PART III

The Earnings Puzzle: Why Do Women Earn Less than Men?

This middle part of the book turns to the central question facing economists interested in gender
issues: why do women earn less than men? Here we explore the different causes of the gender
earnings gap. In particular, we want to abstract from the facts about labor force participation -
such as the facts that many women do not do any market work and that many women work fewer
hours than most men - and directly address the question of why women who do choose to work
in the market make lower hourly earnings on average than do men.

Chapters 6 through 9 address the question of what theories can account for gender earnings
differences, both within and between jobs. Chapter 6 explores the links between the jobs that
men and women hold and the pay that each group receives in those jobs. Differences in the
distribution of men and women across jobs are discussed in detail, and the gender-related
earnings patterns within and between jobs are discussed as well. The exclusion and crowding
models, which link gender pay differences to segregation practices, are discussed.

Chapters 7 through 9 take up, in turn, three groups of arguments for possible variance in pay
within and between jobs. Chapter 7 develops the human capital model, which implies that
differences in education, experience, and training can account for differences in pay between
persons. Chapter 8 examines the model of compensating differentials, which implies that
differences in preferences for aspects of work other than pay can account for differences in pay
between persons. Chapter 9 considers several models of discrimination, all of which attribute
differences in pay to differences in persons' treatment by employers. However, the
discrimination models differ in whether they attribute prejudice to employers, employees, or
customers, and in whether or not prejudice is assumed at all.

III-1
Part III concludes with a policy application that draws on material from all four chapters: the
study of the rationales for and possible effects of comparable worth policies.

III-2
6
Gender Segregation in the Workplace

Gender segregation in the labor force is a pervasive phenomenon with deep roots in the gender
division of labor in both modern and historical societies. Segregation occurs within and between
firms, occupations, and industries. This chapter characterizes the current and past nature of
gender segregation of workers, explores theories of why segregation occurs and persists,
examines the link between segregation and the gender earnings differential, and considers how
various policies might affect segregation patterns.

The situation in various occupations

As the proportion of women in the labor force has increased, women have entered both
traditionally female and traditionally male occupations. Let us discuss a variety of different
occupations to illustrate the range of different patterns found in terms of sex composition.
General trends for blue-collar, pink-collar, professional, and managerial occupations are
considered. The special cases of apparently stable, highly segregated occupations are considered
in terms of how the minority group within them fares. Then different occupations that have had a
sharp rise in percentage of females are considered.

Blue-collar occupations
Blue-collar occupations remain the most heavily male area of the occupations. These include
production jobs, craft jobs such as carpenters and electricians, and transportation-related jobs, as
well as less-skilled jobs such as garbage collectors and construction laborers. Most of the
occupations in which women posted disproportionately small gains in employment are in this
area.1

6-1
For example, truck driving has seen a large increase in women, but the percentage is still low,
rising only to around 4 percent today compared with under 2 percent in 1978.2 Female drivers
indicate that besides facing male hostility and being treated as sex objects, they have had to
overcome strong doubts about their capabilities from prospective employers.3 One woman
related that when she walked up to the loading dock:4

All the guys would glare, as if to say, ``What is she doing here?'' They gave me four 55-
gallon drums and said, ``If you can load 'em, you can haul 'em.'' So I got the dolly while
they stood back and watched. I had to climb up to push that thing over, but I passed the

test. I knew that I couldn't break down or quit, because I was representing any other
women who would come along.

Pink-collar occupations
Women continue to enter the clerical and service occupations in large numbers. While these
occupations were already feminized by 1970, they continued their feminization process
throughout the 1970s and 1980s.5 Meanwhile men continued to shun these jobs. Only three
service occupations had noticeable gains in male representation: cooks, kitchen workers, and
house servants.6 Notably, many of the men entering these areas are members of minorities, and
many are recent immigrants.

The professions
In the professions as a whole, there has been a noticeable trend of rising female participation
rates. In fact, it is startling how many of the younger members of the various professions are
women now, compared with 30 years ago. Table 6.1 compares percentages of bachelor's and
master's degrees awarded in various professions to women in the 1974, 1984, 1994, and 2002
school years. In most of these cases, there was a sharp rise from 1974 to 1984 and a continuing
rise from 1984 to 2002 at both levels. Education has the flattest pattern, in part no doubt because

6-2
it already had a higher proportion of women in 1974. Several occupations, including accounting,
journalism and pharmacy, have shifted so far as to become majority women. Computer science
is the only occupation in the table that shows a reduced level of women's entry relative to 1984.

In 1956 only 3 percent of law degrees were awarded to women. By 2002 the percentage had
risen to 48 percent.7 This is sizable feminization of a prestigious profession. Women, however,
are overrepresented among the less-well-paying jobs in law, such as jobs in legal clinics, and
appear not to rise to the top even in the most lucrative area of large law firms.8 In 2000, the
percentage of law firm associates who were female was 42 percent, and the percentage of law

firm partners who were female was 16 percent.9

Female lawyers may, however, be having their biggest impact on the law profession, as well on
the working world as a whole, through the efforts of some of them to ``redraw the landscape of a
legal world by replacing the `male' value system of rights, rules, and hierarchies with `female'
values based on relationships, responsibility and caring.''10 Women lawyers are found in large
numbers in government, the nonprofit sector, and in family law, and 20 percent district and
circuit court judges were women in 2000,11 implying that there may be many more sympathetic
ears for hearing a recasting of the law to embody more ``female'' values.

In 1950 10 percent of medical degrees were awarded to women. By 2002 the percentage had
risen to 44 percent.12 While women earn less than men in the profession, as shown in Table 6.10,
it is still an extremely attractive option in terms of the pay and employment possibilities. To
older women in the profession, in particular, the increased openness of the profession is an
impressive change. One 42-year-old, who went to medical school at age 34, commented: ``I
always wanted to be a doctor, but I was always told to do something more `acceptable.' Women
of my generation had to make conscious choices between career and family. But since then, the
trail has been blazed by many women, and today you can be a doctor and have a family.''13

6-3
Managers
A 1990 survey of 4012 senior managers in 800 major corporations found that only 19 were
women - under 0.5 percent.14 Meanwhile, women filled over a third of all management positions,
compared with 19 percent of such positions in 1972.15 Studies by Columbia and Stanford
Universities of women MBAs show that starting salaries are similar between the sexes, but that
seven years out the door, the women are 40 percent behind the men.16 This difference cannot be
explained by arguments that women overall do not yet have the experience to rise high in the pay
and responsibility structure. The term ``glass ceiling'' has been coined to describe the

phenomenon that while it now appears possible for women to rise to the top, few actually make
it. The term has been used to describe the situations found in law firms and academic medicine,
as well as in corporations. It appears to many that a relatively impenetrable wall, built from
covert discriminatory practices on the part of top-level managers, exists. An alternative
explanation is that few women, compared with men, are willing to make the sacrifices in terms
of time and effort necessary to rise to the top of a major corporation.17

Various programs have been suggested for trying to help women ``crack the glass ceiling.'' They
include reducing workloads for women after they give birth, and offering training sessions to
promote leadership development.18 The implications of such programs will be considered at
length in Chapter 7.

Tokens: Women in predominantly male occupations

In 1975 the percentage of women in U.S. Congress was 4 percent; in 2005 it was 14 percent. A
similar small rise is found in other corridors of power, such as major corporations and the top
echelons in professions. A small number of women in an occupation may be ``tolerated'' in the
sense of not leading to notice on anyone's part that the occupation is becoming female-dominated
or even integrated.19 In fact, they may be viewed as an asset in the sense of receiving disparate

6-4
attention, thereby drawing attention away from the lack of rank-and-file members and/or the
failure of a proportionate number of women to make the rise from the rank-and-file. About
medicine, one female psychiatrist said: ``Right now, women doctors in positions of power are
window dressing, with one scattered here and one scattered there - without enough votes to make
a change in academia, hospital administration or professional associations.''20 Nevertheless, it
appears that women in male-dominated occupations make substantially less relative to the men
in the occupation, even though they make more than most members of female-dominated
occupations.

FOCUS
Directors and officers at Fortune 500 companies

Women are not yet proportionately represented at middle management, but they are clearly
underrepresented among top management, even in consumer industries that sell mainly to
women. In 1992, Fortune magazine surveyed 201 chief executives of companies (CEOs) in the
Fortune 500 and found that only 2 percent considered it very likely and 14 percent somewhat
likely that the company would have a female CEO in the next ten years.21 Table 6.2 contains
some recent data illustrating the small numbers of women at the highest levels of corporate
America.

Top dogs: Men in predominantly female occupations

At colleges that I visit in the course of lecturing, I often bet students that the librarians in charge
at the college library are men. By ``in charge'' I mean the head librarian, the head of reference,
and the heads of most of the special collections - in particular, any that are particularly valuable
and/or prestigious. I am rarely proved wrong. Librarians as a whole are over 80 percent female.

6-5
However, a disproportionate percentage of the most prestigious and influential positions, which
are generally also the best-paying, are held by men. In fact, any position that involves
supervisory capacities is disproportionately male.

One female-dominated profession where this tendency of men to rise to the ``top dog'' positions
has met with particular dismay from the rank-and-file is nursing. Many nurses actively resent
men who enter nursing because they are viewed - probably accurately - as tending to rise to
supervisory positions more often than women. Similarly, men in clerical positions are often
considered by themselves and their peers to be just ``stopping through'' on their way up to

management; this impression is generally validated by their later career paths.22 Sociologist
Christine Williams uses the phrase ``glass escalator'' to describe men's apparently accelerated
progress to the top in female-dominated occupations.23

Interpretation of large changes in the proportion of women in some occupations

Studies of occupations that have shown recent rises in percentage of females have concluded that
these rises in female participation may be accompanied by increased intraoccupational
segregation and by job de-skilling. The general pattern for intraoccupational segregation is that
men tend to become more and more concentrated in the higher-paying jobs within the occupation
or within the higher-paying firms.24 De-skilling is the process by which jobs in an occupation
become more routinized and generally involve less responsibility. This may or may not be a
situation in which persons in the occupation have either less formal education or less on-the-job
training. De-skilling may be a primary reason why wages fall as these occupations become more
female-dominated.

It appears that the question upon seeing a rise in percentage of females in a traditionally male
occupation is: is resegregation, ghettoization, or integration occurring? Observed rises in

6-6
percentage of females may be an intermediate step in the process of tipping, which is the
eventual abandonment by men of male-dominated occupations that are entered by a critical mass
of women, leading to the creation of a newly female-dominated occupation. (In the opposite
case, a male-dominated occupation is created.)25 The final outcome of the tipping process is
resegregation of the occupation. Alternatively, the percentage of females may eventually
stabilize at a higher level, but the women may be predominantly in particular jobs within the
occupation, specifically the lower-paying, less prestigious jobs - i.e., ghettoization. The other
possibility, an occupation with a stable, integrated population, appears to be the rarest outcome.
Specific occupations are discussed below to illustrate these various phenomena.

The tipping phenomenon

Examples of job tipping abound. Although there were almost no women bank tellers before
World War II, over 90 percent of tellers were female in 1980. Meanwhile, salaries and career-
advancement possibilities dropped precipitously.26 Clerical professions, in general, were
predominantly male when they first came into existence in large numbers as the industrial
revolution generated more need for paper processors; all of these occupations are now female-
dominated and generally considered to be the female ghetto of jobs.27

It is rare to find an occupation that tips from predominantly female to predominantly male. The
most interesting example may be the occupation of baby-delivering, once performed almost
exclusively by female midwives, then performed almost exclusively by male doctors, and only
recently becoming a more heavily female subspecialty within the medical profession. In Britain,
the invention and use of forceps by a particular set of related male doctors in the seventeenth
century led in part to this development.28 Forceps deliveries had a high success rate, and women
were soon demanding this new technology to aid their births. In the United States as well as in
Britain, the male-dominated medical profession was instrumental in urging that women turn
away from use of midwives for home deliveries in favor of hospital deliveries performed by

6-7
doctors. Whereas in 1910, about 50 percent of babies were delivered by midwives, fewer than 1
percent were by 1970.29

De-skilling
Insurance adjustment appears to be a clear case of de-skilling accompanying the feminization of
the occupation. This occupation changed from being 30 percent female in 1970 to 72 percent by
1988.30 A case study of this occupation concluded that the feminization was a direct result of the
routinization and automation of the work process, primarily through the introduction of
computerized claims-processing. Claims processing has become more of an indoor office job,

instead of one that entails traveling to claim sites and exercising individual judgment. Men
abandoned these jobs, and firms hired women to replace them. Concurrently, the pay associated
with this occupation declined.31

Different skills, but not necessarily lesser skills?

