The Paradox of
Meritocracy in
Organizations
Emilio J. Castilla
Massachusetts Institute of
Technology
Stephen Benard
Indiana University
© 2010 by Johnson Graduate School,
Cornell University.
0001-8392/10/5504-0543/$3.00.
•
We are grateful for the financial support
received from the Institute of Work and
Employment Research at the MIT Sloan
School of Management and the Center for
Human Resources at the Wharton School,
University of Pennsylvania. We have
benefited enormously from the detailed
comments of Roberto F. Fernandez, Ezra
Zuckerman, Thomas Kochan, and Paul
Osterman. We also thank Lotte Bailyn,
Lucio Baccaro, John Carroll, Shelley
Correll, Jared Curhan, Mauro F. Guillén,
Alexandra Kalev, Kate Kellogg, Anne
Marie Knott, Ko Kuwabara, Denise Loyd,
Joe Magee, Mark Mortensen, Devah
Pager, Ray Reagans, Cecilia Ridgeway,
Cat Turco, Christopher Wheat, Robb
Willer, and many other colleagues at the
MIT Sloan School of Management and
New York University (both at the Wagner
School and the Sociology Department).
We also greatly benefited from feedback
by seminar participants in the Institute for
Work and Employment Research and the
Economic Sociology Seminars at MIT, the
Laboratory Experimental Group at Cornell
University, the Social Psychology, Health,
and the Life Course Workshop at Indiana
University, the Inequality Workshop at
Stanford University, and the Management
Department Seminar at the Wharton
School. Finally, we also thank Patty
Curley, Amy DiMattia, John-Paul
Ferguson, Helen Hsi, Eun-Suk Lee, Jackie
Martelli, and Kate Parsons for their
invaluable research assistance in different
parts of the project. The views expressed
here are exclusively those of the authors.
Direct all correspondence to Emilio J.
Castilla, Massachusetts Institute of
Technology, Sloan School of Management, 100 Main Street, Room E62-332,
Cambridge, MA 02142; e-mail: ecastilla@
mit.edu.
In this article, we develop and empirically test the
theoretical argument that when an organizational culture
promotes meritocracy (compared with when it does not),
managers in that organization may ironically show greater
bias in favor of men over equally performing women in
translating employee performance evaluations into
rewards and other key career outcomes; we call this the
“paradox of meritocracy.” To assess this effect, we conducted three experiments with a total of 445 participants
with managerial experience who were asked to make
bonus, promotion, and termination recommendations for
several employee profiles. We manipulated both the
gender of the employees being evaluated and whether the
company’s core values emphasized meritocracy in evaluations and compensation. The main finding is consistent
across the three studies: when an organization is explicitly
presented as meritocratic, individuals in managerial
positions favor a male employee over an equally qualified
female employee by awarding him a larger monetary
reward. This finding demonstrates that the pursuit of
meritocracy at the workplace may be more difficult than it
first appears and that there may be unrecognized risks
behind certain organizational efforts used to reward merit.
We discuss possible underlying mechanisms leading to
the paradox of meritocracy effect as well as the scope
conditions under which we expect the effect to occur.•
The idea of meritocracy as a social system in which “merit or
talent is the basis for sorting people into positions and
distributing rewards” (Scully, 1997: 413) has received great
attention since the term was popularized in 1958 by Young
(1994). Advocates of meritocracy stress that in true meritocratic systems everyone has an equal chance to advance and
obtain rewards based on their individual merits and efforts,
regardless of their gender, race, class, or other non-merit
factors. In the United States, for example, survey research
repeatedly reveals that Americans endorse the meritocratic
ethos. Most believe that meritocracy is not only the way the
system should work but also the way the system does work
(Kluegel and Smith, 1986; Ladd, 1994; Ladd and Bowman,
1998). Because meritocracy has been culturally accepted as a
fair and legitimate distributive principle in many advanced
capitalist countries and organizations (Scully, 1997, 2000;
McNamee and Miller, 2004), scholars have sought to assess
the extent to which equal opportunity and meritocratic
outcomes have been successfully achieved in society (e.g.,
Arrow, Bowles, and Durlauf, 2000; Dench, 2006).
Inside organizations, a significant strand of this research
concerns how organizational practices and procedures affect
employees’ opportunities and careers, especially those
practices designed to reduce disparities for women and
ethnic minorities (e.g., Edelman, 1990; Baron, Mittman, and
Newman, 1991; Dobbin et al., 1993; Edelman and Petterson,
1999). Recent empirical studies have found, however, that
workplace inequality persists even with the adoption of
merit-based pay programs (Castilla, 2008), affirmative action
and diversity policies (Kalev, Dobbin, and Kelly, 2006), or
certain popular team and cross-training arrangements (Kalev,
2009). These findings are not surprising to neo-institutional
543/Administrative Science Quarterly, 55 (2010): 543–576
theorists, who have long argued that organizational practices
are adopted in part for symbolic reasons and consequently do
not always accomplish their stated purposes (Edelman, 1992;
Sutton et al., 1994; Kelly and Dobbin, 1998; Edelman, Uggen,
and Erlanger, 1999; Stinchcombe, 2001).
What remains an open question, however, is whether gender
and racial inequality persists in spite of management’s efforts
to promote meritocracy or even because of such meritocratic
efforts. This is an important question given the fundamental
shift to meritocratic employment strategies, such as payfor-performance or merit-based reward practices, over the
past two decades (Heneman and Werner, 2005; Miller, 2006;
Noe et al., 2008). Although these merit-based efforts are
intended to link employees’ rewards directly to their performance, rather than to factors such as seniority or demographic characteristics, there is a growing concern that these
efforts may not actually result in meritocratic outcomes (e.g.,
Roth, 2006; Castilla, 2008). A number of scholars have argued
that organizational pay practices can increase gender and
racial disparities because they introduce bias into employee
compensation decisions (Reskin, 2000; Elvira and Graham,
2002). It may also be the case that not only merit-based
practices but also meritocracy as a cultural value can serve as
an “environmental trigger” (DiMaggio, 1997: 279) or be part
of a “tool kit” of habits (Swidler, 1986: 273) that unleashes
individual cognitive biases. Because employment decisions
are made by managers embedded in organizational cultures,
unintended adverse effects may result from employers’
efforts to reward merit or other practices meant to increase
fairness in the workplace.
Consistent with these research insights, recent scholarship
has demonstrated that merit-based pay practices in particular
may fail to achieve race or gender neutral outcomes, with
results showing that women and minorities (in the same job
and work unit, with the same supervisor, and the same
human capital) received lower salary increases than white
men, even after they are given the same performance
evaluation score (Castilla, 2008). Because previous empirical
studies have evaluated workplace inequality after the introduction of these practices (e.g., Castilla, 2008; Manning and
Swaffield, 2008), however, research has not been able to
successfully answer the question of whether the introduction
of organizational cultures and practices aimed at promoting
meritocracy can cause bias in organizations.
The goal of this article is to investigate the causal link
between merit-based organizational efforts and their employment outcomes at the level of individuals involved in making
these decisions. We develop and test our key hypothesis that
managers making decisions on behalf of organizations that
emphasize meritocracy will ironically show greater bias in
favor of men over equally performing women in the translation of performance into bonuses than managers in organizations that do not emphasize meritocracy. Drawing on the
culture and cognition tradition, we suggest that organizations
promoting meritocracy as a cultural value can lead to unintended behaviors, in part by triggering managers’ stereotypes
and other schematas (Swidler, 1986; DiMaggio, 1997) when
544/ASQ, December 2010
Paradox of Meritocracy
making their employment decisions. This is what we call the
“paradox of meritocracy,” in which emphasizing meritocracy
as an organizational value to reward employees fairly may
result in the opposite outcome. We test our paradox of
meritocracy hypothesis directly with three different experiments (with a total of 445 participants across all three studies)
in which individuals with managerial experience are asked to
play the role of managers in a hypothetical organization and to
evaluate and compensate employees based on their performance reviews. In our study, we experimentally manipulate
both the gender of the employees being evaluated and
whether the company’s core values emphasize meritocracy in
the organization.
