The robustness of these analyses is tested with several alternative analytic strategies, which are described in detail in Methods and Materials. Supplemental decomposition analyses examining intersectional variation in these premiums among non-WAHM are discussed in the “Supplemental analysis” section in Materials and Methods.
H2: WAHM advantages in social inclusion, respect, rewards, and persistence intentions will not be fully accounted for by differences between WAHM and non-WAHM in human capital, background characteristics, job characteristics, work effort and attitudes, and family responsibilities.
Using decomposition analysis, a second set of models examines the extent to which these privileges by WAHM status can be accounted for by systematic differences between WAHM and non-WAHM STEM professionals along five categories of explanatory predictors: human capital, background characteristics, job characteristics, work effort and attitudes, and family responsibilities. Table 1 lists the specific predictors included in each category. Although differences in these characteristics may account for some of the variation in work experiences between WAHM and others, given the cultural and structural processes of privilege and bias noted above, WAHM advantages in inclusion, respect, rewards, and persistence may act in part as premiums—unearned benefits attached to WAHM status that are not accounted for by these explanatory factors.
If such WAHM status advantages are identified, the next analytic task to understand WAHM status privilege is to examine whether WAHM’s more positive work experiences can be attributed to differences between WAHM and non-WAHM in work and employment characteristics like human capital, work effort, and job type. Perhaps, WAHM tend to work more hours, are more dedicated to their jobs, or have more education on average than other STEM professionals, and this explains their greater inclusion, respect, rewards, and persistence intentions.
H1: Compared to WAHM, members of 31 other intersectional gender, race/ethnicity, LGBTQ status, and disability status groups will be less likely to report experiencing social inclusion and more likely to encounter harassment at work, will report less professional respect and career opportunities, will have lower average salaries, and will have lower persistence intentions (controlling for variation among respondents in education level, age, STEM field, and employment sector).
Third, WAHM may enjoy greater work rewards than other STEM professionals. Previous research has identified salary discrepancies in STEM by gender and race ( 45 46 ), and systematic salary gaps may exist by disability and LGBTQ status as well. Alongside monetary rewards, WAHM may be more likely to report career advancement opportunities in their STEM jobs (e.g., access to leadership roles) than their non-WAHM peers. Fourth, due in part to such differences in inclusion, respect, and rewards, WAHM may have the highest persistence intentions of any other group ( 8 16 ). These differences in inclusion, respect, rewards, and persistence may be evident even after accounting for possible variation among groups by education level, age, STEM field, and employment sector. Formally stated:
The culmination of evidence from single-axis paradigm research referenced above would suggest that WAHM may be more likely on average than members of all other groups to experience social inclusion (e.g., to feel like they fit in among colleagues) and less likely to encounter harassment in their STEM jobs ( 44 ). Reflecting gendered, racialized, heteronormative, and ableist notions of STEM competence and excellence noted above, WAHM may also be more likely on average than members of the 31 other intersectional demographic groups to report that their professional contributions and expertise are respected.
To assess intersectional variation along these dimensions, the sample is disaggregated into 32 intersecting demographic groups by gender (men and women), race [Asian, Black, Latinx and Native American/Pacific Islander (NAAPI), and white], disability status (persons with disabilities and persons without disabilities), and LGBTQ status (LGBTQ and non-LGBTQ). The analyses below compare WAHM’s work experiences to those of STEM professionals in the 31 other intersectional groups. See Methods and Materials for detailed operationalization and results from more fine-grained desegregation by gender (including transgender and gender nonbinary persons), race/ethnicity (including multiracial persons), sexual minority status, and disability status.
Intersectional privileges of WAHM status may manifest within many aspects of STEM work. To capture the possible scope of this privilege, this study examines four dimensions of work experience identified by past research as particularly consequential for STEM careers ( 9 14 ): (i) social inclusion and harassment experiences; (ii) professional respect by colleagues; (iii) career rewards, including salary and advancement opportunities; and (iv) intentions to stay in one’s STEM career long term (i.e., persistence intentions). These dimensions are interconnected and mutually reinforcing but conceptually distinct. Their breadth reveals whether WAHM privilege exists only in specific domains (e.g., respect) or whether this privilege is evident across a spectrum of STEM work experiences.
