Sociology of Education The Gender Income Gap and the Role of Education On behalf of: American...

http://soe.sagepub.com/Sociology of Education

http://soe.sagepub.com/content/80/1/1The online version of this article can be found at:

DOI: 10.1177/003804070708000101

2007 80: 1Sociology of EducationDonna Bobbitt-Zeher

The Gender Income Gap and the Role of Education

Published by:

http://www.sagepublications.com

On behalf of:

American Sociological Association

can be found at:Sociology of EducationAdditional services and information for

http://soe.sagepub.com/cgi/alertsEmail Alerts:

http://soe.sagepub.com/subscriptionsSubscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.com/journalsPermissions.navPermissions:

http://soe.sagepub.com/content/80/1/1.refs.htmlCitations:

What is This?

- Jan 1, 2007Version of Record >>

at Macquarie University Library on November 13, 2013soe.sagepub.comDownloaded from at Macquarie University Library on November 13, 2013soe.sagepub.comDownloaded from

http://soe.sagepub.com/


http://soe.sagepub.com/content/80/1/1

http://soe.sagepub.com/content/80/1/1



http://www.asanet.org

http://www.asanet.org

http://soe.sagepub.com/cgi/alerts

http://soe.sagepub.com/cgi/alerts

http://soe.sagepub.com/subscriptions

http://soe.sagepub.com/subscriptions

http://www.sagepub.com/journalsReprints.nav

http://www.sagepub.com/journalsReprints.nav

http://www.sagepub.com/journalsPermissions.nav

http://www.sagepub.com/journalsPermissions.nav

http://soe.sagepub.com/content/80/1/1.refs.html

http://soe.sagepub.com/content/80/1/1.refs.html

http://soe.sagepub.com/content/80/1/1.full.pdf

http://soe.sagepub.com/content/80/1/1.full.pdf

http://online.sagepub.com/site/sphelp/vorhelp.xhtml

http://online.sagepub.com/site/sphelp/vorhelp.xhtml





The Gender Income Gap and the Role ofEducation

Donna Bobbitt-ZeherThe Ohio State University

Education is thought to be the pathway to success for disadvantaged groups. Given that

young women now match or surpass men’s educational achievements on many measures,

how do they fare in terms of equal earnings? Would further educational changes matter for

closing any existing gap? Analyzing data from the National Educational Longitudinal Survey,

the author found that college-educated men in their mid-20s already earn, on average, about

$7,000 more per year than do college-educated women. The findings suggest that this gap

would still be substantial--about $4,400 per year--if women and men had similar educational

credentials, scores on standardized tests, fields of study, and degrees from colleges of similar

selectivity. Although women’s gains in education may have been central to narrowing the gen-

der gap in income historically, gender differences in fields of study continue to disadvantage

women. Moreover, gender differences in work-related factors are more important than are

educational differences for understanding contemporary income inequality among young

workers.

Sociology of Education 2007, Vol. 80 (January): 1–22 1

Given the obvious connection betweeneducational success and labor marketoutcomes, many consider education

to be key to reducing group inequalities. Inparticular, schooling is thought to play a piv-otal role in the success of racial/ethnicgroups, such as Asian Americans, and thecontinuing struggles of others (e.g., blacks,Native Americans, and Hispanics). But whatrole does education play in lessening genderdisparities in the larger society? Withwomen’s educational attainment andachievement patterns now matching or sur-passing those of men on many measures, arewomen on their way toward gender equalitymore broadly?

The literature has documented theunprecedented success of women in theclassroom (e.g., National Center forEducation Statistics, NCES 2005) and hassuggested that education may have played

an important role in reducing gender wagegaps over the past few decades (e.g., Gill andLeigh 2000; Loury 1997). Some studies haveisolated educational influences on persistinggender wage gaps, most notably gender dif-ferences in skills (e.g., Farkas et al. 1997), col-lege majors (e.g., Bradley 2000), and collegeselectivity (e.g., Davies and Guppy 1997),while others have concentrated on nonedu-cational influences on wage disparities, suchas family formation and patterns of labormarket participation (e.g., Kilbourne,England, and Beron 1994; Marini and Fan1997). Yet few studies have systematicallyanalyzed the mediating role of education ingender disparities in earnings, along with thepotentially confounding family and employ-ment factors. In addition, little attention hasbeen paid to the current level of gender dis-parities in earnings among the beneficiaries ofthese changing patterns of educational

at Macquarie University Library on November 13, 2013soe.sagepub.comDownloaded from



2 Bobbitt-Zeher

accomplishment or to the prospects of edu-cation contributing to further reductions insuch inequalities should the patterns ofwomen’s educational success persist.

This article, drawing from and extendingprior work, addresses the role of education inmediating gender inequality in earnings foryoung college-educated workers today. Suchanalyses are important because they empirical-ly test both the level of contemporary genderdisparities in earnings and the potentially equal-izing influence of education among those whoare the most likely to benefit from recent edu-cational changes. I begin by discussing chang-ing educational patterns for women and theirpotential implications for gender inequality inthe contemporary labor market. I then turn tothe potentially confounding role of familyresponsibilities and workplace stratification. Theanalyses, which draw from the NationalEducational Longitudinal Survey of 1988 (here-after NELS), assess the degree to which educa-tion mediates gender inequality in income foryoung college-educated workers. I conclude bydiscussing the implications of the importantbut limited role of education in reducing earn-ings inequality for the prospects of broadergender equality.

BACKGROUND

Consequences of ChangingEducational Patterns

While scholars in the 1970s and 1980s high-lighted ways in which schools shortchangedgirls (e.g., Sadker and Sadker 1994), the focushas shifted in the past decade because youngwomen now outperform young men onmany indicators of educational achievement(see NCES 2005; Riordan 2003). Not only arewomen enrolling in college in greater num-bers than men, they are also outpacing menin graduating from high school, attendingcollege, and attaining college degrees (NCES2004; Sum, Fogg, and Harrington 2003). Inaddition, gender gaps in enrollment anddegree attainment favoring women areexpected to widen further in the next decade(NCES 2003a:6; Sum et al. 2003).

These dramatic and well-documented

educational changes have been fodder formuch public debate regarding a “war againstboys” (see Riordan 2003); however, thepressing issue that has been overlooked in theliterature is the degree to which the educa-tional success of young women today leads togender equality later in life. In particular,there has been little research on the implica-tions of changing patterns in higher educa-tion for gender equality in labor market out-comes. A notable exception is Loury’s (1997)work, which suggested that women’s educa-tional success had a direct effect on the nar-rowing gender gap in earnings in the early tomid-1980s. In particular, Loury concludedthat gendered changes related to collegegrades and fields of study explain virtually allthe 6-percentage-point decrease in the gen-der gap in earnings for college graduatesfrom 1979 to 1986.1

Such work connecting women’s educa-tional accomplishments to the declining gen-der gap in income suggests an optimistic pic-ture for young, college-educated women atthe time of their entry into careers. In partic-ular, the pattern of women’s increasing par-ticipation in higher education, given theincreasing importance of a college educationfor labor market success (Cappelli et al. 1997;Farley 1995), suggests that young womenwho are entering careers may be well posi-tioned, finally, for gender equity in pay.

However, to what extent do young college-educated women reap equal returns in thelabor market? The weight of the empirical evi-dence suggests lingering inequality, withwomen earning about 15 percent less thanmen early in their careers (e.g., Blau 1998:129;Marini and Fan 1997). Yet other research hassuggested no gender gap in wages for youngengineers (Morgan 1998). Understanding thedegree of gender inequality in earnings in theearly years of careers is important because ini-tial income disparities tend to grow over time(see Marini 1989). And for those who are inter-ested in examining the equalizing effects ofeducation, it is during these early years of acareer—when differences in employment histo-ries, life experiences, and accumulated skills areminimized—that educational credentials andschool experiences are likely to matter themost.




Gender Income Gap 3

In addition to the need to determine themagnitude of the income differences for col-lege-educated men and women early in theircareers, a second question also looms large:Given the changing gendered patterns ofeducational success, to what extent do edu-cational factors contribute to any gender dis-parities in earnings for young workers today?I turn now to this question.

Educational Explanations for theGender Income Gap

The recent focus on women’s educationaladvances, which are both impressive anddeserving of scholarly attention, tends toobscure the ways in which women remaindisadvantaged on several educational mea-sures. Gender differences on four of thesemeasures, in particular, are implicated in thegender income gap: (1) choice of a collegemajor, (2) skills as measured by standardizedtests, (3) amount of education, and (4) selec-tivity of the college attended.

