A Survey Experiment on “Bad Bosses”: The Effect of Social Networks on Gender Solidarity
University of Maryland
University of Maryland
Forthcoming in the Latin American Research Review (LARR)
Abstract: Are women and men in position of authority judged differently? If a gender evaluation gap
exists, is it due to persistent stereotypes or notions of gender solidarity? We explore gender differences
in judgement through a survey experiment in Argentina with a national sample of 4,068 employed
respondents and four rotations randomly assigned in roughly equal samples (approximately 1000
respondents per treatment). Respondents were asked to recommend a salary increase to a “bad boss”
who was characterized as aggressive and toeing the line of what is fair and appropriate. The survey
experiment measures the extent to which respondents punish and reward female and male managers
differently, and to the best of our knowledge, it is the first to measure gendered penalties for improper
behavior by authorities. The main finding of the experiment is that women are more likely to punish
male bad bosses and men are more likely to punish female bad bosses, although the former are more
sensitive to treatment. We explain variation in the propensity to penalize bad behavior by men and
women in position of authority as a function of respondents’ social and personal networks. Study
findings carry significant implications for the study of the gender pay gap.
“I also think it is important for women to help one another. I have a saying: There is a special
place in hell for women who don't” ― Madeleine Albright
“The gender stereotypes introduced in childhood are reinforced throughout our lives and
become self-fulfilling prophesies. Most leadership positions are held by men, so women don't
expect to achieve them, and that becomes one of the reasons they don't” ― Sheryl
Sandberg, Lean In: Women, Work, and the Will to Lead
Are women and men in position of authority evaluated differently? Two leaders of the political
and the corporate world, cognizant that women in positions of authority are judged more harshly than
their male counterparts, offer different prescriptions to the problem. Madeleine Albright, the first
female American Secretary of State, has stated repeatedly that women, especially those in power, must
support one another in order to overcome the systemic discrimination they face. As Secretary Albright
explains, “What we have begun to see is that, as women get into a variety of positions, networking is
very important. We really understand that we have to help each other…The bottom line is, the more
we have a cadre of women moving up the scale, and it doesn't seem threatening, and people realize
that women actually work much harder than men, and realize that they need more women in these
jobs, I think that [discrimination] goes away.”1 On the other hand, Sheryl Sandberg, chief executive
officer (CEO) of Facebook and one of the few women at the top of the tech industry, recommends that
women “lean in” and become more assertive and bold –in a way, mimic the stereotypical male
overconfidence in the workplace. Sandberg acknowledges that this strategy is not without negatives,
as more aggressive women are less liked. As she has affirmed time after time, “Success and likeability
are positively correlated for men and negatively for women. When a man is successful, he is liked by
both men and women. When a woman is successful, people of both genders like her less.”2 Ms.
Sandberg believes gender differences in how female and male leaders are evaluated will begin to
disappear once women are able to close the ambition (and fear) gap that hampers their careers and
reduces the number of women in charge. The prescriptions advanced by Secretary Albright and Ms.
Sandberg beg the question, are women and men as bosses judged differently, and if so, is it a function
of differences in personal ties or gender stereotypes? We theorize that personal networks, as essential
sources of interpersonal norms and information, shape (gendered) evaluations of authority. 3
The question of an “evaluation gender gap” is significant. Competitive markets, such as the
labor and the political ones, rely heavily on the evaluation of candidates for selection, promotion, and
dismissal. If there is a gender gap in how equivalent female and male candidates are perceived, the
implications for political success, professional advancement, and income disparity are critical. Indeed,
a wealth of studies in the social sciences point to differences in the way women and men are rewarded
and punished (Rudman and Glick 1999, Lawless 2004, Weikart, Chen et al. 2007, Lammers, Gordijn
et al. 2009, Mendez and Osborn 2010, Dolan and Lynch 2013, Esarey and Chirillo 2013, Morgan and
Buice 2013). Overwhelmingly, the scholarship focuses on the different conditions under which gender
stereotypes shape individuals’ perceptions and evaluations (Rudman and Glick 1999, Lawless 2004,
Weikart, Chen et al. 2007, Lammers, Gordijn et al. 2009, Mendez and Osborn 2010, Abers and Keck
2013, Dolan and Lynch 2013, Morgan and Buice 2013). Few examine how men and women, as
evaluators, differ in the normative assessments they make (Eckel and Grossman 1996, Andreoni and
Vesterlund 2001, Eckel and Grossman 2008, Esarey and Chirillo 2013, Barnes and Beaulieu 2014).
Yet, the information and values that shape individuals’ decision whether to reward or punish the
behavior of others, are formed and disseminated through social interaction, which in turn
unequivocally is gendered (Erickson 2006). Thus, drawing from the literatures on the gendered nature
of social networks (McPherson and Smith-Lovin 1986, Molyneux 2002, Caiazza and Gault 2006,
Erickson 2006, Norris and Inglehart 2006, Carpenter and Moore 2014, Djupe, Mcclurg et al. 2016),
we expect to find a connection between the size of personal networks of evaluators and their gendered
evaluations of women and men in positions of authority. We anticipate gender solidarity in leadership
evaluation among women with larger networks given that past scholarship has found that “women are
more responsive to their networks than men” since their “subordinate position in our gender
stratification system makes gendered issues more costly for women, and hence women are more
attentive to information and influence on these topics” (Erickson 2006: 305).
We explore gender differences in judgement through a survey experiment in Argentina. We
believe Argentina is an excellent case to conduct our survey experiment as a middle income country
with a strong yet highly stratified female labor force. Reported in the 2011 Census Bureau, women
make up 38.5% of all private sector employees and 53% of all public sector employees.4 However,
there are fewer women in leadership roles, representing 30.3% of total managers (directores) and 27%
of section chiefs (jefes). Thus, even though the country in the early 1990s was one of the first to
implement legislative gender quotas and in 2015, at the time of this survey, had a female president
serving out her second term, women remain underrepresented in positions with the power to make
hiring and promotion decisions.
