Pursuing an Engineering Major: Social Capital of Women and Underrepresented Minorities

As part of a longitudinal research effort that examines the influence of social capital on differential persistence and retention among undergraduate engineering majors, this study examines how engineering degree-related social capital differs for first-year engineering students by gender and ethno-racial groups. Social capital is operationalized as a person's network of relationships with individuals who hold influential positions (e.g. parent, teacher, advisers) and access to resources that support persistence in engineering programs. Our data comprise survey responses from 2186 first-year engineering students, from eleven diverse colleges and universities, who provided information about their participation in engineering-related courses, activities, and programs while in high school as well as the individuals who influenced their decision to pursue an engineering major. We found few differences in social capital between men and women, but found many differences among ethno-racial groups, which suggests that different levels of social capital could influence students' persistence and retention.

Keywords: Social capital; persistence; retention; ego-networks; minority/majority groups

In the United States, women and minorities continue to be underrepresented in science, technology, engineering, and mathematics (STEM) fields compared to the population as a whole. The National Science Foundation (NSF [18]) categorizes Blacks, Hispanics, and American Indians and/or Alaskan Natives as underrepresented minorities (URM) in STEM. In 2017, women earned 57% (n = 1119,987) of all undergraduate degrees (n = 1956,032). However, of all undergraduate degrees (n = 115,640) earned in engineering, engineering technologies, and engineering-related fields, women only earned 22% (n = 24,904), followed by Hispanics who earned 10% (n = 11,871), Blacks who earned four percent (n = 4505) and American Indian/Alaska Natives who earned 0.3 percent (n = 301) (Snyder, de Brey, and Dillow [28]). The dearth of women and URM in engineering stymies technological innovation and advancement by limiting diverse perspectives. Understanding the factors that contribute to the persistence of women and URM can inform attempts to increase their participation in STEM fields.

Numerous studies suggest that while academic preparation and socio-economic status are contributing factors, the low numbers of women and URM in STEM are likely due to implicit factors unrelated to aptitude and interest (Margolis, Fisher, and Miller [16]; Hill, Corbett, and St. Rose [11]; Sax [23]; Shapiro and Sax [25]; Summers and Hrabowski III [29]). For instance, Seymour and Hewitt ([24]) found that switchers and non-switchers were not distinguishable in their high school preparation, effort expended, or performance scores. Ultimately, it was the problems students encountered related to the STEM degree program's structure and culture that often had the most influence on their persistence.

Previous research suggests that students' entering levels of social capital may influence academic outcomes. For instance, URM students might not know that academic preparation for the sciences and engineering may need to start by middle school (Hilton et al. [12]; Clewell, Anderson, and Thorpe [5]). Further, access to STEM career information (Clewell, Anderson, and Thorpe [5]) influenced preparation (Seymour and Hewitt [24]). In college, students of color often lack faculty role models and mentors, do not have peers of color (Brown [3]), and feel isolated along with a lack of belonging and thus have different educational experiences than White male peers (Litzler and Samuelson [15]; Marra et al. [17]; Seymour and Hewitt [24]). Marra et al. ([17]) found that non-White students had similar reasons as women and other groups for deciding to switch out of engineering, which includes poor teaching and advising, curriculum difficulty, and lack of sense of belonging. However, non-White students more often cited curriculum difficulty and lack of belonging as being influential in their switching than other groups. The persistence of URM was influenced by their ability to acquire certain attitudes and strategies and whether they found personal and academic support from influencers to keep them motivated (Seymour and Hewitt [24]).

Tyson, Smith, and Ndong ([31]) found that students who switched out of engineering, administrators, and staff differed on their views about influences on degree attainment. Faculty and staff believed that academic preparation in high school was crucial to academic attainment. However, some students who reported appropriate academic preparation switched majors because they lacked information about what to expect in their engineering program. Similar to Seymour and Hewitt ([24]), Tyson, Smith, and Ndong ([31]) found that some students who switched out of engineering did not realize how important achievement in high school mathematics courses was or did not take the appropriate sequence of courses because no one provided the needed guidance. These findings suggest that the social capital gained from knowledgeable individuals can play a critical role in students' preparation and persistence.