The most persuasive educational explana-tion of gender income inequality is thatwomen major in fields that lead to jobs that arenot rewarded with higher incomes (Bradley2000; Davies and Guppy 1997; Gerber andSchaefer 2004). Individuals who major insuch fields as engineering and computer sci-ence tend to earn more than do those whomajor in education and the humanities(Daymont and Andrisani 1984; Gerber andSchaefer 2004). However, in spite of thetrend toward the integration of fields ofstudy, college majors are still quite gendersegregated (see Bradley 2000; Charles andBradley 2002; Jacobs 1995, 1996). For exam-ple, women received 20 percent of the engi-neering degrees and 77 percent of the edu-cation degrees in 2000–01 (NCES 2004:78).Given that men are more concentrated in thehigher-earning fields and women are moreconcentrated in the less rewarded ones, gen-der segregation in fields of study appears tocontribute to gender differences in income.

Indeed, studies that have considered fieldsof study have found that the choice of collegemajors explains between roughly one-quarterto one-half of the gender gap in wages forcollege graduates (Brown and Corcoran

1997; Daymont and Andrisani 1984). Bradley(2000) concluded that this horizontal dimen-sion of gender segregation is pivotal in under-standing gender inequality in wages globally.

But what is it about some fields that leadto their being better compensated than oth-ers? Is it the content of the field or the gendercomposition that suggests its worth? It is pos-sible that “certain majors and courses maydevelop more valuable job-related humancapital than do other majors and courses”(Brown and Corcoran 1997:432, citing Paglinand Rufolo 1990). In this view, the labor mar-ket rewards this investment in human capitalwith higher earnings. Alternatively, the gen-der dominance of the major (i.e., percentagefemale) may be the most salient difference infield of study, since traditionally male fieldshave been rewarded more than have tradi-tionally female ones. While most studies havetested the impact of fields of study by usingdummy variables for majors, making it hardto know what about the majors is leading tothe earnings disparities, NCES (1998; see alsoJoy 2000) considered women’s proportionalrepresentation in each field. It found thatamong the college educated, workers withfemale-dominated majors averaged 20 per-cent less in annual earnings in the first yearafter graduation than did workers with male-dominated majors. Such research suggeststhat the gender composition of fields shouldnot be overlooked when considering why col-lege majors matter for gender disparities inincome.

Another education-related explanation forincome inequalities concerns gender differ-ences in cognitive skills. Measured using stan-dardized test scores, cognitive skills arethought to affect the gender gap directly aswell as indirectly through the choice of col-lege major and access to jobs (Farkas et al.1997; Paglin and Rufolo 1990). Research hassuggested that as the U.S. economy hastransformed since the 1970s, math and sci-ence abilities have become more predictive ofsalaries (Murnane, Willett, and Levy 1995),and math skills translate into higher earningsfor all types of workers (Mitra 2002). Indeed,Mitra’s (2002) study found that the gendergap in income disappears among profession-al men and women with the highest math




4 Bobbitt-Zeher

skills. Thus, the gender income differential is aresult of differences in highly valued skills—generally math skills—which lead to lower-paying jobs for women. Although evidencesuggests that differences in boys’ and girls’performance on standardized math and sci-ence tests are shrinking (Willingham and Cole1997), persistent differences on standardizedtests, including the SAT, favoring men(College Board 2003; NCES 2004) may con-tinue to play a role in gender disparities inincome, particularly in today’s economy, inwhich skills are increasingly predictive ofsalaries (Murnane et al. 1995).

While the strongest educational influenceson gender disparities in income are likely tobe gender segregation in college majors anddifferences in standardized test scores, twoadditional schooling-related factors may con-tribute to these disparities to a lesser extent.The first is the vertical dimension of gendersegregation,2 or the level of degree attain-ment. Although women now surpass men inundergraduate degrees that are awarded(NCES 2005), gender parity in the highestdegrees has yet to be realized. Womenreceive approximately 45 percent of all pro-fessional and doctoral degrees (NCES2004:82). Although this is not a large differ-ence, the reality that greater educationalattainment leads to higher wages for bothwomen and men (Blau 1998; Kilbourne et al.1994) suggests that men’s advantage inreceiving the highest degrees may contributein a small way to women’s lower averageearnings.

In addition, some scholars have suggestedthat women’s attendance at less selectiveschools contributes to women’s disadvan-taged position in the labor market. Collegeprestige has a positive relationship with earn-ings later in life (Jacobs 1999), and men aresignificantly more likely to attend selectivepostsecondary institutions than are women,net of background and academic factors(Davies and Guppy 1997, replicating the find-ings of Hearn 1991). Women’s attendance atless selective postsecondary educational insti-tutions may be the result of institutional biasfavoring men, more selective schools tendingnot to offer traditionally female-dominatedprograms, and/or parental choices to invest

more financially in sons (Davies and Guppy1997; see also Jacobs 1999). Although therehave been declines in gender differences inthe selectivity of postsecondary institutionattended (see Jacobs 1999; Karen 1991),institutional selectivity remains a potentiallysalient influence on gender disparities inincome for today’s college graduates.

With regard to young adults graduatingfrom college and entering the labor markettoday, these findings suggest that lingeringgender differences in schooling (e.g., fields ofstudy, measured cognitive skills, and collegeselectivity) may explain persistent gender dis-parities in earnings. With women’s postsec-ondary education rates surpassing men’s, par-ticipation in higher education is not likely tocontribute to earnings disparities, except atthe highest levels, where men’s degree attain-ment continues to surpass that of women.

Education’s Limited Role in theGender Income Gap

The literature just discussed has indicated thateducational forces contribute to gender dis-parities in earnings. However, noneducationalfactors related to family, employment, andaspirations—generally in the periphery ofstudies on the effects of education on incomedifferentials—also play a part in genderincome inequality. The effects of family for-mation, particularly marriage and parent-hood and their impact on participation inpaid labor, are implicated in gender incomedisparities. For example, net of other factors,such as education, women with childrenmake 10 percent to 15 percent less than dowomen without children (Korenman andNeumark 1992; Waldfogel 1998), and there isa 7 percent wage penalty for each child thata young woman has (Budig and England2001). The penalty for having children isgreater for married women than for nonmar-ried women (Budig and England 2001:218).The same patterns do not hold for men;fathers experience no comparable wagepenalty for their parental status (Waldfogel1998). Furthermore, married men receivehigher pay than do unmarried men, whilethere is some evidence of a wage disadvan-tage for married women (Kilbourne et al.




Gender Income Gap 5

1994; see, however, Budig and England 2001for evidence challenging a marriage penaltyfor women).

The impact of family formation on genderdifferences in earnings appears to operatethrough women’s decreased labor force par-ticipation (Korenman and Neumark 1992).Both length of job experience and part-timeemployment contribute to lower earnings(Budig and England 2001; Shelton andFirestone 1989). And women historically havehad less job experience and have engaged inpart-time work more often than have men(Blau and Kahn 1997; Rosenfeld andBirkelund 1995).

Indeed, much research has suggested thatthe conditions of employment for men versuswomen contribute to income inequality.Perhaps the most thoroughly discussed expla-nations for the gender income gap are occu-pational sex segregation and women’s con-centration in female-dominated occupations(e.g., Blau and Kahn 2000; England 1992;Huffman 2004; Kilbourne et al. 1994;Macpherson and Hirsch 1995). England(1992:181) found that “the sex compositionof an occupation affects the pay it offers, suchthat both men and women earn less if theywork in a predominantly female occupation.”Given that women are concentrated in tradi-tionally female occupations, the gender gapin wages can be partially explained bywomen’s overrepresentation in jobs that payless.

Similarly, the content of women’s jobsseems to matter, in that women are concen-trated in jobs that are devalued as a result oftheir nurturing character (England 1992;England et al. 1994; Kilbourne et al. 1994).These jobs pay less, and women’s greaterrepresentation in them contributes to thewage gap (England 1992; Kilbourne et al.1994). Given the large degree of occupa-tional sex segregation—half the womenwould have to switch occupations for genderintegration to occur (Padavic and Reskin2002:67)—and that women’s jobs tend topay less and offer fewer other rewards likebenefits and promotional opportunities (seeReskin 1993), Tomaskovic-Devey (1993:10)concluded that “employment segregation iscurrently one of the central rules by which

male employment advantage is created andmaintained.”3

In addition to occupational sex segrega-tion, other work-related factors contribute toinequality in earnings. Men tend to havelonger tenure with their employers, greaterfull-time work experience, and more training,all of which contribute to their higher earn-ings relative to women (Blau and Kahn 1997;Marini 1989; Wellington 1994). Similartrends hold true for young workers: Women’sfirst jobs tend to be of a lesser quality thanmen’s, and women are more likely than mento be employed part time (Joy 2000). Whileoccupations have received the most attentionin studies of income disparities, research onwage differentials across sectors (Moulton1990) and industries (Fields and Wolff 1995;Groshen 1991), along with gendered pat-terns of labor force participation, suggest theneed to consider gender variations at thesebroader levels when explaining gender differ-ences in income (see Fields and Wolff 1995;Macpherson and Hirsch 1995; Marini 1989).

