http://www.jstor.org
Investigating the "Glass Ceiling" Phenomenon: An Empirical Study of Actual Promotions to
Top Management
Author(s): Gary N. Powell and D. Anthony Butterfield
Source: The Academy of Management Journal, Vol. 37, No. 1, (Feb., 1994), pp. 68-86
Published by: Academy of Management
Stable URL: http://www.jstor.org/stable/256770
Accessed: 26/04/2008 11:49
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=aom.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We enable the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org.
?
Academy of Management Journal
1994, Vol. 37, No. 1, 68-86.
INVESTIGATING THE "GLASS CEILING"
PHENOMENON: AN EMPIRICAL STUDY OF
ACTUAL PROMOTIONS TO TOP MANAGEMENT
GARY N. POWELL
University of Connecticut
D. ANTHONY BUTTERFIELD
University of Massachusetts
Although a "glass ceiling" is said to keep women from the top management
levels of organizations, no research has investigated actual decisions
about promotions to such positions. This study examined promotion
decisions for U.S. federal government Senior Executive Service
positions in a cabinet-level department. Contrary to hypotheses, the
job-irrelevant variable of gender worked to women's advantage, both
directly and indirectly, through job-relevant variables. However, an
applicant's employment in the hiring department had the greatest effect
on promotion decisions.
Women have made considerable progress in entering the managerial
ranks of U.S. enterprises in recent years, but not at the highest levels. The
proportion of women managers increased from 16 percent in 1970 to a 1992
level of 42 percent (U.S. Department of Labor, 1992). However, the proportion
of women who hold top management positions increased only slightly
during the last decade, from less than 3 percent in 1979 to less than 5 percent
in 1991 (Fisher, 1992; Korn/Ferry International, 1990). All but one of the
chief executive officers of the Business Week Top 1,000 corporations for
1992 were men (Segal, 1992). The phenomenon that keeps women from
reaching the top levels of organizations has been labeled the "glass ceiling"
(Morrison, White, Van Velsor, and the Center for Creative Leadership, 1987).
Morrison and colleagues described the glass ceiling as a "transparent
barrier that [keeps] women from rising above a certain level in corporations"
(1987: 13). They considered it a barrier for women as a group, barring individuals'
advancement simply because they are women rather than because
they lack the ability to handle jobs at higher levels. These authors placed the
glass ceiling at just below the general manager level. The U.S. Department of
This research was supported by faculty research grants from the University of Connecticut
and the University of Massachusetts. We express our deep appreciation to the federal department
studied for granting us access to its promotion records and to its people, whose cooperation
and hard work, especially in copying promotion files, made this study possible. We also
thank our research assistants, Laurel Goulet and Dale Finn, for their help.
68
Powell and Butterfield
Labor defined the glass ceiling as "those artificial barriers based on attitudinal
or organizational bias that prevent qualified individuals from advancing
upward in their organization" (1991: 1). Although the glass ceiling could
exist at different levels in different organizations or industries, the term is
typically used to suggest a barrier to entry into top-level management positions.
No field studies to date have investigated whether actual promotion
decisions for top management positions reflect the glass ceiling phenomenon.
The present study was designed to fill this void.
Research on women in management began in earnest in the mid-1970s
(e.g., Rosen &Jerdee, 1974; Schein, 1973), about the time that women were
becoming managers in significant numbers. Although a great volume of research
has accumulated since then (cf. Freedman & Phillips, 1988; Powell,
1993), research that focuses on women's movement into top management
positions has been scant. Much of what has been published on why women
"still don't hit the top" (Fierman, 1990) has been based on anecdotal data
from individual managers (e.g., Freeman, 1990) or on assessments of influences
on women's career and advancement experiences in general (e.g.,
Larwood & Gattiker, 1987; Solomon, Bishop, & Bresser, 1986; Stewart &
Gudykunst, 1982). Virtually all the empirical research on women's advancement
up the corporate ladder has focused on lower or middle management
positions (e.g., Eberts & Stone, 1985; Rosen &Jerdee, 1974; Shenhav, 1992;
Stroh, Brett, &Reilly, 1992). To date, there has been more speculation than
hard evidence offered as to what happens when women are considered for
top management positions.
Moreover, there has been very little research of any type, not just genderbased,
on how actual promotion decisions are made for top management
positions. Stumpf and London (1981b) identified criteria that are commonly
used when decisions are made about management promotions. They specified
both job-irrelevant criteria, such as gender, race, and appearance, and
job-relevant criteria, such as related work experience, being a current member
of the organization offering the position, past performance, education,
and seniority. However, they observed that "the relative importance of these
factors in a particular promotion decision or across decisions has received
little attention" (1981b: 543). This observation still holds true. Most studies
of decisions about management positions have focused on selection rather
than promotion decisions (e.g., Barr &Hitt, 1986; Hitt &Barr, 1989) or have
examined criteria for decisions made about hypothetical rather than real
applicants (e.g., London & Stumpf, 1983; Stumpf & London, 1981a). For
example, Hitt and Barr (1989) found that both job-irrelevant and job-relevant
variables influenced ratings of hypothetical applicants for middle and top
management positions.
