Andrew J. COGNARD-BLACK* RIDING THE GLASS ESCALATOR TO THE PRINCIPAL'S OFFICE** Sex-atypical Work among Token Men in the United States Abstract. This study investigates the effect of workplace sex composition on men who do "women's work." Using data from 5,734 U.S. school teachers, I conduct multinomial logistic regression analyses to test hypotheses regarding men's and women's relative likelihood of moving out of teaching and advancing up into administrative positions. Results from these analyses provide no evidence that Kanter's tokenism proposition is generalizable to token men who teach in elementary schools; rather than suffering disadvantages, evidence supports Williams's "glass escalator" proposition that men enjoy privilege in predominantly female jobs and are more likely than 878 women to be promoted into administrative positions. Keywords: gender, work, occupations, teaching, sex composition, glass escalator, tokenism In 1977 Rosabeth Moss Kanter published one of the most widely read organizational studies in sociology, Men and Women of the Corporation. In it she proposes that token workers—individuals who are in an extreme numerical minority along some salient dimension such as race or sex—are subject to workplace difficulties above and beyond those their nontoken colleagues face. Such individuals, Kanter argues, are more visible by virtue of their difference, are marginalized in everyday workplace activities, and are thus subject to more work-related scrutiny, criticism, and performance pressure. Although based on the experience of women at "Indsco," the tenor * Dr. Andrew J. Cognard-Black, Assistant professor St. Mary's College, Maryland. ** Reprinted in part from "Will They Stay, or Will They Go? Sex-Atypical Work among Token Men Who Teach," by Andrew Cognard-Black, published in The Sociological Quarterly 45(1), 113-139, by permission of John Wiley and Sons. © 2004 The Midwest Sociological Society. I would like to thank the following individuals for their helpful comments during completion of this project: Kathryn Barry, Kristina de los Santos, Mikaela Dufur, Carl Klarner, Julie Kmec, Debra McBrier, Michael McCrary, Cynthia Pelak, Shana Pribesh, Rachael Woldoff, and three anonymous reviewers. Special thanks go out to Robert L. Kaufman for both commentary and counsel. I would also like to thank Joan Arnfield and Marieke van Willigen for technical assistance and computer support. Though these individuals have assisted me in improving this paper, any errors are my own. Completion of this project was supported by a grant from the Graduate School at The Ohio State University to the Department of Sociology research fellows program. and theoretical thrust of Kanter's study suggests that a marginal, scrutinized existence would be the experience of anyone who is a token in whatever context; women would suffer these consequences in other predominantly male settings, African Americans in predominantly white settings, and so, too, would men in predominantly female settings. Since its publication, however, several scholars have criticized Kanter's general theory, arguing that it is misconceived as "gender-neutral" (Williams, 1995) and that it does not adequately account for sex differences in the experience of sex-atypical work (Zimmer, 1988; Budig, 2002). The fundamental criticism is that, while men in predominantly female workplaces may suffer disadvantages as a result of token status, they may also be advantaged due to a general cultural preference for men and masculinity. Kanter (1977) does acknowledge this possibility: in the case of tokens whose master status [in society] is higher than their situational status [as workers], leveling can work to their advantage, as when male nurses are called "Dr." (p. 232) However, she does not pursue this line of reasoning in the course of her study, and its implications do not make their way into her conclusions about sex-atypical work. 879 On one level, then, this paper interrogates these two competing notions about men's experience in sex-atypical work. On another level, this paper contributes to our limited understanding of men's movement into and out of sex-atypical work and the implications of these phenomena for occupational sex segregation—a workplace modus operandi for perpetuating income, wealth, authority, and prestige inequality between men and women (England, 1992; Marini, 1989; Reskin, 1993, 1988; Reskin and Roos, 1990). To examine these phenomena I focus on men who do "women's work." More specifically, I focus on men who do a particular kind of "women's work"— elementary school teaching—and ask the question: Do male elementary school teachers' careers respond more to the disadvantages or the advantages associated with sex-atypical work? If tested empirically, will they stay, or will they go? Methodologically, school teaching provides a unique case with which to examine men's experience in sex-atypical work. At the more aggregated, occupational level, it might appear as if teaching is somewhat sex-integrated (what Kanter calls "tilted"), with about 28 percent of teachers being male at the beginning of the 1990s (U.S. Department of Education, 1994a). However, looking more closely at the major divisions within the occupation, we see that school teaching is quite segregated by sex (see Bielby and Baron, 1986 for a discussion of the importance of level of aggregation when studying workplace sex segregation). In the decades up to and including 1990, men represented only about 15 percent of elementary school teachers (Biedenkapp and Goering, 1971; Williams, 1995)—Kanter's "token" threshold. In contrast, about 42 percent of secondary school teachers were men in 1990 (U.S. Bureau of the Census, 1992). Thus, more so than the handful of other predominantly female professions (e.g., librarianship, social work, and nursing), school teaching is meaningfully divided into two occupations similar in length of training and requisite skills but dramatically different in sex composition. This fact makes school teaching ideal for comparing the effect of varying sex composition on men's career patterns to that on and women's career patterns. Most research on workplace sex segregation has focused on ways in which premarket forces (e.g., childhood socialization) and market forces (e.g., statistical discrimination in hiring) sort men and women unevenly throughout the occupational structure (see England, 1992; Reskin, 1993). However, we know less about how experiences while in the occupational structure contribute to segregation through patterns of uneven occupational attrition (Jacobs, 1989), and we know even less about men's attrition from predominantly female workplaces, where they are already few in number (Williams, 1993). As Reskin (1993) points out in an Annual Review 880 of Sociology article on workplace sex segregation, more research is needed to explore "the factors that influence workers' moves into and out of sex-atypical jobs" (p. 265; emphasis added). Though Jacobs (1989) has examined this process specifically as it pertains to women's movement into and out of sex-atypical work, there is comparatively little research exploring men's movement into and out of sex-atypical work. Such an oversight is particularly problematic for a theoretical understanding of workplace sex segregation since it is arguably men's resistance to integrated work that is the overwhelming force behind perpetual workplace sex segregation (Reskin, 1988). In order to understand more fully the existing sex composition of occupations, it is important to study not just what encourages men to work in particular occupations but also what discourages them from working in others. However, few scholars have investigated why men are so underrepresented in jobs defined as "female" or what happens to those men who do take such jobs (Rosenfeld, 1984; Williams and Villemez, 1993). This study is a response to Reskin's (1988) appeal to bring the men back into our research and understanding of workplace sex segregation—to better understand the parameters of men's resistance