The present study uses data from a large national-level survey of 25,324 STEM professionals employed full time in the Unites States, collected as part of the STEM Inclusion Study [SIS; principal investigators (PIs): E.A.C. and T. Waidzunas]. The SIS dataset is composed of representative samples of the members of 21 STEM professional societies, including 8 national flagship societies in the physical, natural, and life sciences and mathematics; 6 interdisciplinary STEM societies; 5 national flagship disciplinary societies in engineering; and 2 STEM teaching–focused societies. The survey encompasses detailed demographic measures, multiple dimensions of work experiences and rewards, and a robust set of job characteristics. Questions include previously validated items as well as novel measures that were pretested and validated for this survey. See Methods and Materials for details.
This study examines whether, in the aggregate, WAHM experience intersectional privileges compared with members of other demographic groups along multiple dimensions of STEM workplace treatment. Empirical limitations of existing data have previously made comprehensive investigation of intersectional privilege difficult ( 17 ). Past national surveys (e.g., Scientists and Engineers Statistical Data System and U.S. Census supplemental surveys) have insufficient sample sizes of STEM professionals, lack a full suite of demographic measures (e.g., do not include LGBTQ status), or include too few work experience measures to allow for the analysis undertaken here.
It is not a settled matter that WAHM would experience the greatest levels of inclusion, respect, and rewards in STEM, however. Groups marginalized along multiple axes of difference may experience certain intersectional freedoms that lead to better work experiences, rewards, and respect ( 40 ), while socially dominant groups may experience constraints on their behaviors and affect that can lead to more negative outcomes ( 41 42 ). Also, not every heterosexual white man without disabilities will experience advantages associated with intersectional privilege. Some WAHM may encounter prejudice and discrimination on the basis of other disadvantaged statuses (e.g., age, nationality, and/or socioeconomic background). Others may be targets of negative workplace treatment disconnected from social status, such as generalized bullying or incivility ( 43 ).
Investigating the experiences of WAHM vis-à-vis other professionals also requires explicit consideration of processes of privilege in STEM. In most STEM inequality research and policy efforts, WAHM are taken as the neutral (often unspoken) standard against which the experiences of women, people of color, LGBTQ persons, and persons with disabilities are believed to deviate. However, privilege is not simply an absence of the disadvantages experienced by marginalized and minoritized persons ( 30 32 ); it involves distinct opportunities and benefits that only members of that group have full access to. In STEM, such privilege may be anchored in the historical and contemporary overrepresentation of WAHM, which facilitates processes such as homophily, opportunity hoarding, and “old boys” networks ( 33 34 ). WAHM privilege may also be tied to particular embodiments that are historically and culturally believed to most closely align with definitions of excellence, genius, and objectivity ( 35 36 ). For example, WAHM in STEM are more likely to be assumed by default to be intelligent and to produce work that is free from ulterior motives ( 37 39 ). Thus, WAHM privilege is not only the outcome of being spared from the sexism, racism, ableism, and heterosexism that non-WAHM STEM professionals may encounter, but WAHM may also experience unearned advantages that are culturally attached to their demographic status.
In contrast to the single-axis paradigm, the analytic tool of intersectionality considers how inequalities and privileges operating simultaneously along multiple axes of social status create an intersecting matrix of (dis)advantage ( 19 20 ). Rooted in theoretical advancements originating in Black feminist scholarship ( 21 ), intersectionality emphasizes the role of interconnected power relations and social and cultural structures in processes of inequality ( 22 23 ). Intersectional approaches to inequality attend both to divergent experiences within specific axes of disadvantage (e.g., how gendered processes are also racialized) ( 22 ) and to the convergence of experiences of disadvantage among different marginalized or minoritized groups compared to intersectionally dominant groups ( 22 24 ). Intersectional approaches have been used by social scientists to study other patterns of social inequality for decades ( 25 26 ) but have only recently begun to make their way into investigations of inequality in STEM ( 17 29 ). Yet, intersectionality is indispensable for understanding how sexism, racism, ableism, and heteronormativity are entwined in ways that reinforce intractable patterns of inequality in STEM.