HYPOTHESES AND RESEARCH QUESTION
Despite the paucity of research on promotions to top management positions,
with or without gender as a variable, authors have proposed various
1994 69
Academy of ManagementJournal
reasons for the lack of women in such positions. Two general explanations
have been offered. One is that the job-irrelevant variable of applicant gender
has a direct effect on decisions about promotions to top management positions,
to women's disadvantage. The other is that the job-irrelevant variable
of applicant gender has an indirect effect on promotion decisions for top
management positions through its relationship with job-relevant variables,
also to women's disadvantage. These two explanations do not compete with
each other; applicant gender may have both direct and indirect effects on
promotion decisions. In addition, the gender of decision makers may have a
moderating effect on the relationship between applicant gender and promotion
decisions for top management positions.
Direct Effect of Applicant Gender
Numerous theories have proposed the existence of outright discrimination
against women in organizations because of their gender. For example,
theories that focus on patriarchy suggest that men's desire to keep women in
a dependent status contributes to limitations on the latter's employment
opportunities, including access to top management positions (Strober,
1984). Kanter's (1977) theory of sex discrimination focused on the power of
numbers and the desire for social certainty. Promotion decisions for top
management positions involve subjective appraisals as to whether a given
candidate will fit in with incumbent top managers. Individuals who are
dissimilar to the incumbents in a particular type of job or to those making
promotion decisions (or to both), as female candidates for a top management
position in a male-dominated organization would be, are at a disadvantage
compared to individuals who are similar to incumbents and decision makers
(male candidates). Further, rational bias theory suggests that gender discrimination
results from intentional bias in decisions by managers, acting in their
own self-interest, who see their organizations as not interested in eliminating
gender discrimination or alleviating its effects (Larwood, Szwajkowski, &
Rose, 1988). Such managers may find a personal advantage in gender discrimination
even if they themselves do not care for it and consider the
women being discriminated against just as capable as the men being favored.
Although such theories disagree on the underlying rationale, they agree that
women are discriminated against when decisions are made for promotions
to top management positions.
However, gender discrimination need not be intentional. Motowidlo
(1986) suggested that unconscious biases resulting from fear of selecting an
unacceptable candidate, personal attraction to candidates similar to oneself,
and holding a stereotype of the ideal candidate for a position may distort
decision makers' judgments and lead job-irrelevant variables such as gender
to influence their selection decisions. Nieva and Gutek (1980) documented
the existence of pro-male bias in laboratory studies involving evaluation of
applicants for managerial positions.
70 February
Powell and Butterfield
Hypothesis 1: The job-irrelevant variable of applicant
gender directly influences promotion decisions for top
management positions in such a way that women receive
less favorable decision outcomes than men.
Indirect Effect of Applicant Gender
The second explanation for the low proportion of women in top management
positions is that applicant gender has an indirect effect on promotion
decisions through its relationship with job-relevant variables. Jobrelevant
criteria, such as experience, education, seniority, past performance,
and being a current member of the promoting organization, are to be expected
to influence promotion decision outcomes (Stumpf & London,
1981b). Women could fare poorly when assessed according to such criteria
for various reasons. Gender discrimination may have directly influenced
past personnel decisions about them, making it more difficult for them than
for men to obtain the necessary preparation for top management positions.
Also, there could be gender differences in the distribution of job market
skills like education and work experience. Human capital theory (Becker,
1971) suggests that workers make rational choices regarding investment in
their own education and careers. If women have built up less human capital
over time than men, then their career progression, including ascendancy to
top management positions, would be more restricted (Stroh et al., 1992).
Finally, gender discrimination may have influenced prior evaluations of
employees' human capital, with women given less credit for their education,
work experience, and so on than men. Thus, women who come up short on
job-relevant criteria for top management promotions, whatever the reason
may be, are likely to be at a disadvantage, even if they are not directly
discriminated against when such decisions are made.
Hypothesis 2: The job-irrelevant variable of applicant
gender indirectly influences promotion decisions for top
management positions in such a way that women are
rated less favorably than men on job-relevant variables
that influence decision outcomes.
The Moderating Effect of Decision Makers' Gender
As noted earlier, Kanter (1977) concluded that the desire for social certainty
leads decision makers to prefer to work with individuals like themselves.
The "similar-to-me" effect (Rand &Wexley, 1975) may be based on
perceived similarity (e.g., Dalessio &Imada, 1984; Pulakos &Wexley, 1983),
on actual similarity, or on both. Pfeffer (1983) argued that the demographic
composition of organizations influences many behavior-based events, including
managerial successions. Research has found that similarity in demographic
characteristics influences outcomes such as superiors' ratings of
subordinates' performance (Tsui & O'Reilly, 1989), turnover within work
teams (Jackson, Brett, Sessa, Cooper, Julin, & Peyronnin, 1991; O'Reilly,
Caldwell, & Barnett, 1989), and employees' organizational attachment
1994 71
Academy of Management Journal
(Tsui, Egan, & O'Reilly, 1992). However, no empirical research has investigated
whether the gender of decision makers interacts with applicant gender
to influence promotions to top management positions. Thus, we also explored
the question, Does the gender of decision makers moderate the relationship
between applicant gender and promotion decisions for top management
positions?
METHODS
Population
Empirical research on promotions to top management is scarce in part
because organizations are reluctant to reveal their practices in such a sensitive
area. However, as a result of the Civil Service Reform Act of 1978, the
United States government has a systematic and accessible procedure for
making top management promotions. Thus, the federal government seemed
a good place to begin field research on actual promotion decisions to top
management.
A large, cabinet-level federal department was approached as a possible
research site in 1989. We received permission to obtain all available data on
decisions for career appointments to the Senior Executive Service (SES),
which consists of all top management positions except those reserved for
political appointees. Approximately 1 percent of all positions that government
employees can attain without political sponsorship are SES positions.
