Content uploaded by Theodore Masters-Waage
Author content
All content in this area was uploaded by Theodore Masters-Waage on Mar 12, 2025
Content may be subject to copyright.
1
Women Advocates and Men Critics:
How Referees’ Gender Influences Candidates’ Likelihood of Receiving a Promotion
2
Abstract
External review letters (ERLs) play a critical role in the promotion and tenure (P&T) process.
However, recently, scholars have questioned their validity, given the strong relationships between
letter writer characteristics and letter content. Building on Madera et al. (2024), we develop a social
role-based theory of how letter writer gender affects letter content and voting outcomes. Results
from within-candidate analysis find gender differences in letters written for the same candidate, 1)
men letter writers used more personal pronouns (I/me/myself) and women used more other-
pronouns (she/he), 2) letters written by women - compared to men - had a more positive tone
overall, and 3) used less doubt language. Collectively, this suggests that women write more
supportive and candidate-focused letters than men in the P&T process. Confirming this, we find
that - controlling for school, discipline, scholarly productivity, and demographics - candidates with
a higher proportion of women letter writers (i.e., lower proportion of men) have more positive
P&T outcomes. These findings underscore the need for P&T reforms to improve equity in the
external review letter-writing and writer selection process.
Keywords: Science Careers, Promotion and Tenure, Gender, External Review Letters, Bias,
Social Role Theory, Decision Making
3
Highlights:
● Gender differences bias the content of external review letters (ERLs) in the promotion and
tenure (P&T) process.
● Women (vs. men) letter writers use less doubt language, fewer personal pronouns (e.g.,
I/me), more other pronouns (e.g., he/she/they), and an overall more positive tone.
● The gender composition of a candidate’s letter writers is linked with P&T outcomes, such
that a greater proportion of women letter writers was associated with fewer negative votes
at the college level and a higher likelihood of a positive provost vote.
4
External review letters (ERLs) are one of the single most important career-influencing elements
of researchers’ promotion and tenure (P&T) portfolios (Abbott et al., 2010), supporting the
continued career success of some faculty while stalling or ending the careers of others. By sourcing
referees who are at arm’s length from the candidate, these letters are lauded for providing an
“outsider view” that is less tainted by the subjective biases of the candidates' colleagues or
collaborators, and influenced by expertise in the candidates’ specific research domain. However,
recent research has cast significant doubt on this assumption (Bellamy et al., 2022; Cervato et al.,
2024). Most notably, Madera et al. (2024) found that the content of letters was more associated
with the characteristics of the letter writer than that of the candidate. This begs the question, given
that faculty have little to no control over who writes their letters, do the gatekeepers who make
these decisions (e.g., department chairs choosing between a man or a woman letter writer or a
professor at a high- or low-ranked university) indirectly influence a candidate's chance of receiving
a promotion? To investigate this question we examine to what extent letter writer gender is
associated with the linguistic content of letters and ultimately a candidate's P&T outcomes.
Gender is a key psychological factor determining social perception and behavior. Social
role theory argues that these gender differences originate from the division of labor within social
groups (Eagly et al., 1987). These social roles in turn influence women’s and men’s traits and
behaviors. According to this theory, masculinity is associated with agency, which is centered
around goal pursuit, power, and achievement. Conversely, femininity is associated with
communion, which is associated with a social orientation, desire for connection, and other-focus.
These two dimensions (agency and communion) are so profoundly linked to gender that research
has found that when anthropomorphizing inanimate objects, one of the first things people do is
assign the object a gender in order to align it with one of these two dimensions (Martin & Mason,
5
2022). This line of research views gender as a psychological construct for understanding the world
and for determining social role expectations that people typically conform to in society (Eagly et
al., 2000; Martin & Slepian, 2018, 2021).
In this paper we focus on a particular component of social role theory, which is that there
are gender differences in social perception, and as a result in language use. Whereas men are more
likely to view the social-world through an agentic lens, being more self-focused and achievement-
oriented, women are more likely to view the social world through a communion lens, being more
other-focused and affiliation-oriented. We argue that these differences in social perception lead to
differences in how women vs. men ERL writers evaluate candidates in P&T processes, with
consequential links to candidates' P&T outcomes.
This paper contributes to research on the role of gender in the psychology of language use.
Past research has suggested that women and men use different language in numerous domains
(Iosub et al., 2014; Park et al., 2016; Pitt, 2021; Schwartz et al., 2013). However, language use is
context-dependent, meaning that gender differences in language use may manifest differently
across different domains. This paper examines a particularly important context: when individuals
are writing evaluative assessments of the contributions of another person. Focusing on this specific
context, this paper makes hypotheses that contradict but supplement past literature on gender
differences in language use (e.g., the use of “I” language; Newman et al., 2008), providing a more
nuanced account of how gender differences manifest across contexts. Further, given the extremely
consequential nature of external review letters (and reference letters more broadly) for individuals’
careers (Abbott et al., 2010; Madera et al., 2024), this paper highlights how the role of gender in
the psychology of language influences not only writers themselves but also other people.
6
This research also makes significant contributions to research on the role of gender in
academia. Past scholarship has focused on how including women in the evaluation process affects
the progression of women within academia, however, results have been inconsistent. For example,
including more women on committees did not improve outcomes for women applicants to
professorships (Bagues et al., 2017), nor did women reviewers provide more favorable assessments
of women’s papers in the peer review process (Abrevaya & Hamermesh, 2012; Hengel, 2022).
However, research has found that increasing the representation of women on search committees
increases the representation of women applicants to faculty positions (Kazmi et al., 2022). In this
paper we approach the role of gender in academia from an alternative perspective, focusing on
gender differences in how women and men review P&T candidates irrespective of their gender. In
other words, instead of examining whether women (vs. men) are more supportive of other women
candidates, we assess whether women overall are more likely to advocate for a P&T candidate
irrespective of their gender. In doing so, we develop a theory of gender difference in how letter
writers fundamentally perceive and evaluate other faculty.
This paper also provides an extension to the work by Madera et al. (2024). First, Madera
et al. (2024) employed an exploratory approach to examine how letter writer and candidate
characteristics affected letter content. In the current paper, we develop theory-driven hypotheses
from psycholinguistics, the psychology of language use, and the role of gender in perception based
on social role theory. Guided by this theory, the majority of these hypotheses focus on different
linguistic variables than Madera et al. (2024), each of which has been studied heavily in research
on the psychology of language use, e.g., pronoun use (Carey et al., 2015; Kacewicz et al., 2014;
Pennebaker, 2011b). Second, the ten-fold increase in available data allows this paper to more
rigorously test these hypotheses and replicate the findings of Madera et al. (2024). Third, we test
7
whether the characteristics of the letter writer directly affect P&T voting outcomes by examining
whether the gender composition of a candidate’s letter writers (i.e., proportion of men vs. women)
is associated with their P&T outcomes.