These gaps have meant that a fundamental assumption of STEM inequality research has largely gone untested: that the most well-represented and presumed most socially advantaged population in STEM—white heterosexual men without disabilities—experience the most respect and rewards in STEM, compared with all other demographic groups. Although the culmination of existing research would seem to suggest that persons privileged along each of these axes would be best able to avoid negative work experiences in STEM, little research has investigated these intersectional (dis)advantages directly. Do white able-bodied heterosexual men (WAHM) really experience privileges in the STEM workforce unique to their intersectional status? If so, can these privileges be explained by systematic differences between WAHM and non-WAHM in qualifications and job type, or do these differences operate in part as premiums on WAHM status—social rewards that accompany this particular demographic status that cannot be accounted for by differences in human capital, job characteristics, work effort, or other factors? Drawing insight from intersectionality and social privilege literatures, this study uses a large, national-level dataset of U.S. STEM professionals to compare the work experiences of WAHM to STEM professionals in 31 other intersectional gender, race, LGBTQ status, and disability status categories.
Despite these advancements, STEM inequality research has largely operated within a single-axis paradigm, focusing on only one dimension of inequality (e.g., gender or race or LGBTQ status) at a time. Scholarship in the single-axis paradigm has been vital for revealing sexism, racism, heteronormativity, and ableism in STEM education and STEM workplaces, yet reliance on this paradigm has facilitated two gaps in STEM inequality research: lagged attention to intersectional patterns of inequality and lack of investigation into structural and cultural processes of privilege ( 17 18 ).
Scholars have made important strides in documenting the extent and forms of disadvantage that women and members of minoritized racial/ethnic groups face in STEM ( 6 10 ) and have started to demonstrate that similar disadvantages exist for lesbian, gay, bisexual, transgender, and queer (LGBTQ)–identifying persons and persons with disabilities ( 11 15 ). Such research reveals that inequality in STEM not only is an issue of (under)representation but also involves processes of marginalization and devaluation at the structural, cultural, and interpersonal levels ( 9 16 ).
The diversification of science, technology, engineering, and math (STEM) fields has largely stagnated over the past 2 decades, despite substantial national and institutional investments aimed at recruiting and retaining underrepresented populations ( 1 2 ). This trend is concerning not only because more diverse groups of problem solvers tend to produce more innovative and creative solutions ( 3 5 ) but also because it indicates that STEM is failing to live up to its goals of equity in opportunities and outcomes.
RESULTS
x axis), so that the height of each bar represents the divergence of that group’s mean from the mean for WAHM. Intersectional groups are listed in order from smallest to largest differential. Error bars indicate 95% confidence intervals (CIs = 1.96 × SE), illustrating the significance of the difference in means between WAHM and each focal group, net of controls. As described in detail in the “Operationalization of demographic measures” section in Materials and Methods, several smaller groups were aggregated (Native American and Pacific Islander with Latinx respondents, persons across LGBTQ categories, and persons across disability types) to protect the confidentiality of respondents in particularly minoritized groups and to ensure statistical power. Table S6 presents disaggregated means on each outcome for subgroups within these aggregated categories. Figures 1 to 6 present results from the tests of H1. The bar graphs represent means on each outcome for the 32 intersectional demographic groups, holding constant variation in respondents’ education level, age, STEM field, and employment sector. The values in each figure are centered at the mean for WAHM on that measure (represented by theaxis), so that the height of each bar represents the divergence of that group’s mean from the mean for WAHM. Intersectional groups are listed in order from smallest to largest differential. Error bars indicate 95% confidence intervals (CIs = 1.96 × SE), illustrating the significance of the difference in means between WAHM and each focal group, net of controls. As described in detail in the “Operationalization of demographic measures” section in Materials and Methods, several smaller groups were aggregated (Native American and Pacific Islander with Latinx respondents, persons across LGBTQ categories, and persons across disability types) to protect the confidentiality of respondents in particularly minoritized groups and to ensure statistical power. Table S6 presents disaggregated means on each outcome for subgroups within these aggregated categories.
Predicted means for each category, holding constant variation by STEM field, employment sector, highest education, and age. Scale on the “social inclusion” measure ranges from 1 (strongly disagree) to 5 (strongly agree), with higher number representing stronger agreement. Values represent the average divergence of each group’s experiences from those of WAHM. Values were produced by ordinary least squares (OLS) regression models with gender × race × LGBTQ status × disability status interaction terms. See the “Supplemental analysis” section in Materials and Methods for details. Error bars represent 95% confidence intervals. N = 25,324. WAHM, white heterosexual men without disabilities; NA, Native American and Pacific Islander.