THE PARADOX OF MERITOCRACY
The concept of meritocracy as a distributive mechanism
resting on equal opportunity and merit has broad cultural
appeal (Scully, 1997, 2000; McNamee and Miller, 2004). As a
result, many scholars have been interested in understanding
to what extent equal opportunity and meritocratic outcomes
have been achieved.
Inside organizations, employment strategies aimed at linking
merit to employees’ careers, such as pay-for-skill and
pay-for-performance reward systems, are often portrayed as
variations on meritocracy (Scully, 1997: 413). Merit pay is
seen as an important symbol of an organization’s culture,
emphasizing that work is to be rewarded on the basis of
performance alone, rather than other considerations, such as
equality, need, or seniority (Heneman and Werner, 2005: 9).
But results of empirical studies that control for employee
performance have recently called into question whether the
introduction of meritocratic (or merit-based) reward practices
and routines in organizations helps to remedy gender and
racial disparities in wages in the workplace (e.g., Elvira and
Graham, 2002; Castilla, 2008).
The persistence of gender and racial inequality in wages is
especially puzzling given the claims that some type of meritbased or incentive pay practices are widespread among
employers (Heneman and Werner, 2005; Miller, 2006; Noe
et al., 2008). According to a comprehensive survey of personnel procedures used in 826 firms in the United States, there
has been a sharp rise in the percentage of companies using
performance evaluations at the workplace, from approximately
45 percent in 1971 to more than 95 percent in 2002 (Dobbin,
Schrage, and Kalev, 2008). According to the Hewitt Associates
salary survey in 2002, 90 percent of the large organizations
surveyed already had a merit pay plan in place (Hewitt Associates, 2002, cited in Heneman and Werner, 2005).
These organizational strategies aimed at promoting meritbased reward systems in companies have also received great
support in both scholarly and practice-oriented communities.
Some practitioners encourage employers to use performancereward systems (Scharinger, 2002) and highlight the idea that
strengthening the tie between rewards and performance
evaluations increases job satisfaction and motivates
employees to work hard (Lazear, 1998; Martocchio, 2004;
545/ASQ, December 2010
Milkovich and Newman, 2004). These programs can also
attract more able workers by paying them a wage that better
reflects their performance (Lazear, 2000). Many workers find
that these practices give them greater opportunities for
advancement (Osterman, 1999) or at least create an “illusion
of opportunity” that can also be motivational at the workplace
(Ospina, 1996).
Less well understood is whether these merit-based reward
practices successfully link employees’ compensation directly
to their performance evaluations and productivity, thereby
reducing the influence of stereotypes and other workirrelevant factors. In particular, we know little about the
impact of promoting meritocratic cultures and practices on
inequality in employee wages and attainment. The suspicion
that adopting these merit-based pay practices in organizations, especially those that promote meritocracy, may not
solve inequality in the workplace is not new (e.g., Kalev,
Dobbin, and Kelly, 2006; Castilla, 2008; Kalev, 2009). Broadly,
the claim that organizational bureaucracies and routines may
even serve to exacerbate or institutionalize gender and racial
inequality in the workplace has long been established (e.g.,
Kanter, 1977; Edwards, 1979; Acker, 1989, 1990). Scholars
interested in studying the transformation of the employment
relationship and new “market-driven” employment arrangements have also raised equity and fairness concerns about
these practices (e.g., Jacoby, 1985; Kochan, Katz, and
McKersie, 1986; Cappelli et al., 1997; Cappelli, 1999;
Osterman et al., 2001; Dencker, 2009). However, past work
has not tested the causal effect promoting meritocracy might
have on biases in reward decisions.
Under certain circumstances, organizations that emphasize
meritocratic values and beliefs may unintentionally introduce
bias and create inequity in the distribution of employee
rewards. In a recent examination of pay practices, Castilla
(2008) showed that the implementation of an ostensibly
meritocratic performance-reward system, designed to give
workers extra compensation based on their performance, did
not eliminate gender and racial bias in earnings. The large
service organization studied had recently introduced a twostage performance-reward process. In the first stage, supervisors meet employees annually and evaluate their performance.
In the second stage, based on those performance evaluations,
the employee may be recommended for a bonus by a manager superior to the rater. Castilla (2008: 1479) found what he
called “performance-reward” bias: even though performance
evaluations were the most important predictors of employees’
salary increases and bonuses every year (in stage 2), significant effects for demographics were found on salary growth.
Overall, salary increases were significantly lower for women,
ethnic minorities, and non-U.S.-born employees when compared with white men with the same performance evaluation
scores, in the same job and work unit, with the same supervisor, and the same human capital. Notably, this penalty
occurred even after the organization signaled that it strongly
valued and supported meritocracy at the workplace by implementing a performance-reward program that linked employees’ performance with the size of pay increases. Similarly,
546/ASQ, December 2010
Paradox of Meritocracy
using data from a financial corporation, Elvira and Graham
(2002) reported a 25 percent difference in performance-based
bonuses, also distributed at managers’ discretion, between
women and men in the same jobs.
Because these field studies focused exclusively on organizations after the introduction of the merit-based bonus system,
however, these findings cannot determine whether ascriptive
inequality in the distribution of bonuses persisted in spite of
management’s efforts to introduce a merit-based reward
system or because of these efforts. It could also be that the
race or gender effects found in real settings reflect some
unobserved heterogeneity, either for the employees or for the
features of the organization. Our goal in this article was to
experimentally test whether emphasizing meritocratic values
at the organizational level may actually introduce bias in favor
of men over equally performing women in translating performance into bonus amounts. In our study, we focused specifically on how organizations may attempt to promote particular
meritocratic values among their managers and employees,
which is consistent with one dimension of the broad definition of culture in DiMaggio (1997).
Although our prediction of greater bias in monetary rewards
under meritocratic cultures may seem counterintuitive, it is
consistent with broader scholarship in this area. For example,
the important “myth and ceremony” argument made by
Meyer and Rowan (1977) highlighted that organizational
procedures and structures are often designed to be “rituals.”
They are adopted symbolically to gain legitimacy but can be
inefficient or ineffective, not necessarily accomplishing their
stated purpose (e.g., Edelman, 1992; Sutton et al., 1994; Kelly
and Dobbin, 1998, 1999; Stinchcombe, 2001). Consistent
with these neo-institutional predictions, studies have shown
that organizational practices aimed at reducing ascriptive
inequality do not always work (e.g., Edelman, 1990; Baron,
Mittman, and Newman, 1991; Dobbin et al., 1993; Edelman
and Petterson, 1999). Recent empirical work has shown that
workplace inequality remains even after the adoption of
affirmative action and diversity policies (Kalev, Dobbin, and
Kelly, 2006). Although institutional accounts suggest that
practices may fail to accomplish their stated purpose, they
generally do not go so far as to predict that these practices
may accomplish the opposite.
The prediction that emphasizing meritocracy may actually
have a paradoxical effect is in accordance with research on
the link between culture and cognition. The insight is that
cultures play a key role in shaping cognitive processes (e.g.,
Swidler, 1986; DiMaggio, 1997), with studies showing that
specific elements of local cultures can trigger individual
cognitive and interactional biases against low-status groups
(Ridgeway, 1997; Correll and Ridgeway, 2003; Turco, 2010).
