1

Thinking about Occupational Sex-Typing

One of the fundamental concerns in feminist social science has been specifying the relationship between capitalism and patriarchy. Since Heidi Hartmann’s seminal essay “Capitalism, Patriarchy, and Job Segregation by Sex” was written in 1976, it has generally been accepted that the oppression of women by men is facilitated and maintained by the economic institutions of modern society. However, specifying the mechanisms by which this occurs has not been easy; within the 1970’s and 1980’s, there has been a resurgence of interest in generating a materialist explanation of patriarchy that is consistent with rational economic theory. This new scholarship has taken a wide variety of forms. There have been attempts to develop a theory of patriarchy that closely parallels the form and logic of Marxist theories of exploitation (Secombe 1974; Beechey 1978; O’Brien 1981; Power 1983). Economists have used neo-classical economics, human capital theory, and, more recently, dual labor market theory to explain women’s concentration in low-paid employment and production within the home (Barron and Norris 1976; Blau and Jusenius 1976; Polachek 1979). There has also been excellent empirical research on the history of women’s experiences in the work world that has attempted to uncover inductively the obstacles impeding women’s advancement (Kanter 1977; Dublin 1979; Kessler-Harris 1982; Milkman 1982). Finally, there has been a flowering of demographic research and family history that has documented how larger economic changes affected women’s domestic roles, occupational lives, and self-perceptions. (M. Anderson 1971; Zaretsky 1973; Tilly and Scott 1978). Each of these styles of work makes its own distinctive contribution towards answering the question. There is one phenomenon, however, that has such an overwhelming effect on women’s status and opportunities that writers of nearly every theoretical persuasion have made at least tacit acknowledgment of its importance. This phenomenon is occupational sex-typing.

Occupational sex-typing is the process by which certain occupations are designated as being primarily male or female. Sex-typing has produced a virtual bifurcation of the labor market into male/female sectors; in 1960, fully 50 percent of all women were working in occupations that were over 70 percent female (Oppenheimer 1970). This division of labor has had enormous consequences for women’s economic opportunities and prospects. Valerie Kincaide Oppenheimer has compellingly demonstrated that occupational sex-typing has placed an upper limit on the extent to which women can participate in work outside the home. Between 1900 and 1960, the increase in the number of female sex-typed jobs far outweighed any supply-side variable in predicting the increase of female labor force participation (Oppenheimer 1970).

Women’s wages are also affected by occupational sex-typing. The most compelling explanation of women’s inferior wage attainments is the overcrowding hypothesis. This argument claims that sex-typing forces a large number of women into a small, restricted set of occupations, while men may select from a much wider range. The limited demand for female labor intensifies competition among women for jobs, thus lowering their overall market wage (Fawcett 1918; Edgeworth 1922; Stevenson 1975a, 19756). The combination of limited job opportunities and low wages also serves to promote women’s subordination in the domestic sphere. The dismal economic opportunities faced by women in the marketplace induce many of them to drop out of the labor force to become full-time homemakers. Furthermore, their limited financial opportunities increase their dependence on male providers, which in turn makes them vulnerable to domestic and sexual exploitation. Thus, if one is to understand changes in women’s work and family roles, it is essential to understand the process by which employers designate which jobs will be open to women and which jobs will be reserved for men. While occupational sex-typing may not be the only mechanism by which economic forces shape those of patriarchy, it surely is one of the most central and important.

Common Explanations of Occupational Sex-Typing

Thus far, however, no one has developed a clear and convincing theory of what makes a given job male or female. One of the major problems has been that discussions of women’s work have been dominated by a series of propositions that are for one reason or another unsatisfying. The major issue that confronts any theory of occupational sex-typing is that both economic and patriarchical considerations affect the hiring decisions made by managers. Managers are basically businessmen, whose rewards and life chances depend to a large extent on the growth and profitability of their enterprises. As such, hiring policy should be a tool for lowering costs, raising productivity, and furthering the organization’s economic goals. At the same time, the firm exists in a society in which patriarchy is pervasive; commonly held sex-role ideologies might be expected to affect the hiring decisions of managers who are being ostensibly objective. The intellectual challenge has been to develop an explanation of occupational sex-typing that incorporates both economic and ideologically based variables into a synthesis that is both theoretically coherent and consistent with the available evidence.

Most theories of women’s work do deal with both types of factors, but the strategies for combining them have often left something to be desired. It has often been argued that patriarchy affects hiring decisions through one of two following intermediating mechanisms: restricting female labor supply or creating normative and cultural limits on female employment. Let us consider each of these in turn.

Restricting Female Labor Supply

Many models of occupational sex-typing argue that women voluntarily choose to limit their labor supply in order to spend their time as full-time mothers and housewives rather than as paid employees. These models are used by authors of many different intellectual persuasions, but they have been most systematically developed by human capital theorists (Becker 1964; Zellner 1975; Polachek 1979; Rotella 1981).¹ Human capital theorists start with the reasonable assumption that both employers and potential employees are economically rational. Employers do not discriminate, but instead fill their job openings in such a way as to maximize productivity and minimize cost.² Decisions about occupational sex-typing are based on employers’ choosing between two types of workers: men, who are committed to long-term employment, and women, whose labor force participation is likely to be intermittent. There are two reasons why women supposedly choose to withhold themselves from the labor force. On the one hand, traditional gender ideologies encourage women to stay home, thus giving them a greater taste for leisure. On the other hand, women are hypothesized as having a natural aptitude for home production and as having a comparative advantage in child care, cooking, and other domestic tasks. Thus, it will be rational for women to concentrate in home rather than market production, particularly in those stages of the family life cycle when there is a high demand for child care.

Human capital theorists argue that women’s domestic obligations result in their being excluded from high-status jobs. High-status jobs tend to be relatively skilled and require substantial amounts of both formal education and learning on the job. Because women anticipate interrupting their careers, they will choose neither to spend money on formal education nor to select occupations that require extensive training in the early stages of the career.³ Thus non-discriminatory employers of skilled labor will find relatively few qualified female workers applying for their vacancies, and they will be forced by default to give these skilled jobs to men. Furthermore, even if a supply of female candidates exists, employers will be reluctant to give these positions to women. There are many situations in which the employers themselves finance the cost of educating workers for high-level jobs. In these cases, employers will choose to hire workers with long anticipated tenures so they can minimize the number of times they have to provide basic training. If they hire short-term workers, such as women, there is a risk that the workers will leave employment before producing enough to cover the cost of their human capital.

There are several possible objections to this account. First, it is not at all clear that women’s domestic obligations keep them from wanting to obtain good jobs; in general the obstacles to women’s entering high-status jobs have come from management rather than from women themselves. Second, the evidence does not support the claim that, controlling for occupation, women have higher turnover than men. Women seem as likely as men to remain in responsible positions. Third, even if women voluntarily choose to have discontinuous careers, it does not follow from this that employers would choose to bar women from jobs with high firm-specific training. There are countervailing advantages to hiring women that can outweigh the costs associated with human capital loss. Let us consider each of these arguments in further detail.

