Pay for Performance or Pay Harmony?
Peer Comparison and Compensation in Multi-business Firms
Claudine Gartenberg*
New York University
Julie Wulf**
Harvard University and NBER
October 7, 2015
Abstract:
This study provides evidence that social comparison among employees in similar positions affects wage
setting within firms. We analyze a panel of division manager compensation within 270 firms over a 14year period. Our findings suggest that increases in peer comparison lead to increases in pay co-movement
and decreases in pay-for-performance compensation. We also find evidence that peer comparison affects
productivity, particularly for those managers at the low end of the wage distribution. We make use of
variation in the degree of geographic dispersion of firms in our sample and a 1991-1992 controversy over
executive compensation to develop predictions that isolate the effects of social comparison from omitted
factors that may also affect pay. Taken together, our evidence supports the notion that peer comparison of
pay matters when designing senior manager wage contracts and that this comparison dampens
performance pay and has productivity consequences for firms.
Keywords: Executive Compensation, Pay-for-Performance, Internal Labor Markets, Peer Comparison,
Firm Geography
* NYU Stern School of Business, 40 West 4th Street, New York, NY 10012, cgartenb@stern.nyu.edu; ** Harvard
Business School, Soldiers Field Road, Boston, MA 02163, jwulf@hbs.edu. Acknowledgements: We appreciate
comments from Iwan Barankay, Luis Cabral, Constanca Esteves-Sorenson, Carola Frydman, Denis Gromb, Maria
Guadalupe, Peter Kuhn, Mark Leary, Ian Larkin, Francine Lafontaine, Chris Malloy, Andrew McElheran, Stephan
Meier, Kevin Murphy, Antoinette Schoar, Kelly Shue, Tim Simcoe, Geoff Tate, and Todd Zenger and participants at
the MIT Organizational Economics brownbag, NYU Economics of Strategy summer workshop, Columbia Strategy
Conference, CRES conference at Washington University, SOLE, and seminars at USC, Berkeley, Kellogg,
INSEAD, University of Illinois, MIT, Rochester and Cornell.
1
I. Introduction
Social comparison is one of the most fundamental behaviors among humans. Since Sherif (1936),
Hyman (1942) and particularly Festinger (1954), social scientists have long recognized that humans make
sense of their worth and standing via pervasive comparisons with others. However, despite this insight,
research has only recently begun to focus on social comparison within organizations (Buunk and Gibbons
2007). This newer research has uncovered a range of consequences arising from negative comparisons
within organizations, including increased turnover (Card et al 2012), deception (Gino and Pierce 2010),
theft (Chen and Sandino 2011), and reduced performance (Cohn et al 2014).
Less is known, however, about how social comparison interacts with the organization’s structure
itself. Nickerson and Zenger (2008) highlight the importance of this interaction in their discussion of how
firm boundaries define the referent group against whom employees compare themselves. They describe
situations in which firms that incorporate businesses with large pay differences can seed envy among
employees, resulting in a host of pay and productivity consequences. In a related argument, Larkin, Pierce
and Gino (2014) propose that pay-for-performance, which may successfully incentivize employees at an
individual level, can create greater pay disparity inside firms, which may in turn impose psychological
costs on workers. It follows from this argument that there may be a tradeoff between the incentive effects
of performance pay and effects of peer comparison that arise from unequal pay (Rebitzer and Taylor,
2011).
Despite these theoretical contributions, and likely due to data limitations, there has been
surprisingly limited systematic evidence of these types of wage comparisons within actual organizations.
In a two-firm study of engineers, Zenger (1992) shows that salary changes and turnover are distributed
unevenly throughout the performance distribution and argues the results are consistent with costs
associated with pay comparison between similarly–performing peers (other engineers). Among more
senior employees, Wade, O’Reilly and Pollock (2006) find evidence of upward CEO pay comparison by
managers at various organizational levels as measured by pay co-movement across time. Beyond these
studies, and to the best of our knowledge, no research has examined across-firm and over-time evidence
of how peer comparison between managers in comparable positions within the same organization affects
pay. Furthermore, no research has systematically isolated the predicted tradeoff between the incentive
benefits of performance-pay and internal pay equity or “pay harmony.” 1
In this paper, we aim to fill these gaps by studying patterns of pay co-movement among division
managers within 270 large multi-business firms over a 14-year period. Our results suggest that peer
1
The term harmony derives from the Greek ἁρµονία (harmonía), meaning "joint, agreement, concord.” This term
was introduced to us by a compensation consultant who emphasized how equity considerations regularly lead to
similar compensation packages for senior managers within his client firms.
2
comparison leads to increased pay co-movement while reducing pay-for-performance within firms. We
also find evidence suggesting that it affects performance, particularly at the low end of the wage
distribution where negative peer comparisons predominate.
We focus on division managers for several reasons. First, this role provides us with a rich context
to examine peer comparison within a firm: these managers represent each other’s salient referents as they
occupy roles of similar responsibility and compete for the same pools of resources and positions for
promotion. Furthermore, peer comparison is likely to affect pay in these positions since conditions under
which pay disparities are more easily tolerated, according to equity theory (e.g., Adams 1965), are likely
not met in this context. Specifically, procedural fairness across managers is hard to assess: despite size
and division performance being widely reported, the ratio of division outcomes to manager effort is not
easily knowable, given the broad and discretionary responsibilities of the role. Moreover, other managers
might rightly question the relative difficulties of managing businesses in different industries or
geographies and would therefore question compensation based purely on standard performance
benchmarks.
Aside from being a good context for peer comparison, we also focus on division managers
because of the overall importance of their role in large, multi-business firms. Division managers are the
effective CEOs of their constituent businesses and control the deployment of the majority of firm assets
(Bower and Gilbert 2007). After the CEO, they are the main protagonists in vast literatures about multibusiness firms, including work on influence costs (Milgrom and Roberts 1990), divisional rent-seeking
and internal capital markets (Scharfstein and Stein 2000) and vertical integration (Grossman and Hart
1986, Hart and Moore, 1990). As such, peer comparison within this position may have profound
consequences for firm strategy and performance, per the discussion by Nickerson and Zenger (2008).
We adopt the following empirical strategy. Following Wade et al (2006), we first argue that comovement in pay is consistent with peer comparison; that is, changes in one’s own pay is positively
associated with changes in referent pay. We then replicate their main result that manager pay co-moves
with CEO pay and show that peer pay co-movement is larger than pay co-movement between manager
and CEO. In fact, the degree of peer pay co-movement is about four times that of the co-movement with
CEO pay. One challenge with this initial approach, as with many studies of peer effects, is that pay comovement may also be driven by external factors that could simultaneously affect one’s own pay and
referent pay (Manski 1993). Since it is impossible to observe peer comparison directly in this nonlaboratory context, we instead develop three additional predictions of how peer comparison affects payco-movement and pay-for-performance -- over time and across firms -- that are not easily explained by
these outside forces.
3
First, we appeal to research on referent selection (Goodman 1977; Kulik and Ambrose 1992) to
predict that, as the overall awareness and salience of pay increases during our study timeframe, the
intensity of peer comparison--and hence pay co-movement--will increase within firms (and pay-forperformance will correspondingly decrease). To explore this proposition, we take advantage of a national
debate about executive compensation that unexpectedly and dramatically arose during 1991 and 1992, the
middle of our study period. Second, we predict that peer pay comparison occurs more in geographically
concentrated firms relative to dispersed firms, since concentrated firms do not experience the same
“social chasm” (Tzabbar and Vestal 2015) as do dispersed firms. Lastly, we combine these ideas to
predict that a rise in pay salience has a greater effect within dispersed firms than in concentrated firms
since managers in concentrated firms are likelier to share pay information prior to this rise. Support for
these predictions, as well as a finding of a negative relationship between peer pay co-movement and payfor-performance, provides evidence beyond prior studies that simply considered pay co-movement as
indicative of social comparison.
We also look for a performance impact on dispersed firms. We find that division productivity
increases less after the 1991-1992 pay controversy in dispersed firms than in concentrated firms and that
this effect is most pronounced for those at the low end of the wage distribution of division managers
within the firm. This result is consistent with managers expending less productive effort upon discovering
that their pay is low relative to peers, a potential real cost of social comparison.
This paper makes four contributions to research on peer comparison. First, to the best of our
knowledge, this is the first paper with data that allows the use of a multi-firm, long panel analysis with
position fixed-effects to identify peer comparison within organizations. Moreover, the longitudinal and
cross-sectional variation in our panel allows us to develop tests that further isolate peer comparison
effects from other factors that may simultaneously affect pay. As such, this paper provides empirical
evidence consistent with the proposition that firm boundaries define referent groups that in turn lead to
social comparison (Nickerson and Zenger 2008).
Second, we provide evidence in support of a tradeoff between social comparison and incentive
pay that has not been explored in prior empirical research. As the degree of social comparison increases,
we find an increase in peer pay co-movement and a commensurate decrease in performance pay as
predicted by Rebitzer and Taylor (2011) and Larkin et al (2012). Multi-business firms – the firms in our
sample and generally the most economically significant firms – may be particularly subject to these costs,
since diverse divisions may have different performance profiles. Within these firms, CEOs may find
themselves especially constrained in the degree to which they can incentivize and reward top performing
managers. As such, our results have direct implications for managers involved in designing pay practices.
4
Third, our finding on the negative impact on divisional productivity at the bottom of the wage
distribution provides evidence that this peer comparison – aside from influencing the incentives of some
of the most influential employees of a firm – also appears to have a measurable performance effect. This
finding is consistent with other research on peer comparison that shows that most negative effects are
concentrated among those at the bottom of the distribution (Pierce and Gino 2009; Card et al 2012; John
et al 2014).
Finally, this paper contributes to the ongoing policy debate on the consequences of transparency
and mandatory information disclosure to investors (e.g., Hall, and Murphy, 2003; Greenstone, Oyer and
Vissing-Jorgensen, 2006; and Hertzberg, Liberti, and Paravisini, 2011), and contributes to the literature
on pay secrecy (e.g., Lawler, 1965; Bewley, 1999) which suggests that employers may have an incentive
to not disclose pay.
The remainder of the paper is organized as follows. Section II includes our theoretical discussion
and development of five hypotheses. Section III describes our empirical strategy and the data. Section IV
outlines and discusses our results. Section V discusses study limitations and implications for theory and
practice.
II. Theoretical Background: Pay-for-Performance, CEO and Peer Pay Comparison
Within economics and strategic management, the principal-agent model (Jensen and Meckling,
1976) has traditionally been the workhorse model of how firms set executive and employee pay (Larkin et
al, 2012; Murphy 2012). The main theoretical implication for firms is that, to elicit optimal effort from
agents, principals should link pay to performance.
Wage contracts are typically assumed to have the following linear form, in which payperformance sensitivity is defined as b1 and the optimal choice of b1 depends on the underlying model:
(1) 𝑤 = 𝑎 + 𝑏! 𝑦
where 𝑤 is the wage and 𝑦 is the measurable output (or performance) of the agent.
However, even within this tradition, there is increased acknowledgement that the process is more
social than this model suggests. An early mention of wage fairness dates to Hicks (1963) in The Theory of
Wages (pg. 317):
“The labor market is…a very special kind of market which is likely to develop ‘social’
as well as purely economic aspects…For the purely economic correspondence between wages
paid to a particular worker and his value to the employer is not a sufficient condition of
efficiency: it is also necessary that there should not be strong feelings of injustice about the
relative treatment of employees since these would diminish the efficiency of the team.”
More recent research has focused on social comparison within organizations as a key psychological
factor in determining employee happiness and performance outcomes. Theoretically, Akerlof and Yellen
5
(1990), Mui (1995), Nickerson and Zenger (2008) and Larkin et al (2012) have focused on various
negative effects of envy resulting from workplace social comparison, either implicitly or explicitly
resulting from a pay-for-(individual) performance compensation policy. Empirically, social comparison
has been linked to competitive behavior (Johnson 2012), reduced cooperation (Armstrong and Collopy
1996; Poortvliet 2012), theft and deception (Gino and Pierce 2009, 2010; Chen and Sandino 2012; Moran
and Schweitzer 2008; Argo, White and Dahl 2006; Edelman and Larkin 2014) and lower satisfaction and
increased turnover intentions (Card et al 2012).