External Review Letters (ERLs) and Voting Outcomes
Academic policy for selecting letter writers is variable across institutions and even across
departments in the same institution (Cervato et al., 2024; Hannon & Bergey, 2024). The main
similarity is that at most universities, P&T candidates provide their department chair or the chair
of the P&T committee with a list of prospective ERL writers with in-depth content knowledge of
the sub-disciplinary area of the promotion candidate. Then, department or committee chairs
supplement that list with names of senior faculty they consider disciplinary experts. Ensuring these
individuals are at arm’s length, letters are then solicited. Therefore, within the selection of letter
writers within the P&T process, letter writer gender is largely incidental.
However, given the critical role of language use in ERLs in P&T decision-making (Madera
et al., 2024), these decisions would potentially be consequential if gender is related to ERL content.
If gender was found to influence how letter writers perceive a candidate’s portfolio - as is expected
by social role theory (Eagly et al., 2000) - this could mean that men and women focus on different
content and represent strengths/weaknesses differently. In Madera et al. (2024), inductive analysis
found that women and men letter writers used different language; specifically, women wrote letters
with more clout and men wrote letters with more authenticity. In this paper, we investigate this
question more deeply. Using a deductive approach, we examine whether gender differences in
social perception result in women (vs. men) writing more supportive letters for P&T candidates,
thus increasing their likelihood of receiving a positive P&T outcome. To do so, we first focus on
8
specific linguistic differences we expect to see based on social role theory in how men and women
write letters.
Gender Differences in Language Use: Agentic vs. Communal Evaluations
Numerous studies have examined the relationship between gender and language
production. Most of these have focused on situations where men and women are writing about
themselves or issues that interest them. Large-scale studies have shown gender differences in how
men and women write on Facebook (Park et al., 2016; Schwartz et al., 2013), when being
interviewed about their organization (Pitt, 2021), and when collaborating with colleagues (Iosub
et al., 2014). Although the findings from these studies are informative, we argue that the effects of
gender on language use are unique to the context being studied. This is because letter writers are
providing an evaluative assessment of someone other than themselves as opposed to writing about
themselves. Therefore, gender differences in social perceptions play a greater role in determining
the language letters writers use. This can help explain why contrasting results are seen in different
contexts (Leaper & Robnett, 2011; Newman et al., 2008).
External review letters are a particularly unique context. Unlike the majority of past studies
on linguistic gender differences that examine how individuals talk about themselves or aspects of
their life (e.g., Babal et al., 2019; Leaper & Robnett, 2011; Lenard, 2016; Newman et al., 2008;
Pérez-Rosas & Mihalcea, 2014; Pitt, 2021), ERL writers are providing an evaluative assessment
of another person. This is innately a more interpersonal process involving the social perceptions
of others. Further, ERLs are unique because there is a power dynamic between the writer and the
recipient, where the writer is in a high-power position and can significantly influence the
candidate's career outcomes (Madera et al., 2024). Finally, the nuance of ERLs also means that
understanding gender differences in language use in this context requires novel theorizing.
9
Therefore, instead of developing hypotheses based on past research on gender differences
in language use, this paper uses social role theory - specifically the contrast between agentic men
and communal women - to explain how men and women differ in their social perceptions of others
(Eagly et al., 2000). This leads to some contrasting predictions compared to past results, such as
women’s use of tentative language (Leaper & Robnett, 2011) and pronoun use (Newman et al.,
2008). In doing so, we underline how gender differences in language use can vary between writing
contexts, specifically when writing a social evaluation of another person (e.g., reference letters)
compared to writing about oneself (e.g., social media posts).
Gender-Based Language Difference in ERLs. Evaluating a P&T candidate through an
ERL is a challenging task that requires letter writers to process a complex portfolio of
informational elements and communicate their conclusions in a succinct evaluation. The process
is inherently subjective, meaning that the letter will be a reflection of both the candidate's qualities
and the letter writer’s background, beliefs, and professional experiences. Ideally, ERLs should
reflect a candidate's qualities as a researcher, scientist, and educator. However, Madera et al.
(2024) showed that ERL content was more influenced by letter writer characteristics than
previously thought, calling into question the meritocratic basis of one of P&T decision-making’s
most crucial factors (Abbott et al., 2010). Specifically, ERL writers’ research productivity, status,
and gender were more predictive of letter content than the evaluated promotion candidates'
research productivity metrics. Through this paper, we propose and examine how, when, and why
agentic vs. communal lenses for perceiving the social world will relate to linguistic differences in
external review letters (ERLs) in the academic P&T process.
Agency and communion are viewed as two independent dimensions of social perception.
In their most general terms, agency can be viewed as a desire for independence and separation
10
from other organisms whereas communion is a desire for connection and unity with other
organisms (Bakan, 1966; Hsu et al., 2021; Wiggins, 1991). Therefore, in their original description
of the terms, Bakan (1966) equated agency with being competent, independent, competitive,
power-seeking, and communal with being warm, honest, agreeable, and generous. These two
dimensions - which align with masculinity and femininity, respectively (Martin & Slepian, 2021)
- have been shown to affect how individuals perceive other people around them. For example, an
agentic person may see another person as a potential threat, whereas a communal person may see
them as an ally. We argue that these differences in social perceptions, in turn, will affect how ERL
letter writers evaluate P&T candidates.
First, we propose that men and women letter writers will represent themselves and the
candidate in different ways, aligning with their respective gender roles (i.e., men-agentic and
women-communal). Research in linguistics has shown that the use of personal pronouns (e.g., I,
me, and myself) vs. the use of other or collective pronouns (e.g., he/she or we) serve an important
role in communication (Pennebaker, 2011a). For example, they can indicate social dominance
(Kacewicz et al., 2014), personality traits (Pennebaker, 2011b), and extreme traits such as
narcissism (Carey et al., 2015). In particular, increased use of other or collective pronouns indicates
that the writer is more actively thinking about people other than themselves (Pennebaker, 2011a).