to working as equals with women. It is at once both an attempt to address the theoretical disagreement about men's experience in sex-atypical work and an examination of the role that men's experience with sex-atypical work plays in workplace sex segregation. Theory and Evidence on Men's Sex-Atypical Work The gender-neutral Kanterian perspective suggests that the few men who enter "female jobs" (i.e., occupations or jobs that are numerically dominated by women) will be more dissatisfied by virtue of their token status and thus suggests that such men will be more likely than their female colleagues to leave these jobs. More recently, Jacobs (1989, 1993) has formulated an addendum to this perspective that focuses more explicitly on mechanisms discouraging sex-atypical work. Using a revolving door as a metaphor for the process, Jacobs theorizes that lifelong encounters with social control constrain individuals' occupational choices. Individuals may encounter control that discourages entry into an occupation, and more to the point, persisting social control can serve as a mechanism that encourages men and women to quickly leave sex-atypical work that they have come to occupy despite earlier social controls, as if they had come and gone through a single turn of a revolving door. For instance, typical workers could sabotage the work of an individual who is sex-atypical (see also Cockburn, 1983 and 1991 for examples of sabotage among token women), or dominant workers could exclude them from break room conversations and otherwise make them feel as if they don't belong. According to the Kanterian perspective 881 we would expect that the cluster of disadvantages encountered by men in "female jobs" would discourage them so much that the tendency for men in those positions would be disproportionate outward movement. In contrast, a second perspective recognizes that men's experience of sex-atypical work may be different from that of women because of the cultural power associated with masculinity and that they are often rewarded in "female jobs" (see Acker, 1990; Williams, 1992, 1989; Williams and Villemez, 1993; Yoder, 1991; Yoder and Sinnett, 1985; Zimmer, 1988; Budig, 2002). Subsequently, this perspective proposes that these advantages outweigh any disadvantages that men may suffer as token workers. To accommodate this possibility theoretically Williams (1992) proposes a "glass escalator" model of men's experience in predominantly female work, arguing that, among other benefits, there is an informal glass escalator on which token men in predominantly female settings are encouraged to rise upward into positions of authority. Although these men may suffer some disadvantage in the form of societal suspicion—chiefly about their masculinity and sexual orientation—economic benefits and opportunity assuage those disadvantages. Thus, we would have little reason to expect token men to leave "female jobs" with any greater likelihood than women, except for those few for whom we would expect a greater likelihood of upward movement into occupationally related administrative jobs. In addition to these two competing explanations of men's experience as tokens in predominantly female workplaces, research also suggests a third possibility. Drawing on Blalock's (1967) and Blau's (1977) theories of minority-group relations, Wharton and Baron (1987, 1991) use 1973 Quality of Employment Survey data in two companion articles to examine men's and women's work satisfaction across a range of workplaces of varying sex composition. Because the quality and quantity of intergroup relations decline as groups become more balanced, they argue, men will be least satisfied in integrated settings. Looking specifically at men's satisfaction at work, Wharton and Baron (1987) find that men are least satisfied in sex-integrated workplaces and that "the satisfaction levels of men in predominantly female and mixed but segregated work settings ... do not differ significantly from men in all-male work settings" (p. 582). This evidence suggests a parabolic relationship between sex composition and men's disadvantage (i.e., dissatisfaction) at work: Token men in predominantly female settings might not necessarily suffer disadvantages in the way Jacobs (1989) and Kanter (1977) suggest they will, and thus such men may have little reason to leave predominantly female workplaces. Advantages, Disadvantages, and Outward Movement among Token Men 882 These are three quite-different theoretical expectations, and to date the evidence bearing on them is contradictory and inconclusive. Research on token men, much of it based on qualitative data from nonrepresenta-tive samples, provides evidence of both advantages and disadvantages for these men. Some studies indicate that men have advantages in hiring, often acknowledged as administrative attempts to integrate "female jobs" (Allan, 1996; Williams, 1989, 1995). Other studies find that token male nurses are taken more seriously on the job than female nurses (Floge and Merrill, 1986; Heikes, 1991). There is also some qualitative evidence of a glass escalator for men that encourages their advancement (Pierce, 1995; Williams, 1989, 1992, 1995), but in one study the authors ask if there is "another ceiling" and conclude that there is antimale bias in promotion among social workers (Atwater and Van Fleet, 1997). Indeed, the bulk of studies in this area focus on identifying the difficulties associated with being a man in a "female job." When one compares what they could have in "male jobs," men in female jobs suffer both prestige and wage penalties (Jacobs and Powell, 1984; England and Herbert, 1993); men may encounter suspicion about their masculinity and sexual orientation, arguably large components of many men's identities (Allan, 1996; King, 1998; Murray, 1996; Sargent, 2000; Skelton, 1994; Williams, 1993, 1992, 1989); and there is some qualitative evidence based on research on elementary school teachers that token men perceive interpersonal resistance from female colleagues (Allan, 1993, 1996), though other research on nurses provides contradictory evidence that women are not resistant to the idea of male colleagues (Fottler, 1976). Studies specifically about men in elementary school teaching cite some additional unique disadvantages. Many studies indicate that token men in elementary schools have severe role conflict: The job requires a degree of nurturing and physical contact with small children, but men's gender role expectations preclude such behavior and invoke suspicion of pedophilia if they meet the nurturing expectations of the job (Allan, 1996; DeCorse and Vogtle, 1997; King, 1998; Murray, 1996; Sargent, 2000; Skelton, 1994). To a lesser extent, men in these studies also describe role encapsulation, whereby they are pressured to take on responsibilities as the resident disciplinarian (Allan, 1996; Sargent, 2000). Finally, in Allan's (1993) research using qualitative data from 15 male elementary school teachers, he reports that men expressed feelings of marginalization. These men reported alienation from female colleagues and believed that the alienation was partially due to the women's resentment toward the men for their privileges in hiring. However, if respondents tried to resist preferential treatment in order to gain trust among the other teachers, they then risked alienation from male principals. 