The proposition that female labor supply has been limited by domestic considerations may have some validity for the early nineteenth century, but it is indefensible for the modern period. Valerie Kincaide Oppenheimer’s work (1970) has compellingly shown that supply-side theories do not convincingly account for changes in American female labor force participation from 1900 to 1960. Women’s work rates during this period did not correlate in any meaningful fashion with changes in fertility, the availability of labor-saving home appliances, or changes in national attitudes on traditional sex roles. Female labor force participation rose dramatically during the 1950’s, despite the countervailing forces of the baby boom and a generally conservative social climate. In the fifties, women’s economic opportunities increased, as employment in clerical work, education, and the female semi-professions expanded. Women responded to these opportunities and left their homes for paid employment even as the baby boom added to their domestic obligations (Oppenheimer 1970). Cynthia Epstein’s research on law shows that even women who had relatively conservative attitudes on sex roles entered law school when law schools changed their admissions policies and began to accept women. Despite traditional socializations, many women were capable of making the ideological adjustments necessary to prepare themselves for professional careers (Epstein 1981). Both these studies suggest that women are willing and eager to enter market employment and that the primary barriers to female labor force participation come from management.

Furthermore, it is not at all clear that women have desisted from obtaining human capital. Historically, women have been more likely than men to seek higher education. Between 1870 and 1930, between 55 and 60 percent of all high school graduates were women (Rotella 1981). If women invested in long-term job skills in the late Victorian era when traditional sex roles were quite prevalent, it is implausible that modern women would be unwilling to make such commitments when the potential economic returns to such commitment have been substantially improved.

There are similar difficulties with the claim that women are disproportionately likely to quit their jobs. This proposition has received only lukewarm support from the empirical literature on turnover and exit rates. Some studies do find significant differences between men and women in career lengths (Parnes 1954; Burton and Parker 1969; Pencavel 1970); however, far more studies have found no sex differences whatsoever (Stoikov and Raimon 1968; Fry 1973; Armknecht and Early 1972; Begnoche Smith 1979; Blau and Kahn 1981). The size of the relationship is not robust to minor changes in the control variables or to changes in the strategies of measurement. An example of the complexities involved in the relationship between sex and turnover can be found in an analysis of quit rates by K. P. Viscusi. Viscusi found that global rates of turnover for female employees were higher than the equivalent rates for men. However, there were no significant sex differences in tenure-specific exit rates; furthermore, women were far more likely to decrease their quit rates in response to organizational incentives than were men (Viscusi 1980). The evidence does not support the conclusion that there are no sex differences in turnover whatsoever. Given the lack of consistent findings on the very existence of a turnover differential, however, let alone any consensus that the size of this turnover differential is large, it would seem inappropriate to base the explanation of such a pervasive institution as sex-typing on such an ephemeral phenomenon.^4,5

Even if women do systematically interrupt their careers to get married and raise a family, there are still critical logical problems with most of these supply-side theories. It is not at all clear that the most rational economic response by either women or employers to short anticipated careers is the elimination of women from high-status jobs. Consider this problem from the point of view of a female worker first. The primary reason a woman supposedly chooses a low-status position is that the job allows her to minimize her loss of human capital when her career is interrupted. The gains that come from the preservation of deteriorating skills would seem to be trivial compared with the overall benefits involved in taking a high-paying rather than a low-paying job. Furthermore, once a person has received his or her basic training in an occupation, skills that have been lost through short- or middle-term absences can often be regained fairly quickly. Adult students who return to school after several years in the job market usually regain their academic skills within the first year. Job skills are likely to have a similar rate of regeneration. And even if the rate of skill loss were catastrophic, the wages received during retraining in a high-paying job are likely to exceed the wages a fully trained worker receives in a low-paying job. Thus, it is very difficult to explain the concentration of women in low-status jobs in terms of a model of voluntary choice.⁶

The argument is equally hard to justify from the employer’s point of view. It is not clear that the benefits of conserving scarce human capital outweigh the costs of operating with a relatively inexpensive labor force. A woman will work at the rate of 59 cents for every male dollar. Even if the use of women requires higher levels of training costs due to more frequent replacement of personnel, retraining would have to be very expensive to exhaust the 40 percent wage subsidy associated with using cheap female labor. Admittedly, there are personnel who are so scarce as to be irreplaceable; the Manhattan Project would have paid anything not to lose Enrico Fermi. However, the skills of most workers are not so rare as to require conservation at any price. Many skilled workers are readily replaceable from the marketplace. If one loses a high school principal, one merely hires a new one from another system. Other vacancies can be easily filled from within. Most advanced positions are filled not by neophytes but by subordinates in the same workplace who have some relevant experience and need only a partial retraining. It is not implausible to imagine that, when an assistant is promoted to a position formerly held by her supervisor, she might work at 60 percent efficiency the first year, 80 percent efficiency the second year, and full efficiency for all years thereafter. From the point of view of the employer, this would represent substantial training costs, but hardly an overwhelming burden. In particular, it would hardly justify turning down a female candidate who was willing to work for 59 percent of a male salary. The benefits from exploiting women’s inferior position in the marketplace would seem to far outweigh the benefits that accrue from the conservation of human capital.⁷

Even if one grants that women are likely to leave their jobs because of family responsibilities, however, and that the costs of retraining these workers is greater than the savings of using cheaper labor, it is still logically difficult to argue that such problems will induce employers to bar women from high-status jobs. This is because only younger women are actually subject to the kinds of domestic pressures that could force withdrawal from the labor force. Older women can readily be hired for any job involving firm-specific learning and will have low rates of turnover, similar to those of men. The one domestic duty that might require withdrawal from the labor force is taking care of infants and children under five. Cooking, cleaning, and laundry may all be deferred or obtained through the market, but small children require constant attention and nurturance, which is not easily provided by market labor. Nevertheless, only women in their twenties and thirties are likely to have young children in their households. Because fertility tends to be concentrated in the years just after marriage, by age forty most women have completed their childbearing and most of their children are old enough to free their mothers for work outside the home. Furthermore, because pre-industrial mortality schedules closely resemble modern mortality rates for ages after twenty, historically a forty-year-old woman could have looked forward to good health until her late sixties or seventies. She thus could have entered the labor force anticipating an uninterrupted career of twenty-five to thirty years. Most jobs could be filled quite profitably by twenty-five or thirty-year workers, even if they required extremely long periods of initial training.⁸

Occupational sex-typing divides jobs into subsets that are either all male or all female; there are very few jobs that are given to men and older women, but not to younger women. If employers were primarily concerned about conserving scarce human capital, such patterns would occur relatively frequently as employers differentiated between long-term and short-term employees. In fact, employers have hired girls in their teens and early twenties quite readily, while discriminating against older women; the labor force participation rates of teen-aged women have always been high, while those of women over forty-five only increased in the 1950’s (Oppenheimer 1970). Conservation of human capital does not predict a division of labor that lumps all men into one category and all women into another. The phenomenon of occupational sex-typing that treats all women as mutually interchangeable, though different from men, must be explained by a wholly different dynamic.

