PART TWO: CO-CONSTRUCTION THEORY AND ANALYSIS
4.1 Theoretical Background
As discussed in the introductory chapter, there are forceful normative trends of choosing career paths for self-expressive reasons. I illustrated in chapter 2 the
interconnection between perceptions of the gender structure and perceptions of self—
namely, respondents‘ gender schemas significantly predict their self-conceptions. If perceptions of self significantly predict the extent of sex-segregation in the
post-graduation career choices men and women make, then I will have identified a mechanism by which sex-segregation is reproduced at the individual-level—reproduction within a culturally-legitimated realm of individualistic self-expression.
Career Launch and Occupational Sex Segregation
This chapter seeks to find out whether the trends identified between
self-conceptions and respondents‘ contributions to sex-segregation continue once they have left college and entered the workforce or graduate school. The individual-level forces that reproduce sex segregation in college majors are somewhat similar to those that reproduce segregation in the workforce, and the segregation in the former certainly contributes to segregation in the latter (Jacobs 1989). While there is some slippage that allows one to enter different occupations than that in which he or she received an
undergraduate education, the specialization inherent in a college degree necessarily limits the types of occupations and graduate programs he or she is qualified to enter. A college graduate with a degree in art history cannot get a job as an engineer, for example. For these reasons, college majors are an important factor in the reproduction of occupational sex segregation.
However, the career decisions men and women make just after leaving college have effects that arguably extend throughout their careers (Jacobs, 1989). Once one has established a career trajectory, career track changes are possible, but costly in terms of time, income, and job prestige. Additionally, due to the differential valuation of female-dominated versus male-female-dominated occupations (Charles & Grusky, 2004; England,
1984), if one chooses a female-dominated career track, they have entered a bracket of lower-paid, less prestigious jobs. Individuals in these occupations may certainly end up in jobs that are far above the median income for male-dominated occupations, women and men working in fedominated jobs make, on average, less than people in male-dominated jobs (England, 1984).
I call this early stage of career decisions ―career launch.‖ While career launch could encompass the first several years of one‘s career, I study the sex-typing of men‘s and women‘s early occupations or of the graduate school program in which they have enrolled eighteen months after they have graduated. This time lag is purposeful. The first year after college graduation is a period of adjustment (Astin, 1993); some men and women take a year off from employment to ―figure out what they want to do with their lives.‖ Others might work a temporary job while they search for a more long-term position in their field of interest. The respondents in my study also hit the beginning of the 2007-2008 economic downturn as they exited college, likely making entrance into graduate school programs more competitive and job searches a multi-year process for many. I exclude from this analysis the four respondents who were either traveling or looking for a job at the time of the year 5 survey. The lag between graduation and the time in which I measure career launch allows for this volatility to stabilize, while not allowing enough time to elapse that they have the opportunity to make another large career move.
To the extent that they have freedom to make decisions about their trajectories at career launch, I expect men and women to make such decisions along self-expressive lines. As it was for college major selection, self-expression is an important normative
consideration for decisions about one‘s career: young men and women are still expected to give self-expressive answers to the question, ―what will you do with your degree?‖
I am particularly interested in two aspects of the career launch decision process in this chapter: whether respondents choose female-dominated or male-dominated
occupations after graduation, and the extent to which men and women move into more male-dominated or more female-dominated occupations than that the fields in which they graduated. As in the previous chapter, I use a measure I call respondents‘ career launch sex-segregation score. Instead of a dichotomous variable predicting whether or not respondents enter a female-dominated (or male-dominated) occupation, the
sex-segregation score captures the percent female of the occupation respondents choose. This is a purely demographic measure of sex-segregation: the cultural sex-typing of
occupations is not necessarily completely determined by the representation of men and women therein. However, it is a useful indication of whether respondents‘
self-conceptions influence where they locate themselves along the spectrum of
sex-segregated occupations. It is in indication as well of their individual-level contribution to reproducing or undermining occupational sex-segregation.
4.2 Hypotheses
To investigate the possibility that gendered self-expression leads to gendered decisions at career launch, I examine the effects of feminine, unsystematic, and people-oriented self-conceptions on the probability that men and women will choose male-dominated or female-male-dominated career tracks. I use several types of self-conceptions in an attempt to show that any effects I find are likely not specific to a narrowly-defined type of self-conception. Again, my purpose is to show examples of mechanisms by
which conceptions influence career launch decisions. These are not the only self-conceptions that influence career launch; rather they are useful exemplars of substantive categories of self-beliefs. If I find an effect of self-conceptions here, it is likely there are many other self-conceptions that follow the same pattern.
I expect self-expressive patterns of career selection to follow those identified in chapter 3; namely, that students with self-conceptions traditionally stereotyped as feminine will be more likely to enter female-dominated career launch paths and be more likely to move into an even more female-dominated field after college graduation.
Similar patterns should exist for unsystematic and people-oriented self-conceptions.
This chapter tests the assumption that respondents‘ career launch decisions are made, in part, as gendered expressions of self. The analysis will help determine whether these gendered self-concepts contribute to the reproduction of occupational sex
segregation.