Contrast the previous case illustrating de-skilling to the case of court reporting. Modern court
reporting involves the use of a 22-key stenotype machine. One needs to be able to record in
excess of 200 words per minute to become a court reporter. Previous to development of the
stenotype, shorthand systems were developed for pen and paper and are still in use in many
situations. However, as the percentage of pen writers has declined, the percentage of females has
been rising. In 1942 the profession was 74 percent male, and only 10 percent of the members
used a stenotype machine.32 By 1975 the profession was only 45 percent male,33 and 87 percent
of the members used a stenotype machine.34 Additionally, a survey of schools teaching the
subject in 1975 indicated that the student body was only 10 percent male.35

By 1990 there were practically no pen writers left in the courts, and the profession was moving
toward the increased use of computer systems to aid the cumbersome process of transcribing the
paper tape record produced by the stenotype machine into a finished transcript of court or

6-8
deposition proceedings. A 1990 survey of Tennessee court reporters found that 81 percent were
women, 7 percent were pen writers, and 84 percent reported using a computer-aided transcription
system at least part of the time.36 It appears that the skills required to enter this profession are
actually greater now than in the past and/or that they are so different from those in the pen-
writing days - including the demand for greater manual dexterity - that a smaller percentage of
the male population than of the female population is capable of doing the work.

Meanwhile, it appears that salaries continue to be high for this occupation, despite compositional
changes. Court reporting requires only a high school degree, plus the necessary training to

develop speed and accuracy in reporting.37 Yet court reporters' average earnings were greater
than those for many professionals requiring a B.S. or B.A. degree, including teachers, engineers,
and pharmacists.38 It is impossible to answer the question of how high earnings would have been
had the profession remained male-dominated. It is notable that in states with higher proportions
of female court reporters, the wages appear to be lower. However, these also tend to be the states
such as Arkansas and Mississippi where wages are lower in general.39

Ghettoization
Examples were given above of occupations in which women are found in the less prestigious
jobs. In some occupations, the differential placement of men and women is less obvious to the
outsider. For instance, in baking, which would appear to be a relatively undifferentiated
occupation in terms of jobs and pay, the percentage of female bakers has risen from 25.4 percent
in 1970 to 42 percent in 2004.40 However, men continue to dominate production baking, meaning
that they are more likely to be found in automated factories, where the jobs are higher-paying.
The absolute number of men has declined due to automation; meanwhile, women continue to be
found in the roles of finishing and packaging baked goods, in the lower-paying retail sector. The
evidence suggests that technological innovations in preserving dough has allowed for more
``bake-off'' bakeries, where prepared dough is delivered for baking rather than dough being

6-9
prepared from scratch on the spot. The expansion of the baking workforce in these retail outlets
has accommodated the rising female employment.41

Integration
It is notably difficult to find an example of a truly integrated occupation, where the proportion of
women closely matches their representation in the workforce, where the rate of change in the sex
ratio is small, and where women are not ghettoized. Bartending appears to be such a case.
Female bartenders were 21 percent of the total in 1970, rising sharply to 44 percent in 1980, but
leveling off at 50 percent by 1988 and 52 percent by 2004. The unique characteristics of the

profession are worth noting. Exclusionary laws that existed prior to 1970 at both the state level
and within the bartenders' union were struck down. There is no job ladder in bartending to allow
for segregation by rank and no range of industries for bartenders to separate into. (Four-fifths of
bartenders work in the single industry of food and beverage service.) Finally, since only one or
two bartenders work per shift, there is little or no opportunity for division of labor or for the
distinction of a ``managing'' bartender. However, one might well expect to find traditions of only
men or only women being hired into particular bars or shifts.42

Segregation index values

For many purposes, it is useful to create a summary measure that characterizes the overall level
of segregation, such as a segregation index. This allows us to compare changes over time in the
degree of segregation and differences between countries and between demographic groups and
sectors within a country. The most commonly calculated index, used to measure segregation for
two types of people (e.g., men and women, or whites and nonwhites) between any number of
different classifications (e.g., occupations), is the Duncan index of dissimilarity.43 This index
number is calculated as:

6-10
 Xi Yi
N X - Y
100 * ∑ 2
i=1

where there are N occupations and Xi is the number of persons of a group in occupation i, X is
the total number of persons in this group, Yi is the number of persons in a comparison group in

occupation i, and Y is the total number of persons in the comparison group. The index sums up
the absolute value of the differences between the proportions of each group (measured relative to
each group's total employment) for each occupation. The index ranges from 0 (complete

integration) to 100 (complete segregation). Notice that this index uses the implicit definition of
integration as a situation where the proportional representation of each group is the same in all
occupations as for the national workforce. For instance, if 30 percent of the national workforce is
female, then the index would measure 0, or complete integration, only if each occupation were
30 percent female. An interpretation of the index is that it shows what percentage of either group
would have to switch occupations in order to achieve complete integration. If the index equals
40, either 40 percent of men would have to switch into relatively female-dominated occupations
or 40 percent of women would have to switch into male-dominated occupations. While other
indexes can also be used to characterize segregation, the Duncan index has been widely used and
has thereby become a standard of comparison across different researchers' efforts.44

To see how this index is calculated, consider the following set of hypothetical cases. Assume that
there are only two occupations in the country, blue-collar and pink-collar. Table 6.3 illustrates
four cases for the distribution of employed men and women between these two occupations.
Note that we do not need to know how many women and men there are in toto in the workforce;
we only need to know the percentage distributions of men and women across occupations. At
one extreme, if all men are blue-collar workers and all women are pink-collar workers, there is
complete occupational segregation. At the other end of the spectrum, if the same percentage of

6-11
employed men as employed women are found in each occupation, there is complete integration.
Two intermediate cases are illustrated as well, which show that the index increases as the
percentages of women and men diverge more extensively across occupations.

Turning now to actual numbers for the U.S. workforce, Table 6.4 shows occupational
segregation indexes for 1960 to 2000. Sex segregation indexes are calculated for the whole
workforce, as well as separately for whites and nonwhites. There has been a notable downward
trend in segregation since 1960.45 While in 1960, 64 percent of either men or women would have
had to change jobs to drive the overall index down to 0, by 2000 only 52 percent would have had

to switch. The patterns for whites and nonwhites are similar, but with a more notable decline
among nonwhites. However, little additional decline occurred between 1990 and 2000.

The rise from 1960 to 1970 both overall and for whites may be explained by the fact that the
number of occupational categories available in the data rose sharply between 1960 and 1970,
from 291 occupations to 438. The number of categories rose again in 1980, to 503, but the
classification system did not change between 1980 and 1990, and the total number of categories
only rose to 505 in 2000 (although groupings were altered). Increasing the number of categories
over time imparts an upward bias to the index calculations. Opinion is divided as to whether it is
better to attempt to standardize over time by using a smaller number of matched, more inclusive
categories or whether the increasing number of classifications is a legitimate reflection of
segregation.46

Using 1990 Census data, one can clearly see the effect on the index of increasing the number of
occupational categories. Census data contain several levels of detail for describing occupational
and industrial categories. As shown in Table 6.5, using six very broad occupational classes, the
index value is 33.47 It climbs to a level of 59 if the fullest possible job class detail - 503
occupations within 224 different industries - is used, which yields a total of 36,669 occupied job

6-12
classifications (because most occupations are not represented in all industries).48 The biggest
jump in the index is found in moving from 16 to 83 occupational categories, when the index
climbs from 37 to 48. So, even at a very nondetailed level of occupational classification,
segregation is quite high.

Another way to look at the data is to calculate the index for subsets of the labor force. Using
2000 Census data, Table 6.6 presents segregation indexes for four age groups, five schooling
groups, the public and private sectors separately, the private sector broken down between goods-
producing and service-oriented industries, and the public sector broken down between federal,

state, and local governments. Index values rise slightly with age. By educational attainment,
values drop off greatly for the top two educational groups. Differences between the public and
private sectors in occupational segregation are small, although the public sector is slightly lower
overall. No real difference is found in the private sector between the goods-producing and
service industries, but segregation declines as the level of government rises from local to federal.
Table 6.6 demonstrates that segregation is remarkably consistent across subsectors of the
economy. This consistency implies that sectoral and demographic shifts cannot be counted on to
create much downward movement in the segregation index calculated for the entire labor force.

As a contrast to these trends in occupational sex segregation, consider the patterns for race
segregation over this time period. Table 6.7 contains occupational race segregation indexes
calculated separately by sex. These indexes show a marked decline in race segregation since
1960 for both women and men, although at a diminishing rate over the last decade. Additionally
the race segregation series of indexes both start and end at lower levels than the sex segregation
series. By 2000 only 23 percent of either white or nonwhite men and 20 percent of white or
nonwhite women would have had to switch jobs in order to achieve racial integration within each
gender group.

6-13
How much sex segregation are we overlooking by concentrating on the occupational detail
available in Census data? Case studies of particular occupations often show a great deal of
intraoccupational segregation, and it appears that men tend to be concentrated in the higher-
paying subspecialties in both highly male and highly female occupations. For instance, among
economists - a predominantly male occupation - there are large sex differences in research
specialties, with most of the women in the more applied fields such as labor economics, and few
women working on the theoretical side of the discipline.49 Among librarians - a predominantly
female occupation - men are proportionately overrepresented in administrative positions in
research libraries.50

There is scant evidence regarding segregation index values at more detailed levels of job
classification than simply occupations, but research in this area is quite unanimous in finding
high levels of segregation. When individual firms or establishments51 are studied and attention
paid to segregation by job classes, intrafirm values of the Duncan index commonly rise to 90. In
one study of 373 establishments, men and women in 60 percent of the establishments were
completely segregated by job title, i.e., there was no job type in which both men and women
were found.52

It appears that without drastic social change, no more than a slow downward trend in sex
segregation is likely for the foreseeable future. Factors that could maintain the currently observed
high levels of sex segregation include the growth of jobs in occupations that are presently quite
segregated, such as clerical jobs, and the possible slowdown of further rises in female
participation in occupations such as those that have experienced an influx of women in the recent
past. One should not, however, look only at new positions as providing the possibility of
reallocation of women and men across occupations. Mobility between occupations and jobs is
quite high, and desegregation could conceivably be brought about primarily by job changers
rather than by new entrants and exits from the labor market.53 It appears, however, that there is

6-14
little possibility in the near future of achieving levels of occupational sex segregation as low as
those currently found for race segregation.

[insert cavepeople cartoon about here]

© 1994: Reprinted courtesy of Bunny Hoest and Parade Magazine

Cross-cultural segregation data

How representative is the modern U.S. experience of sex segregation patterns in general? This

section sketches some general patterns found in collected ethnographic data on a wide variety of
societies, some historical and some contemporary.54

Research on both other industrialized nations and preindustrial cultures provides interesting
similarities and contrasts. Industrialized countries that have been studied extensively include the
larger Western European nations, Japan, Israel, and Russia.55 In these societies, which can be
considered roughly comparable to the United States in terms of occupational distribution and
female labor force participation, it is clear that despite substantial variability in age patterns of
labor-force participation and in the extent to which women engage in market work, there is
substantial similarity across cultures in the high level of sex segregation. Moreover, the same
occupations tend to be dominated by men or women in all these countries.

This pattern of an occupation's being dominated by the same sex in different industrial societies
does not carry over as a good description across nonindustrial societies. In examining evidence
derived from ethnographic reports on societies from different time periods and from all parts of
the world, it is striking both how few activities are integrated within any culture and how
activities vary in their assignment to one or the other sex, depending on the culture.56

6-15
Illustrating these points, Table 6.8 uses data from these 863 societies to create a simple
breakdown for 11 activity groups by whether only or mostly men, both sexes equally, or only or
mostly women perform chores in the activity group. Only metal working and hunting are
exclusively male activities; no activities are exclusively female; and few societies assign any of
the activities to both sexes equally. Additionally, this tabulation conceals sex segregation that
may take place within any activity group. For example, in agriculture, women may handle only
certain crops or animals, while men tend to others.

