WOMEN HELPING WOMEN?
EVIDENCE FROM PRIVATE SECTOR DATA ON WORKPLACE HIERARCHIES
Astrid Kunze and Amalia R. Miller
Abstract—We study gender spillovers in career advancement using eleven
years of employer-employee matched data on white-collar workers at
over 4,000 private sector workplaces in Norway. Our data allow us to
define seven hierarchical ranks that are consistent across plants and over
time and track promotions even for individuals who change employers.
We find positive spillovers across ranks (flowing from higher-ranking to
lower-ranking women) but negative spillovers within ranks. The finding
of narrower gender gaps in promotions for workers with more female
bosses suggests that policies that increase female representation in corpo-
rate leadership can have spillover benefits to women in lower ranks.
I. Introduction
A LTHOUGH women comprise nearly half of the labor
force across OECD countries, they remain dramati-
cally underrepresented among business leaders. Eight per-
cent of employed men in OECD countries have managerial
responsibilities, but only 4% of employed women do. At
the highest levels of leadership, women hold only 5% of
CEO positions in Fortune 1000 companies.1 The gender
gap in leadership is even present in Nordic countries, where
only 3% of CEOs at the 145 largest companies are women
(Zander, 2014) and women are just 6% of general managers
in public limited companies in Norway.
This paper studies white-collar private sector workers in
Norway to examine a potential source of the gender gap in
leadership: gender differences in career progression as a
consequence of path dependence from male dominance of
workplace hierarchies. Specifically, we measure the spill-
over effects of increasing female representation at different
organizational ranks on gender differences in promotion
rates using comprehensive data on over half a million
worker-year observations across 4,000 plants from 1987 to
1997.2 Rather than focusing exclusively on how the highest-
ranking women (top executives or board members) affect
lower-ranking workers, we consider the entire organiza-
tional hierarchy of white-collar workers and changes in the
female share of two groups of coworkers: peers at the same
rank and bosses at a higher rank.
Gender spillovers from bosses might be positive because
higher-ranking women serve as mentors, role models, and
advocates for their lower-ranking coworkers (Athey, Avery,
Received for publication June 10, 2015. Revision accepted for publica-
tion January 9, 2017. Editor: Bryan S. Graham.
* Kunze: NHH Norwegian School of Economics; Miller: University of
Virginia, IZA, and NBER.
We are grateful to Francine Blau, Albrecht Glitz, David Matsa, Øivind
Anti Nilsen, Carmit Segal, Geoffrey Warner, and various seminar and
conference participants for helpful comments and discussions.
1
http://www.catalyst.org/knowledge/women-ceos-fortune-1000 lists
female CEOs.
2
Kunze and Miller (2014) demonstrate the empirical relevance of both
promotion gaps as a source of gender differences in leadership and gender
differences in hierarchical rank in explaining the gender pay gap in our
sample.
The Review of Economics and Statistics, December 2017, 99(5): 769–775
Ó 2017 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology
doi:10.1162/REST_a_00668
& Zemsky, 2000) or because promotion decisions are
affected by the taste-based or statistical discrimination of
current leaders (statistical discrimination may be affected
by the gender of bosses if people are better at evaluating the
work performance of others who are more similar to them;
Aigner & Cain, 1977). Increasing female representation at
higher ranks could also improve promotion rates for lower-
ranking women by weakening the associations of leadership
with masculinity (Koenig et al., 2011). These theories for
positive spillover suggest that men’s historical dominance
of business leadership is self-perpetuating, but that it can be
disrupted by increases in female shares at higher ranks.
Although policymakers and advocates often assume that
gender spillovers from bosses are positive, it is worth noting
that they need not be. The spillovers from female bosses can
be negative if the ‘‘queen bee’’ phenomenon is common.
This happens when a woman who achieves career success in
a male-dominated field blocks other women from advancing
(Staines, Tavris, & Jayarante, 1974). Although women are
often expected to be more favorable judges of other
women’s work,3 there is evidence of the opposite pattern in
some male-dominated fields, which could also lead to nega-
tive spillovers, as in Bagues and Esteve-Volart (2010).
