American Sociological Review
2017, Vol. 82(1) 179 –210
© American Sociological
Association 2017
DOI: 10.1177/0003122416683393
journals.sagepub.com/home/asr
Despite women’s rising participation in
higher education in recent decades, gender
inequalities in the labor market persist in the
United States and other Western societies.
Most prominently, women’s hourly wages
remain approximately 80 percent of men’s
wages (Blau and Kahn 2007; Ridgeway
2011). Even after extensively controlling for
human capital and working hours, a remark-
able gender pay gap remains, with the ratio of
female-to-male earnings being .90 (for the
United States, see Blau and Kahn 2007; for
Germany, see Gartner and Hinz 2009). Moreover,
683393ASRXXX10.1177/0003122416683393American Sociological ReviewAuspurg et al.
2017
a
LMU Munich
b
University of Konstanz
c
Radboud University Nijmegen
Corresponding Author:
Katrin Auspurg, Department of Sociology,
LMU Munich, Konradstr. 6, DE-80801 Munich,
Germany
E-mail: Katrin.Auspurg@lmu.de
Why Should Women Get Less?
Evidence on the Gender Pay
Gap from Multifactorial Survey
Experiments
Katrin Auspurg,
a
Thomas Hinz,
b
and Carsten Sauer
c
Abstract
Gender pay gaps likely persist in Western societies because both men and women consider
somewhat lower earnings for female employees than for otherwise similar male employees
to be fair. Two different theoretical approaches explain “legitimate” wage gaps: same-gender
referent theory and reward expectations theory. The first approach states that women compare
their lower earnings primarily with that of other underpaid women; the second approach argues
that both men and women value gender as a status variable that yields lower expectations
about how much each gender should be paid for otherwise equal work. This article is the first
to analyze hypotheses contrasting the two theories using an experimental factorial survey
design. In 2009, approximately 1,600 German residents rated more than 26,000 descriptions
of fictitious employees. The labor market characteristics of each employee and the amount of
information given about them were experimentally varied across all descriptions. The results
primarily support reward expectations theory. Both men and women produced gender pay
gaps in their fairness ratings (with the mean ratio of just female-to-male wages being .92).
Respondents framed the just pay ratios by the gender inequalities they experienced in their
own occupations, and some evidence of gender-specific evaluation standards emerged.
Keywords
gender pay gap, same-gender referent theory, reward expectations theory, double standard
theory, factorial survey experiment
180 American Sociological Review 82(1)
policy efforts in recent years have not materi-
ally decreased gender inequalities (Akchurin
and Lee 2013).
In this article, we argue that an important
reason why notable gender pay inequalities
persist is that both men and women perceive
women’s lower wages as fair, at least to a cer-
tain degree. There are several reasons to
assume that wage distributions that observers
consider fair will persist in labor markets. Fol-
lowing the fair wage hypothesis, workers pro-
portionally reduce their efforts as their wages
fall short of their perceptions of fair wages
(Akerlof and Yellen 1990). Indeed, there is
evidence that perceptions of unfairness cause
low job satisfaction, low organization com-
mitment, high turnover rates, absenteeism,
and shirking (see, e.g., Colquitt et al. 2001). In
the employer’s perspective, perceptions of
unfairness increase a firm’s transaction costs,
including searching for, bargaining with, and
monitoring employees (Fehr, Goette, and
Zehnder 2009; Husted and Folger 2004).
But why would women consider lower pay
for women to be fair? To address this puzzle,
we contrast two prominent explanations for
just gender pay gaps. First, women might
consider lower earnings to be fair because of
gender-specific referents. Both the general
tendency to compare oneself with others who
are similar (Festinger 1954) and homophily in
social networks (McPherson, Smith-Lovin,
and Cook 2001) suggest that women compare
themselves primarily with other women or
with employees in female-typed occupations
that have below-average wages. As a conse-
quence, women might be partially unaware of
the pay gaps between themselves and their
male colleagues (Major 1989, 1994; Major
and Forcey 1985). Second, women’s inputs
into the labor market might be perceived as
being of less value than men’s. This hypoth-
esis of gender status beliefs is advocated by
reward expectations theory (Berger et al.
