Pathways to Mobility: Using a Multidimensional Approach to Examine Postgraduation Outcomes for Promoting College Student Success

Theoretical Groundings

Many studies interrogating the outcomes associated with attending postsecondary education draw on the assumptions of human capital theory, proposing that an individual’s skills and knowledge are advanced by obtaining higher levels of education, which are then rewarded in the labor market by improved socioeconomic outcomes. Studies using human capital theory assume that education helps develop productive skills valued in the labor market, inferring causality between higher education and economic outcomes (Becker, 1964; Mincer, 1974). These enhanced skills and knowledge can include accomplishing tasks more effectively, learning new skills more efficiently, and developing desirable dispositions such as reliability or leadership (Tomaskovic-Devey et al., 2005), which are rewarded in the labor market. However, researchers have also suggested that returns associated with higher education are more complicated than merely acquiring desirable skills and experiences (Tomaskovic-Devey et al., 2005).

Human capital theory also treats education as a largely homogenous experience, frequently considering the level of education an individual receives and how that impacts singular measures of socioeconomic status (Torche, 2015). However, as the American Psychological Association (APA, n.d.) describes, socioeconomic status is more complex than a single measure.

Socioeconomic status (SES) encompasses not just income but also educational attainment, financial security, and subjective perceptions of social class. Socioeconomic status can encompass quality of life attributes as well as the opportunities and privileges afforded to people in society (para. 1).

Mueller and Parcel (1981) further emphasize the relational aspects of socioeconomic status and articulate that socioeconomic status:

. . . describes social systems (usually society or community) in which individuals, families, or groups are ranked on certain hierarchies or dimensions according to their access to or control over some combination of valued commodities such as wealth, power, and social status (p. 14).

Despite the complexity of socioeconomic status highlighted above, single measures remain the primary means of analyzing the socioeconomic returns of higher education. In this paper, we seek to explore the complexity of higher education’s impact on socioeconomic outcomes by examining it as a multidimensional construct.

Measurement of Socioeconomic Returns From Higher Education

Scholars across sociology, economics, and education disciplines have studied the impact of higher education on individuals’ socioeconomic status from various fields and perspectives. Sociologists and economists have examined these concepts from the perspective of intergenerational mobility at the macro level (Chetty et al., 2017b; Torche, 2015), while the field of higher education research has focused more on the micro-level interactions between students and institutions (Benson et al., 2015; Thomas & Zhang, 2005), and career counseling has examined non-monetary aspects such as satisfaction and major alignment (Melguizo & Wolniak, 2012). These fields contributed to our understanding of the impact of higher education attendance. However, few studies have considered the complexity of socioeconomic outcomes.

Data Sources and Sample

Data for this research study comes from the third cohort of the U.S. NCES Baccalaureate and Beyond Longitudinal Study (B&B:08/12). In creating the sample for this study, we limited institutions to four-year degree-granting institutions included in the B&B:08/12 study to focus our examination on those graduates who received a bachelor’s degree. Additionally, to focus our analysis on first-time bachelor’s degree recipients, the student sample was limited to those who had not previously received a bachelor’s degree before their 2007/2008 graduation and those who completed both the 2009 and 2012 surveys. Individuals who were enrolled in school in 2012 were also removed from the final sample as their socioeconomic status could have been impacted by their return to school, resulting in n = 7,250.

Measures

To expand the conceptualization of socioeconomic status, we used multiple measures rather than a single measure, such as occupation, income, class status, or wealth (Table 1). In selecting the variables to include in our measures of socioeconomic status, we selected items that captured different dimensions of graduates’ socioeconomic well-being and success. The first dimension focuses on financial success and is captured by graduates’ salaries, home ownership, and the percentage of income that goes toward student loan repayment. Homeownership has historically been considered a financial and socioeconomic achievement marker and an essential component of the “American Dream” (Hirschl & Rank, 2010, p. 125). The ability to own a home, especially shortly after graduation from college, demonstrates substantial socioeconomic security. Additionally, as the student loan crisis continues to hamper the socioeconomic success of college graduates (Kim & Chatterjee, 2019), loan repayment is a crucial element to consider.