Likewise, experiences of engineering majors during their first years in the program are important in persistence (Tyson, Smith, and Ndong [31]). Because students in their study population did not declare engineering majors until their junior year, they did not have opportunities to work in a lab or engage in engineering coursework.[1] In fact, Black students often experienced a department where homophily influenced how study groups and relationships with faculty were formed. Seymour and Hewitt ([24]) assert that social supports help engineering students cope with problems they encounter. Indeed, a majority of women and URM believed that their memberships in professional societies contributed to their persistence by providing access to resources (e.g. free tutoring, information about professors' teaching and exam styles) (Borman, Halperin, and Tyson [1]; Chanderbhan-Forde, Heppner, and Borman [4]). In contrast, the analysis of switchers' responses indicated that they 'did not know what to expect' before entering the engineering program (Tyson, Smith, and Ndong [31]). Thus, the influence of more knowledgeable individuals was essential in providing the critical information that influenced students' persistence, particularly for women (Seymour and Hewitt [24]) and students of color (Borman, Halperin, and Tyson [1]; Chanderbhan-Forde, Heppner, and Borman [4]), when compared to men.

STEM department cultures that did not espouse student-teacher relationships because of the 'weed-out system' appeared to be the biggest difficulty for women in their early years. Seymour and Hewitt ([24]) noted 'many [interviewees] offered 'fork-in-the-road' stories in which, having plummeted into depression, confusion and uncertainty, they sought the counsel of faculty about whether they should or should not continue. They were prepared to accept their professor's assessment ... ' (272). 'Care-less' responses of faculty often proved to be the last straw for women switchers. A strong need of women who were STEM majors was a supportive, personal relationship with faculty, and the gender of this instructor was less important than whether this need was met (Seymour and Hewitt [24]). Poor and Brown ([21]) found that establishing a mentoring program that provided women engineering majors with mentors who were alumni women engineers increased their retention. Not only did these students feel a stronger connection to engineering as a discipline, their relationship with their mentors also helped build their confidence about succeeding academically and professionally in the engineering workplace. Their mentors were also sources of insider knowledge about their courses and instructors, succeeding in the engineering workplace, networking, and internship and employment opportunities.

Collectively, the research cited above reveal factors beyond academic preparation that influence outcomes for women and URM, which suggest that they must acquire social capital and receive guidance and strategies from more knowledgeable persons to navigate their engineering programs and address challenges that are encountered. Because of this, it is important to understand the role that social capital plays in students' experiences and how they acquire additional social capital as they engage in their engineering programs. Thus, the research question guiding this study is, what systemic differences in social capital acquired prior to enrolling as engineering majors are there between women and URM and dominant groups?

To understand factors in persistence that include social connections and the environment, STEM researchers are increasingly studying social capital (see for example, Foor and Walden [8]; Prewitt, Daily, and Eugene [22]; Trenor et al. [30]). Social capital enables individuals to achieve goals they are not able to achieve on their own without access to resources through social ties and organizational participation. However, a common scholarly definition of social capital does not exist. We posit that social capital can be conceptualized in two broad categories: network-based and participatory social capital.

The first conceptualization of social capital emphasizes the networks of social ties in which an individual is embedded. Coleman ([6]) offers a 'collective' view of social capital as a function of societal structure, a group asset embedded in the relationships of individuals. Healy, Haynes, and Hampshire ([10], 111) express this 'collective' view in the claim that social capital refers 'to the norms and networks that enable people to work collectively to mobilize resources and achieve common goals.' Alternatively, Bourdieu ([2]) uses a more 'individualistic' view and defined social capital as 'connections' or access to a well-established network of useful relationships (e.g. who one knows) and material resources (e.g. engineering program information) that benefit group members. Similarly, Lin ([14]) argues that the focus should be on access to resources through ties to others that derive from group memberships. Members of the group gain access to resources (e.g. knowledge, materials, or privilege) available through their social networks. By activating these resources, they are able to attain benefits such as access to employment opportunities. Regardless of whether the conceptualization of social capital is collective or individualistic, these authors agree that its analysis must consider how socio-demographic attributes such as gender, ethno-racial grouping, age, education, etc. and the interactions between them are related to network position.