883 Based on this evidence, many scholars agree that both advantages and disadvantages exist for token men in elementary school teaching and other female-dominated professions, but there is mixed evidence as to how the weight of the advantages and disadvantages affect mobility. Two quantitative studies on attrition from teaching indicate that men and women have the same tendency to leave after only a few years in the profession (Sch-lechty and Vance, 1983; Talbert, 1986), but neither of these studies distinguishes between elementary and secondary school teaching, so there is no evidence regarding male mobility specifically out of elementary school teaching, the level in which they are tokens. Studies using representative quantitative data of the labor force also provide contradictory evidence regarding men's likelihood of attrition from female occupations in general. In support of the gender-neutral tokenism thesis, Jacobs (1993) offers evidence from National Longitudinal Surveys data of individuals who entered the workforce during the 1960s and 1970s; Jacobs finds that the employment of men in "female jobs" is "unusual and often brief' (p. 61). Two more recent studies find evidence of promotion advantages for men in female occupations, but the findings from these two studies lead to different conclusions about the unique effect of tokenism among men. Maume (1999) uses Panel Study of Income Dynamics data from 1981-87 to examine the effect of occupational sex and race composition on men's and women's likelihood of moving into managerial positions. Modeling sex composition as a linear effect, Maume finds that "men are more likely to move into management as the percentage of females in the origin occupation increases" (p. 499). Using the National Longitudinal Survey of Youth, Budig (2002) examines the effect of sex on wage promotions across occupation/industry "jobs" of female-skewed, male-skewed, and balanced compositions. She finds that men do have better wage levels and wage trajectories than women in female-skewed occupation/industry cells, but that advantage is no greater for these men than for those in balanced and male-skewed cells. Thus, Budig concludes, "token status appears to be irrelevant in terms of wage levels and wage trajectories" (p. 274): Men are uniformly advantaged across male, female, and balanced work. Since Maume (1999) models sex composition as a linear effect, his study does not speak to the possibility that men are advantaged across different levels of composition (male versus balanced versus female). Both of these studies do provide evidence of glass escalators for token men, but one suggests a statistical interaction between sex and token status, while the other suggests a uniform advantage for men across the occupational structure. 884 Hypotheses Where recent studies have examined the effects of sex and sex composition across the structure of heterogeneous occupations, I focus on school teaching to test hypotheses from the gender-neutral tokenism perspective and competing perspectives. Central Gender-Neutral Tokenism Hypotheses If men's token status in elementary schools leads to their higher relative disadvantage and higher attrition, then two things must be true. First, disadvantages will outweigh any advantages of token status, and thus men in elementary schools will be more likely to leave than women: Hypothesis 1: Men who teach at the elementary level are more likely to leave teaching for paid work outside a school than women who teach at the elementary level (i.e., the effect of teachers' sex on leaving may vary by the level of school at which they teach). Second, men will encounter more disadvantages than advantages in predominantly female occupations, and they will thus be more likely to leave these occupations than they would occupations with a smaller relative proportion of women. Because elementary school teaching is a predominantly female occupation and secondary school teaching is a more integrated occupation, men in elementary schools will be more likely to leave teaching than men in secondary schools: Hypothesis 2: Men who teach at the elementary level are more likely to leave teaching for paid work outside a school than men who teach at the secondary level (i.e., the effect of school level on leaving will vary by teachers' sex). If there were support for these two hypotheses, this would provide evidence that disadvantages outweigh advantages for men in sex-atypical occupations, thus supporting the gender-neutral tokenism explanation in the Kanter tradition... . Competition Hypotheses * Wharton and Baron's (1987) research suggests an effect of workplace sex composition other than that from the gender-neutral social control perspective: a nonlinear one. . In direct contrast to the predictions in Hypotheses 1 and 2, this perspective would also lead us to expect that men in secondary 885 schools will be more likely to leave than women in secondary schools and that men in secondary schools will be more likely to leave teaching than men in elementary schools, thus: Hypothesis 6: Men who teach at the secondary level are more likely to leave teaching for paid work outside a school than women who teach at the secondary level (i.e., again, the effect of teachers' sex on leaving will vary by the level of school at which they teach); and Hypothesis 7: Men who teach at the secondary level are more likely to leave teaching for paid work outside a school than men who teach at the elementary level (i.e., the effect of school level on leaving will vary by teachers' sex). The Male Privilege Hypothesis Alternatively, Williams's (1992) proposition suggests that men encounter more advantages than disadvantages in predominantly female work, and * Hypotheses 3-5 and related text have been omitted in this reprinting. that one of these advantages is a better chance at promotion. If this is true, we would expect to find that, among elementary school teachers, men will be more likely than women to move up to administrative positions: Hypothesis 8: Male elementary school teachers are more likely than female elementary school teachers to move up to a principalship or assistant principalship (i.e., male sex will have a direct positive effect on moving up among elementary school teachers). Data and Methods To test these hypotheses I use data from the Teacher Survey (1990-91), Teacher Followup Survey (1991-92), and School Administrator Questionnaire components of the Schools and Staffing Survey (U.S. Department of Education, 1994a, 1994b). The original Teacher Survey sample of 53,347 is representative of public and private school teachers in the United States. The Schools and Staffing surveys are useful for this research because they were designed to provide nationally representative data for estimating 886 teacher turnover and attrition; in addition to data on teachers' jobs at two points in time, these surveys include measures of a variety of other variables, including advanced degrees, job tenure, demographic characteristics, and family-oriented obligations. Respondents were considered teachers if they were either full-time or part-time and had a primary assignment in grades kindergarten through twelfth. Short-term substitutes, student teachers, nonteaching specialists (e.g., guidance counselors or librarians), teachers' aides, and other support staff were excluded. Approximately one year after the Teacher Survey, the Teacher Followup component resurveyed a subsample of approximately 7,200 teachers from the original 1990-91 teacher sample, of whom 6,733 responded. The study used a stratified multistage cluster sampling design. In the initial stages of the followup, researchers attempted to obtain information on the mobility status of all teachers from the original Teacher Survey; the Followup Survey then took a subsample of those who stayed in teaching, while those who left teaching were sampled with certainty (Kerry Gruber, National Center for Education Statistics, personal communication, May 27, 2003). Consequently, those who left are overrepresented in the Followup Survey data. Thus, this design allows for comparison of the great percentage of people who stayed in teaching with the relatively rare cases who left teaching altogether. In addition, the stratification procedure was designed to allow for comparisons of those types of occupational moves within sector (i.e., public/ private), across experience groups (i.e., new versus experienced teachers), and between elementary and secondary levels. I weight all analyses using a simple transformation of a weight variable that adjusts for the sampling design.1 Table 1. MEANS, STANDARD DEVIATIONS, AND DESCRIPTIONS OF STUDY VARIABLES Variable Description and Metric Mean s Dependent Variable Mobility