For this reason, we expect that women, who are more likely to be communal letter writers, are
more likely to perceive the social world through the lens of empathy, other-focused concern, and
the collective (Bem, 1981; Sczesny et al., 2018). Alternatively, men being less communal letter
writers means they are more likely to perceive the social world through the lens of self-focused
concern and the individual. Subsequently, we expect women will be more likely to use other
pronouns in their reference letters whereas men are more likely to use personal pronouns.
11
Hypothesis 1: Gender differences in pronoun use such that men letter writers are more
likely to use pronouns referring to themselves (H1a) and women are more likely to use
pronouns referring to the candidates (H1b) and collective pronouns (H1c).
Agency and communion are also expected to affect individuals’ use of positive (vs.
negative) language in ERLs. According to social role theory (Eagly & Wood, 2012), communion
is aligned with the concept of warmth, which suggests that this dimension will lead to a more
positive outlook on a candidate’s qualities (Winquist et al., 1998; Sczesny et al., 2018). For
example, when evaluating a candidate’s research papers, letter writers with a more communal
social perception would be more likely to describe the paper as “good” or “great”, whereas an
agentic view might describe the paper in more neutral terms (Pietraszkiewicz et al., 2019). Along
with increasing the likelihood of using positive terms, communion may also decrease the
likelihood of using negative terms (Winquist et al., 1998; Sczesny et al., 2018). Communion is
also associated with increased compassion in social perception (Bakan, 1966); therefore, if a part
of a candidate’s portfolio reflects negatively on them, more letter writers with a more communal
social perception would be more likely to use neutral terms than agentic letter writers.
Hypothesis 2: Gender differences in tone use such that women (vs. men) letter writers will
use a more positive tone overall (H2a), are more likely to use positive words in their letters
(H2b), and are less likely to use negative words (H2c).
12
Another possible effect of agency and communion on language use is the extent to which
letter writers use doubt language. Doubt language is defined as the use of terms, phrases, or
statements that question a candidate's aptness for a position (Madera et al., 2019). Using such terms
in letters, such as “may” and “potentially,” has been shown to play an important role as they
provide a key insight for the promotion decision makers on the mindset of the letter writer (Madera
et al., 2019). Interpreting letters is not a straightforward process for promotion decision-makers
due in part to individuals' tendency to avoid negativity when evaluating others, termed the MUM
(Minimizing Unpleasant Message) effect (Dibble & Levine, 2010). Therefore, to infer the “true
evaluation,” decision-makers must often “read between the lines.” Doubt language is one way a
letter writer can avoid saying something negative but not give a full endorsement of a candidate.
For example, instead of saying, “This candidate should not receive tenure,” someone could say,
“This candidate might not be suited for a tenured position.”
Madera et al. (2019) found that women candidates received more doubt language in letters
of recommendation for academic positions. However, within the ERL context, we hypothesize that
women letter writers will use less doubt language. It is expected that because women are socialized
to be more communal than men, ERLs from women writers will be associated with less use of
doubt language, as it is associated with a greater desire to be supportive of others. In contrast, we
expect more agentic letter writers - i.e., men - might question a candidate’s abilities more because
agentic writing is more likely to entail elements of social comparison and competition.
Hypothesis 3: Gender differences in doubt use such that women (vs. men) letter writers
are less likely to use doubt language in ERLs.
13
The Woman Advocacy Advantage
The difference in letters written by women and men is interesting on a linguistic level, but
it could be instrumental from a decision-making point of view. ERLs have been touted as one of
(if not the) most important documents influencing P&T decisions (Abbott et al., 2010) and have
been empirically linked to voting outcomes (Madera et al., 2024; Masters-Waage et al., 2024). A
common feature of the hypotheses above is that women - by perceiving the social world through
more of a communion lens - write more supportive letters for P&T candidates than men.
Specifically, these are letters that a) have a more positive (and less negative) tone, and b) use less
doubt language. In contrast, men by perceiving the world through a more agentic lens - would
write more negative letters and use more doubt language, which could be associated with their
increased competitiveness relative to women (Niederle & Vesterlund, 2007; Saccardo et al., 2018).
Beyond these specific components, it is also likely that individuals who are more agreeable and
generous would also be more likely to view a candidate positively and write a supportive letter
(Habashi et al., 2016). As a result, we hypothesize that having more women letter writers will
benefit candidates in the P&T process.
Hypothesis 4: The gender composition of referees a candidate has is associated with voting
outcomes, such that greater proportion of women (vs. men) letter writers for a P&T
candidate improves voting outcomes at the department, college, and provost level.
This research builds directly on the research of Madera et al. (2024), using a large
consortium dataset derived from understanding equity and validity in promotion and tenure
decision-making, specifically external review letter use. This new dataset includes promotion and
14
tenure decisions across six institutions, including external review letters (ERLs) for the candidates.
The language used in letters was analyzed using Pennebaker et al.’s Linguistic Inquiry and Word
Count (LIWC) software (Pennebaker et al., 2015), and data on voting outcomes was provided by
the Provost office at each institution. We were also granted access to the gender of the candidate
and the letter writer, which were self-reported and available in the institutional records. Additional
data, such as candidate performance (specifically research output), discipline, and tenure, were
also provided and utilized as control variables.
Methods
Data collection methods for this project followed the methods developed by Madera et al.
(2024).
Sample
The present study uses ERLs submitted for P&T decisions from across 6 research-intensive
institutions. The sample consists of 10,056 ERLs submitted for P&T portfolios of 1,748 candidates
(64% men) seeking promotion to Associate (n = 920; 62% men) or Full Professor (n = 828; 66%
men) from 2015-2022. ERL writers were majority male (72%) and at the full professor rank (90%).
The mean number of external letters per candidate was 6.25 (SD = 1.77). On average, the ERLs
contained 1083.27 words (SD = 619.61).
Measures
ERL linguistic features. Consistent with the methodological approach outlined by Madera
et al. (2024), we used a computer text analysis program, the Linguistic Inquiry and Word Count
program (LIWC), to examine the linguistic features of the ERLs (Tausczik & Pennebaker, 2010).
The LIWC program has been used extensively in organizational behavior. For example, LIWC has
been used to analyze large qualitative data sets focused on individual behaviors, scholars have
15
made significant strides with LIWC to study leadership behaviors, and team performance has also
been analyzed using LIWC (Short et al., 2018). LIWC has also been used to examine ERLs in
academia (Han et al., 2021; Miller et al., 2019).