Furthermore, as a result of gender social-ization, young men and women have differ-ent values and occupational aspirations, andthese gender differences appear to influencethe gender income gap via occupationalchoices (Daymont and Andrisani 1984;Wilson and Boldizar 1990; see also Corcoranand Courant 1985; Reskin 1993; Shu andMarini 1998). By their last year of highschool,

men were more likely than women to feelthat making a lot of money is very importantin selecting a job or career. Consistent withsocietal expectations that men be assertiveand dominant, they were also more likely tofeel the importance of choosing a job orcareer that provides an opportunity to be aleader. Women, on the other hand, weremore likely to feel the importance of oppor-tunities to be helpful to others or to society,and of opportunities to work with peoplerather than things. (Daymont and Andrisani1984:414)

These different occupational aspirations affectdecisions regarding higher education, which,in turn, affect the occupations that theseyoung adults enter (Wilson and Boldizar1990).




6 Bobbitt-Zeher

Taking all these patterns into considera-tion, then, the question is this: How much doeducation-related factors—particularly field ofstudy and standardized test scores—con-tribute to gender disparities in earnings earlyin young workers’ careers, relative to family,work, and aspiration influences? Much of theresearch on educational contributions toearnings disparities has not rigorously consid-ered work and family factors, while much ofthe work and family literature has deempha-sized the role of education, which makes itdifficult to understand the weight of the twosets of influences relative to one another.Also, much of this literature has focused onwomen as a whole and has offered littleinsight into women and men in the earlyyears of their careers.

Yet, a focus on young workers is importantfor understanding the salience of educational,work, and family factors for gender equalityin the contemporary social context. Genderdifferences in family responsibilities, labormarket participation, and other human capi-tal-related characteristics that grow over timeare greatly minimized for this group of work-ers. However, little empirical work has exam-ined the impact of such factors specifically onyoung workers.

The most compelling research on workersat their entry into adult careers comes fromMarini and Fan (1997). Using data from theNational Longitudinal Survey of Youth(NLSY), Marini and Fan found that on entryinto careers, women earn 84 percent of whatmen earn. Their findings suggest little reasonto believe that differences in educationalcharacteristics drive gender inequality inincome for workers at career entry. Years ofeducation and field of study explain only 2.8percent of this gender wage differential andother educational factors have only “negligi-ble effects” (p. 600).4

Marini and Fan (1997) found greaterexplanatory power in factors that are relatedto work and aspirations. Indeed, they con-cluded that 22 percent of the gender wagegap is attributable to differences in men’s andwomen’s occupations, and another 28 per-cent is due to differences in industrial place-ment. Part-time employment experienceexplains 9 percent of the gender wage gap,

while gender differences in occupational aspi-rations account for 10 percent. At careerentry, family obligations do not seem to affectthe gender wage gap directly.

Building on the lessons learned from thispast research, this study extends this line ofinquiry by overcoming some of the data limi-tations of past studies and by giving greaterconsideration to academic factors that arethought to influence the gender gap in earn-ings for young workers. While Marini and Fan(1997) focused on early career entry, theirdesign used data on first jobs begun at vari-ous times over a period of 12 years. Althoughthis approach has the advantage of capturingthe gender wage gap for younger workers ata similar place in their career development, ithas the disadvantage of including workers atvarious ages and in different periods. Thistrade-off makes it hard to know if the patternsare consistent across age cohorts and overtime. By following a cohort of young adultsthrough their educational and early labormarket experiences, as my study does, I canspeak more specifically to the gender incomegap for young workers who are experiencingthe same historical labor market conditions.

Furthermore, because of the expansive setof indicators of educational characteristicsavailable in NELS, this study is well posi-tioned to provide a more comprehensiveassessment of the educational factors thatshape the gender gap in earnings. AlthoughMarini and Fan (1997) explored the effects ofseveral academic characteristics on incomedisparities, their analysis was limited by thedata. For example, they considered broadfields of study but did not explicitly test thegender composition of the field, which otherresearch has suggested is important. Theyalso had no measure of level of education orschool selectivity and limited indicators ofmath and reading skills. With better mea-sures of all these educational factors, thestudy design I used can provide a more in-depth assessment of the importance of edu-cational characteristics (i.e., college major,skills as captured by standardized test scores,graduate and professional education, andinstitutional selectivity) for income equalityat the point in one’s career when educationis likely to matter most.




Gender Income Gap 7

Moreover, given the remarkable reportthat young female engineers now enjoy nogender wage gap (Morgan 1998), a freshlook at the factors that influence the genderincome gap is merited with more recent datafor workers in all types of academic fields. Aswe move toward a more highly educatedsociety with fewer educational factors disad-vantaging young women, assessing claimsabout the educational arena’s impact on thegender income gap with contemporaryyoung workers provides an especially com-pelling test.

DATA AND METHODS

Data

NELS offers a rich database for exploring therelationships between educational factors andthe earnings gap between young workingmen and women. It followed students whowere eighth graders in 1988 through highschool and into their early adult lives.5 Thebaseline data are nationally representative,based on a sample of almost 25,000 studentsfrom 1,052 public and private schools. The2000 data contain interviews with 12,144 ofthese individuals.6

This data set is particularly well suited foraddressing my research question because itincludes information from the students’ tran-scripts for both secondary schools and anypostsecondary institutions attended, familyformation history, and labor market participa-tion and earnings.7 NELS offers the opportu-nity to explore potential explanations for agender gap in income in the early years ofmen’s and women’s careers, since the respon-dents had been out of high school forapproximately eight years, had often com-pleted bachelor’s degrees and some graduatedegrees as well, and were frequently in thelabor force by the fourth follow-up in 2000.

For comparison with national data fromthe Current Population Survey on the genderwage gap and to avoid part-time and incon-sistent workers from biasing the analysis, Ilimited my analysis to college graduates whowere full-time, year-round workers (that is,those with four-year degrees who were work-

ing 35 or more hours per week for all 52weeks in 1999) and had annual income dataavailable for 1999 (N = 1,946).8 I focused oncollege graduates for two reasons. First, post-secondary educational attainment is one areain which women have made the greatestgains. Given the relationship between post-secondary educational success and labor mar-ket outcomes, it is especially important toknow to what degree college-educatedwomen enjoy income equity with their malecounterparts. Second, many educationalcharacteristics, such as institutional selectivity,are meaningful or applicable only to the col-lege educated. In supplementary analyses, Ifound that the gender gap in income is larg-er among those without a college degree.9 Byfocusing on college graduates, therefore, Ideliberately concentrated on a group forwhich educational characteristics arguablyhave the greatest chance of explaining gen-der gaps in earnings.10 Missing values werehandled through multiple imputation11

(Allison 2002).

Measures

In all the analyses, the dependent variable isincome, measured by the respondent’s report-ed annual income in 1999.12 This variablecaptures the respondents’ answers to the fol-lowing question: “Including all of the wages,salaries, and commissions you earned in1999, about how much did you earn fromemployment before taxes and all otherdeductions?”

Education-Related Variables The indepen-dent variables correspond to the possiblecausal factors discussed in the literaturereview. As a measure of cognitive skills, I usedstandardized test scores from the students’college entrance examinations. This measureof the composite SAT score captures mathand verbal scores on the SAT, as well as scoreson the ACT, converted to the same scale asthe SAT. As an alternate measure of skills,13 Ialso tested undergraduate grade point aver-age (GPA) a self-reported measure of the stu-dent’s GPA.

Factors related to educational field, at Macquarie University Library on November 13, 2013soe.sagepub.comDownloaded from



8 Bobbitt-Zeher

degree, and institution were also considered.College major was measured in two ways.First, consistent with Marini and Fan (1997),the general field of study is captured withfour dummy variables for the type of major:business; math, natural science, and engi-neering; social sciences and humanities; andeducation (with education majors the refer-ence category).14 Second, to test gendercomposition effects specifically, I measuredthe percentage female of each field in a man-ner that was consistent with NCES (1998). Icalculated the percentage female for eachmajor in the data set using national data onthe percentage of degrees awarded in eachfield of study to men versus women.15

Although the sample was limited to thosewith four-year degrees, I also consideredgraduate and professional degrees. Threedummy variables captured the highestdegree attained—bachelor’s degree, master’sdegree, and doctoral or professional degree(with the bachelor’s degree as the referencecategory).

In addition, I included measures of theselectivity of the higher education institutionattended. The NCES, using methodologyfrom the Cooperative Institutional ResearchProject (see NCES 2003b:72 for further infor-mation on this classification scheme) rankedeach institution that the NELS respondents’attended. Selectivity was measured in aninterval fashion ranging from not selective (1)to highly selective (4).