The survey has a national sample of 4,068 employed respondents and four treatments randomly
assigned in roughly equal samples of approximately 1000 respondents per treatment. We provide
respondents with four vignettes depicting female or male managers, all of which have increased profits
for the firm yet have transgressed by either yelling at employees or denying them overtime pay. These
are aggressive company leaders, who toe the line of what is fair and appropriate; hence, “bad bosses.”
The purpose of an aggressive manager is to present respondents with cases that clearly go against
dominant female stereotypes so as to offer the most stringent conditions to test our theory of gender
solidarity stemming from the size of personal networks and the dominant hypothesis of the literature
regarding gendered likability (i.e., both women and men will disapprove of aggressive female leaders).
The survey question proposes a base increase of 10% in salary to reward the managers for improving
earnings. Respondents then are asked to recommend their own salary increase. The experiment is
designed to measure the extent to which respondents penalize and reward female and male managers
differently. To the best of our knowledge, our survey experiment is the first to measure gendered
penalties for improper behavior by authorities.
In contrast to past studies on leadership and candidate evaluation, which overwhelmingly find
evidence that women and men share similar biases based on gender stereotypes, our survey experiment
offers little support for this. Rather, we find that female and male respondents judge women and men
managers differently, with women punishing male bad bosses more and female bad bosses less. In
other words, we find evidence of gender solidarity shaping the evaluation of women by women.
Interestingly, gender solidarity does not extend to men judging other men in power. We explain this
variation in gender solidarity as a function of how men and women draw differently from their social
and personal networks, whereby women rely on them more to decide behaviors and attitudes (Erickson
2006; Carpenter and Moore 2014; Barnes 2016). Our findings carry implications for the study of the
gender pay gap, showing that the gender of the individual recommending the rise has a large and
significant effect on future wages. Furthermore, our empirical findings contribute to the rich research
on social networks. As we will describe in detail in the methods section, the survey we conducted
extensively measures respondents’ networks. Thus, the study gages in a novel way the link between
the size of an individual’s network and the individual’s assessment of leaders. This novel connection
should be of interest to students of Social Network Analysis (SNA), who may be persuaded to test this
in other policy areas.
The organization of the paper follows. First, we situate the question of an evaluation gender
gap in the literature. Second, we examine past literature on perceptions, evaluations, and punishment
of women and men in positions of authority. To flesh out our argument and expectations, we draw
from recent scholarship on social networks and gendered social behavior. Third, we present the
research design of our survey experiment and describe the experimental results. Fourth, we augment
model results by adding covariates that account for attitudinal and contextual determinants of gendered
raises. In the final section, we conclude with some thoughts on future research and study implications.
2. Is There an Evaluation Gender Gap?
The first question our survey experiment answers is whether women and men in positions of
authority are judged differently by respondents. There is a wealth of anecdotal data that points to
differences in how women in positions of leadership are evaluated in relation to equally skilled men.5
It is widely accepted that there are persistent double-standards in perceptions of competence that
overwhelmingly tend to benefit men. Increasingly, dominant gender biases are being exposed and
debated publicly in the political and business worlds, as well as in the cultural, sports, and education
Much of the research concentrates on examining how ceteris paribus women and men are
rewarded and punished differently in the workforce (Blau and Kahn 2000, Cohen, Huffman et al. 2009,
Iversen and Rosenbluth 2010). Gender-based evaluations may help explain consistent findings across
the world, such as the unrelenting gender wage gap –whereby women at equal level of skills and
responsibility systematically make less than men– and the recognition of a glass ceiling that tacitly
hinders women from being promoted to the highest ranks. Recent research has revealed the existence
of a gender wage gap favoring men even in traditionally female-dominated sectors, such as nursing
and education (Budig 2002, Smith 2012). These studies advance gendered differences in evaluation to
explain wage inequality and the so-called glass elevator phenomenon, whereby the men in these
occupations –fewer and on average less senior than their female counterparts– are promoted swiftly
to positions of authority (Smith 2012):
“Research indicates that stereotypes about what a prototypical man is match with
stereotypes about what a prototypical manager is” […as] men tend to be perceived as
more assertive. “Because of the stereotype matching, men more readily fulfill our
notions of what a manager should look like. And when you’re in a female-dominated
profession, there are fewer people that have the ability to match it.”7
Within the political representation scholarship, we also find the conventional wisdom that
candidates are perceived (and might be rewarded) differently according to their gender (Huddy and
Terkildsen 1993, Sanbonmatsu 2002, Thames and Williams 2010, Dolan and Lynch 2013). Often,
when it comes to the leadership traits that appeal to voters, female politicians seem to be judged by a
different standard than male ones. For example, voters tend to “evaluate the same candidate differently
solely because of the candidates gender: …male candidates are perceived as better able to handle
crime, defense, and foreign policy issues; and female candidates are rated as more likely to be liberal,
Democratic, and feminist” (Sanbonmatsu 2002: 21).8 Other research, such as Esarey and Chirillo’s
recent study on the gender of politicians and citizens’ perception of corruption, argue:
“This may be because women are punished more harshly for corruption than men
because of different social expectations for their behavior, as has been anecdotally
observed in American politics: Celinda Lake, a Democratic strategist, says female
politicians are punished more harshly than men for misbehavior. “When voters find
out men have ethics and honesty issues, they say ‘Well, I expected that.’ When they
find out it’s a woman, they say, ‘I thought she was better than that’” (Esarey and
Chirillo 2013: 365).
Based on these past studies, we hypothesis that we will find gender differences in the
evaluation of female and male bad bosses. Yet despite a strong consensus of an evaluation gender gap
in the literature, the question of whether women and men are judged by different standards remains an
empirical one.9 If there are gender differences in how women and men are punished and rewarded in
the workplace, are they driven by attitudinal gender-based biases, by stereotypes associated with the
position that is evaluated, and/or by their social context and personal networks? Next, we discuss
recent scholarship that addresses the reasons behind gender differences in evaluation and lay out our
expectations of causality.