Gender spillovers among peers can also be positive or
negative. Women may be more willing to compete against
other women (Gneezy, Niederle, & Rustichini, 2003), lead-
ing to increasing female performance and promotion rates
when the peer group has more women. However, unlike a
laboratory setting with individual tasks, workplaces often
involve both individual and cooperative tasks. While it is
possible that women collaborate more with other women, it
is also possible that women feel that their closest competi-
tors for promotion are other women. This could happen,
for example, in response to tokenism at workplaces or
informal limits on the number of women in higher ranks
(Kanter, 1977; Dezs}o, Ross, & Uribe, 2016). In that case,
women might be less cooperative with one another than
with men, which could lower their relative promotion rates.
This form of sex discrimination in promotions, through
which female workers are competing most intensively with
their female peers to be promoted into positions occupied
by female bosses, would imply negative gender spillovers
within ranks but positive spillovers from higher-ranking to
lower-ranking women.
We find that women in our sample experience a signifi-
cantly lower annual likelihood of advancing a rank than do
3
Even unconscious biases against women in performance ratings are
found to be less severe when women comprise a larger share of the team
(summarized in Valian, 1998).
770 THE REVIEW OF ECONOMICS AND STATISTICS
their male counterparts, after controlling for a wide range
of individual characteristics and workplace fixed effects.
The gap is reduced when there are more female bosses in
the next highest rank, but it is increased when there are
more female peers at the same rank. These results are con-
sistent across a range of models with different sets of con-
trol variables and fixed effects and are robust to alternative
definitions of female leadership shares.
Although previous studies have measured gender spill-
overs in hierarchical organizations, we are first to measure
effects of both types of coworkers on promotion rates in
corporations. The only other paper we are aware of measur-
ing effects of the gender mix of peers and superiors on pro-
motion rates is Karaca-Mandic, Powell, and Maestas’s
(2013) study of gender and racial spillovers among the
lowest-ranking enlisted members of the U.S. Army.
Although that setting is quite different from ours, it is inter-
esting that the pattern of gender spillovers observed in that
study is the same: positive effects of female superiors but
negative effects of female peers.
Papers that have focused exclusively on downward-flow-
ing gender spillovers on promotion rates, such as Blau and
DeVaro (2007) and Giuliano, Levine, and Leonard (2011),
tend not to find significant effects. In contrast with those
studies, we use panel data on workers across all ranks and
include controls for the worker’s current rank and the
female share among peers. This last variable is positively
correlated with female leadership, but its estimated effect
on female promotion rates is negative. Its omission can
therefore be a source of downward bias in the estimated
effect of female bosses. Our finding of positive spillovers
from female leaders is more in line with the results from
studies that examine the representation of women at differ-
ent organizational ranks (e.g., Kurtulus & Tomaskovic-
Devey, 2012; Matsa & Miller, 2011). Our paper also contri-
butes to a broader literature on how female business leaders
affect lower-ranking workers, including recent studies of
compensation levels, gender pay gaps (Flabbi et al., 2014;
Tate & Yang, 2015; Cardoso & Winter-Ebmer, 2010),
employment, and downsizing (Matsa & Miller, 2013, 2014).
II. Data Description
Our data set is derived from Statistics Norway’s register
data and from plant-level job surveys compiled by the Con-
federation of Norwegian Enterprise (NHO, Næringslivets
Hovedorganisasjon), the primary employer association in
Norway, for 1987 to 1997.4 Our estimation sample includes
workers at private sector firms that are NHO members,
which overrepresents manufacturing but retains broad
4 6
The job survey was conducted to obtain an overview of earnings
among employees; response rates among surveyed plants were very high
because employers were legally bound to participate. We use the longest
time period that our key variables are available and include workers born
between 1936 and 1969. For more details about the data source, see
Lønnsstatistikk for funksjonærer, an annual publication.
coverage. In 1997, NHO employees represented 37% of all
male workers and 27% of all female workers in Norway.
We restrict our sample to white-collar workers to ensure a
comparable set of jobs with a substantial presence of both
male and female workers. Women comprise 30% of the
sample of workers below the top rank, who have a chance
of promotion. Like any other employment-based sample,
ours is limited to current workers and excludes people who
are unemployed or out of the labor market. Workers on
temporary leave from a job, such as parental leave, how-
ever, are included in the data.