Background, Family, Work, and ValuesVariables While my primary interest was tounderstand the contribution of educationalfactors to the gender gap in earnings foryoung workers, I also considered nonacadem-ic characteristics suggested in the literature.In regard to family formation characteristics,the variables included marital status, single-parent status, and the number of hoursworked per week. Three dummy variables—married, divorced, and single—were used tocapture a respondent’s marital status (the ref-erence group was those who are single). I alsoconsidered single-parent status, a dummyvariable (1 = a single parent, 0 = not a singleparent) and the number of dependents who arechildren, measured in a continuous fashion.

Number of hours worked was a continuousvariable capturing the total number of hoursthe respondent worked at all jobs in an aver-age week in 1999, which was also the year forthe income variable.

I used several job characteristics in theanalysis. A dummy variable for for-profit sectorgauged the type of employer for whom therespondent worked at a private, for-profitcompany. Not-for-profit, governmental, andmilitary employers were the reference catego-ry for sector. The NELS data also containcodes for the industry in which the respon-dent was employed in 1999. I createddichotomous dummy variables for each ofthe 17 industries represented in the sample.See the Appendix for the list of industries.Retail trades is the reference category in allthe analyses.

To capture the respondents’ occupations,NELS aggregated the subjects’ responses tothe question of their current or previous jobtitle into broad occupational categories. Ithen created dichotomous dummy variablesfor each of the 31 occupational categoriesthat were represented in the sample. By sam-ple design, all unemployed persons wereexcluded. See the Appendix for a list of all theoccupations. The regression models used thecriminal justice/military category as the refer-ence group.

To gain a better understanding of the rela-tionship between job quality and incomeequality noted by Reskin (1993) and Wellington(1994), I included job training and job autono-my in the analysis. Job training is a dichotomousvariable in which 1 indicates that the respon-dent received job training in the previous 12months and 0 indicates that the respondentreceived none. Job autonomy measures the levelof autonomy the respondents perceived theyhad on their jobs, with values of 1 to 4; highervalues signify higher levels of autonomy.

Furthermore, corresponding to Daymontand Andrisani’s (1984) study, I used a mea-sure of values, the importance of having lots ofmoney. This is an ordinal variable with rangesof 1 (not important) to 3 (very important)that was measured during the students’ 12th-grade year.

In addition to these theoretically guidedindependent variables, I included measures of




Gender Income Gap 9

the socioeconomic status of the family of origin(SES) and the respondent’s race as controls.SES is NELS’s socioeconomic composite score,constructed from information on parents’education, income, and occupation. Race ismeasured with dichotomous dummy vari-ables for white, black, Latino, Asian American,and other race; the reference category in allthe analyses is white.

ANALYTIC STRATEGY ANDRESULTS

I used two regression strategies in the analy-sis. First, I used estimated generalized least-squares (EGLS) regression to ascertain theeffects of the independent variables of inter-est on income. Because of NELS’s clusteredsampling design, EGLS is an appropriate ana-lytic approach to correct for estimates of stan-dard errors.16 In these equations, the interestfor interpretative purposes is on the coeffi-cient for the binary variable female (1 =female, 0 = male) in each model. This coeffi-cient represents the effect of being femaleversus being male on income, net of theother variables in the model. A negative coef-ficient for female indicates an income gapadvantaging men. The coefficient for femalein the bivariate baseline model may be com-pared with the coefficient for female in eachadditional model. I was especially interestedin the extent to which the coefficient forfemale changed across the various modelspecifications, which allowed me to assess theextent to which each of the possible explana-tions offered earlier contributes to an under-standing of the gender income gap. I beganby testing the individual effects of the educa-tion-related variables of interest. These indi-vidual tests suggest which educational vari-ables merit further consideration. I then test-ed a series of additive models that captureyoung workers’ typical life course trajectory—moving from ascriptive qualities of race andsocial class through socialization of values andhigher education to the formation of familiesand entry into careers.

Second, I used regression decompositionsto enhance my capacity to capture the influ-

ence of each individual factor in the compre-hensive regression model. A standard tech-nique among scholars of wage inequality,regression decomposition suggests theamount of the total gender income gap thatcan be attributed to a given variable. Thetechnique does so by partitioning the influ-ence of gender differences in the characteris-tics (i.e., the mean differences between menand women on each independent variable)while considering the return for a one-unitincrease in that characteristic (i.e., the slopefor the independent variable for women andmen) (see, e.g., England 1992). To get thenecessary information, I ran the equationfrom the best regression model separately forwomen and men in the sample. I calculatedthe decompositions using, alternatively,men’s and women’s coefficients as the stan-dard, since the choice of standard may influ-ence the effects that are found (see, e.g.,Daymont and Andrisani 1984; Marini and Fan1997). Following Marini and Fan, I took theaverage of the estimates from these alternatespecifications. The amount of the income gapnot explained by the attributes of the genderdifferences in observed characteristics is con-sidered the unexplained portion of theincome gap. The gender income differential,then, can be represented by the followingformula:

using the male standard,

X—

mi – X—

fi = Σbm(X—

m – X—

f) + ε;

using the female standard,

X—

mi – X—

fi = Σbf (X—

m – X—

f) + ε;

where the left-hand side represents the meandifference in income between men andwomen and the right-hand side captures thedifference in men’s and women’s means mul-tiplied by the slope from the male (bm) orfemale (bf) model, respectively, plus the dif-ferences in income left unexplained.

Table 1 presents descriptive statistics forincome and educational measures for menand women. As expected, the patterns sug-gest some gender differences in educationaloutcomes, with women generally garneringbetter grades and men scoring better onstandardized tests in math and science. While




10 Bobbitt-Zeher

patterns of course taking in high school sug-gest some statistically significant gender dif-ferences, these gaps are small in magnitude.However, at the collegiate level, there are sig-nificant differences in the majors that menand women choose. Men are significantlymore likely to major in business, math, natur-al science, and engineering, whereas womenare significantly more likely to major in socialsciences, humanities, and education. Theaverage college major for women is 63 per-cent female, while the average college majorfor men is 48 percent female.

These patterns suggest considerable gendersegregation of college majors; yet, it is alsoclear that a large number of men and womenchoose fields that are neither male- nor femaledominated. Although few men and womenhave completed graduate or professionaldegrees (8 percent of the sample), women aresignificantly more likely to have attained a mas-ter’s, doctorate, or professional degree. Thereare no significant gender differences in theselectivity of the bachelor’s degree institutionsthat women and men attend.

The Appendix presents descriptive statisticsfor noneducational conditions and suggestsseveral significant differences by gender.Women, on average, are more likely than aremen to be married or in a marriagelike rela-tionship and to be single parents. Although allthe workers are full-time employees, men aver-age more hours worked in a typical week. Also,men and women are going into different occu-pations and industries, and women are signifi-cantly more likely to work in the public sector.17

By the time these college graduates wereworking full time in 1999, there were sizablegender differences in average income. In fact,the women earned $6,938 less than similarlyaged, college-educated male workers. Statedanother way, in the years immediately follow-ing college graduation, full-time workingwomen with a bachelor’s degree earn, onaverage, 83 percent of what their male coun-terparts earn. This pattern is in line withnational data on the earnings’ ratios of youngadults (U.S. Department of Labor, Bureau ofLabor Statistics 2004) and suggests a sub-stantial gender gap in income in contempo-rary American society among young workerswith college degrees.

Given that women are making significant-ly less money than are men, I now turn toregression analysis to explore which factorsexplain the gender income gap for young col-lege graduates. Table 2 presents the results ofEGLS regressions of income on the variouseducation-related independent variables.18

The baseline model (1) shows that womenaverage $6,938 less than do men per yearwhen no factors are controlled. Given thatgender, race, and social class are all ascribedstatuses that significantly influence income, Iuse controls for race and parental SES in allthe models, although race and backgroundSES have little direct influence on the gendergap in earnings19 (see Model 2).

The additional models suggest that thegender composition of college majors is theprimary educational influence on gender dis-parities in earnings. When general fields aremeasured (Model 3), the income gap de-creases 22 percent. More robust, however, isthe measure of the gender composition offields of study; the percentage female of themajor accounts for 39 percent of the incomedifferential between men and women whenrace and background SES are controlled.

The other educational factors consid-ered—cognitive skills, higher degrees, andcollege selectivity—have modest or littleeffect on the gender gap in income. Whenmen and women have the same SAT/ACTscores (Model 5), the female coefficient isreduced to -$6,247, suggesting that differ-ences in cognitive skills explain 10 percent ofthe income gap among the college educat-ed.20 An alternate measure of skills, under-graduate GPA (Model 6), accounts for noneof the gender income gap and suggests thatwomen’s higher grades suppress the gapfrom being even larger than it is. Selectivity ofthe bachelor’s degree institution (Model 7)has a small effect on the income gap, explain-ing 4 percent when only race and back-ground SES are controlled. Most important,when college-educated workers have thesame fields of study, same test scores, andsame levels of graduate education and attendthe same types of educational institutions,women still earn $4,436 less per year thantheir male counterparts do.