3. What Explains the Evaluation (and the Evaluators’) Gender Gap?
If there are gender-based differences on how female and male leaders are evaluated, what
factors explain the gap? This question has been examined across the social sciences and business
studies, with the scholarship roughly falling into one of two camps. There are those that explain this
gap as a matter of gender stereotypes shared by all evaluators. Others, in contrast, find evidence that
female and male evaluators differ in how they judge women and men in power. In this section, we
survey these literatures and present our argument.
A significant number of studies focus on the effect gender stereotypes have on how individuals
evaluate women and men in position of power (Rudman and Glick 1999, Lawless 2004, Weikart, Chen
et al. 2007, Lammers, Gordijn et al. 2009, Mendez and Osborn 2010, Abers and Keck 2013, Dolan
and Lynch 2013, Morgan and Buice 2013). The extant research finds that under conditions of crises
or high risk, gender stereotypes shape individuals’ perceptions. Often, the salience of these stereotypes
results in a belief that men are more effective leaders than women. In her 2004 study of gender
stereotypes in the post 9-11 world, Lawless explains:
“I find that a clear bias favoring male candidates and elected officials accompanies
the ‘war on terrorism.’ Citizens prefer men’s leadership traits and characteristics,
deem men more competent at legislating around issues of national security and
military crises, and contend that men are superior to women at addressing the new
obstacles generated by the events of September 11, 2001. As a result of this
stereotyping, levels of willingness to support a qualified woman presidential
candidate are lower than they have been in decades” (Lawless 2004: 479-480).
Similarly, Rudman and Glick find empirical evidence that when management jobs are
“feminized” (i.e., described with stereotypical female characteristics) women who do not live up to
those criteria are punished more than their male counterparts (1999). Lammers et al (2008) and Dolan
and Lynch (2014) advance and test explanations of an evaluation gap based on context. Lammers et
al find that a country’s “Most Important Problem,” such as the threat of terrorism, determines whether
voters will prefer male or female candidates, according to the particular gender stereotypes that are
associated with the resolution of the problem (Lammers et al 2008).10 Dolan and Lynch (2014) assert
that gender stereotypes can be contingent on type of policy and shape voters’ attitudes towards
candidates.11 Finally, Morehouse Mendez and Osborn (2010) find that regardless of their objective
levels of political knowledge, women are perceived as less informed about politics than men, which
in turn better positions men to participate in politics.
The scholarship discussed above, however, presuppose little difference in how women and
men judge. Yet, there is a growing body of research that lends support to the view that women and
men’s evaluations may also differ systematically (Gilligan 1982, Eckel and Grossman 1996, Andreoni
and Vesterlund 2001, Dollar, Fisman et al. 2001, Eckel and Grossman 2008, Esarey and Chirillo 2013,
Friesdorf, Conway et al. 2015, Gottlieb, Grossman et al. 2016). Scholars have explained the observed
differences between female and male evaluators as resulting from women and men’s different
perception of risk, whereby the former are more risk averse than the latter (Olsen and Cox 2001; Eckel
and Grossman 2008; and Esarey and Chirillo 2013) as well as resulting from gendered differences in
the evaluation of moral dilemmas (Friesdorf et al. 2015).
Increasingly, research on behavioral differences between women and men identify the
individual’s personal and professional networks as critical sources of information and values
(McPherson and Smith-Lovin 1986; Molyneux 2002; Caiazza and Gault 2006; Erickson 2006; Norris
and Inglehart 2006; Carpenter and Moore 2014; Barnes 2016; Djupe et.al. 2016). In the workforce,
networks matter enormously for the professional advancement of women precisely because
individuals’ ties vary in number and type by gender (Ibarra 1997, Campion and Shrum 2004). Research
has shown also that solidarity ties tend to be stronger for women than for men (Fajak and Haslam
1998, Esarey and Chirillo 2013, Morgan and Buice 2013, Barnes 2016) given that women “are more
powerfully subject to social norms because systematic discrimination against them makes their
position more tenuous” (Esarey and Chirillo 2013: 365).
Network theories of social capital tell us that face-to-face ties lead to inter-personal trust, which
in turn results in increased cooperation (Putnam 2000, Norris and Inglehart 2006). This is particularly
true of the denser bonding social capital, based on close and discriminative ties, which tends to
promote “specific reciprocity and mobilizing solidarity” (Putnam 2000: 22). Accordingly, we expect
evaluators with stronger and more social connections to evaluate managers of the same gender more
favorably. We expect, especially, evaluations that reflect gender solidarity to be stronger for female
respondents because “as a dominated rather than a dominant group in society, women’s gender-based
identity tends to be chronically salient, because they (unlike men) are forced to define themselves, and
be defined, in relation to the opposite sex” (Fajak and Haslam 1998: 73; see also Barnes 2016).
In the following sections we test our argument and answer whether women and men are judged
differently and why.
4. “Bad Bosses”: A Survey Experiment
To measure biases in the evaluation of women and men in positions of authority, we designed
a survey experiment whereby respondents had to recommend a wage raise for a manager that increased
company profits by 20% but, in the process of doing so, s/he also mistreated employees. Different
from most research on implicit bias that center on positive traits, our objective was to measure
differences in the respondent’s assessments of “bad bosses.” That is, as Sandberg’s suggested, whether
successful yet aggressive male and female bosses become “less likable.” The experiment allows us to
test whether respondents penalize women that transgress and misbehave to a larger extent than their
male counterpart. The experiment was embedded in a national level survey conducted in Argentina,
eliciting answers from 7,494 randomly selected respondents, 4,068 of which were 18 or older and
employed or seeking employment at the time of the survey.12
Each individual in the survey received only one of four possible treatment assignments of our
survey experiment. The questions were worded as follow:
“I will now ask you to make a decision as if you were the CEO of a company. Evaluate the following
situation: ‘Since s/he was selected as section chief, just 12 months ago, [María Laura Padilla/José
Morales] increased her/his division’s profits by 20%. As a reward, you have decided to give [her/him]
a raise. However, [María Laura Padilla/José Morales] has been criticized by [her/his] employees, who
accuse [her/him] [of asking them to work extra-hours without compensation/of yelling at and
insulting them]. You had thought about giving [her/him] a 10% raise, as you have done in similar
cases before. However, you do not want to leave unattended the complaints of your employees’. What
raise do you think will reward [María Laura Padilla/Jose Morales] but will also acknowledge complains
by company employees?