Several features of our data enable us to study how gen-
der spillovers affect promotions. The data contain detailed
job information that allow us to assign workers to one of
seven hierarchical ranks that are defined consistently across
plants and over time. Hence, our sample includes a wide
range of employers, unlike many studies of gender gaps in
promotions that focus on a single firm (e.g., Ransom &
Oaxaca, 2005; Giuliano et al., 2011; Jones and Makepeace,
1996). We observe the population of workers at each plant,
which enables our analysis of gender spillovers. We are
able to track workers over time and across plants in the
sample, meaning that we can also observe promotions that
involve a change of employer. These features are not typi-
cally available in samples based on household or firm sur-
veys that include multiple firms (Olson & Becker, 1983;
Blau & DeVaro, 2007),5 and they allow us to control for
unobservable differences in promotion rates between work-
places and for the worker’s current rank using a wide range
of fixed effects. We also have detailed information on
worker characteristics that we include as controls.
The main weakness of our data source relative to sources
used in some previous studies of promotions (such as Blau
& DeVaro, 2007, and several of the single-firm analyses) is
that we do not observe performance reviews or evaluations
for workers. Although it is possible that these (often subjec-
tive) evaluations themselves reflect (possibly unconscious)
gender bias on the part of evaluators,6 having that informa-
tion would have enabled us to distinguish between promo-
tion gaps driven by differences in evaluation scores and
those caused by women needing to meet higher perfor-
mance benchmarks to advance. This latter situation can
arise when the positive signal from a high evaluation is per-
ceived as less precise when applied to women: risk-averse
supervisors may then prefer to promote men over equally
scoring women (Aigner & Cain, 1977). Without this infor-
mation, the gender spillovers we measure could potentially
include both of these effects.
5
Winter-Ebmer and Zweimuller (1997) use a cross-sectional sample
and define prior promotions implicitly using rank and education.
Ibarra, Carter, and Silva (2010) discuss gender bias in evaluations in
the corporate setting. Bagues and Esteve-Volart (2010) also find evidence
of gender bias, but in their setting, evaluators are harsher on candidates of
their same sex. Evaluations for promotions into management positions
may be further influenced by stereotypical associations between masculi-
nity and leadership (Koenig et al., 2011).
 WOMEN HELPING WOMEN? 771

TABLE 1.—SUMMARY STATISTICS

Male Female Total Total Total
Mean Mean Mean SD Observations
Entire sample
Rank ¼ 1 (low) 0.024 0.27 0.094 744,531
Rank ¼ 2 0.18 0.15 0.17 744,531
Rank ¼ 3 0.28 0.4 0.31 744,531
Rank ¼ 4 0.29 0.14 0.25 744,531
Rank ¼ 5 0.079 0.012 0.06 744,531
Rank ¼ 6 0.13 0.025 0.1 744,531
Rank ¼ 7 (top) 0.015 0.0011 0.011 744,531
Main analysis sample
Female 0 1 0.3 597,552
Any promotion 0.073 0.066 0.071 597,552
Internal promotion 0.059 0.056 0.058 597,552
Stay at plant 0.74 0.74 0.74 597,552
Age 40.2 37.1 39.3 9.04 597,552
Years of schooling 12.6 11.7 12.3 2.62 597,552
Years of work experience 18.4 14.5 17.3 7.33 597,552
Years of tenure 7.41 6.52 7.15 5.28 597,552
Rank-specific tenure 4.29 4.04 4.22 2.76 597,552
Working part time 0.076 0.26 0.13 597,552
Any children 0.7 0.56 0.66 597,552
Children under 7 0.33 0.27 0.31 597,552
Female boss share 0.14 0.25 0.17 0.24 597,552
Female peer share 0.18 0.58 0.3 0.29 597,552
Data from NHO (1987–1997) and Statistics Norway. Summary statistics are reported for the entire sample and the main analysis sample that includes workers in ranks 1 to 6 and for whom we calculated the pro-
portion of female bosses at the next highest rank. The proportions of workers within each of the ranks (1 through 7) are reported using the entire sample. Female boss share is defined for workers one rank higher.