Since Festinger’s (1954) early insight that people engage in rampant social comparison, one of
the key questions in this research has been: who is the salient referent? One line of research has focused
on upward comparison (Taylor and Lobel 1989; Collins 1996) and translated this upward orientation to
referents that are above the employee in the vertical hierarchy, particularly the CEO. Of particular
relevance to our study, Wade et al (2006) attributes co-movement in employee salaries with CEO salaries
to this upward comparison. They find a strong association between employee compensation and CEO
over- and under-payment, particularly for more senior employees, and increased turnover when this
association breaks down (i.e., when CEOs are overpaid relative to employees). Siegel and Hambrick
(2005) similarly find evidence of CEO pay comparison within top management groups (senior executives
that are firm officers), while Carpenter and Sanders (2002) find top management team pay to be related to
CEO pay, which they attribute, in part, to equity considerations.
Despite this evidence of the CEO as referent, Festinger’s original proposition is that the most
likely referents would be selected from “similar others.” This proposition has been subsequently
supported by a number of empirical studies of referent selection (Suls and Miller 1977; Lamertz 2002;
Lawrence 2006). Examining wages explicitly as outcome, several studies have shown evidence for this
peer pay comparison, albeit still using CEOs as referents (O’Reilly et al 1988; Shue 2013).
To provide a structure for our analysis, we expand equation 1 into the following linear-in-means
wage contract:
2 𝑤 = 𝑎 + 𝑏! 𝑦 + 𝑏! 𝑤!"# + 𝑏! 𝑤!!
Where 𝑤!"# is the CEO pay and 𝑤!! is the average pay of one’s peers.
Throughout the paper and consistent with the executive compensation literature, we refer to b1 as
pay-performance sensitivity (PPS), the degree to which division manager pay changes with changes in
division performance. We label b2 as CEO pay sensitivity, the degree to which division manager pay
changes with changes in CEO pay. Finally, we refer to b3 as pay-referent sensitivity (PRS), the degree to
which division manager pay changes with changes in the average pay of all other division managers
within the firm. We consider how both the level of and changes in PRS over time and between firms
provide evidence of peer pay comparison within organizations.
6
Using the wage model in equation 2 is advantageous for a number of reasons. First, it is closely
related to the model estimated by Wade et al (2006), thus allowing us to build on their findings when
interpreting our results.2 Second, it allows us to explore the tradeoff between performance pay sensitivity
(PPS) and pay co-movement (PRS) that is an important prediction of recent research incorporating social
comparison into agency models (e.g., Larkin, et al 2012, and Rebitzer and Taylor 2011) and has not been
explored in prior studies.3
From equation 2 and the discussion above, we develop several predictions for pay setting. First,
consistent with prior literature on pay-for-performance, we expect 𝑏! > 0, meaning that manager pay
should co-vary with his/her division’s performance. Second, following prior research on CEO pay
comparison within firms, we expect that 𝑏! > 0, indicating that division manager pay will co-move with
CEO pay. Similarly, if peer pay comparison affects wages, then we expect 𝑏! > 0, that is, we expect to
find positive co-movement of division manager pay. Our first hypothesis is, therefore:
H1: Holding pay performance sensitivity (PPS) and CEO pay sensitivity constant, increases in
manager pay are positively associated with increases in peer pay; that is, pay referent sensitivity (PRS)
will be greater than zero.
II.1 Pay Salience
As noted by Manski (1993), pay co-movement may result from other factors aside from simple
peer comparison, including common market conditions or events that are not captured in controls
available to the researcher. We therefore now develop three additional hypotheses about division manager
pay co-movement that are most consistent with the presence of peer comparison, rather than these other
factors.
First, we focus on the salience of pay and the relationship to peer comparison of wages. We
predict that an increase in the overall awareness of pay among employees will induce relatively more pay
comparison. The intuition underlying this prediction comes from Goodman (1977): the greater the
salience of whatever characteristic forms the basis of comparison (pay, in our case), the greater the search
for information about that characteristic among referents and the greater the degree of comparison overall.
In a similar vein, Kulik and Ambrose (1992) isolate two main factors that determine the intensity of social
comparison: i) availability of information and ii) the relative attractiveness of the comparison. They
2
Wade et al (2006) use a residual model, which is equivalent in intent to our model. In Appendix Table A4, we
replicate Table 2 using their residual approach and obtain the same results.
3
In a related paper, Zenger (1992) focuses explicitly on the relationship between salary increases and individual
performance outcomes, however does not consider pay co-movement.
7
discuss how situational or environmental changes can influence the attractiveness of comparison and
therefore increase the motivation of individuals to seek out information about referents. Therefore,
H2a: Holding pay performance sensitivity (PPS) and CEO pay comparison constant, peer
referent sensitivity (PRS) increases as the salience of pay increases.
Next, we consider the relationship between peer comparison and performance-based pay. Larkin
et al (2012) develop a theory of compensation that incorporates effects of social comparison between
peers in an agency- theoretical framework. They argue that social comparison costs reduce the
effectiveness of individual performance-based pay because of the psychological costs of introducing pay
inequity into an organization. This perceived inequity offsets the effort benefits of performance-based pay
and introduces additional costs from sabotage, costs that are not typically accounted for in agency theory.
One testable prediction is that performance-based pay will be observed less often and with lower power in
the presence of social comparisons and that team-based compensation can reduce the associated costs. In
our context, team-based compensation implies pay co-movement or, equivalently, increases pay referent
sensitivity (PRS). It follows that there is a tradeoff between the incentive effects of performance pay and
effects of peer comparison that arise from unequal pay. Given our discussion above that an increase in
pay salience increases the intensity of peer comparison, it follows, therefore, that it should have an
opposite effect on pay-performance sensitivity:
H2b: Holding peer referent sensitivity (PRS) and CEO pay comparison constant, pay
performance sensitivity (PPS) decreases as the salience of pay increases.
II. 2 Geographic Concentration
Physical proximity has been shown to have a considerable effect on social comparison. As noted
by Tzabbar and Vestal (2015), geographic distance creates a “social chasm.” This chasm is associated
with fewer planned and chance meetings, reducing in-person interactions (Kraut et al 2002). Instead,
these interactions are replaced by virtual meetings mediated by technology such as telephones, email and
videoconferencing that has been shown to lead to more perfunctory sharing of information (Hinds and
Bailey 2003). Relatedly, physical distance is also associated with weaker interpersonal relationships and
friendships and lower overall levels of trust (Hinds and Bailey 2003).
Both of these features of physical distance -- less aggregate time spent together and lower levels
of trust – lead to predictions of reduced social comparison. Fewer in-person meetings logically lead to
fewer word-of-mouth opportunities (Hong et al 2000; Monge and Contractor 2000) and, specifically,
8
fewer opportunities to share pay information. Beyond this, lower levels of trust reduce the likelihood that,
even when the opportunity does arise, managers will be inclined to exchange this sensitive information.
Lastly, in firms with frequent meetings, the salience of nearby managers increases since managers are
reminded of their peers more often, which in turn increases the degree to which individuals will gather
reference information on their peers for comparison purposes (Kulik and Ambrose 1992; Garcia et al
2006).
Multi-business firms vary considerably in their degree of geographic concentration. Within some
firms, divisions are all co-located within the same state, while in others divisions are distributed
throughout the country or internationally, at substantial physical distance to each other. Within these latter
firms, according to the logic above, there should be less sharing of pay information. Consider a case
described to us by the Chief Development Officer (CDO) of a large U.S. investment bank during the early
to mid-1990s. On one day, near the end of the fiscal year, two investments bankers working at the New
York City headquarters were individually informed by the CDO about their own annual compensation
(bonuses and proposed salary increases). The following day, the lower-paid banker was in the office of
the CDO to ask: “Why am I paid less?” While the CDO told us that this dynamic was the norm in
headquarters, it was much less common with MDs working in the firm’s satellite offices, such as Los
Angeles or Tokyo.4
Given this discussion, we expect the degree of peer pay comparison to vary by the geographic
concentration of a firm leading to the following related hypotheses.
H3a: Holding pay performance sensitivity (PPS) and CEO pay comparison constant, peer
referent sensitivity (PRS) will be lower in geographically dispersed firms than in concentrated firms.
H3b: Holding peer referent sensitivity (PRS) and CEO pay comparison constant, pay
performance sensitivity (PPS) will be greater in geographically dispersed firms than in concentrated
firms.
Furthermore, these effects should differ by the degree of pay saliency. Since proximate managers
already exchange pay information due to low sharing costs, distant managers should learn relatively more
when pay becomes more generally salient. This differential effect could arise because the attractiveness of
knowing peer pay itself increases (per Kulik and Ambrose 1992) or because the costs of acquiring
information decrease. For example, the link between compensation levels and social status may
4
More broadly, the notion that pay is more likely to be shared between managers that are geographically proximate
was confirmed based on interviews with compensation consultants and senior executives.
9
strengthen as a result of the broader attention on pay, inducing distant managers to seek out pay
information. More simply, distant managers may exert more effort to acquire information when they are
primed more often on pay when reading the newspaper or listening to the radio (whereas proximate
managers are already primed by seeing their peers and word-of-mouth effects). Also, the cost of acquiring
information may fall as information frictions are reduced. This fall could occur because social taboos on
discussing pay fall, which in turn eases communication between managers with fewer chance meetings
and weaker interpersonal relationships - or simply because the frequency of discussing pay on aggregate
increases, thereby increasing word-of-mouth effects even for managers at a distance.
We therefore predict the following differences-in-differences results, the strictest of our four tests
for peer effects:
H4a: Holding pay performance sensitivity (PPS) and CEO pay comparison constant, peer
referent sensitivity (PRS) will increase more in geographically dispersed firms than in concentrated firms
as a result of greater pay salience.
H4b: Holding peer referent sensitivity (PRS) and CEO pay comparison constant, pay
performance sensitivity (PPS) will decrease more in geographically dispersed firms than in concentrated
firms as a result of greater pay salience.
II.3 The 1991-1992 Executive Pay Controversy
We now turn to a discussion of an event that we argue led to increased salience of pay both for
senior executives and employees more generally. During 1991 and 1992, public and media attention on
executive pay became an “explosive issue.”5 The attention from the 1992 presidential candidates was
supplemented by high profile books and media articles on CEO pay:
“CEO pay did not really become a public “issue” until 1991. Feature stories on CEO pay aired on
the nightly news broadcasts of the three major networks in the spring of 1991, and CNN, 60
Minutes and Nightline devoted segments to CEO pay. The controversy heightened with the
November 1991 introduction of Graef Crystal’s (1991) expose on CEO pay, In Search of Excess,
and exploded following President George Bush’s ill-timed pilgrimage to Japan in January 1992,
accompanied by an entourage of highly paid US executives.” (Murphy, 1999, pg. 50)
In October 1992, in response to the heightened visibility of CEO pay, the Securities and
Exchange Commission (SEC) “announced sweeping new rules affecting the disclosure of top executive
compensation in the annual proxy statement” of publicly-traded firms (Murphy, 1999, pg. 50) particularly
performance-based pay. To validate the timing of heightened media attention, in Figure 2 Panel A, we
5
Wall Street Journal, April 22, 1992.
10
plot the number of articles appearing in the Wall Street Journal between 1986 and 1999 with any of the
following phrases: “executive compensation,” “executive pay,” “CEO compensation,” and “CEO pay.”
As the figure illustrates, the peak period of media coverage was in 1992, with the run-up beginning in the
prior year, a pattern that is similar to that documented in Kuhnen and Niessen (2012).6
We offer that this national debate on compensation significantly increased the salience of pay, not
only of CEOs, but also of employees more generally. The logic behind this connection is that CEO pay is
judged to be high only relative to employee pay, a version of the upward comparison argument offered by
Wade et al (2006). Therefore, as the debate about CEO pay increased, focus on employee pay as a general
topic also increased. To demonstrate this broader interest, we plot the number of articles with any of the
following phrases: “employee pay,” “salaries,” “compensation” or “wages.” We find a definite and
sustained increase in the number of articles with these phrases from 1992 onward. This time trend
supports the view that the post-1992 period was marked by a heightened national interest in compensation
as a general topic and wage benchmarking more specifically, both at the executive and employee level.
