Early-Career Assignments and Workforce Inequality in Engineering

ABSTRACT

Positioned as part of leadership development in many organizations, ‘stretch assignments’ are a type of work assignment that can prove someone’s readiness to advance in their career. Informed by status characteristics theory, our research investigates the frequency and expected outcomes of stretch assignments among recent engineering graduates in the workforce. Findings suggest that early-career stretch assignments, especially assignments involving new and unfamiliar areas, potentially intensify gender and racial/ethnic workforce inequality. Other types of assignments that may be more familiar and clearly-scoped to early-career engineers show a different and less inequality-intensifying pattern. We discuss why early-career engineers’ assignments may be sites of inequality and the need for more focus on organizational processes around career-advancing work.

KEYWORDS:

• Gender
• race/ethnicity
• status characteristics theory
• early-career engineers
• stretch assignments

Introduction

It is well known that gender and racial/ethnic inequality increases up the job ladder in U.S. engineering occupations, from entry to managerial and leadership ranks.¹ Studies of inequality in and beyond engineering show how sameness is systematically rewarded within organizations and how stereotypes undermine the advancement of minoritized gender, racial, and ethnic groups.² Promotion disadvantages start to accrue to women, particularly Black, Latinx, Indigenous, and Pacific Islander women, in entry-level engineering ranks, controlling for performance and time in position.³

The present study investigates how the assigned work tasks of early-career engineers reinforce gender and racial/ethnic inequality at the earliest stages of workforce participation and foreshadow increasing inequality up the job ladder. Our focus is on ‘stretch assignments’, defined as levers of leadership development in organizations that are informally allocated and can prove someone’s readiness to advance in their career. Drawing on longitudinal survey data collected from more than 600 recent U.S. engineering graduates, we apply sociological theory of status-based inequality to examine the frequency of stretch assignments and the anticipated rewards engineers may associate with them. We show that more than other types of early-career assignments, stretch assignments, especially those involving new and unfamiliar areas, may advantage higher-status gender and race/ethnicity groups in engineering and carry long-term workforce consequence.

Background

Contexts of early-career work assignments

Little research has examined how organizational practices around early-career work assignments systematically contribute to inequality in engineering.⁴ However, studies of engineering work identify three broader contexts that set the stage for understanding why stretch assignments in particular could be so differentiating in early-career engineering in particular: ambiguity embedded in engineering problem-solving; value ascribed to technical merit and the belief in (technical) meritocracy; and unevenness in career networks and information in the early career. We note our definition of ‘early-career engineers’ incorporates both degree and time-since-degree, i.e. those graduating with an engineering bachelor’s degree and entering the workforce into an array of jobs, no more than a few years out of their undergraduate programs (this includes ‘non-traditional-aged’ graduates and non-linear educational pathways).

First, engineering work broadly involves technical problem-solving in collaborative, distributed teams, where problems are open-ended with few precedents and many possible solutions.⁵ More so than in undergraduate engineering settings, the initial work experiences of early-career engineers tend to be ‘informal, relatively unstructured [and] based largely on self-directed trial and error’⁶, underscoring the ambiguity and space for interpretation even in early-career problem-solving.

Second, engineering culture emphasizes technical merit and achievement as definitive of the field and more so than other critical dimensions of professional practice. Engineers’ professional self-concepts and rewards center on making technical contributions⁷; organizational learning and social dynamics are downplayed in engineers’ professional preparation, identity, values, and day-to-day work.⁸ Advancement and success are believed to be highly (techno-) meritocratic, in step with merit-based performance systems of many organizations.⁹ In practice, however, performance evaluation of engineers also involves reputation and consensus,¹⁰ and promotion outcomes differ by gender and race/ethnicity even among equally qualified engineers.¹¹

Third, early-career engineers often develop understandings of workplace practices and possibilities incrementally, through everyday interactions with colleagues, managers, and other organizational stakeholders.¹² Organizational division of labor and how teams roll up to a ‘bigger picture’ can be obscure.¹³ Positive and negative work experiences track with perceptions of support from co-workers,¹⁴ with early-career women reporting fewer instances of support and social connectedness at work than do early-career men.¹⁵ Differences in networks by race/ethnicity or at the intersection of gender and race/ethnicity are less studied but may be stark.¹⁶

Together, these contexts indicate that early-career work takes place in ambiguous problem and people settings, where technical achievement is highly valued even if daily practice involves more than technical tasks, and support and resources are more sporadic than systematized. We now turn to specific organizational levers of career advancement, namely stretch assignments, that may operate especially well within these three contexts to advantage some groups of early-career engineers more than others. We explain how this advantage may work over the next two sections, beginning with description of a ‘stretch’.

Stretch assignments in engineering work

Stretch assignments can be characterized as developmental work tasks with consequences for career mobility. They are closely related, if not interchangeable, with such terms as ‘challenging’, ‘developmental’, or ‘growth’ assignments. In practice, they are informal levers of leadership development and succession planning in organizations.¹⁷ To build and promote talent, managers allocate and reward stretch assignments with a high degree of discretion. Tobias Neely et al. describe how stretch assignments are part of ‘individualized career maps’ curated by smaller manager ranks in the flexible ‘new economy’.¹⁸ A successful assignment can be linked to promotion and monetary bonuses, but these are outcomes usually decided by managers on a case-by-case basis, rather than protocol or policy.

Operationally, stretch assignments have an element of newness or unfamiliarity and an element of high visibility and high stakes for individuals doing them.¹⁹ Software engineering managers identified two primary components of a stretch assignment in Tobias Neely et al.’s engineering case: visibility in the organization (the assignment is ‘known to others and gets you known’), and capacity-building learning (the assignment involves acquiring a new skill and is ‘outside of your comfort zone’).²⁰ These assignments were embedded in collaborative technical project work. The visibility and new skills built into the stretch assignment proved, to the manager and their peers, or to a set of leaders with vested interest in project outcomes, a person’s readiness for the next position or role, provided it was completed successfully.

Retrospective, field, and experimental studies of career development show the importance of stretch assignments for professional advancement and their role in differential career outcomes for women and men. Although not specific to engineering fields, these studies collectively demonstrate that women do not have the same access to stretch assignments, equally scoped assignments, and the same career returns on such assignments compared with men.²¹ While few have studied racialized dynamics of stretch assignments in professional settings, scholarship on racialized work tasks and institutions is relevant. In one study of Latinx and White social service professionals in a non-profit organization, tasks were divided in ways consistent with ‘ethnoracial logics’ where Latinx employees were ‘tethered’ to minoritized markets while their White colleagues enjoyed career flexibility and racial abstraction.²² Intersectional analysis indicates that ‘desirable assignment’ disparities are salient by race and gender,²³ with minoritized groups facing additional, unrecognized work of ‘fitting in’ at their organizations.²⁴

Yet disparities alone do not show how advantage can systematically accumulate in stretch assignment conditions, particularly among early-career engineers. Studies of early-career stretch assignments are scarce overall. We draw from sociological study of group inequality and status characteristics theory for deeper understanding of cumulative advantage and application to early-career engineering work.