A. Gender Differences in Promotion Rates
We define a promotion in year t as a job change between
years t and tþ1 that entails an increase in rank. Worker ranks
include seven hierarchical levels that are determined using
information drawn directly from NHO survey responses
about each worker’s occupational group, task level, and job
description. Within each occupation,7 up to seven task levels
are distinguished, ranging from unskilled and routine tasks
in the lower levels to high-skill and leadership tasks in the
higher levels. Not all occupations span all task levels; some
are concentrated in the lower levels, while others are in the
higher levels. Because the task levels are designed for com-
parisons within occupational groups, we also use informa-
tion on the detailed job descriptions to assign ranks in a way
that ensures comparability across occupational groups.
Table 1 reports the distribution of workers across ranks.
Following the empirical literature on promotions, we
focus on binary measures of progress. We separately con-
sider measures of internal promotions limited to workers
who remain at the same plant in period tþ1 and any promo-
tions, which also include rank increases involving a change
in plant. Internal promotions more closely match the defini-
tion used in the literature (e.g., Booth, Francesconi, & Frank,
2003), where changing employers (with an increase in rank)
is not counted as a promotion. The advantage of our second
measure is that it captures a wider range of possible paths
for career progression. We are able to measure this outcome
because our data track workers across plants and because
7
The occupational groups are (a) technical; (b) production supervision;
(c) administrative, including clerical, accounting, and shipping; (d) sales;
(e) storage; and (f) other.
our seven-rank scale for hierarchies is consistent across
plants and time. In our data, annual promotions of both types
generally involve increases of one or two ranks.
Table 1 reports additional summary statistics for our sam-
ple. Annual promotion rates are higher for men (7.3% for
any promotions and 5.9% for internal promotions) than for
women (6.6% for any promotions and 5.6% for internal pro-
motions), but these raw differences actually understate the
gender gaps for workers of the same rank. This is because
promotion rates decline dramatically at higher ranks, drop-
ping from 13% to 1.7% for any promotions and from 11%
to 1.3% for internal promotions,8 and women are vastly
overrepresented in the lowest rank (over 80% female) and
underrepresented in the highest ranks (over 90% male).
Men and women in the sample also differ along other
observable dimensions that may be related to productivity
and promotions. Men are, on average, older and have more
formal schooling, work experience, tenure at their current
workplace, and rank-specific tenure (time since initial entry
or since the last promotion, whichever is more recent). They
are also less likely to be working part time, which we define
as usual weekly work hours under 37 over the year (defined
based on contracted hours and not affected by leave taking).
Not accounting for these differences could overstate the
gender gap for comparable workers.
In our regression analysis, we therefore include all of
these variables as controls in our models, as well as year,
occupation, and plant fixed effects. After accounting for
8
The pattern of lower rates of promotion from higher ranks is a typical
feature of workplace hierarchies. Other evidence that our rank measure is
capturing meaningful variation in job status is the increasing hourly
wages, age, and years of education at ascending ranks and large year-on-
year wage growth concurrent with promotions.
772 THE REVIEW OF ECONOMICS AND STATISTICS
these variables, we find a highly statistically significant
gender gap in promotions (coefficient on a Female indica-
tor), of 3.3 percentage points, or 50% of the female pro-
motion rate. The gender gap for internal promotions is 2.6
percentage points (46% of the female rate, also highly sig-
nificant). When we estimate expanded versions of the
model with controls for family status—indicators for hav-
ing any children under the age of 17 and under the age of 7
and their interactions with the Female indicator—we still
find substantial and significant gaps of 2.5 (for any promo-
tions) and 2.0 (for internal promotions) percentage points
for childless women. These promotion gaps are present for
workers in each of the six lower ranks in the hierarchy.
B. Gender Mix of Coworkers
We use two measures of coworker gender: (a) the female
share among peers, who are workers at the same rank, plant,
and year, and (b) the female share among bosses working at
the next highest rank at the same plant and year. The mean
female peer share is 30%, which is higher than the mean
female boss share of 17%, reflecting women’s underrepre-
sentation at higher ranks. Across all observations in our
sample, over 25% of workers have no female bosses and
only 1% of workers have all female bosses.
Two features of our measure of female bosses are worth
noting. First, the measure is based only on plant-level rates;
we are not able to identify workers’ immediate supervisors
(as in Karaca-Mandic et al., 2013, and Blau & DeVaro,
2007). If the effects of female bosses run exclusively within
the chain of authority, then our estimates, based on a noisy
measure, will tend to understate the importance of female
bosses. Second, our measure of female bosses is rank speci-
fic and focuses on female representation at the next highest
rank. This has the advantage of measuring effects of female
leadership deep within organizational hierarchies. In a
robustness analysis, we also found similar positive spill-
overs on promotion rates from increasing female shares two
ranks higher or among the plant’s top leadership, defined
based on rank, earnings, or leadership job tasks.