Promotions to the SES are made systematically and are based on the same
types of information across positions. Records of the decision process are
kept for at least two years. Data were obtained for decisions made from
January 1987 to February 1992.
In 1990, the federal department studied had 150 career SES positions, of
which 29 (19%) were filled by women. Federal employees who are not in
SES positions have federal civil service grade levels of up to 15. Grades 13
to 15 are considered the pipeline to the Senior Executive Service. Approximately
4,700 department employees held grade 13, 14, or 15 positions. The
proportion of women was 25 percent in grade 15, 26 percent in grade 14, and
27 percent in grade 13 positions in the department. For the federal government
as a whole, the proportion of women was 10 percent in SES positions,
11 percent in grade 15, 15 percent in grade 14, and 20 percent in grade 13
positions.
Procedure for Filling SES Positions
When a Senior Executive Service position becomes open, a position
announcement is circulated that specifies the criteria by which applicants
will be judged. Six of the criteria are the same for every SES position-an
example is "ability to integrate internal and external program/policy issues"-and
the other criteria are position-specific. Interested individuals
submit formal applications that provide background and career history data.
Applicant's current supervisors are asked to rate how well the applicants
February72
Powell and Butterfield
would meet the specified criteria for the applied-for position if they were
selected. The applicants' most recent performance appraisals in their current
positions are also obtained if available. The personnel office in which the
vacancy is located then screens out applicants who are considered obviously
unqualified because they did not meet minimum eligibility criteria for the
position.
The person who makes the final decision on a given position, known as
the selecting official, is typically the future manager of the person to be
selected for the position. The selecting official asks a panel of one to three
senior individuals who are familiar with the demands of the position to
review the credentials of the remaining applicants. The review panel rates
each applicant on each of the specified criteria as high (3), medium (2), or
low (1) in qualifications and arrives at an overall rating of the applicant as
highly qualified or qualified. The panel also decides which applicants to
refer to the selecting official for final consideration. In most cases, applicants
rated as highly qualified are referred and those rated as qualified are not
referred. Once the selecting official has a list of referred applicants from the
review panel, he or she selects the applicant who will receive the job.
Thus, once applicants pass the initial screen and are evaluated by the
review panel, up to two decisions are made about them: (1) a review panel
decides whether to refer them to the selecting official and (2) the selecting
official decides whether to choose them for the position.
Data
From January 1987 to February 1992, 32 open SES positions were filled
within the department. Review panels regarded 438 applicants as qualified
or highly qualified for the positions they applied for. We obtained review
panel and application data for these 438 applicants. The study was restricted
to applicants who were considered to be legitimate contenders for the positions,
and thus we did not obtain data for applications whom personnel
offices had screened out as not meeting minimum eligibility criteria.
Eighty-eight percent of the 438 applicants were men. Their mean fulltime
work experience was 22 years, and their mean age was 47 years. For 67
percent of the applicants, the highest degree obtained was a graduate degree;
it was a bachelor's degree for 30 percent and less than a bachelor's degree for
3 percent. Virtually all the applicants had some federal work experience at
some time; those who had no federal experience were not included in the
data analysis. The highest grade was 13 or less for 4 percent, grade 14 for 16
percent, grade 15 for 65 percent, and SES for 15 percent of the applicants.
The mean time at the highest grade attained was six years. Forty-three percent
were currently employed by the hiring department.
Review panels referred 258 of the 438 applicants (59%) to selecting
officials. Of the 258 referred applicants, 32 (12%) were selected to fill the 32
open positions.
Sixty-two percent of the applicants were evaluated by an all-male review
panel, 36 percent by a mixed-gender panel, and 2 percent by an all-
1994 73
Academy of Management Journal
female panel. For men serving on review panels, the mean age was 49 years,
and for women it was 48 years. The selecting officials were men with a mean
age of 50 years for 84 percent of the referred applicants and women with a
mean age of 60 years for the other 16 percent of the referred applicants.
Measures
Three variables served as measures of outcomes of promotion decisions.
A review panel's evaluation of an applicant's qualifications for a position
was measured by the average of its ratings of the applicant on the specified
criteria for the position. Whether the applicant was referred for the position
by the review panel (0 = no, 1 = yes) and, if referred, selected for the
position by the selecting official (0 = no, 1 = yes) were the other two
outcome measures.
The job-irrelevant variable of applicant gender (effects coding: -1 =
men, 1 = women) and six job-relevant variables were included in analyses
as potential predictors of the outcome measures. The job-relevant variables
were whether the applicant for a given job was currently employed in the
hiring department (0 = no, 1 = yes), the highest grade the applicant had
held in the federal government (1 = grade 13 or less, 2 = grade 14, 3 = grade
15, 4 = SES), the number of years the individual had been at the highest
grade, the number of years of full-time work experience, the highest degree
obtained (1 = less than bachelor's degree, 2 = bachelor's degree, 3 = graduate
degree), and the summary performance appraisal rating (five-point
scale, with 5 the highest rating) on the applicant's most recent performance
appraisal.1 Stumpf and London (1981b) described all these job-relevant variables
as criteria commonly used by organizations for management promo-
tions.
The gender compositions of the review panels (-1 = all men, 1 =
mixed gender)2 and the gender of the selecting official (-1 = man, 1 =
woman) were also included as measures.