Why stretches differentiate: status characteristics theory

Status characteristics, such as gender and race, are culturally based and reflect widely held beliefs about group differences. A status characteristic denotes higher perceived value and greater perceived competence, i.e. higher status, ascribed to members of one group more so than members of another group in everyday social interactions.²⁵

Status characteristics are more or less influential depending on the type of social setting and the degree to which group membership is closely associated with being successful there. In settings where the status characteristic differentiates people or is strongly linked to a task at hand, the status characteristic is salient and powerful.²⁶ The characteristic sustains and even intensifies the status structure in two ways: first, through one’s self-assessments of their ability, wherein individuals evaluate their own skills in line with cultural beliefs about those skills, not their actual performance²⁷; and second, through broader performance expectations, wherein others expect people in the higher status category to perform better than those in the lower-status category.²⁸ The performance expectation alone can drive bias – groups assumed to be more competent are more positively evaluated, while groups assumed to be less competent are less positively evaluated, even among individuals who otherwise have identical qualifications. The biasing effect of cultural beliefs about group differences is heightened when those qualifications or the evaluative context are ambiguous; people draw from group stereotypes, making ‘third-order inferences’ based on what they think others think when there is perceptibly little else on which to base a decision.²⁹

Engineering work is an exemplary context to consider how status characteristics intensify inequality. Gender is a salient status characteristic in engineering settings, rooted in the social construction of gender as binary.³⁰ The salience of gender derives from two conditions: it starkly differentiates the field and is assumed to be highly relevant to highly valued technical tasks, i.e., masculinity is equated with mathematical, scientific, and technical skill, and femininity with the presumptive opposite, social skill.³¹ Race/ethnicity acts in similar but not identical ways. White and Asian engineers are more prevalent in most U.S. technical settings than are Black, Latinx, Indigenous, and Pacific Islander engineers,³² and Asian groups, in particular, are culturally associated with having high technical skill.³³ Black, Latinx, Indigenous, and Pacific Islander groups are perceived as having neither high technical nor high social skill in this cultural frame; the technical-social dualism is both gendered and racialized (mapping to White-centered gender stereotypes).³⁴ Lower status for Black, Latinx, Indigenous, and Pacific Islander engineers – especially women – may derive more from being ‘non-prototypical’ of who is culturally understood to be an engineer (Asian/White, male) and thus rendered invisible in engineering settings.³⁵ As a result, we would expect Asian and White men to be given the benefit of the doubt, and to give themselves the benefit of the doubt, more so than would any other group of engineers under (especially ambiguous, novel) problem-solving conditions involving evaluation of their technical performance.³⁶

It follows that stretch assignments in engineering offer theoretically rich conditions to investigate status-based bias. By design, they are novel and unfamiliar and almost completely capture the idea of performance expectation. They represent a managerial bet on someone’s ‘potential’, executed and evaluated in a strongly gendered and racialized technical context. In fact, the more the assignment involves novel technical work, the more we might expect status characteristics to influence who gets a stretch assignment and how they are evaluated on it, with which prospective rewards. Critical to our argument, the biasing role of status characteristics in stretch assignments may be especially relevant to early-career engineers, who are most unproven in already ambiguous work settings where technical identities and contributions are primary and organizational processes/networks can be obscure – three decisive contexts in early-career engineering work.

While we do not experimentally test these scenarios in our study, we use status characteristics theory as an interpretive frame to explain why stretch assignments may intensify social inequalities among early-career engineers. Our study focuses on gendered and racialized patterns in early-career engineers’ tasks and expected rewards and considers how those patterns reflect ‘status at work’. In doing so, we extend research on early-career engineering practice into new sociological space.

Research questions

Our study explores early-career stretch assignments for two sets of groups: one based on gender, where men and women compose, respectively, socially dominant and minoritized engineering groups, and one based on race and ethnicity, where Black, Latinx, Indigenous, and Pacific Islander people compose a socially minoritized engineering group, and Asian and White people compose a socially dominant engineering group. Dominant/minoritized labels intentionally denote historical and ongoing social-structural inclusion/exclusion in engineering practice.³⁷ Status – including widely held cultural beliefs about technical competence, extending to beliefs about potential – is ascribed to each group. While we hope that grouping in this way helps to begin a conversation about systematic biasing of stretch assignments, there are assumptions and limitations built into our categorizations, discussed further in sections below.

We pose two sets of research questions (RQs). The first concerns access to four types of early-career work assignments: two measures of stretch assignments (visibility and new area assignments), and two measures of assignments selected for comparison (documentation and skill application assignments). We treat group differences in assignment frequency as signaling differences in access, adjusting for individuals’ performance, confidence, and experience.

RQ1: Assignment access: How often do early-career engineers work on stretch assignments, compared with other types of assignments? To what extent does assignment frequency vary across gender and race/ethnicity groups? How do college experiences and job characteristics explain between-group differences in assignment frequency?

The second concerns four possible outcomes of these assignments, holding that past and current assignments will influence expected career outcomes. We investigate if stretch assignments have intended effects, for which gender and race/ethnicity groups, and/or if other types of assignments act as (potentially more equitable) advancement levers.

RQ2: Assignment outcomes: What are the net effects of stretch and comparison assignments on engineers’ anticipated career outcomes? To what extent do these effects vary across gender and race/ethnicity groups? Under which conditions is an assignment inequality-intensifying, equality-intensifying, or inequality-preserving?

We define each italicized term in RQ2 as follows:

• Inequality-intensifying: The positive effect of an assignment is stronger for a dominant engineering group and weaker and/or negative for a minoritized engineering group, i.e., the assignment potentially fast-tracks groups more likely to advance to leadership roles. Inequality-intensifying scenarios can reflect status at work vis-a-vis evaluative advantages to higher-status versus lower-status groups (regardless of experience, performance, and so on).

• Equality-intensifying: The positive effect of an assignment is stronger for a minoritized engineering group and weaker and/or negative for a dominant engineering group, i.e., the assignment potentially corrects for systemic bias and leadership gaps. Equality-intensifying scenarios challenge status-affirming conditions and/or are made possible in the absence of those conditions.

• Inequality-preserving: The effect of an assignment is statistically equivalent for dominant and minoritized engineering groups and does not substantively alter what we would expect in status quo (unequal) circumstances.

By classifying our pattern of effects this way, separating access questions from expected outcomes questions, and breaking down stretch assignments into sub-parts with comparison measures, we can explore how stretch assignments are, or are not, the sites of status-based bias we conceive them to be. Specifically, this paper addresses the extent to which stretch assignments are an inflection point for inequality in early-career engineering, more so than other early-career assignments, and whether there are aspects of early-career engineering practice that might counteract status advantages.

Methods

Sample

Our study sample comprises 625 employed individuals who, as of Fall 2017, had earned an engineering bachelor’s degree up to three years earlier. All 625 individuals participated in the first and third waves of the Engineering Majors Survey (EMS), a National Science Foundation (NSF)-funded longitudinal investigation of engineering students’ transition from school to work.³⁸ The first wave of the survey, known as EMS 1.0, was administered to over 35,000 undergraduate engineering students across a stratified sample of 27 U.S. engineering schools in Winter/Spring 2015, yielding 7,197 respondents. The third wave, EMS 3.0, was administered in Fall 2017 to approximately 3,000 EMS 1.0 respondents who had agreed to participate in follow-up survey research, yielding 1,058 respondents. Of these, 625 met the main criteria for inclusion in our analyses: employed and not simultaneously enrolled as undergraduate students as of EMS 3.0 and, given our study’s focus, reported both gender and race/ethnicity on EMS 1.0.³⁹

Table 1 compares our sample and the U.S. degree-earning population using our aggregate gender and race/ethnicity groupings. The Institutional Review Board at Stanford University approved all research procedures (IRB #31803/#35539), and participants provided informed consent on all surveys.

Table 1. EMS respondents by gender and race/ethnicity groups¹.

	EMS (2015–2017)	U.S. undergraduate engineering degree-earners (2016)^2
	% (n)	%
Women	35 (216)	21
Men	65 (409)	79
Asian/White	88 (552)	83
Black/Latinx/Indigenous/Pacific Islander	12 (73)	17
Asian/White women	30 (189)	17
Asian/White men	58 (363)	66
Black/Latinx/Indigenous/Pacific Islander women	4 (27)	4
Black/Latinx/Indigenous/Pacific Islander men	7 (46)	13

¹ Refer to Variables for coding details.