The second conceptualization of social capital emphasizes the resources accessible to an individual through their participation in organizations and/or institutions that have as part of their mission making available such resources. For example, in many U.S. high schools, school activities and programs, clubs, study groups, tutoring, and so on constitute resources that students can choose to access and activate. This form of social capital is less about the particular ties that a person has and more about the resources that are available through social group memberships. Small ([26]) is a proponent of this form of social capital, which is available through participation in organizations, emphasizing that organizations can differ in the resources they make available.

This study adheres to Lin's ([14]) individualistic conceptualization, which contends that social capital is conveyed through ties that connect an individual to influential others that control access to valuable knowledge and resources. To measure this type of social capital, Lin developed a position generator instrument in which respondents are not asked about ties to specific other persons (as is typical in many social network studies) but about any ties they have to persons in specific social positions (e.g. doctor, lawyer, professor, police officer). As Crossley et al. ([7]) posit, the instrument measures 'ties to potentially influential others ... to whom [respondents] might turn in an effort to secure various resources' (30). This conceptualization of social capital drives our interest in the personal networks of individuals formed by the persons they name as influential on their academic and career pathways.

This study is part of a longitudinal research effort that examines how social capital contributes to the retention and degree attainment of women and URM majoring in engineering. Our research combines network-based social capital with participatory social capital. We believe that resources must be available, accessible, and, finally, activated or taken advantage of in order for positive outcomes to occur and for an individual's social capital to develop. Thus, we examine the role played by an individual's network in the provision of available, accessible, and activated resources while they were high school students as it relates to their gender and ethno-racial group membership. A major premise of our work is that success in any endeavor requires appropriate knowledge and resources. This study addresses how women and URM may differ, as compared to dominant groups (e.g. White and Asian men) with respect to availability, access, and activation of engineering resources. Specifically, we ask about the availability, access, and activation of resources contingent on participation and assess students' personal social capital by the associates they view as influential on their career choice.

We administered an online survey that inquired about respondents' participation in STEM-related activities and programs and their experiences in STEM-related courses while in high school. We also asked respondents who they felt were influential on their decision to select engineering as a major. In the technical language of social network analysis, each respondent is an 'ego' and the persons they name as influential are 'alters' with the connections from ego to his or her alters constituting an 'ego-network' (Perry, Pescosolido, and Borgatti [20]). Consistent with Lin ([14]), we did not ask for the names of influencers, instead we asked whether someone in a particular status (e.g. parent or teacher) was influential in their decision. This protocol is an intentional departure from a standard ego-network design which typically asks for names in response to prompts like 'with whom important issues have been discussed in the last six months' and then asks respondents to classify the relationship in terms of standard social statuses like friend, co-worker, family member, and the like (Crossley et al. [7]). Our protocol reduces cognitive burden on respondents and reflects the range of alter-types identified as contributing to the success of engineering students through a free listing exercise conducted with engineering faculty, advisors, and students (Smith et al. [27]).[2]

