February 01, 2017 | By Kenneth C. Tsang
NACE Journal, February 2017
In recent years, the issue of the gender pay gap has become more prominent than ever. However, varying approaches to studying this issue have created a rather muddled picture of the ways in which the pay gap manifests.
Perhaps the most widely cited explanation for the gap is that women simply are more likely to “self-select” into fields of study and professions associated with (relative) lower pay. For example, engineering is a profession dominated by men, while social work is a profession dominated by women; engineers get paid more than social workers. Therefore, in this scenario, the proposed remedy for the gender pay gap is for women to shift into more lucrative lines of work. While there is some truth to the self-selection argument, it has been criticized for precluding the notion that gender bias has anything to do with the gender pay gap.1
Among several competing explanations for the gap, one of the most intriguing is the theory of valuative bias which argues that, as certain fields of study and professions become more widely represented by women, employers—as well as society at large—characterize those areas as less valuable (i.e., “women’s work”) and in turn lower the associated wages (and vice versa). It is important to point out that valuative bias theory does not aim to explain why the demographics of fields of study and occupations change (as other theories do); it only aims to explain what happens when they do. Several studies have found evidence for valuative bias as a contributing—although not exclusive—factor to the gender wage gap in Germany, the Netherlands, Sweden, and the United States. One of the most exhaustive of these studies comes from Paul Allison, Paula England, and Asif Levanon (of Stanford University and the University of Pennsylvania)2 who argue that valuative bias has remained a pervasive force affecting wages in the United States since at least the 1950s and may have actually intensified during the 1980s. On the other hand, a study by Fabian Ochsenfeld (of Goethe University)3 questions the validity such findings on the basis that there exists no macro-level longitudinal data (as opposed to cross-sectional data) that sufficiently capture all of the necessary elements to “prove” valuative bias theory. Altogether, a review of the literature highlights three key reasons for why compelling evidence of valuative bias can be somewhat elusive, or at the very least, incomplete. (A number of other recent studies4, 5, 6, 7, 8, 9, 10 are listed in the End Notes.)
First, its effects may be undetectable when examining a relatively short period of time as these effects manifest gradually or can be obscured by idiosyncratic conditions. Second, wages are affected by a multitude of other factors, including, but not limited to, job type, industry, location, and education/skill requirements. A fair test of this theory must adequately control for these potential confounders and, to add further methodological complexity, their effects may amplify or diminish over time due to—among other things—evolving cultural, economic, and technological dynamics.11 For example, as the cost-of-living in certain metropolitan areas shifts over time, so too does the degree to which location confounds wages. Beyond these methodological obstacles, a third reason is that the effect of valuative bias is often accelerative (or decelerative) which means that wages may still be rising “naturally,” but at a slower rate than otherwise possible. As such, valuative bias, in some instances, cannot be discretely observed and can only be revealed through correlative analysis.
This article will apply the theory of valuative bias in a more specific context than has been much of the focus of past research: It examines starting salary offers for recent graduates of bachelor’s degree programs in three STEM fields—biology, computer science, and engineering. For the period from 1974 to 2011, the analyses herein will compare NACE’s starting salary offer data for recent bachelor’s degree graduates (which control for the aforementioned potential confounders) against the National Center for Education Statistics’ (NCES) degree completions data by gender.
Just to illustrate the broad imbalance in earning potential between men and women on the basis of academic field of study, Figure 1 displays the average starting salary offers of Class of 2015 bachelor’s degree recipients across 27 majors. In terms of degrees conferred in 2015, majors represented by 50 percent or more women are displayed in blue, while those represented by less than 50 percent women are displayed in orange. Even this simple comparison makes it clear that fields of study—STEM or otherwise—with wider male representation, for the most part, have far greater earning potential upon graduation. The median of the average starting salary offers for female-dominated majors was $38,125, while the median for male-dominated majors was $44,104—a difference of 14.5 percent.