² Authors’ tabulations of data reported in NSF, Table 5–7, 2016. Denominator (N = 92,414) limited to U.S. citizen/permanent resident degree-earners providing race/ethnicity data that could be classified into EMS categories.

Note. n’s for stricter-criteria RQ2 models: 179 Asian/White women, 350 Asian/White men, 25 Black/Latinx/Indigenous/Pacific Islander women, 45 Black/Latinx/Indigenous/Pacific Islander men. Percentages may not add to 100.0 due to rounding.

Variables

Assignments

We examine four EMS 3.0 work assignment measures as dependent variables (RQ1) and independent variables (RQ2) in our statistical models. The EMS question prompt read, ‘Since starting your current job, how often have your assignments involved … ’ and each item was measured on a five-point scale (0 = Never, 1 = Rarely, 2 = Sometimes, 3 = Often, 4 = Very Often):

• Working on something that generates interest and feedback from others in your organization (Stretch Assignment: Visibility)

• Working in areas or domains that are unfamiliar to you (Stretch Assignment: New Area)

• Documentation (Comparison Assignment: Documentation)

• Skills that you learned as part of your undergraduate education (Comparison Assignment: Skill Application)

Our two stretch assignment measures were expressly designed to capture facets of a stretch assignment as conceived by engineering managers in previous research,⁴⁰ but adapted to an early-career work setting. For instance, popular business phrases such as ‘high-visibility work’ or ‘high-stakes work’ might not resonate with early-career engineers, semantically or experientially. Work that ‘generates interest and feedback’, by contrast, is a plainer phrase and experience that more early-career respondents would likely be able to evaluate while still capturing the visibility-to-others theme.⁴¹ We note that prior studies of stretch assignments do not specify the ratio of visibility work to new area work in stretches. We explore descriptive associations between visibility and new area assignments, but purposefully model each measure separately.

Our two comparison assignments, documentation and skill application, consider early-career work that could be more familiar to recent graduates in new jobs but may not fall within the range of a stretch assignment as understood by managers and organizations. For instance, documentation may exemplify routine engineering work with organizational value; most engineers, including recent graduates, are responsible for some documentation as part of everyday practice.⁴² Yet, it often has little recognition as a source of technical value creation, cast as administrative work distinct from design and innovation and not ‘real’ engineering.⁴³

The frequency with which graduates report using skills learned during college may interest engineering educators, economists, and students.⁴⁴ However, like documentation, it is unclear whether managers and organizations recognize and reward skill application,⁴⁵ perhaps because it is an assumed condition of hire and job placement.⁴⁶

Career outcomes

We address RQ2 by examining four measures of respondents’ expected career outcomes. Presented after the assignment questions on EMS 3.0, these survey questions were not linked to specific work experiences in their wording. The question prompt read, ‘Looking ahead, how likely is it that you will do each of the following within one year?’ and items were measured on a five-point scale (0 = Definitely Not, 1 = Probably Not, 2 = Maybe, 3 = Probably Yes, 4 = Definitely Yes):

• Get a salary raise

• Receive a promotion

• Leave your organization (Intentions to Stay, reverse-coded)

• Switch career fields (Career Commitment, reverse-coded)

The first two items, expected salary raise and promotion, are possible outcomes we might expect to follow a stretch assignment but are not formally promised. They are the end goals of a ‘stretch assignment well done’, proving one’s readiness for the next career milestone.⁴⁷ Because decisions about raises and promotions rest with managers, we understand survey responses to these measures as at least partially rooted in respondents’ perceptions of how others in their organization perceive and reward them. We might anticipate that status characteristics will strongly come to the fore here, as respondents must self-assess their ability and connect these assessments to performance expectations and anticipated rewards in a workforce differentiated and typed by gender and race/ethnicity.

The second two items, intentions to stay and career commitment, are less strongly tied to stretch assignments. They are one step removed from evaluation and reward systems (one does not ‘get’ intentions or commitment as one ‘gets’ raises or promotions) and do not necessarily represent the type of advancement foreseen by a stretch. We include these measures as possible counterfactual devices, helping to show (or not) that stretch assignments are a distinctive, conditional lever of advancement.⁴⁸

Gender and race/ethnicity

We created our gender and race/ethnicity groups based on respondents’ self-reported identification at EMS 1.0 (gender and race/ethnicity were not asked in subsequent surveys). Our gender groups (and coding for analysis) are ‘women’ (=0) and ‘men’ (=1). These groups include individuals specifying ‘trans man/male’ or ‘trans woman/female’ in an open-ended option on the survey. The seven EMS 1.0 respondents who wrote ‘non-binary’ or ‘trans’ without specifying man/woman did not respond to EMS 3.0 and are (therefore) not represented in the longitudinal dataset. Our race/ethnicity groups compare respondents reporting ‘Black or African American’, ‘Hispanic or Latino/a’, ‘Native Hawaiian or Pacific Islander’, and/or ‘American Indian or Alaska Native’ backgrounds (shortened to Black, Latinx, Indigenous, and Pacific Islander) (=1) and respondents reporting ‘White’ and/or ‘Asian or Asian American’ backgrounds only (shortened to Asian and White) (=0).

Our groupings constitute broad markers of gender, race, and ethnicity with cultural and structural significance in engineering settings; they do not speak to how and why different racial/ethnic and gender sub-groups (e.g. Asian women, Black men) experience labor markets and workplaces differently. The EMS design as a whole also falls short in representing early-career engineers who do not identify with binary gender categories or the categories enumerated for race and ethnicity. For instance, the 2015 survey measurement for gender (three categories: ‘female’, ‘male’, and write-in for ‘other’) does not match what we now know to be more robust ways of asking about gender identity and could have signaled to survey-takers that our research was not inclusive.⁴⁹ Even if a single measure was more robust, we would need a follow-up question to understand others’ perceptions of one’s gender identity and the mis-gendering that results from the persistence of the socially constructed binary. Our groupings are starting rather than ending points for analysis of gendered and racialized patterns in assignments.

Controls: College and work measures

Eight control measures adjust for other potentially salient factors in assignment access (RQ1) and expected outcomes (RQ2). Accounting for these measures, we can identify relationships between our focal variables (gender, race/ethnicity, assignments) among statistically similar respondents in terms of college, labor market, and work experiences. However, many of these measures are themselves gendered and racialized, e.g., engineering task self-efficacy,⁵⁰ meaning our estimates of gender and race/ethnicity group effects are conservative.

Engineers’ experiences in the labor market and new jobs may vary by employer demands for particular skills associated with a sub-field. We group bachelor’s degree field as of EMS 3.0 into three categories: ‘electrical/mechanical engineering’ (an indicator of older fields with steady employer demand), ‘computer science (CS) in engineering’ (an indicator of newer fields with high demand), and ‘all other engineering fields’.⁵¹ We operationalize the first two categories as binary variables (=1) with ‘all other engineering fields’ as our reference group (=0). Two additional college measures, collected at EMS 1.0, are ‘Career Conversations’ (calculated as the mean response to four survey items asking respondents how often they talked with professors and students about professional and business options, each measured on a 5-point scale, 0 = Never to 4 = Very Often, α = .74) and cumulative grade point average (GPA) (measured on a 6-point scale, from 0 = C + (2.2–2.4) to 5 = A/A + (3.9+)). Career Conversations is a potentially important source of career capital as graduates enter the workforce. GPA is a performance measure that stands in for competencies across an engineering curriculum and may serve as a proxy for employers when assessing the merit of recent engineering graduates.⁵²

Two of four EMS 3.0 job measures ask respondents whether they had previously interned at their current company (0 = No, 1 = Yes) and would describe their job as an ‘engineering job’ (=1) or a ‘non-engineering job’ (=0). A third measure codes for working in what we designate as ‘core’ engineering business units (R&D, Design, and/or Production/Manufacturing) to the exclusion of other business units (=1) versus non-exclusion or in other units (=0). A fourth measure is Engineering Task Self-Efficacy (ETSE), respondents’ self-rated confidence in their ability to conduct engineering tasks (calculated as the mean response to five self-efficacy task items, each measured on a 5-point scale, 0 = Not Confident to 4 = Extremely Confident, α = .83). All of these measures could pattern access to stretch assignments (e.g. how managers evaluate potential candidates for stretches) and respondents’ expected career outcomes.