We used LIWC to measure the frequency of words related to our key hypotheses (see for
example terms see Table 1). To measure personal pronouns we used the dictionaries for the use of
“I” which includes terms like me and myself, for other pronouns we included the “she/he” and
“they” dictionary (note, we didn’t include “you” as in ERLs when letter writers say you they are
more likely referring/talking to the P&T committee members, not about the other candidate), for
collective pronouns we used the “we” dictionary. To measure emotional tone we used the “tone”
dictionary, which captures overall tone by calculating the amount of positive words and then
deducting the amount of negative words. In addition, we also included dictionaries for “positive
tone” and “negative tone” separately. To capture doubt language, we used the “doubt” dictionary
developed and validated by Madera et al. (2019). In addition to our hypothesis tests, we also
included the “communion” and “agency” dictionaries developed by Madera et al. (2009) in order
to see if there are gender differences in communion and agency language overall.
Promotion and/or tenure votes. We used the outcome of three P&T recommendations
based on the vote outcomes at the departmental, college, and provost levels. For the department
and college level we calculate the percentage of votes for the candidate that were negative
(negative vote percentage). At the provost level, we used a dummy variable indicating if the vote
was positive “1” or negative “0”.
Letter writer gender. Letter writer demographics were recorded using the same method
employed as Madera et al. (2024). This involved coders using google scholar profiles and
16
institution pages to determine each letter writer's gender. A dummy variable was created indicating
if participants were women (coded as “1”) or men (coded as “0”).
Data analysis strategy
The data used in the current study has a nested data structure. Specifically, external letter
writers were nested within candidates. Therefore, external letter writers’ characteristics, including
letter writers’ gender, school rank, and h-index, and external letter linguistic terms were considered
level-1 variables. Candidates’ characteristics, including gender, h-index, research field (i.e., STEM
or not), and the department-, college-, and university-levels P&T decisions were level-2 variables.
Analyses were conducted at both levels. To examine gender differences in language use,
analyses were conducted at level-1. Linguistic variables (e.g., tone, pronouns, doubt) were group
mean-centered at the candidate level. To group mean-center, we subtracted each candidate’s mean
language score across their ERLs from each of their ERL data points (for each linguistic variable,
e.g., doubt, tone, other pronouns). Doing so allows us to focus on within-candidate differences in
letter content - how two different letter writers’ language varied for the same candidate (i.e., man
vs. woman) - while parceling out the between-candidate differences. In other words, the analyses
examined how men vs. women wrote letters for the same candidate. In addition, to account for
differences between letter writers, we also controlled for letter writers' discipline, rank, and the US
news ranking of their home institution. We controlled for letter writer discipline to account for
differences in norms across academic disciplines, for example, it might be that letters written for
mathematicians are linguistically different to those written for sociologists. We controlled for US
news ranking as Madera et al. (2024) found this factor significantly affected linguistic content.
Finally, we controlled for the rank of the letter writer primarily due to gender disparities in women
17
vs. men letter writers. For example, 74% of the letter writers who were full professor were men,
but only 53% of those at letter writers who were associate professor were men
1
.
To examine the relationship between letter writer gender and voting outcomes, analyses
were conducted at level-2. The proportion of women letter writers for each candidate was
calculated (0% to 100%) and this was used to predict voting outcome at the three levels. Controls
at this level included candidate characteristics: institution, discipline, gender, underrepresented
minority (URM) status, rank being promoted to, h index, total citations, and number of grants
received as a PI. These control variables were used to account for differences across disciplines
and institutions, along with factors that have been shown to impact P&T decisions, i.e., scholarly
productivity and grant funding. URM status was controlled for given that past research has shown
these individuals face additional barriers in the P&T process (Masters-Waage et al., 2024).
Note, women were coded as “1” and men as “0” across all analyses, meaning that results
must be interpreted accordingly. In other words, a significant positive relationship means women
letter writers used more of a given language variable, whereas a negative relationship means men
used more.
Data Availability Statement: Data used in analyses is available at this OSF page,
https://osf.io/2hjgr/?view_only=19141930271d4c11b533469a26e1cded
Results
Descriptive statistics and correlations for all study variables are provided in Table 2. Before
analyzing the study hypotheses, we examined gender differences in communion and agency
1
Note supplemental analyses were conducted including letter writers' h-index a control variable (see
Table S1). The conclusions from these supplemental analyses are broadly the same as those without this
control variable, the only difference is that the effect on “tone” becomes marginally non-significant.
However, we suspect this is a result of reduced power. Due to incomplete data on the h-index for letter
writers using this as a control excludes over 2,500 letter writers from the analyses due to incomplete data.
18
language using the dictionary developed by Madera et al. (2009). In their research they found that
advertisements targeted at women included more communion language but did not investigate
whether women use more communion language when writing. Consistent with social role theory,
in our dataset women as compared to men were found to use significantly more communion
language in their ERLs (b = .05, SE = .01, p < .001). However, gender was not related to agency
language (b = .01, SE = .02, p = .37), which could be attributable to the primacy of communion in
social cognition (Wojciszke & Abele, 2008).
Gender-Based Language Difference in ERLs
The first set of analyses focused on gender differences in pronoun use. Results found that
women used personal pronouns (e.g., “I”, “me”, “myself”) significantly less (b = -.16, SE = .02, p
< .001); this provides support for hypothesis 1a, in other words, men use personal pronouns more.
In terms of other pronouns, we examined the use of they and she/he terms. Women letter writers
were not more likely than men to use “they” in letters (b = .003, SE = .01, p = .56) but were
significantly more likely to use the terms “he/she” (b = .11, SE = .02, p < .001). However, given
that the incidence of “they” was much lower in letters than the use of “she/he”, we created a
composite measure of both and found that overall women were more likely to use other pronouns
in their letters (b = .12, SE = .02, p < .001), therefore there was support for hypothesis 1b. Finally,
in terms of collective pronouns, women were not more likely than men to use the collective term
“we” in letters (b = -.005, SE = .01, p = .38), which does not support hypothesis 1c. In sum,
consistent with prediction, men letter writers used personal pronouns more and women letter
writers used other pronouns more, but counter to predictions, women did not use collective
pronouns more.
19
Next, we examined gender differences in tone. We evaluated both overall tones (positive
words - negative words), along with specifically whether letter writers used more positive or
negative words. Results found that women letter writers used a significantly more positive tone
overall in their letters (b = .64, SE = .30, p = .03). However, this difference was only at the
aggregate level as women neither use positive words (b = .03, SE = .02, p = .17) nor negative
words (b = -.01, SE = .01, p = .35) differently from men. In sum, there was support for hypothesis
2a, that women in general use a more positive tone in their letters.
Our final linguistic analyses focused on gender differences in the use of doubt language.