To demonstrate this point and underline the earlier point about the lack of exclusively female

economic activities, Table 6.9 contains data from 224 nonindustrial societies for which detailed
descriptions of work activities are available. Activities are classified for each society by whether
they are done exclusively or predominantly by one sex or the other. The first column (M) gives
the number of tribes for which the particular activity is only performed by males. The second
column (M-) gives the number of tribes in which women engage in the occupation only relatively
infrequently or in a subordinate capacity. The third column (=) counts those tribes in which the
occupation is either carried on indifferently by either sex or cooperatively by both (which still
implies that complete sex segregation of the activity may be occurring at any point in time). The
fourth column (F-) counts those tribes where women predominate, and the fifth column (F)
counts those tribes where women exclusively perform the activity. The sixth column (Percent)
gives a rough index of the degree of masculine involvement in the occupation, where the first
five columns are scored 100, 75, 50, 25, and 0, respectively, and the scores averaged for each
activity.

Even activities that are predominantly female, such as water carrying, are undertaken exclusively
by men in some societies. Still, Table 6.9 conceals the fact that specific subactivities within an
activity group may be regarded as exclusively performed by males or females. For example, in

6-16
the course of building a house, men may raise the timbers, while women prepare and affix the
roofing material.

The economic type of society - whether foraging (hunter-gatherer), horticultural (simple

agriculture), pastoral (herding), or agrarian (advanced agriculture) - does not appear to influence
this pattern of the relative scarcity of integrated activities, although exceptions can be found at
the foraging level, as discussed in Chapter 12.57 Segregation in tribal societies is much more
prevalent than integration, and it is so extreme in some societies that almost all work activities
are defined as either male or female, with the result that the sexes congregate in ``sexual ghettos''

while carrying out their daily work routines.58 Widespread sex segregation, rather than the
prevalence of sex-dominance of particular activities, is the tie bridging these preindustrial
cultures' segregational practices to the industrial societies discussed above.

Theories of why segregation occurs and persists

The above data make a convincing case for the view that sex segregation is widespread. This
section of the chapter will examine different explanations of why this phenomenon occurs and
persists: (1) gender differences in tastes for work activities, (2) gender differences in abilities for
work activities, leading to the exploitation of comparative advantage through division of market
labor, (3) efficiency in separating the sexes so as to reduce work disruptions related to sexual
tensions, (4) needs to balance market with nonmarket labor and other familial concerns, (5)
imperfect information about relative abilities between the sexes on the part of employers, and (6)
exploitation of most women by men or by another subset of society.

Gender-linked differences in tastes and/or abilities

Given the prevalence of segregation, it is natural to look for factors common to all human
societies to explain this phenomenon. The first factors to consider are differences in tastes and

6-17
abilities between the genders, regardless of the type of society examined, that lead to different
job choices for men and women.

Considering differences in abilities first, a piece of evidence supporting the idea that differences
in abilities alone cannot be the cause of segregation is found in the data presented above on
preindustrial societies. As noted, there are only a few cases among those that have been studied
(such as hunting) where men predominate in all societies, and no activities are ascribed solely to
women. This implies also that sex-linked differences in tastes between people, while a possible
cause of sex segregation, must be generally society-specific. It appears that differences between

the genders in tastes for various job characteristics are a strong determinant of the labor supplies
to different occupations.59 Still, the very fact that tastes do appear to differ among societies (for
work as well as for other things, e.g., food) casts doubt on whether they can be considered as
exogenous, sex-linked forces leading to occupational segregation.

The importance of efficient allocation of human resources

If women are, indeed, predisposed to be more efficient at some tasks and men at others, then any
imposed decline in segregation that caused certain tasks to be performed by the less apt parties
would be inefficient. If this is true, then in an ideal state, men and women would be free to
choose whatever job best suited them, given their tastes and abilities upon entering (and
subsequently relocating within) the labor market. There would be no justification for imposing
an outcome of integration on such a system, or for permitting attitudes and practices that impose
segregation.

Even if men and women have similar distributions of similar tastes and abilities, a highly
segregated situation could be more efficient than an integrated one. One argument, which has
appeared frequently in the popular press, is that any situation in which men and women work
together has sexual ramifications that may not be conducive to maintaining an efficient

6-18
workplace. Concern is expressed that sexual harassment, disruptive office romances, and
nepotism accompany the increase of women in the workplace.60 However, this is an argument for
segregating worksites, not for segregating by job class within a worksite. These problems arise
with or without the existence of men and women in the same job class. While men and women
are currently highly segregated by occupation, they frequently come into contact during the
workday. The stereotypical female secretary-male boss situation has been in existence for
decades, with the accompanying closeness sometimes leading to unpleasant situations, but also
often leading to mutually pleasant and productive situations. In addition, some writers cite
positive aspects of increased workplace contact between men and women to be reaped by both

employers and employees, such as increased loyalty to the firm and more cooperation between
workers.61

Finally, the different structures of occupations, in conjunction with market-nonmarket work

divisions, may motivate segregation. In societies that maintain a division between market and
nonmarket activities, such as the United States today, whichever sex has primary responsibility
for the nonmarket activities, such as childrearing or household chores, may take on market tasks
that are compatible with the demands these nonmarket activities create.62 This may involve
choosing occupations in which part-time employment is more likely63 or in which intermittent
labor force participation is not heavily penalized.64 However, if no difference in tastes or abilities
is postulated, it is still indeterminate which sex will be the one to choose nonmarket over market
activities.

Segregation as a response to market imperfections

Alternatively, could sex segregation arise because of labor market imperfections, such as the
situation in which employers have imperfect information about an individual's abilities?
Employers may hire women only for female-dominated jobs if individual ability is unobservable.
If women are not observed in nontraditional jobs, then there is no reason to update the belief that

6-19
they are not found in those positions because they cannot do such jobs effectively. In this
situation, no individual employer may have incentive to change, and there would be room for
intervention which would improve overall efficiency. Increasing female representation in male-
dominated jobs (and vice-versa) would enable employers to revise their beliefs based on actual
experience. With more information, workers could be better allocated among jobs. This approach
suggests placing persons in jobs nontraditional for their sex for the purpose of allowing people to
see how they do in such positions, rather than of having a large effect on the relative earnings of
the sexes in the short run. This might well be a useful approach for those dissatisfied with the
current state of sexual integration to take. Since we do not have a large sample of integrated

occupations, let alone integrated societies, it might be worth running a large social experiment to
see what happens in an integrated society. Any possible loss in efficiency during the trial period
might be offset by the value of running such an experiment in order to observe how an integrated
society operates.

The importance of distributional considerations

The preceding two sections were concerned with allocative efficiency, defined as the distribution
of men and women in the workforce so as to maximize total output of goods and services
without regard to the distribution of output. But both the distribution of output in a society and
the total output produced matter in evaluating overall social welfare. Does segregation serve the
needs of both sexes, or does it favor one sex over the other? Or is some other subset of the
society, not necessarily divided along sex lines, favored by this system?

The idea that men are the ones who benefit from sex segregation has been advanced by many
writers.65 Empirical evidence that female-dominated occupations pay less than male-dominated
ones is often cited to show that there are effects of sex segregation on earnings that benefit men
over women. This may be the most forceful argument for changing the system, but it also implies
that sex segregation will be difficult to change using policies that focus primarily on the

6-20
workplace. If segregation is a symptom of underlying discrimination, then concentrating on it
rather than modifying the underlying taste for discrimination is not a long-run solution.

Another idea is that for a social system to persist, the majority of the society must find some
benefits in it, though different benefits depend on one's social situation. For example, while
women may not reap huge direct monetary rewards in the workplace under a system that
relegates them to lower-paying jobs, given an income-sharing system such as marriage, they may
find that other rewards are available, such as the possibility to engage solely in nonmarket labor
during certain periods in their lifespan. At any point in time, this system would benefit married

women more than unmarried women. However, at a point in society in which conditions are
rapidly changing - as is the case now with high divorce rates and lower marriage rates - enough
dissatisfaction may occur for people to question the system.

Immutability vs. changeability

A theme running through the previous sections is the contrast between the viewpoint that
segregation is basically an immutable state versus the view that change is possible. If there are
strong forces pushing a society towards segregation, there is a high probability that policies
designed to lessen segregation will be adapted to in ways that preserve the prevalence of
segregation, with segregation perhaps re-emerging in forms less noticeable, given our present
ways of measuring its extent. We have already seen evidence on the consistency of segregation
patterns across cultures. The previous section of this chapter also provided evidence on changes
within occupations that support this statement.

At the same time, public policies affecting other social practices appear to have had substantial
effects, which argues that some social practices can be changed through active intervention. One
example is the apparently large effect of antidiscrimination laws on nonwhite hiring and earnings
patterns. However, gender differences in both earnings and hiring have proved to be more

6-21
resistant to change. Some observers use this contrast between eroding race differences and more
constant gender differences as evidence that gender differences are immutable - i.e., essential.
Others reject this view and argue that gender and race differences are essentially different
phenomena and that appropriate remedies differ as well.

Theoretical structures used by those seeking to justify desegregation will be inherently

unsatisfying if they do not build in an avenue for change in underlying motivations. If
segregation reflects learned feelings, in contrast to being the reflection of a set of immutable
basic feelings, then it may be possible to change those feelings. An empirically plausible model,

in which employers and employees become progressively more gender-blind in their work
behavior as they are exposed to more men and women working in nontraditional jobs, is needed
before a good case can be made that desegregational policies can really have long-term effects in
the workplace. A normative decision that gender-blindness is desirable, either because
discrimination is undesirable in itself or because the outcomes under discrimination (such as
women receiving lower wages) are unsatisfactory, is still required before one should attempt to
modify underlying tastes. These two aspects would comprise a coherent position: gender-blind
behavior is desirable, and integration in all aspects of society leads to gender-blind behavior as
tastes and beliefs are changed through seeing men and women perform capably in a wide variety
of roles.

FOCUS
Blind selection processes

One might expect that use of a truly gender-blind selection process in selecting new hires might
result in more women in various occupations. Recently, economists Claudia Goldin and Cecilia
Rouse documented the effects of the increased implementation of blind auditions in selecting

6-22
symphony orchestra members.66 Potential hires for major orchestras now play their audition
while sitting behind a screen; some orchestras even roll out a carpet to muffle footsteps or ask
players to remove their shoes while walking onto the stage so as not to have them reveal their
sex. Meanwhile, female musicians have increased their share of top positions notably,
comprising less than five percent of players in the top five symphony orchestras in 1970, but
twenty-five percent as of 1996. Goldin and Rouse demonstrate that use of a screen significantly
increases the probability that a woman will win the opening, accounting for as much as 46
percent of the increase in percentage female since 1970.

An interesting contrast is provided to this strong result by the findings from a study of double-
blind versus single-blind refereeing processes in journal article selection.67 The American
Economic Review carried out a two-year experiment in which every other submitted paper was
sent to referees with the authors’ names removed. Economist Rebecca Blank found that
acceptance rates were lower and referees more critical under double-blind reviewing, but the
patterns were not significantly different by sex.68 Interestingly, Blank found that the reviewers of
the supposedly anonymous papers were able to correctly identify at least one author on a paper
46 percent of the time.69 Nevertheless, The American Economic Review has instituted double-
blind refereeing on all submitted papers.

The relationship between segregation and earnings

Let us now explore the relationship between segregation and the gender earnings gap. It has been
noted by many researchers that jobs with a high percentage of women generally pay less than
jobs with a low percentage of women. Does this imply that there is a causal relationship linking
the two phenomena and, if so, what form does it take?

6-23
To illustrate this phenomenon, Table 6.10 presents data on percentage of females, female-male
median weekly earnings ratios, and median weekly earnings for a variety of occupations. Three
empirical regularities can be noted. The first is that there does appear to be a negative correlation
between percentage of females and median weekly earnings for this set of occupations. In fact,
the simple correlation coefficient for this table is -0.51.70 Second, the female-male earnings ratio
is positively correlated with the percentage of females (the simple correlation coefficient for the
table is 0.50); that is, men still make more than women in every occupation, but the gap is
narrower in the more female-dominated ones. Finally, the male-dominated occupations in the
table include some of the most prestigious ones: medicine, law, and college teaching.