Figure 1: Class of 2015: gender representation and average starting salary offer12, 13
Computer science and engineering are among a handful of majors—STEM or otherwise—that have seen any meaningful growth in average starting salary offer over the decades. By virtue of this, they have for quite some time been among the highest potential earners, if not the highest. In 2011 the average starting salary offer for computer science majors (adjusted for inflation into 2011 dollars14) was 22 percent greater than that of graduates overall; for engineering majors, it was 17 percent greater. By contrast, biology majors are among those with the least earning potential. From 1974 to 2010, their average starting salary offer has remained effectively flat, and in 2010 their average offer was 38 percent less than that of graduates overall. A rather jarring illustration of the plight of biology majors: From 1974 to 2010 their highest average starting salary offer was less than the lowest average offer for graduates overall. (Figure 2.)
Figure 2: Average starting salary offer (in 2011 dollars), by major: 1974-201115
Over the course of the 20th century, women have made enormous strides in educational attainment. During just the 38-year period covered in this analysis (1974 to 2011), the percentage of bachelor’s degrees conferred across all fields of study represented by women rose from 44 percent to 57 percent. However, the extent to which individual fields of study—i.e., biology, computer science, and engineering—did or did not coincide with this overall trend varied widely.
As shown in Figure 3, the field of biology, for the most part, fell in line with the overall trend: In 1974, women represented just one-third of degrees conferred but, in the last two decades, and particularly since 2003, have represented well over half. Women also saw widened representation among engineering graduates, running parallel to the overall trend but doing so at a great distance; women represented just 2 percent of degrees conferred in engineering in 1974, but since the turn of the century have represented nearly one-fifth. The field of computer science is rather interesting in that it did follow the overall trend through to the early-1980s, but since then has starkly defied that trend. Women’s share of computer science completions peaked at 37 percent in 1983, gradually slipped over the course of the 1980s and 1990s, and then plummeted after the turn of the century. Since 2006, women have represented less than one-fifth (18 to 19 percent) of computer science degrees conferred. While we can attribute much of the ups and downs of average starting salary offer to the mechanics of “supply and demand,” is it possible that these demographic shifts also had a valuative (or devaluative) effect on each field of study’s earning potential?
Figure 3: Percentage of bachelor’s degree completions represented by women, by major: 1974-201116
Controlling for the trend in the average starting salary offer for all graduates, from 1974 to 2011 the percentage of bachelor’s degree conferred in computer science represented by women had a strong negative correlation* with the average starting salary offer for computer science graduates. This indicates that, regardless of the broader “up and downs” of the job market, as women’s representation among computer science graduates has declined (at times rapidly), the average starting salary offer for computer science graduates has risen substantially. At the most extreme, computer science graduates had the greatest earning potential when women’s representation among them was the most limited (and vice versa). For biology graduates, there was a modest negative correlation** but, given the relative flatness of biology graduates’ salary offers over the years, this correlation may not be particularly meaningful. For engineering graduates, there was a very strong positive correlation*** but, given the high market demand for engineering graduates, this correlation is likely spurious. (In other words, there is most certainly a correlation, but it is more so just a “natural” coincidence.)
These observations suggest two things. First, while much of the high earning potential for computer science graduates can certainly be attributed to their high demand in a rapidly advancing technological landscape, these observations suggest—if somewhat tenuously—that they have actually received a boost because of the lack of women. Second, while there is no indication of valuative bias in regard to biology and engineering majors, is it clear that, among STEM majors, women are highly concentrated in a field with very low earning potential (biology) and have limited representation in fields with very high earning potential (computer science and engineering).
Despite the considerable focus there has been on the advancement of STEM education and, in turn, the STEM work force, not all STEM graduates are finding equal, or even comparable, success in transitioning into the work force—that is, if they do at all.17, 18, 19 Apart from the simple mechanics of “supply and demand,” there is some evidence that gender does play some ancillary role in maintaining, if not outright exacerbating, the gender pay gap in both a STEM context and the broader context of all fields of study. However, considering the parallel effects of self-selection, the precise extent of the role that valuative gender bias, in particular,may play remains unclear. Indeed, its effects may be modest if they exist at all.