The sample is a cohort of engineering students who responded to the survey from 11 universities serving diverse student populations (7 predominantly white institutions, 3 Hispanic-serving institutions, and 1 historically Black college and university). We used purposive sampling to gather information from diverse engineering programs and contexts. In this study, gender is operationalized as a binary variable and ethno-racial identity choices mix conventionally defined race categories (Black, Asian), ethnic categories (Hispanic), and national origin categories (Japanese, Cuban). We acknowledge the scholarship that denies the binary nature of gender (Oakley [19]) and scholarship that questions the assumption of fixed race and ethnic categories (see Zuberi and Bonilla-Silva [33]). Our treatment of gender and ethno-racial identity is driven by our need to engage with previous research on women and URM in STEM where the gender binary is common and the use of aggregate ethno-racial categories is widely found. We reject the idea that that simple membership in such categories is 'the causal mechanism of social differences' (James [13]). Indeed, the aim of this study is to understand the social underpinnings of the experiences that women and URM face in STEM educational experiences. Therefore, our total sample includes 2186 respondents to the survey whose engineering departments verified they were enrolled, including 1486 (68%) men and 700 (32%) women. Respondents had 17 options available to identify their ethno-racial identity. We collapsed these choices into five conventional race categories as indicated in Table 1, which depicts the ethno-racial and gender makeup of the sample.

Table 1. Ethno-racial grouping and gender of respondents*.

*Eight cases NA on ethno-racial grouping. aNative Hawaiian/Other Pacific Islander. bMiddle Eastern/North African/Arab.

Table 2 presents the three social capital variables used to analyze the data. Set I includes variables that measure participatory social capital on the dimensions of availability, accessibility, and activation. The first subset asks about participation in specific programs and if there was no participation, whether that was due to lack of availability or lack of activation. The second subset of variables asks about the availability of various general types of activities (rather than specific programs). The third subset probes the level at which a respondent activated his or her participation in available activities. When available, participation depends on other factors, such as having the finances to access the activities/programs and related resources. Finally, when a program is available and accessible (i.e. no barriers to participation), the student may choose to not activate participation or may participate at lower versus higher levels. Set II consists of variables that measure the ego-network of a respondent in terms of the statuses the respondent identifies as influential in their decision to pursue engineering (alters), including variables intended to probe how alters were influential in providing resources and advice.

Table 2. variables used in analyses.

We compared gender and ethno-racial identity groups with respect to network-based and participatory social capital using t-tests, ANOVA, and χ2 tests of independence. We first analyzed how gender influences participatory social capital. Tables 3–5 tabulate responses to the variables in Table 2 that ask about program participation (Set 1A), ask about available activities (Set IB), and measure activation in terms of frequency of participation (Set IC).[3] The first two tables use simple χ2 analysis of counts and the third table uses a simple t-test for difference in means and evaluates effect size using Cohen's d.

Table 3. Program participation by gender, frequencies.

Notes: Differences in distribution of responses by gender over the Yes, No, na; No, np categories are not significant (p <.05) with two exceptions: program 1, men underrepresented in Yes and overrepresented in No, NP and in program 2, men overrepresented in Yes and underrepresented in No/NP. aNo/NA = did not participate, not available. bNo/NP = did not participate, although available. cUnsure = Not sure program was available.

Table 4. Frequencies of available activities by gender.

*A significantly (p <.05) larger proportion of men reported availability of this activity (omitting those unsure). ***A significantly (p <.05) larger proportion of women reported availability of this activity (omitting those unsure).

Table 5. Activity activation by gender.

*Statistically significant, p <.001.

Men and women differ very little on the measures of availability and access with respect to participatory social capital though they differ somewhat more on activation. Only two of the 12 program participation variables show significant differences between men and women, with men less likely to participate in AP courses and more likely to participate in engineering courses. Similarly, men and women differ in only two of the 11 available activities variables, with men being more likely to participate in 'robotics club/competition' and less likely to participate in 'visiting science museums, planetariums, or environmental centers.' However, significant gender differences are far more prevalent in the activation variables. Women reported higher frequencies of participation in five of the 11 activities ('math/science tutoring [tutor or tutee]', 'oral and/or writing skills development', 'math/science study group', 'visiting science museums, planetariums, or environmental centers', and 'math/science camp') while men reported higher participation in only one of the 11 activities ('robotics club/competition'). The effect sizes associated with these statistically significant differences are small in conventional terms.[4]