Meanwhile, the female-dominated occupations include the traditional service-oriented

professions, such as nursing and elementary school teaching.

In measuring the overall effect of segregation on the gender earnings gap, one can decompose
the earnings gap into portions due to within-occupation (or firm, establishment, or job) gender
differences and between-occupation gender differences. One such study found that the gender
earnings gap caused by occupational segregation ranges from 11 percent in manufacturing to 26
percent in services (that is, the gender earnings ratio would still be 0.89 in manufacturing if
women were paid the same as men within each occupation but maintained their different
occupational distribution), while the earnings gaps caused by establishment and job segregation
are both about 6 percent across industries.71 Another study estimates that interfirm segregation is
extremely prevalent, and that it can account for up to 50 percent of the gap.72

As shown above, there are several explanations of why gender segregation occurs and persists in
societies, and it is not necessary to conclude that segregation is a bad thing. If people are free to
choose positions based on their tastes and abilities, including their desire for monetary
compensation, then a variety of patterns can result, ranging from complete segregation to
complete integration. However, if segregation is the outcome of a process in which people are

6-24
not free to choose a position because some jobs are closed to persons of a given group, then this
is disturbing, particularly if this process generates lower wages for one group and higher wages
for another. This link between segregation and lower earnings for women has led to much
support for comparable worth policies, which will be discussed in detail in the policy application
that concludes Part III.

The actual process that generates lower wages for women, even if women and men are equally
skilled, could take one (or both) of two forms. One, the exclusion model, is that women are
systematically excluded from higher-paying jobs and hired only to fill lower-paying ones. This

model assumes that pay is not linked to availability of labor - i.e., that pay rates are not
determined by market forces. The second form, the crowding model, is that women are
systematically excluded from more desirable jobs and crowded into less desirable ones.
However, desirability need not be related ex ante to the pay, but might, for instance, involve
status considerations. This model assumes that wage rates are ultimately determined by market
forces, but that female labor supply is artificially reduced in the more desirable jobs. This
practice leads to an increased supply of female labor to the less desirable jobs and depresses
wages in the less desirable jobs below what they would be if crowding were not occurring.73
Simultaneously, wages are increased in the more desirable jobs due to the decreased labor
supply.

Figure 6.1 represents the crowding model, using two occupations - blue-collar and pink-collar.
Before women are excluded from the blue-collar occupation and therefore crowd into the pink-
collar occupation, wages are the same in the two occupations. After the supply of labor increases
to the pink-collar occupation, its wage falls and employment increases. Meanwhile, the supply of
labor to the blue-collar occupation is decreased, causing its wage to rise and employment to
decrease.

6-25
This discussion assumes that men and women have equal abilities in both occupations and that
the occupations have similar structures in terms of the type of work performed. However, one
argument against a causal interpretation for the negative relationship between percentage female
and pay in occupations is that this correlation is picking up other factors that vary systematically
between male-dominated and female-dominated jobs. For instance, if female-dominated
occupations offer more part-time work opportunities, women may be attracted into them for the
ability to better maintain both market and nonmarket work. Alternatively, men may be attracted
into occupations that allow for more overtime work, once they have determined that they should
specialize in market work. This assumes that comparative advantage patterns dictate that women

are more likely than men to do at least some nonmarket work.

Table 6.11 shows that women and men do tend to be found in occupations with different hours
structures. This evidence takes two forms. One, which can be seen in reading across the rows,
shows that within a given set of occupations (classified by percentage female), women are more
likely than men to be found working part-time, and men are more likely than women to be
working full-time. Reading down the columns, we see that there is a positive correlation for both
sexes between the type of occupation, in terms of degree of percentage female, and the
percentage of workers who work part-time, and that there is also a negative correlation between
percentage female and the percentage of workers who work overtime.

To the extent that wages vary systematically with hours worked, this pattern provides a partial
explanation for the observed negative correlation between percentage female and hourly
earnings. It is clear that many part-time jobs pay lower hourly rates than comparable full-time
jobs; this can occur for reasons unrelated to discriminatory behavior on the part of employers. On
the cost side, part-time workers may be less productive than full-time workers per hour worked,
or the costs of using part-time workers may be higher than of using full-time workers. On the
labor supply side, as stated above, workers (particularly women) may be willing to take these

6-26
jobs at a lower rate as a trade-off for having more flexibility between market and nonmarket
work; a large number of such persons would drive down the wage in these jobs relative to jobs
that have few part-time work possibilities. It is even clearer that jobs requiring overtime must
pay higher average rates because higher overtime pay rates are fixed by law for many jobs.

So, one explanation for the observed correlation between percentage female and earnings levels
is that other variables that are also positively correlated with percentage female are the ones
responsible for the observed negative correlation with earnings. These variables could include
other factors about jobs besides the hours structure, such as riskiness or pleasantness of the work,

an idea explored in more detail in Chapter 8. Another explanation is that there are systematic
differences in the types of jobs that women and men tend to take in terms of the skills needed for
the jobs and/or the abilities that the people in those jobs have. Studies of the relationship between
earnings and percentage female generally control for factors such as the average educational
level required in the job (or the average education held by persons in the job) and the amount of
experience people who hold each job tend to have.74 However, there are many factors that are
generally not measurable by researchers - given time, money, and/or current measurement
techniques - that can have an effect on earnings, including physical characteristics such as height,
weight, and agility, as well as psychological characteristics such as leadership and motivation.
Since no researcher can ever hope to account for all possible earnings-linked characteristics that
persons may have, the objection that the researcher has left out some crucial factor and therefore
spuriously linked percentage female with lower earnings for a job classification is unavoidable.
Nevertheless, as we have seen above, evidence can be gleaned from a variety of sources to better
help us understand the mechanisms that lead to particular occupations' becoming male- or
female-dominated and, thereby, allow us to isolate factors that have changed over time to see if
we can explain the relationship between percentage female and earnings.

6-27
While there are exceptions, the evidence is clear that there is a strong correlation between lower
pay and higher percentage female across most occupations. Also, women are more likely to be
found on the lower rungs in any occupation. The question is, why do these patterns occur? Here
the evidence is mixed. While some occupations show evidence of routinization and de-skilling
concurrent with women entering them, others - such as many of the professions - appear to be
becoming more demanding. Also, while some of the pay gap may be attributable to differences
in training and experience, women who have decided to enter the most demanding occupations
in terms of training and yearly hours (e.g., doctors and lawyers) face some of the largest gaps in
earnings relative to men.

The crucial question, which will be examined in the following three chapters, is how much of
this pattern is attributable to discriminatory behavior on the part of employers, leading to
exclusion and crowding, and how much of it is attributable to freely made choices on the part of
women.

FOCUS
University coaches' salaries

Both The Chronicle of Higher Education and the National Collegiate Athletic Association
recently conducted studies that drew attention to the large disparity in salaries paid to men's and
women's teams coaches at universities.75 The Chronicle found that ``at the median Division I
institution in 1995--96, the average salary paid to the head coach of a men's team was 44 per cent
higher than that paid to the head coach of a women's team.''76 The biggest percentage difference
between the average for men's coaches and the average for women's coaches was found at the
University of Dayton: $57,050 and $16,470 respectively, for a difference of 246 percent. more
paid to the men's coaches. This translates into a pay disparity by gender of coach as well, as few

6-28
women are found coaching men's teams - although about 52 percent of women's team coaches
are men.

Effects of workforce policies on segregation

This section considers how segregation patterns might be influenced by various current and
proposed workforce policies aimed at improving women's well-being, including anti-
discrimination policies, comparable worth, and job training programs. These policies are not

generally implemented because of their effects on segregation, but they have such effects
nonetheless.

Policies are categorized according to their primary focus: (1) those designed to affect wages, (2)
those that attempt to modify employer behavior concerning hiring and promotion, (3) those
aimed at workers rather than employers, namely training and other educational programs, (4)
programs such as child care subsidies that affect decisions about whether or not and to what
extent to participate in the labor force, and (5) those that target sex integration as a specific
desired outcome.

Policies affecting wages

Some policies advocate adjusting the wages that come out of present-day market solutions. The
most topical such policy is comparable worth, which essentially involves raising the wages in
female-dominated jobs to equal the wages in male-dominated jobs that are deemed to be
``comparable.'' Comparable worth has been implemented in some settings, mainly state and local
governments, and has been advocated for wider use. Proponents and opponents are divided as to
what the effects of a broad-based implementation of comparable worth would be on occupational

6-29
segregation. In fact, it turns out that there is no way to prove that comparable worth will lead to
either increased or decreased segregation.

There is a clear contrast between supporters and opponents of comparable worth in terms of what
they generally think the effects on segregation will be. Supporters think it will decrease
segregation because higher wages in traditionally female jobs will encourage men to move in.
Opponents think it will increase segregation because higher wages in traditionally female jobs
will discourage women from moving out.77 The logic on both sides is correct in direction for
supply-side effects, but neither allows for changes in labor demand, nor does existing research

allow one to determine the relative magnitudes of these effects. Generally, both women and men
will find a job more appealing if it pays more. The ratio of women to men depends upon who on
the margin finds it most appealing - the new men attracted to the job or the new women - and
also upon whom employers choose to hire.

One study, which created a computer simulation of an economy-wide comparable worth system,
found that under a comparable worth policy in which wages were raised in female-dominated
jobs, women increased their labor supply relative to men in jobs throughout the economy.78
When wages were raised in female-dominated occupations, many women found it worthwhile to
work more, while many married men benefited from their spouse's higher income and therefore
found it optimal to work less. This outcome is consistent with the discussion of substitution and
income effects of wage and nonearned income changes as discussed in Chapter 4. Although in
this model labor force participation is couched in terms of total hours worked rather than by a
head count and unemployment shows up as worksharing rather than loss of jobs, it still has
implications for segregation as it is usually measured in that the women ended up comprising a
larger proportion of hours worked in both female- and male-dominated occupations. Note that in
such a real-life situation, the segregation index might therefore show no change.

6-30
In general, policies that affect wages in one or more jobs have an indeterminate effect on
segregation. Wage increases make a job more desirable to both sexes, and wage decreases have
the opposite effect, but the net change in segregation is impossible to determine unless a
complete specification of the relative changes in female and male supply and the hiring decisions
of employers is made. There is no a priori reason to assume that hiring patterns either respond to
changes in the sex ratio of prospective employees or stay constant.

There is, however, the possibility that hiring patterns can be affected by legal restrictions forcing
employers to consider the make-up of the applicant pool and hire in proportion to the sex and

race proportions found therein. If this rule is enacted, as is the case in strong applications of
antidiscrimination laws, then supply-side decisions will determine occupational make-up, given a
particular job structure. Under strong applications of affirmative action, an employer may even
be required to hire proportionately more women and minorities than their representation in the
applicant pool. If this were the case, then the hiring rule can, in fact, determine the gender make-
up of occupations. One might expect vacancy rates to rise in such a situation, as employers
compete for the few female applicants.

There is still opportunity for an employer to reclassify jobs in such a way as to resegregate the
workforce, but this becomes much harder to do if one is required to list all individual openings in
a firm and consider the applicant pool for each job separately. Before listing rules came into
widespread use, people generally applied to a firm for whatever position might be available and
employers were free to decide where to place such persons. This can still be done to some extent,
but records are often kept as to who has applied for a particular job. In addition, open
advertisement of internal labor market positions within firms also creates a record of the
applicant pool for each promotional possibility that can be monitored for evidence of gender bias
in intrafirm mobility. The possibility of a discrimination lawsuit should serve as a restraining
factor if employers attempt to promote men over women.

6-31
Policies affecting training
If rules such as affirmative action and antidiscrimination policies - which call for hiring and
promotion in proportion to applicant pools - are binding, then observed outcomes are indicative
of the relative supplies of men and women for particular jobs. If the outcomes in this case still
reflect a large degree of sex segregation, then attempts to influence relative supplies must be
undertaken if segregation is to be reduced.