1 Although performance appraisal information was missing for 28 percent of the applicants,
we included it as a potential predictor because Stumpf and London (1981b) specifically described
past performance as a criterion for management promotions. Other job-relevant and
job-irrelevant variables were considered as potential predictors but not included in the final
analyses for various reasons. The current supervisors' average ratings of applicants on the
specified criteria for the applied-for positions were not included because such ratings were
missing for 59 percent of the applicants. The review panels' overall ratings of applicants as
highly qualified or qualified were not included because they were equivalent to the referral
decision outcomes for 84 percent of the applicants. (The other 16 percent were rated as qualified
but were referred; no applicants were rated as highly qualified but not referred.) Applicant age
was not included because it was highly correlated with years of full-time work experience (r =
.75). Finally, we were unable to include the racial-ethnic groups of applicants as a predictor
because it was not in the applicant files as a matter of law. Applicant gender was included on
the application forms "for statistical use."
2
The all-woman panels were omitted since they reviewed only 2 percent of applicants.
February74
Powell and Butterfield
Analysis
Two outcome measures were used in analyses for all 438 applicants: the
review panel's evaluation of each applicant on the specified criteria for the
position and the review panel's decision about whether to refer the applicant
to the selecting official. We used ordinary-least-squares regression analysis
to determine the influence of the predictor variables on panel evaluations for
the applicants. Logistic regression analysis (Norusis, 1990) was used to assess
the influence of the same predictor variables on the dichotomous referral
decisions. We calculated pseudo R2's for the logistic regression model
following the formula recommended by Aldrich and Nelson (1984).
The third outcome measure, the selecting officials' hiring decisions, was
applicable only for the 258 applicants who were referred by the review
panels. We used logistic regression analysis to assess the influence of predictor
variables on the dichotomous hiring decisions for these applicants.
To determine the unique contribution of each predictor variable to each
outcome measure, we conducted a usefulness analysis (Darlington, 1968),
calculating R2 or pseudo R2for the full model and then for a reduced model
that did not contain the variable of interest. The difference in variance accounted
for between the full and reduced models is the unique variance
(usefulness) the removed variable accounts for (Pedhazur, 1982).
Moderated regression analyses were performed to examine the moderating
influence of the gender of decision makers on the relationship between
applicant gender and outcomes. We first regressed panel evaluations and
referral decisions on applicant gender, the six job-relevant predictor measures,
and the gender composition of a review panel. Next, we computed a
model including the interaction of applicant's gender and the gender composition
of the review panel in addition to all other variables and examined
whether the interaction term made a unique contribution to the regression
equation. The same moderated regression procedure was used for selection
decisions, except that the gender of the selecting official was used instead of
the gender composition of the review panel.
RESULTS
Table 1 reports correlations, means, and standard deviations for the
seven predictor and three outcome variables for all 438 applicants studied.
The highest correlation among the predictor variables was .27. Several correlations
were significant only because of the large number of applicants.
Table 2 presents gender differences in job-relevant and outcome measures
for the complete group of 438 applicants and the subgroup of 258
referred applicants. When the applicant group was considered as a whole,
women received significantly higher performance appraisal ratings and
panel evaluations and were significantly more likely to be referred than men.
They were also significantly more likely to be employed in the hiring department
and had significantly less work experience. When only referred
applicants were considered, there were no significant gender differences in
1994 75
TABLE 1
Descriptive Statistics and Correlationsa
Variables Mean s.d. 1 2 3 4 5
1. Gender -0.77 0.64
2. Employed in hiring
department 0.43 0.50 .12
3. Highest grade 2.90 0.69 .04 -.13
4. Years at highest grade 6.07 4.42 -.06 .18 -.08
5. Years of full-time
work experience 22.04 7.31 -.10 .10 .16 .27
6. Highest degree 2.64 0.53 -.02 .00 .04 -.04 -.12
7. Performance appraisal
rating 4.27 0.67 .20 -.17 -.08 -.05 -.17
8. Panel evaluation 2.64 0.42 .15 .37 .14 .21 -.05
9. Referral decision 0.59 0.49 .19 .40 .22 .16 .04
10. Selection decision 0.07 0.26 .14 .25 .04 -.07 -.08
a
Correlations with an absolute value of .09 or higher are significant at the .05 level.
TABLE 2
Gender Differencesa
I s.n. . . ..::v.1.: :: .-.. . ::::I : .. . x L .- X::- qqpir_ -,I J
All Applicants
Variables Men Women t df Men
Percentage employed in hiring department 40.67 58.82
Highest grade 2.89 2.98
Years at highest grade 6.17 5.35
Years of full-time work experience 22.31 19.98
Highest degree 2.65 2.62
Performance appraisal rating 4.22 4.62
Panel evaluation 2.62 2.81
Percentage referred 55.56 84.31
Percentage selected
N 387 51
- 2.47* 435 57.67
-0.82 394 3.01
1.23 396 6.77
2.15* 434 22.64
0.35 429 2.68
-3.67*** 313 4.24
- 3.02** 389 2.80
-3.99*** 436
10.70
215
a
We performed t-tests for all applicants (N = 438) and for referred applicants (N =
258). Degrees of fr
missing data.
*
p < .05
** p < .01
***
p < .001
Academy of Management Journal
predictor or outcome measures except on the performance appraisal rating
variable, where women scored significantly higher than men.
Panel evaluations for all applicants were regressed on the six jobrelevant
measures and applicant gender using ordinary-least-squares regression
(Table 3). Applicant gender had a significant effect on panel evaluations.