<< Insert Figure 1 about here >>
II.4 Productivity Implications of Peer Wage Comparisons
Our prior discussion focused on the impact of pay comparison on pay itself. Here we consider a
separate outcome of strategic importance to firms: the effect on division productivity. While there is a
growing body of evidence suggesting that pay comparison affects one’s happiness, the theoretical
prediction regarding the impact on actual behavior is ambiguous. On the one hand, when a manager
discovers that she is paid less than her peers, she may respond negatively by reducing her effort, exiting
the organization or engaging in a range of retributive actions in response. Nickerson and Zenger (2008)
related stories from various companies, including several situations in which key employees left in
response to compressed pay schedules and another in which management declined to invest productivityenhancing technology to avoid creating pay disparities among employees. Related to this, Wade et al
(2006) find evidence of increased turnover within firms in which the CEOs appear overpaid relative to the
employees. Edelman and Larkin (2014), Gino and Pierce (2009), Moran and Schweitzer (2008) each link
social comparison to employee deception, while Greenberg (1988), Cohn et al (2014), Ockenfels et al
(2014), Nosenzo (2013) and Gacher and Thoni (2010) find evidence of reduced employee effort,
particularly at the bottom end of the spectrum and when inequality is high. On the other hand, a negative
6
Kuhnen and Niessen (2012) employ a more detailed approach to measure public opinion about executive
compensation by analyzing all newspaper articles and using linguistic software to capture the negativity of press
coverage of CEO pay.
11
pay comparison could have the opposite effect: an agent might work harder because she is motivated to
improve her standing.
For managers paid more than peers, the response is similarly unclear. Top paid managers might
slack off because of limited opportunities for raises or become alternately more motivated when they
discover that they are “stars.” Finally, pay disclosure may not affect manager effort at all. Peer pay may
simply be used to negotiate better pay with senior management and have no effect on effort, particularly if
wages are adjusted appropriately in response. Charness and Kuhn (2007) and Bartlink and von Siemens
(2011) find evidence of this null effect.
We therefore approach the direction of the association between peer pay comparison and
productivity agnostically, with Charness and Kuhn (2007) and Bartlink and von Siemens (2011)
supporting a null finding:
H5: Greater peer comparison within firms will affect division productivity, particularly for
managers at the low end of the pay distribution.
III. Empirical Strategy
III.1 Wage Equation
Equation 2, translated into a form that can be empirically tested, becomes a standard wage model
expanded into a modified linear-in-means specification with terms to capture both CEO and peer pay:
(3) 𝑤!" = 𝛼 + 𝛽! 𝑠!" + 𝛽! 𝑤!"# + 𝛽! 𝑤!!" + 𝐷′!" 𝛽!" + 𝐹′!" 𝛽!" + 𝜂! + 𝑑! + 𝜖!"
Here, d indexes divisions within a firm at time t. A division manager’s (log) wage in period t is a function
of division performance, 𝑠!" , log CEO pay (𝑤!"# ) and peer referent pay (𝑤!!" ), defined as the average
(log) wage of all other division managers within the firm, excluding own wage.7 A vector of both
division, 𝐷′!" , and firm characteristics, 𝐹′𝑑𝑡 , are included as controls. Division performance is
measured as log division sales, which can be interpreted as sales growth in our fixed effects
specifications. We use two measures of firm performance: return on assets (%) and log firm sales, also
interpreted as changes in those measures in our fixed effect specifications. Also included as controls are
7
Linear-in-means specifications typically use the average of the whole group, with the underlying assumption that
any one participant has a small marginal impact on the whole group. Because that assumption is not valid in our
context, we use the average of all other division managers to calculate our mean pay variables. This definition
complicates the interpretation of the standard errors because of the potential for correlated errors between
observations within a given firm-year. In addition to clustering by firm and using division fixed effects for our
primary specifications, we also run simulations to test whether this calculation can introduce spurious correlations
and significance levels into the data. We found no such biases or efficiency distortions.
12
division manager characteristics, including tenure in the position, the number of levels between the
managers and the CEO and whether the manager is also an officer of the firm. Firm level controls include
the number of divisions within the firm and three measures of divisional relatedness to control for
coordination between units.8
We also include year fixed effects, 𝑑! , and cluster standard errors at the firm level. Our basic
specification includes division fixed effects, 𝜂! , which control for unobserved, time-invariant, division
and firm heterogeneity and allow us to interpret coefficients as the change in manager pay with changes
in the independent variables. We also include cross-sectional and firm fixed-effects specifications in our
differences-in-differences analyses to examine similarities in pay levels over time, both across all firms
and within firms.
The main coefficient of interest in this specification is 𝛽! , which captures pay-referent sensitivity
(PRS). We compare estimates of 𝛽! across various specifications and subsamples to test Hypotheses H1H4 as well as changes in 𝛽! to test changes in PPS. Our other coefficient of interest, 𝛽! , captures CEO
pay sensitivity.
III. 2. Data
The primary dataset used in this study includes a panel of 270 publicly-traded U.S. firms over the
years 1986-1999, spanning a number of industries.9 The data are collected from a confidential
compensation survey conducted by Hewitt Associates, a leading human resources consulting firm
specializing in executive compensation and benefits.10 In this paper, we focus on the most-senior position
in a division, which is defined in the survey as “the lowest level of profit center responsibility for a
business unit that engineers, manufactures and sells its own products.” One advantage of this dataset is
that it allows us to identify changes in pay within division manager positions over a 14-year period,
substantially longer than prior research on pay comparison, particularly on employees, which has either
been laboratory-based, cross-sectional or over shorter panels of two to five years.
The data for each position include all components of compensation, including salary, bonus paid,
and long term compensation, in the form of ex ante valuation of annual grants of restricted stock and
stock options (computed by Hewitt) and other forms of long-term incentives. An observation in the
dataset is a division managerial position within a firm in a year. To ensure consistency in matching these
8
These controls include a measure of the degree to which the internal capital markets reallocate capital between
units, and two measures of diversification between units: a weighted measure of horizontal and vertical integration
of the firm, as calculated by the SIC codes of the various units and an “entropy” measure that weights how dispersed
the firms assets are across diverse industries.
9
See Appendix Figure 1 for the industry distribution.
10
For a detailed description of the survey, please refer to Appendix A. Based on several analyses, we conclude that
the survey sample is most representative of Fortune 500 firms.
13
positions across firms, the survey provides benchmark position descriptions and collects additional data
for each position, leading to a dataset rich in position characteristics. As a result, in addition to data on all
aspects of compensation for multiple division manager positions, the dataset includes division-specific
characteristics, such as: job title; division sales; number of employees under the position’s jurisdiction;
industry of operation; geographic state of location; number of positions between the division manager
position and the CEO in the organizational hierarchy (division depth); an indicator of the incumbent’s
status as a corporate officer; and the manager’s tenure in the position.
We supplemented the above data with financial and headquarters location information from
Compustat and firm-level information technology investments from Harte-Hanks. Finally, we construct a
number of variables that are used as controls and that we will describe in the results section.
In Table 1 (Panel A), we present descriptive statistics for the firms and divisions in the sample.
While the dataset includes more than 270 firms, the exact number varies over the period, as firms enter
and exit as survey participants. The firms in the sample are large, well-established and profitable, with
average size of sales of $8.5 billion, market capitalization of $9.4 billion, 44,000 employees and return on
assets of five percent. The average number of divisions reported in the survey for the sample firms is 4.6.
Next, turning to divisional statistics, the mean size of divisions is $752 million in sales and approximately
3000 employees. The average tenure in position is 42 months; approximately 23 percent of the division
managers are corporate officers; and there are 1.4 positions between the CEO and division managers, on
average. Average annual division manager pay is $300K (base salary plus bonus paid), and $460K (total
compensation defined as the sum of salary, bonus paid, and the ex ante valuation of annual grants of
restricted stock, stock options and other long-term incentives). Finally, the sample firms span many
industrial sectors of the economy, with some concentration in the food, paper, chemical, machinery,
electrical, transportation equipment, instrumentation, communications and utilities industries.
<< Insert Table 1 about here >>
For our tests of H3 and H4 that use differences in the geographic dispersion of firms, we use the
information on division state of location from the Hewitt dataset and headquarters’ state of location from
Compustat. We construct a firm-level measure of geographic dispersion by computing the proportion of
divisions in the same state as headquarters (mean of 0.48 for the sample) and then construct quintiles of
geographic concentration using this firm measure and assign values ranging from 1 (least concentrated) to
5 (most concentrated). In Table 1 (Panel B), we split the sample into concentrated firms (4th and 5th
quintiles) and dispersed firms (1st and 2nd quintiles) and report key statistics for these two subsamples. We
14
exclude the middle quintile in our analyses that, while reducing the power of the tests, enables us to make
sharper comparisons between firms at the geographic extremes.
As can be seen from this table, firms in these two categories are roughly similar in several
characteristics. The biggest difference is that concentrated firms have fewer and bigger divisions than
dispersed firms, based on means. However, in comparing median sales and employees, division size is
generally comparable across firm geography.11
IV. Results
Our results are structured as follows: we begin by showing the presence of pay-performance
sensitivity (PPS) to demonstrate that our data behave in a way that is consistent with the basic agency
model. We then document evidence consistent with the CEO pay comparison as reported by Wade et al
(2006). Distinct from any prior literature, we then show that, holding both PPS and CEO pay comparison
constant, we find evidence of peer pay comparison, in the form of positive pay referent sensitivity (PRS)
or co-movement in division manager pay, thus confirming H1 (Table 2).12 We then test H2 by estimating
differences in PRS and PPS around the 1991-1992 pay controversy (Table 3). We then test H3 and H4 by
replicating the models from H2 in Table 3 on dispersed and concentrated firms separately and
documenting that PRS changed differently between those subsamples (Table 4 and Figure 3). We then
analyze the effect of pay disclosure on division productivity (H5, Table 5). We use two pay measures
throughout our analysis: base salary plus bonus, following Wade et al (2006), and also total
compensation.
IV.A. Pay-Performance Sensitivity (PPS), CEO Pay Sensitivity and Pay Referent Sensitivity (PRS)
We begin by estimating a standard wage equation for division manager positions (equation 3). In
Table 2, Column 1, we regress the logarithm of the base salary plus bonus on division and firm
performance, while controlling for firms and division manager characteristics described in Section III.
We find evidence that firms link pay to both division and firm performance. The coefficient on
division sales represents the pay-division performance sensitivity, which is positive and significant for
both pay measures. The coefficients on firm performance measures represent the pay-firm performance
sensitivity and are positive and significant. The log-log specification means that PPS coefficient estimate
of 0.09 on division performance can be interpreted as follows: division manager pay increases by 9% for
a 100% increase in division sales. The estimates on our controls are as we might expect: pay is higher for
11
To address concerns about unobserved differences in dispersed and concentrated firms, we replicate Tables 2-4
using propensity score methods and find similar results (reported in Tables A1-A3).
12
Table A4 reproduces this analysis using Wade et al (2006) residual approach, which is conceptually a similar empirical
approach. Using this approach, we find equivalent results. 15
managers with longer tenure in the job, managers that are officers, and managers closer to the top of the
organizational hierarchy (lower depth).
Column 2 now includes the log of CEO base salary plus bonus pay. The coefficient of this term is
positive and significant, replicating the finding from Wade et al (2006) that lower level manager pay comoves with CEO pay. In fact, our coefficient estimates on CEO pay are economically similar to the
magnitude calculated in Wade et al (2006), despite the different time periods and controls in the two
analyses. We find that, as CEO pay doubles, division manager pay increases by 17 percent, while they
find between 11 and 25 percent.13
The coefficients on firm performance (return on assets and revenues) decrease substantially, both
in magnitude and significance, between Columns 1 and 2, while the PPS coefficient on division sales
does not. This difference indicates that the link between division manager pay and firm performance is
largely explained by pay co-movement between division managers and CEOs, while PPS remains driven
by divisional (individual) performance.
Next, we turn to analyzing the data in light of our first hypothesis. Column 3 estimates PRS by
including the log of the average of all other division managers’ base salary plus bonus within the firm for
that year. The PRS coefficient estimate is positive and both economically and statistically significant: a
doubling of peer pay is associated with a 39% increase in manager pay. This result supports H1 - that
PRS is greater than zero, controlling for performance and CEO pay. In fact, PRS appears to be more than
four times greater than the CEO pay sensitivity: a one standard deviation increase in CEO pay is
associated with a 5.3% increase in pay, while a one standard deviation increase in peer pay is associated
with a 23.6% increase.
Columns 4-6 replicate these models, substituting total compensation for base salary plus bonus. If
anything, the link between peer pay strengthens in these models. Overall, Table 2 supports H1: division
manager pay co-moves with peer referent pay, controlling for performance and CEO pay.