Labor supply can also affect women’s work through mechanisms that do not involve human capital. It has been hypothesized that sex-typing is affected by prevailing rates of female labor force participation at the time at which a given occupation is created. If the initial job openings occur in a period when traditional sex-role ideologies are inhibiting female labor force participation, no women candidates will apply for the new vacancies and the occupation will become male. If the job is created in a period of increasing female labor force participation, it will be possible to staff the positions with women. In either case, once these first workers are hired, bureaucratic inertia and the development of self-reinforcing legitimating ideologies tend to perpetuate the sex mix of the original cohort.⁹

This model rarely appears in explicit theories of occupational sex-typing, but it is extremely common in empirical studies of particular occupations. Consider the standard claim that clerical work feminized because office work expanded in an era of rapidly increasing female education and labor force participation. These changes in women’s work behavior ensured that a supply of women would be available to take these new positions (Davies 1975, 1982; Kessler-Harris 1982). Such an insight would not be terribly informative if occupations adapted instantaneously to changes in the labor supply. If this were the case, all types of jobs would have responded to these changes and there would be no such thing as occupational sex-typing. Increasing labor supply becomes an internally consistent theory of occupational sex-typing only if older occupations such as mining do not respond to changes in the labor market and thus remain predominantly male.

Although there are examples that can be used to support the theory, the number of counterexamples is quite striking. The feminization of teaching, clerical work, retail sales, social work, and nursing have been explained in terms of the increasing female labor supply of the late nineteenth and early twentieth centuries. Between 1870 and 1930, however, employers also first began to hire accountants, life insurance salesmen, airline pilots, truck drivers, electricians, elevator operators, advertising consultants, radio announcers, and automobile mechanics. All of these jobs became overwhelmingly male. The expanding turn-of-the-century supply of female labor does not seem to have had the same effect on every occupation that was created during the period. In fact, supply considerations seem to have very poor predictive power in explaining what types of jobs become male or female. In order to understand how new occupations develop their sex types, it is necessary to understand the demand-side determinants of female employment.

Normative and Cultural Limits on Female Employment

Many discussions of women’s work explicitly argue that employers’ hiring decisions are fundamentally shaped by occupational norms derived from the division of labor within the home. Because of this, women are given jobs associated with traditional female tasks or attributes (Caplow 1954; National Manpower Council 1957; Oppenheimer 1970). The training women receive in the household as young girls gives them a natural aptitude for jobs with a domestic component. Likewise, women’s personalities, which are shaped by early gender socialization, makes them more suitable for work that requires “female” psychological attributes. Thus, women work in textile factories because as young girls they receive training in sewing. Women work as waitresses because at home they are more likely than men to serve meals. Women work in the helping professions because they are socialized to be expressive rather than instrumental. Just as women are concentrated in those areas where they have special expertise, men are concentrated in those areas where they have a particular gender-related advantage. Thus men are more likely to work in jobs that require physical strength, are dirty, or demand leadership skills.

Not every writer would insist that men and women differ in their gender-related competencies. Whether women are truly weaker or more understanding than men is a matter of considerable dispute. However, cultural theories of sex-typing need not be based on actual differences in the skills or personalities of men and women. It is sufficient for employers to believe that these differences exist and to hire on the basis of this consistent gender ideology. In their most credible form, normative theories predict the sex-type of occupations in terms of commonly held ideologies that blind employers to the possibility of extending the domain of female employment. Traditional gender roles irrationally inhibit firms from hiring women for jobs characterized as masculine, making the occupational structure reflect patterns of patriarchy in the larger society.

Although these considerations can provide seemingly plausible explanations of particular cases, they nearly always fail to provide a general principle that will correctly predict the overall distribution of male and female jobs. It is very hard to find a norm or a cultural prescription that seems to have been applied consistently by employers in a wide range of occupations. Consider the argument that women take jobs similar to those tasks done within the home. Although women traditionally cook within the home, they have not always obtained such jobs in the marketplace. In 1970, the occupation of cook was roughly evenly split between men and women: just under 40 percent of the cooks were men and just over 60 percent were female.¹⁰ Despite the fact that baking bread and cakes is done by women in the home, only 29 percent of all commercial bakers and 35 percent of bakers’ assistants were female. Only 35 percent of food manufacturing operatives were women. Within this set of occupations, the only occupation that was more than 50 percent female was the canning and preserving of fruits, vegetables, and seafood. It is hard to see how the traditional division of labor in the home would lead women in the labor force to can salmon but not to make butter or freeze TV dinners.

Such incongruities can also be found in work resembling domestic labor. Women invariably do the housecleaning in most families. Nevertheless, only 32 percent of all cleaning workers were women. Likewise, although women made up 69 percent of all food service workers, they comprised only 38 percent of all dishwashers. Although women have been traditionally associated with child care, and 84 percent of all elementary school teachers were women, only 49 percent of all secondary school teachers were women. In the home, women’s responsibility for their children hardly ends at age twelve. In the professions, women do tend to be concentrated in specialties that emphasize child care. In medicine, women are disproportionately likely to be pediatricians, and within mental health, women are more likely to be child psychologists. Nevertheless, none of these jobs have ever been female sex-typed occupations. Women at best have made up a somewhat larger minority than was to be found in other medical specialties.

The same incongruities appear in textiles. The fact that women sew has been used to explain women’s concentration in spinning mills and clothing factories. Nevertheless, only 32 percent of all commercial tailors are women. This is surprising, given that women make the clothing repairs in most households. Although many female jobs require skills similar to those women use within the home, there is a very substantial number of jobs involving domestic arts that traditionally have been monopolized and dominated by men. Thus, the household division of labor actually turns out to be an unreliable guide to predicting occupational sex type.

It is also difficult to find social-psychological correlates of male or female jobs. Although women are traditionally associated with nurturance, they have not always been used in jobs that involve care or sensitivity in dealing with people. Nursing and social work have been traditionally female jobs. However, psychiatry has remained a male occupation. Furthermore, most religions have traditionally provided only limited opportunities for women to act as clergy, despite the fact that this occupation specializes in the most profound emotional and spiritual concerns of human life. It is true that some religions have made use of women in specialized capacities. The Catholic Church has always permitted spiritually inclined women to serve as nuns, but nuns have been confined to a fairly narrow range of duties and rarely heard confessions or provided pastoral care.

It is also difficult to find any physical correlates of men’s or women’s work. Women’s supposedly superior hand-to-eye coordination may be invoked in explaining their concentration in textiles and electronics manufacture, but these skills have never gotten them hired as jewelers, engravers, or surgeons. It is true that women have not tended to appear in occupations that require physical strength, such as baggage porters or freight and material handlers. However, they have also been excluded from manual occupations that involve very little physical exertion at all. Women made up only 3 percent of parking attendants, 11 percent of vehicle washers, and 6 percent of taxi drivers. None of these jobs are particularly physically demanding. Women made up only 2.9 percent of all gardeners and groundskeepers despite the fact that many women routinely garden at home. Since most occupations do not call for very much upper-body strength, it is unclear how much of the difference between men’s and women’s jobs can be accounted for by physical strength.

Although most cultural propositions can be contradicted with numerous examples, one can never prove definitively that ideological factors do not contribute to sex-types. There is no limit to the number of propositions that could be devised by a dedicated believer in sex roles. Furthermore, a cultural determinist could argue that the individual normative factors only work when considered in concert and that an ideological model must contain an enormous number of variables before it can obtain predictive power. Nevertheless, deriving a sex-role socialization model that would be measurable, non-tautological, and consistent with the complexity of most occupational systems is a formidable prospect. The task is made more discouraging by the poor performance most of the components of such a model have when considered individually.

A Framework for Thinking about Occupational Sex-Typing

It is necessary to develop an alternative to supply-side and normative theories that can explain the creation and persistence of occupational sex-typing. Such a theory should be capable of making concrete predictions of the sex-type of a given job, consistent with an assumption of economic rationality on the part of the employer and informed by the actual problems and concerns faced by real-world managers. There are many propositions that at first glance appear to satisfy these conditions. Some after closer examination can be shown to be flawed; others make important contributions to the overall theory. In order to organize the various alternatives to supply-side and normative models, it is helpful to place these models within a larger theoretical framework.