outcome at Time 2a A polychotomous nominal measure indicating whether a teacher: (1) stayed in teaching;b (2) moved up to a principalship or assistant principalship; (3) left to care for family; (4) moved out for employment in a nonschool setting; or (5) moved internally to a job within a school settingc N/A N/A Independent Variables Status characteristics Male A binary measure: 1 = male, 0 = female .276 .447 White A binary measure: 1 = white, 0 = black, Asian, Pacific Islander, American Indian, or Alaskan native .907 .291 Hispanic identity A binary measure: 1 = Hispanic, 0 = not Hispanic .036 .187 Age An ordinal measure of respondent's age at Time 1: 1 = under 30, 2 = 30-39, 3 = 40-49, and 4 = 50 and above 2.730 .915 Structural job characteristics Taught elementary A binary measure of school level taught at Time 1: 1 = elementary, 0 = secondary .529 .499 Salary (in 1000s) A ratio measure of a respondent's base salary at Time 1: ranges from 5 to 70 29.107 10.153 Full-time status A binary measure of status at Time 1: 1 = taught full-time, 0 = taught part-time .906 .292 Taught public school A binary measure of context at Time 1: 1 = taught public, 0 = taught private .881 .324 School's percent female A ratio measure of the percentage of female teachers at Time 1: ranges from 1 to 100 72.804 26.530 Male principal A binary measure for the principal at Time 1: 1 = male, 0 = female .708 .455 Missing principal's sex data A binary measure: 1 = missing data for the male principal variable .035 .184 Human capital characteristics Tenure A ratio measure of the years taught in the current school at Time 1: ranges from 1 to 51 8.854 7.465 Postbaccalaureate degree A binary measure of credential by Time 2: 1 = has a master's, doctorate, professional, or educational specialist degree, 0 = no such degree .457 .498 887 1 The weight used in this analysis is the Schools and Staffing Survey final weight divided by its mean. This procedure results in weights of the same relative size as those in the variable provided, but the mean of this weight is 1. Thus, weighted analyses have the same sample size and the same degrees of freedom as would unweighted analyses... . 888 Variable Description and Metric Mean s Leaves A binary measure of whether a respondent has taken a break in service for a year or more as of Time 1: 1 = has taken at least one break for at least one year, 0 = no such breaks .329 .470 General education credentials A binary measure of a credential in general education as of Time 2: 1 = yes, 0 = no .900 .300 Special education credentials A binary measure of a credential in special education as of Time 2: 1 = yes, 0 = no .123 .328 Administrative credentials A binary measure of a credential in educational administration as of Time 2: 1 = yes, 0 = no .108 .311 Social sciences credentials A binary measure of a credential in social sciences as of Time 2: 1 = yes, 0 = no .118 .323 Humanities credentials A binary measure of a credential in humanities as of Time 2: 1 = yes, 0 = no .009 .093 Foreign languages credentials A binary measure of a credential in foreign language as of Time 2: 1 = yes, 0 = no .040 .196 Natural sciences credentials A binary measure of a credential in natural sciences as of Time 2: 1 = yes, 0 = no .072 .258 Family characteristics Married A binary measure of marital status at Time 1: 1 = married, 0 = not married .723 .447 Young child(ren) A binary measure of young dependent children at Time 1: 1 = at least 1 child < 3, 0 = none .101 .301 New child(ren) A binary measure of new dependent children: 1 = more children at T2 than T1, 0 = no more .063 .244 Community characteristics Urban (omitted) A binary coded 1 = school located in an urban area defined as a "central city" 2 --J oo .448 Suburban A binary coded 1 = school located in an area defined as an "urban fringe" or "large town" .340 .474 Rural A binary coded 1 = school located in an area defined as "rural" or a "small town" .382 .486 Missing urbanity data A binary coded 1 = missing data for the urban/suburban/ rural construct .045 .208 Source: Schools and Staffing Survey, 1990-91 (U.S. Department of Education 1994a) and the TeacherFollowup Survey, 1991-92 (U.S. Department of Education 1994b). Note: Means and standard deviations do not reflect the data of 999 respondents who left teaching for reasons outside the scope of this study (e.g., those who left for retirement, those who left for reasons associated with a disability, those who left for further schooling, or those who left for some unknown "other"' reason not captured under the other headings). These figures are based on weighted data. a The metric presented here matters less than the fact that I use "stayers" as the omitted, reference group in the logit analyses. b In these analyses, if a respondent did not leave the occupation in one fashion or another, he or she was a stayer (i.e., the respondent could have actually moved to another school and still be considered a "stayer"'). c This excludes upward moves and includes occupations such as librarian; counselor; "resource person," such as a curriculum coordinator or department head; "support staff," such as an aide; coach; or some undefined "other." Dependent Variable Table 1 presents details about the measurement of variables I use in analyses as well as the means and standard deviations of those variables. The dependent variable in this analysis is a polychotomous nominal variable indicating whether a teacher remained in teaching at Time 2 or, if not, for what activity he or she left. It indicates whether the teacher (1) remained in teaching, (2) moved up to assume a principalship or assistant principalship, (3) left to care for family, (4) moved out for new employment in a nonschool setting, or (5) moved internally to a job in a school that was not teaching and not a principalship or assistant principalship. Because I am interested in occupational moves, this measurement does not distinguish between those who stayed in their same job and those who may have changed schools but stayed in teaching. Nor does the measurement tap whether a teacher left for retirement, for reasons associated with disability, or to attend some form of postsecondary schooling; these moves are distinct from the others in terms of normative career patterns, so I omit from multivariate analyses people who left for those reasons. Independent and Control Variables Beyond the variables implicit in the hypotheses—teacher's sex, school level, sex of the principal, and school sex composition—research in the sociology of work suggests the need to control for other factors that may affect workplace outcomes. I test five categories of independent measures for inclusion in final statistical models: status characteristics, job characteristics, human capital characteristics, family characteristics, and community characteristics. ... Analytic Approach I use multinomial logit regression to test whether independent variables influence teachers to remain in teaching or to leave for another occupation. Multinomial logit regression allows one to estimate simultaneously the likeli- 889 hood of several different outcomes in one model. In regression analyses I select only those respondents who either stayed in teaching from Time 1 to Time 2 or those who left teaching for one of the following: (1) to move up to a principal-ship or assistant principalship, (2) to take care of family, (3) to move out for employment in a nonschool setting, or (4) to move internally to another job in a school (other than principal or assistant principal). This specification excludes respondents who retired, left to attend college, or left for reasons associated with disability, and analyses thus involve only 5,734 of the original 6,733 cases. Because my hypotheses predict interactions among variables, I specify [an] interaction [term] in my analyses: (1) a two-way interaction between teachers' sex and the level at which they taught ... . Findings Hypothesis Testing* ... . Table 5 presents the final pooled model (with staying as the reference), with both logit coefficients and odds ratios for significant effects.3 * Tables 2-4 and related text have been omitted in this reprinting. 