay over time.

A related contribution to the literature on managerial power is that the current study used four different measures of CEO power to address various aspects of power. Not all measures of power provided equally significant support for the study’s hypotheses. In existing literature on managerial power in the executive compensation context, it is common that researchers use disparate measures of power and report inconclusive findings. Using similar measures of power as in this paper, a meta-analysis of the relationship between CEO power and compensation also showed that not all indicators of power unequivocally support hypotheses drawn from the managerial power perspective (van Essen et al., in press). The inconclusiveness of some of these findings reflects the challenge of studying power as a multidimensional concept. To study the multiple dimensions of managerial power more adequately, future research should adopt further refined measurement strategies that are rooted in sound theory.

Some limitations of this study should be noted. First, peer group membership was indirectly assigned based on industry and size categories. The actual composition of peer groups may differ from those assigned. For instance, critics have indicated the practice of handpicking peer firms to create what is sometimes called “convenience peer groups” or “aspirational peer groups” (Ceron, 2004). Studies measuring actual peer group membership based on proxy statements examine shorter periods and show that traditionally (at least prior to 2006) many firms have simply not disclosed peer group membership (Porac et al., 1999). A change in SEC disclosure requirements in 2006 may enable researchers to gather more detailed information on benchmarking from more recent periods (Faulkender & Yang, 2010).

Second, the effects of CEO power were inferred in this study by using the variables that represent CEO power. They were not measured through direct observation. As Finkelstein and Mooney (2003), Hambrick et al. (2008), and Johnson et al. (1996) have argued, structural indicators such as board composition may not fully represent actual behavior. Although indirect measures are common in field studies using large samples, more work is needed to specify the exact scope and mechanism of these social processes. For example, a survey of top executives may be useful. Such efforts have been relatively rare due to the substantial cost of administrating surveys among top managers with regard to sensitive topics such as compensation and power. Existing surveys of top executives tend to suffer from low response rates, that is, usually no more than 20 percent (Chattopadhyay, Glick, Miller, & Huber, 1999). It is also extremely costly to conduct a repeated longitudinal study with a set panel of respondents. Another way to extend the current research would be to apply a qualitative approach to executive compensation and corporate governance (e.g., Useem, 1996). Such an approach could be useful in that it would provide a thick description of actual social dynamics that is typically unavailable in quantitative research.

Although the use of longitudinal data spanning an 11-year period (1996 through 2006) reduces the possibility of reverse causality and controls for unobserved heterogeneities across firms and CEOs, one potential limitation stems from the fact that major regulatory changes occurred during the study period, such as the introduction of the Sarbanes-Oxley Act of 2002 and the change in the listing requirements of the New York Stock Exchange in 2003. These rule changes have generally increased the importance of board vigilance and challenged executives’ discretionary power. Although this study controls for variations across time periods by using year dummies, its analysis is limited to the test of theory-driven hypotheses for any given time period, rather than an exploration of systematic changes in corporate governance over time.

This study suggests several avenues for future research. Three questions about compensation benchmarking need to be explored. The first question is, why do firms benchmark? The present study examines the predictions of equity theory and the managerial power perspective; however, these two may not be the only approaches that explain compensation benchmarking. For example, an application of institutional theory could emphasize the role of mimetic isomorphism in explaining the adoption of compensation benchmarking as a legitimate management practice. Moreover, an interorganizational network perspective could focus on the way that actual ties between firms—such as strategic alliances—affect the formation of quasi-ties established through compensation benchmarking by grouping peer firms together.

The second question is, which firms are included in the peer groups? The question of how referents are selected is important in pay equity theory (Brown, 2001; Harris et al., 2008) and relates to the issue of categorical structure in organizational fields (DiMaggio & Powell, 1983; Zuckerman, 2000). This article assumes that most firms rely on predefined categories (the combination of industry and firm size) to define peer groups; however, this may not be the case for all firms. Faulkender and Yang (2010) recently conducted an innovative study that discovered valuable information on the issue of peer selection; however, more research is required to better understand the process. Studying how firms select peers outside the predefined categories is a fruitful area for future research (Porac et al., 1999).

The third question is, which benchmarks do firms use (for example, the median or the 75th percentile of the distribution)? Thus far, no systematic study has been conducted on this topic. This article uses the peer group mean for the benchmark; however, if many firms pay above the mean or median, then average pay will rise even further (DiPrete et al., 2010). This may be an interesting area for study in terms of the self-perception of performance and relative self-worth. Media reports have speculated on this topic. For example, The Economist (2003: 13) asserted, “No board wants to pay the average for the job. The above-average candidate which directors have just selected as CEO, they invariably reason, deserves more. And so bosses’ pay spirals upwards.” Further research on this issue would shed light on executive compensation and the causes of the rise in CEO pay.

Finally, future research should explore whether the findings of this study are specific to the U.S. context or can be generalized to other countries. Although the principles of equity and power relations may be quite general across societies, executive compensation systems are shaped by institutional and cultural differences across societies (Aguilera & Jackson, 2010). In the U.S. context, peer information is readily available because there is an abundance of publicly traded firms that are subject to the SEC’s disclosure requirements; this facilitates peer benchmarking. However, in countries where privately owned firms are more common, peer information is not as readily available. Furthermore, the relationships between the managements and boards vary in different countries, which suggests the existence of cross-national variations in power relationships. In countries where boards or owners have an effective means of monitoring management, or where top managers own a considerable part of a company, a CEO’s tendency to shape the process of pay-setting may be relatively limited. Future research should explore international variations in the role of peer benchmarking in determining executive compensation.