Advice for Employers and Recruiters
Part 5: As the world interns: The impact of identity and social, economic, and cultural capital on college student internship engagement | Results, Research Questions 1 and 2 & Summary
This is the fifth of six articles in this series, click here to go to the first article. If you’re searching for a remote internship, go to our search results page that lists all of the remote internships and other entry-level jobs advertised on College Recruiter and then drill down as you wish by adding your desired category, location, company, or job type.
CHAPTER FOUR: RESULTS
This chapter details the descriptive and statistical results that I obtained through my analysis. I begin with an overview of descriptive statistics and demographic information about the sample. To answer my first research question, I review the results of my t-test and chi-square analyses. I address research question two using five separate regression analyses to predict internship participation. The chapter then concludes with a summary.
4.1. Descriptive Statistics
The population for the current study consisted of 350 senior students at the University of Vermont who took the National Survey of Student Engagement in 2017. Of the 350 students in the sample, 231 of those students (66%) engaged in an internship during their time in college and 119 students (34%) did not. The distribution of categorical demographic variables (sex, disability status, sexual orientation, race, residence, and first-generation status) is provided in Table 3
Information about the distribution of continuous variables (i.e., GPA, median household income, and the eight NSSE engagement indicators: Quality of Interactions, Supportive Environment, Student-Faculty Interaction, Collaborative Learning, Higher-Order Learning, Reflective and Integrative Learning, Learning Strategies, and Quantitative Reasoning) is provided in Table 4.
4.2. Research Question One
The first research question was “Do students’ identities and/or economic, social, and cultural capital indicators significantly differ depending on participation in an internship?” To answer this question for continuous variables, I conducted an independent samples t-test comparing students who engaged in internships to students who did not engage in an internship. I measured effect size using Cohen’s d and summarized results in Table 5.
Results of the t-test demonstrated that six of the indicators significantly differed depending on internship engagement. There was a significant difference between GPA for students with internship engagement (M=3.38, SD=.37) compared to no internship engagement (M=3.20, SD=.47); t(348)=3.93, p<.001, d=.43. Cohen (1988) suggested that a Cohen’s d of .8 is considered a large effect size; .5 is a moderate effect size; and, .2 is a small effect size. In the case of GPA, the effect size (d=.43) was moderate.
While a much smaller effect size, there was also a significant difference for the NSSE engagement indictor Reflective and Integrative Learning t(348)=2.26, p=.024, d=.26. Students who interned (M=40.42, SD 11.92) had significantly higher scores for Reflective and Integrative Learning compared to those who did not engage in an internship (M=37.41, SD=11.60).
Scores on the NSSE engagement indicator Collaborative Learning also significantly differed between interning (M=37.22, SD=13.05) and non-interning (M=34.16, SD=14.90) students; t(348)=1.97, p=.049, d=.22. A Cohen’s d of .22 demonstrated only a small effect size for Collaborative Learning scores, but a significant one nonetheless.
Scores for Student-Faculty Interaction had the greatest effect size between students who interned (M=26.59, SD=14.97) and those who did not (M=20.35, SD=13.46); t(348)=3.82, p<.001, d=.44. The effect size for Student-Faculty interaction (d=.44) was a moderate one.
Supportive Environment NSSE engagement indictor scores were also significantly higher for interning students (M=35.23, SD=11.78) compared to students 69 without internships (M=32.51, SD 12.04); t(348)=12.04, p=.044, d=.23. The effect size for Supportive Environment would be considered small (d=.23).
Lastly, Median Household Income differed significantly between students who interned (M=94250.20, SD=36358.95) and those who did not intern (M=81664.26, SD=33267.37); t(348)=3.06, p=.002, d=.36. While the effect size was small, this result still indicated a significant different in income between the two groups.
To examine differences between categorical variables (i.e., sex, disability, sexual orientation, race, first generation status, and residence) for interning and non-interning students I used a chi-square analysis, the results of which are displayed in Table 6.
I observed significant differences between interning and non-interning groups for two variables: first-generation status and residence. Students who were not first generation college students were significantly more likely to engage in an internship experience compared to their first-generation peers. (χ2 =4.2, df=1, p=.036). Students from outside of the state of Vermont were also significantly more likely to intern compared to their in-state peers (χ2 =10.38, df=1, p=.003).
My first research question asked whether students’ identities and/or economic, social, and cultural capital indicators significantly differed depending on participation in an internship. Results demonstrated that several factors significantly differed when comparing interning and non-interning students. Students with internship experience were significantly more likely to have a higher cumulative GPA and median household income, higher scores on NSSE indicators Reflective and Integrative Learning, Collaborative Learning, Student-Faculty Interaction, and Supportive Environment, and were significantly more likely to be from outside the state of Vermont and not to identify as a first-generation college student.
4.3. Research Question Two
My second research question was “What are the factors that significantly relate to undergraduate students’ participation in internships?” I conducted five separate regression analyses to examine whether individual factors significantly predicted internship engagement, as well as whether the larger combined themes of identity, social capital, economic capital, and/or cultural capital predicted participation in an internship. Results demonstrated that although only a few individual factors were significant predictors of interning, all larger themes were significant predictors.
I used binary logistic regression to estimate the impact of 16 individual factors on internship participation. Binary logistic regression models use data to predict the likelihood of a particular event (in this case, internship participation). For all analyses, I used a type I error rate of .05 to establish statistical significance. I used pseudo R2 measures Cox and Snell’s R2 and Nagelkerke R2 to explain the amount of variation that is accounted for in each block, which is an indication of the model’s power (Hu, Shao, &
Palta, 2006). Both Cox and Snell R2 and Nagelkerke R2 range from zero to one, with zero indicating that the model has no predictive power and one indicating that the model perfectly predicts results (Hu et al., 2006). The likelihood-ratio chi-square indicates the statistical significance of the overall model, and the Wald statistic demonstrates the statistical significance of each individual predictor variable (Howell, 2010).