3 The odds ratio is an exponential transformation of the logit coefficient and can be interpreted as the change in the odds of the outcome for an appropriate unit of the independent variable. Because they failed to significantly improve model specification in preliminary modeling, this final model excludes a number of variables: white (race), Hispanic identity, full-time (employment status), prior leaves of absence, general education training, special education training, social sciences training, humanities training,4 foreign languages training, natural sciences training, missing male principal data, and urban context. Given their theoretical centrality, I have left the percent female and male principal variables in the final model, despite the fact that they fail to be statistically significant... . Figure 1 presents a visual representation of the effect of sex on all mobility outcomes for the pooled model. This figure presents an odds ratio plot of the effects of being male on all the mobility outcomes simultaneously, controlling for other relevant factors (see Long, 1987 and 1997 for an explanation of this method of presentation). Whereas Table 5 presents coefficients and statistical significance of the effects of variables on four mobility outcomes vis-à-vis staying, the figure summarizes the significance of sex on mobility for all the relationships (i.e., while Table 5 shows significance of effects for four kinds of movement relative to staying, Figure 1 indicates significance and the magnitude of difference for all 10 relationships, such as moving out relative to moving up). In the Figure, lines between out-890 comes indicate the absence of a statistically significant difference in the effect of being male on whether a teacher has one or the other of those two outcomes. Thus, it visually joins together those effects that are statistically indistinguishable. I plot along both a logit scale (on the bottom) and a factor change, or odds, scale (on the top). Both the logit coefficients and odds ratios provide estimates of the effect for men versus women in a given mobility outcome relative to staying in the occupation (i.e., staying is set at a factor of 1), but the magnitude of difference in the log-odds (i.e., logit) or odds can be easily calculated for the difference between any two outcomes by figuring the absolute difference along either of the scales.5 Vertical distance in Figure 1 has no meaning and only provides room to distinguish more easily the connecting lines among effects. . 4 Preliminary models do suggest that academic training in the humanities does increase the likelihood of moving upward relative to staying. However, there is not enough variation in humanities training across all the mobility outcomes in this analysis to retain the humanities credential measure in the final models. 5 For instance, the absolute difference in the effect of sex for takingfamily leave and moving up among all teachers is 4.682—the absolute difference between 1.138 and -3.544 (these numbers can be found in Table 5). Therefore, the log-odds that a teacher will move up rather than take family leave is 4.682 greater for men than for women. The exponential of4.682 is 107.986. Thus, men are about 108 times more likely than women to move up versus to take family leave. Conversely, the exponent of -4.682 is .009; being male reduces the odds of takingfamily leave versus moving up by 99.1 percent. Similarly, being female increases the odds of taking family leave versus moving up by a factor of 108 and reduces the odds of moving up versus takingfamily leave by 99.1 percent. Any of the other relationships can be calculated in a similar way. Table 5. PREDICTORS OF DIFFERENT TYPES OF OCCUPATIONAL MOVES AMONG TEACHERS WHO LEFT TEACHING VIS-À-VIS THOSE WHO STAYED (MULTINOMIAL LOGITREGRESSION) Moved Left for Moved Moved Variable Up Family Care Out Internally Logit Odds Logit Odds Logit Odds Logit Odds Constant -8.257 -7.716 1.443 -4.633 Status Characteristics Male 1.138 * (.657) 3.122 -3.544 tt (.946) .029 .407 (.327) -2.013 tt (.731) .134 Age 1.145 (2.455) 3.272 f (1.577) 26.361 -1.327 f (.803) .265 1.687 (1.734) Age2 -.339 (.511) -.830 f (.365) .436 .263 (.161) -.360 (.342) Structural Job Characteristics Taught elementary -9.640 (7.531) 5.075 tt (1.862) 159.999 -1.700 (1.454) 2.752 (2.571) Salary (in 1000s of dollars) -.043 (.035) -.047 f (.022) .954 -.062 tt (.019) .939 -.005 (.026) Taught public school -.270 (.770) -.691 (.382) -1.172 tt (.302) .310 .215 (.694) School's percent female .009 (.014) .003 (.007) -.010 (.006) -.016 (.010) Male principal .932 (.746) .580 (.320) -.088 (.287) -.456 (.418) Human Capital Characteristics Tenure (logged) -.544 (.359) -.450 f (.222) .638 -.682 " (.201) .505 -.765 " (.295) .465 Postbaccalaureate degree 1.655 f (.807) 5.234 -.208 (.340) .075 (.305) .511 (.474) Administrative credentials 1.911 tt (.557) 6.763 -.284 (.633) .136 (.470) 1.731 tt (.445) 5.647 Family Characteristics Married 1.223 (.881) 2.125 tt (.670) 8.374 -.054 (.284) .049 (.434) Young child(ren) -1.228 (1.134) 1.390 tt (.309) 4.013 -.437 (.460) .178 (.621) New child(ren) -1.060 (1.795) 2.696 tt (.289) 14.819 -.031 (.519) 1.172 f (.510) 3.228 Interactions Taught elementary * Age 6.637 (5.576) -5.529 tt (1.779) .004 1.223 (1.358) -2.686 (2.181) Taught elementary * Age2 -1.045 (1.030) 1.296 tt (.402) 3.653 -.275 (.281) .529 (.433) Log-likelihood Model c2 Degrees of freedom Number of cases -801.912 440.262 64 5,734 891 Source: Schools and Staffing Survey, 1990-91 (U.S. Department of Education 1994a) and the TeacherFollowup Survey, 1991-92 (U.S. Department of Education 1994b). Note: Numbers in parentheses are standard errors. The model is significant at the .01 level. *p < .05 "p < .01 (one-tailed tests) V < .05 "p < .01 (two-tailed tests) Figure 1 ODDS RATIO PLOT FOR THE EFFECT OF MALE ON DIFFERENT MOBILITY OUTCOMES AMONG SCHOOL TEACHERS Factor Change Scale .007 .018 .050 .135 .368 1.000 2.718 7.389 -5.000 -4.000 -3.000 -2.000 -1.000 0.000 1.000 2.000 Logit Scale Note: S = staying, U = moving up, F = taking family leave, O = moving out, and I = moving internally. Connecting lines indicate the absence of statistically significant differences between the effects of male (i.e., sex) on the odds of moving to those two occupational outcomes. Vertical distance in the figure has no meaning and only provides room to more easily distinguish the lines between effects. Turning to the final model presented in Table 5, the determinants of mobility out of teaching are as one might expect. Possession of advanced degrees and formal educational training in educational administration and 892 counseling are among the best predictors of movement upward to become a principal or assistant principal. For leave to care for family, the most important factors are family characteristics (especially having a new child), sex, and age, an effect which varies in size and trajectory depending upon whether one teaches at the elementary or secondary level. Of particular interest is the effect of sex on leaves to care for family: Men are exceedingly unlikely to take family leave. Looking at Table 5 and Figure 1, we can see that women are 35 times more likely than men to leave for family reasons (e3544 = 34.6). Together, the results presented in these tables provide no support for the tokenism hypotheses that men suffer particular disadvantages in predominantly female occupations. . [T]he multiplicative term for a sex by school level interaction failed a test of statistical significance in preliminary models. . [A]dding the sex by school level interaction term contributes little to the final model presented in Table 5 and does not improve it in any significant way. This result, of course, suggests that the effect of sex on mobility among teachers is not conditioned by level: Whether men leave teaching for an occupation outside a school does not seem to vary depending upon the level at which they teach, and whether elementary