Relevant to this prediction, recent studies on cognitive bias
and stereotyping have found that in contexts in which people
are led to feel that they are unbiased, fair, or objective, they
are more likely to then behave in biased ways (Monin and
Miller, 2001; Crandall and Eshleman, 2003; Uhlmann and
Cohen, 2005, 2007; Effron, Cameron, and Monin, 2009;
Kaiser et al., 2009). For example, people given a chance to
547/ASQ, December 2010
disagree with a set of sexist statements (Monin and Miller,
2001) or primed to feel objective (Uhlmann and Cohen, 2007)
have been found to be more likely to recommend a male over
a female candidate in experimental hiring scenarios.
Drawing on the culture and cognition tradition, we suggest
that employers’ efforts to promote meritocratic beliefs or
cultures in organizations may ironically yield unintended
negative consequences, perhaps by leading individuals to feel
unbiased, fair, or objective, and as a result become more
likely to express individual bias toward low-status groups of
employees. In the case of gender, we thus predict that
managers making decisions on behalf of an organization that
emphasizes meritocracy will show greater bias in favor of
male employees than managers making decisions on behalf
of an organization that does not emphasize meritocracy. In
particular, we identify and test this “paradox of meritocracy”
effect, whereby emphasizing meritocracy has the causal
effect of increasing ascriptive bias in the distribution of
monetary rewards. Our main hypothesis is as follows:
Hypothesis: Participants in an organization that emphasizes meritocracy as a core organizational value will show greater levels of
ascriptive bias in translating employee performance evaluations into
monetary bonuses than participants in an organization that does not
emphasize meritocracy.
We conducted three experimental studies designed to test
our hypothesis, the first focusing specifically on whether
there is a paradox of meritocracy. The next two studies
further assess the paradox of meritocracy finding. Because an
empirical examination of the potential underlying mechanisms
leading to the paradox of meritocracy effect is beyond the
scope of our study, we consider them theoretically in the
discussion section.
STUDY 1: THE PARADOX OF MERITOCRACY
We first tested our hypothesis with an experimental study in
which participants, who played the role of employee managers in a fictitious large service organization in the United
States, read a set of employee performance reviews and
evaluated the employees on a number of career dimensions.
The study employed a 2 × 2 mixed factorial design that
manipulated (1) the apparent extent to which a performance
evaluation system was meritocratic (meritocratic or nonmeritocratic, between subjects) and (2) the gender of the
person being evaluated (male or female, within subjects).
Participants were asked to make compensation decisions
based on yearly employee performance reviews. Participants
were randomly assigned to receive one of two different sets
of organizational core values, one set that emphasized
meritocracy (the “meritocratic” condition) versus another
(neutral) set that did not emphasize meritocracy (the “nonmeritocratic” condition).
Participants then examined three employee profiles. Two of
the profiles were “test profiles” and included one male
employee and one female employee with similar performance
evaluations. These test profiles formed the basis of our
analysis. We also included one “filler” profile, a male
548/ASQ, December 2010
Paradox of Meritocracy
employee with a lower performance evaluation score. The
filler profile was included to reduce suspicion that the study
was about gender bias. Participants decided the size of the
bonus, if any, each employee should receive. Participants also
evaluated the profiles on other measures, including recommendations about promotion and retention. This design
allowed us to test whether believing that the organization is
meritocratic increases the level of gender bias in the managerial decision-making process.
Method
Participants. The data for this study were collected in three
sessions at a business school in a private university in the
northeastern United States. Session 1 was conducted as an
optional in-class exercise for masters of business administration (MBA) students and included 95 participants (67 male
and 26 female; two did not answer the question on gender).
Sessions 2 and 3 were conducted as an optional prepresentation exercise for a group of students and managers
attending either an MBA program or a similar business degree
program. Attendees were interested in learning about
performance-reward systems in the workplace. Sessions 2
and 3 were conducted at the same university but, by design,
did not include any of the participants in session 1. Session 2
included 68 participants (48 male and 20 female). Session 3
included 66 participants (48 male and 18 female). The final
sample in our first study thus included 229 individuals
(163 male and 64 female).
Unlike many social science experiments, which rely on undergraduate participants, this study employed MBA students with
substantial work and managerial experience. Although these
participants are in limited supply, thus not permitting the
extensive experimental permutations possible in some
research using undergraduate samples, this approach potentially offers more realism in its assessment of the ways
managers with different professional backgrounds evaluate and
compensate their workers. Additionally, one of the many goals
of these MBA programs is to prepare MBA students to fill
positions with supervisory and managerial responsibilities and
to play an active role in employee performance evaluations.
The average age of participants was 29.71 (with a standard
deviation of 3.89 years); they had an average of 5.80 years of
work experience (with a standard deviation of 3.36 years).
Approximately 4 percent of the respondents had already
earned an MBA degree; the remainder were currently enrolled
in an MBA program. Additionally, 80.4 percent of participants
had previously worked as a manager, and the average participant had 2.4 years of management experience (with a standard
deviation of 2.6 years). About 78.3 percent of respondents
reported liking jobs with supervisory responsibilities (with
5.3 percent not liking them, and 16 percent not knowing yet
whether they would like jobs with supervisory duties).
Procedure. The procedure across sessions was identical.
Participants were invited to participate in a “Management
Personnel Decision-Making Exercise” as part of an educational unit on a similar topic. Participants who volunteered to
549/ASQ, December 2010
participate received a packet that included the study
materials. The experimenter gave a verbal overview of the
study and then participants read and completed the packet.
The instructions asked participants to play the role of a
manager at “ServiceOne,” a large service-sector employer,
and to make personnel decisions similar to those made at
actual firms. Following the instructions, participants read a
company description for ServiceOne. They were informed,
accurately, that the description was based on a real firm that
one of the authors had worked with (with many details
changed in order to protect the firm’s identity). In this section,
participants also read a description of how performance
evaluations and compensation decisions are made at
ServiceOne, described in more detail below.
Following the company description was the experimental
manipulation of meritocracy. We manipulated the apparent
level of meritocracy by providing participants with a list of
“Core Company Values” that either did or did not emphasize
meritocracy in the performance-reward system (see below).
Participants then examined three employee profiles, including
two equivalent “test profiles” that varied in gender, as well as
the filler profile, which was always male. After participants
examined the three profiles, they first evaluated each
employee on a range of measures, including our key dependent variable (the bonus amount decision), and then they filled
out a set of “final reflections” that included our manipulation
checks. Participants were informed from the outset that the
profiles were fictional. After the experiment was conducted,
the exercise was integrated into a class discussion and was
immediately followed by a cautionary lesson for participants
to learn about the unintended consequences of using
performance-reward systems.
Company description. The company description for
ServiceOne included both general details about the firm and
specific details about the performance evaluation process.
ServiceOne was described as a large private service-sector
organization in an urban area in North America, focusing on
research and information technology. The description included
information about the types of jobs available at ServiceOne
and age and tenure demographics for employees. Participants
were asked to play the role of a manager in charge of a small
work group of consultants.
The company description also included information about the
evaluation procedure that participants would use when
considering the employee profiles. We focused on two-stage
evaluation processes such as those described by Castilla
(2008) in which (1) one manager or immediate superior
evaluates an employee’s performance, and then (2) a second,
different manager uses this evaluation to determine whether
the employee receives a raise and, if so, how much. Participants played the role of the second manager. This meant that
participants received the performance evaluations of three
employees in their work unit and used them to make
managerial decisions about the bonuses, promotion, and
termination for these employees at the end of the fiscal year.
We used this company setting for several reasons. First,
laboratory research has generally focused either on
550/ASQ, December 2010
Paradox of Meritocracy
performance evaluations alone or on simultaneous performance-reward evaluations. Examining the two-stage evaluations allowed us to study a frequently used organizational
procedure about which relatively little is known in the inequality literature. More importantly, the two-stage procedure
replicated in our study is widely advocated by employers and
human resource specialists for making pay decisions (e.g.,
Campbell, Campbell, and Chia, 1998; Mathis and Jackson,
2003; Burke, 2005; for a review, see Bretz, Milkovich, and
Read, 1992; Heneman and Werner, 2005).