Online Supplemental Material (OSM) Appendix A provides additional information on EMS items.

Statistical methods

Following descriptive analysis of our focal assignment variables using parametric techniques (independent t-tests, one-way ANOVAs, and Pearson’s correlations), we construct two sets of ordinary least squares (OLS) regression models: those predicting frequency of assignments 'since starting one's job' (RQ1 access models) and those predicting expected career outcomes ‘in the next year’ as a function of assignments (RQ2 outcomes models).

For RQ1, each of four assignment measures is tested as a function of: (Model 1) gender group and race/ethnicity group; (Model 2) gender group, race/ethnicity group, and eight control variables; and (Model 3) Model 2 with a two-way interaction term for gender and race/ethnicity groups. For RQ2, each expected career outcome measure is regressed on each assignment measure in a four-model sequence, for a total of sixteen models for each outcome: (Model 1) assignment measure, gender group, race/ethnicity group, and eight control variables (identical to those tested in RQ1); (Model 2) Model 1 with a two-way interaction term for gender group and assignment measure; (Model 3) Model 1 with a two-way interaction term for race/ethnicity group and assignment measure; and (Model 4) Model 1 with a three-way interaction term for gender group, race/ethnicity group, and assignment measure (controlling for all two-way terms).

Our descriptive and statistical analyses are conducted on complete-case samples where only valid responses to every measure are included. On any given assignment measure, 2–7 respondents have missing data (Table 2, ‘Total’ row). With controls added to RQ1 models, we lose ∼3–4% of total cases using listwise deletion (Table 3). RQ2 models draw from a smaller starting sample than do RQ1 models because RQ2 models are restricted to the valid respondent sample for all four career outcome questions, asked after assignment questions and subject to slightly higher survey attrition. For RQ2 models, therefore, listwise deletion results in a ∼6–7% case loss (Table 4). We do not impute missing values in RQ1 and RQ2 models given respondents’ voluntary nonresponse to questions and negligible gains in statistical power with imputation. Examining group means on all variables in RQ1 to RQ2 samples suggests that narrowing in RQ2 does not introduce substantive sample bias; the proportions of Asian and White and Black, Latinx, Indigenous, and Pacific Islander women and men are nearly identical in the RQ1 and RQ2 samples (refer to Table 1 and OSM Appendix A).

Table 2. Assignments by gender and race/ethnicity groups.

	Stretch assignment – visibility			Stretch assignment – new area^1			Comparison assignment – documentation^2			Comparison assignment – skill application
	n	M	SD	n	M	SD	n	M	SD	n	M	SD
Asian/White women	187	2.84	1.02	188	3.26	0.82	187	3.14	0.97	188	2.54	1.02
Asian/White men	359	2.83	1.04	362	2.96	0.92	360	3.11	1.05	362	2.48	1.08
Black/Latinx/ Indigenous/ Pacific Islander women	27	3.00	0.96	27	3.22	0.93	27	3.63	0.69	27	2.30	1.20
Black/Latinx/ Indigenous/ Pacific Islander men	45	2.71	0.94	46	3.00	1.01	45	3.33	0.90	46	2.74	0.93
Total	618	2.83	1.02	623	3.06	0.91	619	3.16	1.01	623	2.51	1.06

¹ F(3,619) = 4.830, p = .002; post hoc comparisons using Bonferroni test: Asian/White women > Asian/White men p = .002.

² Welch Robust Test (3,92.45) = 4.756, p = .004; post hoc comparisons using Games-Howell test: Black/Latinx/Indigenous/Pacific Islander women > Asian/White women p = .012, > Asian/White men p = .005.

Table 3. RQ1 access models: fully adjusted models (B coefficients shown with standard errors in parentheses).

	Assignment (dependent variable)
Independent variables	Visibility	New area	Documentation	Skill application
Gender group: Men	−.13 (.09)	−.31 (.08)***	−.13 (.09)	−.15 (.09)
Race/ethnicity group: Black/Latinx/Indigenous/ Pacific Islander	−.01 (.13)	−.03 (.12)	.21 (.11)	.16 (.13)
Elec/Mech Engr Degree	−.08 (.09)	.02 (.08)	−.03 (.09)	.12 (.09)
CS (in Engr) Degree	−.04 (.14)	.17 (.11)	−.49 (.15)**	.42 (.14)**
Sch1.0: Career conversations scale	.21 (.06)***	.15 (.05)**	.05 (.05)	.13 (.05)*
Sch1.0: College GPA	.01 (.03)	−.03 (.03)	−.12 (.03)***	.08 (.03)*
Wk3.0: Interned at current company	.20 (.08)*	.05 (.07)	.00 (.08)	−.02 (.09)
Wk3.0: Position: Engineering	.23 (.14)	.16 (.12)	.58 (.14)***	.50 (.13)***
Wk3.0: Functional Area: ‘Core engineering’ (R&D/Des/Prod/Mfg)	−.21 (.08)*	−.15 (.08)*	−.22 (.09)*	.14 (.09)
Wk3.0: Engr Task Self-Efficacy	.18 (.07)**	−.01 (.06)	.06 (.07)	.25 (.06)***
Constant	1.87 (.23)***	3.01 (.20)***	3.03 (.22)***	.89 (.21)***
R-square	.09	.06	.10	.13
Listwise n	602	607	604	607

***p < .001, **p < .01, *p < .05.

Variables lists coding schemes/reference groups. ‘Sch1.0’: EMS 1.0 college measure. ‘Wk3.0’: EMS 3.0 Job/workforce measure. All interactions between gender group and race/ethnicity group were non-significant. OSM Appendix B includes RQ1 full model sequence.

Table 4. RQ2 outcomes models: focal main effects and interaction term results, fully adjusted models (B coefficients shown with standard errors in parentheses; significant interaction terms in bold).