One such policy is the attempt, using government-sponsored training programs, to prepare

women for employment in traditionally male-dominated occupations. The actual effect of these
programs on sex segregation appears to have been minimal, although individual participants may
have benefited from receiving higher wages in these occupations than they would have in other
areas. Training these women did not ensure that they would find an opportunity to apply their
skills, as they were still subject to the decisions of employers regarding specific positions and the
general availability of positions.79

A more fundamental policy is to increase awareness of work options among the young at a stage
when they are able to adjust their secondary education so as to take advantage of a chosen
option. This means both informing children about different occupations and telling them what
background is required in order to enter these occupations, so that they can train appropriately
for a chosen field. There is little argument to be made against increasing the information
available to young people; the question is how much to devote to stressing particular
occupations, such as engineering to women or teaching to men, in an attempt to influence the sex
ratios in such fields. Still, it appears that gender differences in educational attainment, in both the
decision of what type, if any, of postsecondary training to get and the choice of college major
among college attendees, have large influences on subsequent differences in occupational
choice.80

6-32
Policies affecting labor force participation decisions
Many other policies beside basic education and training affect women's and men's decisions
about what type and amount of labor force participation to seek. This decision may indirectly
affect the level of segregation. For example, availability of cheap, high-quality child care would
affect many women's decisions as to whether and how much to participate in the labor market.
Other policies of current interest include the availability of on-site child care, maternity leave,
and flex-time. However, policies that either increase or decrease labor force participation need
have little or no effect on segregation. To the extent that women on the margin are more likely to

enter female-dominated occupations, an increase in their employment will increase workforce

segregation. If, however, women are more likely to enter male-dominated occupations under
these circumstances, then such policies would have the opposite effect.

Other workforce policies

Other policies that do not fall under the categories considered above include directly rewarding
employers who act to reduce segregation. For instance, instead of concentrating on penalizing
firms that discriminate, one could provide subsidies to firms that show marked improvement in
integrating women into traditionally male jobs and/or in improving women's pay relative to
men's. Such subsidies could provide a market advantage to these firms and make them relatively
more profitable than firms who make no such efforts, or at least counter any costs incurred from
implementing these changes.

Still other policies can augment existing workplace institutions, creating new employment
possibilities without directly utilizing existing firms. One such policy is to increase the number
of self-employed persons by small business funding through loans, subsidies, or outright grants.
For this policy to affect segregation, women (and/or men) have to enter nontraditional lines of
business, such as plumbing and home improvement, rather than setting up shop in more

6-33
traditional fields such as beauty care and catering. To the extent that the distribution of
independent retailers and service providers is affected by differing costs of capitalizing such
businesses, making the costs of entering, say, plumbing and catering equivalent on the margin
would tip more women into the former than are presently found. Still, basic tastes and training
would affect the number of persons suited to enter each trade.

Conclusions about the effects of various policies

From the empirical evidence to date, it appears that current policies have had a limited effect on
segregation. While there have been some shifts in the relative numbers of men and women to

certain occupations (as reflected in changes in the percentage female for those occupations), it is
difficult to attribute these changes to any particular policy.

The main point to be gained from the arguments made above is that one should be wary of any
claims that a particular policy will definitely either increase or decrease sex segregation. The
answer in each case depends crucially on the particular assumptions made about labor supply and
demand responses to the policy. To the extent that hiring preferences of employers are
suppressed through imposition of hiring rules, the relative supply of male and female workers
effectively determines the level of segregation. Many policy analysts, however, assume the
former point and concentrate on supply responses.

If achieving a greater degree of integration in the workplace is desired, then clearly new
strategies need to be tried if the rate of integration is to be accelerated. One point to be
considered in advocating particular strategies is whether the time frame for desired change is
short or long. Some strategies, such as stepping up efforts to create a sex-blind educational
environment, will take longer for their effects to be felt in the workplace, but they may have
more lasting and widespread effect than ``quick fixes'' such as retraining women already in the
labor force to enter nontraditional occupations. Additionally, policies need to be undertaken

6-34
simultaneously if the probability of success in achieving integration is to be increased
substantially. In light of the discussion above, a combination of policies designed to affect both
hiring practices and the relative supplies of women and men to different jobs is required so that
pressure on both the demand and the supply side of the labor market is tuned towards integration.

Policy Application: Affirmative action

The implementation of affirmative action policies provides an interesting case study in the
contrast between such policies' sizable effect on minority employment - particularly for blacks

and Hispanics - and the small effect on white female employment. It appears that workforce sex
segregation in the form of distribution of workers between the covered and uncovered sectors,
where the sector covered by affirmative action is currently more male-dominated (and appears to
contain relatively better-paying jobs), has not been greatly influenced by these policies.

Affirmative action policy originally focused on improving employment possibilities for blacks.
Starting in 1941 with Franklin Roosevelt's Executive Order 8002, a series of executive orders has
barred race discrimination by federal contractors. In 1961 John Kennedy's Executive Order
10925 required contractors to implement affirmative action and established sanctions, including
contract termination, against noncompliers. Lyndon Johnson's Executive Order 11246 beefed up
enforcement, and his Executive Order 11375 in 1967 expanded affirmative action to include
women. However, effective regulations enforcing this expansion did not reach full stride until
after the Equal Employment Act of 1972. Since 1972, all federal contractors and first-tier
subcontractors with 50 or more employees (or a contract worth $50,000 or more) must agree to
``take affirmative action to ensure that applicants are employed and that employees are treated
during employment without regard to their race, color, religion, sex, or national origin.'' These
actions include maintaining written affirmative action plans containing goals and timetables for
correcting deficiencies in equal employment opportunity. In addition, many organizations that

6-35
are not required to have an affirmative action plan nonetheless have them, in varying degrees of
formality.

It is not clear that affirmative action policies have been particularly successful in increasing the
representation of women in better-paying positions, nor do researchers agree as to how much of
the increase that has occurred can be attributed to such policies. Enforcement of affirmative
action regulations has been sketchy.81 The Equal Employment Opportunity Commission is
limited to responding to discrimination complaints, while the Office of Federal Contract
Compliance Programs, the main government agency concerned with affirmative action in the

workplace as the overseer agency for federal contractors, is understaffed to the point where it
would take 38 years for it to investigate each contractor workplace once.82 Out of the employers
it checked in 1994--95, 75 percent were in substantial noncompliance.83

How might one measure the impact of affirmative action on women's employment patterns? One
way is to compare changes in female employment in the covered sector with changes in the
uncovered sector. For instance, several studies compare federal contractors, who are covered by
affirmative action, with similar employers who are not covered.84 While studies concentrating on
the effects of affirmative action on women's employment in the early 1970s found little effect -
or even net employment losses - for white women,85 by the late 1970s black women appear to
have benefited substantially, and white women somewhat. Table 6.12 shows that in 1974, 27.6
percent of the average federal contractor workforce were white women, compared with 39.4
percent among noncontractor employees.86 Because federal contractors are more likely than
noncontractors to be in the manufacturing sector and less likely to be in the service or retail trade
sector, we would expect women's employment share to be lower among contractors. However,
women's employment growth in the contractor sector outstripped growth in the noncontractor
sector. By 1980 employment of white women had increased by 1.2 percentage points among
contractors, but only by 0.6 percentage points among noncontractors. The author of this study

6-36
concludes that affirmative action has had a modest effect in increasing white women's
employment: ``If white female employment share among contractors had grown only at the
slower noncontractor rate, roughly 2 percent fewer white females would have been employed
among federal contractors in 1980.'' Black women have made more sizable employment gains
overall, but less of a difference in growth rates appears between sectors.

Studying the effect of affirmative action on federal contractors overstates the impact of
implementing affirmative action throughout the economy to the extent that contractor gains
represent a reshuffling from noncontractors that could not occur if there were no uncovered

sector. It could also understate the effect if there are spillover effects on noncontractors that
raises their female employment as well, but this appears unlikely.87 This methodology also
misses the effects on the uncovered sector of voluntary affirmative action or affirmative action
mandated as a remedy following litigation under Title VII of the Civil Rights Act of 1964, which
applies to all private employers with 15 or more employees and bars employment discrimination
on the basis of race, color, religion, sex, or national origin.

A recent paper critiques this methodology, pointing out that there are no specific statistical
controls for wage and technology differences between the sectors.88 Economic theory predicts
that firms subject to affirmative action will have higher costs than firms who are not if they find
hiring restrictions binding in the sense of limiting their ability to substitute between inputs (e.g.,
between black workers and machinery). The paper finds that constraints were binding in 1980
and that ``the cost of complying with affirmative action averaged 6.5 percent of total costs for
constrained firms.''89 This indicates that such restrictions may cause supply to contract in the
covered sector, thereby reducing the sector's demand for labor. One could interpret this as a
social cost of enforcing integration.

SUMMARY

6-37
As the female labor force grows, women are entering many occupations where they were
underrepresented in the recent past, particularly the professions. However, they also continue to
cluster in the service and clerical sectors and are making slower inroads into blue-collar
occupations. As the percentage of females rises in different occupations, these occupations
become resegregated, ghettoized, or integrated.

Sex segregation indexes have been dropping since 1960, but the rate of change is much less than
for race segregation indexes, and the levels are much higher. Segregation is prevalent in all

societies. Across industrialized societies, men and women are found in similar occupations;
across nonindustrialized societies, occupations vary in their assignment to men or women. Six
explanations are offered for the prevalence of sex segregation, ranging from choice-based
arguments to discrimination. These explanations are not mutually exclusive.

The interpretation of the relationship between pay and percentage female for a job is
problematic. While there is a negative correlation between the two, it may be due to factors about
the different jobs in which men and women are found or to discrimination on the part of
employers that leads women to be barred from entering particular jobs without difficulty.

The effects of particular workplace policies on segregation are generally unpredictable and hard
to discern. Affirmative action policies appear to have had little effect on sex segregation,
particularly for white women.

ENDNOTES

1. Barbara F. Reskin and Patricia A. Roos, Job Queues, Gender Queues: Explaining Women's
Inroads into Male Occupations (Philadelphia, Pa.: Temple University, 1990): 19.

6-38
 2. Hank Whittemore, ``I Want Respect, That's All,'' Parade (July 10, 1988): 4—5;
Employment and Earnings 52, no. 1 (January 2005) (Table 39)..
3. See Jean R. Schroedel, Alone in a Crowd: Women in the Trades Tell Their Stories
(Philadelphia, Pa.: Temple University, 1985), for a collection of interviews with women working
in predominantly male occupations.
4. Whittemore, op. cit.: 4.
5. See Louise Kapp Howe, Pink Collar Workers (New York: Putnam, 1977), for a collection
of interviews with women working in these occupations during the mid-1970s.

6. Reskin and Roos, op. cit.: 18.

7. Digest of Education Statistics (2003): (Table 260).
8. Cynthia Fuchs Epstein, Women in Law, Second Edition (Urbana, Ill.: University of Illinois
Press, 1993). See also American Bar Association Commission on Women in the Profession,
Unfinished Business: Overcoming the Sisyphus Factor (1996).
9. American Bar Association Commission on Women in the Profession, ``A Current Glance
of Women in the Law'' (2001). Accounting, another profession in which large partnership firms
are dominant, shows similar patterns by sex; see Shari H. Wescott and Robert E. Seiler, Women
in the Accounting Profession (New York: Markus Wiener, 1986).
10. Ted Gest, ``Feminizing the Law: The New Meaning of Equality,'' U.S. News & World
Report (June 17, 1991): 48.
11. American Bar Association Commission on Women in the Profession, loc. cit.
12. Digest of Education Statistics (2003): (Table 260).
13. Carol Kleiman, ``Number of Women Doctors No Longer So Anemic,'' Chicago Tribune
(May 20, 1991), Section 4: 5.
14. Carol Kleiman, ``3 Firms Serve as Catalysts for Women,'' Chicago Tribune (January 23,
1991), Section 3: 3.