Consistent with hypothesis 3, we found that letters written by women (vs. men) contained
significantly less doubt language (b = -.03, SE = .01, p < .001).
In sum, the results from linguistic analyses are largely aligned with our hypotheses, finding
that men used more personal pronouns, and women used more other pronouns, a more positive
overall tone, and less doubt language; these within-candidate analyses of the relationship between
letter writer gender and letter content can be found in Table 3. Further, the results for tone - which
were in the same direction in Madera et al. (2024)
2
but weren’t significant - were significant in
these analyses, which is likely attributable to the increased statistical power of the analyses.
The Women Advocacy Advantage
These analyses examined the relationship between the gender composition of letter writers
that a candidate had and voting outcomes. These analyses were conducted at level two (i.e., the
candidate level). These analyses controlled for candidate discipline, institution code, rank being
promoted to, candidate demographics (URM status and gender), and scholarly productivity (h-
2
Given that the dataset used in this study is an extension of one used in Madera et al. (2024) we also
sought to replicate the relationship between gender and clout and authenticity observed in that paper. The
results reported in that study were replicated in this larger dataset with women using more clout language
(b=2.13, SE=.23, p<.001) and less authentic language (b=-1.09, SE=.18, p<.001).
20
index, total citations, and grants received as PI). The independent variable was the proportion of
letter writers for the candidates who were women.
Analyses were conducted at the department, college, and provost level using negative vote
percentage. There were no differences at the department level (b = -.03, SE = .03, p = .33).
However, at the college level, an increased proportion of women letter writers was associated with
a lower negative vote percentage (b = -.07, SE = .03, p = .047). To estimate effect size, we
calculated the standardized beta coefficients for the relationship between the proportion of women
letter writers and college-level negative vote percentage. One standard deviation change in the
proportion of women letter writers was associated with a decrease of .07 standard deviations in
negative voting percentage.
Finally, at the provost level, which is the penultimate and typically most important vote in
the P&T process, a higher proportion of women letter writers was associated with greater
likelihood of receiving a positive provost vote (b = 2.10, SE = .94, p = .02; Odds Ratio = 8.21).
This provides support for hypothesis 4 (see Table 4), that P&T candidates benefit from having a
higher proportion of women letter writers.
General Discussion
In an ideal world promotion decisions would be based solely on the relevant achievements
of promotion candidates; however, this paper shows that the external review letter writer’s gender
plays an instrumental role in the process. When reviewing the same candidate, women (compared
to men), on average, wrote letters that were more other-focused, had a more positive tone, and
used less doubt language. Gender differences in letter writing are not just artifacts of different
modes of social perceptions; they also are related to promotion decisions. We found that the gender
composition of candidate’s letter writers influenced one of the most high-stakes decisions in their
21
career. Specifically, candidates who had a higher proportion of women letter writers received
fewer negative votes at the college level and were more likely to receive a promotion (i.e., positive
provost vote), a phenomenon we term the women advocacy advantage. Building on the exploratory
findings of Madera et al. (2024), this paper highlights the biases inherent in the external review
process.
The gender differences in letters found in this paper are aligned with the proposition of
social role theory (Eagly & Wood, 2012), that women are more communion-focused and men are
more agency-focused. A good example of this is that men were more likely to use “I” language,
whereas women were more likely to use “she/he”, suggesting that there is a difference in how the
genders perceive the self and others when evaluating others. Given the critical role that reference
letters play in organizations - particularly the academic P&T process - these results shed light on
bias in letter writers.
This study advances research on the psychology of language, which has primarily focused
on gender differences in what language individuals use (Leaper & Robnett, 2011; Lenard, 2016;
Newman et al., 2008; Pérez-Rosas & Mihalcea, 2014), but less so on how these differences
influence others' decision making. By writing letters that are more other-focused, positive, and less
doubtful, we find that women are ultimately writing letters that are more supportive of the
candidate, i.e., they are advocates. By virtue, this means men are writing letters that are less
supportive, i.e., they are critics. Analyzing the data at the candidate level, consistent with this, we
found that candidates with a higher proportion of women letter writers were more likely to receive
positive votes at the college and provost level. Critically, this is in a situation where the decision-
makers knew the gender of the letter writer, so even if the decision-makers were adjusting their
22
evaluation based on the letter writer’s gender, there was still variation beyond that accounted for
by the committee.
A key contribution of this paper is the identification of a women advocacy advantage,
which could equally be termed the critical man disadvantage. This is the observed phenomenon
that women provide more supportive social evaluations of others, which is a finding that has
implications beyond just the context of external review letters. Review letters are an example of a
much more general phenomenon; the evaluation of other people (or their work). This kind of
evaluation is very common in academia, most notably in the peer-review process. Peer review is
another domain in which it is assumed that reviewers can remain objective or that the extent of
their subjective biases is limited. However, in light of the current results, future researchers should
examine whether there are also gender differences in the linguistic content of reviews, and, perhaps
more critically, if this affects publishing decisions. In addition, in the model predicting the provost
vote, the only two significant predictors were gender composition and h-index. The influence of
research output on P&T decisions is well documented, but given the critiques of the h-index
measure (Bornmann & Daniel, 2007), it is notable that this metric is associated with the final
provost decision.
These results also contribute to research on gender differences in the psychology of
language use (Leaper & Robnett, 2011; Lenard, 2016; Newman et al., 2008; Pérez-Rosas &
Mihalcea, 2014). In a large meta-analysis of studies Newman et al. (2008) found that women used
“I” words more than men, opposite our results. Similarly, in a dataset of 25,000 college admissions
essays collected by Pennebaker et al. (2014), women were found to use more “I” words and use
negative tone words more. To explain the conflicting results we highlight the unique nature of
review letters. The studies in Newman et al. (2008) and data from Pennebaker et al. (2014) include
23
contexts where people are speaking about themselves or things in their life. Alternatively, review
letters are based on social perception. Therefore, while “I” language may reflect self-doubt in a
college admissions essay (Pennebaker, 2011a), it may reflect something very different when
someone is reviewing another person’s credentials. In sum, this study shows that theories of gender
difference in language use must be adapted to different contexts. Further, it highlights review
letters as an interesting area of research in the psychology of language use, encouraging future
scholars to understand what makes this context unique.