Unlike with gender, we found many significant differences for ethno-racial identity on the availability, access, and activation items. Tables 6–8 tabulate responses to the variables in Table 2 that ask about program participation (Set IA), available activities (Set IB), and measure activation in terms of frequency of participation (Set IC). Expected levels of availability are calculated from the marginal distributions of availability and do not refer to levels expected on the basis of substantive considerations related to different social positions of the groups. Table 6 shows that Asian respondents participate more than expected based on the marginal distribution of participation in five of the 12 programs and less than expected in one. White respondents participate more in four of the 12 and less than expected in another four. Black/African American respondents participate more than expected in three of 12 and less than expected in none. Hispanic respondents participate more than expected in three of 12 and less than expected in five of the 12 programs. Table 7 shows that six of the 11 activities are more available than expected for Asian and White respondents while two are less available than expected for Black respondents and eight of 11 are less available than expected for Hispanic respondents.

Table 6. Program participation by ethno-racial grouping: summarya.

aExpected levels of participation are calculated from the marginal distributions of participation and do not refer to levels expected on the basis of substantive considerations related the different social positions of the groups.

Table 7. Available activities by ethno-racial grouping: summarya.

aExpected levels of availability are calculated from the marginal distributions of availability and do not refer to levels expected on the basis of substantive considerations related the different social positions of the groups.

Table 8. Activity activation by ethno-racial grouping: summary.

These findings suggest Asian and White students are advantaged (i.e. activities are more available to them), when compared to Black/African American and Hispanic students. However, when we examine activation with respect to the same set of activities in Table 8, Hispanic respondents are more engaged with the activities than other groups. They report more frequent participation in seven of the 11 activities than at least one other group and in two cases all other groups ('visiting science museums, planetariums, or environmental centers' and 'making industry tours and visits'). In nine of 11 activities, White students report less frequent engagement as compared to at least one other group (Hispanic students seven times, Asian students three times, and Black students once). The values of partial η2 (the effect size for an ANOVA) are small.[5]

To elicit an ego-network of influencers, respondents were asked to 'THINK BACK to when [they] were considering engineering as a college major while in high school or middle school and indicate the person(s) who influenced [their] decision-making process at that time in some way.' Respondents could pick one or more alters from the following list: (1) parent/guardian; (2) sibling/other family member; (3) peer/classmate/friend; (4) high school teacher; (5) high school counselor; (6) club/organization contact; and (7) other. A respondent's ego-network consists of a combination of one or more of these alter-types, for example, 'parent|sibling' is the ego-network of a respondent who indicated that only a parent and a sibling were influential.

Table 9 displays the top ten types of ego-networks reported by the respondents. There are 128 = 27 possible combinations of alter-types that could constitute an ego-network if we include the possibility of the empty network, i.e. no one mentioned as influential. The type of ego-network most often identified has one alter in it, the respondent's parent/guardian. Interestingly, the next most frequent type has six alters in it, all but the alter-type of other. Parents are alters in seven of the 10 top types, siblings in five of the 10, and teachers in five of the 10. Of those ego-network types that have just one alter in it, the most common among the top 10 were parent, followed by teacher, then by sibling, and finally by other. The top 10 types of alters covered 1517 or 69% of the respondents, with the remaining 31% distributed over 48 remaining types of ego-networks.

Table 9. Top ten types of ego-networks of influencers.

Tables 10–12 compare the ego-networks of men and women with respect to which alter-types they include, their size, and the top ten profiles in each gender group. As in the analysis of participatory social capital, gender has little effect. Women mention peer as influential more often than expected and men less often than expected. For both men and women, the most frequently cited alter is parent/guardian (cited by 79.3% of all respondents), with other contact as the least frequently cited alter (cited by 8.8% of all respondents). There is no significant difference in the size distributions of the ego-networks, and both gender groups have almost same top 10 ego-network profiles with parent identified as the most common single influencer. Eight of the profiles in the top 10 are the same for both groups. In fact, the first five are identical just differently ordered. The profiles 'other' and 'parent|sibling|peer|teacher' are in the top 10 for men but not the in the top 10 for women, while the types 'peer' and 'parent|sibling|peer|teacher|counselor' are in the top 10 for women but not the top 10 for men.