However, the direction of the effect, favoring women, was opposite to
that predicted by Hypothesis 1. Thus, findings did not support Hypothesis 1
for the panel evaluation outcome.
Four of the six job-relevant measures-employment in the hiring department,
highest grade, years at highest grade, and years of full-time work
experience (a negative effect)-significantly influenced panel evaluations.
As Table 2 reports, women differed from men on two of the four significant
job-relevant predictors of panel evaluations, employment in the hiring department
and work experience. However, the indirect effect of gender on
panel evaluations through these predictors favored women rather than men,
contrary to Hypothesis 2. Women scored significantly higher than men on
employment in the hiring department, which was positively related to panel
evaluations, and significantly lower than men in work experience, which
was negatively related to panel evaluations. Thus, Hypothesis 2 was not
supported for the panel evaluation outcome. In addition, although the effect
of applicant gender was significant, the unique variance it accounted for
(1%)was small relative to that of employment in the hiring department (8%).
Table 4 presents the results of the logistic regression analysis for the
referral decision for all applicants. The logistic regression procedure in
SPSS, Version 4.1 (Norusis, 1990) was employed. Applicant gender had a
significant effect on referral decisions. However, as for panel evaluations,
the direction of the effect, favoring women, was opposite to that predicted in
Hypothesis 1. Thus, findings did not support Hypothesis 1 for the referral
decision outcome.
TABLE 3
Results of Regression Analysis for Panel Evaluations for All Applicantsa
Variables P Unique R2
Gender .13* .01*
Employed in hiring department .30 ** .08 **
Highest grade .15* .02*
Years at highest grade .22 *** .04 **
Years of full-time work experience -.17** .02**
Highest degree .02 .00
Performance appraisal rating .09 .01
F7,265 7.66***
R2 .17
Adjusted R2 .15
aN is 273 because of missing data.
p < .05
*p < .01
** p < .001
February78
Powell and Butterfield
TABLE 4
Results of Regression Analysis for Referral Decisions for All Applicantsa
Variables b s.e. Unique Pseudo R2
Gender .80** 0.30 .02 *
Employed in hiring department 2.54*** 0.37 .16**
Highest grade 1.25 ** 0.32 .04 **
Years at highest grade .12** 0.04 .03**
Years of full-time work experience -.08** 0.02 .03**
Highest degree .52 0.32 .01
Performance appraisal rating .44 0.25 .01
Constant - 5.96** 2.00
Chi-square 94.01***
df 7
Pseudo R2 .26
aN is 273 because of missing data.
p < .05
* p < .01
**p < .001
Four of the six job-relevant measures-employment in the hiring department,
highest grade, years at highest grade, and years of full-time work
experience (a negative effect) -significantly influenced referral decisions.
These effects were similar to those obtained for predicting panel evaluations.
Thus, results did not support Hypothesis 2 for the referral decision outcome.
In addition, as for panel evaluations, the unique variance accounted for by
applicant gender (2%) was small relative to what employment in the hiring
department accounted for (16%).
Table 5 presents the results of the logistic regression analysis for the
selection decision, conducted only for applicants whom the review panels
referred to the selecting officials. Since the review panels transmitted evaluations
of applicants as well as referral decisions, panel evaluations were
included as a predictor variable.
Applicant gender in itself did not have a significant effect on selection
decisions; thus, Hypothesis 1 was not supported. Three of the six jobrelevant
measures-employment in the hiring department, years at highest
grade (a negative effect), and performance appraisal rating-significantly
influenced selection decisions. As Table 2 shows, the first two variables did
not favor referred applicants of either gender, whereas women received significantly
higher performance appraisal ratings than men. Since the indirect
effect of gender on selection decisions through performance appraisal ratings
favored women rather than men, findings did not support Hypothesis 2.
Moderated regression analyses revealed that the interaction between
applicant gender and the gender composition of the review panels did not
make a unique contribution to the regression model for either panel evaluations
(AR2= .00) or referral decisions (AR2= .01); the relationship between
applicant gender and each outcome was not significantly different for all-
1994 79
Academy of Management Journal
TABLE 5
Results of Regression Analysis for Selection Decisions for
Referred Applicantsa
Variables b s.e. Unique Pseudo R2
Gender .06 0.26 .00
Employed in hiring department 1.81* 0.72 .04*
Highest grade .51 0.56 .00
Years at highest grade -.13* 0.06 .02*
Years of full-time work experience .00 0.04 .00
Highest degree .01 0.44 .00
Performance appraisal rating .80* 0.40 .02*
Panel evaluation .58 1.30 .00
Constant -9.04* 4.77
Chi-square 19.10*
df 8
Pseudo R2 .10
aN is 175 because of missing data.
* p < .05
male and mixed-gender review panels. Also, the interaction between applicant
gender and the gender of the selecting official did not make a unique
contribution to the regression model for selection decisions for referred applicants
(AR2 = .01); the relationship between applicant gender and selection
decisions was not significantly different for male and female selecting
officials. These results suggested that the gender of decision makers did not
moderate the relationship between applicant gender and promotion decision
outcomes. Complete results of these analyses are available from the senior
author.
DISCUSSION
The hypotheses of the study, which propose that applicant gender will
directly and indirectly influence promotion decisions for top management
positions to the disadvantage of women, were rejected. Although the effect
sizes were small, applicant gender directly influenced the two earlier promotion
decision outcomes, panel evaluations and referral decisions, to women's
advantage and did not directly influence final selection decisions.