Although this look at the individual, direct at Macquarie University Library on November 13, 2013soe.sagepub.comDownloaded from



Gender Income Gap 11

Table 1. Means for Income and Education Variables, including Significance Tests of theMeans for Women versus Men, from Imputation 3

Sample Women’s Men’s t-Test ofMean Mean Mean Mean

Variable (N = 1,946) (N = 1,053) (N = 893) Difference

Income in 1999 36,137 32,953 39,891 11.08***

Standardized Test ScoresReading (12th grade) 55.98 56.98 54.81 -5.83***Math (12th grade) 58.17 57.30 59.19 5.52***Science (12th grade) 56.14 54.48 58.11 9.47***SAT math 510.03 492.42 530.80 7.69***SAT verbal 445.10 445.18 445.01 -0.04SAT/ACT combined 974.93 956.52 996.63 4.36***

GradesEnglish (12th grade) 8.04 8.48 7.53 -11.07***Math (12th grade) 7.30 7.40 7.17 -2.18*Science (12th grade) 7.62 7.74 7.49 -2.49*Undergraduate GPA 2.99 3.06 2.91 -7.05***

Course Work (High School)Units in English 4.20 4.27 4.11 -4.38***Units in math 3.75 3.71 3.81 2.99**Units in science 3.47 3.41 3.53 2.54*

College MajorBusiness major .22 .20 .24 1.99*Math, natural science, or .28 .22 .36 6.89***

engineering majorSocial science or .42 .48 .36 -5.15***

humanities majorEducation major .07 .10 .04 -5.53***Percentage female of major 56.09 62.98 47.96 -17.66***

Highest DegreeBachelor’s degree .92 .90 .94 3.25**Master’s degree .07 .09 .06 -2.63**Professional or doctoral degree .01 .01 .00 -2.62**

Institutional Selectivity 2.29 2.28 2.31 0.84

* p < .05, ** p < .01, *** p < .001 (two-tailed tests).




12 Bobbitt-Zeher

influences of education on the gender gap inincome suggests a substantial effect ofschooling on earnings’ inequality, these mod-els do not consider confounding influences ofvalues, family formation, and work. Table 3begins with the average difference in incomebetween men and women and adds clustersof independent variables, following the gen-eral life-course sequence. These findings sug-gest that when schooling is considered along-side background factors and values regardingthe importance of earning lots of moneyexpressed during junior high school (Model4), the gender earnings gap decreases 44 per-cent. The addition of family-formation factorsyields little reduction in the income gap

(Model 5); however, consideration of workfactors appears to add greatly to an under-standing of the income disparities. This bestmodel (Model 6) explains 69 percent of thetotal gap in earnings.

Given that this best model explains morethan two-thirds of the income disparitybetween these young men and women, I per-formed a regression decomposition on thismodel. Presented in Table 4, the regressiondecomposition suggests how much of thetotal difference in men’s and women’s earn-ings can be attributed to differences in thecharacteristics of the two groups. The contri-bution of the educational qualities is substan-tially smaller than indicated by the regression

Table 2. EGLS Regression Coefficients for Female and Percentage of the Gender IncomeGap Explained with Alternate Models, including Educationa

Model Income Gap Percentage of Number Model Description (bfemale) Gap Explained

1 Female -6,938 —

2 Female, background -6,643 4.2

3 Female, background, field of college major -5,418 21.9

4 Female, background, percentage female of college major -4,244 38.8

5 Female, background, SAT/ACT score -6,247 10.0

6 Female, background, undergraduate GPA -7,215 0

7 Female, background, graduate degree -6,815 1.8

8 Female, background, college selectivity -6,634 4.4

9 Female, background, percentage female of college major, SAT/ACT score, graduate degree, college selectivity -4,436 36.1

a Background factors are controls for race and parental SES.





models that did not control for nonschoolingeffects. The percentage female of the collegemajor explains 14 percent of the income gap,while scores on standardized tests explainanother 5 percent. However, almost half thedifferences in men’s and women’s earningscan be attributed to work-related characteris-tics, especially occupation, sector, industry,and hours worked per week. Values appear tomatter only modestly, while family formationhas virtually no effect on the income gap forthis sample of young workers.

Taking into consideration the results fromthe EGLS regression models presented inTables 2 and 3, along with the decompositionfindings from Table 4, it appears that the gen-der composition of college majors has a con-siderable direct effect on the gender gap inearnings. However, employment-related fac-tors are more salient for understanding theincome disparities.

DISCUSSION

Girls’ success in school is a hot topic amongboth academics and the general public.Generally absent from this dialogue, however,has been the key question of the conse-quences of these patterns for the genderequality of young adults as they transitionfrom educational institutions to the work-force. Now that young women’s educationalperformance is so strong relative to men’s,how does this newfound “advantage” mat-ter? Given the long-standing debate over therole of education in mediating group inequal-ities, this discussion is overdue.

The findings of this study suggest thateducation continues to contribute to genderstratification in a meaningful way despitewomen’s overall success in educationalrealms. The educational factor that appears tomatter most is college major, and college

Table 3. EGLS Regression Coefficients for Female and Percentage of the Gender IncomeGap Explained with Alternate Models, including Background, Values, Education, FamilyFormation, and Work Factors

Model Model Description Income Gap Percentage ofNumber (bfemale) Gap Explained

1 Female -6,938 —

2 Female and background -6,643 4.2

3 Female, background, and values -6,166 11.1

4 Female, background, values, and education -3,903 43.7

5 Female, background, values, education, and family formation -3,854 44.4

6 Female, background, values, education, family formation, and work -2,132 69.3

Note: Background factors are controls for race and parental SES. Values are measured by theimportance of having lots of money. Education factors are the percentage female of the fieldof study, SAT/ACT scores, highest degree earned, and selectivity of degree granting institution.Family formation factors are marital status and single-parent status. Work factors are the num-ber of hours worked per week, occupation, industry, sector, job autonomy, and job training.




14 Bobbitt-Zeher

major appears to affect inequality in earningsin two ways. As one may expect, field of studycontributes to earnings inequality via occupa-tional choices: People tend to work in jobsthat are related to their fields of study, andsome occupations are better rewarded thanare others. Yet, the regression decompositionpresented here suggests that college majorsplay a meaningful role in women’s lesser

income independent of later work factors.Indeed, even when work-, family-, and val-ues-related factors are considered alongsideeducation, 14 percent of the gender gap inincome is still attributable to field of study.

What is it about college majors, and whydo they matter for the gender gap in earningsbeyond their relationship to occupations? It isoften argued that some fields are more high-

Table 4: Regression Decompositions Showing Contributions of Background, Values,Education, Family Formation, and Work Characteristics to the Gender Income Gap

Percentageof

Men’s Women’s Average Total Gap Rank of Characteristic Slopea Slopeb Slope Explained Influence

Background SES and Race 124 33 78 1.1 9

Importance of Having Lots of Money 472 294 383 5.5 6

Education RelatedSAT/ACT scores 434 226 330 4.8 7

Percentage female of college major 1,087 842 964 13.9 2

Institutional selectivity 18 59 39 0.6 10

Graduate/professional degree 57 -244 -93 -1.3 12

Family Formation 15 3 9 0.1 11

Work RelatedHours worked per week 613 816 715 10.3 3

Occupation 1,544 1,319 1,431 20.6 1

Industry 497 492 495 7.1 5

Sector 770 453 611 8.8 4

Other work factors 145 135 140 2.0 8

Total Explained with These Factors 5,776 4,428 5,102 73.5

Total Unexplained 1,162 2,510 1,836 26.5

Total Income Gap 6,938 6,938 6,938

a Σbm (X—

m – X—

f).b Σbf (X

—m – X

—f).





ly compensated because they develop skillsthat are more valued in the labor market.Although the content of the field of studyseems to have an important relationship withearnings inequality, the gender compositionof the field appears to be much more salient21

(see Table 2). This devaluation of majors asso-ciated with women is consistent with thefinding of a general devaluation of jobs asso-ciated with women (see, e.g., England 1992).And it appears that the lesser value assignedto majors in which women are more heavilyconcentrated continues to affect one’s earn-ings even when young workers enter compa-rable occupations. This pattern indicates thateducational sex segregation by field of studycontinues to be an impediment to genderequality beyond its relationship to occupa-tional sex segregation and will have to beaddressed directly if gender disparities are tobe eliminated.