<< Insert Table 2 about here >>
IV.B. Pay, Geography, and the 1991-1992 Pay Controversy
As discussed earlier, one challenge is that this analysis may capture events that simultaneously
affect the pay of all division managers and the CEO that are not captured in the controls available to the
researcher. This challenge is common to research on peer effects and we attempt to address it here with
our tests of hypotheses H2-H4. Tables 3 and 4 show results of these tests.
13
Using their Table 3A, Model 4, which we approximate to be the closest level as our division managers in their
sample.
16
In Table 3, we test H2 by estimating differences in PPS, CEO pay sensitivity and PRS before and
after the pay controversy for all firms in our sample. Columns 1-3 use base plus bonus as the
compensation measure, while Columns 4-6 use total compensation. We begin with a cross-sectional
specification (Columns 1 and 4), then add firm fixed effects (Columns 2 and 5) and then divisional fixed
effects in place of firm fixed effects (Columns 3 and 6). The coefficients on Log referent pay*Post 1992,
our measure of the change in PRS after the pay controversy, are positive in all six models and significant
in four. This positive estimate indicates that pay co-moved more with peer pay after the 1992 pay
controversy. Conversely, PPS, the coefficient on Log division sales*Post 1992, is negative and significant
in five of the six specifications, supporting the notion that pay-for-performance decreases as social
comparison increases. Moreover, the magnitude of these effects are quite substantial. For example, in
Column 5, an increase in PRS of 0.08 represents a 30% increase on a baseline PRS of 0.27, and a PPS
decrease of 0.03 is an 18% decrease from the baseline value of 0.17.
These results support H2a and H2b and are consistent with the tradeoff as argued by Larkin et al
(2012) and Rebitzer and Taylor (2011).
We also find that CEO pay sensitivity generally decreases (the coefficient on Log CEO pay*Post
1992 is negative and generally significant). While this result may seem counterintuitive at first, given that
the controversy was focused on CEO pay, recall that one of the outcomes during this period was a large
escalation of CEO pay, likely enabled by external benchmarking of CEO peer pay via compensation
consultants (Murphy 1999). As such, any increase in CEO comparison within firms during this time could
plausibly be swamped by pay dynamics in the CEO market.
<< Insert Table 3 about here >>
Table 4 tests H3 and H4. Here, we reproduce the analysis of Table 3 in subsamples of dispersed
(Panel A) and concentrated firms (Panel B) and measure the changes in PRS, PPS and CEO pay
sensitivity around the 1991-1992 pay controversy. First, it is important to note that in both panels, and
confirming H1, we find that both measures of division manager pay are sensitive to referent pay (PRS),
CEO pay and division performance (PPS). We also see support for H2, in that Log referent Pay*Post
1992 is generally positive while Log division sales*Post 1992 is generally negative in both panels,
although with different magnitudes and statistical significances, which we discuss below in the context of
H4.
Next, we turn to discussing the results relevant to H3. We find support for H3a: the coefficient
estimates on Log referent pay are larger in Panel B for concentrated firms than they are in Panel A for
dispersed firms, indicated greater pay co-movement. For example, PRS for base plus bonus with division
17
fixed effects (Column 3) is 0.49 for concentrated firms (Panel B) and only 0.25 for dispersed firms (Panel
A). Similarly, PRS for total compensation in the same specification (Column 6) is 0.51 for concentrated
firms (Panel B) and only 0.34 for dispersed firms (Panel A). H3b similarly is supported by the analysis in
this table. Using Columns 3 and 6 again, the coefficient estimates on Log division sales is 0.06 and 0.07
for concentrated firms (Panel B) and 0.11 and 0.12 for dispersed firms (Panel A), indicating greater payfor-performance sensitivity baseline levels in dispersed firms.
Finally, we examine H4, the differences in how PRS and PPS change after the 1991-1992 pay
controversy in dispersed and concentrated firms. In concentrated firms (Panel B), while the coefficient
estimates on Log referent pay*Post 1992 are positive, they are generally insignificant or marginally
insignificant and of low magnitude. Similarly, while the coefficients on Log division sales*Post 1992 are
mostly negative, they are insignificant and close to zero. As such, concentrated firms do not respond to
increases in the public attention to pay. In contrast, pay practices in dispersed firms significantly differ
after 1992 (Panel A). The coefficient on Log referent pay*Post 1992 is positive and significant (and
economically meaningful) for both pay measures across all pay specifications and the coefficient on Log
division sales*Post 1992 is negative and much larger in magnitude than the estimates for concentrated
firms in Panel B and generally significant.
<< Insert Table 4 about here >>
In Figure 2, we plot PRS and PPS to explore the dynamics of pay changes over time to see if the
changes are consistent with the timing of the 1991-1992 pay controversy. We focus on the dispersed firms
and include a set of dummy variables representing different time periods. Figure 3 shows a clear increase
in PRS beginning in 1991 and 1992 and a drop in PPS during the same period for both measures of pay.
<< Figure 2 about here >>
Taken together, this evidence is consistent with greater pay salience intensifying peer pay
comparison across division managers (H2a) and reducing pay for performance (H2b), primarily in
dispersed firms (H4a and H4b). It is also consistent with H3, that dispersed firms structure pay taking into
account less peer comparison with a lower baseline level of pay co-movement and a greater importance of
pay-for-performance (H3a and H3b).
IV. C Peer Comparison and Division Productivity
18
In this section we ask: Do we see evidence consistent with peer pay comparison affecting
performance? In the prior analyses, we argue that peer comparison affects pay setting in firms. Here, we
explore whether peer comparison has a productivity impact within firms. If peer comparison exists
among division managers and that it increased after the 1991-1992 pay controversy within dispersed
firms, then we might expect these divisions to experience some form of performance impact during this
period.
In this section, we explore the productivity response of divisions within dispersed and
concentrated firms around the period of the 1991-1992 pay controversy. To measure performance, we use
division productivity (defined as sales/employee). We regress the log of division productivity in period t
on prior-period indicator variables for whether the firm is dispersed, whether the year is after 1992, and
an interaction term between the two, as well as firm and year fixed effects.14 Results are shown in Table
5, Panel A. We find a negative and statistically significant coefficient on the interaction term—evidence
that is consistent with divisions in dispersed firms exhibiting reduced productivity after 1992 relative to
divisions in concentrated firms. The magnitude of the effect is significant. For example, the coefficient 0.1483 represents 30% of a “within” standard deviation of productivity. The coefficient is larger with a
higher level of significance when we restrict the sample to 1990-1997, a smaller window that excludes
outlying years.
In Table 5 Panel B, we use the restricted sample to evaluate whether divisional productivity
differs by whether the manager is below or above the mean salary in the year prior to the year in which
productivity is measured. We find that the division managers at the bottom end of the wage distribution
largely drive the lower productivity in the dispersed firms (Column 1). In fact, when we split the sample
into above- and below-mean division managers, we find no productivity differences between dispersed
and concentrated firms among division managers paid above-mean wages (Column 2), while we see large
differences within the below-mean sample (Column 3). This result indicates that the relative change in
productivity between dispersed and concentrated firms after 1992 is driven by a relative drop in
productivity among the low-wage earners in dispersed firms.
<< Insert Table 5 about here >>
These results suggest that division productivity was negatively impacted after 1992 within the
firms most affected by increased pay salience, particularly for managers at the low end of the pay scale.
While there may be many alternative explanations, the evidence is consistent with managers reducing
productive effort when they discover they are paid less than their peers.
14
We use prior-period indicator variables in this analysis because we assume that the effort response lags realized
pay by one period. Note that the sample size is reduced by approximately 50 percent because of this restriction.
19
V. Discussion
In this study, we find evidence consistent with pay comparison between senior divisional
managers inside multi-business firms. Our underlying approach is to measure peer and CEO pay comovement and pay-for-performance sensitivity during our fourteen year panel using division manager
fixed effects. We then exploit variation in geographic dispersion within firms and a 1991-1992 public
controversy about executive pay to separate peer effects from potential other factors that could drive pay
co-movement.
We present the results of one fixed effects and two difference-in-difference analyses that, taken
together, support the argument that firms’ pay policies respond to peer comparison by increasing peer pay
co-movement and decreasing pay-for-performance sensitivity, consistent with predictions by Rebitzer and
Taylor (2011) and Larkin et al (2012). We find positive pay referent sensitivity (PRS), even after
controlling for division and firm performance and CEO pay. We also find that PRS increases and PPS
decreases after the 1991-1992 pay controversy, and that PRS is lower and PPS higher in geographically
dispersed firms. Most strongly, we find that the changes in PRS and PPS are driven almost exclusively by
geographically dispersed firms and that these changes occur during and immediately following the
controversy. Finally, we find that division productivity in dispersed firms increases less after the pay
controversy relative to productivity in concentrated firms and that this effect is driven by managers at the
low end of the wage distribution. Altogether, our findings suggest that wage comparisons between peers
within firms and concerns for “pay harmony” affect firms’ policies on setting pay for executives. The
results are also consistent with principals facing a tradeoff between the incentive effects of performance
pay and the effects of peer comparison that arise from unequal pay.
V.1 Limitations
The biggest challenge with our study, as with most studies of peer effects, is that we do not
actually observe social comparison occurring in our firms, but instead have to infer it from pay patterns
across firms and time periods. One concern with our tests using the 1991-1992 pay controversy and
geographic variation between firms is that other differences between firms or changes during the time
period are causing the same pay patterns in our data. We cannot rule this out entirely. Ultimately,
however, any plausible alternative must explain both the increase in PRS and decrease in PPS in
dispersed, but not concentrated firms and specifically during the period immediately around 1992. This
bar is fairly high. In unreported regressions, we test the following candidate explanations, including i)
differential technology adoption between concentrated and dispersed firms, ii) differences in manager
rotations between concentrated and dispersed firms, iii) industry time trends that are related to differences
20
in geographic concentration, iv) differential recovery from the 1990 recession across dispersed and
concentrated firms, v) different time trends by local geography also related to differences in concentration
and vi) differences in industry time trends. We find no support for any of these explanations and, in sum,
we do not find compelling evidence that these or other alternative explanations are driving our results.
V.2 Implications for Theory and Practice
Our findings are relevant for both managerial practice and for policy. In practice, firms’ internal
pay structures are commonly based on systems that specify pay ranges by job and level (e.g., Hay points)
(Baron and Kreps, 1999) to ensure some semblance of equity in pay. Compensation consultants that
specialize in designing executive pay emphasize the importance of “pay harmony” among managers
within a firm. However, this effort to maintain harmony among peers is offset by an increased emphasis
on performance-based executive pay over the past several decades (e.g., Frydman and Jenter 2010;
Murphy 2012). Consultants cite the critical tradeoff in structuring pay between the incentive effects of a
strong pay-for-performance compensation system and the effects of peer comparison within firms – a
shared theme with recent research (Larkin et al 2012). Our paper finds evidence for this tradeoff, which
suggests that comparison costs do in fact limit the degree to which employees can be incentivized within
firms.
On the policy front, our findings may have implications for the extensive debate about the sources
and remedies for rising pay inequality over the past several decades. CEO pay has increased dramatically,
while employee wages have remained relatively stagnant, a dynamic that has led to the sharp increase in
pay inequality within firms. In direct response to this trend, the Securities and Exchange Commission
issued a rule in 2013 that requires publicly traded firms to report the difference between CEO pay and that
of the median worker. One objective of this legislation is to “embarrass” CEOs publicly into narrowing
the wage gap.15 While Wade et al (2006) do find direct evidence that employee – CEO comparison affects
wages within firms, their result presents a puzzle: if CEO pay is an important anchor in determining
manager pay, why has inequality, even with senior employees, risen so dramatically within firms?
Between 1986 and 1999, the total compensation of CEOs increased from four times the pay of division
managers to nearly 14 times as high (See Figure 3).
<< Insert Figure 3 about here >>
15
Morgenson, G., (4/10/2015)“Despite Federal Regulation, CEO-Worker Pay Gap Data Remains Hidden, ” The
New York Times. “…the Securities and Exchange Commission in 2013 proposed a specific pay-ratio rule that
companies disclose the median annual total compensation of all their employees …and compare that with the
amount awarded to the chief executive. ‘The pay ratio was designed to embarrass…’ said Charles Elson, director of
the John L. Weinberg Center for Corporate Governance at the University of Delaware.”
21
A multitude of forces potentially drive the rise in intra-firm wage inequality. One explanation is
that CEOs are concerned about fair pay, but any attempt to narrow the CEO-worker pay gap is dwarfed
by the economic forces in external labor markets that drive up CEO pay (Murphy 1999, 2012). A related
explanation – and one that is implicitly supported by our findings -- is that the salient referent of
employees is not primarily the CEO, but is the employees in similar positions. If true, then the CEOemployee pay ratio may not be the most relevant benchmark used by employees to determine whether
they are receiving fair pay.