Table 10. Influential alters by gender.

aWomen mention Peer significantly more often and men less often than expected (given marginal distribution) at p <.05 level.

Table 11. Network size by gender (NMen = 1486, NWomen = 700)a.

a χ 2 = 10.43, df = 6, p = 0.11.

Table 12. Top ten types of ego-networks of influencers by gender (NMen = 1486, NWomen = 700).

Tables 13 and 14 compare ego-networks of alters by ethno-racial grouping. With respect to ethno-racial grouping, there are significant differences for three of the seven alter-types. Asian respondents mention school-related alter-types less often than expected as compared to the Black and Hispanic respondents. For four of the alter-types, including the most often mentioned influencer (parent/guardian), there are no differences between ethno-racial groups. Finally, there are significant differences in the mean size of the ego-networks such that Asian respondents have a smaller network on average when compared to the Black/African American, Hispanic, and White respondent groups. There are no significant differences between pairs of these latter three groups.

Table 13. Influential alters by ethno-racial grouping, frequencies.

*Significantly different at.05 level with Asian students underrepresented and Black and Hispanic students overrepresented. ***Significantly different at.001 level with Asian students underrepresented and White students overrepresented with respect to High School Teacher; Asian students and White students underrepresented and Black and Hispanic students overrepresented with respect to High School Counselor.

Table 14. Frequency of network size by ethno-racial groupinga.

aANOVA analysis of mean difference in size with Tukey HSD adjustment shows Black, Hispanic, and White students have significantly larger networks than Asian students at p <.05 level.

Respondents were asked to describe whether each alter-type (e.g. parents) influenced them in 15 different areas listed in Table 2, Set III. We measured the amount of alter-type's influence by determining the total number respondents who answered yes for each areas of influence for the whole sample and by gender and ethno-racial categories. Figure 1 shows that the parent/guardian alter-type that has the most influence on average for the entire sample. We found significant differences by gender and by ethno-racial grouping in the amount of influence that alter-types have. Tables 15 and 16 report these results along with Cohen's d for the t-tests and partial η2 for the ANOVA results. Sibling/other family member has more influence for women than for men and teacher has more influence for Hispanic respondents when compared to Asian and White respondents. However, the effect sizes are small even for the two cases where there is a statistically significant t value and F ratio.

Graph: Figure 1. Distribution of Amount of Influence by Alter-types in Table 2, Set III.

Table 15. Comparison of alter influence by respondents' gender.

Table 16. Comparison of alter influence by respondents' ethno-racial grouping.

Consistent with other researchers, we assert that factors other than academic preparation influence outcomes for women and URM in engineering programs (Margolis, Fisher, and Miller [16]; Hill, Corbett, and St. Rose [11]; Sax [23]; Shapiro and Sax [25]; Summers and Hrabowski III [29]). In particular, we argue that social capital obtained from more knowledgeable others support students' decisions to pursue an engineering degree.

Although the focus on social capital is a useful framework for our study, some researchers believe such interpretation of social capital inherently focuses on the 'deficits' of women, and in particular, of members of URM population (Yosso [32]). We acknowledge that women and URM who enroll in engineering programs bring with them rich knowledge and understanding that can benefit a field dominated by White and Asian men. With the inclusion of women and URM in the profession, solutions to societal problems can be addressed in ways that are inclusive and that might not be considered otherwise. For example, women and URM may provide insights that ensure that approaches take accounts of the ways in which individuals from different cultural and linguistic background may interact with developed ideas, tool, or resources. For this to happen, women and URM must have access to and thrive in a culture currently dominated by White and Asian men. Therefore, women and URM must have access to the social capital needed to not only decide to pursue an engineering major, but also to navigate the dominant engineering culture, of which they may have limited insights.