III. Gender Spillovers in Promotions
Our main estimates of gender spillovers are reported in the
top panel of table 2. They are from estimating variants of this
model with different sets of controls and fixed effects (in Xijt):
Promotionijt ¼ bF Femalei þ bX Xijt
þ bP FemaleSharePeersijt
þ bPF FemaleSharePeersijt  Femalei
þ bB FemaleShareBossesijt
þ bBF FemaleShareBossesijt  Femalei
þ eijt :
The variable FemaleSharePeers is the female share of
worker i’s peers at the same rank, plant j, and year t.
FemaleShareBosses denotes the female share of worker i’s
bosses who are one rank higher at the same plant j in year t.
Each of these variables is interacted with the Female vari-
able, which is set equal to 1 if the worker is female and 0
otherwise. The coefficients of the interacted variables are
the differential effects of interest. The outcome variable is
an indicator for an increase in rank between period t and
(t þ 1), regardless of whether the worker also changed
plants during the time. The explanatory variables are all
measured in period t.
Column 1 has the basic controls for year, industry, occu-
pation, age, schooling, experience, tenure, rank-specific
tenure, part-time status, and rank; column 2 adds plant fixed
effects; and column 3 adds the two indicators for having
children (under the ages of 17 and 7) and their interactions
with the Female indicator. Column 4 addresses concerns
that individuals on different promotion trajectories might
sort differentially into workplaces with different coworker
demographics by adding a full set of individual worker
fixed effects. Across all of the models, we find significant
gender spillovers in promotions. More female workers in
the next highest rank are related to a smaller gender gap in
promotion rates, and more female coworkers in the same
rank are related to a larger gap. The female boss share inter-
action ranges from 0.018 to 0.021, implying a standard
deviation (or 0.24) increase in female bosses increases
women’s relative promotion rates by about 0.5 percentage
points, or 7% of their promotion rate. The female peer share
interaction ranges from 0.039 to 0.062 and implies that
a standard deviation (or 0.29) increase in female peers
would reduce female promotion rates relative to men’s by
at least 1.7 percentage points, or 25%.
The positive interaction effect from female bosses is
consistent with women helping women below them in the
hierarchy through mentoring, serving as role models or
reducing discrimination by supervisors. The finding pro-
vides indirect support for the theory that demand-side
factors play a role in creating the observed gender gaps in
promotion rates. It also provides some support for a key
mechanism underlying public policies aimed at promoting
women in business leadership positions: that female leaders
tend to help lower-ranking women in their organizations.
The finding of negative gender interactions among
women at the same rank, however, complicates this story of
women helping women. Because many of the theories of
positive gender spillovers (mentoring, sponsorship, role
models, access to powerful networks) are more important
for bosses than for peers, the absence of positive peer
spillovers may not be surprising. As discussed above, the
negative effects of female peers may come from greater
competition (and less cooperation) among peers of the same
sex or from women in lower ranks facing greater competi-
tion for scarce sex-specific resources such as female mentors
and sponsors.