Predicted rates of harassment experiences for each category, holding constant variation by STEM field, employment sector, highest education, and age. Values represent the average divergence of each group’s experiences from those of WAHM. Values were produced by logistic regression models with gender × race × LGBTQ status × disability status interaction terms. See the “Supplemental analysis” section in Materials and Methods for details. Error bars represent 95% confidence intervals. N = 25,324.
Predicted means for each category, holding constant variation by STEM field, employment sector, highest education, and age. Scale on the experiences of professional respect measure ranges from 1 (strongly disagree) to 5 (strongly agree), with higher number representing stronger agreement. Values represent the average divergence of each group’s experiences from those of WAHM. Values were produced by OLS regression models with gender × race × LGBTQ status × disability status interaction terms. See the “Supplemental analysis” section in Materials and Methods for details. Error bars represent 95% confidence intervals. N = 25,324.
Predicted means for each category, holding constant variation by STEM field, employment sector, highest education, and age. Values represent the salary differences of each group compared to WAHM. Values were produced by OLS regression models with gender × race × LGBTQ status × disability status interaction terms. See the “Supplemental analysis” section in Materials and Methods for details. Error bars represent 95% confidence intervals. N = 25,324.
Predicted means for each category, holding constant variation by STEM field, employment sector, highest education, and age. Scale on the career advancement opportunities measure ranges from 1 (strongly disagree) to 5 (strongly agree), with higher number representing stronger agreement. Values represent the average divergence of each group’s experiences from those of WAHM. Values were produced by OLS regression models with gender × race × LGBTQ status × disability status interaction terms. See the “Supplemental analysis” section in Materials and Methods for details. Error bars represent 95% confidence intervals. N = 25,324.
Predicted means for each category, holding constant variation by STEM field, employment sector, highest education, and age. Scale on the persistence intentions measure ranged from 1 (less than 5 years) to 5 (“the rest of my career”), with higher number representing stronger agreement. Values represent the average divergence of each group’s experiences from those of WAHM. Values were produced by OLS regression models with gender × race × LGBTQ status × disability status interaction terms. See the “Supplemental analysis” section in Materials and Methods for details. Error bars represent 95% confidence intervals. N = 25,324.
Figure 1 presents results on the social inclusion scale for each intersectional group, centered at the mean for WAHM. As illustrated by the negative values and CIs, members of all other 31 intersectional groups experience significantly less social inclusion in their STEM jobs on average than WAHM experience, net of differences by STEM field, sector, education level, and age. There is wide variation in each group’s average departure from WAHM’s inclusion experiences. The divergence from WAHM’s social inclusion experiences is smallest (but still significantly more negative) for heterosexual Asian men without disabilities, and largest for LGBTQ Black women with disabilities. Patterns across specific gender, race, LGBTQ, and disability status groups are summarized below and detailed in the “Supplemental analysis” section in Materials and Methods.
P < 0.10, two-tailed test]). Fourteen percent of heterosexual Latinx/Native American men without disabilities (the next lowest) experienced harassment, while more than one in three (38%) LGBTQ-identifying Latinx and Native American women with disabilities faced harassment in the last year—over three times the harassment rate experienced by WAHM. As shown in Fig. 2 , 30 of the 31 intersectional groups experience significantly higher rates of harassment than WAHM (LGBTQ white men without disabilities experienced slightly higher rates of harassment [11.8%] than WAHM [9.9%], but this difference is only marginally statistically significant [< 0.10, two-tailed test]). Fourteen percent of heterosexual Latinx/Native American men without disabilities (the next lowest) experienced harassment, while more than one in three (38%) LGBTQ-identifying Latinx and Native American women with disabilities faced harassment in the last year—over three times the harassment rate experienced by WAHM.
A second hypothesized dimension of WAHM privilege is greater access to professional respect. Figure 3 presents bar graphs for group averages on the professional respect scale, centered at the mean for WAHM. Consistent with H1, all 31 other intersectional groups reported experiencing significantly less professional respect in their STEM jobs than WAHM reported experiencing, net of controls.