Second, practitioners increasingly view the two-stage evaluation system as more desirable than single-stage evaluation
systems. Many have defended separating performance
appraisals and salary discussions into two separate stages
mainly because decoupling these two processes and
strengthening the tie between the performance evaluations
of employees and their career outcomes are generally seen
as more meritocratic. Work has suggested that this decoupling encourages employees’ perception of merit, increases
job satisfaction, and is motivational (Gerhart and Rynes, 2003;
Martocchio, 2004; Milkovich and Newman, 2004). Finally,
findings of bias in the experimental study would dovetail with
recent case studies showing that real-world organizations that
use these two-stage performance-reward procedures exhibit
pay gaps based on workers’ race, gender, and national origin
(e.g., Castilla, 2008, in the United States; Manning and
Swaffield, 2008, in the United Kingdom).
Meritocracy manipulation. To manipulate whether the
organization was presented as meritocratic, we varied the
information that participants received about ServiceOne’s
company values. For each condition, participants read a form
describing five “Core Company Values at ServiceOne.” To be
as realistic as possible, we drew on information from a real
organization’s core values introduced to emphasize meritocracy at the workplace as one of the most basic aspects of
an organizational move toward achieving meritocracy and
also as a cultural symbol signaling that work was to be
rewarded on the sole basis of performance. This approach
also allowed us to test directly the potential causal effect of
promoting a merit-based culture on employee bonuses. In
the meritocratic condition, the core values emphasized
fairness and compensation based on performance. These
meritocratic core values statements were as follows: (1) “All
employees are to be rewarded fairly”; (2) “whether employees deserve a raise is determined by their performance”;
(3) “raises and bonuses are based entirely on the performance of the employee”; (4) “promotions are given to
employees when their performance shows that they deserve
it”; and (5) “ServiceOne’s goal is to reward all employees
equitably every year.”
In the non-meritocratic condition, the core values did not
indicate meritocratic values; instead, they emphasized the
regularity of evaluation and managerial autonomy. We refer to
this condition as the non-meritocratic condition simply
because this condition does not emphasize fairness or compensation based on employee performance as the meritocratic
condition did. To be conservative, the non-meritocratic
551/ASQ, December 2010
condition was designed to be neutral but not explicitly antimeritocratic; thus the possibility of bias or discrimination in
evaluations was not raised. The non-meritocratic core values
statements were the following: (1) “All employees are to be
evaluated regularly”; (2) “whether an employee deserves a
raise is determined by their manager”; (3) “raises and bonuses
are to be given based on the discretion of the manager”;
(4) “promotions are to be given to employees when their
manager decides that they deserve it”; and (5) “ServiceOne’s
goal is to evaluate all employees every year.”
To make sure participants read and considered each of the
core values statements carefully, we asked them to indicate
whether they agreed with each value by placing a check mark
on a line next to each statement. Participants were asked to
indicate whether they agreed with the values statements so
that they would feel as if their goals were the same as those
of the company and thus would behave like an actual manager at that firm. Requesting that participants indicate agreement makes the manipulation similar to the “moral
credentials” manipulation introduced by Monin and Miller
(2001). One distinction between the latter study and ours is
that our participants agreed with statements about the
general values of the company rather than about their specific
beliefs about gender or other bases of moral credentials.
Following the meritocracy manipulation, participants examined the three employee profiles.
Employee profiles. Participants examined three profiles,
including two equivalent test profiles that varied in gender
and one filler profile that was always a low-performing male
employee. Each profile was presented using a “Performance
and Staff Development Evaluation Form,” which included a
quantitative assessment of each employee on a 1–5 scale as
well as qualitative comments from the employee’s immediate
supervisor. All employees had the same title, “Consultant,”
worked in the same unit, “Product Development,” and had
the same supervisor.
We manipulated the employee’s gender by using maleand female-typical names on the profiles. We chose
gender-typical names from a list tabulating the most common
names for men and women in the United States and then
paired them with common last names (from the genealogy of
names in the census.gov Web site). The names of our test
profiles were Patricia Anderson and Michael Taylor, and the
name of our filler profile was Robert Miller. To test our
hypothesis, it was key that the test profiles were of equivalent merit, but not so similar as to raise participants’ suspicion
that studying gender bias was a goal of the research. To
accomplish this, we gave each test profile equal quantitative
performance scores, similar but not identical qualitative
comments, and counterbalanced the qualitative comments
across profiles. The 5-point quantitative assessment scale
was labeled “Summary of Performance,” and each of the
5 levels was labeled with a descriptive phrase. The two test
profiles received a score of 4 on the 5-point scale. This score
included the descriptive label “Staff member’s performance
consistently meets and frequently exceeds all established
goals/expectations for the position.”
552/ASQ, December 2010
Paradox of Meritocracy
The profiles also included two types of qualitative feedback:
areas in which the employee performed well (praise) and
areas in which the employee’s work needed improvement
(criticism). The test profiles each included three sentences of
praise and two sentences of criticism. For one test profile,
the praise read, “Michael/Patricia is hardworking and quick to
find ways to solve clients’ problems. He/She is also generally
popular with the clients. Michael/Patricia reliably completes
projects on time.” For the other test profile, the praise read,
“Michael’s/Patricia’s proposals are always well thought-out
and highly detailed. He/She always does an excellent job of
communicating technical aspects of the proposals to clients.
Clients respect and enjoy working with Michael/Patricia.” The
criticism for one of the test profiles read, “While the quality of
Michael’s/Patricia’s work is excellent, several projects this
year have gone over budget. In the next appraisal cycle, he/
she needs to work on keeping costs down.” The criticism for
the other test profile read, “Michael/Patricia is a valuable
team member, but sometimes tries to take on too many
projects at once. In the next year, he/she needs to work on
staying focused.” The qualitative comments were counterbalanced across the two test profiles: each set of comments
was randomly assigned to the male test profile for half of the
sample and to the female test profile for the other half. This
ensured that any differences in the evaluations of the qualitative comments did not bias the results because the
employee’s gender was uncorrelated with which comments
he or she received.
We included a third filler profile, named “Robert Miller,” to
further reduce suspicion that gender bias was a focus of the
study. With three profiles, gender may less obviously differentiate the profiles than if participants rated two very similar
profiles that only differed by gender. The third profile was
designed to be clearly less qualified than the two test profiles
so as not to compete with the test profiles on the ranking
variables. The test profile was always rated a 3 out of 5 on
the quantitative evaluation. The “3” rating was labeled “Staff
member’s performance consistently meets established goals/
expectations for the position.” The praise for the filler profile
was similarly lukewarm, reading, “Robert does a good job of
listening to the clients and meeting their expectations. His
work has been consistently solid, but not spectacular.” The
criticism for the filler profile always read, “Robert has a
tendency to miss minor deadlines when things get busy. He
needs to do a better job of staying on top of his projects.”
Dependent measures. Our hypothesis predicted that people
will be more likely to engage in gender bias in the translation
of performance evaluations into rewards when the organization presents itself as meritocratic. To test this argument, we
asked participants to assign a yearly bonus to each employee.
They were told that they had a limited pool of resources
($1,000) from which to assign the bonus, to be divided among
the three employees. To determine whether other employee
outcomes are also affected by perceptions of meritocracy, we
asked participants to rate each employee on four additional
items, using a set of 7-point Likert-type scales. The first of
these questions read, “Do you think hiring this employee was
the right decision?” and was anchored at “definitely wrong
553/ASQ, December 2010
decision” and “definitely right decision.” Similar questions
asked to what extent the employee should be considered for
promotion or termination and whether the employee would
be successful in the future. We also collected variants of
these questions, which asked participants to choose only one
employee profile as most deserving of a bonus, promotion, or
retention, and as having the greatest potential for success.