V.3 Future Directions
Taken together, our findings suggest that social comparison may limit the ability of firm
leadership to incentivize critical employees (and, by extension, influence their performance). Our results
also raise several interesting questions for future theory and empirical research: first, to the extent that pay
and other aspects of employee performance are becoming increasingly transparent and quantified within
firms, what are consequences to job design and employee performance? Also, as organizational
boundaries become more fluid, how is referent selection and peer comparison influenced? Lastly, how do
horizontal and vertical comparisons interact within firms and to what degree is this interaction within the
control of a firm’s leaders? These questions represent several of the interesting avenues for future
research that have yet to be explored.
22
References
Adams, J. Stacy. "Inequity in social exchange." Advances in experimental social psychology 2.267-299 (1965).
Akerlof, George A. and Janet L. Yellen. 1990. “The Fair Wage-Effort Hypothesis and Unemployment.”
Quarterly Journal of Economics, 105(2), 255-83.
Argo, Jennifer J., Katherine White, and Darren W. Dahl. "Social comparison theory and deception in the
interpersonal exchange of consumption information." Journal of Consumer Research 33.1 (2006): 99-108.
Armstrong, J. Scott, and Fred Collopy. "Competitor orientation: Effects of objectives and information on managerial
decisions and profitability." Journal of marketing research (1996): 188-199.
Baron, James N. and David M. Kreps. 1999. “Employment as a Social Relation,” in Strategic Human Resources:
Frameworks for General Managers. New York: John Wiley & Sons, Inc., 102-04; 256-64.
Bartling, Björn, and Ferdinand A. von Siemens. "Wage inequality and team production: An experimental
analysis." Journal of Economic Psychology 32.1 (2011): 1-16.
Bewley, Truman F. 1999. Why Wages Don't Fall During a Recession. Cambridge, MA, and London: Harvard
University Press.
Bower, Joseph L., and Clark G. Gilbert. "How managers' everyday decisions create-or destroy-your company's
strategy." Harvard Business Review 85.2 (2007): 72.
Buunk, Abraham P., and Frederick X. Gibbons. "Social comparison: The end of a theory and the emergence of a
field." Organizational Behavior and Human Decision Processes 102.1 (2007): 3-21.
Card, David; Alexandre Mas; Enrico Moretti and Emmanuel Saez. 2012. “Inequality at Work: The Effect of Peer
Salaries on Job Satisfaction.” American Economic Review, 102(6), 2981-3003.
Carpenter, Mason A., and W. M. Sanders. "Top management team compensation: The missing link between CEO
pay and firm performance?."Strategic Management Journal 23.4 (2002): 367-375.
Charness, Gary and Peter Kuhn. 2007. “Does Pay Inequality Affect Worker Effort? Experimental Evidence.”
Journal of Labor Economics, 25(4), 693-723.
Chen, Clara Xiaoling, and Tatiana Sandino. "Can wages buy honesty? The relationship between relative wages and
employee theft." Journal of Accounting Research 50.4 (2012): 967-1000.
Cohn, A., Fehr, E., Herrmann, B., & Schneider, F. (2014). Social comparison and effort provision: Evidence from a
field experiment. Journal of the European Economic Association, 12(4), 877-898.
Collins, Rebecca L. "For better or worse: The impact of upward social comparison on self-evaluations."
Psychological bulletin 119.1 (1996): 51.
Edelman, Benjamin, and Ian Larkin. "Social comparisons and deception across workplace hierarchies: Field and
experimental evidence." Organization Science (2014).
Festinger, Leon. 1954. “A Theory of Social Comparison Processes.” Human Relations, 7(2), 117-40.
Frydman, Carola and Dirk Jenter. 2010. “CEO Compensation.” Annual Review of Financial Economics, 2, 75-102.
Gächter, Simon, and Christian Thöni. "Social comparison and performance: Experimental evidence on the fair
wage–effort hypothesis." Journal of Economic Behavior & Organization 76.3 (2010): 531-543.
23
Garcia, Stephen M., Avishalom Tor, and Richard Gonzalez. "Ranks and rivals: A theory of
competition." Personality and Social Psychology Bulletin 32.7 (2006): 970-982.
Gino, Francesca, and Lamar Pierce. "The abundance effect: Unethical behavior in the presence of
wealth." Organizational Behavior and Human Decision Processes 109.2 (2009): 142-155.
Gino, Francesca, and Lamar Pierce. "Lying to level the playing field: Why people may dishonestly help or hurt
others to create equity." Journal of Business Ethics 95.1 (2010): 89-103.
Goodman, Paul S. "Social comparison processes in organizations." Published in New directions in organizational
behavior. St Clair, 1977 (Eds. Barry M. Staw and Gerald R. Salancik).
Greenstone, Michael; Paul Oyer and Annette Vissing-Jorgensen. 2006. “Mandated Disclosure, Stock Returns, and
the 1964 Securities Acts Amendments.” Quarterly Journal of Economics, 121(2), 399-460.
Grossman, Sanford J., and Oliver D. Hart. "The costs and benefits of ownership: A theory of vertical and lateral
integration." The Journal of Political Economy (1986): 691-719.
Hall, Brian J. and Kevin J. Murphy. 2003. “The Trouble with Stock Options.” Journal of Economic
Perspectives, 17(3), 49-70.
Hart, Oliver, and John Moore. "Property Rights and the Nature of the Firm."Journal of political economy (1990):
1119-1158.
Hertzberg, Andrew; José Maria Liberti and Daniel Paravisini. 2011. “Public Information and Coordination:
Evidence from a Credit Registry Expansion.” Journal of Finance, 66(2), 379-412.
Hicks, John R. 1963. The Theory of Wages. London: Macmillan.
Hinds, Pamela J., and Diane E. Bailey. "Out of sight, out of sync: Understanding conflict in distributed
teams." Organization science 14.6 (2003): 615-632.
Hong, Harrison, Jeffrey D. Kubik, and Jeremy Stein, 2005. “Thy Neighbor’s Portfolio: Word-of-Mouth Effects in
the Holdings and Trades of Money Managers. Journal of Finance, Vol. LX, No. 6, 2801-2824.
Hyman, Herbert Hiram. "The psychology of status." Archives of Psychology (Columbia University) (1942).
Jensen, Michael C. and William H. Meckling. 1976. “Theory of the Firm: Managerial Behavior, Agency Costs and
Ownership Structure.” Journal of Financial Economics, 3(4), 305-60.
John, Leslie K., George Loewenstein, and Scott I. Rick. "Cheating more for less: Upward social comparisons
motivate the poorly compensated to cheat."Organizational Behavior and Human Decision Processes 123.2
(2014): 101-109.
Johnson, Camille. "Behavioral responses to threatening social comparisons: From dastardly deeds to rising
above." Social and Personality Psychology Compass 6.7 (2012): 515-524.
Kraut, Robert E., et al. "Understanding effects of proximity on collaboration: Implications for technologies to
support remote collaborative work." Distributed work (2002): 137-162.
Kulik, Carol T., and Maureen L. Ambrose. "Personal and situational determinants of referent choice." Academy of
Management review 17.2 (1992): 212-237.
Kuhnen, Camelia M. and Alexandra Niessen, 2012. “Public Opinion and Executive Compensation.” Management
Science, Vol. 58, No. 7, July 2012, ppl. 1249-1272.
24
Lamertz, Kai. "The social construction of fairness: Social influence and sense making in organizations." Journal of
Organizational Behavior 23.1 (2002): 19-37.
Larkin, Ian; Lamar Pierce and Francesca Gino. 2010. “The Psychological Costs of Pay-for-Performance:
Implications for the Strategic Compensation of Employees,” Strategic Management Journal, 33(10), 11941214.
Lawler, Edward E. 1965. “Managers’ Perceptions of Their Subordinates' Pay and of Their Superiors’ Pay.”
Personnel Psychology, 18(4), 413-22.
Lawrence, Barbara S. "Organizational reference groups: A missing perspective on social context." Organization
Science 17.1 (2006): 80-100.
Manksi, Charles. 1993. “Identification of Endogenous Social Effects: The Reflection Problem.” The Review of
Economic Studies, vol. 60, Issue 3, Jul. 1993, 531-542.
Milgrom, Paul, and John Roberts. "Bargaining costs, influence costs, and the organization of economic
activity." Perspectives on positive political economy57 (1990): 60.
Monge, Peter R., and Noshir S. Contractor. Theories of communication networks. Oxford University Press, 2003.
Moran, Simone, and Maurice E. Schweitzer. "When better is worse: Envy and the use of deception." Negotiation
and Conflict Management Research 1.1 (2008): 3-29.
Mui, Vai-Lam. "The economics of envy." Journal of Economic Behavior & Organization 26.3 (1995): 311-336.
Murphy, Kevin J. 1999. “Executive Compensation,” O. Ashenfelter, P. R. G. Layard and D. E. Card, Handbook of
Labor Economics. Amsterdam; New York: North-Holland, 2485-563 (Chapter 38).
Murphy, Kevin J. 2012. “The Politics of Pay: A Legislative History of Executive Compensation,” in Jennifer Hill
and Randall Thomas (eds.), Research Handbook on Executive Pay, Edward Elgar Publishers.
Nickerson, Jack A. and Todd R. Zenger. 2008. “Envy, Comparison Costs, and the Economic Theory of the Firm.”
Strategic Management Journal, 29(13), 1429-49.
Ockenfels, Axel, Dirk Sliwka and Peter Werner, 2014. “Bonus Payments and Reference Point Violations.”
Management Science, Published Online.
O'Reilly III, Charles A., Brian G. Main, and Graef S. Crystal. "CEO compensation as tournament and social
comparison: A tale of two theories."Administrative Science Quarterly (1988): 257-274.
Palepu, Krishna. "Diversification strategy, profit performance and the entropy measure." Strategic management
journal 6.3 (1985): 239-255.
Poortvliet, P. Marijn. "Harming others’ task-related efforts: The distinct competitive effects of ranking information
on performance and mastery goal individuals." Social Psychology 44.6 (2013): 373.
Rebitzer, James B. and Lowell J. Taylor. 2011. “Extrinsic Rewards and Intrinsic Motives: Standard and Behavioral
Approaches to Agency and Labor Markets,” in O. Ashenfelter and D. Card (eds), Handbook of Labor
Economics. North Holland: Elsevier B.V., 701-72.
Seru, Amit. "Firm boundaries matter: Evidence from conglomerates and R&D activity." Journal of Financial
Economics 111.2 (2014): 381-405.
Scharfstein, David S., and Jeremy C. Stein. "The dark side of internal capital markets: Divisional rent-­‐seeking and
inefficient investment." The Journal of Finance 55.6 (2000): 2537-2564.
25
Sherif, Muzafer. The psychology of social norms. Oxford, England: Harper. (1936)
Shue, Kelly. 2013. “Executive Networks and Firm Policies: Evidence from the Random Assignment of MBA
Peers,” SSRN eLibrary. Chicago: University of Chicago - Booth School of Business,
Siegel, Phyllis A., and Donald C. Hambrick. "Pay disparities within top management groups: Evidence of harmful
effects on performance of high-technology firms." Organization Science 16.3 (2005): 259-274.
Suls, Jerry Ed, and Ladd Ed Wheeler. Handbook of social comparison: Theory and research. Kluwer Academic
Publishers, 2000.
Taylor, Shelley E., and Marci Lobel. "Social comparison activity under threat: downward evaluation and upward
contacts." Psychological review 96.4 (1989): 569.
Tzabbar, Daniel, and Alex Vestal. "Bridging the Social Chasm in Geographically Distributed R&D Teams: The
Moderating Effects of Relational Strength and Status Asymmetry on the Novelty of Team Innovation."
Organization Science (2015).
Wade, James B.; III, Charles A. O’Reilly and Timothy G. Pollock. 2006. “Overpaid CEOs and Underpaid
Managers: Fairness and Executive Compensation.” Organization Science, 17(5), 527-44.
Zenger, Todd R. "Why do employers only reward extreme performance? Examining the relationships among
performance, pay, and turnover." Administrative Science Quarterly (1992): 198-219.