Figures 4 and 5 present results for the salary and professional opportunity measures. Holding constant variation in respondents’ STEM field, sector, education level, and age, WAHM have significantly higher salaries on average compared with every other intersectional group ( Fig. 4 ). These salary gaps are largest for Latinx and Native American women and men across disability and LGBTQ statuses: each of these groups has average salaries that are at least $30,000 lower than WAHM employed in the same STEM fields and sectors and with the same education level and age. In addition, persons with disabilities across gender, race, and LGBTQ status experience a $20,000 average salary deficit compared with WAHM peers. Figure 5 illustrates similar patterns in respondents’ access to career opportunities: compared with WAHM, all other intersectional groups reported significantly lower advancement opportunities in their STEM jobs, holding constant employment sector, field, age, and education level.
The final outcome of interest in H1 is persistence intentions. Figure 6 presents the results for STEM professionals’ intentions to remain in their STEM field for the rest of their career by demographic group, centered at the mean for WAHM. Here, all other groups except heterosexual white men with disabilities have significantly lower persistence intentions than WAHM. Persistence intentions are particularly low for LGBTQ-identifying nonwhite STEM professionals compared with WAHM, and for nonwhite persons with disabilities.
The intersectional patterns of disadvantage revealed by these figures are discussed in more detail in the “Supplemental analysis” section in Materials and Methods. Together, these figures show that the work experiences of LGBTQ-identifying Black women, Latinx and Native American women, and persons with disabilities tend to diverge the most from the experiences of WAHM. Yet, these intersectional processes are not consistently additive: marginalization along the greatest number of demographic axis is not always accompanied by the highest degree of difference from the experiences of WAHM. Furthermore, supplemental analyses assessed the effect sizes of individual identity dimensions in the context of the other intersectional categories. Those results revealed that which identity category had the greatest consequence for workplace experiences depends on the outcome in question. Variation by gender incurred the greatest consequences for experiences of harassment and turnover intentions, while race/ethnicity incurred the greatest consequence for experiences of social inclusion, respect, and professional opportunities (see table S5). These analyses are discussed in more detail in Materials and Methods and underscore the need for more research into the nuances of intersectional patterns.
Together, Figs. 1 to 6 illustrate that WAHM, compared to 31 other intersectional demographic groups, are advantaged across all four workplace experience dimensions. Can these privileges be explained by variation in employment-related factors such as human capital and work commitment, or WAHM’s potentially greater likelihood of working in sectors where positive work experiences are more abundant?
H2 hypothesized that WAHM advantages across the four work experience dimensions operate in part as premiums—benefits that accompany WAHM status that do not accrue to other groups of STEM professionals when they have identical traits like human capital, job characteristics, and work effort. Blinder-Oaxaca decomposition analysis allows for the partitioning of the gap between WAHM and their non-WAHM peers on each outcome into an explained portion (attributed to variation between groups) and the portion of the gap that cannot be accounted for by variation in these characteristics.
9, 47– Each decomposition model includes explanatory predictors from five categories of factors shown in STEM inequality literature to be important drivers of work experiences: human capital (e.g., education level, organizational tenure); job characteristics (e.g., employment sector, supervisory status, primary work responsibility); background characteristics (e.g., parental education, whether born in the United States); work effort and attitudes (e.g., hours worked, personal commitment to STEM job); and family responsibilities (e.g., having a young or school-aged child or eldercare responsibilities) ( 2 50 ). See Table 1 for the list of factors in each category. It may be, for example, that WAHM work more hours on average than non-WAHM professionals, are more committed to their work, or tend to be more likely to do work aligned with their highest degree than non-WAHM STEM professionals, and this explains WAHM’s greater likelihood of experiencing inclusion, respect, and rewards.
Figure 7 presents bars that partition the gap between WAHM and non-WAHM on each outcome into the portion accounted for by variation in the explanatory factors (the shaded segments of each bar) and the portion of the differences that remain after these factors are accounted for (the unshaded segments). Table S2 presents detailed decomposition results, including the proportion of the explained variation attributable to each specific explanatory factor.
Shaded segments represent the portion of the difference between WAHM and non-WAHM STEM professionals explained by variation in human capital, background characteristics, job characteristics, work effort and attitudes, and family responsibilities (shaded segments), and the unshaded segments and accompanying percentages represent the portion that remains unexplained by variation in these factors. N = 25,324. See table S2 for full decomposition models.