	Expected career outcome (dependent variable)
With controls …	Expected salary raise	Expected promotion	Intentions to stay	Career commitment
Visibility assignment sequence
Visibility	0.07 (.04)	0.10 (.05)*	0.12 (.05)*	0.09 (.04)*
Gender group: Men	0.08 (.08)	−0.07 (.10)	0.06 (.10)	0.08 (.09)
Race/ethnicity group: Black/Latinx/Indigenous/Pacific Islander	0.04 (.11)	0.11 (.14)	0.06 (.15)	0.09 (.13)
Constant	2.30 (.23)***	1.43 (.26)***	1.20 (.24)***	1.24 (.23)***
R-square	0.04	0.04	0.13	0.20
Two-way interactions
Visibility	−0.07 (.06)	0.03 (.08)	0.02 (.08)	0.00 (.08)
Visibility*Gender group: Men	0.22 (.08)**	0.10 (.10)	0.15 (.10)	0.13 (.09)
Visibility	0.08 (.05)	0.11 (.05)*	0.08 (.05)	0.07 (.05)
Visibility*Race/ethnicity group: Black/Latinx/Indigenous/Pacific Islander	−0.07 (.14)	−0.12 (.15)	0.37 (.16)*	0.21 (.15)
Exploratory three-way interaction	ns	−0.65 (.26)*	ns	ns
New area assignment sequence
New area	0.09 (.05)	0.19 (.06)**	−0.01 (.05)	−0.01 (.05)
Gender group: Men	0.09 (.08)	−0.02 (.10)	0.03 (.10)	0.06 (.09)
Race/ethnicity group: Black/Latinx/Indigenous/Pacific Islander	0.03 (.11)	0.12 (.14)	0.09 (.15)	0.08 (.13)
Constant	2.16 (.27)***	1.02 (.30)**	1.47 (.28)***	1.48 (.27)***
R-square	0.05	0.06	0.11	0.19
Two-way interactions
New area	−0.05 (.07)	0.07 (.10)	−0.14 (.09)	−0.20 (.08)*
New area*Gender group: Men	0.20 (.09)*	0.18 (.12)	0.19 (.11)	0.27 (.10)**
New area	0.13 (.05)*	0.26 (.06)***	−0.04 (.05)	−0.02 (.05)
New area*Race/ethnicity group: Black/Latinx/Indigenous/Pacific Islander	−0.26 (.11)*	−0.47 (.14)**	0.29 (.19)	0.08 (.15)
Exploratory three-way interaction	ns	ns	ns	ns
Documentation assignment sequence
Documentation	0.11 (.05)*	0.15 (.05)**	0.14 (.05)**	0.10 (.05)*
Gender group: Men	0.09 (.08)	−0.06 (.10)	0.05 (.10)	0.07 (.09)
Race/ethnicity group: Black/Latinx/Indigenous/Pacific Islander	0.02 (.12)	0.09 (.14)	0.04 (.15)	0.07 (.13)
Constant	2.11 (.27)***	1.14 (.29)***	1.05 (.27)***	1.14 (.26)***
R-square	0.05	0.05	0.13	0.19
Two-way interactions
Documentation	−0.01 (.08)	0.06 (.08)	0.08 (.08)	0.07 (.08)
Documentation*Gender group: Men	0.17 (.09)	0.13 (.10)	0.08 (.09)	0.04 (.09)
Documentation	0.13 (.05)**	0.18 (.05)**	0.13 (.05)**	0.07 (.05)
Documentation*Race/ethnicity group: Black/Latinx/Indigenous/Pacific Islander	−0.26 (.11)*	−0.32 (.18)	0.12 (.20)	0.29 (.19)
Exploratory three-way interaction	ns	ns	ns	ns
Skill application assignment sequence
Application	0.08 (.04)*	0.14 (.05)**	0.08 (.05)	0.13 (.04)**
Gender group: Men	0.08 (.08)	−0.06 (.10)	0.05 (.10)	0.07 (.09)
Race/ethnicity group: Black/Latinx/Indigenous/Pacific Islander	0.01 (.11)	0.09 (.14)	0.08 (.15)	0.06 (.13)
Constant	2.36 (.22)***	1.48 (.25)***	1.37 (.23)***	1.32 (.22)***
R-square	0.05	0.05	0.12	0.20
Two-way interactions
Application	0.04 (.06)	0.16 (.08)*	0.05 (.08)	0.14 (.07)
Application*Gender group: Men	0.07 (.08)	−0.04 (.09)	0.05 (.09)	−0.02 (.09)
Application	0.10 (.04)*	0.16 (.05)**	0.09 (.05)	0.14 (.05)**
Application*Race/ethnicity group: Black/Latinx/Indigenous/Pacific Islander	−0.13 (.11)	−0.18 (.14)	−0.05 (.17)	−0.08 (.14)
Exploratory three-way interaction	ns	ns	0.72 (.29)*	0.70 (.25)**

***p < .001, **p < .01, *p < .05, ns = not significant.

Visibility-Outcomes listwise–n = 584; New Area-Outcomes listwise-n = 588; Documentation-Outcomes listwise-n = 587; Application-Outcomes listwise-n = 588.

Statistical methods describes RQ2 model sequence. OSM Appendix C lists control coefficients. Only significant coefficients presented for exploratory three-way interactions. Marginal effects for significant interactions reported in text. Significant interaction terms generally increase R-squares by one point (e.g., .05 to .06).

Using Stata 15.1, we specify models with robust standard errors clustered by person because the EMS data are longitudinal. We also compute marginal effects (net controls) for all significant interaction terms (p < .05) using the Stata command margins to clarify findings of our interaction term analyses.

Our main effects for gender group, race/ethnicity group, and assignments, and two-way interaction terms between these measures, directly address our research questions. We include three-way interaction terms in our RQ2 models as exploratory measures meriting additional research. We are exploratory because of group sizes (especially for Black, Latinx, Indigenous, and Pacific Islander women) and conventions of statistical work. A three-way interaction term in our model, tested with all controls, creates extremely small statistically equivalent comparison groups and potentially large standard errors. Yet, while we are cautious in interpretation of our three-way terms, we do not want to exclude them from analysis. We reason that most statistical conventions leave understudied and marginalized populations exactly that: understudied and marginalized.⁵³

Results

Rq1: access

In aggregate, respondents were most frequently involved in documentation, new area, visibility, and skill application assignments in that order (Table 2). Women engineers (M = 3.25, SD = .83) were more likely than men engineers (M = 2.96, SD = .93) to report new area assignments (t(480) = 3.933, p < .001, equal variances not assumed). Black, Latinx, Indigenous, and Pacific Islander engineers (M = 3.44, SD = .84) were more likely than Asian and White engineers (M = 3.12, SD = 1.02) to report documentation assignments (t(101) = −2.967, p = .004, equal variances not assumed). The mean gender difference in new area assignments was larger among Asian and White engineers than among Black, Latinx, Indigenous, and Pacific Islander engineers, and Black, Latinx, Indigenous, and Pacific Islander women reported, on average, the highest frequency of documentation assignments, significantly higher than the averages among Asian and White respondents (Table 2). Between-group differences in self-reported rates of skill application and visibility assignments were non-significant.

Simple correlations between the four assignment measures were modest. The two components of a stretch assignment, visibility and new area, were moderately correlated for all gender and race/ethnicity groups (r = ∼.2–.3) (OSM Appendix B).

Table 3 summarizes the results of the multivariate models in which we regressed each assignment measure on gender group, race/ethnicity group, and eight controls (corresponding to RQ1 Model 2; all interactions between gender group and race/ethnicity group, corresponding to RQ1 Model 3, were non-significant). Consistent with descriptive findings, multivariate analysis showed that frequency of visibility and skill application assignments did not significantly vary by gender or race/ethnicity group. Women engineers reported higher rates of new area assignments than men engineers even among those comparable in field of degree, college grades, current jobs, and engineering task confidence. The descriptive difference in documentation assignments between Black, Latinx, Indigenous, and Pacific Islander engineers and Asian and White engineers was statistically non-significant in our fully adjusted regression models. Across RQ1 models, college GPA was associated with both comparison assignments but neither measure of stretch assignments.

RQ2: expected outcomes

RQ2 models examined the net predictive power of our focal assignments to identify how and for whom these assignments operate, using our ‘inequality-intensifying’, ‘equality-intensifying’, and ‘inequality-preserving’ interpretive lens. Descriptive calculations for the four outcome measures showed that respondents were, on average, uncertain to positive concerning their anticipated career outcomes in the next year (OSM Appendix A). Mean differences on these four items by gender and race/ethnicity groups were non-significant using one-way ANOVA tests. Overall, expected raises and promotions were moderately correlated (r = ∼.3–.5 by group). Intentions to stay and career commitment were strongly correlated (r = ∼.6–.8 by group) (OSM Appendix C).