26
Wall Street Journal Articles Mentioning Executive Compensation
0
50
# articles
100
150
Figure 1: Press Mentions
1984
1989
1994
1999
Wall Street Journal Articles Mentioning General Compensation Topics
1200
1400
# articles
1600
1800
Annual counts of articles appearing in the Wall Street Journal with any of the following phrases: “executive
compensation,” “executive pay,” “CEO compensation,” and “CEO pay.” Source: Factiva.
1985
1988
1991
1994
1997
2000
Annual counts of articles appearing in the Wall Street Journal with any of the following phrases: “salaries,”
“employee pay,” “compensation,” and “wages.” Source: Factiva.
27
Figure 2: Pay Referent Sensitivity versus Pay Performance Sensitivity over Time
0.4
0.14
0.35
0.12
0.3
0.1
0.25
0.08
0.2
0.06
0.15
1986
1988
1990
1992
Pay Referent Sensitivity (PRS)
1994
1996
1998
PPS
PRS
Base + bonus
0.04
2000
Pay Performance Sensitivity (PPS)
Total compensation
0.5
0.14
0.12
0.4
0.1
0.35
PPS
PRS
0.45
0.08
0.3
1986
1988
1990
1992
Pay Referent Sensitivity (PRS)
1994
1996
1998
0.06
2000
Pay Performance Sensitivity (PPS)
Pay referent sensitivity and pay performance sensitivity for division manager positions over
time within dispersed firms.
28
Figure 3: Pay Inequality Within Firms: CEO vs. Division Manager
1500
1000
0
500
$ '000
2000
2500
Base+bonus
1985
1990
CEO
1995
2000
Division manager
4000
0
2000
$ '000
6000
8000
Total compensation
1985
1990
CEO
1995
2000
Division manager
4
6
8
10
CEO/Division Manager
12
14
Ratio of CEO to Division Manager Pay
1985
1990
Base+bonus
1995
2000
Total compensation
29
Table 1 Panel A: Summary Statistics: Firm and Division
Variable
Obs
Mean
Std. Dev.
Table 1 Panel B: Concentrated v Dispersed firms (Means)
Variable
Concentrated
Dispersed
Firm
Firm
Sales ($millions)
2,315
8,508
13,548
Obs
1,003
865
Market capitalization
2,248
9,366
16,502
Firm sales ($millions)
8992.64
7607.12
Employees (000s)
2,307
44.12
69.46
Market capitalization
9349.69
7757.12
Assets ($millions)
2,319
8,868
17,933
Firm employees (000s)
Firm assets
Return on assets
2,315
0.0512
0.0674
Average no. div / firm
2,362
4.55
4
Return on assets
Proportion div in HQ state
2,362
0.48
0.39
Average no. div / firm
Internal capital reallocation
2,362
0.09
0.02
Proportion div in HQ state
Horizontal relatedness
2,362
0.27
0.31
Division
Total firm entropy
2,362
0.68
0.53
Obs
44.91
44.39
9764.51
7507.56
0.0554
0.0443
3.83
5.1
0.89
0.05
3,837
4,346
Sales (mean) ($millions)
919
562
1,448
Sales (median)
379
258
10.14
Employees (mean) ($000s)
3.3
2.6
1435
Employees (median)
0.9
1.1
Division manager
Productivity (mean) ($000s)
881
399
Tenure (months)
Productivity (median)
319
190
40.33
43.66
Base salary ($)
216,397
197,225
Base plus bonus ($)
312,907
280,139
Total compensation ($)
498,911
415,225
Division
Sales ($millions)
Employees (000s)
Productivity (sales/emp)
9,871
9,790
9,373
752
3.02
590
8,899
42.23
42
Officer corp
10,731
0.23
0.42
Division manager
Division depth
10,706
1.42
0.81
Tenure (months)
Base salary ($)
10,731
208,849
82,439
Base plus bonus ($)
10,731
300,088
151,459
Total compensation ($)
10,731
459,640
357,865
Firm statistics are obtained from Compustat and Hewitt Associates for 296 firms with 2572 divisions between 1986 and 1999. Division and division manager statistics are from
Hewitt. Internal capital reallocation refers to a measure of the different of capital allocation between internal and external markets (Seru 2014). Horizontal relatedness is a
measure of the degree to which the division industries are in related industries. Total firm entropy represents a measure of total firm diversification (Palepu 1985).Concentrated
firms are defined as firms in the top two quintiles of firms with the highest proportion of divisions in the same state as HQ, and dispersed firms are those firms in the bottom two
quintiles. Tenure refers to the number of months a division manager has been in the position. Officer corp is equal to one if the division manager is also an officer of the firm.
Division depth is the number of levels between the division and the CEO. Base salary refers to annual salary, Base plus bonus refers to salary plus annual bonus and Total
Table 2: Division Manager Pay-for-Performance Sensitivity (PPS), CEO Pay Sensitivity and PayReferent Sensitivity (PRS)
Log pay type:
Base+bonus
(2)
(3)
0.0934***
(0.0117)
0.1730***
(0.0285)
0.0944***
(0.0112)
0.3922***
(0.0456)
0.1037***
(0.0188)
0.0886***
(0.0103)
0.5321***
(0.1407)
0.0820***
(0.0285)
0.2736**
(0.1153)
0.0506*
(0.0266)
-0.0007
(0.0017)
0.0011***
(0.0001)
0.1055***
(0.0230)
-0.0595***
(0.0102)
-0.3435
(0.2714)
0.0086
(0.0388)
0.0067
(0.0339)
12.2239***
(0.1599)
0.4576
Y
10,312
-0.0010
(0.0016)
0.0010***
(0.0001)
0.1063***
(0.0223)
-0.0655***
(0.0104)
-0.2261
(0.2693)
-0.0068
(0.0362)
0.0037
(0.0320)
9.7828***
(0.4286)
0.5151
Y
10,312
(1)
Pay Sensitivity
Log referent pay (PRS)
Log CEO pay
Log division sales (PPS)
Firm Performance
Return on Assets
Log firm revenues
Other Controls
# divisions in firm
Tenure
Officer of firm
Division depth
Internal capital reallocation
Horizontal relatedness
Total firm entropy
Constant
R-squared
Division and Year FE
Observations
(4)
Total compensation
(5)
(6)
0.1108***
(0.0139)
0.3217***
(0.0282)
0.1102***
(0.0126)
0.4556***
(0.0372)
0.1750***
(0.0209)
0.1035***
(0.0116)
0.1668**
(0.0766)
0.0251
(0.0173)
0.6138***
(0.1654)
0.1453***
(0.0402)
0.0960
(0.1460)
0.0596*
(0.0350)
0.0517
(0.0879)
0.0261
(0.0208)
0.0028**
(0.0014)
0.0010***
(0.0001)
0.0956***
(0.0178)
-0.0523***
(0.0083)
-0.1478
(0.1786)
-0.0069
(0.0241)
0.0008
(0.0200)
5.9604***
(0.5487)
0.6653
Y
10,312
-0.0007
(0.0025)
0.0011***
(0.0002)
0.1529***
(0.0289)
-0.0891***
(0.0138)
-0.3637
(0.2421)
0.0312
(0.0612)
0.0058
(0.0496)
12.6707***
(0.4165)
0.5025
Y
10,312
-0.0017
(0.0027)
0.0010***
(0.0001)
0.1543***
(0.0261)
-0.1027***
(0.0136)
-0.1830
(0.2211)
-0.0102
(0.0504)
0.0156
(0.0435)
8.2955***
(0.4690)
0.6100
Y
10,312
0.0040**
(0.0020)
0.0010***
(0.0001)
0.1308***
(0.0206)
-0.0755***
(0.0105)
-0.1075
(0.1418)
-0.0072
(0.0312)
0.0040
(0.0256)
4.4861***
(0.4126)
0.7409
Y
10,312
Standard errors clustered by firm. *** = p<0.01, ** = p<0.05, * = p<0.1. Log referent pay refers to the log mean pay of division managers, excluding the
focal division, within a given firm and year. CEO pay refers to base+bonus in (1)-(3) and total compensation in (4)-(6). Log referent pay is calculated
using base+bonus in (1)-(3) and total compensation in (4)-(6). # divisions in firm refers to the number of divisions in a firm-year. Refer to footnote in
Table 1 for additional definitions.
31
Table 3: Division Manager Pay-for-Performance Sensitivity (PPS), CEO Pay Sensitivity and Pay-Referent Sensitivity (PRS) by 1991-1992
Pay Controversy
Log pay type:
1991-1992 pay controversy
Log referent pay (PRS)*Post 1992
Log CEO pay*Post 1992
Log division sales*Post 1992
Baseline levels
Log referent pay (PRS)
Log CEO pay
Log division sales (PPS)
Constant
R-squared
Controls
Fixed Effects
Observations
(1)
Base+bonus
(2)
(3)
(4)
Total compensation
(5)
(6)
0.1017**
(0.0394)
-0.0783***
(0.0278)
-0.0377***
(0.0105)
0.0675*
(0.0401)
-0.0376+
(0.0247)
-0.0352***
(0.0081)
0.0205
(0.0310)
-0.0328*
(0.0186)
-0.0127*
(0.0074)
0.1050***
(0.0399)
-0.0810***
(0.0311)
-0.0334***
(0.0122)
0.0812**
(0.0400)
-0.0497*
(0.0293)
-0.0304***
(0.0097)
0.0242
(0.0307)
-0.0117
(0.0203)
-0.0043
(0.0085)
0.5634***
(0.0304)
0.1668***
(0.0207)
0.1270***
(0.0091)
5.1135***
(0.3797)
0.734
Y
Year
10,312
0.2062***
(0.0479)
0.1666***
(0.0204)
0.1467***
(0.0078)
9.4427***
(0.5999)
0.6789
Y
Firm, Year
10,312
0.3865***
(0.0420)
0.1209***
(0.0162)
0.0918***
(0.0090)
7.2386***
(0.5209)
0.705
Y
Division, Year
10,312
0.5696***
(0.0295)
0.2000***
(0.0197)
0.1467***
(0.0102)
2.7415***
(0.3351)
0.766
Y
Year
10,312
0.2744***
(0.0385)
0.2338***
(0.0235)
0.1689***
(0.0091)
5.9371***
(0.4485)
0.7276
Y
Firm, Year
10,312
0.4328***
(0.0326)
0.1773***
(0.0187)
0.1030***
(0.0100)
4.7239***
(0.3760)
0.742
Y
Division, Year
10,312
Standard errors clustered by firm. *** = p<0.01, ** = p<0.05, * = p<0.1, +=p<0.15. Log referent pay refers to the log mean pay of division managers, excluding the focal
division, within a given firm and year. CEO pay refers to base plus bonus in (1)-(3) and total compensation in (4)-(6). Log referent pay is calculated using base plus bonus
in (1)-(3) and total compensation in (4)-(6). All controls from Table 2 included.