6-39
 15. ``Women Make Slow Progress up the Corporate Ladder,'' The Economist (March 14,
1987): 61--62. See also Rochelle Sharpe, ``The Waiting Game,'' Wall Street Journal (March 28,
1994): A1, A10, and related articles in the same issue, for information from a 1992 study of
firms reporting to the U.S. Equal Employment Opportunity Commission.
16. Kleiman (January 23, 1991).
17. Those women who do rise to the top receive great attention from the press. Cf. Joseph B.
White and Carol Hymowitz, ``Watershed Generation of Women Executives is Rising to the
Top,'' Wall Street Journal (February 10, 1997): A1, A8.

18. Kleiman (January 23, 1991).

19. For the classic discussion of the tokenism phenomenon, see Rosabeth Moss Kanter, Men
and Women of the Corporation (New York: Basic Books, 1977).
20. Kleiman (May 20, 1991), quoting Dr. Michelle Harrison.
21. Anne B. Fisher, ``When Will Women Get To the Top?'' Fortune 126, no. 6 (September
21, 1992): 44--56. On a brighter note, see Joseph B. White and Carol Hymowitz, ``Broken Glass:
Watershed Generation of Women Executives Is Rising To the Top,'' Wall Street Journal
(February 10, 1997): A1, A8.
22. Carl Hoffman and John Shelton Reed, ``Sex Discrimination? The XYZ Affair,'' Public
Interest 62 (Winter 1981): 21--39.
23. Christine Williams, Still a Man's World: Men Who Do ``Women's Work'' (Berkeley,
Calif.: University of California, 1995).
24. Francine D. Blau, Equal Pay in the Office (Lexington, Mass.: Heath, 1977).
25. Barbara F. Reskin and Patricia A. Roos, ``Status Hierarchies and Sex Segregation,''
Ingredients for Women's Employment Policy, eds. Christina Bose and Glenna Spitze (Albany,
N.Y.: State University of New York, 1987): 3--21.

6-40
 26. Myra H. Strober and Carolyn L. Arnold, ``The Dynamics of Occupational Segregation
among Bank Tellers,'' Gender in the Workplace, eds. Clair Brown and Joseph A. Pechman
(Washington, D.C.: Brookings, 1987): 107--148.
27. The clerical labor situation in Great Britain is described in Samuel Cohn, The Process of
Occupational Sex-Typing: The Feminization of Clerical Labor in Great Britain (Philadelphia,
Pa.: Temple University, 1985).
28. Diane Mason, ``Facts About Forceps,'' American Baby (October 1990): 48--51.
29. Barbara Ehrenreich and Deirdre English, ``Witches, Midwives, and Nurses: A History of

Women Healers'' (Detroit, Mich.: Black and Red, 1973): 33.

30. Reskin and Roos (1990): 18.
31. Polly A. Phipps, ``Occupational Resegregation among Insurance Adjusters and
Examiners,'' Job Queues, Gender Queues: Explaining Women's Inroads into Male Occupations,
eds. Barbara F. Reskin and Patricia A. Roos (Philadelphia, Pa.: Temple University, 1990): 225--
240.
32. Mary Louise Gilman, ``Changing Times,'' National Shorthand Reporter (December
1976): 26--27.
33. Wesley V. Gales, ``?Men Wanted?'' National Shorthand Reporter (May 1976): 26--27.
34. Gilman, loc. cit.
35. Sandra W. McFate, ``Men Wanted,'' National Shorthand Reporter (February 1976): 13--
15.
36. Hollie W. Sharpe and Vincent W. Smith, ``A Profile of the Tennessee Shorthand
Reporter,'' Journal of Court Reporting, 52, no. 10 (August 1976): 30--32.
37. Gary M. Cramer, ``How Long Does it Take?'' Caligrams 17 (July 1989): 12--13.
38. McFate, loc. cit.
39. Ibid.

6-41
 40. Reskin and Roos (1990): 18; Employment and Earnings 52, no. 1 (January 2005) (Table
39).
41. Thomas Steiger and Barbara F. Reskin, ``Baking and Baking Off: Deskilling and the
Changing Sex Makeup of Bakers,'' Job Queues, Gender Queues: Explaining Women's Inroads
into Male Occupations, eds. Barbara F. Reskin and Patricia A. Roos (Philadelphia, Pa.: Temple
University, 1990): 257--274.
42. Linda A. Detman, ``Women behind Bars: The Feminization of Bartending,'' Job Queues,
Gender Queues: Explaining Women's Inroads into Male Occupations, eds. Barbara F. Reskin

and Patricia A. Roos (Philadelphia, Pa.: Temple University, 1990): 241--255.

43. Otis Dudley Duncan and Beverly Duncan, ``A Methodological Analysis of Segregation
Indexes,'' American Sociological Review 20, no. 1 (1955): 210--217.
44. For a discussion of other useful measures, see Christina Jonung, ``Patterns of
Occupational Segregation by Sex in the Labor Market,'' Sex Discrimination and Equal
Opportunity, eds. Gunther Schmid and Renate Weitzel (New York: St. Martin's, 1988): 44--68;
and Victor R. Fuchs, ``A Note on Sex Segregation in Professional Occupations,'' Explorations in
Economic Research 2, no. 1 (Winter 1975): 105--111.
45. These patterns are corroborated and discussed in several papers, including Judith Fields
and Edward N. Wolff, ``The Decline of Sex Segregation and the Wage Gap, 1970--80,'' Journal
of Human Resources 26, no. 4 (Fall 1991): 608--622; and Mary C. King, ``Occupational
Segregation by Race and Sex, 1940--88,'' Monthly Labor Review 115, no. 4 (April 1992): 30--36.
46. For a discussion of this point, see Jerry Alan Jacobs, The Sex-Segregation of Occupations
and the Career Patterns of Women (Ann Arbor, Mich.: University Microfilms International,
1983).
47. The classes are: Managerial and Professional; Technical, Sales, and Administrative
Support; Service; Farming, Forestry, and Fishing; Precision Production, Craft, and Repair; and
Operators, Fabricators, and Laborers.

6-42
 48There are only 500 inhabited occupations as the three military occupation codes are
excluded.
49. Myra H. Strober and Barbara B. Reagan, ``Sex Differences in Economists' Fields of
Specialization,'' Women and the Workplace, eds. Martha Blaxall and Barbara Reagan (Chicago,
Ill.: University of Chicago, 1976): 292--307.
50. Betty Jo Irvine, Sex Segregation in Librarianship (Westport, Conn.: Greenwood, 1985).
51. A firm consists of one or more establishments, which are physical locations where the
firm operates. For instance, a chain of restaurants is a firm, while each individual restaurant is an

establishment.
52. William T. Bielby and James N. Baron, ``A Woman's Place Is with Other Women: Sex
Segregation Within Organizations,'' Sex Segregation in the Workplace: Trends, Explanations,
Remedies, ed. Barbara F. Reskin (Washington, D.C.: National Academy, 1984): 27--55.
53. Rachel A. Rosenfeld, ``Job Changing and Occupational Sex Segregation: Sex and Race
Comparisons,'' Sex Segregation in the Workplace: Trends, Explanations, Remedies, ed. Barbara
F. Reskin (Washington, D.C.: National Academy, 1984): 56--86. Jerry A. Jacobs, Revolving
Doors: Sex Segregation and Women's Careers (Stanford, Calif.: Stanford University, 1989),
documents high mobility between male-dominated, neutral, and female-dominated fields on the
part of individual women and a slow net rate of accumulation of women in male-dominated
areas.
54. This theme will be developed further in Parts IV and V.
55. Patricia A. Roos, Gender and Work: A Comparative Analysis of Industrial Societies
(Albany, N.Y.: State University of New York, 1985), considers Japan, Israel, the United States,
and nine countries in Western Europe. The situation in Russia (the Soviet Union) is examined by
Gail Warshofsky Lapidus, ``Occupational Segregation and Public Policy: A Comparative
Analysis of American and Soviet Patterns,'' Women and the Workplace, eds. Martha Blaxall and

6-43
Barbara Reagan (Chicago, Ill.: University of Chicago, 1976): 119--136. Sweden is examined in
Christina Jonung, loc. cit.
56. These data are derived from the cross-cultural files maintained at the University of
Pittsburgh, as reported by George Murdock, Ethnographic Atlas (Pittsburgh, Pa.: University of
Pittsburgh, 1967); and George P. Murdock and Douglas R. White, ``Standard Cross-Cultural
Sample,'' Ethnology 8, no. 4 (October 1969): 329--369. A study of these data in terms of sexual
division of labor is George P. Murdock and Caterina Provost, ``Factors in the Division of Labor
by Sex: A Cross-Cultural Analysis,'' Ethnology 12, no. 2 (1973): 203--225.

57. See Charlotte G. O'Kelly and Larry S. Carney, Women and Men in Society, Second
Edition (Belmont, Calif.: Wadsworth, 1986).
58. Peggy Sanday, Female Power and Male Dominance: On the Origins of Sexual Inequality
(Cambridge, United Kingdom: Cambridge University, 1981).
59. Mary Corcoran and Paul Courant, ``Sex Role Socialization and Labor Market Outcomes,''
American Economic Review 75, no. 2 (May 1985): 275--278; Randall K. Filer, ``The Role of
Personality and Tastes in Determining Occupational Structure,'' Industrial and Labor Relations
Review 39, no. 3 (April 1986): 412--424.
60. For discussion of sexual harassment and sexuality in organizations, see Barbara A.
Gutek, Sex and the Workplace: The Impact of Sexual Behavior and Harassment on Women, Men,
and Organizations (San Francisco, Calif.: Jossey-Bass, 1985); Jeff Hearn, Deborah L. Sheppard,
Peta Tancred-Sheriff, and Gibson Burrell (eds.), The Sexuality of Organization (London: Sage,
1989).
61. Barbara Kantrowitz, ``Love in the Office,'' Newsweek (February 15, 1988): 48--52.
62. This was discussed in Chapter 3; see Gary Becker, ``Human Capital, Effort, and the
Sexual Division of Labor,'' Journal of Labor Economics 3, no. 1 (January 1985): 33--58.
63. This argument is advanced by Victor R. Fuchs, ``Women's Quest for Economic
Equality,'' Journal of Economic Perspectives 3, no. 1 (Winter 1989): 25--41.

6-44
 64. For a lively debate on this last hypothesis, see Solomon Polachek, ``Occupational Self-
Selection: A Human Capital Approach to Sex Differences in Occupational Structure,'' Review of
Economics and Statistics 63, no. 1 (February 1981): 60--69; Paula England, ``Wage
Appreciation and Depreciation: A Test of Neoclassical Explanations of Occupational Sex
Segregation,'' Social Forces 62, no. 3 (March 1984): 726--749; Solomon Polachek,
``Occupational Segregation: A Defense of Human Capital Predictions,'' Journal of Human
Resources 20, no. 3 (Summer 1985): 437--439; Paula England ``Occupational Segregation:
Rejoinder to Polachek,'' Journal of Human Resources 20, no. 3 (Summer 1985): 441--443; and

Solomon Polachek, ``Occupational Segregation: Reply to England,'' Journal of Human

Resources 20, no. 3 (Summer 1985): 444. This issue is discussed in more detail in Chapter 7.
65. Cf. Barbara R. Bergmann, ``Occupational Segregation, Wages and Profits When
Employers Discriminate by Race or Sex,'' Eastern Economic Journal 1, no. 2 (April-July 1974):
103--110; Heidi Hartmann, ``Capitalism, Patriarchy, and Job Segregation by Sex,'' Women and
the Workplace, eds. Martha Blaxall and Barbara Reagan (Chicago, Ill.: University of Chicago,
1976): 137--169; Jean Lipman-Bluman, ``Toward a Homosocial Theory of Sex Roles: An
Explanation of the Sex Segregation of Social Institutions,'' Women and the Workplace, eds.
Martha Blaxall and Barbara Reagan (Chicago, Ill.: University of Chicago, 1976): 15--31; Myra
H. Strober, ``Toward a General Theory of Occupational Sex Segregation: The Case of Public
School Teaching,'' Sex Segregation in the Workplace: Trends, Explanations, Remedies, ed.
Barbara F. Reskin (Washington, D.C.: National Academy, 1984): 144--156.
66. Claudia Goldin and Cecilia Rouse, ``Orchestrating Impartiality: The Impact of `Blind'
Auditions on Female Musicians,'' American Economic Review 90, no. 4 (2000): 715--41.
67. In double-blind processes, neither the author nor the reviewer knows each other's
identity. In single-blind processes, the reviewer knows the author's identity. Journals vary in their
policy both within and between academic disciplines, with natural science and medical journals
tending to be single-blind, political science and sociology journals tending to be double-blind,

6-45
and economics journals split, with a higher proportion using single-blind. A few journals use a
nonblind process (in economics only Review of Radical Political Economy, to the best of my
knowledge), where the referee signs the report given to the author.
68. Rebecca M. Blank, ``The Effects of Double-Blind versus Single-Blind Reviewing:
Experimental Evidence from The American Economic Review,'' American Economic Review 81,
no. 5 (December 1991): 1041--1067.
69. Blank, op. cit.: 1051.
70. A correlation coefficient for variables X and Y is defined as:

Vxy
rxy =
VxxVyy

1 n
where Vxy = n ∑
(Xi - X )(Yi - Y ) , the covariance of X and Y,
i=1
1 n 1 n
Vxx = n ∑ (Xi - X )2 and Vyy = n ∑(Yi - Y )2 , are the variances of X and Y
i=1 i=1
respectively,
1 n 1 n
and X = n ∑Xi and Y = n ∑Yi are the means of X and Y respectively.
i=1 i=1

A correlation coefficient ranges from -1 to 1, where 1 indicates perfect positive correlation

between the two variables (i.e., when one variable increases in value, the other also increases), 0
indicates no correlation between the two variables, and -1 indicates perfect negative correlation
between the two variables (i.e., when one variable increases in value, the other decreases).
71. Erica L. Groshen, ``The Structure of the Female/Male Wage Differential: Is It Who You
Are, What You Do, or Where You Work?'' Journal of Human Resources 26, no. 3 (Summer
1991): 457--472.