school teachers leave teaching for an occupation outside a school does not seem to vary depending on teachers' sex. This being the case, I find no support for Hypothesis 2; despite their token status, there is no evidence that men who teach in elementary schools are any more likely than men who teach in secondary schools to leave teaching for another occupation. Though I also use the sex by school level interaction to evaluate Hypothesis 1, the absence of this interaction is not enough in and of itself to dismiss the Hypothesis. It could be that men are more likely than women at either level to leave teaching for paid work outside a school; in other words, there could be a direct effect of sex on outward movement among teachers—one that applies for elementary and secondary school teachers alike. Looking at ... [Table] 5, however, there is no evidence of disproportionate outward movement among men. The effect of sex on outward movement fails a test of statistical significance, and thus I find no support for Hypothesis 1. In the absence of significant effects for these terms, the findings provide no convincing evidence that token men who teach in elementary schools suffer any disadvantages that would encourage them to leave teaching with any greater likelihood than either their majority female colleagues or the nontoken men who teach in secondary schools. Indeed, .the effect of sex, though not significant, is in a direction opposite that predicted by gender-neutral tokenism [results not shown]. The combination of nonsignificance and an effect for sex opposite of that predicted by Hypothesis 2 makes me suspicious about the generalizability of the gender-neutral tokenism perspective. . So, too, must we summarily find that there is no support for Hypotheses 893 6 and 7. In essence, these two hypotheses are the inverse of Hypotheses 1 and 2, and they are thus predicated on the same notion that there is a significant interaction between teachers' sex and the school level at which they teach or, for the case of Hypothesis 6, that there is a significant direct effect of sex on outward movement. As mentioned in the discussion of Hypotheses 1 and 2, however, there is no evidence of such an interaction effect ... or of a direct sex effect on outward movement among teachers (Table 5). Again, to the extent that there is a difference in the effect of sex across level, the difference is more consistent with the competition perspective than with the Kanterian perspective: Men appear to have slightly higher odds than women of moving out of secondary schools (relative to staying), and men appear to have slightly lower odds than women of moving out of elementary schools (relative to staying). However, the evidence is not persuasive (i.e.,p > .05). The only hypothesis for which these results provide convincing support is Hypothesis 8, the proposition that token men who teach in elementary schools will be more likely than women to move up to administrative positions. Table 5 and Figure 1 show that, controlling for other mobility-relevant variables, there is a significant direct effect of being male on moving up to administrative positions (vis-à-vis staying in teaching) in both elementary and secondary schools. Controlling for other relevant factors, the log-odds of moving up (relative to staying) are 1.138 greater for men than for women. This translates into significantly greater odds of promotion for male teachers: Men are 3.1 times more likely than women to move up. Thus, we can say with confidence that there is a greater likelihood of men being promoted upward in school settings and that this general finding applies to the token men who teach in elementary schools. This being the case, these data do provide convincing support for Hypothesis 8: There is evidence that male elementary school teachers are more likely than female elementary school teachers to move up to a principalship or assistant principalship.6 Discussion and Conclusion Despite the importance of understanding men's relationship to workplace sex composition (Reskin, 1988), we have until recently had little evidence bearing on the appropriateness of applying gender-neutral social control propositions to token men's experience with sex-atypical work. In this study I make use of a large set of systematically collected, representative data on teachers in the United States, and I find no evidence whatever that men suffer disadvantages as sex-atypical workers to an extent that would lead them to leave the field. There is no evidence that men move out of the 894 predominantly female occupation of elementary school teaching at a greater rate than women, and there is no evidence that male elementary school teachers move out at a greater rate than men who teach in the sex-integrated occupation of secondary school teaching—both things we would expect to find if men suffered some unique disadvantage in predominantly female settings. Nor do I find compelling evidence that men suffer disadvantages in sex-integrated workplaces, as is predicted by the competition perspective. What I do find strong evidence for, however, is Williams's (1992) alternative proposition: There does appear to be a glass escalator upward for token men who teach. Despite any disadvantages that may exist for men who teach in elementary schools, they benefit in one real, important way from 6 It should be noted that this analysis is based on a conceptualization of staying that includes job movers (i.e., movers from one school to another) as stayers. This conceptualization would not pose a problem, since I am interested in occupational moves, except that there is a small but noteworthy minority of stayers who remained in teaching but switched level. Given dataset inconsistencies in variable construction from Time 1 to Time 2, it is difficult to identify the teachers who made such a move, but based on roughly equivalent measures that I was able to construct, I identified only 57 teachers who stayed in the teaching occupation but switched from the elementary to secondary level, and I identified another 74 teachers who stayed in teaching but switched from the secondary to elementary level. It is inappropriate to model these additional outcomes in a pooled model similar to that presented in Table 5, but I did conduct a supplementary analysis where I estimated (1) an elementary-only model with a dependent variable that accommodated elementary-to-secondary moves and (2) a secondary-only model with a dependent variable that accommodated secondary-to-elementary moves. Though these specifications do change the significance levels for some variables, this analysis reveals the same effects for the central theoretical variables and interactions and would lead to the same conclusions regarding all eight hypotheses I test in this study. their status as men: They are significantly more likely to advance upward into prominent school administrative positions. Although it is possible that this effect could be the result of unmeasured gender differences in aspirations to become principals, evidence on the effects of gender and organizational structure on promotion aspirations makes me dubious of this as a complete explanation. Cassirer and Reskin (2000) provide convincing evidence that men and women in comparable organizational structures place the same importance on promotion. If women's aspirations to become principals are lower, this could be the result of an occupational structure that encourages men's aspirations with a history of placing them in headships rather than a result of sex, per se. Moreover, to the extent that pursuit and completion of requirements for administrative credentials reflects aspirations to become a principal or assistant principal, these models statistically control for gender differences in aspirations. Among elementary school teachers men are about two times more likely to have such credentials (not presented), but even when controlling for administrative credentials men are still three times more likely than women to move up. Thus, although gender differences