Respondents would only see one out of four possible scenarios: (1) the manager was a woman
(María Laura Padilla) and required employees to work extra hours without compensation; (ii) the
manager was a man (José Morales) and required employees to work extra hours without compensation;
(iii) the manager was a woman and she yelled at and insulted her employees; and (iv) the manager was
a man and he yelled at and insulted his employees. No other features distinguish the four treatments
at any point of the survey or in this question. Further, all questionnaires were randomized prior to their
being assigned to the surveyors, ensuring randomization throughout the collection of responses. The
question was posed after a general module on media preferences, and ahead of our other control
questions. Tests show balanced samples across all four treatments, as shown in the online
Supplemental Information File (SIF). All questionnaires, data, and the replication code are available
at (omitted information on the data repository and personal webpages).
By providing a baseline raise of 10% and eliciting a recommended raise, we sought to assess
the extent to which respondents reward or penalize bad bosses conditional on the gender status of the
boss, the gender status of the respondent, and the type of transgression committed against employees.13
We considered a non-violent abuse of power (not paying overtime for extra work hours) and a more
aggressive abuse of power (yelling at and insulting employees). Each treatment had a large number of
respondents, which ensures sensible samples for all relevant combinations of respondents, bosses, and
types of transgression. Among the independent variables, we include a measure of personal network
size, following McCarthy et.al. (2006) and Zheng et.al. (2006), as described in the next section and
detailed in the Supplemental Information File (SIF) that accompanies this article. We also include a
number of other explanatory and control variables.
Descriptive results of the four treatments are provided in Figure 1. Graphs include raises in the
original scale (percent) and in logs, to linearize the data and minimize the effect of a small number of
relatively large raises in the overall means.14 Results from the experiment show that bad bosses receive
on average a 6.5% raise rather than the suggested baseline of 10%. The difference is large and
statistically significant, indicating that respondents understood what was being asked and penalized
transgressions as expected. The mean raise for all bad bosses among female respondents was 6.49%
while among male respondents was 6.51%. That is, both male and female respondents penalized bad
bosses at roughly similar rates, reducing raises from the 10% baseline to a 6.5%. However, as shown
in Figure 1 and discussed ahead, that is where similarities between male and female respondents end.
Figure 1 - Difference in Proposed Raises for Bad Bosses: All respondents by type of
treatment and split samples by women and men respondents
Note: Mean wage raise and log (raise) offered by respondents to the “bad boss”, down from a 10%
(2.3 in the log scale) suggested raise. Responses truncated below 0 (3 cases) and above 30 (168 cases),
which represent 2.8% of respondents. A 43.9% of respondents are men while 56.1% are women.
Descriptive results show statistically significant differences in the wage raises given by
respondents to María Laura Padilla and to José Morales. While respondents on average recommended
a 6.75% raise for María Laura Padilla, the raise for José Morales was a lower 6.25%. The difference
across means was substantive and statistically significant at p<.01. In all, results indicate that a female
bad boss was given a 0.5 percentage point premium over her male counterpart. Further, men that yell
at and insult employees were more heavily penalized by all respondents than equally aggressive female
bosses. Results would seem to cast doubt on the dominant assumption that aggressive behavior is
expected and less damaging for men in positions of authority than for women. On average, men did
worse both when failing to pay overtime and when verbally abusing their employees.
While results cast doubt that male bad bosses fair better than female bosses when exercising
arbitrary authority, results do lend at least some limited support for this hypothesis when we split the
sample by gender. The survey experiment shows that women penalize transgressions by male bosses
more than male respondents. On average, women gave María Laura Padilla a 7.03% raise when she
withheld wages from employees compared to a raise of just 6.17% for José Morales. The difference
is statistically significant and substantively important among women respondents. By contrast, male
respondents gave a 6.84% raise to José Morales compared to a 6.56% raise to María Laura Padilla.
Results for male respondents, however, were not statistically significant.
Female respondents also gave María Laura Padilla an average raise of 6.78% when she yelled
at and insulted her employees, compared to a 5.99% for José Morales. Again, the difference was large
and statistically significant at the 95% in the original scale (p= 0.0310) as well as in the log scale
(p<0.01). Finally, when a bad boss yelled at and insulted her/his employees, male respondents offered
María Laura Padilla a 6.48% raise but just a raise of 6.06% for José Morales, although the difference
was not statistically significant. In all, results show that women penalized male bad bosses more than
they penalize female bosses, while male respondents were insensitive to the treatment.
5. Why do respondents Evaluate Female and Male Bad Bosses Differently?
Descriptive plots and t-tests from the four treatments of the survey experiment show that: (i)
respondents penalize bad bosses for transgressions on the rights of employees; (ii) female and male
bad bosses were evaluated differently for misbehaving; and (iii) female and male respondents
sanctioned bad behavior differently. Women respondents evaluated female managers significantly less
harshly than male managers. The results seem to indicate that there is gender solidarity in wage raises
among women, in support of our argument and in contrast to the dominant hypothesis of gender
stereotypes. Next, we seek to test whether differences in the size of personal networks of respondents
as well as in their preference to interact with male or female authorities, explain differences across all
four treatments and in the overall raises offered.
Dependent and Independent Variables
As in the previous section, the dependent variable is the percent raise (log) that respondents
gave to the two bad bosses (María Laura Padilla and José Morales). We control for the four different
frames of our experiment, holding the first treatment as a baseline (María Laura Padilla docking
overtime pay). We include a number of covariates that explore the effects of: (i) the size of the
respondents’ personal networks; (ii) prior experiences with sexual harassment and personal safety at
work; (iii) the perceived level of autonomy and support for professional growth (training); (iv)
proclivities to interact with male or female bosses at work; and (v) socio-demographic variables such
as the gender of the respondent and their educational achievement. The most important variables are
those that describe the relative size of the respondent’s personal network and their perceived
propensity to engage with men and women bosses in their own jobs (descriptive information of the
variables is available in the online Supplemental Information File.