6-46
 72. William J. Carrington and Kenneth R. Troske, ``Gender Segregation in Small Firms,''
Journal of Human Resources 30, no. 3 (November 1992): 503--533.
73. This model was first developed by Bergmann, loc. cit.
74. Cf. David A. Macpherson and Barry T. Hirsch, ``Wages and Gender Composition: Why
Do Women's Jobs Pay Less?'' Journal of Labor Economics 13, no. 3 (July 1995): 426--471, who
control for a large number of skill-related occupational characteristics and are able to reduce the
measured effect of gender composition on earnings substantially, but not completely. Notably,
Barry Gerhart and Nabil El Cheikh, ``Earnings and Percentage Female: A Longitudinal Study,''

Industrial Relations 30, no. 1 (Winter 1991): 62--78, find use of longitudinal data to estimate
earnings equations drives the percentage female penalty effect down substantially for women,
but does not reduce the penalty for men.
75. Jim Naughton, ``A New Debate Over Disparities in Coaches' Salaries,'' Chronicle of
Higher Education 43, no. 36 (May 16, 1997): A35--A36.
76. Naughton, op. cit.: A35.
77. Henry J. Aaron and Cameran M. Lougy, The Comparable Worth Controversy
(Washington, D.C.: Brookings, 1986).
78. Perry C. Beider, B. Douglas Bernheim, Victor R. Fuchs, and John B. Shoven,
``Comparable Worth in a General Equilibrium Model of the U.S. Economy,'' Research in Labor
Economics 9 (Greenwich, Conn.: JAI, 1988).
79. See Linda J. Waite and Sue E. Berryman, ``Occupational Desegregation in CETA
Programs,'' Sex Segregation in the Workplace: Trends, Explanations, Remedies, ed. Barbara F.
Reskin (Washington, D.C.: National Academy, 1984): 292--307; Irmtraud Seeborg, Michael C.
Seeborg, and Abera Zegeye, ``The Impact of Nontraditional Training on the Occupational
Attainment of Women,'' Journal of Human Resources 19, no. 4 (Fall 1984): 452--471.

6-47
 80. Arthur E. Blakemore and Stuart A. Low, ``Sex Differences in Occupational Selection:
The Case of College Majors,'' Review of Economics and Statistics 66, no. 1 (February 1984):
157--163.
81. Barbara R. Bergmann, In Defense of Affirmative Action (New York: BasicBooks, 1996):
8.
82. Bergmann, op. cit.: 53--54.
83. Bergmann, op. cit.: 44.
84. Morley Gunderson, ``Male-Female Wage Differentials and Policy Studies,'' Journal of

Economic Literature 27, no. 1 (March 1989): 62 (Table 3), summarizes studies.
85. Morris Goldstein and Robert S. Smith, ``The Estimated Impact of the Anti-discrimination
Program Aimed at Federal Contractors,'' Industrial and Labor Relations Review 29, no. 4 (July
1976): 523--543; James J. Heckman and Kenneth I. Wolpin, ``Does the Contract Compliance
Program Work? An Analysis of Chicago Data,'' Industrial and Labor Relations Review, 29, no. 4
(July 1976): 544--564.
86. Jonathan S. Leonard, ``Women and Affirmative Action,'' Journal of Economic
Perspectives 3, no. 1 (Winter 1989): 61--75.
87. Ibid.
88. Peter Griffin, ``The Impact of Affirmative Action on Labor Demand: A Test of Some
Implications of the Le Chatelier Principle,'' Review of Economics and Statistics 74, no. 2 (May
1992): 251--260.
89. Griffin, op. cit.: 259.

FURTHER READING

Blaxall, Martha, and Barbara B. Reagan (eds.) (1976). Women and the Workplace: The
Implications of Occupational Segregation. Chicago, Ill.: University of Chicago. Older book

6-48
 containing some classic articles on sex segregation, particularly of a more discursive
theoretical nature.

Brown, Clair, and Joseph A. Pechman (eds.) (1987). Gender in the Workplace. Washington,
D.C.: Brookings. Collection of scholarly papers on various gender issues, including part-time
work, and women and unionism.

Federal Glass Ceiling Commission (1995). Good for Business: Making Full Use of the Nation's
Human Capital: The Environmental Scan. Washington, D.C.: U.S. Government Printing

Office. Fact-finding report from the Glass Ceiling Commission, 1991--96, which conducted
studies and issued recommendations concerning the advancement of women and minorities,
particularly into management positions. Lists all commissioned reports in an appendix. All
the reports are available for downloading at the School of Industrial & Labor Relations,
Cornell University website (www.ilr.cornell.edu).

Jacobs, Jerry A. (1989). Revolving Doors: Sex Segregation and Women's Careers. Stanford,
Calif.: Stanford University. Study using national data of women's entrance and exit patterns
into occupations.

Michael, Robert T., Heidi I. Hartmann, and Brigid O'Farrell (eds.) (1989). Pay Equity: Empirical
Inquiries. Washington, D.C.: National Academy. Collection of scholarly papers on various
issues related to occupational segregation and its link with pay.

O'Kelly, Charlotte G., and Larry S. Carney (1986). Women and Men in Society: Cross-Cultural
Perspectives on Gender Stratification, Second Edition. Belmont, Calif.: Wadsworth. Very
useful textbook discussing gender roles in a wide variety of modern and premodern societies.

6-49
 Provides valuable world view of sex segregation patterns and gives a picture of how
anthropologists and sociologists approach the puzzle of sex segregation.

Reskin, Barbara F. (ed.) (1984). Sex Segregation in the Workplace: Trends, Explanations,
Remedies. Washington, D.C.: National Research Council. Committee report summarizing
findings from a major research initiative; clearly lays out issues and patterns.

Reskin, Barbara F., and Heidi I. Hartmann (eds.) (1986). Women's Work, Men's Work: Sex
Segregation on the Job. Washington, D.C.: National Research Council. Collected articles on

sex segregation that accompany and provide background for preceding committee report.

Reskin, Barbara F., and Patricia A. Roos (1990). Job Queues, Gender Queues: Explaining
Women's Inroads into Male Occupations. Philadelphia, Pa.: Temple University. Contains
many useful chapter-length case studies of particular occupations that have experienced
gender recomposition.

Roos, Patricia A. (1985). Gender and Work: A Comparative Analysis of Industrial Societies.
Albany, N.Y.: State University of New York. Useful empirical comparison of developed
countries, mainly European data, on a number of labor force measures, including percentage
female by occupation.

Scott, Alison MacEwen (ed.) (1994). Gender Segregation and Social Change: Men and Women
in Changing Labour Markets. Oxford, United Kingdom: Oxford University. Interesting set of
related studies of labor market patterns in six cities in the United Kingdom carried out by an
interdisciplinary research team.

6-50
Tomaskovic-Devey, Donald (1993). Gender & Racial Inequality at Work: The Sources &
Consequences of Job Segregation. Ithaca, N.Y.: ILR. Describes original survey of North
Carolina adults; forceful discussion of shortcomings of neoclassical economic theory and
advocacy of social closure and status explanations for segregation patterns.

Wilkinson, Carroll Wetzel (1991). Women Working in Nontraditional Fields: References and
Resources 1963--1988. Boston, Mass.: G.K. Hall & Co. Bibliography for works, mainly
historical and case study, on women in male-dominated occupations.

Williams, Christine (1989). Gender Differences at Work: Women and Men in Nontraditional
Occupations. Berkeley, Calif.: University of California. Examines empirical data on
nontraditional occupations; no particular theoretical underpinning.

DISCUSSION QUESTIONS

1. Consider a job that you currently hold or held in the past. Describe the segregation patterns
that you found within the firm, the occupation, and the industry. What do you think the
causes of these patterns were?

2. Table 6.1 shows that the percentage of bachelor degrees awarded to women in computer
science and economics has declined since 1984. What might have caused these declines?
Why might the two fields diverge in their pattern when it comes to master's degrees (still
rising in economics, falling in computer science)?

3. Consider the Focus on directors and officers at Fortune 500 companies. Do you think that
company policies would change if a higher proportion of these persons were women?

6-51
4. Consider two countries, A and B. In each country there are only two occupations: white-collar
and blue-collar. White-collar jobs pay twice as much as blue-collar jobs. In Country A, 30
percent of the working women and 70 percent of the working men are in white-collar jobs. In
Country B, 35 percent of the working women and 90 percent of the working men are in
white-collar jobs.

(a) What is the Duncan index for each country?

(b) What is the female-male earnings ratio for each country (assuming that everyone works
the same number of hours)?

(c) In your opinion, in which country are women better off?

5. In the country of X, men and women can work in any of three occupations (farmer, cook, or
hunter). The distribution of men and women across the occupations and their wage rates are
listed below:

Distribution (%) Wages ($)

Occupation Women Men Women Men
Farmer 40 40 10 8
Cooker 50 20 15 10
Hunter 10 40 20 12

(a) Compute the Duncan index.

(b) What is the average wage of women? Of men? What is the ratio? What is the gap?
(c) What would women earn if they had men's jobs (i.e., if their distribution across the
occupations was the same as men's)? Now what is the ratio? In this case, how much of the
gap is explained by occupational differences?

6-52
 (d) What would men earn if they had women's jobs? Now what is the ratio? In this case, how
much of the gap is explained by occupational differences?

6. Consider the Focus on blind selection processes. Why might the study of refereeing have
found less of a gender effect than the study of orchestral auditions? Can you think of other
selection processes that could be turned into gender-blind selection processes? In your
example, would it likely increase the proportion of women selected?

7. Using the data in Table 6.9, discuss which activities are likely to contain a high degree of

sexual segregation by subactivity and which activities are likely to also involve physical
separation for the sexes during the workday.

8. Some of the occupations in Table 6.10 appear to be ``outliers;'' in other words, they do not
conform well to the overall pattern. For example, computer programmers are predominantly
male, yet women are paid very nearly equally with men in this occupation. Why might this
be the case? Are there other apparent outliers?

9. If an occupation is completely segregated, could it appear in Table 6.10? Why or why not? Is
this a problem in studying the link between segregation and earnings?

10. Consider the Focus on coaching salaries. Give some potential explanations for why men's
team coaches are paid more. If salaries are raised for women's team coaches relative to men's
team coaches, what might happen to the percentage of women coaching each type of team?

11. Draw diagrams similar in style to Figure 6.1 to show what would happen to the labor markets
for nurses and doctors if women are at first barred from attending medical school and then
the bar is dropped.