Applicant gender indirectly influenced panel evaluations and referral decisions
through its relationship with two job-relevant variables, employment
in the hiring department and work experience, with each indirect effect
favoring women. Applicant gender indirectly influenced selection decisions
through its relationship with one job-relevant variable, the performance appraisal
rating, with the indirect effect again favoring women. The gender of
decision makers did not moderate the relationship between applicant gender
and promotion decision outcomes.
80 February
Powell and Butterfield
Direct and Indirect Effects of Applicant Gender
Why were all the significant direct and indirect effects of gender in this
study favorable to women? To address this question, we need to consider the
unique nature of the organization studied. The very fact that the federal
government is unlike most organizations may have accounted for the absence
of decision making that favored men. The federal government places
a high degree of emphasis on procedural fairness in making promotion decisions
for SES positions. First, it requires that all open positions be made
known through a public announcement. Second, it requires that all promotion
decisions be made using the same basic procedure. Third, it requires
that records be kept of the entire decision-making process. By providing
structure to the decision-making process and enabling identification of decisions
that are not properly made, these requirements make decision makers
accountable for how their promotion decisions are made. Stumpf and
London (1981b) suggested increasing accountability and creating standardized
procedures for making decisions as interventions to remedy biased
decision-making processes.
In addition, the federal government is particularly concerned with issues
regarding equal employment opportunity. This concern is evident in
the Department of Labor's (1991) "glass ceiling initiative" for private corporations
as well as in established federal policies and practices. It may lead to
women in the federal government benefiting more from promotion decisions
than men rather than being victimized by the glass ceiling phenomenon.
First, women may receive more favorable performance evaluations and referral
decisions than men simply because of their gender. Second, they may
be in a better position to be promoted to top management jobs than women
in other types of organizations because they have been treated favorably, or
at least not been discriminated against, in past decisions regarding open
positions at lower levels.
Finally, women may have benefited from a special commitment to equal
opportunity in the particular federal department examined. The department's
proportions of women in the pipeline grades of 13 to 15 (25-27%)
were above average for the federal government, with less of a drop-off in the
proportion of women from grade 13 to grade 15, suggesting that the department
may have been particularly committed to hiring and promoting women
throughout its ranks. However, the proportion of women applicants in the
study (12 percent overall, 16 percent for department employees) was smaller
than the proportion of women in pipeline grades in the department, further
suggesting that the women who applied may have been stronger applicants
than the men who applied. This unevenness could account for women's
higher performance appraisal ratings, which indirectly led to their receiving
more favorable selection decisions. Perhaps only women who were extremely
high performers applied for open SES positions; perceptions of a
glass ceiling, whether warranted or not, may have inhibited women with
lesser performance records from applying.
1994 81
Academy of Management Journal
These results suggest that when decision making is open and a systematic
procedure is used, decisions that foster the glass ceiling phenomenon
may be averted. When procedures for promotion decisions are standardized
and criteria for decisions are well established, qualified women may fare at
least as well as qualified men. When procedures are not standardized, or
when criteria for promotion decisions are unspecified or vague, there may be
more occasion for gender-related biases favoring men to affect the outcomes
of the promotion process. To what extent are the decision-making conditions
found in this study present in the private sector and other public agencies?
Additional research in differing organizational contexts is necessary to answer
this question.
No matter how systematic decision making is, however, perceptions of
gender-related biases in decision making are likely to influence which individuals
actually apply for top management positions. In particular,
women who believe that the glass ceiling phenomenon will operate to their
disadvantage may be less likely to apply for open positions than equally
qualified men. Such behavior on their part may lead to the average female
applicant for top management positions having stronger credentials and faring
better in promotion decisions than the average male applicant, as appeared
to be the case in the present study. However, a smaller number of
female applicants may also lead to a smaller number of promotions of
women to top management positions. Thus, to truly avert the glass ceiling
phenomenon, organizations need to address perceptions about how promotion
decisions are made as well as the reality.
Effects of Job-Relevant Variables
The study confirmed the importance of several of the job-relevant variables
previously described as common criteria for management promotions.
It also enabled a rough assessment of the relative importance of the variables
examined, which had been lacking (Stumpf &London, 1981b). Being a current
member of the organization filling the open position explained the
greatest amount of unique variance in each decision outcome; it was also the
only variable that had a significant, positive effect on all three outcomes.
Years at the highest grade had a positive effect on the early judgments assessed
(panel evaluations and referral decisions), but a negative effect on the
final selection decision. Two other job-relevant variables, highest grade
achieved and work experience, had a significant effect on the early judgments
regarding applicants (positive for highest grade and negative for work
experience) but not on the selection decision. Performance appraisal ratings
had a significant effect only on the selection decision. Education was the
only variable that did not have a significant effect on any outcome measure,
perhaps because there was little variance in education in this group; twothirds
of the applicants had advanced degrees and only 3 percent lacked
four-year college degrees.
Being a current member of the hiring organization could have had a
particularly strong effect on promotion decisions for several reasons. Ac-
February82
Powell and Butterfield
cording to Kanter's (1977) reasoning, decision makers may see insiders as
more similar to themselves than outsiders, and thus more preferable as candidates.