The leftmost bar in Fig. 7 represents the decomposition of the WAHM/non-WAHM gap in social inclusion experiences into explained (shaded) and unexplained (unshaded) portions. Here, the total portion of the gap between WAHM and non-WAHM in social inclusion experiences that can be accounted for by all explanatory measures combined is less than 14%. Variation in work attitudes and effort (segment 3) explains the greatest portion of this gap. The unexplained portion of the gap in social inclusion experiences is more than six times as large as the explained portion; even if non-WAHM were identical to WAHM on each factor across the five categories, 86% of the gap in inclusion experiences would still remain. As shown in table S2 and explained below, variation in background characteristics, particularly age, has an offsetting contribution to the gap in social inclusion (i.e., WAHM are older on average than non-WAHM, but older STEM professionals tend to experience less social inclusion than younger professionals), and thus, this segment sits below the zero line.
Similar to social inclusion, only small portions of the differences in harassment experiences (second bar in Fig. 7 ), professional respect (third bar), and career opportunities (fifth bar) by WAHM status can be explained by variation between WAHM and non-WAHM in human capital, job characteristics, work effort, family responsibilities, and the other factors. The rest of the variation (81.1% for harassment experiences, 83.3% for professional respect, and 59.0% for career opportunities) are benefits accompanying WAHM status that cannot be attributed to these factors.
The average salary gap between WAHM and non-WAHM STEM professionals is $24,994. Variation in work-related characteristics between WAHM and non-WAHM accounts for 68.7% of this gap. Yet, 31.3% of the salary gap remains unexplained: WAHM earned $7831 more on average than non-WAHM STEM Professionals even when non-WAHM had the same human capital, job characteristics, work effort, background work-related characteristics, and family responsibilities and when they worked in the same STEM fields and sectors.
Last, regarding persistence intentions, variation between WAHM and non-WAHM on the explanatory predictors accounts for less than a third of WAHM’s greater intentions to stay in STEM. Thus, even with identical values on these work-related characteristics, non-WAHM are still significantly less likely to intend to persist in their STEM jobs than their WAHM peers.
In sum, and supporting H2, sizable portions of the WAHM advantages in inclusion, respect, rewards, and persistence intentions shown in Figs. 1 to 6 appear to operate as premiums attached to WAHM status itself—benefits that cannot be attributed to these differential job and work characteristics between WAHM and non-WAHM. Differentials were most fully accounted for by the explanatory predictors in the case of salary because salary is heavily determined by structural and labor market positions, which are also highly sociodemographically differentiated. Yet, even there, nearly a third (over $7800) of the $25,000 average salary differential remained once these differences were accounted for. The unexplained portions of the gap on the other outcomes ranged from 59 to 86% of the total differential—two to six times larger than the explained portions. Although decomposition analysis cannot capture all possible factors that may help account for these differences, in well-crafted decomposition models with robust sets of explanatory predictors, such large portions of unexplained variance are typically attributed to premiums attached to membership in a privileged social category ( 51 53 ).
The “Supplemental analysis” section in Materials and Methods below explores variability within the non-WAHM groups with a series of decomposition models that examine WAHM premiums vis-à-vis specific disaggregated non-WAHM groups. These models reveal similar patterns of intersectional variability as those highlighted above. Specifically, the premium for WAHM status is largest in comparison to Black women across LGBTQ and disability statuses, and for persons with disabilities across gender, race/ethnicity, and LGBTQ status. These analyses provide further motivation for nuanced, multimethod intersectional analyses of workplace experiences among STEM professionals.
The “Supplemental analysis” section in Materials and Methods also reports the results of robustness tests that rerun the central models of the study using several alternative analytic approaches. Results are fully consistent with those above and provide greater detail on these intersectional patterns. For example, to test the alternative explanation that these patterns are driven by uniformly more negative work attitudes among non-WAHM STEM professionals (rather than a reflection of WAHM privilege), all models were rerun controlling for respondents’ job satisfaction. Doing so does not explain away these patterns (see table S3). Supplemental analysis also examines the contribution of WAHM’s advantages in social inclusion and respect to their greater persistence intentions. The Supplementary Materials provides additional details on these intersectional and disaggregated patterns among non-WAHM (see tables S4 and S5).