After completing and submitting the rating part of the experiment, participants were asked to fill out a “final reflections”
questionnaire. They were asked to indicate their beliefs about
the performance evaluation process and about ServiceOne as
a company. This included our key manipulation check questions, the extent to which ServiceOne as a company was
meritocratic and fair (again using 7-point scales). We expected
participants to rate the company as more meritocratic and fair
in the meritocratic condition than in the non-meritocratic
condition. To determine whether this belief was limited to the
company, we also asked participants to evaluate (1) the
extent to which the particular supervisor’s evaluations were
accurate, competent, and fair, and reflected knowledge of the
employees and effort, (2) whether the performance evaluation process itself was viewed as accurate and fair, and
(3) whether, as an employee, they would like to be evaluated
using a similar process.
Results
Manipulation check. We first checked whether our manipulation successfully convinced participants that ServiceOne was
more meritocratic and fair in the meritocratic condition
compared with the non-meritocratic condition and found that
participants did rate ServiceOne as more meritocratic under
the meritocratic condition than under the non-meritocratic one
(meritocratic condition mean = 4.05 vs. non-meritocratic
condition mean = 3.57, t-value = 2.609, significant at the
.01 level). Participants also rated the company as more fair in
the meritocratic condition (meritocratic condition mean = 3.67
vs. the non-meritocratic condition mean = 3.25, t-value =
2.565, significant at the .01 level). This indicates that our
manipulation of meritocracy was successful. Further checks
determined that impressions of the company did not generalize to the performance appraisal process or to the supervisor.
Thus we did not find any significant differences in the supervisor’s accuracy, competency, knowledge of employee, or
fairness across conditions, nor did we find any differences in
ratings of the accuracy and fairness of the company’s evaluation process or willingness to be evaluated using this process.
This indicates that the manipulation successfully shaped
beliefs about the organization, not beliefs about the supervisor or the evaluation process itself.
The paradox of meritocracy effect. We assessed our main
hypothesis by examining the bonus-amount decision for each
of the test profiles by apparent meritocracy condition. We
expected to find greater levels of gender bias in the meritocratic condition than in the non-meritocratic condition in the
form of a lower bonus for women. To test this claim, we
began by comparing the bonus amounts for the equally
performing male and female test profiles within each
554/ASQ, December 2010
Paradox of Meritocracy
Figure 1. The paradox of meritocracy in the distribution of rewards by employee gender, study 1 (N = 229).
$430
$419.69
$420
$418.75
Male
Female
Mean Bonus in $
$410
$400
$390
$380
$372.37
$370
$368.16
$360
$350
$340
Non-meritocratic
Condition
Meritocratic Condition
2 × 2 factorial design: ANOVA F-test (Gender × Meritocracy interaction) = 18.79 (p = .000).
condition. We also calculated a paired t-test to determine
whether the difference varied within each condition. Given
the known problems of using difference scores as dependent
variables (Edwards, 2001), we interpreted the ANOVA results
using bonus amount as the dependent variable: we hypothesized that the meritocracy manipulation would interact with
the gender of the employees who were being evaluated to
influence the bonus. Such interaction is properly tested by an
ANOVA using bonus amount as the dependent variable.
1
We found no significant difference in the
bonus amount assigned to the filler profile
between the two meritocracy conditions.
The average bonus for the filler profile
was $159.23 in the meritocratic condition
and $150.07 in the non-meritocratic
condition. The t-test for the bonus
difference of $9.16 was not significant
(t-value = –.739, p = .46), suggesting that
the meritocracy condition did not
significantly affect the bonus rating of the
filler profile across conditions.
The results of our analyses are presented in figure 1. In the
non-meritocratic condition, we found that women, on average,
earned a bonus $51 higher than equally performing men,
significant at the .01 level. By contrast, we found the opposite
pattern in the meritocratic condition: men earned, on average, a
bonus $46 higher than equally performing women, significant at
the .01 level.1 Because the t-tests evaluate differences within
each condition only, to test for differences across condition, we
also performed a 2 × 2 factorial ANOVA (repeated measures for
the male and female test profiles) to test our main hypothesis
regarding the bonus. Consistent with our hypothesis, this
analysis yielded an interaction effect for gender and meritocracy
such that women were paid less than men in the meritocratic
condition, but not in the non-meritocratic condition (F = 18.792,
p = .000). Consequently, participants in a performance-reward
system that emphasizes meritocracy significantly favored men
over women in the translation of employee performance into
bonus amounts. There was neither a significant gender main
effect (F = .052, p = .82) nor a meritocracy main effect
(F = .027, p = .87) on the bonus. We therefore find strong
support for our main hypothesis.
555/ASQ, December 2010
We did not predict a tendency for participants to give women
a higher bonus than men in the non-meritocratic condition.
Though this result does not contradict our main prediction,
we considered reasons for this pattern in detail below, in our
third study.
Robustness checks. We conducted several additional analyses to ensure that the results were robust as well as to further
investigate our findings. First, we estimated the analyses
separately for each of the three experimental sessions. We
found results substantially similar to those reported. Additionally, we analyzed the data separately by the gender of the
participants. Regardless of participants’ gender, we found
strong support for our main hypothesis. The interaction effect
for gender and meritocracy on the bonus was significant for
male participants (N = 163; F = 11.121, p = .001) and female
participants (N = 64; F = 7.273, p = .01).2
To further evaluate the robustness of our findings, we estimated a series of multivariate regression models that included
participants’ characteristics as control variables. For each
participant, regardless of meritocratic condition, we computed
the difference in the amount of the bonus between the male
and the female test profiles and then used that difference as
the dependent variable for the regression analyses. This
difference provides an absolute measure of “rating” bias in
favor of men. In addition to examining the main effect of
meritocracy, we included a number of control variables,
including participants’ gender, age, and years of management
experience, as well as their ratings of the sources of
employees’ success. Consistent with our main hypothesis, the
meritocracy manipulation was always statistically significant
and in the predicted direction for the bonus amount.3
2
We ran ANOVA (repeated measures for
the male and female test profiles) to test
for the three-way interaction of gender of
the employee × gender of the participant ×
meritocracy condition. This interaction
was not significant, providing further
support for the finding that the interaction
effect for gender and meritocracy on the
bonus does not depend on the gender of
the participant (p = .98). We also ran our
ANOVA analysis controlling for the bonus
amount that participants gave to the filler
profile. Though the effect of this variable
was negative and significant at the
.05 level—signaling that the higher the
amount of money given to the control
profile, the lower the amount of money
participants gave to our two test profiles,
which is not surprising, given the $1,000
budget constraint—including such control
variables in the models did not change our
main results at all. This model’s
interaction effect for gender and
meritocracy on the bonus was still
significant (F = 17.725, p = .001).
3
Because the regression results were
largely redundant with the findings
presented in prior ANOVA analyses, they
are omitted but are available upon
request.
Although the results of study 1 support our hypothesis, one
alternative explanation for our findings is that participants
might have made certain gender attributions and interpreted
the language in the organizational values statement differently
in the meritocracy condition. In particular, participants may
have interpreted the emphasis on “equity” and “fairness” in
the meritocratic condition as a rhetorical device actually
signaling a preference for women. Along these lines, experimental research has found that, under certain conditions,
preferential selection methods can produce a backlash toward
the beneficiaries (Heilman, Block, and Lucas, 1992; Heilman,
McCullough, and Gilbert, 1996). If so, participants in the
meritocratic condition might have assumed that the female
test profile was held to more lenient standards and was more
likely to have achieved her performance rating through help
from others or some source other than her own productivity,
ability, or effort. For example, research suggests that men’s
performance tends to be attributed to skill, while women’s
tends to be attributed to luck (Deaux and Emswiller, 1974).