From a practical perspective the results of this paper strengthen calls for reforming the
letter-writing process in P&T decision-making (Bellamy et al., 2022; Cervato et al., 2024; Hannon
& Bergey, 2024; Madera et al., 2024). The majority of gender bias studied in academia has focused
on discrimination, specifically against women (Aragón et al., 2023; Ceci et al., 2023; Moss-
Racusin et al., 2012). This paper instead examined gender differences in social perceptions in an
academic context. Although institutions are not able to alter gendered social perceptions in the
letter-writing process, they could implement structure into the P&T process that reduces the role
of gender. Typically, biases are more pronounced in weak situations where there is a less clear
directive of what is required from people (Snyder & Ickes, 1985). To address this, departmental
and university policy could require structured ERLs from letter writers to present a clearer directive
of what is required from them. An example of where standardized structure has helped reduce bias
is in the interview process (Kausel et al., 2016). Therefore, applying this same approach to ERLs
is a promising practical intervention to mitigating the role of gender bias in letter content.
A lot of responsibility also falls on the individuals in charge of selecting letter writers
(usually department chairs) who should apply transparent rubrics for deciding who will be a
candidate’s letter writers. Hannon and Bergey (2024) reviewed ERL policy at R1 PhD institutions
24
with very high research activity and found a large degree of variation. Along with the number of
letters requested varying from three to ten, there was also variation in what level of influence the
candidate had in letter writer choice along with the network-distance and anonymity of the letter
writer. Given the role that letter writer characteristics play in the content of letters - evident in this
study and Madera et al. (2024) - one method departments and universities should pursue for
reducing the role of extraneous factors in the P&T process would be to minimize this variation. In
addition, department chairs should be cognizant of how their decisions are likely to affect P&T
candidates. For example, instead of soliciting from the first letter writers to accept an invitation,
chairs should ensure that a candidate's set of letter writers is reasonably balanced in terms of the
rank of their university, scholarly productivity (see Madera et al., 2024), and - as this paper shows
- gender. Further, more research should be conducted on how the women advocacy advantage
interacts with the characteristics of the candidate, as it is possible that it may benefit some
candidates more than others.
Like all papers, this paper should be viewed in light of its strengths and weaknesses. A
limitation of this paper is that we focused on a specific context, academic promotion, and tenure,
limiting the generalizability of these findings across all organizations. However, there are also
benefits of focusing on a specific context and, in particular, academia. First, the conducting of
reviews at arm’s length means that the candidate's personal relationship did not affect the content
of the letters and the anonymity of the letter writer to the candidates means there was no benefit
for the letter writer to be more positive. Second, each candidate received multiple letters, meaning
that the linguistic content could be centered at the candidate level, allowing us to conduct within-
candidate analyses to examine gender differences in letters written for the same candidate. Another
strength of this study is the consortium-based nature of the data collected, which has resulted in a
25
large dataset with a high degree of disciplinary and institutional variability, increasing the
generalizability of these results. In sum, results suggest that the letter writer's gender affects letter
content and candidates’ P&T outcomes, highlighting the need for university policy revisions to
promote equity and fairness in faculty advancement and that promotion candidate evaluations are
in the eye of the letter writer.
26
References
Abbott, A., Cyranoski, D., Jones, N., Maher, B., Schiermeier, Q., & Van Noorden, R. (2010).
Metrics: Do metrics matter? Nature, 465(7300), Article 7300.
https://doi.org/10.1038/465860a
Abrevaya, J., & Hamermesh, D. (2012). Charity and Favoritism in the Field: Are Female
Economists Nicer to Each Other? The Review of Economics and Statistics, 94(1), 202-
207.
Aragón, O. R., Pietri, E. S., & Powell, B. A. (2023). Gender bias in teaching evaluations: The
causal role of department gender composition. Proceedings of the National Academy of
Sciences, 120(4), e2118466120.
Babal, J. C., Gower, A. D., Frohna, J. G., & Moreno, M. A. (2019). Linguistic analysis of
pediatric residency personal statements: Gender differences. BMC Medical Education,
19, 1–9.
Bagues, M., Sylos-Labini, M. & Zinovyeva, N. (2017). Does the gender composition of
scientific committees matter? American Economic Review 107 (4), 1207-1238.
Bakan, D. (1966). The duality of human existence: An essay on psychology and religion (p. 242).
Rand Mcnally.
Bellamy, P., Haynes, C. L., Martin, L., Mirabile, S., Niv, Y., Rose, J. K., & Ross, R. A. (2022).
A guide for writing anti-racist tenure and promotion letters. Elife, 11, e79892.
Bem, S. L. (1981). Gender schema theory: A cognitive account of sex typing. Psychological
Review, 88(4), 354.
Bornmann, L., & Daniel, H. D. (2007). What do we know about the h index?. Journal of the
American Society for Information Science and technology, 58(9), 1381-1385.
27
Carey, A. L., Brucks, M. S., Küfner, A. C., Holtzman, N. S., Back, M. D., Donnellan, M. B.,
Pennebaker, J. W., & Mehl, M. R. (2015). Narcissism and the use of personal pronouns
revisited. Journal of Personality and Social Psychology, 109(3), e1.
Ceci, S., Kahn, S., & Williams, W. (2023). Exploring gender bias in six key domains of
academic science: An adversarial collaboration. Psychological Science in the Public
Interest : A Journal of the American Psychological Society, 15291006231163179.
https://doi.org/10.1177/15291006231163179
Cervato, C., Bilen‐Green, C., Johnson, C. A., Koretsky, C., & Minerick, A. (2024). External
Review Letters in Promotion and Tenure: Recommendations for Chairs. The Department
Chair, 34(4), 14–17.
Niederle, M., & Vesterlund, L. (2007). Do women shy away from competition? Do men compete
too much?. The quarterly journal of economics, 122(3), 1067-1101.
Dibble, J. L., & Levine, T. R. (2010). Breaking good and bad news: Direction of the MUM effect
and senders’ cognitive representations of news valence. Communication Research, 37(5),
703–722.
Eagly, A. H. (1987). Sex differences in social behavior: A social-role interpretation. Lawrence
Erlbaum Associates.
Eagly, A. H., & Wood, W. (2012). Social role theory. Handbook of Theories of Social
Psychology, 2, 458–476.
Eagly, A. H., Wood, W., & Diekman, A. B. (2000). Social role theory of sex differences and
similarities: A current appraisal. In The developmental social psychology of gender (pp.
123–174). Lawrence Erlbaum Associates Publishers.
Habashi, M. M., Graziano, W. G., & Hoover, A. E. (2016). Searching for the prosocial
28
personality: A big five approach to linking personality and prosocial behavior.
Personality and Social Psychology Bulletin, 42(9), 1177-1192.