To further muddle this picture, the concept of self-selection itself is less straightforward than one might expect. For example, while there may be merit to the socio-cultural argument that women have something of an aversion—perhaps a growing aversion—to the field of computer science (and the profession of programming),20 it has also been argued that a deliberate “masculinization” of the programming profession over the decades has gradually pushed women out of the field.21
Ultimately, the persistence of the gender pay gap is, whether purposeful or coincidental, the result of a more complexly layered blend of socio-cultural, organizational, and economic forces than one might realize. For these reasons, it is recommended that future research on this topic not be a “key hole” for any individual hypothesis but rather a broader, simultaneous exploration of the numerous complementary forces affecting the pay gap.
* r = −0.551, p < 0.01 ** r = −0.369, p < 0.05 *** r = 0.796, p < 0.01
1 Thompson, Stéphanie. “The simple reason for the gender pay gap: work done by women is still valued less.” World Economic Forum: April 2016.
2 Allison, Paul; England, Paula and Levanon, Asaf. “Occupational Feminization and Pay: Assessing Causal Dynamics Using 1950-2000 U.S. Census Data.” Social Forces: December 2009.
3 Ochsenfeld, Fabian. “Why Do Women’s Fields of Study Pay Less? A Test of Devaluation, Human Capital, and Gender Role Theory.” European Sociological Review: June 2014.
4 Allison, Paul; England, Paula and Wu Yuxiao. “Does bad pay cause occupations to feminize, Does feminization reduce pay, and How can we tell with longitudinal data?” Social Science Research: September 2007.
5 Blau, Francine and Kahn, Lawrence. “The Gender Wage Gap: Extend, Trends, and Explanations.” Forschungsinstitut zur Zukunft der Arbeit Institute for the Study of Labor. January 2016.
6 Bobbitt-Zeher, Donna. “The Gender Income Gap and the Role of Education.” Sociology of Education. January 2007.
7 de Ruijter, Judith; Schippers, Joop and van Doorne-Huiskes, Anneke. “Size of Causes of the Occupational Gender Wage-gap in the Netherlands.” European Sociological Review. September 2003.
8 Petersen, Trond and Saporta, Itzhak. “The Opportunity Structure for Discrimination.” American Journal of Sociology. January 2004.
9 Leuze, Kathrin and Strauss, Susanne. “Female-typical Subjects and Their Effect on Wage Inequalities Among Higher Education Graduates in Germany.” European Societies. 2014
10 Magnusson, Charlotta. “Mind the Gap: Essays on Explanations of Gender Wage Inequality.” Stockholm University: Swedish Institute for Social Research. 2010.
11 Ochsenfeld, 2014.
12 National Center for Education Statistics, Integrated Postsecondary Education Data System.
13 NACE Salary Survey, Spring 2016.
14 U.S. Bureau of Labor Statistics, CPI Inflation Calculator.
15 NACE Salary Survey Reports, 1974 - 2011.
16 Olson, Randal S. “Percentage of bachelor’s degrees conferred to women, by major (1970 - 2012).” June 2014.
17 NACE First-Destination Survey Reports, 2014 - 2015.
18 NACE Salary Survey Reports, 1974 - 2011.
19 NACE Student Survey Reports, 2007 - 2016.
20 Jacobs, Jerry A.; Kanny, Allison; Lehman, Kathleen; Lim, Gloria; Paulson, Laura; Sax, Linda and Zimmerman, Hilary B. “Anatomy of an Enduring Gender Gap: The Evolution of Women’s Participation in Computer Science.” UCLA Women in Technology Initiative: September 2016.
21 Frink, Brenda D. “Researcher reveals how ‘Computer Geeks’ replaced ‘Computer Girls.’” Stanford University Clayman Center for Gender Research: June 2011.
Kenneth C. Tsang is a research associate at NACE. He can be reached at email@example.com.
Overall unemployment rate
Bureau of Labor Statistics
Unemployment rate, bachelor’s degree grads age 20 – 24
Bureau of Labor Statistics
Mean starting salary, Class of 2021 bachelor’s degree graduate
Dashboard - Class of 2021 First Destinations
Projected hiring increase for the Class of 2023
Job Outlook 2023