1985; Berger, Wagner, and Webster 2014).
Gender, as a diffuse status characteristic, is
assumed to create cultural beliefs in men’s
higher competence and status worthiness,
thus entitling them to higher rewards.
To date, little research contrasts these dif-
ferent approaches within the same research
design. Such research, however, would be
particularly promising for advancing knowl-
edge on gender inequality and designing pos-
sible interventions. For instance, if just gender
pay gaps are the result of gender-specific
comparisons to others, attempts to reduce
gender segregation in the labor market or to
provide more information on actual wages
would be helpful in reducing gender inequali-
ties (Desmarais and Curtis 2001; Major
1989). From the perspective of reward expec-
tations theory, however, increasing exposure
to pay inequalities might instead promote the
internalization of gender status beliefs. From
this perspective, only additional measures
(e.g., female role models earning high pay)
would help destabilize gender inequalities
(Ridgeway 2011).
We implemented two important innova-
tions over prior research. First, using an exper-
imental factorial survey approach, we varied
the amount of information that described
hypothetical employees. This ensured that gen-
der differences in fairness evaluations of hypo-
thetical employees’ earnings were not caused
simply by gender functioning as a proxy for
lack of information on performance-related
criteria, such as work experience (for such
assumptions on statistical discrimination, see,
e.g., Arrow 1998; Phelps 1972). Second, we
collected information about respondents’ occu-
pations, thus enabling us to observe the extent
to which fairness evaluations were framed by
respondents’ own social contexts. In 2009,
approximately 1,600 residents of Germany
participated in our experiments, resulting in
more than 26,000 evaluations of fair earnings.
Germany seemed well-suited for this research,
because prior work has speculated whether
subtle processes of discrimination could be the
cause of the substantial gender pay gaps in
Germany (see, e.g., Gangl and Ziefle 2009).
This article contributes to the still under-
researched question of the extent to which
comparison processes and (status) beliefs play
out not only in the lab but also in the real-
world macro-context of societies.
Auspurg et al. 181
THEORETICAL BACKGROUND
AND EMPIRICAL EVIDENCE
Justice Evaluations
When analyzing how individuals make jus-
tice evaluations, researchers typically assume
combinations of equality-, need-, and equity-
based distribution rules (Deutsch 1985). The
equality and need principles represent ideal
standards, and the principle of equity is an
existential standard in which the relation
between rewards and inputs for rewardee A
must equal the relation observed for a referent
person or referent group B (Adams 1965;
Shepelak and Alwin 1986; Walster, Bers-
cheid, and Walster 1973). Just rewards are
based on the typical reward for other people
with similar reward-relevant characteristics,
which means ideas about just rewards are
linked to the existent distribution of rewards.
For economic exchange situations, such as
employment relationships, this existential
standard is seen as being dominant, trumping
ideal standards in their relevance (Shamon
and Dülmer 2014).
Two theories explaining just gender pay
gaps stand out. First, women might use other
reference groups when making comparisons.
Second, gender might be an input variable on
its own, legitimizing higher earnings for peo-
ple with higher status in this variable. This
situation also includes the possibility of gen-
der-colored evaluations of performance, as
assumed in the extension of the double stand-
ard hypothesis (Foschi 1996, 2000; Foschi,
Lai, and Sigerson 1994).
Gender-Specific Referents
Social comparison theories suggest that indi-
viduals prefer referents with similar charac-
teristics, such as gender or occupation
(Festinger 1954). One rationale is that these
characteristics likely relate to performance
and, as a result, provide ideal information for
what one can expect (Gibson and Lawrence
2010). Another reason is opportunity struc-
tures: due to gender segregation in the labor
market (Charles and Grusky 2004) and gen-
der homophily in social networks (McPher-
son et al. 2001), male and female employees
have different opportunities for comparison.
Women typically work in sectors and profes-
sions characterized by lower wage levels
(Charles and Grusky 2004). In combination
with the use of same-gender referents, actual
gender pay gaps translate into similar just
gender pay gaps. In summary, one can first
assume that women have generally lower pay
expectations than do men. Second, gender
differences in fair wages mirror actual pay
gaps, and third, this equivalence exists, in
particular, in areas where women have little
contact with male co-workers (i.e., in highly
segregated occupations).