To address this alternative explanation, we collected additional
measures right after the administration of the employee
reward questionnaires. Following Pazy (1986), we asked
participants to evaluate, using 7-point Likert type scales, why
each employee was successful, along five dimensions:
(1) ability and talent, (2) effort and hard work, (3) luck or
556/ASQ, December 2010
Paradox of Meritocracy
chance, (4) easiness of their job, and (5) help they have
received. Participants were also asked to indicate which factor
was most responsible for each employee’s success. If the
meritocratic condition did lead participants to believe women
were evaluated according to more lenient standards, women
in the meritocratic condition should be evaluated as more
likely to have been successful due to factors other than ability
and talent or effort and hard work. The results of our analyses
of these employee attributions did not support this alternative
explanation. We found no significant gender differences in the
perceptions of sources of employees’ success between
respondents in the meritocratic and non-meritocratic conditions. Consistent with earlier work (Pazy, 1986), we only found
significant that women’s success was more likely than men’s
to be attributed to hard work and effort, but this pattern did
not differ by meritocratic condition. These results suggest that
participants did not perceive women to be evaluated more
leniently in the meritocratic condition.
Other employee outcomes. Our hypothesis focused on the
effect of meritocratic values or beliefs on bias in the distribution of bonuses. One important question is the extent to
which this effect may be found for employees’ other career
outcomes. In general, past research suggests that gender
bias may affect decisions on a wide range of outcomes, such
as hiring, promotion, and salary (e.g., Steinpreis, Anders, and
Ritzke, 1999; Foschi, 2000; Biernat and Fuegen, 2001; Eagly
and Karau, 2002). This is because gender stereotypes draw
on broad-based beliefs about women’s and men’s differential
competence, assertiveness, and other traits generally thought
to be needed for high status or traditionally male occupations
(Eagly and Karau, 2002; Biernat, 2003; Correll and Ridgeway,
2003). As a result, these stereotypes should apply to other
employment outcomes that are related to an individual’s
competence and productivity. Consequently, to the extent
that emphasizing meritocracy at the organizational level
increases the expression of gender stereotypes, we would
also expect to find greater levels of bias in favor of male
employees in a variety of career outcomes.
Recent fieldwork suggests that the effects of emphasizing
meritocracy may be greatest for salary and bonus increases.
Empirically, in studying a company that emphasized the
meritocratic aspect of its performance-reward system, Castilla
(2008) found significant penalties for women, minorities, and
non-U.S. citizens in bonus amounts but not in promotions,
terminations, or the binary decision of whether an employee
deserves a bonus. Theoretically, Castilla argued that this
difference arises at least in part because employee hiring,
promotion, and termination are more visible employment
outcomes (consistent with Petersen and Saporta, 2004).
Employees may not know how much their salaries changed
relative to other members of their unit, but information about
who was hired, promoted, or terminated is more easily
available and observable. Because bias in these career
outcomes is more manifest and therefore easier to detect,
Castilla predicted that it should be less likely to occur in the
workplace. This is also consistent with Kalev, Dobbin, and
Kelly’s (2006) finding that managerial accountability is associated with higher female and minority representation in
557/ASQ, December 2010
managerial jobs. Since the development of employers’
compliance with Title VII and the human resource profession
in the late 1960s (see Reskin and McBrier, 2000;
Stinchcombe, 2001; Kalev, Dobbin, and Kelly, 2006; Dobbin,
2009), we would expect managers to feel more accountable
for their decisions on hiring, base salaries, promotions, and
terminations than for decisions regarding bonuses (Castilla,
2008). This is also in accordance with considerable work in
social psychology indicating that bias is more likely when decision makers feel that their judgment is unlikely to be closely
scrutinized (Tetlock, 1983a, 1983b; Lerner and Tetlock, 1999).
The transparency argument made in these previous studies
aligns closely with the moral credentials argument (Monin and
Miller, 2001), according to which people do not wish to appear
prejudiced to others, or even to themselves (i.e., privately
acknowledge that bias might shape their evaluations of others).
Thus they should be less likely to express bias when that bias
could more obviously call their moral credentials into question.
For example, recommending a somewhat smaller salary
increase for a woman over a similarly qualified man may be
more easily rationalized, and thus pose a smaller threat to
one’s view of oneself as unbiased, than choosing to hire,
promote, or terminate a man over a similarly qualified woman.
Though emphasizing meritocracy should increase bias, the
manifestation of such bias should be stronger for outcomes in
which disparities would be subtler or less noticeable to others.
There are also practical reasons to expect that emphasizing
meritocracy in organizations when implementing
pay-for-performance programs will have the greatest effect
for salary and bonus increases. Such programs typically rely
on performance evaluations for making pay decisions (Institute of Management and Administration, 2000; Burke, 2005),
but they explicitly require additional supply- and demandrelated factors, such as job openings and/or employees’
tenure in the company, and skills, when making promotion or
termination decisions at the firm level (Miller, 2006).
Based on these reasons presented in prior work, our ancillary
prediction is therefore that participants in an organization that
emphasizes meritocracy as a core organizational value will show
lower levels of bias in the translation of employee performance
evaluations into other more observable career decisions than
monetary bonuses. To test this prediction, we collected and
examined four other employee ratings of the test profiles by
meritocracy condition. Because these other career outcome
variables are measured using a different metric than the bonus,
we computed a standardized measure of the paradox of meritocracy effect for each rating variable (i.e., beta coefficients).
This approach allowed us to directly compare the effect of our
meritocracy manipulation on employee bonus versus the
other career variables. For each participant, regardless of
meritocratic condition, we computed the difference in ratings
between the equally performing male and female test profiles
and then used the standardized values of these differences as
the main dependent variables. These differences provide a
standardized measure of the level of bias in favor of men in
the translation of performance evaluations into each
558/ASQ, December 2010
Paradox of Meritocracy
employee rating score. Similar results were obtained when
estimating Cohen’s D (Cohen, 1988). Regardless of the
method used to compare the paradox of meritocracy effect
across the different variables of different scales and magnitudes collected (including ANOVA coefficients and marginal
effects), we consistently found that the levels of bias (in favor
of men) were larger for the bonus amount than for the other
four career outcome variables.
Results are presented in figure 2. Supporting our ancillary
prediction, we found that the tendency for participants to
express bias in favor of men in the meritocratic condition was
large and highly significant for the bonus measure (B = .278,
p = .000). But we found smaller levels (in absolute magnitude)
of gender bias in the meritocratic condition for hiring (B = .101,
p = .064), promotion (B = .082, p = .062), termination
(B = –.123, p = .031), and success in the future (B = .127,
p = .028; all one-sided tests).
Table 1 reports differences in ratings and the relevant paired
t-tests comparing the unstandardized employee career
ratings for the male and female test profiles in each experimental condition; for convenience, we also include the
analysis of the bonus amount in the table. Once again, we
also ran 2 × 2 factorial ANOVAs (repeated measures for the
male and female profiles) and report the interaction term
Figure 2. The paradox of meritocracy: Standardized coefficients by employment career outcome.
0.30 0.278 (p = .000)
Standardized Paradox of Meritocracy Effect
0.25
0.20
(p = .028)
0.127
0.15
0.101 (p = .064)
0.10
0.082 (p = .062)
0.05
0.00
Bonus
Hiring
Promotion
Termination
Success
−0.05
−0.10
−0.123 (p = .031)
−0.15
Note: For the purpose of comparing across career outcomes of different scales and magnitudes, the
standardized measure of the paradox of meritocracy effect is reported for each rating variable (i.e., beta
coefficients). Consistent with our prediction, the penalty for the bonus amount is larger than for hiring,
promotion, termination, or success. P-values (in parentheses) reflect statistical differences from zero and
are calculated for each career outcome variable.