Most theories of occupational sex-typing attempt to specify what is distinctive or unique about women’s work. Human capital theory argues that male workers are acceptable for all jobs, while female workers must be confined to a more limited subset. The theory then attempts to specify the properties of this female subset. Normative theories could concentrate on the attributes of jobs of either sex. In practice, however, they have tended to focus on women’s domestic roles and personality and how these shape the universe of jobs that become defined as female. While there is nothing incorrect about proceeding in this fashion, it is more informative to proceed in the opposite fashion and ask what it is that is distinctive about those jobs that are defined as male.

Employers should prefer to hire women rather than men, all other things being equal, because women work for substantially lower wages. At the global level, this phenomenon is caused by occupational sex-typing and by the overcrowding of the labor market that results from discrimination. For an individual firm, however, national patterns of discrimination are a given. An employer deciding how to fill a particular vacancy knows that, because of the hiring policies of other employers, women can be forced to accept a lower level of wages; this should make women the employees of preference for all positions until occupational sex-typing completely disappears.

This argument was originally made in 1957 by Gary Becker in his Economics of Discrimination. Becker argued that any firm that chose to discriminate was arbitrarily restricting its labor supply and forcing itself to operate with a higher level of market wages. This economic handicap would restrict the firm’s ability to compete effectively and could ultimately lead to the closing or sale of the firm. Becker thus concluded that discrimination was economically irrational; if prejudice persisted, it was because either the employer or the workforce had a “taste for discrimination.” There are many grounds on which this argument can be challenged, such as its overreliance on the principles of neo-classical economics and its strong prediction that in the long run discrimination would disappear. Nevertheless, Becker’s main argument that employers pay a premium to employ men is basically sound. The important theoretical task is to move beyond the vague term “taste for discrimination” in order to specify the economic basis for not making use of cheap female labor.¹¹

Logically, there are four different categories of reasons for employers not to hire women. No one particular category has any intrinsic logical advantage over any other. Although each type of argument has the capacity to be the basis of a reasonable, internally consistent theory of occupational sex-typing, individual arguments within each category will vary in their logical adequacy and empirical support. The four different strategies for explaining the non-hiring of women are:

1. Identifying a countervailing economic advantage to using men.

2. Identifying some environmental constraint that prevents employers from hiring women.

3. Identifying some sector of the male population that will work for lower wages than the average salary demanded by women.

4. Identifying some factor that buffers employers from the requirement of minimizing labor costs.

Let us consider each of these in turn.

Countervailing Advantages

Employers could hire men because men are associated with countervailing advantages that can compensate for their relatively higher pay rates. Both human capital theory and cultural theory argue that men are inherently more productive than women in performing certain kinds of tasks, which justifies paying a premium to obtain male labor. These two specifications are the most common forms of countervailing advantage arguments; however, there are a variety of alternative specifications. Consider Michael Reich’s theory of the racial division of labor markets. He would argue that by dividing the labor force into two hostile groups and carefully rationing employment between the two sectors, employers can lower the level of working-class militancy (Reich 1981).¹² Such a strategy requires the hiring of both white adult males and some other category of workers.

A model will be presented in the analysis called the “synthetic turnover hypothesis” that links occupational sex-typing to motivation problems in internal labor markets. Under a narrow set of conditions, employers within a bureaucracy can find themselves forced to reduce the levels of rewards to some workers within the system arbitrarily, while maintaining these rewards for the rest of their labor force. Under such conditions, it is necessary to create a caste system within the organization in which differential placement and treatment are legitimated in terms of pre-existing social and demographic distinctions. Under such conditions, it is necessary to hire both a set of victims, such as women or blacks, and a set of elite members, namely white adult males. These conditions do not apply to all employers; in those settings where synthetic turnover operates, however, men are hired because of the countervailing advantage they provide in maintaining segmented reward systems within the firm.

Environmental Constraints

Even if employers prefer to hire women for a given position, they may find themselves prevented from using women and forced to use men. Labor supply arguments make use of such environmental constraints. In these models, employers wish to hire women because of their lower wages; however, because women are choosing domestic production over market employment, female job candidates are scarce, and employers are forced to use men as a more expensive alternative. Two other environmental constraints that are often discussed in theories of occupational sex-typing are exclusionary demands from organized labor and legislative limitations on female employment.

The exclusion of women from certain occupations can be caused by a strong, well-organized, sexist male union. Gary Becker argued that discrimination can come from two sources: employers and employees (Becker 1957). Employee discrimination is actually easier to account for than employer discrimination. The traditional worker strategy for increasing wages is to limit the supply of potential workers. Thus, organized labor has attempted whenever possible to enforce extremely restrictive limits on who may obtain employment, a policy that has encouraged racism, sexism, ethnic prejudice, and the development of barriers to exclude some categories of white men. The hostility of male workers to women has been fueled by patriarchical concerns as well. Thus, it is not surprising that historically there are many examples of male unions’ actively agitating for the formal restriction or abolition of female employment (Hartmann 1976; Kessler-Harris 1979; Foner 1979, 1980; Milkman 1980).

While organized labor represents an important source of exclusionary pressure, its importance in determining sex-type is limited. Not all unions have the strength to impose a male workforce on an employer who is committed to resisting these demands. In most union-management confrontations, management has bargained from a position of strength. Union security is a very recent concession for many workers; obtaining economic benefits has often been problematic. It has been extremely difficult to win concessions on issues not involving wages, such as limits on the rate of work or protection from redundancy. Most unions have very little control over hiring decisions and only limited control over promotions and job assignments. Without such power, unions can contribute little to the process of occupational sex-typing, and gender decisions remain a matter of managerial prerogative.

There are, however, some exceptions to this. Workers in very strong positions, notably craft workers, professionals, and managers, have historically maintained tight control over recruitment into their occupations. The exclusionary powers of doctors, lawyers, and plumbers is well known. Thus, the concentration of men in the most skilled or powerful occupations can be plausibly explained in terms of the environmental constraint of male organized labor, as long as one compares average jobs with those of the very elite.

Many employers have been unable to use women because hiring them for a particular job was illegal. In the nineteenth century, women were barred from many manual occupations by protective legislation. In some cases, these laws limited the number of hours and times of the day that women could work when long night shifts were a substantive requirement for employment. In other cases, women were forbidden from workplaces that were legally defined as dangerous (Creighton 1979; Kessler-Harris 1982). The prevalence of these laws can be exaggerated. Most occupations were wholly unregulated by the state, and even within the most heavily regulated industries, women were barred from employment in only a limited number of localities (Brenner and Ramas 1984). Nevertheless, protectionist legislation provides an obvious environmental restriction on female employment that should have a measurable effect on occupational sex-typing.

Alternative Sources of Secondary Labor

The discussion thus far has assumed that there are only two kinds of workers, cheap women and expensive men. Thus, the theory of sex-typing devolves into a theory of what jobs can be filled with cheap labor. However, women are not the only source of secondary labor available to employers. An employer who is seeking inexpensive labor can hire blacks, hispanics, or local teenagers, all of whom work for less than the average white male. Some white males will also work for a marginal wage, particularly if they are unqualified, desperate, or looking for very temporary employment. Given that women do not have a monopoly on the market for cheap labor, it is necessary to specify the conditions under which secondary jobs will be given to women as opposed to members of other groups whose wages are depressed.