A similar-to-me effect has been found in previous research on ratings
of others (e.g., Rand & Wexley, 1975; Tsui & O'Reilly, 1989). Perhaps
such an effect extends to promotion decisions regarding top management
positions and is based on qualities of applicants besides their gender. However,
there may also be legitimate reasons for choosing an applicant who has
first-hand experience with an organization over applicants who are otherwise
equally qualified. Familiarity with an organization may make the applicant
better prepared to assume the responsibilities of the position.
Indeed, familiarity with the selecting official may also help. Selecting
officials may know applicants from their own department better than outsiders
and feel more confident choosing applicants whose work is well
known to them and to the department. Also, selecting officials occasionally
interview applicants, and impressions gathered in these interviews could
affect their decisions. Further research that assesses the influence of decision
makers' first-hand knowledge of candidates is necessary to reveal more
about how such decisions are made. Such research is also necessary to
determine whether the strong insider effect is present in other types of organizations
and, if so, why it is so pronounced.
The significant, negative effects of job-relevant variables on outcomes
call for special consideration. First, years of work experience negatively
influenced review panels' evaluations of applicants and referral decisions.
Review panels may have been impressed by accomplishments and performance
achieved within a limited time frame. Applicants at the threshold of
a top management position with less experience may have been seen as
"fast-trackers" who offered greater potential for success precisely because
they had reached that threshold sooner than the average applicant. At the
other extreme, applicants with many years of experience who had still not
been promoted to the top rank may have been viewed as over the hill and not
deserving of further advancement. The women studied tended to have fewer
years of work experience than the men, so they may have benefited from
such perceptions.3
A similar explanation may apply for selection decisions, which were
not affected by years of work experience but were affected negatively by the
number of years an applicant had spent at the highest grade achieved. Selecting
officials may have wanted new blood and therefore favored candi-
3 Post hoc analysis revealed that the relationship between work experience and panel
evaluations was curvilinear, with the highest evaluations received by applicants with ten years
of experience and lower evaluations received as years of experience increased beyond ten years.
Because 95 percent of the applicants had more than ten years of work experience, the relationship
between work experience and panel evaluations was essentially negative for the range of
experience studied. There was no evidence of a curvilinear relationship between years of work
experience and referral decisions or between years at the highest grade achieved and selection
decisions.
1994 83
Academy of Management Journal
dates who had not been at their highest grade for very long. Applicants with
many years at their highest grade may have been seen as essentially "plateaued"
(Veiga, 1981) and not destined for advancement to the top rank.
Additional research is necessary to determine the conditions under which
time-related criteria, such as years at a highest grade and years of work
experience, have positive or negative effects on top management promo-
tions.
CONCLUSIONS
The present study was subject to the difficulties inherent in conducting
field research on management promotions (London &Stumpf, 1983). In particular,
we had a limited study group of female applicants and aggregated
data across promotion decisions with different task requirements, contexts,
and decision makers. However, various definitions of the glass ceiling phenomenon
suggest that factors that are common to different types of organizational
contexts keep women from reaching the top levels of organizations.
The only way to determine whether common factors affect the advancement
of women is to aggregate data on promotion decisions from different con-
texts.
What would it take for the glass ceiling phenomenon to disappear? This
study suggests some tentative answers. Organizations may be able to avert it
by revising procedures for making promotion decisions in ways that increase
the accountability of decision makers and impose uniformity on the promotion
process, and by making these procedures well known to all potential
applicants. In addition, organizations that pursue a strong promote-fromwithin
policy and fill positions immediately below the top level with highly
qualified women and men who progressed to that level without delay may
eventually find their glass ceilings shattered.
REFERENCES
Aldrich, J. H., & Nelson, F. D. 1984. Linear probability, logit, and probit models. Newbury
Park, CA: Sage.
Barr, S. H., & Hitt, M. A. 1986. A comparison of selection decision models in manager versus
student samples. Personnel Psychology, 39: 599-617.
Becker, G. S. 1971. The economics of discrimination (2d ed.). Chicago: University of Chicago
Press.
Dalessio, A., & Imada, A. S. 1984. Relationships between interview selection decisions and
perceptions of applicant similarity to an ideal employee and self: A field study. Human
Relations, 37: 67-80.
Darlington, R. B. 1968. Multiple regression in psychological research and practice. Journal of
Applied Psychology, 69: 161-182.
Eberts, R., & Stone, J. 1985. Male-female differences in promotions: EEO in public education.
Journal of Human Resources, 20: 504-521.
Fierman, J. 1990. Why women still don't hit the top. Fortune, July 30: 40-62.
Fisher, A. B. 1992. When will women get to the top? Fortune, September 21: 44-56.
84 February
Powell and Butterfield
Freedman, S. M., & Phillips, J. S. 1988. The changing nature of research on women at work.
Journal of Management, 14: 231-251.
Freeman, S. J. M. 1990. Managing lives: Corporate women and social change. Amherst: University
of Massachusetts Press.
Hitt, M. A., &Barr, S. H. 1989. Managerial selection decision models: Examination of configural
cue processing. Journal of Applied Psychology, 74: 53-61.
Jackson, S. E., Brett, J. F., Sessa, V. I., Cooper, D. M., Julin, J. A., & Peyronnin, K. 1991. Some
differences make a difference: Individual dissimilarity and group heterogeneity as correlates
of recruitment, promotions, and turnover. Journal of Applied Psychology, 76: 675-
689.
Kanter, R. M. 1977. Men and women of the corporation. New York: Basic.
Korn/Ferry International. 1990. Executive profile 1990: A survey of corporate leaders. New
York: Korn/Ferry International.