559/ASQ, December 2010
Table 1
Mean Employee Ratings in Meritocratic and Non-meritocratic Conditions by Employee Gender*
Non-meritocratic Condition (N = 116)
Male test
profile
(Michael)
Bonus amount
$368.16
(Total of $1,000 to
(116.03)
distribute)
Hiring decision
5.84
(1 = Definitely
(1.06)
wrong decision;
7 = Definitely right
decision)
Promotion decision
4.86
(1 = Definitely do
(1.31)
NOT promote;
7 = Definitely
promote)
Termination decision
2.12
(1 = Definitely do
(1.21)
NOT terminate;
7 = Definitely
terminate)
Success in the
5.58
future
(0.97)
(1 = Will NOT be
successful at all;
7 = Will be highly
successful)
ANOVA
F-test
(Gender x
Meritocracy
Difference Interaction
(t-value)
Term)†
Meritocratic Condition (N = 113)
Female
test profile
(Patricia)
Difference
(t-value)
Male test
profile
(Michael)
Female
test profile
(Patricia)
$419.69
(122.30)
$51.53•••
(–2.948)
$418.75
(111.38)
$372.37
(102.75)
$46.38•••
(3.267)
6.03
(0.91)
–0.19•
(–1.545)
6.04
(0.68)
5.99
(0.85)
0.05
(0.501)
2.34
5.02
(1.41)
–0.16
(–0.923)
5.18
(1.07)
5.06
(1.23)
0.12
(0.826)
1.52
1.85
(1.01)
0.27••
(2.028)
2.01
(1.19)
2.06
(1.26)
–0.05
(–0.495)
3.51•
5.75
(0.98)
–0.17•
(–1.406)
5.84
(0.76)
5.70
(0.92)
0.14•
(1.319)
3.70•
18.79•••
• p < .10; •• p < .05; ••• p < .01; one-sided t-tests.
* Standard deviations are in parentheses. We calculated paired sample t-tests for the difference in the rating variables
between the male and the female test profiles in each meritocracy condition separately, reported under the Difference (t-value) columns.
† F-values are reported in this column. Based on our ancillary prediction, we expected participants in the meritocratic condition to show lesser levels of bias in the translation of employee performance evaluations into other key
employee career outcomes (when compared with the translation of performance evaluations into bonus amounts)
than participants in the non-meritocratic condition.
between the meritocracy manipulation and the gender of the
employees in column 7 of table 1. The only significant
interaction effects for gender and meritocracy were found for
the termination (F = 3.51, p = .062) and success ratings
(F = 3.70, p = .056).
In addition, we examined whether participants were more likely
to rank the male test profile first across all outcome variables,
by meritocracy condition and gender. We also expected that the
effect of the meritocracy manipulation on gender bias would
be smaller than on the bonus for the more visible measures.
Across all variables, men were always preferred more under the
meritocratic condition than under the non-meritocratic condition.
The preferences for men over equally performing women, while
substantively large, were not significant.
STUDY 2: GENDER COMPOSITION OF THE
EMPLOYEE PROFILES
Study 2, designed to test whether the gender composition of
the employee profiles matters, was nearly identical to study 1,
560/ASQ, December 2010
Paradox of Meritocracy
with the exception that we used a female name (Linda
instead of Robert) for the filler profile. We conducted this
second experiment to rule out one important alternative
explanation for our findings in study 1: the fact that our filler
profile was male in study 1 may have shaped the comparisons made by the participants, leading to a preference for the
male employee in the meritocratic condition.
Individuals often use gender to determine salient referents
for comparison when making evaluations regarding pay and
other career outcomes (Kulik and Ambrose, 1992), comparing males with other males and females with other females
(for a review of work showing how one’s numerical representation in a group affects individual judgment and influence, also see, e.g., Reagans, 2005; Loyd and Phillips, 2006;
Duguid, Loyd, and Tolbert, 2010). In study 1, participants
may have rewarded our male test profile more highly in the
meritocracy condition because they implicitly compared him
with Robert, our low-performing filler profile. When making
such a comparison, it could have seemed more fair or
meritocratic to give the male test profile a larger bonus. In
contrast, our female test profile could not have benefited
from comparison with a low-performing female test profile
and instead may have been compared with the male test
profile, which was virtually identical in terms of apparent
quality. If this alternative argument is true, then changing
the gender of the filler profile from male to female should
reverse the results, producing a greater bonus for women in
the meritocracy condition.
Method
Participants. In study 2, the participants were again
recruited at a business school in the northeastern United
States. The study included 115 participants (70 male and
45 female). Age and managerial experience were similar to
those in the previous study. Participants were on average
29.29 years old (with a standard deviation of 4.20 years) and
had an average of 6.07 years of work experience (with a
standard deviation of 3.76 years). Approximately 8 percent
of the respondents had already earned an MBA. As in study 1,
most respondents (almost 75 percent) reported liking jobs
with supervisory responsibilities, with 4.3 percent reporting
not liking them.
Procedure. With the exception of substituting the
low-performing female filler profile for the low-performing
male filler profile, the procedure was identical to that used in
the prior study.
Results
Manipulation check. Similar to study 1, our manipulation led
participants to perceive ServiceOne as more meritocratic and
fair in the meritocratic condition than in the non-meritocratic
condition (both significant at the .01 level). Participants rated
ServiceOne as more meritocratic (the mean difference
between the two conditions is .738, t-value = 2.641, p < .01)
and more fair (the mean difference was .708, t-value = 2.980,
p < .01) under the meritocratic condition than under the
non-meritocratic one.
561/ASQ, December 2010
Figure 3. The paradox of meritocracy in the distribution of rewards by employee gender, study 2 (N = 115).
$420
$410.50
$412.30
$410
Male
Female
Mean Bonus in $
$400
$390
$378.37
$380
$370
$363.60
$360
$350
$340
$330
Non-meritocratic
Condition
Meritocratic Condition
2 × 2 factorial design: ANOVA F-test (Gender × Meritocracy interaction) = 10.125 (p = .001).
The paradox of meritocracy effect. The results for the bonus
measure are summarized in figure 3. Consistent with study 1,
in the non-meritocratic condition we found that women earn
on average a bonus $47 higher than equally performing men
(p < .01). By contrast, in the meritocratic condition, men earn,
on average, a bonus $34 higher than equally performing
women (p < .01). Also consistent with the analyses of study 1,
this yielded an interaction effect for gender and meritocracy
such that women were paid less than men in the meritocratic
condition but not in the non-meritocratic condition (F = 10.125,
p = .001). The findings thus strongly support our main
hypothesis, demonstrating that the effect does not depend on
the gender of the filler profile.4
4
Once again, we found no significant
difference in the bonus amount assigned
to the female filler profile between the
two conditions (the mean bonus
difference was $21.57, t-value = –.775,
p = .44). The average bonus for the filler
profile was $150.42 in the meritocratic
condition and $128.85 in the
non-meritocratic condition.
As in study 1, the analyses of the ratings variables in study 2 on
hiring, promotion, termination, and success in the future also
supported our ancillary prediction of less ascriptive bias in the
translation of performance evaluations into these other
employee career outcomes. The standardized coefficients
were similar to those reported in table 1: consistent with our
prediction, participants also tended to rate women more
favorably than men in the non-meritocratic condition and to rate
men more favorably than women in the meritocratic condition
on hiring, promotion, termination, and success decisions,
although the effect sizes were lower than for bonuses. For
hiring, promotion, and termination, the interaction effects of
gender and meritocracy were non-significant (for hiring,
F = .762, p = .385; promotion, F = .237, p = .628; and termination, F = 1.668, p = .199). The only significant interaction effect
found was for employee success rating (F = 4.389, p = .038).
Comparing high- and low-performing employee profiles.