There are several factors that can determine whether women will be used to fill a secondary position. One consideration is the human capital requirements for the job. Some secondary jobs require few skills and little education. Others, despite their low pay, require substantial amounts of human capital. Virtually anyone can qualify to be janitor; nursing requires years of formal study and practical experience. Many workers would be unlikely to have the qualifications for the more demanding secondary positions. Juveniles can not take any position that requires advanced education; recent immigrants can not take a job that requires fluency in the host language. For many secondary jobs that require substantial human capital, women may be the only source of cheap labor having the necessary skills.

A second consideration of equal importance is the relative supply of alternative sources of labor. Many desirable sources of inexpensive labor are not available to particular employers. There are legal restrictions on certain kinds of workers. Child labor laws and school attendance requirements remove certain juveniles from the labor market entirely. Immigration policy restricts the minority labor supply. The juvenile labor supply will be affected by the availability and importance of schooling. Where higher education is open to the general public, many teenagers will voluntarily delay labor market entry in order to attain a degree. There may be regional strictures on the supply of labor as well. Very few communities aside from frontier communities have a shortage of women; however, some areas are ethnically homogenous and lack minority labor. In each of these cases, an employer may be limited in his options for filling the secondary positions within his firm. Women should be hired in those situations where alternative sources of labor are scarce.

There may be other considerations that affect the trade-off between competing secondary labor forces. The wages for these different groups are unequal. Juveniles make less than women, while the relative cost of ethnic labor depends on the amount of discrimination suffered by any particular group. The least expensive workers should be relatively more attractive to employers. Likewise, there may be differential psychic costs to employers in hiring women as opposed to hiring minorities. The theory of the competition between women and other disadvantaged groups is far less developed than the theory of competition between majority males and majority females. Nevertheless, both are important to a fully predictive theory of sex-typing.

Buffering from Labor Costs

Men are more likely to be hired into occupations in which employers are under little pressure to economize on labor costs. The prime consideration that makes women more desirable than men as employees is that women work for substantially lower wages. If wages cease to be a significant concern for management, then hiring can be based on other principles that can be either gender-blind or overtly discriminatory. Either way, fewer women will be likely to be hired under these conditions than would have been the case had the firm been trying to minimize labor costs.

There are several conditions under which employers can be released from the obligation of having to seek the cheapest possible labor. Most of these involve situations in which labor costs are relatively unimportant in determining organizational survival or rates of organizational growth. Consider the case in which labor makes up a relatively small proportion of costs within the firm. In labor-intense industries, staff expenses make up the overwhelming majority of operating costs. Employers in such firms must work aggressively to contain levels of salaries and wages or risk seriously overrunning their total budget. This problem is far less pressing for capital-intense employers. Although they may have total wage bills that are absolutely high, these expenses are relatively small compared to expenditures on raw material and technology. Technological increases in productivity or savings from obtaining cheaper sources of raw materials can all subsidize a relatively expensive staff.

The same argument holds for isolated subgroups of workers within a larger population of workers who are being managed efficiently. If an employer obtained 95 percent of his staff at the cheapest possible rates, he could afford to hire somewhat more expensive workers for the remaining 5 percent. Thus, if one could identify a set of positions that for exogenous reasons an employer might prefer to staff uneconomical^ with men, such positions might remain male so long as they represented such a small share of total staff that their cost would not significantly raise the total wage bill. Most firms do not control remuneration to executives with the same rigor that they control the other aspects of corporate expenditure. High salaries, generous expense accounts, and lavish fringe benefits have been justified in terms of the high productivity of top administrators. To some extent, however, the extremely generous executive remuneration found in most corporations stems from managers’ choosing to reward themselves when they know that such payments will have little effect on the overall profitability of the firm.

Another case of buffering is when organizational success depends less on the conservation of costs than on other criteria of performance. A firm with a solid cost-plus contract need not worry about hiring female labor. Likewise, a firm whose profits depend on the irregular success of particular products may put all of their resources into product development and few into obtaining cheap factors of production. In Hollywood, where a single hit can finance the expenses of a studio for years, one would expect that film executives would emphasize artistic values and marketing and be less concerned with the costs of studio labor. Another example is the political organization that runs on patronage. The purpose of a patronage organization is to provide lucrative jobs for the supporters of the incumbent political party. Because of this, it is less important to economize on labor costs and more important to make salaries as attractive as possible. Women should appear in those firms that engage in price competition, and men should be hired in those firms that are protected from such considerations.

The Study of Occupational Sex-Typing

General Considerations

The preceding discussion of the theory of occupational sex-typing suggests two general conclusions. First, there are a very large number of variables that play a major role in determining the sex-type of a given occupation. Second, there are an equally large number of variables whose relationship to sex-typing is either non-existent or spurious. This implies that empirical studies of the economics of women’s work need to use data of sufficient richness and detail to permit the identification of the complex historical and contextual factors that could be contributing to the determination of a particular sex-type. At the same time, the research methodology needs to be sufficiently rigorous and scientific to permit the testing of alternative hypotheses and the rejection of those variables that do not play a significant role.

In general, most of the recent research on women’s work has been either historically detailed or methodologically rigorous, but rarely both. Great historical works such as Margery Davies’ Women’s Place Is at the Typewriter and Alice Kessler-Harris’ Out to Work represent sophisticated treatments of women’s economic life that are based on enormous masses of historical and archival material. These authors have, with good reason, emphasized intellectual and historical breadth rather than the niceties of social-scientific methodology. Their analyses are compelling because of the large number of examples they provide, but they rarely apply alternative models to their data to see if different perspectives provide a better explanation. In the interest of maintaining an integrated perspective, they incorporate as many variables as they can into their explanations, without attempting to test particular theories with an eye to disconfirmation so that a simpler, more parsimonious explanation can be extracted.¹³ These accounts, while often convincing, leave the reader curious about how their implicit models would behave when subjected to rigorous quantitative examination.

The quantitative work on occupational sex-typing has been extremely informative, but limited in the types of questions it could address. Most of these analyses have used census data or other comparable national aggregate statistics to study inter-occupational variance in the use of women. David Snyder and Paula Hudis in the late 1970’s and William Bridges in the 1980’s have provided stimulating analyses using this technique. However, most national economic data do not provide information on many factors that could reasonably affect the feminization or de-feminization of an occupation. Union membership data are available, but little is known about which male workers accept or resist the extension of female employment. Statistical measurements of skill are available, but these hardly describe whether a given job requires upper-body strength. Salary data are available, but little is known about the effect of female employment on promotion chains or job queues. Most of the interesting questions concerning women’s work require detailed information on organizational functioning that is hard to synthesize from national statistics.