The fact that we find opposite effects of female peers and
bosses helps rule out alternative stories for the positive
 WOMEN HELPING WOMEN? 773

TABLE 2.—GENDER SPILLOVERS IN PROMOTIONS AND MOBILITY

(1) (2) (3) (4)
Any promotion
Female  Female Boss Share 0.020*** 0.019*** 0.018*** 0.021***
[0.0044] [0.0047] [0.0047] [0.0095]
Female  Female Peer Share 0.060*** 0.060*** 0.062*** 0.039***
[0.0054] [0.0057] [0.0057] [0.011]
Female Boss Share 0.0048 0.0058 0.0054 0.0036
[0.0037] [0.0049] [0.0049] [0.0072]
Female Peer Share 0.050*** 0.036*** 0.037*** 0.030***
[0.0048] [0.0059] [0.0059] [0.009]
N 597,552 597,552 597,552 597,552
R2 0.050 0.073 0.073
Internal promotion
Female  Female Boss Share 0.022*** 0.019*** 0.019*** 0.032***
[0.0040] [0.0043] [0.0043] [0.0088]
Female  Female Peer Share 0.055*** 0.052*** 0.053*** 0.033***
[0.0051] [0.0053] [0.0053] [0.011]
Female Boss Share 0.0019 0.0068 0.0065 0.0077
[0.0033] [0.0044] [0.0044] [0.0065]
Female Peer Share 0.045*** 0.031*** 0.031*** 0.0308***
[0.0045] [0.0056] [0.0056] [0.0089]
N 597,552 597,552 597,552 597,552
R2 0.041 0.061 0.061
Stay at plant
Female  Female Boss Share 0.015* 0.0044 0.0046 0.0027
[0.0080] [0.0068] [0.0068] [0.0120]
Female  Female Peer Share 0.045*** 0.012 0.013 0.006
[0.010] [0.0086] [0.0086] [ 0.0130]
Female Boss Share 0.016** 0.0016 0.0017 0.0072
[0.0082] [0.0080] [0.0080] [ 0.0117]
Female Peer Share 0.031*** 0.0032 0.0038 0.0094
[0.0093] [0.0098] [0.0098] [0.0130]
N 597,552 597,552 597,552 597,552
R2 0.32 0.44 0.44
Basic controls including rank FE Yes Yes Yes Yes
Plant FE No Yes Yes Yes
Children FE No No Yes Yes
Individual FE No No No Yes
OLS estimates with standard errors clustered at the plant-rank-year level. Female Boss Share is the female share of workers in the next highest rank at the same plant-year. Peer Share is for workers at the same rank
in the same plant-year. Outcome variables are measured as changes from period t to t þ 1. The outcome in the top panel is an indicator for having an increase in rank, with or without a change in workplace. The out-
come in the middle panel is an indicator for a rank increase within the same plant. The outcome in the bottom panel is an indicator for changing plants, with or without a change in rank. Basic controls in all columns
are measured in period t and include year, rank, industry, occupation, age, and years of schooling fixed effects, as well as quadratic controls for experience and tenure, a linear control for rank-specific tenure, and an
indicator for part-time status. Significant at ***1%, **5%, *10%.

spillovers. Specifically, the pattern is not consistent with
workers sorting based on workplace-specific gender-related
omitted variables (such as female friendliness or skill in
identifying and recruiting female talent) that create spurious
correlations between female representation at all levels,
including bosses and peers, and promotion outcomes for
women.
Although female bosses are associated with higher pro-
motion rates for female workers, they appear to have no sig-
nificant effects on promotion rates for males. This gender
difference provides evidence against the presence of an
omitted variable (time-varying, because our model includes
plant fixed effects) that is related to promotion rates overall
(such as corporate growth and expansion) and to female
boss shares (e.g., if growing firms are able to increase diver-
sity at higher ranks more rapidly). Unlike the zero effect of
female bosses on male promotion rates, female peers are
estimated to have a positive effect on male promotion rates.
These positive effects for men also suggest that the negative
peer effects on women are not driven by an omitted variable
that is related to the female share in a rank and overall
lower promotion opportunities, which would be the case if
women tended to sort into dead-end job groups identifiable
by their work plant, rank, and year combination.
We also conducted a wide range of robustness checks
that all confirmed the main findings. These are described in
detail in Kunze and Miller (2014). One concern is that
female- and male-dominated ranks differ in size, absolutely
or relative to the next highest rank, and that these size dif-
ferences constrain the available slots for promotions. We
address this by adding controls for the size of the current
and the next highest ranks. Another concern is that the
effects of female peers in particular are coming from larger
gender gaps in promotion at lower ranks, which we address
by adding a full set of interactions between the Female and
rank indicators. We also account for the possibility that
women are disproportionately working in occupational
groups with limited promotion opportunities by controlling
for the share of workers in the next highest rank in the orga-
nization from the worker’s current occupational group. We
774 THE REVIEW OF ECONOMICS AND STATISTICS
also found that our estimates are robust to controlling for
plant-by-year, plant-by-rank, or rank-by-year fixed effects.