Earlier, we discussed two possible types of social capital. The first, network-based social capital, highlights the social ties in which a person is embedded (Coleman [6]; Healy, Haynes, and Hampshire [10]). This type of social capital places emphasis on the knowledge and resources available from an individual's network, which is shaped by their sociocultural experiences (Lin [14]). Women and URM enter their engineering programs with strong social and cultural networks, however because of the historical underrepresentation of women and minorities in engineering fields their networks might not include individuals who have access to insights, knowledge, and resources that are directly supportive of their pursuit of engineering, which may be readily available to White and Asian men who dominate the disciplinary culture. For example, we found that parents/guardians, followed by school-based alters (e.g. teachers, counselors) were identified as the most influential alters in URM' decision to pursue engineering when compared to others. This suggests that the type of social capital obtained from these individuals, which depends on their historical, social, economic and experiential backgrounds, influences the types of knowledge they are able to convey (see for example, Hardie [9]). If the alters' backgrounds are not embedded in the socio-cultural norms of engineering programs, they can provide encouragement and support, but not the context- or disciplinary-specific social capital students need to thrive in engineering programs. Therefore, it might be important to engage parents and school-based alters in interventions designed to develop their understanding of STEM and academic requirements that support STEM so that they can better support their children/students.

In contrast to network-based social capital, we frame participatory social capital as an approach to make needed resources, not readily accessible from a persons' social network, accessible through their participation in organizations and institutions that have as part of their mission to make available such resources. Such an approach acknowledges the historical and social development of culture that may have excluded individuals from certain groups and allows for proactive measures to support more inclusive and thoughtful practices, thereby providing a means for obtaining the social capital needed to thrive engineering programs. This approach is less reliant on a person's social network and more focused on their promise or potential by intentionally providing opportunities to uncover needed social capital. Participatory social capital expands the social capital of groups by focusing on areas of common interest, growth, and development. Through participation in STEM-related activities, for example, groups of individuals from different socioeconomic and sociocultural backgrounds can access information and resources that can support their decision making and success.

Although men and women differed little on availability and access to STEM-related resources, we found significant differences in activation based on the ethno-racial group. These differences may influence the ability of URM to obtain participatory social capital through engagement in these activities, many of which are designed to provide the needed support to thrive academically and/or generate interest in STEM. This suggests a need to attend to influences on activation as interventions are developed and implemented to support STEM interest, recruitment, and retention and determine approaches for encouraging the engagement of all students.

In the prior section, we assert that engaging in academic and STEM-related activities may be a more productive approach to providing women and URM access to social capital that they may not develop simply from their social networks. To determine the efficacy of such an approach, research is needed to test this hypothesis and examine what types of social capital women and URM gain from participating in activities, particularly those designed to engender interest or success in STEM. Further, many interventions are designed to engender students' interest in STEM and to provide academic supports to encourage their success. Our findings related to students' influencers suggest that the audience for STEM-related interventions should be expanded to include the individuals most likely to influence students' decisions, which include parents and school-based personnel.

We compared women and URM to more the dominant group in engineering, White and Asian men. Although this provided some significant results, the effect sizes were small and therefore the findings are inconclusive. Additional research is needed to examine differences between and among ethno-racial groups. For example, it might be important to compare the experiences of each group independently. Rather than consider White women and URM as a collective group, it might be important for researchers to examine social capital differences between each group against the dominant group as well as differences between White women and URMs. Similarly, researchers should explore differences between women and URM from low- and high-SES backgrounds. Such comparison might highlight nuanced differences that are not visible when the groups are combined. Our results highlight the differences among groups, but do not explain why such differences exist. Research is therefore needed that examines causes behind those differences, particularly as related to activation of available and accessible resources.

Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.

No potential conflict of interest was reported by the authors.

Rebecca Campbell-Montalvo http://orcid.org/0000-0003-2671-8056