6-53
12. Which explanations of society-wide sex segregation do you find least plausible? Which do
you find most plausible? Can you think of particular areas of the U.S. workscene that appear
to lend evidence for each explanation?

13. Consider the effects on segregation of expanding the use of the employment policies of
provision of on-site child care and maternity leave and availability of flex-time. In what ways
would the increased use of these policies tend to increase or decrease sex segregation of the
labor force?

14. Consider Table 6.12: can you think of reasons other than affirmative action for the rise of
female employment in the covered sector from 1974 to 1980?

15. A company decides to expand, so it opens a factory, creating 455 new openings.i For the 70
white collar positions, 200 men and 200 women applied, of which 30 men and 40 women
were hired. For the 385 blue collar positions, 400 men and 100 women applied, of which 300
men and 85 women were hired. A federal Equal Employment enforcement official notes that
many more men were hired than women and decides to investigate. The company president
denies discrimination, arguing that in both the white collar and blue collar fields, the
percentage of female applicants hired was greater than the percentage of male applicants
hired. But the government official argues that a woman applying for a position at the firm
was less likely to be hired than a man applying for a position at the firm. As the current law
is written, if women have a greater chance of being denied employment, then this is a
violation.

iThis puzzle was posed by a correspondent in the ``Ask Marilyn'' column, Parade (April 28,
1996): 6. It is an example of what is referred to in probability and statistics texts as Simpson's
Paradox

6-54
(a) Calculate the relevant percentages to check each side's assertions.
(b) Can both sides be right? Explain what causes this apparent paradox.

6-55
Table 6.1
Percentage of bachelor's and master's degrees awarded to women, selected fields, 1974, 1984,
1994, and 2002
Bachelor's Master's
Field 1974 1984 1994 2002 1974 1984 1994 2002
Accounting 13.7 47.0 55.1 61.8 10.5 36.6 45.7 62.1
Architecture 6.6 23.5 29.6 35.6 10.1 25.8 32.3 36.6
Business management 9.8 41.5 47.2 50.5 3.2 29.5 35.2 38.9
Computer science 16.4 37.1 28.4 27.6 12.9 29.3 25.8 33.3

Economics 14.6 34.1 29.5 34.0 14.2 23.5 32.7 39.5

Education 73.5 76.0 77.3 77.4 59.9 72.0 76.7 76.4
Engineering 1.6 14.1 16.4 20.7 2.3 10.4 15.4 21.2
Journalism 45.7 62.5 63.6 68.0 38.0 56.1 63.9 70.2
Pharmacy 25.9 49.4 63.0 62.6 18.3 35.4 44.9 59.7

Source: U.S. National Center for Education Statistics, Digest of Education Statistics (1976)
(Table 112), (1987) (Table 152), (1996) (Table 244), (2003) (Table 255). Year refers to end of
academic year, e.g., 1972--73.

6-56
Table 6.2
Number of men and women directors and officers at Fortune 500 companies by industry, 1990
Directors Officers
Women Men %Female WomenMen %Female
Aerospace 8 213 3.6 7 272 2.5
Apparel 6 84 6.7 10 89 10.1
Chemicals 17 609 2.7 14 629 2.2
Computers 12 221 5.2 15 461 3.2
Electronics 23 434 5.0 7 564 1.2

Food 26 522 4.7 12 698 1.7

Industrial and farm
equipment 9 329 2.7 7 372 1.8
Metals 6 267 2.2 8 269 2.9
Pharmaceuticals 18 203 8.1 11 270 3.9
Publishing and printing 18 235 7.1 13 244 5.1
Soaps and cosmetics 17 84 16.8 8 185 4.1
Textiles 4 103 3.7 4 127 3.1
Transportation equipment 0 51 0.0 1 66 1.5

Total* 254 5384 4.5 175 6502 2.6

Source: Joann S. Lublin, ``Rights Law to Spur Shifts in Promotions,'' Wall Street Journal
(December 30, 1991): B1. Figures provided by Mary Ann Von Glinow, University of Southern
California. Reprinted by permission of Wall Street Journal, © 1991 Dow Jones & Company, Inc.
All rights reserved worldwide.
*For all Fortune 500 companies.

6-57
Table 6.3
Examples of calculating the Duncan dissimilarity index
Blue-collar Pink-collar Index values Outcome
Extreme cases:
|100-0| |0-100|
(a) 100% of all 100% of all D= 2 + 2 = 100 complete segregation

employed men employed women

|40-40| |60-60|
(b) 40% of men, 60% of men, D = 2 + 2 =0 complete integration

40% of women 60% of women

Intermediate cases:
|60-40| |40-60|
(c) 60% of men, 40% of men, D = 2 + 2 = 20 low segregation

40% of women 60% of women

|80-20| |20-80|
(d) 80% of men, 20% of men, D = 2 + 2 = 60 high segregation

20% of women 80% of women

6-58
Table 6.4
Occupational sex segregation indexes, all and by race, 1960 to 2000
1960 1970 1980 1990 2000
All 64 66 59 53 52
Whites 63 66 59 55 53
Nonwhites 70 64 56 50 50

Sources: Calculated by the author using data from Census of Population and Housing, 1960,
Public-Use Microdata 1/100 Sample; Census of Population and Housing, 1970, Public-Use

Microdata 1/100 B States Sample; Census of Population and Housing 1980, Public-Use
Microdata 1/100 C Sample; Census of Population and Housing 1990 and 2000, Public-Use
Microdata 1/100 Samples.

6-59
Table 6.5
Sex segregation indexes by level of disaggregation, 1990
Number of job categories: 6 16 83 503 36,669
Index value: 33 37 48 53 59

Source: Calculated by the author using data from the Census of Population and Housing 1990,
Public-Use Microdata 1/100 Sample.

6-60
Table 6.6
Occupational sex segregation indexes by age, education, and sector, 2000
Age group: 25--34 35--44 45--54 55--64
Index value: 52 54 55 57
Years of school: 0--8 9--11 12--15 16 >16
Index value: 54 55 57 43 40
Sector: Private Public
Index value: 52 49
Private/Public: Goods Services Federal State Local

Index value: 50 49 39 46 55

Source: Calculated by the author using data from the Census of Population and Housing 2000,
Public-Use Microdata 1/100 Sample.

6-61
Table 6.7
Occupational race segregation indexes by sex, 1960 to 2000
1960 1970 1980 1990 2000
Men 45 38 28 24 23
Women 50 36 26 22 20

Sources: Same as Table 6.4.

6-62
Table 6.8
Percentage distribution of societies by activity group segregation
Societies in which the activity is performed by:
Only or Both Only or No. of societies
Activity mostly men men and womenmostly women in sample
Hunting 100 0 0 738
Metal working 100 0 0 360
Boat building 96 3 1 215
Fishing 79 15 6 562

House building 75 10 15 457

Animal husbandry 64 22 14 412
Leather working 46 5 49 280
Weaving 30 12 58 265
Pottery making 9 5 86 328
Gathering 8 14 78 396
Agriculture 32 32 36 639

Source: Calculated by the author using data from George P. Murdock, Ethnographic Atlas
(Pittsburgh, Penn.: University of Pittsburgh, 1967).

6-63
Table 6.9
Number and percentage of societies in which a particular activity is strongly or weakly gender-
typed
Activity M M- = F- F Percent
Metal working 78 0 0 0 0 100.0
Weapon making 121 1 0 0 0 99.8
Pursuit of sea mammals 34 1 0 0 0 99.3
Hunting 166 13 0 0 0 98.2
Manufacture of musical instruments 45 2 0 0 1 96.9

Boat building 91 4 4 0 1 96.0

Mining and quarrying 35 1 1 0 1 95.4
Work in wood and bark 113 9 5 1 1 95.0
Work in stone 68 3 2 0 2 95.0
Trapping or catching of small animals 128 13 4 1 2 94.9
Work in bone, horn and shell 67 4 3 0 3 93.0
Lumbering 104 4 3 1 6 92.2
Fishing 98 34 19 3 4 85.6
Manufacture of ceremonial objects 37 1 13 0 1 85.2
Herding 38 8 4 0 5 83.6
House building 86 32 25 3 14 77.0
Clearing of land for agriculture 73 22 17 5 13 76.3
Net making 44 6 4 2 11 74.1
Trade 51 28 20 8 7 73.7
Dairy operations 17 4 3 1 13 57.8
Manufacture of ornaments 24 3 40 6 18 52.5
Agriculture -- soil preparation and planting 31 23 33 20 37 48.4
Manufacture of leather products 29 3 9 3 32 48.0

6-64
Body mutilation, e.g., tattooing 16 14 44 22 20 46.6
Erection and dismantling of shelter 14 2 5 6 22 39.8
Hide preparation 31 2 4 4 49 39.4
Tending of fowls and small animals 21 4 8 1 39 38.7
Agriculture -- crop tending and harvesting 10 15 35 39 44 33.9
Gathering of shellfish 9 4 8 7 25 33.5
Manufacture of non-textile fabrics 14 0 9 2 32 33.3
Fire making and tending 18 6 25 22 62 30.5
Burden bearing 12 6 33 20 57 29.9

Preparation of drinks and narcotics 20 1 13 8 57 29.5

Manufacture of thread and cordage 23 2 11 30 73 27.3
Basket making 25 3 10 6 82 24.4
Mat making 16 2 6 4 61 24.2
Weaving 19 2 2 6 67 23.9
Gathering of fruits, berries and nuts 12 3 15 13 63 23.6
Fuel gathering 22 1 10 19 89 23.0
Pottery making 13 2 6 8 77 18.4
Preservation of meat and fish 8 2 10 14 74 16.7
Manufacture and repair of clothing 12 3 8 9 95 16.1
Gathering of herbs, roots and seeds 8 1 11 7 74 15.8
Cooking 5 1 9 28 158 8.6
Water carrying 7 0 5 7 119 8.2
Grain grinding 2 4 5 13 114 7.8

Source: George P. Murdock, ``Comparative Data on the Division of Labor by Sex,'' Social
Forces 15, no. 4 (May 1937): 551--553.

6-65
Table 6.10
Percentage female, gender earnings ratio, and median weekly earnings for selected occupations,
2004
Occupation Female (%) Gender earnings ratio Median weekly earnings
Architects and engineers 13 0.77 1098
Security guards 22 0.89 457
Computer programmers 28 0.87 1118
Physicians and surgeons 31 0.52 1680
Lawyers 33 0.73 1561

Marketing and sales managers 39 0.62 1213

Postsecondary teachers 41 0.76 1034
Sales occupations 44 0.62 604
Editors 49 0.80 856
Waiters and waitresses 67 0.82 348
Elementary/middle teachers 80 0.85 806
Nursing aides 88 0.91 388
Registered nurses 92 0.87 904
Secretaries and admin assts 97 0.92 552

Source: Employment and Earnings 52, no. 1 (January 2005) (Table 39).

6-66
Table 6.11
Part-time and overtime work, employed persons, ages 25 to 64, 1990
% of occupation Usual hours under 30 (%) Usual hours over 40 (%)
that is female WomenMen WomenMen
0--24.9 12.4 3.8 24.6 38.8
25--49.9 12.7 3.8 22.9 43.9
50--74.9 18.1 5.6 16.1 36.2
75--100 20.5 7.8 11.9 26.9

Source: Calculated by the author using data from the Census of Population and Housing 1990,
Public-Use Microdata 1/100 Sample.

6-67
Table 6.12
Female employment share (percentage) by federal contract status, 1974 and 1980
White female Black female Other female
Contractor? Yes No Yes No Yes No
1980 28.8 40.0 4.5 5.9 2.8 3.6
1974 27.6 39.4 3.0 4.7 1.6 2.4
Change 1.2 0.6 1.5 1.2 1.2 1.2

Source: Jonathan S. Leonard, ``Women and Affirmative Action,'' Journal of Economic

Perspectives 3, no. 1 (Winter 1989): 63. Reprinted with permission.

6-68
Figure 6.1
Effects of crowding

(i) Before crowding occurs

Pink-collar Blue-collar
wage wage
S0 S0

W0

D D

E0 employment E0 employment

(ii) After crowding occurs

Pink-collar Blue-collar

wage wage S1
S0 S0
S1
W1
W0
W1
D D

E0 E employment E1 E0 employment
1

6-69