The middle panel reports estimates using the narrower
promotion outcome that focuses on internal promotions
within the same plant. The point estimates for the gender
spillovers are consistent in sign between the two panels,
with equal or larger coefficients for female bosses and
somewhat smaller ones for female peers in the middle
panel. However, as we note in section IIA, the gender gap
in promotions, conditional on our basic controls and plant
fixed effects, is larger for any promotions (3.3 percentage
points) than for internal promotions only (2.6 percentage
points). When computed as a proportion of the overall gen-
der gap in promotions for each outcome, the gender spill-
overs for both bosses and peers are in fact consistently
larger for internal promotions. The proportionally larger
internal effects are plausible because the spillovers we
study are largely internal to the firm.
The results in the bottom panel of table 2 show no signifi-
cant gender spillovers in the probability of staying at the
same workplace, after plant fixed effects are included in the
model (columns 2 to 4). The significant spillover coeffi-
cients in column 1 seem to suggest that there is some sort-
ing out of workplaces based on gender compositions of
peers and bosses. However, their dramatic reduction in size
and significance in the latter two columns indicates that this
sorting is captured with plant fixed effects. As discussed
above, these same fixed effects leave the spillover effects
on promotions unchanged.
IV. Conclusion
This paper uses unique data on the gender composition
of workplaces within hierarchical ranks from an eleven-
year employer-employee matched panel on white-collar
workers at over 4,000 workplaces in Norway to study how
gender differences in promotion rates are affected by the
presence of female bosses and female peers. Several theories
predict that increasing female leadership can improve out-
comes for lower-ranking women: lower-ranking women
may gain access to better mentors and role models and pro-
fessional networks. Consistent with this prediction, we find
that higher shares of female workers at the next highest
rank are associated with significantly smaller gender gaps
in promotion. This is not due to higher relative promotion
rates for women overall at firms that hire more women in
all ranks (because they are more ‘‘female friendly’’).
Rather, we find that having a greater share of female
coworkers at the same plant and rank actually depresses
promotion rates for women, relative to men. This negative
effect may be caused by greater competition within gender
(either between men or women) among peers, either for
promotions or for gender-specific mentoring and support.
The finding that greater female representation at higher
ranks narrows the gender gap in promotion rates at lower
ranks supports the theory that male dominance in corporate
leadership continues to present barriers to women’s advance-
ment. It also suggests that policies that promote greater
female representation in corporate leadership will have spill-
over benefits to women in lower ranks. However, because we
focus on measuring spillovers within organizations, our
estimates may understate the overall importance of gender
spillovers by failing to capture benefits to women at other
firms. These can include mentoring that crosses firm bound-
aries or women helping other women progress in their careers
by hiring them into their firms. Future research that combines
data of the type used in this study with social network infor-
mation could illuminate those cross-firm spillovers.
REFERENCES
Aigner, Dennis J., and Glen G. Cain, ‘‘Statistical Theories of Discrimina-
tion in Labor Markets,’’ Industrial and Labor Relations Review 30
(1977), 175–187.
Athey, Susan, Christopher Avery, and Peter Zemsky, ‘‘Mentoring and
Diversity,’’ American Economic Review 90 (2000), 765–786.
Bagues, Manuel F., and Berta Esteve-Volart, ‘‘Can Gender Parity Break
the Glass Ceiling? Evidence from a Repeated Randomized Experi-
ment,’’ Review of Economic Studies 77 (2010), 1301–1328.
Blau, Francine D., and Jed L. DeVaro, ‘‘New Evidence on Gender Differ-
ences in Promotion Rates: An Empirical Analysis of a Sample of
New Hires,’’ Industrial Relations 46 (2007), 511–550.
Booth, Alison L., Marco Francesconi, and Jeff Frank, ‘‘A Sticky Floors
Model of Promotion, Pay, and Gender,’’ European Economic
Review 47 (2003), 295–322.
Cardoso, Ana, and Rudolf Winter-Ebmer, ‘‘Mentoring and Segregation:
Female-Led Firms and Gender Wage Policies,’’ Industrial and
Labor Relations Review 64 (2010), 143–163.
Dezs}o, Cristian L., David Gaddis Ross, and Jose Uribe, ‘‘Is There an
Implicit Quota on Women in Top Management? A Large-Sample
Statistical Analysis,’’ Strategic Management Journal 37 (2016),
98–115.