To measure the size of the respondent’s personal networks, we draw from recent developments
in social network analysis that use indirect survey questions such as “how many X’s do you know” to
estimate the gregariousness of individuals and to uncover the social structure in individual-level data
(Zheng, Salganik et al. 2006, Calvo and Murillo 2013). As in Calvo and Murillo (2013), we asked
each respondent to indicate the number of individuals that they know, and with whom they interact on
a yearly basis, from a variety of social categories.15 We then estimate individual level parameters for
personal network and group level parameters of prevalence using a Poisson distribution with
overdispersion (see online supplemental file for estimation details). The personal network parameter
returned by the Poisson specification has a normal distribution, with negative numbers indicating that
a respondent i knows fewer people than the average respondent (a smaller personal network) and with
positive numbers indicating that the respondents knows more people than the average respondent (a
larger personal network).
To measure the respondent’s inherent propensity to engage with male or female bosses, we
consider two different items in the survey. The first question asks respondents to indicate on a five-
point scale the extent to which they agree with the statement: “When I have problems at work, I prefer
to talk about them with a man”. The second question asks respondents to indicate their agreement with
the statement: “In all work environments, men tend to help men and women tend to help women.”
These two variables seek to measure a baseline inclination to prefer a male or a female boss while at
work and are estimated on split samples of men and women.
Controls include a five-point autonomy question (low to high), measuring the respondents’
perceived level of autonomy to buy goods that are needed at work. We also include a question that
asks on a five-point scale whether the respondent felt physically insecure at work during the last year,
whether s/he felt sexually harassed, or resigned within the previous year because of being harassed.
These controls sought to assess whether respondents had a higher sensitivity to misbehavior by bad
bosses. Two survey items inquire on whether on-the-job training was important for the individual and
whether employers should bear the responsibility of paying for such training. This sought to assess the
extent to which employees consider that career advancement is the responsibility of the employer or
of the employee. Finally, we introduced controls for education and gender.
5.1 Personal Networks and Inter-Personal Work Ties
Results are presented in Table 1, with 5 different specifications that explain the wage raises
recommended by either men or women respondents (a total of 10 models). The first four columns (1
through 4) present the basic estimates of wage raises proposed by men and women respondents using
only the experimental treatments, with the baseline set to María Laura Padilla docking extra hours to
employees. As described in the previous section, differences in the raises proposed by men where less
sensitive to framing (Models 1 and 3) while female respondents penalized more heavily male bad
bosses both when they failed to pay extra hours and when they yelled at and insulted employees
(Models 2 and 4).
Given that the samples for all four treatments are well balanced (see the online SIF), we expect
differences across treatments to hold even when adding all model controls. Readers can verify that
this is the case, with women being more likely to penalize bad behavior by male bosses in all
Table 1: Determinants of Wage Raises
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
WRaise WRaise ln(WRaise) ln(WRaise) ln(WRaise) ln(WRaise) ln(WRaise) ln(WRaise) ln(WRaise) ln(WRaise)
José fails to pay
0.284 -0.858** -0.00350 -0.134** 0.0949 -0.169** 0.0906 -0.182** 0.188 -0.418**
(0.422) (0.366) (0.0705) (0.0624) (0.0851) (0.0777) (0.0863) (0.0796) (0.202) (0.181)
María Laura shouts
-0.0690 -0.246 -0.0473 -0.0445 0.0353 -0.00648 0.0517 -0.0484 0.200 0.0531
(0.427) (0.361) (0.0713) (0.0616) (0.0856) (0.0756) (0.0875) (0.0781) (0.201) (0.176)
José shouts at
-0.495 -1.040*** -0.126* -0.163*** -0.0824 -0.149** -0.0799 -0.156** 0.360* 0.0587
(0.423) (0.363) (0.0707) (0.0619) (0.0856) (0.0756) (0.0867) (0.0783) (0.198) (0.176)
-0.00240 0.0736 -0.00288 0.0765 -0.00608 0.0640
(0.0752) (0.0689) (0.0753) (0.0689) (0.0753) (0.0690)
0.341 0.0320 0.336 0.0248 0.353 0.0495
(0.229) (0.169) (0.230) (0.169) (0.230) (0.169)
-0.260* -0.122 -0.262* -0.119 -0.249* -0.130
(0.150) (0.132) (0.150) (0.132) (0.151) (0.132)
Size of Personal
-0.0485 -0.0641** -0.0391 -0.127** -0.0518* -0.0667**
(0.0311) (0.0293) (0.0576) (0.0571) (0.0312) (0.0294)
Age(LN) 0.0388 -0.0552 0.0384 -0.0595 0.0358 -0.0440
(0.0845) (0.0791) (0.0846) (0.0791) (0.0846) (0.0791)
Education 0.0327** 0.0157 0.0330** 0.0152 0.0322** 0.0180
(0.0158) (0.0141) (0.0158) (0.0141) (0.0159) (0.0142)
Autonomy 0.0385** 0.0249* 0.0379** 0.0244* 0.0379** 0.0260*
(0.0175) (0.0138) (0.0175) (0.0138) (0.0175) (0.0138)
Employer pays for
-0.0309 -0.102*** -0.0322 -0.105*** -0.0311 -0.102***
(0.0274) (0.0262) (0.0274) (0.0263) (0.0274) (0.0262)
-0.0347 -0.0346 -0.0337 -0.0347 -0.0320 -0.0359
(0.0309) (0.0308) (0.0310) (0.0308) (0.0310) (0.0308)
Likes to talk to
-0.0357 -0.0370* -0.0366* -0.0370* 0.00867 -0.0672*
(0.0220) (0.0193) (0.0221) (0.0193) (0.0443) (0.0372)
Likes to talk to
Likes to talk to -0.0578 0.0463
Likes to talk to
males *José 4
Men w/Men vs
0.0202 0.0412** 0.0207 0.0422** 0.0397 0.0797**
(0.0217) (0.0194) (0.0217) (0.0194) (0.0428) (0.0374)
M-M vs W-W
M-M vs W-W
*María Laura 3
M-M vs W-W
*María Laura 3
Constant 6.558*** 7.033*** 1.609*** 1.655*** 1.492*** 2.280*** 1.494*** 2.323*** 1.313*** 2.218***
(0.300) (0.252) (0.0502) (0.0430) (0.373) (0.350) (0.373) (0.351) (0.390) (0.363)
Observations 1,784 2,279 1,784 2,279 1,273 1,517 1,273 1,517 1,273 1,517
LogLik -5820 -7387 -2628 -3355 -1885 -2233 -1884 -2231 -1881 -2228
Note: Models estimate the wage raise for a bad boss proposed by respondents. Cells report coefficients with standard errors in
parenthesis. ∗∗∗ 𝑝 0.01, ∗∗ 𝑝 0.05, ∗𝑝0.1.