Larwood, L., &Gattiker, U. E. 1987. A comparison of the career paths used by successful women
and men. In B. A. Gutek & L. Larwood (Eds.), Women's career development: 129-156.
Newbury Park, CA: Sage.
Larwood, L., Szwajkowski, E., & Rose, S. 1988. When discrimination makes "sense": The rational
bias theory. In B. A. Gutek, A. H. Stromberg, &L. Larwood (Eds.), Women and work:
An annual review, vol. 3: 265-288. Newbury Park, CA: Sage.
London, M., &Stumpf, S. A. 1983. Effects of candidate characteristics on management promotion
decisions: An experimental study. Personal Psychology, 36: 241-259.
Morrison, A. M., White, R. P., Van Velsor, E., & the Center for Creative Leadership. 1987.
Breaking the glass ceiling: Can women reach the top of America's largest corporations?
Reading, MA: Addison-Wesley.
Motowidlo, S. J. 1986. Information processing in personnel decisions. In K. M. Rowland &G. R.
Ferris (Eds.), Research in personnel and human resources management, vol. 4: 1-44.
Greenwich, CT: JAI Press.
Nieva, V. F., &Gutek, B. A. 1980. Sex effects on evaluation. Academy of Management Review,
5: 267-276.
Norusis, M. J. 1990. SPSS advanced statistics student guide. Chicago: SPSS.
O'Reilly, C. A., III, Caldwell, D. F., & Barnett, W. P. 1989. Work group demography, social
integration, and turnover. Administrative Science Quarterly, 34: 21-37.
Pedhazur, E. J. 1982. Multiple regression in behavioral research (2d ed.). New York: Holt,
Rinehart &Winston.
Pfeffer, J. 1983. Organizational demography. In L. L. Cummings &B. M. Staw (Eds.), Research
in organizational behavior, vol. 5: 299-357. Greenwich, CT: JAI Press.
Powell, G. N. 1993. Women and men in management (2d ed.). Newbury Park, CA: Sage.
Pulakos, E. D., & Wexley, K. N. 1983. The relationship among perceptual similarity, sex, and
performance ratings in manager-subordinate dyads. Academy of Management Journal, 26:
129-139.
Rand, T. M., & Wexley, K. N. 1975. Demonstration of the effect "similar to me" in simulated
employment interviews. Psychological Reports, 36: 535-544.
Rosen, B., &Jerdee, T. H. 1974. Effects of applicant's sex and difficulty of job on evaluations of
candidates for managerial positions. Journal of Applied Psychology, 59: 511-512.
Schein, V. E. 1973. The relationship between sex role stereotypes and requisite management
characteristics. Journal of Applied Psychology, 57: 95-100.
Segal, A. T. 1992. Corporate women. Business Week, June 8: 74-78.
1994 85
Academy of Management Journal
Shenhav, Y. 1992. Entrance of blacks and women into managerial positions in scientific and
engineering occupations: A longitudinal analysis. Academy of Management Journal, 35:
889-901.
Solomon, E. E., Bishop, R. C., & Bresser, R. K. 1986. Organization moderators of gender differences
in career development: A facet classification. Journal of Vocational Behavior, 29:
27-41.
Stewart, L. P., & Gudykunst, W. B. 1982. Differential factors influencing the hierarchical level
and number of promotions of males and females within an organization. Academy of
Management Journal, 25: 586-597.
Strober, M. H. 1984. Toward a general theory of occupational sex segregation: The case of public
school teaching. In B. F. Reskin (Ed.), Sex segregation in the workplace: Trends, explanations,
remedies: 144-156. Washington, DC: National Academy Press.
Stroh, L. K., Brett, J. M., & Reilly, A. H. 1992. All the right stuff: A comparison of female and
male managers' career progression. Journal of Applied Psychology, 77: 251-260.
Stumpf, S. A., & London, M. 1981a. Capturing rater policies in evaluating candidates for promotion.
Academy of Management Journal, 24: 752-766.
Stumpf, S. A. & London, M. 1981b. Management promotions: Individual and organizational
factors influencing the decision process. Academy of Management Review, 6: 539-549.
Tsui, A. S., Egan, T. D., &O'Reilly, C. A., III. 1992. Being different: Relational demography and
organizational attachment. Administrative Science Quarterly, 37: 549-579.
Tsui, A. S., & O'Reilly, C. A., III. 1989. Beyond simple demographic effects: The importance of
relational demography in superior-subordinate dyads. Academy of Management Journal,
32: 402-423.
U.S. Department of Labor. 1991. A report on the glass ceiling initiative. Washington, DC: U.S.
Department of Labor.
U.S. Department of Labor. 1992. Employment and earnings, vol. 39 no. 5: Table A-22. Washington,
DC: U.S. Department of Labor.
Veiga, J. F. 1981. Plateaued versus nonplateaued managers: Career patterns, attitudes, and path
potential. Academy of Management Journal, 24: 566-578.
Gary N. Powell is a professor of management in the School of Business Administration
at the University of Connecticut. He received his Ph.D. degree in organizational behavior
from the University of Massachusetts, Amherst. His research interests include the
study of gender-related issues in organizations and the study of recruitment processes.
D. Anthony Butterfield is a professor in and the chairperson of the management department
in the School of Management at the University of Massachusetts, Amherst. He
received his Ph.D. degree in organizational psychology from the University of Michigan.
His research interests center on issues of leadership and gender. Currently he is
president-elect of the Eastern Academy of Management.
86 February