Table 4 summarizes our regression sequence. We focus on main effects of each assignment, net controls (i.e. among early-career engineers with statistically equivalent degrees, grades, current positions, and engineering task confidence), statistically significant two-way interactions between assignments and gender group and assignments and race/ethnicity group, and implications of these effects for suggesting equality-intensifying or inequality-intensifying relationships (or, in absence of such effects, inequality-preserving relationships). There was no statistically significant main effect of gender or race/ethnicity group on any of the four expected career outcomes, holding all control variables constant. For all significant interactions, we report marginal effect coefficients and p-values in text, with standard errors in parentheses.

Visibility. In aggregate, visibility assignments were associated with intentions to stay at one’s organization and career commitment in the next year, with a borderline significant main effect on expected promotion (p = .048) and a non-significant main effect on expected raise (Table 4). However, interaction term and subsequent marginal effects calculations revealed a positive, significant relationship between visibility assignments and expected raise for men (B = .15 (.05), p < .01) but not women (B = −.07 (.06), ns), suggesting an inequality-intensifying effect of visibility work. Note the difference in marginal effect size (.15 vs. – .07), which supports the interpretation of a differential effect; if the coefficients were equally sized and in the same direction, but the larger group coefficient reached statistical significance and the smaller one did not, ‘difference’ clearly would be an artifact of limited statistical power for the smaller group. We extend this effect size logic to the interpretation of all subsequent marginal effects.

This sequence yielded our only instance of a possible equality-intensifying effect across all assignment models: visibility assignments linked to intentions to stay for Black, Latinx, Indigenous, and Pacific Islander engineers (B = .45 (.16), p < .01) and not Asian and White engineers (B = .08 (.05), ns). We observed one statistically significant, exploratory three-way interaction in the expected promotion model. Marginal effects suggest that Asian and White men might anticipate a promotion boost from visibility more than other groups; the coefficient for our small group of Black, Latinx, Indigenous, and Pacific Islander women was large but statistically non-significant (Asian and White men B = .17 (.06), p < .05, Black, Latinx, Indigenous, and Pacific Islander women B = .37 (.19), ns, Black, Latinx, Indigenous, and Pacific Islander men B = −.13 (.14), ns, Asian and White women B = .01 (.08), ns).

New area. New area assignments have the largest number of potential inequality-intensifying effects across all models. We observed one positive main effect of new area assignments on expected promotion net controls. However, interaction term and marginal effects calculations showed that this relationship holds for Asian and White engineers (B = .26 (.06), p < .001) but not Black, Latinx, Indigenous, and Pacific Islander engineers (B = −.22 (.13), ns). We also observed a positive relationship between new area work and expected raise for men engineers (B = .15 (.06), p < .05) but not women engineers (B = −.05 (.07), ns), and Asian and White engineers (B = .13 (.05), p < .05) but not Black, Latinx, Indigenous, and Pacific Islander engineers (B = −.14 (.09), ns). Finally, there was a negative relationship between new area assignments and career commitment for women (B = −.20 (.08), p < .05) but not men (B = .07 (.06), ns). The career commitment finding is not as clear-cut in terms of inequality-intensifying: new area work had no effect for men, and although we control for engineering (versus non-engineering) job, we did not test from which specific field women intend to switch.

Documentation. Documentation assignments were positively related to all four expected career outcomes. Effects were not conditional on race/ethnicity or gender group for the most part, suggesting inequality-preserving effects. However, one potentially inequality-intensifying relationship was observed: documentation and expected salary raise were associated for Asian and White engineers (B = .13 (.05), p < .01) but not Black, Latinx, Indigenous, and Pacific Islander engineers (B = −.13 (.10), ns).

Skill application. Skill application assignments linked with three of the four career outcomes: expected raise, expected promotion, and career commitment. There were no statistically significant two-way interactions, suggesting inequality-preserving patterns. There was one significant, exploratory three-way interaction with significant marginal effects in the career commitment model (marginal effects for the significant term in the intentions model were non-significant, refer to OSM Appendix C): application assignments positively linked to commitment for both Asian and White women (B = .22 (.08), p < .01) and Black, Latinx, Indigenous, and Pacific Islander men (B = .39 (.17), p < .05). The relationship was less consistent for Asian and White men (B = .10 (.05), p = .05) and especially Black, Latinx, Indigenous, and Pacific Islander women (B = −.20 (.16), ns).

Discussion

Our RQ1 results suggest that access to stretch assignments may not systematically favor early-career engineers from dominant engineering gender and race/ethnicity groups (men engineers and Asian and White engineers, respectively). Women, in fact, report being involved in more new area stretch assignments than men, even among those with comparable degrees, grades, confidence, and jobs. Access to other assignments that may carry lower organizational recognition⁵⁴ appears to work differently, with racial/ethnic minoritized engineers from Black, Latinx, Indigenous, and Pacific Islander backgrounds reporting higher-frequency documentation assignments than do those in the racial/ethnic dominant group. Differences between dominant and minoritized groups on our control measures appear to explain differences in documentation access, hinting that it is less unequal task allocation than other unequal school, labor market, and job experiences influencing why racial/ethnic minoritized engineers are even more involved in documentation. Access results raise the possibility that routine work is a convergence point of racial/ethnic inequality in engineering practice, dominant and minoritized groups have access to visibility stretches in comparable measure, and stretches involving new and unfamiliar areas are availed to women more than men.

However, a more complex picture emerges in RQ2 when we consider how stretch assignments link to respondents’ expectations of raises and promotions – the organizational traction thought to come from stretches. Whenever we observe stretch assignments acting conditionally on expected raises/promotions (i.e. only for some groups, not others), the relationships favor engineers in dominant groups.⁵⁵ This pattern is most salient in our analysis of new area assignments and expected salary raises: more frequent assignments involving new and unfamiliar areas are associated with expectations of a raise among men and Asian and White engineers, not equally confident, equally experienced women and Black, Latinx, Indigenous, and Pacific Islander engineers. In two model sequences, we observe relationships that are stronger for minoritized groups in engineering, but they are not about raises or promotions: visibility links to organizational loyalty for Black, Latinx, Indigenous, and Pacific Islander engineers, not Asian and White engineers, and new areas link to prospective changes in career fields for women, not men. The patterns suggest that stretch assignments, especially those in new and unfamiliar areas, may intensify inequality in the early career via expected outcomes and anticipated rewards – and status characteristics provide a frame for understanding why.

We conceived status characteristics as relevant to early-career stretch assignments for engineers based on previous research highlighting three central contexts to early-career engineering practice, how engineering managers define career-advancing stretch work, and the salience and function of status characteristics themselves in these types of conditions. Widely held cultural beliefs about gender, race, ethnicity, and technical competence would be expected to shape how engineers doing stretch work assess their performance and then link those assessments to how others evaluate and reward them. Our empirical results repeatedly reveal positive relationships between stretch assignments and expected career gains for gender- and racial/ethnic-dominant, not minoritized, engineering groups, despite statistical equivalence in academic performance, engineering confidence, and jobs. Academic performance, in fact, is not even predictive of access to stretch assignments (but it is predictive of access to comparison assignments, suggesting that merit narratives in engineering⁵⁶ are especially incomplete in describing access to highest-leverage, highest-profile tasks). Placing what we theoretically mapped out next to our empirical findings, we can make some assumption that status characteristics are driving increasing advantage for dominant groups in these stretch settings.

Yet it is not any stretch assignment. New area assignments yield even more effects that could be considered inequality-intensifying than do visibility assignments, suggesting that novel and unfamiliar work particularly heightens status salience in already ambiguous early-career engineering contexts. This finding is a key extension of prior research on and our conceptual linking of engineering practice, stretch assignments, and status characteristics – and it is not entirely unsurprising given how essential ambiguity is to the foregrounding of status.⁵⁷ Higher-status groups may be given the benefit of the doubt in new area conditions and/or evaluate their own performance and prospects more favorably; they also may be given better new area assignments, with more support in executing the assignment, to begin with. Lower-status groups may face extra scrutiny from others and/or evaluate their own performance and prospects less favorably in these same conditions.⁵⁸ Doing novel work has deep learning promise. How novice standing is read in an organization demands careful study.