Our hypothesis, and by extension our study design, focused
562/ASQ, December 2010
Paradox of Meritocracy
on comparing equally performing male and female
employees, but another possible way of ascertaining the
effect of a culture of meritocracy on merit-compensation
decisions is to compare the same-gender high and low
performers in the meritocratic and non-meritocratic conditions. For simplicity, “low performer” refers to the filler profile
of the employee who received a 3 rating versus “high performer,” which refers to the test profiles of employees who
received a 4 rating. This approach allowed us to assess to
what extent performance differentials are less effective at
generating rewards for women than men in the meritocratic
condition, compared with the non-meritocratic condition. In
other words, we could also examine whether greater performance translates into greater rewards in the meritocratic
condition, regardless of the gender of the employees.
To explore this possibility, we ran some additional analyses.
For study 1 (low-performing male filler profile), we ran an
ANOVA analysis with only the two male profiles, estimating
(a) the main effect of meritocracy, (b) the main effect of being
the test or filler profile (“performance”), and (c) the interaction
of these two. The latter interaction effect tells us whether the
performance effect is significantly greater in the meritocratic
condition than in the non-meritocratic condition for male
profiles. This interaction effect was positive and significant
(F = 4.015, p = .046, two-sided). Both the “performance”
direct effect and the effect of meritocracy were significant for
men (p < .001). We took a similar approach for study 2
(low-performing female), with only the two female profiles.
The interaction effect was not significant for female profiles
(F = 1.422, p = .236). Although the “performance” direct
effect was significant (p < .001), the effect of meritocracy was
not significant for female profiles (F = .041, p = .84).
5
Although a direct comparison cannot be
made between studies 1 and 2 because
participants were not randomly assigned
across studies, we can still approximate
this comparison by merging both datasets
in studies 1 and 2 and running an ANOVA
comparing the same-gender high and low
performers in the meritocratic and
non-meritocratic conditions. We examined
(a) the main effect of meritocracy; (b) the
main effect of being the test or filler
profile (“performance”); (c) the main
effect of whether the compared profiles
are male (study 1 data) or female (study 2
data); (d) all two-way interactions; and in
particular, (e) the three-way interaction of
(a), (b), and (c). The estimated three-way
interaction coefficient (d) indicates that
the “performance” effect is significantly
greater in the meritocratic condition for
men than for women (F = 4.818, p = .029).
Though the “performance” direct effect
(term b) was significant at the .001 level,
the effect of meritocracy and the two-way
interactions were not significant.
Participants in the meritocratic condition
thus showed greater levels of bias in
favor of men in translating employee
performance differentials into bonuses
(in comparison with participants in the
non-meritocratic condition). We thank an
anonymous ASQ reviewer for suggesting
this analysis.
Overall, these findings are consistent with our paradox of
meritocracy hypothesis and indicate that the effect of meritocracy on monetary rewards is significant for men but not
women. In addition, the effect of performance on rewards is
significantly greater in the meritocratic condition than in the
non-meritocratic condition for men, but there is no evidence
of a similar boost for women in the meritocratic condition.5
STUDY 3: THE FEMALE ADVANTAGE IN THE
NON-MERITOCRATIC CONDITION
The results of studies 1 and 2 supported our prediction that
women would receive smaller average bonuses than men in
the meritocratic condition. One unpredicted finding in both
studies, however, was that women received greater average
bonuses in the non-meritocratic condition. Although this
finding does not contradict our hypothesis, it is surprising and
warrants additional attention in a third study.
One possible explanation is that the language about discretion
used in the non-meritocratic condition may have signaled the
possibility of bias on the part of the evaluating supervisors. If
the participants believed that managerial bias in the evaluation
system disadvantaged women, they may have felt they
needed to compensate or correct for this bias by favoring
women (consistent with Petty and Wegener, 1993;
563/ASQ, December 2010
Wegener and Petty, 1995).6 In particular, the participants in
studies 1 and 2 may have been responding to the language
emphasizing managerial discretion in the “Core Values”
statement of the non-meritocratic condition (e.g., “raises and
bonuses are to be given based on the discretion of the
manager”). If participants perceived more managerial
discretion in the non-meritocratic condition, they may have
suspected that the performance evaluations they received
were biased in favor of male employees and compensated for
this bias by awarding a larger bonus to the female test profile.
If this explanation is correct, then removing the emphasis on
managerial discretion in the non-meritocratic condition should
result in equal bonuses for the male and female test profiles.
Method
Updated non-meritocratic condition. To test the effect of
an emphasis on discretion, we first constructed a new
non-meritocratic condition designed to be less discretionary
than the non-meritocratic control condition used in studies 1
and 2. In the new condition, the statements on the “Core
Company Values” form read as follows: (1) “All employees
are to be evaluated regularly”; (2) “performance evaluation
forms include a quantitative as well as qualitative component
about the employee’s performance”; (3) “performance
evaluations are part of the employee’s official personnel file”;
(4) “performance evaluations are discussed with each
employee every year”; and (5) “ServiceOne’s goal is to
evaluate all employees every year.” We refer to this as the
“updated non-meritocratic condition.”
6
Another possibility is suggested by work
on “aversive racism” (Gaertner and
Dovidio, 1977; Gaertner et al., 2005),
which has found that individuals favor
stigmatized groups when concerned
about appearing prejudiced. Because
gender discrimination is frowned upon in
organizations, especially among managers
(see, e.g., Dobbin, 2009), participants in
the non-meritocratic condition may have
awarded the female employee a larger
bonus to avoid the perception that they
were biased. Given the language of
discretion used in the non-meritocratic
condition, however, we decided to first
test the overcorrection explanation
described above in study 3.
We next conducted a pretest of all three “core values”
statements to evaluate two key assumptions: first, that the
original non-meritocratic condition was perceived as more
discretionary than the meritocratic condition; and second, that
the updated non-meritocratic condition and the meritocratic
condition would be perceived as equally discretionary. We
asked 21 participants (undergraduate students at a public
university in the Midwest) to read and rate the three “Core
Company Values” statements. They were asked to assess
the level of discretion that managers working for an organization with each set of values would possess. The values
statements were rated on a scale of 1 (“Very little discretion”)
to 7 (“A great deal of discretion”). As expected, participants
rated managers in the original non-meritocratic condition as
having significantly greater discretion than managers in the
meritocratic condition (mean = 6.24 vs. mean = 3.67, p < .01,
paired t-test, two-tailed). Our updated non-meritocratic
condition successfully reduced the perceived level of managerial discretion, being rated as significantly less discretionary
than the original non-meritocratic condition (updated nonmeritocratic condition mean = 3.95, p < .01, paired t-test,
two-tailed). Importantly, the updated non-meritocratic condition and the meritocratic condition were rated as equally
discretionary (p = .52, paired t-test, two-tailed). The results of
the pretest thus confirmed both our assumptions.
Participants. Study 3 included 101 participants (62 men and
39 women), again recruited at a business school in the
northeastern United States. Similar to the previous studies,
564/ASQ, December 2010
Paradox of Meritocracy
participants were 30 years old on average (with a standard
deviation of 3.5 years); they had an average of 5.73 years of
work experience (with a standard deviation of 3.56 years); and
most of them (71.6 percent) reported liking jobs with supervisory responsibilities (with 3.9 percent not liking them, and
24.5 percent not knowing yet whether they would like jobs
with supervisory duties).
Procedure. The procedure in study 3 was identical to that
used for study 1, with the exception that the updated nonmeritocratic condition was used in place of the original
non-meritocratic condition with the five non-discretionary core
values statements, as described above.
Results
Manipulation check. Our meritocracy manipulation was
successful. As in the previous two studies, we found that
participants rated ServiceOne as more meritocratic (the mean
difference between the two conditions was .674, t-value =
2.376, p
Purchase answer to see full
attachment