Figure 2: Proposed Raise Conditional on Size of Personal Network
Note: Marginal effects computed from Table 1 (7) and (8).
As shown in Table 1, larger personal networks have a negative and statistically significant
effect on raises offered to bad bosses. In all specifications, women with larger personal networks
penalize more both female and male bad bosses when they docked overtime payment and when they
yelled at employees. Among men, on the other hand, the negative effect of large personal networks
was only observed in the last model specification (Table 1, Model 10).
The conditional effect of personal network by the type of treatment is extremely interesting,
showing that more gregarious male and female respondents tend to penalize differently female and
male bosses that mistreat their employees. As shown in Figure 2, women with large personal networks
tend to penalize more heavily male bosses, both when he is unfair with payment as when he is
aggressive. Female respondents with larger personal networks (two standard deviations above the
mean) gave both types of male managers a raise of approximately 4%. By comparison, these female
respondents increased the salary of female bad bosses by 6%, a statistically significant difference of
Men with large personal networks, in contrast, offered lower wage raises to bad bosses that
yelled at and insulted employees, but the differences were only statistically significant when compared
to respondents with very small personal networks. Otherwise, the effect of larger personal networks
among male respondents had a negative but statistically insignificant effect. In all, results show that
more gregarious women display stronger gender solidarity and sanction bad behavior to a much larger
extent than both all men and less gregarious women.
Figure 3: Proposed Raise Conditional on Agreement with the Statement: “When I
have problems at work I prefer to talk about them with a male colleague”
Note: Marginal effects computed from Table 1 (9) and (10).
Figure 4: Proposed Raise Conditional on Agreement with the Statement:
“At work, men help men and women help women”
Note: Marginal effects computed from Table 1 (9) and (10).
We seek to shed a light on how prior gendered biases shape wage raises. Results in Table 1
show that a favorable predisposition towards men has important consequences for how respondents
evaluate female and male superiors. Respondents that agreed with the statement “when I have a
problem at the office I prefer to discuss it with a man” where more lenient towards male bosses that
failed to pay extra hours, but more critical of bosses that yelled at or insulted their employees. These
results seem to line up with recent research on the negative effects of betrayed expectations on
sanctioning behavior (Koehler and Gershoff 2003, Bohnet and Zeckhauser 2004, Bohnet, Greig et al.
2008). Koheler and Gershoff (2003) coined the term “betrayal cost” to describe the extra penalty that
is incurred not just for transgressions committed against an individual but, more importantly, resulting
from the defiance of prior expectations. In the case at hand, respondents who confided in a male boss
may feel betrayed by his aggressive behavior. In Figure 3, we show that male and female respondents
who were positively inclined to talk to men at the office were lenient towards bosses that failed to pay
extra hours but harsh towards bosses that betrayed a disposition to dialog.
Results from Table 1 (reported in Figure 4) show also that respondents predisposed to believe
that the work environment is biased along gender lines were more lenient towards bosses that
transgressed. Indeed, the more respondents agreed with the statement that at work men help men and
women help women, the less likely they were to penalize transgressions by bosses. This was the case
for three of the four treatments, again with the noticeable exception of male bosses that yell at and
insult their employees, whose behavior elicited the strongest negative reactions across the board.
Finally, estimates from the control variables behave as expected. More educated respondents
that report higher autonomy at work tend to identify with the boss and provide larger raises. By
contrast, women that agree that training should be paid for by the employer tend to penalize more the
transgressions of bosses. There is no negative effect on raises from female employees that were
sexually harassed yet there is a positive bias in the case of male respondents; both results are somewhat
surprising although the sample of individuals reporting harassment does not allow for reliable
estimates. A total of 94 women reported being harassed at work during the previous year (representing
a 4.6% of total female respondents) compared to 44 men (a 2.7% of male respondents).16
The results of the survey experiment conducted among over 4,000 respondents in Argentina
lends credence to our argument of female gender solidarity and offers little support of the dominant
hypothesis that leadership stereotypes work against women. Study results show that women
respondents punish misbehaving male managers much more than female transgressors. Gender
solidarity is stronger among women than men, as we had anticipated, and overall, female respondents
appear much more sensitive to treatment than men. Interestingly, regardless of gender, respondents
punished the male manager who yelled at his employees the most. Our survey experiment does not
allow us to test the reason for this, but we speculate that respondents might perceive the shouting man
in power as either too threatening or too incompetent. Another critical finding of the study is the
relationship between respondents’ larger personal networks and an increase in their negative
evaluations of bad bosses. Among women, gregariousness in and of itself serves as a source of moral
judgement and increases negative evaluation, especially of men in positions of authority. We interpret
this finding to mean that more socially-oriented individuals are more likely to punish socially-
undesirable behavior. Our survey experiment, however, raises the issue of why gregarious women are
more willing to punish male bad bosses, rather than bad bosses in general. The novel finding in this
research is that larger social networks make punishment more selectively gendered.