The second dimension is job quality, which includes whether the job offers benefits and how satisfied individuals are with their job security. These elements move beyond the direct monetary benefits of employment. Employer benefits are essential to financial security in that they provide both short-term economic security by helping pay for medical care and long-term economic security through retirement savings (Loya et al., 2020). Additionally, being able to receive preventative medical care is important to graduates’ overall well-being (Benach et al., 2014). Job security satisfaction also indicates a level of socioeconomic success, where graduates are employed not just for a paycheck but also do not worry about whether they will continue to have a job. Job security satisfaction is not only important for overall well-being (Bialowolski & Weziak-Bialowolska, 2021), but for subjective views of socioeconomic status (Ren et al., 2022).

The third dimension is socioeconomic stability, which includes the percentage of time unemployed since graduation and the number of jobs graduates hold. During the unstable economic climate that has been present since the housing crisis of 2009, college graduates have had to traverse a more difficult job market than previous graduates. Only considering if graduates are employed during discrete time periods (when surveys are conducted) may miss the socioeconomic instability present before obtaining employment. Many college graduates struggle to secure jobs following graduation, and while they may currently be employed, the impact of being unemployed previously and the duration of unemployment impact socioeconomic success (Western et al., 2012). Additionally, it is also not uncommon for individuals to hold multiple jobs to meet financial obligations such as paying rent or repaying student loans (Caza et al., 2022). Having to work multiple jobs again adds important nuance to socioeconomic success that is not captured by employment or salary alone. Instead, it indicates that more than one job is needed to meet an individual’s financial needs (Caza et al., 2022).

Data Preparation

In order to conduct LCA, we first converted values from the B&B:08/12 dataset that were continuous (loan repayment, salary, and unemployment time) into categorical values to be compatible with LCA (Wang & Wang, 2020). For the loan repayment and income variables, the distribution of the responses was analyzed to establish quantile cutoffs that represented the median, 25^th percentile, and 75 percentile, which we used to create groups of categorical variables. For the percentage of time unemployed variables, the distribution was likewise examined to determine how to best create categories from the data. For this variable, most respondents had 0–2 jobs, with a much lower number having 3 or more. We therefore decided to retain the existing categories of 0, 1, and 2 jobs and then combine those with 3 or more jobs into a single group. Finally, the percentage of time unemployed had a narrow distribution of responses, with the majority falling below 2% in both 2009 and 2012. Due to the limited distribution, we used cut points at the 33^rd and 66^th percentiles to create three categories. Students’ majors were dummy coded for inclusion in the model.

Additionally, NCES uses a variety of codes to denote individuals who did not respond to items within the survey. For B&B:08/12 many respondents were coded as skipped for several of the variables. However, within NCES survey methods, skipped is different from missing. Instead, it denotes that respondents did not receive the question due to their response to a previous question. For example, in 2009, the loan repayment question was asked only for individuals who indicated that they had taken out a loan to pay for college. Meaning that those marked skipped never took out loans to pay for college. As these skipped responses still capture important elements of respondents’ socioeconomic status, we chose to keep responses coded as legitimately skipped as part of the B&B:08/12 interview protocol (NCES, 2012). We instead renamed the response option to reflect the question that resulted in the survey item not being administered (i.e., skipped for loan repayment is now never loan). With these variables recoded into usable values, the remaining missing data for all variables was negligible (< 1%).

There were also some inconsistencies across the two waves of data in terms of how questions were asked and response options. For example, in 2009, respondents were asked, “Do you own a home, pay rent, or both own a home and pay rent?” and response options included: 1 = own a home or pay mortgage, 2 = pay rent, 3 = both own home and pay rent, 4 = neither own home nor pay rent. However, while respondents were asked the same question in 2012, NCES changed the variable options in the dataset to break out the possible responses into dichotomous options. The response option that most closely aligned with the 2009 question had the response options: 1 = paid mortgage, 0 = did not pay mortgage. To align the questions as best as possible, the 2009 and 2012 responses were recoded into own or do not own a home. Table A.1 in the appendix breaks down the recoding for home ownership, job security satisfaction, and benefits. Table 1 shows the final variable codes and both weighted and unweighted frequencies.