32
Table 4: Effect of 1991-1992 Pay Controversy on PRS and PPS in Concentrated and Dispersed Firms
Panel A: Dispersed Firms
Log pay type:
N=4,225
1991-1992 pay controversy
Log referent pay (PRS)*Post 1992
Log CEO pay*Post 1992
Log division sales*Post 1992
Baseline levels
Log referent pay (PRS)
Log CEO pay
Log division sales (PPS)
Controls
R-squared
Fixed Effects
(1)
Base+bonus
(2)
(3)
Total compensation
(5)
(6)
(4)
0.1873***
(0.0536)
-0.0685*
(0.0395)
-0.0520***
(0.0130)
0.2010***
(0.0591)
-0.0264
(0.0349)
-0.0302**
(0.0123)
0.1109**
(0.0534)
-0.0055
(0.0289)
-0.0180+
(0.0110)
0.1592***
(0.0602)
-0.0761+
(0.0467)
-0.0533***
(0.0166)
0.1932***
(0.0633)
-0.0804*
(0.0440)
-0.0322**
(0.0159)
0.1137**
(0.0517)
-0.0303
(0.0316)
-0.0124
(0.0143)
0.5390***
(0.0444)
0.1220***
(0.0258)
0.1479***
(0.0105)
Y
0.733
Year
0.0108
(0.0647)
0.1194***
(0.0290)
0.1560***
(0.0115)
Y
0.6185
Firm, Year
0.2498***
(0.0554)
0.0838***
(0.0233)
0.1068***
(0.0136)
Y
0.6821
Division, Year
0.5519***
(0.0505)
0.1581***
(0.0300)
0.1720***
(0.0130)
Y
0.763
Year
0.0901
(0.0640)
0.2070***
(0.0394)
0.1848***
(0.0134)
Y
0.6892
Firm, Year
0.3437***
(0.0431)
0.1483***
(0.0300)
0.1196***
(0.0143)
Y
0.7321
Division, Year
33
Panel B: Concentrated firms
Log pay type:
N=3,620
1991-1992 pay controversy
Log referent pay (PRS)*Post 1992
Log CEO pay*Post 1992
Log division sales*Post 1992
Baseline levels
Log referent pay (PRS)
Log CEO pay
Log division sales (PPS)
Controls
R-squared
Fixed Effects
(1)
Base+bonus
(2)
(3)
Total compensation
(5)
(6)
(4)
0.0355
(0.0509)
-0.1321***
(0.0352)
-0.0008
(0.0135)
0.0521
(0.0607)
-0.0665*
(0.0371)
-0.0186+
(0.0125)
-0.0072
(0.0417)
-0.0440*
(0.0246)
-0.0024
(0.0113)
0.0915*
(0.0525)
-0.1259***
(0.0469)
0.0022
(0.0163)
0.0976*
(0.0571)
-0.0670
(0.0474)
-0.0172
(0.0148)
0.0193
(0.0404)
-0.0280
(0.0339)
0.0053
(0.0129)
0.5530***
(0.0497)
0.2413***
(0.0316)
0.0942***
(0.0125)
Y
0.735
Year
0.2111***
(0.0780)
0.2275***
(0.0317)
0.1164***
(0.0112)
Y
0.6865
Firm, Year
0.4868***
(0.0628)
0.1389***
(0.0221)
0.0647***
(0.0125)
Y
0.7136
Division, Year
0.5589***
(0.0452)
0.2504***
(0.0319)
0.1121***
(0.0137)
Y
0.766
Year
0.2679***
(0.0617)
0.2776***
(0.0393)
0.1367***
(0.0124)
Y
0.7159
Firm, Year
0.5138***
(0.0483)
0.1783***
(0.0297)
0.0695***
(0.0140)
Y
0.7463
Division, Year
Standard errors clustered by firm. *** = p<0.01, ** = p<0.05, * = p<0.1, +=p<0.15. Log referent pay refers to the log mean pay of division managers, excluding the focal
division, within a given firm and year. CEO pay refers to base plus bonus in (1)-(3) and total compensation in (4)-(6). Log referent pay is calculated using base plus bonus in
(1)-(3) and total compensation in (4)-(6). All controls from Table 2 included.
34
Table 5 Panel A: Divisional Productivity Differences
Log (Sales/employee)
Dispersed* Post 1992
Dispersed
Post 1992
R-squared
Year FE
Firm FE
Observations
(1)
-0.1483*
(0.0793)
-0.0006
(0.0681)
0.3322***
(0.0811)
0.017
Y
Y
4723
All years
(2)
-0.1524**
(0.0774)
-0.1441**
(0.0590)
0.3288***
(0.0811)
0.050
Y
N
4723
(3)
-0.1776**
(0.0863)
0.1244
(0.1307)
0.3635***
(0.1042)
0.013
Y
Y
3139
1990-1997
(4)
-0.1940**
(0.0834)
-0.1236
(0.0925)
0.3253***
(0.1026)
0.057
Y
N
3139
Table 5 Panel B: Divisional Productivity Differences – Asymmetric Effects
Log (Sales/employee)
All firms
(1)
Dispersed*Post 1992*Below mean pay
1990-1997
At or above
mean pay
Below mean
pay
(2)
(3)
-0.0757
-0.2554**
(0.0908)
(0.1103)
-0.2706**
(0.1156)
Dispersed*Post 1992
R-squared
0.019
0.015
0.011
Year FE
Y
Y
Y
Firm FE
Y
Y
Y
Observations
3139
1667
1472
Standard errors clustered by firm. *** = p<0.01, ** = p<0.05, * = p<0.1. Dependent variable, log
Sales/employee, is calculated as division sales divided by the number of division employees. Below mean pay
is a dummy equal to 1 if division manager pay is below mean pay, where mean pay is defined as mean base
salary of division managers in the year prior to the year in which productivity is measured. Firm and division
controls are the same as those in Table 2. Refer to footnote in Table 1 for additional definitions. All dependent
variables lagged by one year and, therefore, sample only includes division observations with adjacent one-year
lagged observation. 35
APPENDIX
NOT PART OF MANUSCRIPT, INCLUDED TO ASSIST REVIEWERS IN ASSESSING PAPER
36
Appendix A: Data Description and Sample Representativeness
The primary dataset used in this study includes a panel of more than 300 publicly-traded U.S.
firms over the years 1986-1999, spanning a number of industries. The data are collected from a
confidential compensation survey conducted by Hewitt Associates, a leading human resources consulting
firm specializing in executive compensation and benefits. The survey is the largest private compensation
survey (as measured by the number of participating firms), and the participants are typically leaders in
their sectors. More than 75 percent of the firms in the dataset are listed as Fortune 500 firms in at least
one year, and more than 85 percent are listed as Fortune 1000 firms. In general, Hewitt survey
participants also participate in other compensation consulting firm surveys (e.g., Hay Associates, Mercer,
Towers Perrin, to name a few) and do so primarily to receive information about pay practices to use as a
competitive benchmark in evaluating their own compensation programs. It is important to note that the
sample includes many more firms than Hewitt’s consulting client base, with at least 50 percent of the
firms as survey participants with no other relationship to Hewitt.
We believe the survey data are accurate for several reasons. First, Hewitt personnel are
knowledgeable about survey participants because they are assigned to specific participants for several
years.
Furthermore, while the participating firms initially match their positions to the benchmark
positions in the survey, Hewitt personnel follow up to verify accuracy and spend an additional eight to ten
hours on each questionnaire evaluating the consistency of responses with public data (e.g., proxy
statements) and across years. Finally, participants have an incentive to match positions correctly and
provide accurate data because they use the survey results to set pay levels and design management
compensation programs.
Clearly, an important issue in datasets such as this one is the question of sample selection and
whether the firms in the dataset are distinct from, or representative of, employers of similar size in their
industry. The survey participants are typically the leaders in their sectors and, in fact, more than 75
percent of the firms in the dataset are listed as Fortune 500 firms in at least one year. We evaluate the
representativeness of the broader sample by comparing key financial measures of our survey participants
to a matched sample from Compustat. We begin by matching each firm in the Hewitt dataset to the
Compustat firm that is closest in sales within its two-digit SIC industry in the year the firm joins the
sample. We then perform Wilcoxon signed rank tests to compare the Hewitt firms with the matched firms.
While the firms in the Hewitt dataset are, on average, have slightly larger sales than the matched sample,
we find no statistically significant difference in employment and profitability (return on sales). We also
find no statistically significant difference in sales growth, employment growth, or annual changes in
profitability for all sample years. We also calculate financial measures for the sample of Compustat firms
with 10,000 employees or greater over the period from 1986 to 1999 (excluding firms operating in
37
financial services). On average, survey participants are more profitable, but growing at a slower rate than
those in the sample of large Compustat firms. This is consistent with our observation that the firms in our
sample are likely to be industry leaders (hence, slightly more profitable) and also large (hence, the slightly
slower growth). In sum, the survey sample is most representative of Fortune 500 firms (for more details,
see Rajan and Wulf, 2006).
38
Table A1: Table 2 replicated on a propensity-score matched sample
Log pay type:
(1)
Base+bonus
(2)
(4)
0.0997***
(0.0049)
0.2007***
(0.0109)
0.1001***
(0.0048)
(3)
0.3254***
(0.0160)
0.1343***
(0.0110)
0.0948***
(0.0046)
0.7325***
(0.0699)
0.0719***
(0.0132)
0.3708***
(0.0705)
0.0387***
(0.0129)
0.2645***
(0.0679)
0.0166
(0.0125)
0.6812***
(0.0888)
0.1520***
(0.0170)
0.1371
(0.0846)
0.0769***
(0.0160)
0.1065
(0.0794)
0.0430***
(0.0151)
-0.0004
(0.0010)
0.0011***
(0.0001)
0.1369***
(0.0114)
-0.0488***
(0.0054)
-0.4379***
(0.1282)
-0.0147
(0.0203)
-0.0111
(0.0150)
12.1123***
(0.1606)
0.4722
Y
6,128
-0.0010
(0.0010)
0.0010***
(0.0001)
0.1390***
(0.0110)
-0.0557***
(0.0053)
-0.2760**
(0.1245)
-0.0220
(0.0197)
-0.0159
(0.0145)
12.2029***
(0.1555)
0.5442
Y
6,128
0.0023**
(0.0009)
0.0010***
(0.0001)
0.1231***
(0.0106)
-0.0468***
(0.0051)
-0.1918
(0.1197)
-0.0229
(0.0189)
-0.0068
(0.0140)
8.2012***
(0.2475)
0.6731
Y
6,128
-0.0034**
(0.0014)
0.0011***
(0.0001)
0.1841***
(0.0147)
-0.0847***
(0.0071)
-0.5132***
(0.1697)
0.0401
(0.0262)
-0.0210
(0.0193)
12.2639***
(0.1987)
0.5205
Y
6,128
-0.0045***
(0.0013)
0.0010***
(0.0001)
0.1860***
(0.0136)
-0.0978***
(0.0066)
-0.2035
(0.1575)
0.0019
(0.0243)
-0.0208
(0.0179)
7.9220***
(0.2437)
0.6412
Y
6,128
0.0018
(0.0012)
0.0010***
(0.0001)
0.1603***
(0.0128)
-0.0760***
(0.0062)
-0.1436
(0.1479)
-0.0042
(0.0228)
-0.0195
(0.0168)
4.8845***
(0.2606)
0.7502
Y
6,128
Log referent pay (PRS)
Log CEO pay
Log division sales (PPS)
Firm Performance
Return on Assets
Log firm revenues
Other Controls
# divisions in firm
Tenure
Officer of firm
Division depth
Internal capital reallocation
Horizontal relatedness
Total firm entropy
Constant
R-squared
Division and Year Fixed Effects
Observations
Total compensation
(5)
(6)
0.3833***
(0.0157)
0.3245***
0.1932***
(0.0120)
(0.0125)
0.1120***
0.1058***
(0.0060)
(0.0057)
0.1175***
(0.0065)
39
Table A2: Table 3 replicated on a propensity-score matched sample
Log pay type:
1991-1992 pay controversy
Log referent pay (PRS)*Post 1992
Log CEO pay*Post 1992