in aspirations may contribute some to men's disproportionate upward movement, I think there is compelling evidence here that men benefit from a chauvinist advantage when it comes to promotion in 895 female-dominated workplaces. If the tokenism thesis is viable at all, the evidence here points to an effect of relative numbers on workplace experience that is conditioned by sex, one that is not gender-neutral, and it thus suggests the need for theoretical revision: While token workers suffer some forms of disadvantage by virtue of contextual minority status, that disadvantage is ameliorated when incumbents enjoy a privileged master status, such as men do. Under this condition token workers may be singled out for special treatment, better monetary and nonmonetary rewards, and better opportunities for career advancement. However, the results presented here, I think, point to a workplace experience of women and men that has less to do with relative numbers than with gender as a major structural stratification mechanism that privileges men in settings of whatever composition. Though tokens may experience some role encapsulation and added attention by virtue of their contextual difference, the kind of attention they get is determined by a gendered organizational structure in which men are more likely to be appreciated and rewarded than women. Thus, men were no more likely in 1992 to have left the gender-skewed occupation of elementary school teaching than the gender-balanced occupation of secondary school teaching, and their better chances of advancement were not different across levels either. This evidence is consistent with a larger feminist gender theory from which the glass ceiling and glass escalator arguments have emerged (Lorber, 1994; Acker, 1990; Reskin, 1988). Although these analyses do not provide any evidence that the tokenism proposition (Kanter, 1977) is generalizable to token men in predominantly female workplaces, they also do not rule out the possibility that token men suffer the alienation and self-doubt identified in previous studies; despite my findings, those things may still hold, and those factors could reduce token men's satisfaction at work without pushing them out. Token men could well stay in teaching while enduring such difficulties, but it is unclear to me why they would do so—why they would not leave with greater rapidity than women who teach elementary school—if they do not also benefit in some other ways that assuage the factors that might lead to dissatisfaction. In addition to greater chances at promotion, male elementary school teachers also make more money than women in the profession (women's base salary is about 90 % of men's; U.S. Department of Education, 1994a), and given this power and privilege, perhaps men enjoy other day-to-day benefits as well. Our theoretical understanding would benefit from research exploring the link between token men's workplace satisfaction and the effect of dissatisfaction with colleagues, supervisors, and professional respect on attrition 896 versus advancement. For instance, we don't know the extent to which token men are satisfied or dissatisfied with various aspects of their work, and the extent to which that dissatisfaction leads them up as an alternative to moving out. Research exploring such issues should be among the next steps in our study of the processes surrounding token men's workplace experience. The implication of these conclusions for our understanding of workplace sex segregation within teaching is fairly clear. Although I offer no real evaluation of the extent to which sex-differentials in attrition across elementary and secondary schools contribute to sex segregation within teaching, these findings suggest that men's attrition cannot contribute much to that segregation. In order for men's attrition to contribute to overall segregation between elementary and secondary school teachers, men's attrition would have to be significantly more likely in elementary school teaching than in secondary school teaching. It is not: Occupational level makes no significant difference in men's likelihood of moving out of teaching. Though men are more likely to leave teaching to move up, this happens for secondary and elementary school teachers alike and doesn't happen enough in either case to account for the drastically different sex composition seen across levels. Instead of attrition, focusing on entrance into teaching will, in all likelihood, be the most rewarding direction for future research on sex segregation within teaching and the larger occupational structure. Based on results from these data, it appears clear that sex segregation across levels within teaching is largely attributable to something that occurs before men enter the profession, discouraging all but a very few from even pursuing elementary school teaching. Jacobs (1989) has already offered evidence that very few men aspire to work as elementary school teachers (see Table 4.1), and Montecinos and Nielsen (1997) provide some additional evidence bearing on the aspirations of young men as they progress through a collegiate elementary education program as well as the time at which such men decide to pursue the occupation (men apparently decide much later than women; see also Allan, 1996). What is now needed is research that examines why this is the case. Most likely, premarket forces that matter are the combination of socialization, college counseling, and other gendered patterns of social control that discourage entrance into this occupation. Such research could improve our understanding of masculinity and boys' and men's general reluctance to pursue jobs seen as "women's work." In addition, the research agenda for this area should also include comparable studies of men in other predominantly female occupations and jobs. Arguably, elementary school teaching provides a site for a relatively conservative test of token theories; along with nursing, elementary school teaching is among the occupations most strongly associated with women and femininity. To the extent that elementary school teaching is among those occupations most strongly associated with women, we would expect 897 that it would be among those most likely to witness male disadvantage predicted by the gender-neutral token thesis. That we do not find evidence of such disadvantage suggests that one would not find it in other predominantly female professions that are not as strongly associated with women and femininity. Nonetheless, the generalizability of the token propositions to other predominantly female occupations and jobs remains an empirical question, and it should be empirically evaluated in other settings. Theoretical understanding may benefit most from research on occupations other than the professions and on jobs where opportunity for advancement is severely limited or nonexistent. It may be that class, training, commitment, and labor process differences between professional and nonprofessional occupations and jobs result in different experiences for token men who occupy them. Investigations into such differences would provide important information as we work toward both a better understanding of the experience of token men across the great diversity of occupations and jobs and a better understanding of the role of that experience in determining workplace sex segregation. REFERENCES Acker, Joan (1990): Hierarchies, Jobs, Bodies: A Theory of Gendered Organizations. Gender and Society (4): 139-58. Allan, James A. (1996): Gender as Opportunity and Constraint in the Work Lives of Men Elementary School Teachers. Ph.D. dissertation, Department of Education, University of Iowa, Iowa City, IA. Allan, Jim (1993): Male Elementary Teachers: Experiences and Perspectives. In Christine L. Williams (ed.), Doing "Women's Work": Men in Nontraditional Occupations, 113-27. Newbury Park, CA: Sage Publications. Atwater, Leanne E. and David D. Van Fleet (1997): Another Ceiling?: Can Males Compete for Traditionally Female Jobs? Journal of Management (23): 603-26. Biedenkapp, Mildred S. and Jacob