Han, A. Y., French, J. C., Tu, C., Obiri-Yeboah, D., Lipman, J. M., & Prabhu, A. S. (2021).
Don't Judge a Letter by its Title: linguistic analysis of letters of recommendation by
author's academic rank. Journal of Surgical Education, 78(6), e19-e27.
Hanek, K. J., Garcia, S. M., & Tor, A. (2016). Gender and competitive preferences: The role of
competition size. Journal of Applied Psychology, 101(8), 1122.
Hannon, L., & Bergey, M. (2024). Policy variation in the external evaluation of research for
tenure at U.S. universities. Research Evaluation, rvae018.
https://doi.org/10.1093/reseval/rvae018
Hengel, E. (2022). Publishing while female: Are women held to higher standards? Evidence
from peer review. The Economic Journal, 132(648), 2951-2991.
Hsu, N., Badura, K. L., Newman, D. A., & Speach, M. E. P. (2021). Gender,“masculinity,” and
“femininity”: A meta-analytic review of gender differences in agency and communion.
Psychological Bulletin, 147(10), 987.
Iosub, D., Laniado, D., Castillo, C., Fuster Morell, M., & Kaltenbrunner, A. (2014). Emotions
under discussion: Gender, status and communication in online collaboration. PloS One,
9(8), e104880.
Kacewicz, E., Pennebaker, J. W., Davis, M., Jeon, M., & Graesser, A. C. (2014). Pronoun Use
Reflects Standings in Social Hierarchies. Journal of Language and Social Psychology,
33(2), 125–143. https://doi.org/10.1177/0261927X13502654
Kausel, E. E., Culbertson, S. S., & Madrid, H. P. (2016). Overconfidence in personnel selection:
When and why unstructured interview information can hurt hiring decisions.
29
Organizational Behavior and Human Decision Processes, 137, 27–44.
Kazmi, M. A., Spitzmueller, C., Yu, J., Madera, J. M., Tsao, A. S., Dawson, J. F., & Pavlidis, I.
(2022). Search committee diversity and applicant pool representation of women and
underrepresented minorities: A quasi-experimental field study. Journal of Applied
Psychology, 107(8), 1414.
Leaper, C., & Robnett, R. D. (2011). Women are more likely than men to use tentative language,
aren’t they? A meta-analysis testing for gender differences and moderators. Psychology
of Women Quarterly, 35(1), 129–142.
Lenard, D. B. (2016). Gender differences in the personal pronouns usage on the corpus of
congressional speeches. Journal of Research Design & Statistics in Linguistics &
Communication Science, 3(2).
Madera, J. M., Hebl, M. R., Dial, H., Martin, R., & Valian, V. (2019). Raising doubt in letters of
recommendation for academia: Gender differences and their impact. Journal of Business
and Psychology, 34, 287–303.
Madera, J. M., Hebl, M. R., & Martin, R. C. (2009). Gender and letters of
recommendation for academia: Agentic and communal differences. Journal of Applied
Psychology, 94(6), 1591–1599. https://doi.org/10.1037/a0016539
Madera, J. M., Spitzmueller, C., Yu, H., Edema-Sillo, E., & Clarke, M. S. F. (2024). External
review letters in academic promotion and tenure decisions are reflective of reviewer
characteristics. Research Policy, 53(2), 104939.
https://doi.org/10.1016/j.respol.2023.104939
Martin, A. E., & Mason, M. F. (2022). What does it mean to be (seen as) human? The
importance of gender in humanization. Journal of Personality and Social Psychology,
30
123(2), 292–315. https://doi.org/10.1037/pspa0000293
Martin, A. E., & Slepian, M. L. (2018). Dehumanizing Gender: The Debiasing Effects of
Gendering Human-Abstracted Entities. Personality and Social Psychology Bulletin,
44(12), 1681–1696. https://doi.org/10.1177/0146167218774777
Martin, A. E., & Slepian, M. L. (2021). The Primacy of Gender: Gendered Cognition Underlies
the Big Two Dimensions of Social Cognition. Perspectives on Psychological Science,
16(6), 1143–1158. https://doi.org/10.1177/1745691620904961
Masters-Waage, T., Spitzmueller, C., Edema-Sillo, E., St. Aubin, A., Penn-Marshall, M.,
Henderson, E., Lindner, P., Werner, C., Rizzuto, T., & Madera, J. (2024).
Underrepresented minority faculty in the USA face a double standard in promotion and
tenure decisions. Nature Human Behaviour, 1-12.
Miller, D. T., McCarthy, D. M., Fant, A. L., Li-Sauerwine, S., Ali, A., & Kontrick, A. V. (2019).
The standardized letter of evaluation narrative: differences in language use by gender.
Western Journal of Emergency Medicine, 20(6), 948.
Moss-Racusin, C. A., Dovidio, J. F., Brescoll, V. L., Graham, M. J., & Handelsman, J. (2012).
Science faculty’s subtle gender biases favor male students. Proceedings of the National
Academy of Sciences, 109(41), 16474–16479.
Newman, M. L., Groom, C. J., Handelman, L. D., & Pennebaker, J. W. (2008). Gender
differences in language use: An analysis of 14,000 text samples. Discourse Processes,
45(3), 211–236.
Park, G., Yaden, D. B., Schwartz, H. A., Kern, M. L., Eichstaedt, J. C., Kosinski, M., Stillwell,
D., Ungar, L. H., & Seligman, M. E. (2016). Women are warmer but no less assertive
than men: Gender and language on Facebook. PloS One, 11(5), e0155885.
31
Pennebaker, J. W. (2011a). The secret life of pronouns. New Scientist, 211(2828), 42–45.
Pennebaker, J. W. (2011b). Your use of pronouns reveals your personality. Harvard Business
Review, 89(12), 32–33.
Pennebaker, J. W., Booth, R. J., Boyd, R. L., & Francis, M. E. (2015). Linguistic inquiry and
word count: LIWC2015. Austin, TX: Pennebaker Conglomerates (www.LIWC.net).
Pennebaker, J. W., Chung, C. K., Frazee, J., Lavergne, G. M., & Beaver, D. I. (2014). When
small words foretell academic success: The case of college admissions essays. PloS one,
9(12), e115844.
Pérez-Rosas, V., & Mihalcea, R. (2014). Gender differences in deceivers writing style. 163–174.