An alternative strategy is to collect one’s own data on the determinants of occupational sex-typing in a form readily amenable to quantitative analysis. Although it would be desirable to do this for an entire national occupational structure, in practice it is very difficult to obtain detailed information on more than a small number of related occupations at one time. In order to do sophisticated multivariate analyses, however, it is essential to have a relatively large case base. One solution to this problem is to abandon the occupation as a unit of analysis and investigate variations in the use of women within a single occupation. One can contrast firms versus firms, departments versus departments, industries versus industries, or periods versus periods, in each case contrasting those units that use men with those units that use women. The concentration on one occupation considerably simplifies the analysis, since it is far easier to grasp the underlying economics of staffing a small number of related positions than it is to understand the complexities of every single job in the economy. At the same time, the smaller units of analysis virtually force the investigator to become aware of the firm-level and workplace-level considerations that are affecting the use of women. Such immersion in the organizational micro-dynamics of the firm can produce unexpected and counterintuitive findings that can have implications for macro-level research. A study based on a single occupation does have the drawback of being of limited generalizibility. If scrupulous attention is paid to the ways in which the central occupation differs from jobs in other sectors of the economy, however, the research can suggest global propositions about sex-typing as a whole.

The Research Design

This is an organizational-level investigation of the feminization of clerical labor in Great Britain. Within the subject of occupational sex-typing, there are two reasons why clerical work is an especially important case to study. First, clerical occupations make up a very large proportion of female employment. In 1950, 15 percent of all American women worked in some sort of clerical occupation (Oppenheimer 1970). Second, clerical work is one of the few occupations that has undergone a drastic shift in sex type. In the mid-nineteenth century it was a virtually all-male occupation. Then, between the 1870’s and the 1950’s, it feminized so rapidly as to become virtually all female. This is a somewhat remarkable transition considering that most occupational sex types remain constant over time. Between 1960 and 1970, fewer than 25 percent of all occupations changed their sex composition by as much as 5 percent (Snyder and Hudis 1976). Clerical work represents a rare opportunity to study significant historical change in a field where most longitudinal changes are relatively trivial.

More specifically, this study is a contrast of the employment policies of two major employers of white-collar workers. One, the British General Post Office (GPO), was one of the first employers in England to use large numbers of female clerks, while the other, the Great Western Railway (GWR), was one of the last employers to maintain a clerical force that was overwhelmingly male. These particular firms are especially important because of their completely different solutions to the problem of clerical hiring. The Post Office was the lead sector of feminization (see Table 1.1). It hired its first women clerks in 1870, and by the late 1870’s nearly a quarter of its clerks were female. By the onset of World War I, the Post Office was 40 percent female, and by the mid-1930’s nearly half of the postal clerks were female. The Great Western Railway did not even begin to hire women clerks until 1906. It had considered hiring women in 1876, but explicitly rejected the possibility. In the pre-World War I period, while the Post Office was feminizing dramatically, the railway hired only a handful of women. The war period significantly increased the GWR’s use of women so that, by 1924, 16 percent of all GWR clerks were women, but this number remained virtually fixed throughout the next fifteen years, so that by World War II, despite the national trends, the Great Western Railway still filled five out of six of its clerical positions with men.

The General Post Office and the Great Western Railway are typical of the behavior of the civil service and of railways overall. Table 1.2 shows the percentage of clerks who were female in various industries in Britain in 1901 and 1921. Following the example of the Post Office, the British civil service and local authorities feminized fairly rapidly. At the turn of the century, the public sector was the only industry in Britain whose clerical force was more than one quarter female. Many important clerical employers, such as insurance and banking, had not begun to feminize at all. By 1920, most of the private sector had caught up with government. A conspicuous exception was railways. The railways had hired no female clerks during the nineteenth century, and even after World War I they continued to be alone in maintaining a clerical force that was more than 85 percent male.^{14, 15}

In order to test competing theories of occupational sex-typing systematically, it is necessary to go beyond a simple comparison of these firms. Because only two cases are being considered, it would be easy to identify many empirical differences between the Great Western Railway and the General Post Office and link them in some convoluted fashion to the use of men or women clerks. To differentiate effectively between the valid and invalid explanations, it is necessary to consider additional data. One strategy for doing this is to examine internal differences within each case. Each firm had several semi-autonomous departments with extremely different functions and technologies. These departments were responsible for setting their own hiring policies, with central intervention occurring only on a limited basis.

Furthermore, there is time-series data for each of the departments in the study. For the Post Office, there are observations of the percentage of females in each department for every year from 1857 to 1937. This is a remarkably continuous body of data that one rarely finds at the organizational level. For economy’s sake, this study uses fifteen panels, spaced roughly five years apart, giving over three hundred usable observations. The railway data are almost as detailed. Since the GWR hired women in substantial numbers only after 1914, the only period of interest for departmental contrasts is the 1914-39 period. Although there are data available for many of these years, the use of women by the various departments is relatively constant over time. Therefore, the study confines itself to one year for which there are exceptionally good data: 1933. In this year, the Great Western Railway took a census of its entire staff, which provides department-by-department summaries of sex ratios, grade levels, and pay rates. This can be combined with the Great Western Railway staff chart of the same year, which provides sex ratios, grade levels, and detailed job descriptions for virtually every clerical work group within each department of the railway. This study will use both departmental and office-level railway data to test the theories suggested by the larger two-case comparison.¹⁶

This quantitative data can be supplemented by an unusually rich stock of archival materials drawn from the records of each firm. The papers of both the Post Office and the Great Western Railway have been preserved by the British government and are available for inspection at Post Office headquarters and the Public Record Office. These two main collections are remarkably intact, even by the relatively high standards of British industrial archives, and include the correspondence, reports, budget estimates, and memos that preceded most of the critical managerial decisions concerning female employment. These are combined with a wealth of material on the normal operation of the various departments within each firm that allows for the uncovering of the relationship between gender policies, personnel problems, and the concrete concerns of daily administration. Furthermore, there are several smaller archives that contain important additional material. The records of the unions of each firm have survived, as well as newspaper accounts of the major strike actions. Parliament made several investigations into both postal and railway operations and published detailed reports on the labor dynamics of each firm in the House of Commons Sessional Papers. Thus, the statistical analysis can be grounded in the historical context of the times; those hypotheses that are suggested by regression analysis can be validated from the actual writings of managers and observers who were intimately familiar with the operation of the firms.

The Historical Background: The Great Western Railway

Because these organizations are relatively unknown to the American reader, some background material on both firms is provided to clarify the material that follows. The Great Western Railway was an extremely large and successful railway that operated the line connecting London with Bristol, South Wales, and the West of England. The system has always been one of the largest in Britain, and effectively dominated inland transport over a significant portion of England. In the 1920’s, it controlled as a subsidiary holding every coal railway and coal-loading dock in southern Wales; all the coal that was shipped out of South Wales after 1921 was transported on some form of Great Western facility. The scope of the Great Western Railway is best appreciated in the light of the events following 1921, when the British government consolidated all the private railways in the country into four private regional cartels. In most cases, this was done to eliminate the inefficiencies of duplication, and most of the holding companies consisted of the merger of two or three major systems. The western region, however, consisted of the Great Western Railway by itself. Although certain smaller lines were incorporated with amalgamation, the GWR had such hegemony over its region that the changes induced by the granting of a territorial monopoly were negligible (Nock 1964; Aldcroft 1968). The GWR was also a major manufacturer. It built most of its own locomotives and rolling stock and rolled many of its own rails. To this end, it operated an enormous complex of factories in Swindon, a rail junction in rural southwestern England. An additional minor holding was a chain of large hotels located near the major metropolitan depots. Furthermore, at various times, the GWR dabbled in bus transport, trucking, and shipping.