While the analysis suggests that segrega-tion of college majors plays an important rolein earnings inequality early in young workers’careers, its contribution should not be over-stated. In their mid-20s, college-educatedwomen make about $4,400 less per year thando men even when they have the same levelof education, college major, cognitive skills,and selectivity of the college from which theygraduated. The regression decompositionshows that work-related factors explain abouthalf the contemporary gender difference inearnings. Here, it seems that gender differ-ences in types of employment—occupations,industries, and sectors—are especially impor-tant.22 For this group of workers, aspirationsfor earning lots of money appear to matteronly modestly for the income gap, and fami-ly formation matters not much at all.23

Research on contemporary gender stratifi-cation provides a useful framework for under-standing these findings and their implicationsfor gender equity (see Charles and Bradley2002; Charles and Grusky 2004). The denialof opportunities for women to attend collegeand attain a degree on the basis of their sex isinconsistent with the contemporary genderideology of equality of opportunity. Yet, asthis vertical dimension of segregation hasdeclined, fields of study—the horizontaldimension—remain resistant to integration.

Because gender differences in college majorsare viewed more as differences than asinequality, the segregation of fields of studycan persist despite a more egalitarian genderideology and the decline of vertical segrega-tion (Charles and Bradley 2002). Thus, theoverall positive picture of women’s educa-tional patterns can mask lingering gender dif-ferences that have important consequencesfor gender inequality later in life. Similarly,occupational sex segregation persists andmay increase despite the greater presence ofwomen in the labor market (Charles andGrusky 2004).

Unfortunately, the ongoing debate overwomen’s growing “advantage” in schoolinghas, by and large, overlooked the conse-quences of these patterns of educationalattainment and performance. A look at oneimportant outcome—the earnings of youngwomen and men—should temper the opti-mism that is generally generated by trendstoward girls’ educational success. On thewhole, the findings of this study suggest littlereason to be optimistic that further educa-tional changes will lead to large declines ingender inequality in income. If women main-tain their current trajectory of improving theireducational credentials relative to men, theywill still face the barrier of sex segregation incollege majors. And while the integration ofmajors could generate important reductionsin the gender gap in income, research hasfound a general stagnation of integration offields of study since the mid-1980s (seeJacobs 1995). Even larger barriers to incomeequity are related to gendered patterns ofemployment, and today’s young college-edu-cated adults continue to confront theseobstacles. Indeed, in spite of women’s educa-tional progress, the gendered organization ofboth higher education and employmentremain substantial impediments to equality inearnings.

NOTES

1. Gill and Leigh (2000) also consideredthe effect of changes in participation in high-er education on declines in the gender gap inwages in the late 1980s and early to mid-




16 Bobbitt-Zeher

1990s. Although they did not consider edu-cational performance (i.e., grades), they con-cluded that gendered changes related to fieldof study played a part in the reduction of thegender gap in wages.

2. For a discussion of horizontal and verti-cal gender segregation in higher education,see Charles and Grusky (2004) or Gerber andSchaefer (2004).

3. See England (1992), Jacobs (2001),Reskin (1993), and Tomaskovic-Devey (1993)for a more thorough overview of occupation-al sex segregation and its causes and conse-quences.

4. Note that Marini and Fan (1997) con-sidered the following education-related fac-tors: years of education and field of study(considered together), high school GPA, scoreon the Armed Forces Qualifications Test as ameasure of verbal and math skills, and par-ents’ education.

5. Waves of data were collected in 1988,1990, 1994, and 2000, providing informationon these young adults from roughly age 14 toage 26.

6. Patterns of sample attrition between1988 and 2000 resulted in a more socioeco-nomically privileged group of young adultsthan would be found at random. However, apreliminary analysis of attrition patterns didnot suggest large gender differences in con-tinuation in NELS.

7. NELS includes information from the stu-dents, as well as supplemental informationfrom their parents, teachers, and schooladministrators. See NCES (2002) for furtherinformation on this data set.

8. To correct for skew, I eliminated respon-dents whose income was four or more standarddeviations above the mean (N = 13) for full-time, year-round workers. I also eliminatedthose who reported $2,000 or less in earningsthat year (N = 5), since it is illogical that work-ing full time for the entire year would yield earn-ings that low. Including these workers wouldsuggest a slightly larger gender income gap.

9. While some scholars (e.g., Blau 1998)have found a similar pattern of larger wagegaps among those with less education, thereis not a clear consensus on this issue.

10. Limiting the sample as I did to collegegraduates who were full-time, nonseasonal

workers captured the most privileged work-ers. For example, these workers tended tocome from families with higher SES scores, tohave performed better in school, and to havefewer children of their own than did the gen-eral sample of NELS participants. Accordingly,I limit the generalizability of these findings tofull-time college-educated workers. In a sup-plemental analysis noted in the Discussionsection and discussed in note 23, I discuss thegender income gap among those with lesseducation.

11. Using multiple imputation allows theanalyst to retain cases that have missing valueson some variables. By using information on therelationships among the variables in the dataset, multiple imputation creates estimates ofwhat the missing values would most likely be ifthey were not missing. The imputation processis repeated multiple times at random to allowfor the correction of estimates of standarderrors. Thus, instead of dropping cases withmissing values, as in listwise deletion, multipleimputation preserves for analysis cases withuseful information, thereby reducing potentialbiases that would result from systematic differ-ences in complete cases and cases with miss-ing information. Furthermore, this method ofhanding missing data “produces estimatesthat are consistent, asymptotically efficient,and asymptotically normal when the data are[missing at random]” (Allison 2002:27). A sup-plemental analysis of the missing data suggest-ed that these assumptions are valid with thesedata, and the findings reported here were notsubstantially altered when I used listwise dele-tion of missing cases.

12. As I noted in the Data section, I han-dled potential skewing by excluding extremecases (those more than 4 standard deviationsabove the mean). Doing so eliminated theneed to transform the income variable byeliminating skew at the highest income levelsand allowed for a more intuitive presentationof results.

13. Loury (1997) suggested that gradesinfluence wage gaps, and the use of GPA isconsistent with the approach of Marini andFan (1997), who tested high school GPA.

14. In a supplemental analysis, I tested amore nuanced measure of field of studiesusing 18 dummy variables for the various





fields represented in the data. I found that thegender gap in income was consistentwhether I used this more refined measure orthe more general four-category classification.See note 21 for the same issue.

15. The NCES Integrated PostsecondaryEducation Data System (IPEDS) CompletionsSurvey provides data on the number ofdegrees awarded by gender and field of studyin 2000–01. The national data on degreesaligns rather consistently with the categoriesused to code majors in the NELS data.

16. NELS used a two-stage, stratified sam-pling design, sampling on schools (by type,region, and size of enrollment) and then indi-viduals in schools. The complex sampledesign leads to less variation than a simplerandom sample would produce (i.e., studentsin the same schools are more similar thanwould be expected if students were sampledrandomly from the population at large). Toadjust for this sampling issue, I used EGLSregression because EGLS yields unbiased esti-mates of the standard errors in such instances(see NCES 2002:97).

17. While on the whole, the gender pat-terns found here are consistent with generalpatterns in education and employment, thefindings related to work experience and grad-uate degrees suggest that the women in thissample were likely the more privileged work-ers—perhaps a bit more so than the maleworkers. This gendered pattern in the sample,then, would suggest a smaller income gapthan would be expected if the respondents’work-history patterns and graduate or profes-sional school participation matched thenational averages. However, given that theearnings differential found here is consistentwith that found with federal data, as well asthe one found by Marini and Fan (1997) intheir study of young workers in their first jobs,I do not consider that these patterns reflectlarge biases in the sampling design.

18. For brevity, the results of all models arenot shown. Full results are available onrequest.

19. In supplemental analyses, I tested forinteraction effects of race and gender andfound no significant interactions. Therefore,the models presented here treat race andgender as additive rather than interactive.

20. I tested 12th-grade math and readingscores as well. Reading scores explain none ofthe earnings gap, while math scores have aneffect similar to that of the SAT/ACT scores.

21. Given that the percentage female ofthe field is a much more refined measure thanis the four-category classification of majors bycontent area, I also tested a much morenuanced measure of content areas using 18dummy variables for fields of study. Theregression results differed only negligibly. Thisfinding supports the contention made herethat the most important difference in fields ofstudy is gender composition.

22. The study attempted to understandgender inequality in income among full-time,year-round workers with bachelor’s degrees.Whether the explanatory factors studied hereoperate similarly among those who workfewer hours per week and/or seasonallyremains to be seen. Arguably, full-time, year-round workers are the most advantagedfinancially and experience less genderinequality in income, given that past researchindicated that part-time and intermittentwork contributes to the gender gap in wages(see Budig and England 2001). Furtherresearch should consider the prevalence,causes, and consequences of gender differ-ences in earnings for workers with variousdegrees of employment (e.g., workers whoare employed full time, part time voluntarily,or underemployed).