Pietraszkiewicz, A., Formanowicz, M., Gustafsson Sendén, M., Boyd, R. L., Sikström, S., &
Sczesny, S. (2019). The big two dictionaries: Capturing agency and communion in
natural language. European Journal of Social Psychology, 49(5), 871–887.
https://doi.org/10.1002/ejsp.2561
Pitt, C. (2021). Gender and the CMO: do the differences make a difference? Journal of Strategic
Marketing, 29(4), 301–315.
Saccardo, S., Pietrasz, A., & Gneezy, U. (2018). On the size of the gender difference in
competitiveness. Management Science, 64(4), 1541-1554.
Schwartz, H. A., Eichstaedt, J. C., Kern, M. L., Dziurzynski, L., Ramones, S. M., Agrawal, M.,
Shah, A., Kosinski, M., Stillwell, D., Seligman, M. E. P., & Ungar, L. H. (2013).
Personality, Gender, and Age in the Language of Social Media: The Open-Vocabulary
Approach. PLOS ONE, 8(9), e73791. https://doi.org/10.1371/journal.pone.0073791
Sczesny, S., Nater, C., & Eagly, A. H. (2018). Agency and communion: Their implications for
gender stereotypes and gender identities. In Agency and communion in social psychology
32
(pp. 103-116). Routledge.
Short, J. C., McKenny, A. F., & Reid, S. W. (2018). More than words? Computer-aided text
analysis in organizational behavior and psychology research. Annual Review of
Organizational Psychology and Organizational Behavior, 5(1), 415-435.
Snyder, M., & Ickes, W. (1985). Personality and social behavior. In Handbook of social
psychology (pp. 883-948). Random House.
Tausczik, Y. R., & Pennebaker, J. W. (2010). The psychological meaning of words:
LIWC and computerized text analysis methods. Journal of Language and Social
Psychology, 29(1), 24–54. https://doi.org/10.1177/0261927X09351676
Wiggins, J. S. (1991). Agency and communion as conceptual coordinates for the understanding
and measurement of interpersonal behavior. In D. Cicchetti & W. M. Grove (Eds.),
Thinking clearly about psychology: Essays in honor of Paul E. Meehl, Vol. 1. Matters of
public interest; Vol. 2. Personality and psychopathology (pp. 89–113). University of
Minnesota Press.
Winquist, L. A., Mohr, C. D., & Kenny, D. A. (1998). The female positivity effect in the
perception of others. Journal of Research in Personality, 32(3), 370–388.
Wojciszke, B., & Abele, A. E. (2008). The primacy of communion over agency and its reversals
in evaluations. European Journal of Social Psychology, 38(7), 1139–1147.
33
Table 1. List of the LIWC dictionaries used in analyses, including a description or the most
frequently used examples provided by Pennebaker et al. (2015).
Dimensions
Description/Most frequently used exemplars
I
I, me, my, myself
She/he
he, she, her, his
They
they, their, them, themsel*
We
we, our, us, lets
Tone (Overall)
Degree of positive (negative) tone
Tone (Positive)
good, well, new, love
Tone (Negative)
bad, wrong, too much, hate
Doubt*
Language that questions the applicant
Communion**
Other-focused language (e.g., helping others/maintaining relationships)
* The doubt dictionary was developed by Madera et al. (2019)
** The communion dictionary was developed by Madera et al. (2009).
34
Table 2. Descriptive statistics and correlations for all study variables.
Note. * p < .05; ** p < .01; *** p < .001.
M
SD
1
2
3
4
5
6
7
8
9
12
1. I
0.00
0.77
–
2. She/He
0.01
0.95
.10***
–
3. They
0.00
0.22
-.06***
-.06
–
4. We
0.00
0.21
.12***
.01
.05***
–
5. Tone
-0.03
12.13
.11***
.31***
-.03**
.06***
–
6. Tone (Pos)
0.00
0.77
.08***
.28***
-.00
.05***
.94***
–
7. Tone (Neg)
0.00
0.25
-.06***
-.14***
.10***
.02*
-.39***
-.09***
–
8. Doubt
0.00
0.24
.13***
.01
.10***
.07***
-.06***
-.03**
.10***
–
9. Communion
0.00
0.52
.07***
.12***
.02*
.25***
.20***
.21***
-.06***
-.02
–
12. Women
0.27
0.44
-.09***
.04***
.01
-.01
.01
.01
-.00
-.05***
.04***
–
13. Ranking
75.81
72.75
.01
.04***
-.03**
.01
.00
-.02
-.04***
-.05***
.03**
.05***
35
Table 3. Within-candidate analyses of the relationship between letter writer gender and letter
content (Level-One).
Predictor
Estimate
S.E.
p-value
Additional analyses of language
Communion
0.05
0.01
<.001
Analyses of the use of personal, other, and collective pronouns.
I
-0.16
0.02
<.001
She/he
0.11
0.02
<.001
They
0.003
0.01
.56
She/he & They
0.12
0.02
<.001
We
-0.005
0.01
.38
Analyses of tone
Tone (Overall)
0.64
0.30
.03
Tone (Positive Only)
0.03
0.02
.16
Tone (Negative Only)
-0.01
0.01
.35
Analyses of doubt language
Doubt
-0.03
0.01
<.001
Note. Estimates in bold are significant at p < 0.05. Gender was coded as men= 0, women = 1.
All analyses controlled for letter writer discipline, letter writer rank, and US news ranking of
their home institution. All outcome variables were group-mean centered.
36
Table 4. Between-candidate analyses of the relationship between letter writer gender and voting
outcomes (Level-Two). Analyses for both “negative vote percentage” (Neg. Vote %) variables
used OLS regression. Analyses for the provost vote (binary variable) used logistic regression.
Depart. Neg. Vote %
College Neg. Vote %
Provost Vote
Variable
b
SE
b
SE
b
SE
% Women Letter Writer
-0.03
0.03
-0.07*
0.03
2.10*
0.94
Candidate Gender
-0.001
0.01
-0.01
0.02
-0.09
0.38
Candidate URM Status
0.004
0.02
0.03
0.02
-0.13
0.57
Candidate H Index
-0.001
0.001
-0.001
0.04
0.09*
0.04
External grants
-0.001
0.001
-0.001
0.30
0.02
0.03
Total Citations
0.000
0.000
0.000
0.000
-0.0001
0.0001
R2
.05
.08
.14
F
1.20
1.95
N
1,077
1,137
762
Note. Coefficients are unstandardized. External grants refers to the number of external grants
awarded as principal investigator. Institution, CIP Code (discipline), and rank being promoted to
were also used as controls but not presented in the table due to the large number of parameters. *
p < .05; ** p < .01; *** p < .001.