Although self-expression may seem to be a likely determinant of career launch decisions, it is plausible that self-conceptions are not significant predictors of such decisions. Given the economic situation during which these young men and women entered their careers, and the structural constraints imposed on people just emerging from their undergraduate education, these constraints may simply be too strong to allow men and women much room to make decisions along self-expressive lines. Or, their decisions may not be patterned in such a way that self-conceptions stereotypically associated with women lead respondents to choose more female-dominated occupations (and self-conceptions associated with men lead respondents to choose male-dominated
occupations). If I find that self-conceptions are significantly predictive of sex-segregation
scores and the change in sex-segregation scores between degree and career launch, it is likely that such effects would be identified in larger samples with more detailed measures of self-conceptions.
4.3 Methods
The dependent variable is a measure of the percent women in respondents‘ post-college occupations or graduate school programs on a scale ranging from 0-100%
women. A positive coefficient predicting this scale means that respondents with that characteristic are more likely to choose female-dominated occupations or graduate programs; negative predictor coefficients would imply that characteristic lead
respondents to choose career launch paths that are more male-dominated. I identified a sex-segregation score for each person by matching their detailed field of work or study (identified by year 5 survey questions) with national statistics of the percent women in each of these occupations. For those in the workforce, I referenced U.S. Department of Labor‘s Bureau of Labor Statistics to find the percent women in respondents‘ field of work. I used statistics computed by the National Science Foundation‘s Division of Science Resource Statistics for the percent women in science and engineering-related graduate programs49 and the National Center for Education Statistics‘ Digest of Educational Statistics for the percent women in non-science or engineering graduate programs.50
I use a measure of the change in respondents‘ sex-segregation scores in descriptive models. To create this variable, I subtracted the percent female in their
49 NSF detailed majors data: http://www.nsf.gov/statistics/wmpd/pdf/tabc-5.pdf
50 Digest of Educational Statistics:
http://nces.ed.gov/programs/digest/d09/tables/dt09_286.asp
undergraduate degree field from the percent female in their career launch field.51 People with a positive change score moved into more female-dominated field than the field in which they earned their degree. People with a negative change score switched into a more male-dominated field than the one in which they graduated. I model change in the demographic analysis and the causal models by including respondents‘ sex-segregation scores of their degrees in the structural equation models predicting sex-segregation scores in their career launch occupations (see below for a description of my analytic strategy).
My key independent variables are the year 2 latent measures of feminine self-conceptions (introduced in chapter 2) and unsystematic and people-oriented self-conceptions
(introduced in chapter 3).
All models control for the following race/ethnicity and school measures: whether respondents identify as African-American (yes=1), Hispanic or Latino (yes=1) or Asian or Asian-American (yes=1), or non-Hispanic white (yes=1; reference category); and whether they attended MIT (yes=1), Olin (yes=1), Smith (yes=1) or UMass (yes=1;
reference category). I also control for whether respondents entered graduate school or the workforce after graduation (1=entered graduate school; 0=entered the workforce). The demographic models in Table 4.2 measure a more complete list of characteristics,
including respondents‘ gender (women=1)52, whether respondents were born in the U.S.
(yes=1), their family‘s income (in dollars), whether they identify as gay, lesbian or
bisexual (yes=1), their political conservatism (1=very liberal to 7=very conservative), and how they rate their religiosity compared to their peers (1=lowest 10% to 5=highest 10%).
51 See Appendix 2 and chapter 3 for information on the calculation of the sex-segregation scores for respondents‘ undergraduate degrees.
52 The structural equation models are ran separately for men and women, so they do not include a control for gender.
Table 4.1 presents descriptive statistics for the mean sex-segregation scores and the change in respondents‘ sex-segregation scores between college and career launch and Table 4.2 lists the most common occupations among those respondents who entered the workforce. I use OLS regression in Table 4.3 and 4.4 to determine the demographic predictors of sex-segregation scores and change in sex-segregation scores over time. The remainder of the analysis in this chapter uses structural equation modeling. I remind the reader that latent variables are meant to represent overarching concepts, components of which are captured by the manifest variables predicted by the latent variable. The benefit of latent variables is that not all possible measures of this concept must be included as manifest variables for the concept to be adequately represented by the latent variable (Byrne, 2010). Table 4.5-4.6 present the structural model (the relationships between the latent independent variable and the dependent variable) and the controls. Table 4.7 uses OLS regression to predict the effect of people-oriented self-conceptions on sex
segregation scores. All structural and OLS models presented in this chapter control for school attended, race/ethnicity, and whether respondents entered the workforce or a graduate school program. (No other demographic factors were significant predictors of sex-segregation scores in the demographic models, thus I did not include them in the models presented below.)
The next section details the demographic trends of respondents‘ career launch paths. The following section presents the structural equation and OLS models predicting percent women in respondents‘ career launch paths and the changes in sex-segregation scores between their degrees and their occupations or graduate school programs.