Log division sales*Post 1992
Baseline levels
Log referent pay (PRS)
Log CEO pay
Log division sales (PPS)
Constant
R-squared
Fixed Effects
Observations
(1)
Base+bonus
(2)
(3)
Total compensation
(5)
(6)
(4)
0.0387
(0.0370)
-0.0369
(0.0314)
-0.0317***
(0.0121)
0.0852**
(0.0392)
-0.0283
(0.0273)
-0.0295***
(0.0109)
0.0109
(0.0338)
-0.0124
(0.0219)
-0.0151+
(0.0095)
0.0446
(0.0409)
-0.0423
(0.0352)
-0.0261*
(0.0143)
0.0879**
(0.0418)
-0.0381
(0.0335)
-0.0168
(0.0121)
0.0113
(0.0335)
-0.0059
(0.0258)
0.0008
(0.0107)
0.5522***
(0.0332)
0.1743***
(0.0234)
0.1303***
(0.0101)
5.3248***
(0.4306)
0.734
Year
6,128
0.1437***
(0.0531)
0.1874***
(0.0218)
0.1480***
(0.0096)
10.1044***
(0.6737)
0.6789
Firm, Year
6,128
0.3317***
(0.0478)
0.1375***
(0.0190)
0.0957***
(0.0114)
7.8278***
(0.5968)
0.705
Division, Year
6,128
0.5486***
(0.0328)
0.2145***
(0.0230)
0.1513***
(0.0112)
2.6616***
(0.3582)
0.766
Year
6,128
0.2110***
(0.0426)
0.2585***
(0.0279)
0.1727***
(0.0108)
5.9828***
(0.4454)
0.7276
Firm, Year
6,128
0.3810***
(0.0376)
0.1960***
(0.0225)
0.1073***
(0.0127)
4.8463***
(0.4098)
0.742
Division, Year
6,128
40
Table A3: Table 4 replicated on a propensity-score matched sample
Panel A: Dispersed Firms
Log pay type:
N=2,427
1991-1992 pay controversy
Log referent pay (PRS)*Post
Log CEO pay*Post
Log division sales*Post
Baseline levels
Log referent pay (PRS)
Log CEO pay
Log division sales (PPS)
R-squared
Fixed Effects
(1)
Base+bonus
(2)
(3)
Total compensation
(5)
(6)
(4)
0.1043+
(0.0696)
0.0181
(0.0573)
-0.0467***
(0.0168)
0.1946***
(0.0612)
0.0425
(0.0401)
-0.0283*
(0.0158)
0.0757
(0.0536)
0.0846**
(0.0364)
-0.0135
(0.0117)
0.1020
(0.0763)
-0.0164
(0.0611)
-0.0386*
(0.0216)
0.1947***
(0.0627)
-0.0327
(0.0536)
-0.0114
(0.0175)
0.0850+
(0.0522)
0.0188
(0.0431)
0.0072
(0.0127)
0.5356***
(0.0478)
0.1201***
(0.0298)
0.1504***
(0.0121)
0.731
Year
-0.0079
(0.0635)
0.1570***
(0.0291)
0.1585***
(0.0132)
0.600
Firm, Year
0.2300***
(0.0554)
0.1146***
(0.0258)
0.0960***
(0.0148)
0.671
Division, Year
0.5447***
(0.0545)
0.1637***
(0.0335)
0.1729***
(0.0145)
0.770
Year
0.0657
(0.0596)
0.2535***
(0.0383)
0.1853***
(0.0154)
0.686
Firm, Year
0.2869***
(0.0479)
0.1921***
(0.0340)
0.1054***
(0.0159)
0.732
Division, Year
41
Panel B: Concentrated Firms
Log pay type:
N=2,506
1991-1992 pay controversy
Log referent pay (PRS)*Post
Log CEO pay*Post
Log division sales*Post
Baseline levels
Log referent pay (PRS)
Log CEO pay
Log division sales (PPS)
R-squared
Fixed Effects
(1)
Base+bonus
(2)
(3)
(4)
Total compensation
(5)
(6)
-0.0114
(0.0583)
-0.1062**
(0.0408)
-0.0031
(0.0163)
0.0168
(0.0593)
-0.0455
(0.0405)
-0.0067
(0.0179)
-0.0483
(0.0424)
-0.0470+
(0.0293)
0.0001
(0.0143)
0.0245
(0.0567)
-0.0813*
(0.0463)
-0.0057
(0.0205)
0.0679
(0.0562)
-0.0348
(0.0456)
-0.0044
(0.0220)
-0.0174
(0.0424)
-0.0170
(0.0362)
0.0140
(0.0176)
0.5267***
(0.0548)
0.2568***
(0.0364)
0.0995***
(0.0118)
0.737
Year
0.1692*
(0.0968)
0.2402***
(0.0329)
0.1202***
(0.0109)
0.673
Firm, Year
0.4557***
(0.0753)
0.1549***
(0.0275)
0.0756***
(0.0150)
0.718
Division, Year
0.5368***
(0.0488)
0.2567***
(0.0359)
0.1197***
(0.0127)
0.770
Year
0.2329***
(0.0769)
0.2732***
(0.0473)
0.1413***
(0.0122)
0.716
Firm, Year
0.4909***
(0.0592)
0.1800***
(0.0355)
0.0815***
(0.0177)
0.751
Division, Year
42
Table A4: Table 2 using the residual method in Wade et al (2006)
Log pay type:
(1)
Base+bonus
(2)
(4)
0.0934***
(0.0117)
0.2496***
(0.0511)
0.2536***
(0.0619)
0.0927***
(0.0109)
0.5321***
(0.1407)
0.0820***
(0.0285)
0.5100***
(0.1073)
0.0728***
(0.0260)
0.3249***
(0.0750)
0.0409**
(0.0178)
0.6138***
(0.1654)
0.1453***
(0.0402)
0.5829***
(0.1498)
0.1434***
(0.0369)
0.3036***
(0.0870)
0.0673***
(0.0210)
-0.0007
(0.0017)
0.0011***
(0.0001)
0.1055***
(0.0230)
-0.0595***
(0.0102)
-0.3435
(0.2714)
0.0086
(0.0388)
0.0067
(0.0339)
12.2239***
(0.1599)
0.458
Y
10,312
-0.0011
(0.0017)
0.0010***
(0.0001)
0.1085***
(0.0219)
-0.0645***
(0.0103)
-0.2936
(0.2341)
-0.0166
(0.0347)
0.0087
(0.0309)
12.2812***
(0.1360)
0.476
Y
10,312
0.0025*
(0.0014)
0.0010***
(0.0001)
0.0978***
(0.0179)
-0.0526***
(0.0083)
-0.1937
(0.1606)
-0.0126
(0.0240)
0.0041
(0.0202)
7.8109***
(0.5765)
0.653
Y
10,312
-0.0007
(0.0025)
0.0011***
(0.0002)
0.1529***
(0.0289)
-0.0891***
(0.0138)
-0.3637
(0.2421)
0.0312
(0.0612)
0.0058
(0.0496)
12.6707***
(0.4165)
0.503
Y
10,312
-0.0020
(0.0028)
0.0010***
(0.0002)
0.1588***
(0.0265)
-0.0983***
(0.0137)
-0.2522
(0.2052)
0.0055
(0.0551)
0.0192
(0.0463)
12.6486***
(0.3298)
0.524
Y
10,312
0.0042**
(0.0020)
0.0010***
(0.0001)
0.1319***
(0.0205)
-0.0715***
(0.0103)
-0.1300
(0.1270)
0.0006
(0.0325)
0.0051
(0.0256)
6.4926***
(0.5017)
0.728
Y
10,312
Log referent pay (PRS)
Log CEO overpayment
Log CEO underpayment
Log division sales (PPS)
Firm Performance
Return on Assets
Log firm revenues
Other Controls
Number of non-focal divisions within firm
Tenure
Officer of firm
Division depth
Internal capital market activity
Horizontal relatedness
Total firm entropy
Constant
R-squared
Division and Year Fixed Effects
Observations
Total compensation
(5)
(6)
0.4846***
(0.0354)
0.4158***
0.2214***
(0.0556)
(0.0362)
0.1927***
0.0887***
(0.0463)
(0.0288)
0.1094***
0.1027***
(0.0129)
(0.0118)
(3)
0.3634***
(0.0460)
0.1602***
(0.0369)
0.1551***
(0.0422)
0.0880***
(0.0102)
0.1108***
(0.0139)
43
Table A5: Alternative Explanations: PRS and PPS and Placebo Break
Panel A: PRS and PPS
Log pay type:
Log referent pay (PRS)*Post break*Dispersed
Log division sales (PPS)*Post break*Dispersed
R-squared
Year FE
Division FE
Observations
Log pay type:
Log referent pay (PRS)*Post break*Dispersed
Log division sales (PPS)*Post break*Dispersed
Year FE
Division FE
Observations
Adjusted R-squared
1987 break (1986-1990)
Base
Base +
Total
salary
bonus
compensation
(1)
(2)
(3)
-0.0414
0.0189
-0.0061
(0.0989)
(0.0928)
(0.0948)
-0.0199
-0.0020
-0.0548**
(0.0184)
(0.0201)
(0.0265)
0.463
0.422
0.446
Y
Y
Y
Y
Y
Y
2829
2829
2829
Base
salary
(1)
-0.0449
(0.0836)
0.0142
(0.0161)
Y
Y
4024
0.419
1989 break (1986-1992)
Total
Base + bonus
compensation
(2)
(3)
-0.0166
-0.0420
(0.0814)
(0.0750)
-0.0020
0.0030
(0.0201)
(0.0235)
Y
Y
Y
Y
4024
4024
0.389
0.434
1995 break (1993-1999)
Base
Base +
Total
salary
bonus
compensation
(4)
(5)
(6)
-0.2002** -0.0120
0.0189
(0.0829)
(0.0991)
(0.1017)
-0.0144
-0.0197
-0.0414
(0.0193)
(0.0257)
(0.0311)
0.279
0.269
0.360
Y
Y
Y
Y
Y
Y
3821
3821
3821
Base
salary
(4)
-0.2002**
(0.0829)
-0.0144
(0.0193)
Y
Y
3821
0.279
1995 break (1993-1999)
Total
Base + bonus
compensation
(5)
(6)
-0.0120
0.0189
(0.0991)
(0.1017)
-0.0197
-0.0414
(0.0257)
(0.0311)
Y
Y
Y
Y
3821
3821
0.269
0.360
44
Panel B: Divisional Productivity
Sales/employee
Dispersed*Post break
Dispersed*Post break*Below mean pay
R-squared
Year FE
Division FE
Observations
1987 break (1986-1990)
(1)
(2)
0.0709
-0.0367
(0.1087)
(0.1358)
0.2346*
(0.1282)
0.077
0.053
Y
Y
Y
Y
1683
1683
1995 break (1993-1999)
(3)
(4)
-0.1141
-0.0927
(0.1077)
(0.1126)
-0.0543
(0.1618)
0.002
0.001
Y
Y
Y
Y
2098
2098
Standard errors clustered by firm. *** = p<0.01, ** = p<0.05, * = p<0.1. Post break is a dummy equal to 1 for all
years after the designated break year (1987 or 1995). In Panel A Log referent pay refers to the log mean pay of
division managers, excluding the focal division, within a given firm and year. Referent pay is calculated using base
salary in columns (1, 4), base+bonus in (2, 5) and total compensation in (3, 6). In Panel B, the dependent variable,
Sales/employee, is calculated as division sales divided by the number of division employees. Below mean pay is a
dummy equal to 1 if division manager pay is below mean pay, where mean pay is defined as mean base salary of
division managers in the year prior to the year in which productivity is measured. All pair-wise interactions and direct
effects included in specification. Firm and division controls are the same as those in Table 2. Refer to footnote in
Table 1 for additional definitions.
45
Table A6: Alternative Explanations: PRS and PPS and Placebo Break
Log pay type:
N=3190
Log referent pay (PRS) * Post * IT
Intensity
Log division sales (PPS) * Post * IT
Intensity
R-squared
Firm and other controls, Year FE
Division FE
Base salary
(1)
(2)
Base + bonus
(3)
(4)
Total compensation
(5)
(6)
-0.1256***
(0.0416)
-0.0232
(0.0397)
-0.0564
(0.0431)
-0.0093
(0.0399)
-0.0584
(0.0433)
-0.0250
(0.0277)
0.0057
(0.0092)
0.693
Y
N
-0.0008
(0.0064)
0.563
Y
Y
0.0001
(0.0109)
0.668
Y
N
-0.0007
(0.0089)
0.694
Y
Y
-0.0017
(0.0121)
0.665
Y
N
-0.0001
(0.0101)
0.738
Y
Y
Standard errors clustered by firm. *** = p<0.01, ** = p<0.05, * = p<0.1. IT Intensity is defined as the quintile for each firm in a given year of ratio of
PCs/employee. Log referent pay refers to the log mean pay of division managers, excluding the focal division, within a given firm and year. Referent pay is
calculated using base salary in columns (1, 2), base+bonus in (3 ,4) and total compensation in (5, 6). All pair-wise interactions and direct effects included in
specification. Firm and division controls are the same as those in Table 2. Refer to footnote in Table 1 for additional definitions.
46
Table A7: Alternative Explanation: Horizontal Job Rotation
Tenure in position (in months)
(1)
(2)
-4.1655
-2.3651
(4.0667)
(4.2742)
Dispersed
8.0865
4.5353
(4.9134)
(3.8029)
Post
3.8150
-1.4083
(4.2263)
(4.0636)
R-squared
0.007
0.005
Year FE
Y
Y
Division FE
Y
N
Standard errors clustered by firm. *** = p<0.01, ** = p<0.05, * = p<0.1. Dependent variable is Tenure in position, Refer to footnote in Table 1 for additional
definitions. Dispersed is defined as those firms in the lowest two quintiles of the proportion of divisions in the same state as headquarters.
N=6424
Dispersed* Post
47
APPENDIX – Additional Figures
Figure A1: Distribution of Geographic Concentration by Industry
200
0
100
# firms
300
400
Concentration by broad industries
ns
Co
No
nd
ura
bl e
n
Co
ur a
sD
ble
Ma
nu
tu
fac
g
rin
s&
Ga
Oil
al
Co
c
Te
h
mm
Co
ti o
ica
un
n
Wh
e
sal
ole
Concentrated
a
He
lth
Uti
litie
s
e
Oth
r
Dispersed
Figure A2: Distribution of Geographic Concentration by HQ State
200
0
100
# firms
300
400
Concentration by region
MidWest
Northeast
Concentrated
South
West
Dispersed