23. In these analyses, I focused on individu-als with college degrees because I wanted toexplore characteristics of education that areapplicable only to this group (e.g., field ofstudy, selectivity of institution, and graduatedegrees). But, what about those without col-lege degrees? My supplemental analysis ofworkers with less than a four-year degree sug-gests that these workers in their mid-20s endurean even larger gender gap in income: Womenwho work full time earn about 75 percent ofwhat their male counterparts make. The resultssuggest that education will matter even less forworkers without a college degree: Educationalfactors (e.g., highest level of education, scoreson standardized tests in high school, andgrades in high school) explain virtually none ofthe gender gap in earnings for this group. Aswith those with college degrees, a big part of




18 Bobbitt-Zeher

the gender gap-in-income puzzle is understoodin terms of work characteristics. Gendered pat-terns of occupational and industrial placementand number of hours worked per week seemcritical. However, unlike the situation for collegegraduates, family-formation factors, particularly

being a single parent, appear to explain more ofthe gender gap in earnings. Furthermore, it isharder to pinpoint the root of the earnings dif-ferential for this group, since my best modelexplains only about half the income disparity forthese workers.

APPENDIX

Means for Background, Values, Family, and Work Variables, includingSignificance Tests of Women’s versus Men’s Means, from Imputation 3



Income in 1999 36,137 32,953 39,891 11.08***

SES of Family of Origin 0.35 0.30 0.40 3.15**

Race of RespondentWhite 0.76 0.74 0.79 2.30*Black 0.06 0.07 0.05 -1.40Latino 0.07 0.08 0.06 -1.24Asian American 0.09 0.09 0.08 -0.93Other race 0.02 0.02 0.02 -0.46

Importance of Having Lots of Money 2.24 2.14 2.36 8.40***

Family CharacteristicsSingle 0.63 0.59 0.68 4.13***Married or marriage-like relationship 0.35 0.38 0.31 -3.64***Divorced 0.02 0.02 0.01 -1.97*Single parent 0.01 0.02 0.01 -1.74Number of children 0.12 0.11 0.14 1.20

Number of Hours Worked, Typical Week 45.78 44.37 47.46 7.95***

SectorFor profit 0.70 0.63 0.78 7.28***Not for profit, government, military 0.30 0.37 0.22 -7.28***

IndustryAgriculture, forestry, fisheries, mining 0.01 0.00 0.02 2.92**Construction and allied 0.03 0.01 0.05 5.36***Manufacturing: Durable goods 0.07 0.06 0.09 2.69**Manufacturing: Nondurable goods 0.03 0.02 0.04 1.30Utilities 0.01 0.01 0.01 1.59Wholesale distribution 0.01 0.00 0.02 3.59***Retail trades 0.06 0.07 0.05 -1.35Finance, insurance, real estate 0.13 0.10 0.17 4.17***Business, personal services 0.08 0.07 0.08 0.81Entertainment, recreation 0.02 0.01 0.02 0.50Professional services 0.17 0.18 0.15 -1.45Public administration, safety, military 0.04 0.03 0.05 2.47*





APPENDIXcontinued



Health care 0.11 0.15 0.06 -7.35***Communications 0.06 0.06 0.07 0.45Transportation 0.01 0.01 0.02 1.79Hospitality 0.01 0.01 0.02 0.75Education 0.14 0.19 0.08 -7.37***

OccupationSecretary, receptionist 0.01 0.02 0.00 -3.75***Cashier, teller, clerk, data entry 0.00 0.00 0.00 -0.68Other clerical 0.01 0.02 0.01 -1.91Farmer, forester, farm laborer 0.00 0.00 0.01 3.21**Personal service, cook, chef, baker 0.01 0.01 0.00 -2.94**Nonfarm laborer 0.01 0.00 0.02 3.54***Mechanic, repairer, service technician 0.00 0.00 0.01 1.71Craftsman, skilled operative 0.01 0.01 0.01 1.69Protective service, criminal justice, military 0.04 0.03 0.05 2.80**Business and financial support services 0.06 0.07 0.05 -1.73Financial service professional 0.11 0.09 0.13 2.42*Sales, purchasing 0.10 0.10 0.10 -0.33Customer service 0.01 0.02 0.00 -2.88**Legal support 0.00 0.00 0.00 -0.42Medical practice professional, services 0.02 0.03 0.01 -3.35***Medical licensed professional 0.04 0.06 0.02 -4.55***Educators (K–12 teachers) 0.09 0.12 0.04 -6.60***Educators, instructors (non-K–12) 0.03 0.04 0.02 -2.88**Human service professional 0.03 0.05 0.02 -4.13***Engineer, architect, software engineer 0.07 0.03 0.11 7.33***Scientist, statistician professional 0.01 0.01 0.01 -0.89Research assistant, lab technician 0.01 0.02 0.01 -0.94Technical, professional worker 0.02 0.02 0.03 1.82Computer systems, related professional 0.07 0.04 0.10 5.79***Computer programmer, other computer 0.02 0.02 0.04 2.79**Editor, writer, reporter 0.03 0.04 0.02 -1.74Performer, artist 0.01 0.01 0.01 1.01Manager, executive 0.01 0.01 0.02 1.96Manager, midlevel 0.04 0.03 0.05 2.36*Manager, supervisory, office 0.10 0.11 0.08 -2.21*Health, recreational services 0.01 0.01 0.01 0.30

Job Training 0.76 0.75 0.78 1.35

Job Autonomy 2.78 2.76 2.81 1.45

Work Experience--Full Time in 1998 0.87 0.86 0.87 0.61

Work Experience--Part Time in 1998 0.10 0.10 0.09 -0.92

* p < .05, ** p < .01, *** p < .001 (two-tailed tests).




20 Bobbitt-Zeher

REFERENCES

Allison, Paul D. 2002. Missing Data. ThousandOaks, CA: Sage.

Blau, Francine D. 1998. “Trends in the Well-Beingof American Women, 1970–1995.” Journal ofEconomic Literature 36:112–65.

Blau, Francine D., and Lawrence M. Kahn. 1997.“Swimming Upstream: Trends in the GenderWage Differential in the 1980s.” Journal ofLabor Economics 15:1–42.

Bradley, Karen. 2000. “The Incorporation ofWomen into Higher Education: ParadoxicalOutcomes” Sociology of Education 73:1–18.

Brown, Charles, and Mary Corcoran. 1997. “Sex-Based Differences in School Content and theMale-Female Wage Gap.” Journal of LaborEconomics 15:431–65.

Budig, Michelle J., and Paula England. 2001. “TheWage Penalty for Motherhood.” AmericanSociological Review 66:204–25.

Cappelli, Peter, Laurie Bassi, Harry Katz, DavidKnoke, Paul Osterman, and Michael Useem.1997. Change at Work. New York: OxfordUniversity Press.

Charles, Maria, and Karen Bradley. 2002. “Equalbut Separate? A Cross-National Study of SexSegregation in Higher Education.” AmericanSociological Review 67:573–99.

Charles, Maria, and David B. Grusky. 2004.Occupational Ghettos: The Worldwide Segregationof Women and Men. Stanford, CA: StanfordUniversity Press.

College Board. 2003, August 26. “SAT Verbal andMath Scores Up Significantly as a Record-Breading Number of Students Take the Test:Average Math Score at Highest Level in Morethan 35 Years.” Press release. Available onlineat http://www.collegeboard.com/press/arti-cle/0,,26858,00.html

Corcoran, Mary E., and Paul N. Courant. 1985.“Sex Role Socialization and Labor MarketOutcomes.” American Economic Review75:275–78.

Davies, Scott, and Neil Guppy. 1997. “Fields ofStudy, College Selectivity, and StudentInequalities in Higher Education.” Social Forces75:1417–38.

Daymont, Thomas N., and Paul J. Andrisani. 1984.“Job Preferences, College Major, and theGender Gap in Earnings.” Journal of HumanResources 19:408–28.

England, Paula. 1992. Comparable Worth: Theoriesand Evidence. New York: Aldine De Gruyter.

England, Paula, Melissa S. Herbert, Barbara StanekKilbourne, Lori L. Reid, and Lori McCrearyMegdal. 1994. “The Gendered Valuation of

Occupations and Skills: Earnings in 1980Census Occupations.” Social Forces 73:65–100.

Farkas, George, Paula England, Keven Vicknair, andBarbara Stanek Kilbourne. 1997. “CognitiveSkill, Skill Demands of Jobs, and Earningsamong Young European American, AfricanAmerican, and Mexican American Workers.”Social Forces, 75:913–38.

Farley, Reynolds. 1995. State of the Union: Americain the 1990s: Volume One. Economic Trends.New York: Russell Sage Foundation.