Access to new and unfamiliar stretches may need reconsideration in this light. Our finding that women in our sample have more access to new areas could be construed as a positive signal – women are connecting to career-advancing opportunities that could challenge other systematic barriers to equality. But the salience of status characteristics in our expected outcomes models raises a different possibility, one synergistic with previous studies showing that women’s leadership-building assignments are different and/or fewer than men’s⁵⁹: women have more access because managers evaluate them as ‘short’ on competence (regardless of grades, internships, etc.) and move them into unfamiliar areas as a way to get experience – only to face doubts about their novice standing that men, as a higher-status group, do not face. As research on gender and teamwork reveals, nor is women’s expertise readily taken up in such status systems,⁶⁰ a double bind that could help explain women’s reconsideration of their career field the more they are placed on assignments in unfamiliar domains.

Other positive signals, such as the possibility that visibility stretch assignments intensify racial/ethnic equality, may need a similarly critical ‘double take’. Doing work that generates interest could mean working on a ‘hot project’, which might be especially significant for otherwise less visible racial/ethnic minoritized groups in their calculations of whether to stay at their organizations. Such an experience could help to counter long-term status advantage. However, intentions to stay are not the same as expected rewards. Organizational efforts to retain people without parallel emphasis on advancement and attention to the role of cultural stereotypes yield little net progress.

The patterns observed for our two comparison assignments help us to rule out the possibility that there is nothing particularly ‘unique’ about early-career stretch assignments, i.e. that we are describing all forms of early-career engineering work. Documentation and skill application assignments, conceived as lower-profile, lower-recognition work familiar to recent graduates, are positively associated with nearly all four expected career outcomes for nearly all gender and race/ethnicity groups. For early-career engineers, applying skills and documenting work could signal ‘jobs solidly done’ for which scope and outcomes are clearer, i.e. the clarity of the task, not the task itself, explains why these measures have fewer inequality-intensifying effects than stretch measures. Such clarity could be brought into stretch work to minimize status-affirming biases associated with ‘getting out of your comfort zone’,⁶¹ by pre-establishing criteria for scope, performance, and rewards, communicating these criteria to all early-career hires, and creating accountability around implementation of criteria.

Even these assignments demand thoughtfulness, however. For instance, higher-frequency documentation is associated with expected raises among engineers in the racial/ethnic dominant group, not those in the racial/ethnic minoritized group. Plus, we argue that a pattern of mostly unconditional positive effects preserves existing inequalities, leaving existing group advantages and disadvantages intact. ‘Better’ stretch assignments would not necessarily entail more skill application and documentation and less novelty and visibility, or vice versa. Instead, organizations and researchers would investigate biases in all types of early-career work that maximize gains for already high-status groups (even by leaving status quo inequality unchallenged).

Limitations

Experimental and qualitative research would help identify causal mechanisms and/or meanings our measures miss. For example, we use longitudinal survey data to control for self-selection into jobs and assignments within those jobs, which we conceive as influencing expected career outcomes. Yet we still have two main sets of self-reported measures collected simultaneously: respondents’ reports of their assignments in the past and respondents’ expected outcomes in the upcoming year. Managers might be guiding some reports towards promotion or raises and ensuring they have the needed assignments to do so.

We do not know if expected outcomes actually happened and how respondents’ performance was evaluated – specifically, whether recent engineering hires doing the same types of stretch assignments in the same role get differential boosts or penalties to their performance scores based on status. Correll et al. show how performance evaluations in companies differentially reward men’s and women’s identical behavior.⁶²

Finally, because our measures of race, ethnicity, and gender are limited given study design, measurement, and aggregation, we understate gender and racial/ethnic inequality in early-career assignments. We do not compare assignment outcomes of racial/ethnic sub-groups of women, probe why Asian men’s pathways diverge from White men’s pathways up the job ladder, or trace misgendering of non-binary engineers in stretch assignment conditions in this study. Additionally, our three-way interaction term findings, reported as exploratory and not treated as the primary evidence base for our research questions, open more questions than answers, e.g., if we had a larger sample of racial/ethnic minoritized women in our study, would visibility assignments significantly link to anticipated promotion for both Asian and White men and Black, Latinx, Indigenous, and Pacific Islander women – offering a possible ‘challenge case’ to status systems in early-career engineering? And why is the relationship between skill application and career commitment weaker for these two groups – what might set them apart on a few of our measures? We encourage future work that uncovers dynamic inequalities, and challenges to inequalities, in ways our study does not.

Conclusion

This study considers access to and outcomes of early-career stretch assignments, an informal advancement mechanism in organizations that could help explain why engineering leadership ranks are even more gender- and race-unequal than are engineering contributor ranks. Expanding what is known about the early-career assignment landscape, we find that stretch assignments, more than other types of early-career assignments, may be a decisive inflection point for inequality in the engineering workforce.

Our results show that stretch assignments do not track with measures of merit and instead suggest inequality-intensification patterns that align with status characteristics. In an equity-minded workspace, all graduates would participate in novel work that people know and care about, find connection with their undergraduate learning, and engage in high-value documentation – in clear, coordinated evaluation and reward systems. We look to both managers and educators to support this emerging matrix of early-career practice and consider how new hires can bring their learning to practice and advance amidst the regressive pulls of status and stereotypes. We call for greater focus on how organizations can disrupt status effects to build a stronger leadership bench.

Acknowledgements

We thank our Engineering Majors Survey participants for their time and engagement with our research project. We are grateful for thoughtful comments and helpful feedback on this manuscript from Shelley Correll, two anonymous reviewers, and the editorial team at Engineering Studies. We thank the National Science Foundation (grant numbers DUE-1125457 and 1636442) for their support of this work. The opinions expressed are those of the authors alone and do not represent views of the National Science Foundation.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by National Science Foundation [Grant Award 1636442 and DUE-1125457].

Notes

1 BLS, “Table 11. Employed Persons by Detailed Occupation”; Tomaskovic-Devey and Han, Is Silicon Valley Tech Diversity Possible Now?

2 Correll, “SWS 2016 Feminist Lecture”; DiTomaso et al., “Effects of Structural Position”; Faulkner, “Doing Gender in Engineering Workplace Cultures I”; Rosette, Leonardelli, and Phillips, “The White Standard.”

3 Corbett and Wullert et al., “Reproducing Inequality in Organizations.” Historically, these four racial/ethnic groups – Black, Latinx, Indigenous, and Pacific Islander – have been underrepresented and minoritized in U.S. engineering fields. White and Asian groups have been overrepresented and compose, in aggregate, a dominant engineering group. While we employ these aggregations in our study, we hope to show nuance and critical treatment of our decision as our argument unfolds. Also, throughout our paper, we sometimes combine ‘race’ and ‘ethnicity’ in a single phrase ‘race/ethnicity’. This decision should not imply interchangeability between the terms, only connectivity in the social phenomena under study.

4 Beddoes, “Gender as Structure”.

5 Bucciarelli, Designing Engineers; DiTomaso et al., “Effects of Structural Position”; Sheppard et al., Educating Engineers.

6 Korte, Brunhaver, and Sheppard, “(Mis)interpretations of Organizational Socialization”, 192.

7 Cardador, “Promoted Up But Also Out?”; Faulkner, “‘Nuts and Bolts and People’”; Trevelyan, “Reconstructing Engineering From Practice.”