Note. Data are from the U.S. Department of Education, National Center for Education Statistics, 2008/12 Baccalaureate and Beyond Longitudinal Study (B&B:08/12).

Data Analysis

To examine socioeconomic status following graduation, we utilized LCA modeling to determine if graduates cluster into distinct groups based on socioeconomic indicators. LCA is an appropriate technique for this analysis because it can divide individuals into unobserved (latent) subgroups or classes based on the observed variables (Oberski et al., 2015). LCA is preferable to simple cluster analysis because the probability modeling that underlies LCA allows formal statistical analysis for determining the number of clusters (Magidson & Vermunt, 2002). This type of modeling lends itself well to examining socioeconomic outcomes since it allows for a multidimensional examination of socioeconomic success rather than unidimensional measures such as income or occupation. Full Information Maximum Likelihood (FIML) was used to address the limited remaining missing or incomplete data. Full information maximum likelihood estimation was used for all models, which allowed for missing data under the missing-at-random assumptions (MAR; Little & Rubin, 2002). NCES data weights were also applied to the analysis to adjust for any nonresponse bias that results from attrition across the different waves of data collection.

To determine the appropriate number of classes in both 2009 and 2012, we conducted LCA using Mplus 8.3 using a series of steps (Muthén & Asparouhov, 2011). The first step was to create unconditional probabilities for class membership for an individual’s socioeconomic status in 2009 and 2012. The unconditional probability indicates the proportion of the population expected to belong to a latent class. We began with one class and then increased the number of classes until fit indices, including Akaike’s information criterion (AIC), Bayesian information criterion (BIC), the sample-size-adjusted BIC (ABIC), Lo-Mendell-Rubin likelihood ratio (LMRLR), and adjusted Lo-Mendell-Rubin likelihood ratio (ALMR LR) did not improve the fit (Wang & Wang, 2020). Once we determined the appropriate number of classes for each year, we utilized the probabilities of belonging to each class for different socioeconomic variables responses to interpret the categories and provide a general description for each class. Finally, after the number of classes were selected, the covariates were added to the model to examine the relationship between college major and family income of the likelihood of students belonging to a specific class.

Limitations

The exploratory focus of the study presents several limitations that should be acknowledged. First, our ability to measure an individual’s socioeconomic outcomes was also limited by data availability within the B&B:08/12 dataset. Some participants in the 2009 study did not complete the 2012 study, meaning we had to exclude them from the analysis. Additionally, while more robust than a single measure, socioeconomic status measures may only encompass some elements of individual success. Other non-monetary aspects of socioeconomic well-being, such as individuals’ assessment of their subjective well-being, were not available within the dataset and, therefore, not considered in the analysis.

Latent Class Identification

For the 2009 LCA analysis, the 2-class model has the largest AIC BIC and ABIC values, indicating that this model fits the data worse than all the other models. In addition, while the 2-class model had perfect entropy, the p-values of the LMR LR and ALMR LR are all < .0001, indicating that additional latent classes could provide a better fit for the data. The 4-class model shows slightly better fits for the fit indices than the 3-class model. However, the p-values for both the LMR LR and ALMR LR are significant, indicating that the fit for the 3-class model is preferable to the 4-class model. Comparative fit indices are shown in Table 2.

Note. Data are from the U.S. Department of Education, National Center for Education Statistics, 2008/12 Baccalaureate and Beyond Longitudinal Study (B&B:08/12).

Based on the fit indices, we selected the 3-class model to examine the classes’ interpretability since the classes’ interpretability in LCA is as important as fit (Wang & Wang, 2020). We examined the interpretability of the latent class membership using the estimated item probabilities and how each item differed across classes (see Table 3). From the probabilities, three distinct groups emerged based on the socioeconomic indicators: SES insecure (Class 1), SES secure (Class 2), and SES unemployed (Class 3).