Flabbi, Luca, Mario Macis, Andrea Moro, and Fabiano Schivardi, ‘‘Do
Female Executives Make a Difference? The Impact of Female
Leadership on Gender Gaps and Firm Performance,’’ IZA discus-
sion paper 8602 (2014).
Giuliano, Laura, David I. Levine, and Jonathan Leonard, ‘‘Racial Bias in
the Manager-Employee Relationship: An Analysis of Quits, Dis-
missals, and Promotions at a Large Retail Firm,’’ Journal of
Human Resources 46 (2011), 26–52.
Gneezy, U., M. Niederle, and A. Rustichini, ‘‘Performance in Competitive
Environments: Gender Differences,’’ Quarterly Journal of Eco-
nomics 118 (2003), 1049–1074.
Ibarra, Herminia, Nancy M. Carter, and Christine Silva, ‘‘Why Men Still
Get More Promotions Than Women,’’ Harvard Business Review
88 (September 2010), 80–85.
Jones, David R., and Gerald H. Makepeace, ‘‘Equal Worth, Equal Oppor-
tunities: Pay and Promotion in an Internal Labour Market,’’ Eco-
nomic Journal 106 (1996), 401–409.
Kanter, Rosabeth. M., Men and Women of the Corporation (New York,
NY: Basic Books, 1977).
Karaca-Mandic, Pinar, David Powell, and Nicole Maestas, ‘‘Peer Groups
and Employment Outcomes: Evidence Based on Conditional Ran-
dom Assignment in the United States Army,’’ working paper
(2013).
Koenig, Anne M., Alice H. Eagly, Abigail A. Mitchell, and Tiina Risti-
kari, ‘‘Are Leader Stereotypes Masculine? A Meta-Analysis of
Three Research Paradigms,’’ Psychological Bulletin 137 (2011),
616–642.
Kunze, Astrid, and Amalia R. Miller, ‘‘Women Helping Women? Evi-
dence from Private Sector Data on Workplace Hierarchies,’’
NBER working paper 20761 (2014).
Kurtulus, Fidan Ana, and Donald Tomaskovic-Devey, ‘‘Do Female Top
Managers Help Women to Advance? A Panel Study Using EEO-1
Records,’’ Annals of the American Academy of Political and Social
Science 639 (2012), 173–197.
 WOMEN HELPING WOMEN?
Lønnsstatistikk for funksjonærer (1997). utarbeidet av Næringslivets
Hovedorganisasjon omr ade for arbeidsmarked og Statistikk. ISSN
0333-2608. Available annually.
Matsa, David A., and Amalia R. Miller, ‘‘Chipping Away at the Glass
Ceiling: Gender Spillovers in Corporate Leadership,’’ American
Economic Review 101 (2011), 635–639.
——— ‘‘A Female Style in Corporate Leadership? Evidence from Quo-
tas,’’ American Economic Journal: Applied Economics 5 (2013),
136–169.
——— ‘‘Workforce Reductions at Women-Owned Businesses in the Uni-
ted States,’’ Industrial and Labor Relations Review 67 (2014),
422–452.
Olson, Craig A., and Brian E. Becker, ‘‘Sex Discrimination in the Promo-
tion Process,’’ Industrial and Labor Relations Review 36 (1983),
624–641.
775
Ransom, Michael, and Ronald L. Oaxaca, ‘‘Intrafirm Mobility and Sex
Differences in Pay,’’ Industrial and Labor Relations Review 58
(2005), 219–237.
Staines, Graham L., Carol Tavris, and Toby E. Jayarante, ‘‘The Queen
Bee Syndrome,’’ Psychology Today 7 (1974), 55–60.
Tate, Geoffrey, and Liu Yang, ‘‘Female Leadership and Gender Equity:
Evidence from Plant Closure,’’ Journal of Financial Economics
117 (2015), 77–97.
Valian, Virginia, Why So Slow? The Advancement of Women (Cambridge,
MA: MIT Press, 1998).
Winter-Ebmer, Rudolf, and Josef Zweimuller, ‘‘Unequal Assignment and
Unequal Promotion in Job Ladders,’’ Journal of Labor Economics
15 (1997), 43–71.
Zander, Christina, ‘‘Even Scandinavia Has a CEO Gender Gap,’’ Wall
Street Journal, May 21, 2014.