Research on social comparison processes
confirms that subjects compare themselves
with similar others (Bylsma and Major 1994;
Major 1994; Major and Testa 1989). In several
experiments, women first showed lower pay
expectations than men, but these differences
disappeared once both genders were provided
the same comparison information (Bylsma
and Major 1992, 1994; Major, McFarlin, and
Gagnon 1984; Major and Testa 1989). Unfor-
tunately, the authors of these laboratory stud-
ies were not able to test whether these effects
are generalizable beyond the gender-neutral
tasks and student participants they used.
Convincing evidence of the same-gender
referent hypothesis would also have to show
that women hold lower reference standards
not only for themselves but also for their male
colleagues. Few studies test this notion. Major
and Konar (1984) found that female manage-
ment students estimated the typical pay of
managers (independent of their gender) as
lower than did their male classmates, and this
difference was one of the main reasons why
female students held lower pay expectations.
However, as the authors stated, the partici-
pants in this study were asked to estimate
typical wages in the fields they planned to
enter, which means women likely anticipated
wages in female-dominated areas such as
personnel management.
182 American Sociological Review 82(1)
Gender as a Status Value: Reward
Expectations Theory
Reward expectations theory is rooted in status
characteristics theories and the status value
theory of distributive justice (Berger et al.
1977; Melamed 2012). These approaches
attempt to explain how status structures
emerge via social interactions and how nomi-
nal characteristics such as gender become
connected to different (reward) expectations
(for overviews, see Berger et al. 2014; Berger
and Webster 2006; Kalkhoff and Thye 2006).
Reward expectations theory, in particular,
focuses on explaining how expectations about
fair rewards are activated within an actor’s
encompassing social framework (Berger and
Webster 2006). A core assumption is that jus-
tice evaluations require stable frames of refer-
ence: “reward expectation inputs include not
simply task related contributions within the
local setting, but also culturally relevant
norms regarding the social worth of different
types of actors” (Fişek and Hysom 2008:771).
These referential standards provide socially
shared beliefs about how rewards are typi-
cally allocated within groups or a society as a
whole (Berger et al. 1985; Fişek and Hysom
2008). Individuals who use these referential
standards come to expect the same compensa-
tion for their services as that provided to
people who show the same (status) character-
istics (Berger et al. 1985; Shepelak and Alwin
1986). Additionally, when a socially valued
reward is distributed unequally, actors infer
respective performance differences based on
these reward differences (Berger et al. 1985).
Berger and colleagues (1985) distinguish
three types of referential structures that might
be activated in social comparison processes:
(1) abilities—what actors can do in a situa-
tion; (2) performance—what actors have
accomplished; and (3) categorical informa-
tion on status characteristics—that is, who the
actors are. Status characteristics can be either
specific or diffuse: specific status characteris-
tics carry widely shared expectations for
competence in limited, well-defined ranges of
ability, whereas diffuse status characteristics
carry very general expectations of higher
competence (Berger et al. 1985; Correll and
Ridgeway 2003). Gender is considered a dif-
fuse status characteristic, that is, men are
commonly expected to be more competent in
most tasks, or at least tasks that “count most”
(e.g., tasks requiring instrumental rationality,
management tasks; Correll and Ridgeway
2003; Thébaud 2015).
To summarize, status beliefs are socially
shared beliefs that consensually value one
category of a social attribute as more worthy
and competent than another, which means
people who are disadvantaged by status
beliefs accept such beliefs (Correll and Ridge-
way 2003). This constitutes the most impor-
tant difference to the same-gender referent
hypothesis: one should expect not an effect of
the observer’s gender (i.e., female observers
consider lower base wages to be fair), but
rather an effect of the rewardee’s gender: both
male and female observers are assumed to
believe in the greater competence of male
employees and thus to assign higher just earn-
ings to male employees.