Fields, Judith, and Edward N. Wolff. 1995.“Interindustry Wage Differentials and theGender Wage Gap.” Industrial and LaborRelations Review, 49(1):105-120.

Gerber, Theodore P., and David R. Schaefer. 2004.“Horizontal Stratification of Higher Educationin Russia: Trends, Gender Differences, andLabor Market Outcomes.” Sociology ofEducation 77:32–59.

Gill, Andrew M., and Duane E. Leigh. 2000.“Community College Enrollment, CollegeMajor, and the Gender Wage Gap.” Industrialand Labor Relations Review 54:163–81.

Groshen, Erica L. 1991. “The Structure of theFemale/Male Wage Differential: Is It Who YouAre, What You Do, or Where You Work?”Journal of Human Resources 23:457–72.

Hearn, James C. 1991. “Academic and NonacademicInfluences on College Destinations of 1980 HighSchool Graduates.” Sociology of Education64:158–71.

Huffman, Matt L. 2004. “Gender Inequality AcrossLocal Wage Hierarchies.” Work and Occupations31:323–44.

Jacobs, Jerry. 1995. “Gender and AcademicSpecialties: Trends Among College DegreeRecipients During the 1980s. Sociology ofEducation 68:81–98.

——. 1996. “Gender Inequality and HigherEducation.” Annual Review of Sociology22:153–85.

——. 1999. “Gender and the Stratification ofColleges.” The Journal of Higher Education70:161–87.

——. 2001. “Evolving Patterns of Sex Segregation.”Pp. 535–50 in Sourcebook of Labor Markets:Evolving Structures and Processes, edited by IvarBerg and Arne L. Kalleberg. New York: KluwerAcademic/Plenum.

Joy, Lois. 2000. “Do Colleges ShortchangeWomen? Gender Differences in the Transitionfrom College to Work.” American EconomicReview 90:471–75.

Karen, David. 1991. “The Politics of Class, Race,and Gender: Access to Higher Education in the





United States, 1960–1986.” American Journalof Education 99:208–37.

Kilbourne, Barbara, Paula England, and Kurt Beron.1994. “Effects of Individual, Occupational,and Industrial Characteristics on Earnings:Intersections of Race and Gender.” SocialForces 72:1149–76.

Korenman, Sanders, and David Neumark. 1992.“Marriage, Motherhood, and Wages.” Journalof Human Resources 27:233–55.

Loury, Linda Datcher. 1997. “The Gender EarningsGap among College-Educated Workers.”Industrial and Labor Relations Review50:580–93.

Macpherson, David A., and Barry T. Hirsch. 1995.“Wages and Gender Composition: Why DoWomen’s Jobs Pay Less?” Journal of LaborEconomics 13:426–71.

Marini, Margaret Mooney. 1989. “Sex Differencesin Earnings in the United States.” AnnualReview of Sociology 15:343–80.

Marini, Margaret Mooney, and Pi-Ling Fan. 1997.“The Gender Gap in Earnings at Career Entry.”American Sociological Review 62:588–604.

Mitra, Aparna. 2002. “Mathematics Skill and Male-Female Wages.” Journal of Socio-Economics31:443–56.

Morgan, Laurie A. 1998. “The Earnings Gap forWomen Engineers, 1982 to 1989.” AmericanSociological Review 63:479–93.

Moulton, Brent R. 1990. “A Reexamination of theFederal-Private Wage Differential in the UnitedStates.” Journal of Labor Economics 8:270–93.

Murnane, Richard J., John B. Willett, and FrankLevy. 1995. “The Growing Importance ofCognitive Skills in Wage Determination.”Review of Economics and Statistics 77:251–66.

National Center for Education Statistics. 1998.Gender Differences in Earnings Among YoungAdults Entering the Labor Market (NCES 98-086). By Suzanne B. Clery, John B. Lee, andLaura G. Knapp. Washington, DC: U.S.Government Printing Office.

——. 2002. National Education Longitudinal Studyof 1988: Base-Year to Fourth Follow-Up Data FileUser’s Manuel (NCES 2002-323). By Thomas R.Curtin, Steven J. Ingels, Shiying Wu, and RuthHeuer. Washington, DC: U.S. GovernmentPrinting Office.

——. 2003a. The Condition of Education 2003 inBrief (NCES 2003-068). By Andrea Livingstonand John Wirt. Washington, DC: U.S.Government Printing Office.

——. 2003b. Postsecondary Attainment,Attendance, Curriculum, and Performance:Selected Results from the NELS:88/2000Postsecondary Education Transcript Study (PETS),

2000 (NCES 2003-394). By Clifford Adelman,Bruce Daniel, Ilona Berkovits. Washington, DC:U.S. Government Printing Office.

——. 2004. Trends in Educational Equity of Girls andWomen: 2004 (NCES 2005-016). By CatherineE. Freeman. Washington, DC: U.S.Government Printing Office.

——. 2005. Gender Differences in Participation andCompletion of Undergraduate Education andHow They Have Changed Over Time. NCES2005-169 by Katharin Peter and Laura Horn.Washington DC: US Government PrintingOffice.

Padavic, Irene, and Barbara Reskin. 2002. Womenand Men at Work (2nd ed.) Thousand Oaks,CA: Pine Forge Press.

Paglin, Morton, and Anthony M. Rufolo. 1990.“Heterogeneous Human Capital, OccupationalChoice, and Male-Female Earnings Differences.”Journal of Labor Economics 8:123–44.

Reskin, Barbara. 1993. “Sex Segregation in theWorkplace.” Annual Review of Sociology19:241–70.

Riordan, Cornelius. 2003. “Failing in School? Yes;Victims of War? No.” Sociology of Education76:369–72.

Rosenfeld, Rachel A., and Gunn ElisabethBirkelund. 1995. “Women’s Part-Time Work: ACross-National Comparison.” EuropeanSociological Review 11:111–34.

Sadker, Myra, and David Sadker. 1994. Failing atFairness: How Our Schools Cheat Girls. NewYork: Simon & Schuster.

Shelton, Beth Anne, and Juanita Firestone. 1989.“Household Labor Time and the Gender Gapin Earnings.” Gender and Society 3:105–12.

Shu, Xiaoling, and Margaret Mooney Marini. 1998.“Gender Related Change in OccupationalAspirations.” Sociology of Education 71:43–67.

Sum, Andrew, Neeta Fogg, and Paul Harrington withIshwar Khatiwada, Shelia Palma, Nathan Pond,and Paulo Tobar. 2003. “The Growing GenderGaps in College Enrollment and DegreeAttainment in the U.S. and Their PotentialEconomic and Social Consequences.” Preparedfor the Business Roundtable, Washington, DC.Boston: Center for Labor Market Studies.Available online at http://www.brtable.org/pdf/943.pdf

Tomaskovic-Devey, Donald. 1993. Gender and RacialInequality at Work: The Sources and Consequencesof Job Segregation. Ithaca, NY: ILR Press.

U.S. Department of Labor, Bureau of LaborStatistics. 2004, September. Highlights ofWomen’s Earnings in 2003. Report 978.Available on-line at http://www.bls.gov/cps/cpswom2003.pdf




22 Bobbitt-Zeher

Waldfogel, Jane. 1998. “Understanding the ‘FamilyGap’ in Pay for Women with Children.” Journalof Economic Perspectives 12:137–56.

Wellington, Alison J. 1994. “Accounting for theMale/Female Wage Gap Among Whites: 1976and 1985.” American Sociological Review59:839–48.

Willingham, Warren W., and Nancy S. Cole. 1997.Gender and Fair Assessment. Mahwah, NJ:Lawrence Erlbaum.

Wilson, Kenneth L., Janet P. Boldizar. 1990. “GenderSegregation in Higher Education: Effects ofAspirations, Mathematics Achievement, andIncome.” Sociology of Education 63:62–74.

Donna Bobbitt-Zeher is a Ph.D. candidate, Department of Sociology, The Ohio State University.Her main fields of interest are gender and racial inequality and the intersection of the two, withparticular emphasis on educational stratification, workplace discrimination, and violence againstwomen. This article draws on her dissertation research on changing gendered patterns of educa-tional success and their implications for gender stratification in later life. In addition, she isresearching gender and pregnancy discrimination as part of the Ohio Discrimination Project.

The author thanks Doug Downey, Vinnie Roscigno, Dennis Condron, Lisa Garoutte, and LisaHickman for their helpful comments on drafts of this work. Direct correspondence to DonnaBobbitt-Zeher, Department of Sociology, 300 Bricker Hall, 190 North Oval Mall, Columbus, OH43210; e-mail: [email protected].




Sociology of Education The Gender Income Gap and the Role of Education On behalf of: American...

Documents

Transcript of Sociology of Education The Gender Income Gap and the Role of Education On behalf of: American...