I began with a binary logistic regression of all 16 factors to examine whether individual items predicted internship engagement, the results of which are displayed in Table 7.
The regression model containing all variables was statistically significant, (16, N=350) = 41.070, p<.001. This demonstrated that the model differentiated between interning and non-interning students and correctly classified 72.2% of cases.
Two individual factors were statistically significant in regression containing all variables. However, though Quality of Interactions (B=-.03, p=.038) and Student-Faculty Interaction (B=.04, p=.002) were both significant predictors of internship engagement, the odds ratio of each was so small that the results did not have much real-world significance. The exponential regression coefficient (odds ratio) indicates the likelihood of an event happening (Howell, 2010). For every one point decrease in Quality of Interactions score, a student is .97 times more likely to engage in an internship. A similarly negligible result showed that for every additional point in Student-Faculty Interaction score, the likelihood of a student engaging in an internship increases by 1.04 times. Overall, the model containing all 16 items had a small amount of explanatory power, accounting for 12.6 to 17.5% of the variance in internship participation.
After examining individual factors, I conducted four separate regressions that organized the 16 variables into separate themes: Identity, social capital, economic capital, and cultural capital (see Table 1 for the categorization of variables). Several individual items within these themes had significant predictive power.
The first theme of identity is displayed in Table 8 and contained the following five variables: sex, disability, sexual orientation, race, and residence. The identity theme accounted for only 3.7 to 5.2% of variance in internship participation. However, residence became statistically significant 75 when looking at only these five factors. Residence had an odds ratio of .46, meaning that students from the state of Vermont were .46 more likely to engage in an internship (p=.002). Put another way, students from outside of Vermont were 2.2 times more likely to engage in an internship than their peers from within the state.
The second theme of social capital contained four items: NSSE engagement indicators Quality of Interactions, Supportive Environment, Student-Faculty Interaction, and Collaborative Learning. This theme correctly classified 67.6% of cases. Regression results for the social capital theme are displayed in Table 9.
The theme of social capital also accounted for only a small amount of variance in internship participation; 4.7 to 6.6%. The social capital theme correctly predicted 66.9% of cases. Student-Faculty Interaction was the one item in theme two that was significant, but with an odds ratio of only 1.03 (p=.003), the result had limited real-world significance.
Theme three was economic capital and contained two items: first-generation status and median household income. Regression results for theme three are presented in Table 10.
The economic capital theme accounted for only 3.3 to 4.5% of variance in internship engagement. The economic capital regression model correctly predicted 65.3% of cases. While median household income had a significant Wald statistic (6.77, p=.012), an odds ratio of exactly 1.00 meant that this statistic was not meaningful.
The last of the four themes was cultural capital, containing the following five items: GPA, and NSSE engagement indicators Higher-Order Learning, Reflective and Integrative Learning, Learning Strategies, and Quantitative Reasoning. Theme four is presented in Table 11.
Cultural capital accounted for only 5.4 to 7.5% of variance in internship participation and correctly predicted 68.5% of cases. There was one significant variable within this theme, which was Cumulative GPA (B=1.01, p<.001). For every 1.0 increase in cumulative GPA, a student was 2.74 times more likely to engage in an internship.
Table 12 displays a summary of the five regression models, including the regression with all variables and the four themed regressions.
All five regression models were significant at the p<.05 level, though the full model of all 16 factors had the best predictive power of all of the models, χ2 (16, N=350) = 41.07, p<.001. Because the smaller themed regressions held relatively low predictive power, and less predictive power than the full model, I cannot conclude that my four themes of identity, social capital, economic capital, and cultural capital significantly predicted internship participation.
The most salient individual factors that emerged as predictors of internship engagement were cumulative GPA in the cultural capital theme and residence in the identity theme. Students with higher GPAs and students who were from outside of the state of Vermont were both significantly more likely to intern when compared to their peers with lower GPAs or those who are from hometowns in Vermont.
4.4. Summary
This chapter summarized the results of my two research questions. Research question one was “Do students’ identities and/or economic, social, and cultural capital indicators significantly differ depending on participation in an internship?” My analysis demonstrated that several different factors significantly differed between interning and non-interning students, including participants’ residence, first-generation status, cumulative GPA, median household income, and results on NSSE indicators Reflective and Integrative Learning, Collaborative Learning, Student-Faculty Interaction, and Supportive Environment.
Research question two asked, “What are the factors that significantly relate to undergraduate students’ participation in internships?” The themes of identity, social capital, economic capital, and cultural capital all significantly predicted internship engagement of the participants, but with low levels of predictive power. Individual factors of Quality of Interactions, Student-Faculty Interaction, residence, median household income, and GPA all significantly predicted participation in an internship, though the two predictors with the most real-world significance were GPA and residence.
— This is the fifth of six articles in this series. Click here to go to the next article. This series of articles are courtesy of Amanda Chase. Amanda Chase is the director of strategic engagement for the collective impact organization Advance Vermont, where she works to increase access to postsecondary education. She also has a private consulting business, and previously worked as the internship coordinator for the University of Vermont. Amanda has worked with a wide variety of businesses to support their hiring goals, from one-person grassroots organizations to Fortune 500 companies. Her hundreds of individual career counseling clients have included high school students applying to first jobs, adults making significant career transitions, retirees seeking encore careers, and everything in between. She received a bachelor of arts in psychology from Hamilton College, a master’s degree in counseling from the University of Vermont, and an Ed.D. in educational leadership and policy studies from the University of Vermont. Her work has always centered on issues of equity and access in education and career development. To learn more about Amanda or to get in touch, visit her website.