D. Goering (1971): How 'Masculine' Are Male Elementary Teachers? Phi Delta Kappan (53)(October): 115. Bielby, William T. and James N. Baron (1986): Men and Women at Work: Sex Segregation and Statistical Discrimination. American Journal of Sociology (91): 75999. Bielby, William T. and Denise D. Bielby (1989): Family Ties: Balancing Commitments to Work and Family in Dual Earner Households. American Sociological Review (54): 776-89. -(1992): I Will Follow Him: Family Ties, Gender-Role Beliefs, and Reluctance to Relocate for a Better Job. American Journal of Sociology (97): 1241-67. Blalock, Hubert M., Jr. (1967): Toward a Theory of Minority-group Relations. New York: John Wiley and Sons, Inc. Blau, Peter M. (1977): Inequality and Heterogeneity: A Primitive Theory of Social 898 Structure. New York: The Free Press. Budig, Michelle J. (2002): Male Advantage and the Gender Composition of Jobs: Who Rides the Glass Escalator? Social Problems (49): 258-77. Cassirer, Naomi and Barbara Reskin (2000): High Hopes: Organizational Position, Employment Experiences, and Women's and Men's Promotion Aspirations. Work and Occupations (27): 438-63. Cockburn, Cynthia (1983): Brothers: Male Dominance and Technological Change. London: Pluto Press. --(1991/1993): In the Way of Women: Men's Resistance to Sex Equality in Organizations. Ithaca, NY: ILR Press. Cohen, Jacob and Patricia Cohen (1983): Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences. 2nd ed. Hillsdale, NJ: Erlbaum. DeCorse, Cynthia J. Benton and Stephen P. Vogtle (1997): In a Complex Voice: The Contradictions of Male Elementary Teachers' Career Choice and Professional Identity. Journal of Teacher Education (48): 37-46. England, Paula (1992): Comparable Worth: Theories and Evidence. New York: Aldine de Gruyter. England, Paula and Melissa S. Herbert (1993): The Pay of Men in 'Female' Occupations: Is Comparable Worth Only for Women? In Christine L. Williams (ed.), Doing "Women's Work": Men in Nontraditional Occupations, 28-48. Newbury Park, CA: Sage Publications. Ferree, Myra Marx (1990): Beyond Separate Spheres: Feminism and Family Research. Journal of Marriage and the Family (52): 866-84. Floge, Liliane and Deborah M. Merrill (1986): Tokenism Reconsidered: Male Nurses and Female Physicians in a Hospital Setting. Social Forces (64): 925-47. Fottler, Myron D. (1976): Attitudes of Female Nurses Toward the Male Nurse: A Study of Occupational Segregation. Journal of Health and Social Behavior (17): 98-110. Heikes, E. Joel (1991): When Men are the Minority: The Case of Men in Nursing. The Sociological Quarterly (32): 389-401. Jacobs, Jerry (1989): Revolving Doors: Sex Segregation and Women's Careers. Stanford, CA: Stanford University Press. -(1993): Men in Female-Dominated Fields: Trends and Turnover. In Christine L. Williams (ed.), Doing "Women's Work": Men in Nontraditional Occupations, 49-63. Newbury Park, CA: Sage Publications. Jacobs, Jerry A. and Brian Powell (1984): Gender Differences in the Evaluation of Prestige. The Sociological Quarterly (25): 173-90. Kanter, Rosabeth Moss (1977/1993): Men and Women of the Corporation. New York: Basic Books. King, James R. (1998): Uncommon Caring: Learning from Men Who Teach Young Children. New York: Teachers College Press. Long, J. Scott (1987): A Graphical Method for the Interpretation of Multinomial Logit Analysis. Sociological Methods and Research (15): 420-46. --(1997): Nominal Outcomes: Multinomial Logit and Related Models. In Regression Models for Categorical and Limited Dependent Variables, 148-86. Thousand Oaks, CA: Sage Publications. Lorber, Judith (1994): Paradoxes of Gender. New Haven, CT: Yale University Press. Marini, Margaret M. (1989): Sex Differences in Earnings in the United States. Annual 899 Review of Sociology (15): 343-80. Maume, David J. (1999): Glass Ceilings and Glass Escalators: Occupational Segregation and Race and Sex Differences in Managerial Promotions. Work and Occupations (26): 483-509. Montecinos, Carmen and Lynn E. Nielsen (1997): Gender and Cohort Differences in University Students' Decisions to Become Elementary Teacher Education Majors. Journal of Teacher Education (48): 47-54. Murnane, Richard J., Judith D. Singer, John B. Willett, James J. Kemple, and Randall J. Olsen (1991): Who Will Teach?: Policies That Matter. Cambridge, MA: Harvard University Press. Murray, Susan B. (1996): 'We All Love Charles': Men in Child Care and the Social Construction of Gender. Gender and Society (10): 368-85. Pierce, Jennifer L. (1995): Gender Trials: Emotional Lives in Contemporary Law Firms. Berkeley: University of California Press. Reskin, Barbara F. (1988): Bringing The Men Back In: Sex Differentiation and the Devaluation of Women's Work. Gender and Society (2): 58-81. Reskin, Barbara F. (1993): Sex Segregation in The Workplace. Annual Review of Sociology (19): 241-70. Reskin, Barbara F. and Patricia A. Roos (1990): Job Queues, Gender Queues: Explaining Women's Inroads into Male Occupations. Philadelphia, PA: Temple University Press. Rosenfeld, Rachel A. (1984): Job Changing and Occupational Sex Segregation: Sex and Race Comparisons. In Barbara F. Reskin (ed.), Sex Segregation in the Workplace: Trends, Explanations, Remedies, 56-86. Washington, DC: National Academy Press. Sargent, Paul (2000): Real Men or Real Teachers: Contradictions in the Lives of Men Elementary Teachers. Men and Masculinities (2): 410-33. Schlechty, Phillip C. and Victor S. Vance (1983): Recruitment, Selection, and Retention: The Shape of the Teaching Force. Elementary School Journal (83): 469-87. Skelton, Christine (1994): Sex, Male Teachers and Young Children. Gender and Education (6): 87-93. Talbert, Joan E. (1986): The Staging of Teachers' Careers: An Institutional Perspective. Work and Occupations (13): 421-43. U.S. Bureau of the Census (1992): 1990 Census of the Population and Housing: Public Use Microdata Samples. 1 % weighted subsample. Online. Ulysses interactive cross-tabulation system, distributed by the Consortium for International Earth Science Information Network. July 1996. world wide web: http://ts2. ciesin.org/plue/ulysses/. U.S. Department of Education, National Center for Education Statistics (1994a): Schools and Staffing Survey, 1990-91: [United States]. Washington, DC: U.S. Department of Commerce, Bureau of the Census [producer]. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor]. U.S. Department of Education, National Center for Education Statistics (1994b): Schools and Staffing Survey, 1990-91 [United States]: Teacher Followup Survey, 1991-92. Washington, DC: U.S. Department of Commerce, Bureau of 900 the Census [producer]. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor]. Wharton, Amy S. and James N. Baron (1987): So Happy Together? The Impact of Gender Segregation on Men at Work. American Sociological Review (52): 574-87. --(1991): Satisfaction? The Psychological Impact of Gender Segregation on Women at Work. The Sociological Quarterly (32): 365-87. Williams, Christine L. (1989): Gender Differences at Work: Women and Men in Nontraditional Occupations. Berkeley: University of California Press. --(1992): The Glass Escalator: Hidden Advantages for Men in the 'Female' Professions. Social Problems (39): 253-67. -(1993): Introduction. In Christine L. Williams (ed.), Doing "Women's Work": Men in Nontraditional Occupations, 1-9. Newbury Park, CA: Sage Publications. -(1995): Still a Man's World: Men Who Do "Women's Work." Berkeley: Univer- sity of California Press. Williams, Susan L. and Wayne J. Villemez (1993): Seekers and Finders: Male Entry and Exit in Female-Dominated Jobs. In Christine L. Williams (ed.), Doing "Women's Work": Men in Nontraditional Occupations, 64-90. Newbury Park, CA: Sage Publications. Yoder, Janice D. (1991): Rethinking Tokenism: Looking beyond Numbers. Gender and Society (5): 178-92. Yoder, Janice D. and Laura M. Sinnett (1985): Is It All in the Numbers?: A Case Study of Tokenism. Psychology of Women Quarterly (9): 413-18. Zimmer, Lynn (1988): Tokenism and Women in the Workplace: The Limits of Gender-Neutral Theory. Social Problems (35): 64-77.