The survey experiment was conducted in Argentina, a middle-income state that has shown
significant leadership in pushing for gender parity in political representation, but has lagged in the
number of women bosses in the private sector. As a country with mixed evidence when it comes to
women in power, we believe that the role of social networks –as a source of information used by
individuals to decide behaviors and attitudes– should matter considerably. Moreover, we believe our
findings help further the study of gendered attitudes towards authority in Argentine society. We
suspect that as gender parity is met across all economic sectors and the workforce in Argentina, we
should see less evidence of gender solidarity among female evaluators of transgressing women in
power. Needless to say, we think more research across countries is required to answer this question.
In the meantime, we see possible implications of our findings in the field of political
representation, especially in the area of recruitment and selection of female candidates. Given the
results of our study, we intuit that if there were more women in party leadership and in general more
party leaders with gendered networks, women would be more likely to be placed in powerful positions,
such as heading a party list of candidates, or obtaining more powerful positions once in office. We
believe scholars focused on the selection and recruitment of female candidates –especially from a
comparative politics perspective- will find our study of interest.
Finally, a significant implication of our study and one area where we hope to see further
research concerns the possible link between our findings and the gender pay gap. Given how men
outnumber women in the top ranks of the corporate world, and what we have learned here about gender
solidarity in determining raises, we can venture a possible causal relationship. That is, if those with
the power to give raises tend to be men, and men tend to evaluate women less favorably, then female
employees suffer a systematic disadvantage. On the other hand, as Secretary Albright has advocated
and our findings seem to corroborate, if there were more female CEOs and board members who
developed bigger professional networks, women employees might be more likely to receive better pay.
Indeed, recent research has found evidence of a link between increased numbers of women on
corporate boardrooms and a narrowing of the gender pay gap.17
Abers, R. N. and M. E. Keck (2013). Practical authority: Agency and institutional change in Brazilian water
politics. New York, Oxford University Press.
Andreoni, J. and L. Vesterlund (2001). "Which is the fair sex? Gender differences in altruism." Quarterly
Journal of Economics: 293-312.
Barnes, T. D. (2016). Gendering Legislative Behavior: Institutional Constraints and Collaboration. New York
City, Cambridge University Press.
Barnes, T. D. and E. Beaulieu (2014). "Gender Stereotypes and Corruption: How Candidates Affect
Perceptions of Election Fraud." Politics & Gender 10(03): 365-391.
Blau, F. D. and L. M. Kahn (2000). Gender differences in pay, National bureau of economic research.
Bohnet, I., et al. (2008). "Betrayal aversion: Evidence from brazil, china, oman, switzerland, turkey, and the
united states." The American Economic Review: 294-310.
Bohnet, I. and R. Zeckhauser (2004). "Trust, risk and betrayal." Journal of Economic Behavior &
Organization 55(4): 467-484.
Budig, M. J. (2002). "Male advantage and the gender composition of jobs: Who rides the glass escalator?"
Social Problems 49(2): 258-277.
Caiazza, A. and B. Gault (2006). Acting from the Heart. Gender and Social Capital. B. O'Neill and E.
Gidengil. New York, Routledge: 99-125.
Calvo, E. and M. V. Murillo (2013). "When Parties Meet Voters Assessing Political Linkages Through
Partisan Networks and Distributive Expectations in Argentina and Chile." Comparative Political Studies
Campion, P. and W. Shrum (2004). "Gender and science in development: Women scientists in Ghana, Kenya,
and India." Science, Technology & Human Values 29(4): 459-485.
Carpenter, D. and C. D. Moore (2014). "When Canvassers Became Activists: Antislavery Petitioning and the
Political Mobilization of American Women." American Political Science Review 108(03): 479-498.
Cohen, P. N., et al. (2009). "Stalled Progress? Gender segregation and wage inequality among managers,
1980‐2000." Work and Occupations 36(4): 318-342.
Djupe, P., et al. (2016). "The political consequences of gender in social networks." British Journal of Political
Dolan, K. and T. Lynch (2013). "It Takes a Survey: Understanding Gender Stereotypes, Abstract Attitudes,
and Voting for Women Candidates." American Politics Research: 1532673X13503034.
Dollar, D., et al. (2001). "Are women really the “fairer” sex? Corruption and women in government." Journal
of Economic Behavior & Organization 46(4): 423-429.
Eckel, C. C. and P. J. Grossman (1996). "The relative price of fairness: Gender differences in a punishment
game." Journal of Economic Behavior & Organization 30(2): 143-158.
Eckel, C. C. and P. J. Grossman (2008). "Differences in the economic decisions of men and women:
Experimental evidence." Handbook of experimental economics results 1: 509-519.
Erickson, B. H. (2006). Persuasion and perception: New models of network effects on gendered issues.
Gender and Social Capital. B. O'Neill and E. Gidengil. New York, Routledge: 293-322.
Esarey, J. and G. Chirillo (2013). "“Fairer Sex” or Purity Myth? Corruption, Gender, and Institutional
Context." Politics & Gender 9(04): 361-389.
Fajak, A. and S. A. Haslam (1998). "Gender solidarity in hierarchical organizations." The British Journal of
Social Psychology 37: 73.
Friesdorf, R., et al. (2015). "Gender Differences in Responses to Moral Dilemmas A Process Dissociation
Analysis." Personality and Social Psychology Bulletin: 0146167215575731.
Gilligan, C. (1982). In a different voice. Cambridge, Harvard University Press.
Gottlieb, J., et al. (2016). "Do Men and Women Have Different Policy Preferences in Africa? Determinants
and Implications of Gender Gaps in Policy Prioritization." British Journal of Political Science: 1-26.
Huddy, L. and N. Terkildsen (1993). "Gender stereotypes and the perception of male and female candidates."
American Journal of Political Science: 119-147.
Ibarra, H. (1997). "Paving an alternative route: Gender differences in managerial networks." Social
Psychology Quarterly: 91-102.
Iversen, T. and F. M. Rosenbluth (2010). Women, work, and politics: The political economy of gender
inequality, Yale University Press.