Railways represent the vanguard of modern-day business administration. As Alfred Chandler has argued quite cogently, railroads were the first corporations to hire large numbers of people in geographical settings that were dispersed over a number of widely scattered physical locations. Funds had to be collected from stations located all up and down the lines and safely delivered and accounted for at central headquarters in London. The enormous potential for fraud and embezzlement implicit in widely decentralized work settings was countered by the early development of a sophisticated coordinated system of financial checks and balances. Most of the modern bookkeeping and accounting procedures that we now view as the basis of sound business practice were first developed by the railways in response to their geographical vulnerability (Pollard 1965; Chandler 1977). Since an integrated accounting system required a clerical force to implement it, by the 1840’s the Great Western Railway had become a national pioneer in developing a large, efficient, highly centralized administrative and clerical staff. Most of these employees were located in the company headquarters at Paddington Station, London. However, a substantial minority was located in the Chief Mechanical Engineer’s Department in Swindon to deal with the western accounts, and many others were scattered among depots all over southern England and Wales.

Although the GWR was large and relatively monopolistic, it must not be thought that it was entirely free from the pressure to run efficiently or conserve on costs. Throughout the nineteenth century, it was perpetually fighting off challenges from aspiring competitors to siphon off traffic onto other lines. The traffic for the southwest of England was evenly divided between the GWR and the London and Southwestern Railway, while the London and Northwestern Railway built numerous feeders into GWR territory. The GWR had to engage in fare competition and invest substantially in subsidiary lines in order to defend its market against these rivals (Nock 1963; Macdermot 1964; Pollins 1971).

By the twentieth century, the GWR’s position relative to rail competitors was essentially secured. However, maturity brought a new and far more devastating set of problems. The development of road transport created the defection of business to busses, trucks, and private cars, all of which operated with significantly lower levels of overhead. The demand for shipping declined even further with the slowing of the British economy, and the depression of the 1920’s and 1930’s especially lowered freight traffic. At the same time, operating costs began to rise significantly. Railway unions became established in the late nineteenth century and exerted a great deal of pressure on wages throughout the twentieth century. Furthermore, the physical capital of most lines had depreciated, requiring ever-increasing amounts of investment in order to maintain the quality of service. The result was that the railways became a high-cost provider of transport services in a period of declining market share (Nock 1964; Aldcroft 1968). Although the Great Western Railway enjoyed some competitive advantages relative to other railways, it was affected by these same adverse industry-wide trends and had to manage aggressively in order to preserve an ever-shrinking margin of profitability.

The Historical Background: The Post Office

The Post Office is one of the few bureaucracies that actually precedes the railways. Mail delivery has existed in England since the Renaissance. However, the nineteenth century gave rise to an expansion of national mail services that far outstripped the rate of change that had gone before. Part of this increase can be attributed to the Industrial Revolution and the general growth of commercial communication. However, the Post Office contributed to this expansion itself by dramatically dropping the price of postage in the 1840’s. Furthermore, in the late nineteenth century, a wide variety of new functions were assigned to the Post Office. Its financial role was expanded by the addition of a money order system and the creation of a national postal savings bank. In the late nineteenth century, the Post Office expanded into telecommunications as well, as Parliament granted it a national monopoly on both telegraph and telephone service. The telegraph service was nationalized in the 1860’s and the phone service in the 1890’s. The Post Office became responsible for the construction, maintenance, and day-to-day operation of these services. As a result, the postal labor force became extremely diversified. Mail delivery required a staff of postmen and sorters, with a modest supporting staff of clerks. The financial divisions trebled the demand for accounting personnel, while the telegraph and telephone divisions created a demand for engineers, linemen, construction workers, telegraph signallers, messengers, and telephone operators. The rapid popular acceptance of both telegraphs and telephones led to dramatic growth in each of these divisions, while a simultaneous increase in the volume of mail produced a parallel although smaller growth in the more traditional postal occupations (GPO Establishment Books 1857–1937; Cohen 1941; Gladden 1967).

Like the Great Western Railway, the Post Office enjoyed some protection from price competition; however, it was still under pressure to conserve costs. The Post Office enjoyed a monopoly of mail, telegraph, and postal communications, which were under no significant pressure from alternative services. Furthermore, because the Post Office was a division of the British government, it could count on public subsidies if required to make up any financial shortfalls. Nevertheless, postal administrators still had to be budget conscious. Only a portion of total postal expenditures was financed by user fees, such as the sale of postage stamps. The remainder had to be requisitioned from the legislature, and Parliament was often reluctant to provide too generous a subsidy. To ensure that postal operations did not produce too great a drain on public revenues, a major check was provided on spending by placing the Post Office under the fiscal control of the Treasury. Every matter of postal policy that implied a change, either positive or negative, in the level of spending, had to be approved by the Treasury. In practice, the Treasury used its veto power frequently. In general, the Treasury rarely objected to technical considerations in the provision of service; the primary ground of contention was cost. Therefore, a major stage of the implementation of any innovation was the development of a proposal that would be financially acceptable to the Treasury authorities, and no staffing decision could be made without precise and defensible estimates of the implication of the new policy for overall costs. The tendency of the Treasury routinely to reject proposals introduced significant financial discipline into postal affairs.¹⁷

Preliminary Considerations

The very nature of the cases being compared rules out certain preliminary explanations of clerical feminization. It would be difficult to argue that clerical work feminized as an automatic response by employers to the presence of cheap female labor. The staffing patterns of the Post Office are consistent with such a claim, but those of the Great Western Railway are not. Despite the availability of women clerks, the GWR waited thirty years before feminizing, and even then used female clerks sparingly. Labor supply can also be ruled out as an explanation for differences in these cases. Both firms recruited from almost identical labor markets. Both firms were hiring from 1870 to 1940 and both firms were headquartered in central London. Both firms also maintained a series of branch offices throughout Britain, in similar although not identical locations. The Post Office hired throughout Britain, while the Great Western Railway hired primarily in southern England and Wales. However, regional factors do not seem to have had any effect on hiring. The London offices of each firm show precisely the same differences in the propensity to use women that are found in the national organizations overall. Furthermore, differences in local labor markets do not seem to have had any effect on sex-typing within each organization. There were no significant regional differences in hiring within the Post Office. There were modest differences between English and Welsh offices in the Great Western Railway, but these invariably disappear when the substantive variables of the analysis are introduced. The distinctive hiring patterns of the Great Western Railway and the General Post Office were created by the firms themselves, not by the external supply of labor.

The reader should be aware that the clerical force employed by these firms was somewhat more diverse than is the case for the average office. These offices used large numbers of traditional clerks, but the Post Office employed a significant number of mail sorters, telegraph signallers, and telephone operators as well. Most commercial offices have used a handful of such workers. Most offices maintained mail-rooms and telephone switchboards; in the nineteenth century, some firms had private telegraph lines for their most urgent communications. Nevertheless, the Post Office had a massive concentration of sorters, signallers, and operators that would have dwarfed the requirements of most private firms. This somewhat specialized labor force makes the so-called “clerks” of the postal analysis not at all comparable with the population of a modern office.