8 Brunhaver et al., “Bridging the Gaps”; Cech, “The (Mis)framing of Social Justice”; Trevelyan, “Reconstructing Engineering From Practice”.

9 Castilla, “Gender, Race, and Meritocracy”; Cech, “The (Mis)framing of Social Justice”; Seron et al., “I am Not a Feminist, But … ”.

10 DiTomaso et al., “Effects of Structural Position.”

11 Corbett and Wullert et al., “Reproducing Inequality in Organizations”; consult also Alegria, “Escalator or Step Stool?”

12 Lutz and Paretti, “Exploring the Social and Cultural Dimensions.”

13 Brunhaver et al., “Bridging the Gaps,” 142.

14 Brunhaver et al., “Supports and Barriers that Recent Engineering Graduates Experience.”

15 Lutz, Canney, and Brunhaver, “‘I Wish I Could Do More’.”

16 E.g. Beddoes, “Examining Privilege in Engineering Socialization.”

17 Ibarra, “To Close the Gender Gap”; McCauley, Eastman, and Ohlott, “Linking Management Selection and Development.”

18 Tobias Neely et al., “The Lifecycle of a Stretch Assignment,” 6–7.

19 De Pater et al., “Gender Differences in Job Challenge”; McCauley, Ohlott, and Ruderman, Job Challenge Profile; Ohlott, Ruderman, and McCauley, “Gender Differences in Managers’ Developmental Job Experiences.”

20 Tobias Neely et al., “The Lifecycle of a Stretch Assignment,” 21.

21 De Pater et al., “Gender Differences in Job Challenge”; Hoobler, Lemmon, and Wayne, “Women’s Managerial Aspirations”; Lyness and Thompson, “Climbing the Corporate Ladder”; Ohlott, Ruderman, and McCauley, “Gender Differences in Managers’ Developmental Job Experiences”; Taylor and Ilgen, “Sex Discrimination Against Women.” Synergistically, Beddoes discusses how early-career engineering women observe their men colleagues being prepared for leadership positions in ways they are not, ‘Gender as Structure’.

22 Abad, “Race, Knowledge, and Tasks,” 112, 116.

23 Williams and Multhaup, “For Women and Minorities to Get Ahead.”

24 Beddoes, “Examining Privilege in Engineering Socialization”; Melaku, You Don’t Look Like a Lawyer.

25 Berger et al., Status Characteristics and Social Interaction; Ridgeway, Framed by Gender; Ridgeway and Correll, “Unpacking the Gender System.”

26 Ridgeway and Correll, “Unpacking the Gender System.”

27 Correll, “Gender and the Career Choice Process,” “Constraints into Preferences.”

28 Correll, “Gender and the Career Choice Process”; Correll, Benard, and Paik, “Getting a Job”; Ridgeway and Correll, “Unpacking the Gender System.”

29 Correll, “SWS 2016 Feminist Lecture”; Correll et al., “It’s the Conventional thought that Counts,” 298–99.

30 This is not to say that gender is a binary. Rather, research shows that people treat gender as a binary in social interactions, making binary classifications (man or woman) in milliseconds. This gender system results in gender non-binary people being subject to misgendering, marginalization, and social penalty. Ridgeway, Framed by Gender.

31 Cech, “Ideological Wage Inequalities?”; Faulkner, “‘Nuts and Bolts and People’”.

32 Corbett and Hill, Solving the Equation; Tomaskovic-Devey and Han, Is Silicon Valley Tech Diversity Possible Now?

33 Shih, Pittinsky, and Ambady, “Stereotype Susceptibility”; Sy et al., “Leadership Perceptions.”

34 Alegria, “Escalator or Step Stool?”

35 Refer to Chavez and Wingfield, “Racializing Gendered Interactions,” 9–10; Purdie-Vaughns and Eibach, “Intersectional Invisibility.”

36 We do not argue that Asian and White men share engineering power equally. Socio-cultural barriers to Asian (versus White) men emerge higher up the ladder, e.g. Rosette, Leonardelli, and Phillips, “The White Standard”; Sy et al., “Leadership Perceptions.”

37 Consult Chase, Dowd, Pazich, and Bensimon, “Transfer Equity.”

38 Gilmartin et al., Designing a Longitudinal Study, and Thompson, Xiao, and Gilmartin, EMS 3.0 Addendum, detail EMS recruitment, samples, and nonresponse bias.

39 We included respondents enrolled in graduate degree programs if they marked concurrent employment. With the gender/race/ethnicity criterion, we lost 20 EMS 3.0 respondents who were employed non-undergraduates but had not voluntarily reported gender and race/ethnicity in EMS 1.0. As of EMS 3.0, 82% of our sample reported being in their first jobs since earning their first engineering bachelor’s, 42% were 2–3 years out from their degrees (versus <1–1 years), and 2% could be classified as non-traditional early-career engineers, having earned other degrees in the years prior to the study. All RQ1 and RQ2 model results were robust to these three measures.

40 Tobias Neely et al., “The Lifecycle of a Stretch Assignment.”

41 These and all other survey items were piloted with multiple engineering student and alumni groups prior to EMS 3.0 administration. Wording was continuously revised for clarity and comprehensibility to our target respondents.

42 Trevelyan, “Observations of South Asian Engineering Practice”; Trevelyan and Tilli, “Longitudinal Study of Australian Engineering Graduates.”

43 Trevelyan, “Observations of South Asian Engineering Practice,” 233; Trevelyan and Williams, “Value Creation in the Engineering Enterprise.”

44 E.g. Passow and Passow, “What Competencies.”

45 Consult Wehner and Wynn, “Can Managers Clarify Criteria?”

46 Rivera, “Employer Decision Making.”

47 Previous research focuses on promotions or promotability from stretches, more so than raises, e.g. De Pater et al., “Employees’ Challenging Job Experiences”, 297–98. We treat both as advancement markers, recognizing monetary bonuses for stretch work in Tobias Neely et al., “The Lifecycle of a Stretch Assignment.”

48 Noting that intent to stay may not be entirely unrelated to stretch assignments if used as retention tools. Tobias Neely et al., “The Lifecycle of a Stretch Assignment.”

49 Magliozzi, Saperstein, and Westbrook, “Scaling Up.”

50 Sterling et al., “The Confidence Gap.”

51 Consistent with aggregation rationale in National Science Foundation, ‘Women, Minorities, and Persons with Disabilities’ reporting, we combine computer and electrical engineering degrees under a single electrical category. We then group mechanical and electrical engineering as a general indicator of older, large engineering fields that may condition labor market and job experiences (e.g. assignment types, promotion opportunities), comparing to newer, high-demand CS fields and all other engineering fields. For CS discussion, consult Fayer, Lacey, and Watson, STEM Occupations.

52 Bills, “Credentials, Signals, and Screens”; Rivera, “Employer Decision Making.”

53 Also consult Slaton and Pawley, “The Power and Politics.”

54 Trevelyan and Williams, “Value Creation in the Engineering Enterprise.”

55 We observe stretches acting unconditionally, too, e.g. visibility is unrelated to expected raises for both race/ethnicity groups. But these (non)relationships are what we categorize as inequality-preserving—they do not change what we would expect in status quo unequal workforce circumstances. Here, we focus on types and implications of inequality-(or equality-)intensifying patterns in our stretch measures.

56 Narratives, and underlying ideologies, described by Cech, “The (Mis)framing of Social Justice” among others.

57 Correll et al., “It’s the Conventional Thought that Counts.”

58 Correll, “SWS 2016 Feminist Lecture.”

59 E.g. Beddoes, “Gender as Structure.”

60 Thomas-Hunt and Phillips, “When What You Know is Not Enough.”

61 Tobias Neely et al., “The Lifecycle of a Stretch Assignment.”

62 Correll et al., “Inside the Black Box.”