As a highly visible status marker, gender
should be salient in most social situations
(Melamed 2012; Ridgeway 2011). However,
variations in effect sizes may exist. On the
one hand, gender may be particularly salient
when actors of different genders frequently
interact with one another, that is, in gender-
mixed occupations (Rashotte and Webster
2005; Thébaud 2015). On the other hand,
gender status beliefs that give men more
credit might be more salient in male-stereo-
typed occupations (Berger and Fişek 2006;
Ridgeway 2011). In such occupations, in
addition to diffuse beliefs about male compe-
tence, individuals also have presumptions
about gender-specific skills (Ridgeway and
Correll 2004).
Gender status beliefs should also be par-
ticularly dominant in social contexts in which
gender inequalities already exist: a core
assumption of reward expectations theory is
that existing inequalities are reproduced. One
can thus assume that fair wage gaps are
shaped by actual pay gaps in observers’ and
Auspurg et al. 183
rewardees’ occupations. Both seem to be
meaningful reference points in third-party
evaluations.
1
Finally, another assumption is built on the
combination of reward expectations theory
with theories of double standards: status char-
acteristics are assumed to prime not only
individuals’ expectations of competence and
performance, but also the standards used to
interpret information about performance as
suggestive of true abilities (Foschi 1996,
2000). Status inconsistencies, such as low-
status group members showing high perfor-
mance, can be reconciled with existing status
beliefs by using a harsher standard. As a
consequence, status beliefs should cause
biased assessments of performance. People
with a disadvantaged status (women) are held
to a stricter standard unless their performance
is judged as being “good” (Foschi 1996,
2000; Thébaud 2015).
Research on expectation states theories
builds on a long tradition of using standard-
ized experimental settings.
2
Dozens of labora-
tory studies have found support for most of
the proposed mechanisms (for overviews, see
Berger et al. 2014; Berger and Webster 2006;
Kalkhoff and Thye 2006). Although one must
conduct research in non-laboratory settings to
determine whether status beliefs are also
primed by real-life experiences or can be gen-
eralized to real (labor market) settings, few
studies do so.
One of the few exceptions is the seminal
field experiment conducted by Correll,
Benard, and Paik (2007). The authors reported
a motherhood penalty, in the sense that moth-
ers are less likely to be invited to a job inter-
view than are equally qualified men or
non-mothers. This finding alone might also
represent evidence of statistical discrimina-
tion (more on this theory will be presented
later). Only in combination with a factorial
survey experiment were the authors able to
provide more direct evidence of status beliefs.
University students reported that mothers
should demonstrate higher test scores in man-
agement abilities before being considered as
hirable as non-mothers (Correll et al. 2007).
This additional evidence for double standards
allowed the authors to conclude that status
beliefs had influenced respondents’ answers.
However, this evidence was again based
solely on university students.
This restriction is also true for Thébaud
(2015), a study in which university students
rated vignettes showing fictitious descriptions
of entrepreneurs. Again, participants held
lower performance expectations for women
and rated their entrepreneurial abilities and
business plans more harshly. These gender
differences were more pronounced in settings
in which entrepreneurship was male-typed
(e.g., in high-tech instead of a gender-neutral
industry; and more so in the United Kingdom
than in the United States, with men’s over-
representation in entrepreneurship actually
being stronger in the United Kingdom).
In addition, some factorial survey studies
have asked respondents to evaluate the fairness
of earnings of hypothetical employees, which
is very close to the empirical design used in the
current study. Studies using general population
surveys consistently find evidence of a just
gender pay gap favoring men (for Germany,
see Sauer et al. 2014; for Switzerland, see Jann
2005; for Ukraine, see Gatskova 2013; for the
United States, see Jasso and Webster 1997).
Jasso and Webster (1997) fit new theoretical
models to the factorial survey module origi-
nally designed and administered by Jasso and
Rossi (1977) to a blocked-quota probability
sampling of 200 white adults in Baltimore in
1974. They also examined underlying mecha-
nisms and found evidence of a fair base-wage
gap (men were generally assigned higher earn-
ings) in combination with double standards for
education (men in the vignettes were assigned
higher fair returns on education). However, as
the authors stated, the study was designed pri-
marily to illustrate the potential application of
factorial surveys rather than to study causal
mechanisms.
Interestingly, factorial surveys with univer-
sity students—or with an overrepresentation
of highly educated, young respondents—did
not find just gender pay gaps or even slight
evidence of just gender pay gaps favoring