These additional workers provide an important analytical opportunity to test competing theories of clerical feminization. Each of these three types of exceptional workers differed from the more traditional clerks in important ways: the telegraphists were highly unionized; the sorters were in departments that were particularly prone to political manipulation; and the telephone operators had unusually low promotion opportunities. By systematically comparing these workers with those in more truly clerical offices, we can isolate the effects of unionization, patronage, and promotion chains in promoting clerical feminization. The semi-clerical workers increase the power of large sample tests using postal data by providing a larger range of variance in the independent variables. Including these extra occupations carries with it a small cost in generalizability; employers with a purely clerical force might not be subject to some of the processes observable in the periphery of this sample.¹⁸

The discussion that follows does not take the form of the theoretical discussion in the first part of this introduction. Rather, the main findings of the empirical analysis are laid out in terms of their own internal logic and then linked to the larger theory at the end of each chapter. In Chapter 2, the interrelation between the use of women and the need to conserve on clerical costs is explored. This leads to a two-part discussion in Chapters 3 and 4 of the relation between skill, internal labor markets, and female employment; Chapter 3 discusses the role of de-skilling in clerical feminization, while Chapter 4 provides a more broad-based attack on human capital theory. Chapter 5 discusses normative and ideological limitations on female employment and how they were related to exogenous economic needs. Chapters 6 and 7 provide a two-part discussion of the relationship between occupational sex-typing and class conflict; Chapter 6 describes the role of male unionists in determining sex type, and Chapter 7 describes the extent to which women were used as a source of labor control. Chapter 8 describes the relationship between female employment and the availability of alternative sources of secondary labor. Chapter 9 restates the main findings of the analysis and discusses their implications for a larger theory of occupational sex-typing.

Notes

1. For examples of the use of such theories in more eclectic presentations, see National Manpower Council (1957). and Oppenheimer (1970).

2. There are of course neo-classical theories of discrimination. However, none of them predict hiring regimes that show significant deviations from those expected under a system of rational self-interest. Becker explicitly argues that employers who discriminate will be driven out of business by the rigors of the market, while Phelps and Aigner and Cain view discrimination as the result of profit-maximizing under conditions of imperfect information (Becker 1957; Phelps 1972; Aigner and Cain 1977).

3. The rationale for the latter is that, in human capital theory, a worker who is being trained is indirectly financing his education by taking a lower salary. Supposedly, when employers provide workers with skills that are of value to competing employers, they force the worker to incur the cost of the training. The long-term worker is willing to make this sacrifice because his education will raise his productivity and thus increase his market wage. However, a short-term worker pays for training costs without receiving a long-term benefit, and thus should be uninterested in acquiring human capital.

4. For an even stronger position against the existence of gender differences in turnover, see the extensive reviews of the empirical literature on quit rates by Anderson (1974) and Price (1977).

5. A reasonable objection can be made that most of the data presented above come from the twentieth century. Brenner and Ramas (1984) argue provocatively that supply-side limits to female employment may have been more important in the nineteenth century. Market employment would have been less attractive to women due to the lower prevailing rates of real wages, while high fertility was an economic necessity due to children’s importance as a source of security in old age. It is true that there were more incentives for women to remain home in the nineteenth century. However, it is still unclear that these incentives were so great as to restrict significantly the supply of female labor open to employers. The evidence that is required to show an independent effect of labor supply would be for employers to attempt to hire women for lucrative positions but be forced to hire men after experiencing a shortage of female job applicants. If an occupation de-feminized after employers failed to find an adequate supply of female labor, or after suffering from excessive rates of female turnover, this would support the Brenner and Ramas hypothesis. Empirically, it is hard to think of examples that fit this model. Those cases where de-feminization has occurred seem to have involved women’s mobilizing into trade unions, the passage of laws banning female labor, or significant changes in the technology of industry (Kessler-Harris 1982). This is not to say supply-based changes did not occur. No definitive settlement of this issue can be made until there is the same sort of rigorous test of supply and demand models for 1820 that there has been for 1970.

6. This argument implicitly assumes that wages are not purely a function of productivity but are affected by exogenous factors as well. The recent literature on dual labor markets, internal labor markets, and the effect of unionization on wages provides more than adequate support for this hypothesis (Parsley 1980; Kalleberg et al. 1981).

7. The savings from using women in high-status positions could be even greater than those associated with normal jobs. The logic behind the overcrowding hypothesis would suggest that since the supply of highly paid jobs for women is far more constrained than the general supply of jobs, employers could receive far greater savings in using women for superior jobs than they could by using them in average positions. The extreme scarcity of supervisory positions for women might induce women to work at 40 cents to the male dollar, instead of the normal 59 cents.

8. It is true that many employers are reluctant to hire older workers. Some managers fear that older workers have suffered a deterioration in their skills because of the long period they have been out of school. This would not be an issue for a women who has gone back to school after completing her childbearing. Another objection is that older workers are often failures and cast-offs from other organizations. This again is not a problem for an older female labor force entrant. A third issue might be the higher chance of serious illness or diminished physical capacities among older workers. However, most middle-aged workers are in perfect health. Furthermore, although a forty-year-old worker only offers a prospective career of twenty-five to thirty years, while a twenty-five-year-old worker offers a prospective career of forty years, it is not at all clear that the extra fifteen years offer any advantage to an employer. Over a forty-year period, most workers will leave their original employer for non-health-based professional or personal reasons anyway.

9. The presence of this last provision makes the theory what Arthur Stinchcombe would call an historicist model. See Stinchcombe’s Constructing Social Theories (1968) for an excellent discussion of the potential uses of such theories. See also Stinchcombe’s 1964 essay on the effect of period of creation on organization structure.

10. All data on the sex composition of current occupations come from the 1970 Census Report on Occupational Characteristics (U.S. Bureau of the Census 1970).

11. In this light, it is hard to accept the argument that particular occupations feminize because of an employer’s need for cheap labor (Oppenheimer 1970; Prather 1971; Davies 1975). Since, at a superficial level, all employers have some interest in obtaining cheap labor, it is necessary that authors use this argument in tandem with other considerations that explain why feminization does not occur universally.

12. This argument is generally more convincing for race and ethnicity than it is for sex; however, it is a non-trivial theory of gender that will be examined in detail in Chapter 7.

13. For an exception to this tendency, see Kessler-Harris’s excellent discussion (1982) of the extent to which the impact of World War II on women’s employment has been exaggerated.

14. The London and Northwestern Railway experimented with hiring women in the 1870’s, but seems to have abandoned the project soon thereafter (GWR Women 1906).

15. American industries closely resemble British industries in their history of using women clerks. The federal government was the first employer to hire women clerks. Women were introduced to the Treasury Department in 1862; by 1870, no other private employers had yet begun to hire female clerks, although the practice had spread to other federal offices (U.S. Bureau of the Census 1870; Aron 1980). By the 1930’s, private industry had caught up with the public sector, and as was the case with Britain, government had an average ratio of males to females. Railways, however, lagged considerably behind. In 1930, fewer than 20 percent of all rail clerks were female, despite the fact that nationally the clerical force was 55 percent female (U.S. Bureau of the Census 1930).

16. See Appendix A for details of each sample.

17. A second source of fiscal review came from Parliament itself. Parliament routinely held committee hearings on the postal budget and operations, in which the administration would be expected to justify its expenditures. In practice, this was sometimes an easier forum than the Treasury, because fiscal conservatives were often counterbalanced with members whose primary concern was with the level of services provided. Nevertheless, many legislators were primarily concerned with reducing government expenditures, and parliamentary votes show significant opposition to a number of costly postal bills, in particular those that would have increased the level of wages.

18. For further details of the occupational base of this sample, see Appendix B.