Koehler, J. J. and A. D. Gershoff (2003). "Betrayal aversion: When agents of protection become agents of
harm." Organizational Behavior and Human Decision Processes 90(2): 244-261.
Lammers, J., et al. (2009). "Iron ladies, men of steel: The effects of gender stereotyping on the perception of
male and female candidates are moderated by prototypicality." European Journal of Social Psychology 39(2):
Lawless, J. L. (2004). "Women, war, and winning elections: Gender stereotyping in the post-September 11th
era." Political Research Quarterly 57(3): 479-490.
McPherson, J. M. and L. Smith-Lovin (1986). "Sex segregation in voluntary associations." American
Sociological Review: 61-79.
Mendez, J. M. and T. Osborn (2010). "Gender and the perception of knowledge in political discussion."
Political Research Quarterly 63(2): 269-279.
Molyneux, M. (2002). "Gender and the silences of social capital: Lessons from Latin America." Development
and change 33(2): 167-188.
Morgan, J. and M. Buice (2013). "Latin American Attitudes toward Women in Politics: The Influence of Elite
Cues, Female Advancement, and Individual Characteristics." American Political Science Review 107(04):
Norris, P. and R. Inglehart (2006). Gendering social capital. Gender and social capital. B. O'Neill and E.
Gidengil. New York, Routledge: 73-98.
Putnam, R. D. (2000). Bowling alone: The collapse and revival of American community, Simon and Schuster.
Rudman, L. A. and P. Glick (1999). "Feminized management and backlash toward agentic women: the hidden
costs to women of a kinder, gentler image of middle managers." Journal of personality and social psychology
Sanbonmatsu, K. (2002). "Gender stereotypes and vote choice." American Journal of Political Science: 20-34.
Smith, R. A. (2012). "Money, Benefits, and Power A Test of the Glass Ceiling and Glass Escalator
Hypotheses." The ANNALS of the American Academy of Political and Social Science 639(1): 149-172.
Thames, F. C. and M. S. Williams (2010). "Incentives for personal votes and women’s representation in
legislatures." Comparative Political Studies 43(12): 1575-1600.
Weikart, L. A., et al. (2007). "The democratic sex: Gender differences and the exercise of power." Journal of
Women, Politics & Policy 28(1): 119-140.
Zheng, T., et al. (2006). "How Many People Do You Know in Prison?: using overdispersion in count data to
estimate social structure in networks." Journal of the American Statistical Association 101: 474.
1 “Madeleine Albright on Barriers Broken and Barriers that Remain” at The Wall Street Journal,
May 7, 2012. Accessed on February 5, 2016 at http://www.wsj.com/articles.
2 Sheryl Sandberg’s December 2010 Ted Talk. Accessed on February 5, 2016 at
3 We operationalize personal network as the number of individuals with whom a respondent
interacts on a yearly basis. This includes not only the close personal ties of an individual, but
also the extended personal ties, which in turn include acquaintances that interact with the
respondent on a given year. Further details of our definition and survey questions to capture
network size are reported in the Supplemental Information File to this article.
4 EPH, National Institute of Statistics and Census (INDEC) at www.indec.gov.ar.
5 For example, see “Even Famous Female Economists Get No Respect” at
respect.html or “Why Aggressive Women Can't Win At Work (And How To Deal)”
6 See, for example, “The Wage Gap in Pro Sports” at http://www.infoplease.com/spot/spt-
wagegap1.html and “The Hollywood gender discrimination investigation is on: EEOC contacts
women directors” at http://www.latimes.com/entertainment/movies/moviesnow/la-et-mn-
7 See http://www.forbes.com/sites/jennagoudreau/2012/05/21/a-new-obstacle-for-professional-
8 Findings are mixed on whether voters’ perceptions based on gender stereotypes shape their vote
choices. See Sanbonmatsu 2002; Thames and Williams 2010; and Dolan and Lynch 2014.
9 Again, findings are mixed (see Lawless and Fox and Dolan and Lynch 2014) and reinforce our
expectation that the question should be settled through a survey experiment.
10 In an interesting twist, when the authors showed pictures of the candidates that were not
prototypically gender-based, the gender effect disappeared (Lammers et al 2008).
11 Like Lammers et al (2008), Dolan and Lynch (2014) do not find evidence that gender
stereotypes matter to voters’ evaluation of real candidates all of the time.
12 The survey was implemented by the PASCAL research center of the University of San Martin
(Argentina) between the months of December 2014 and March 2015. Our survey experiment was
part of a five-question module, attached to the “General Survey for Media and TV Consumption”
(GSMTVC). Of the 7,494 respondents in the Gender Module, 4,068 where 18 or older and
employed or seeking employment when the survey was administered. Four different treatments
included over 1,000 respondents each. The survey included 78 questions answered in under 20
13 It is important to note that we embed a baseline raise of 10% for a “good boss” directly into
the vignette rather than using a six-way-treatment design. Results indicate that respondents
understood that 10% was the baseline for good bosses, as they withheld on average 2 to 6 points
from that baseline as a penalty. We use an embedded baseline to reduce heteroscedasticity from
respondents having different assumptions about what constitutes an acceptable raise, which
would have added significant noise to the experiment.
14 We added a one to all raises when computing the logs, 𝑙𝑛𝑟𝑎𝑖𝑠𝑒 𝑙𝑜𝑔𝑟𝑎𝑖𝑠𝑒 1, to prevent
respondents that gave no raises to bad bosses (1,081 cases) from being dropped from the sample.
15 The groups’ categories are fully described in the online SIF, using questions about the number
of individuals known to the respondent from different name, occupation, and employment
categories. The selection of group categories (Silvia, Patricia, Antonio, Francisco, Angel,
teacher, policemen or policewomen, lawyer, doctor, municipal public employee, provincial
employee, or judges, had-children, died, married, and victims of a crime) follows Calvo and
16 No question was asked on the gender of those who had sexually harassed the 139 respondents.
17 See http://www.voxeu.org/article/female-executives-and-gender-wage-gaps and