Like the 2009 class identification, LCA was run for the socioeconomic responses provided by graduates in 2012 to determine the appropriate number of classes. Table 4 presents summaries of the LCA fit indices for two to four classes. Both the 3-class and 4-class models showed better fits for AIC, BIC, and ABIC than the 2-class model, with the 4-class model providing slightly better fits across all indices. All the p-values for both the LMR LR and ALMR LR for the 3-class model were not significant at p < .001, and the 4-class model was not significant at p < .01. However, the entropy for the 4-class model was higher (0.694) than the 3-class model (0.633).

Note. Data are from the U.S. Department of Education, National Center for Education Statistics, 2008/12 Baccalaureate and Beyond Longitudinal Study (B&B:08/12). SES = socioeconomic status.

Note. Data are from the U.S. Department of Education, National Center for Education Statistics, 2008/12 Baccalaureate and Beyond Longitudinal Study (B&B:08/12).

Since the fit indices did not provide a clear picture of the appropriate number of classes, we examined both the 3-class and 4-class probabilities for interpretability. Like the 2009 LCA, the 3-class model for 2012 (see Table 5) provided a more interpretable grouping of individuals and was selected as the appropriate model. The 3-class model probabilities are provided in Table 5. Based on the socioeconomic indicators, three distinct groups emerged: SES insecure (Class 1), SES secure (Class 2), and SES stable (Class 3).

2009 Latent Classes

The first class, denoted by the blue diamonds in Figure 1, included the highest percentage of respondents (54%) and was categorized as SES secure (i.e., in Class 2). In this group, the probabilities indicated a low likelihood of having loan repayments that were over 9% of their incomes; the highest likelihood of owning a home (23%); a high likelihood of making over $26,000 a year; the highest probability of making over $39,5001; a high likelihood of responding they had a job with benefits; a high probability of responding they had been unemployed for less than one percent of the time since graduation; a high likelihood of responding they were satisfied with the security of their job; but also a high probability of responding they were working two jobs. These probabilities taken together indicate that while the graduates in this category may have some student debt and more than one job, they have achieved a relatively stable level of socioeconomic status. The lower levels of loan repayment are of specific interest, as this is a hotly contested issue across the country (Goldrick-Rab et al., 2016). The relationship between higher salaries and a lower percentage of income makes loan repayment more feasible. Additionally, the high likelihood of holding two jobs suggests that graduates may be taking on more work to manage expenses more easily, such as student loans (Velez et al., 2019).

Compared to the SES secure group, the SES insecure and SES unemployed groups had class probabilities indicating a less stable level of socioeconomic status. The SES insecure group, denoted by the red circles in Figure 1, had the highest probability of responding that they had no student debt; a high likelihood of not owning a home; a high likelihood of only making between $10,000 and $26,000 a year; a high likelihood of having no benefits; a high likelihood of being unemployed for less than one percent of the time; were relatively evenly split between being satisfied with their job security and not satisfied; and a high likelihood of having more than two jobs. These probabilities indicate that these students may have achieved less socioeconomic stability than the SES secure group. One explanation for this may be that students without debt have more flexibility to take a job that pays less but aligns with their interests or a low-paid internship but may take on multiple positions to ensure enough income to cover expenses (American Student Assistance, 2015).

Note. Data are from the U.S. Department of Education, National Center for Education Statistics, 2008/12 Baccalaureate and Beyond Longitudinal Study (B&B:08/12). SES = socioeconomic status.

The SES unemployed group, denoted by the green diamonds in Figure 1, was distinct from the SES insecure group by their high likelihood of not being employed at all in 2009. Interestingly, they also displayed a high likelihood of never having taken out loans to pay for college. Additionally, the response likelihoods were evenly split across the responses for the percentage of time spent unemployed. These results indicate that while the graduates were currently unemployed, some had been employed at some point since graduation. These graduates may personify students that struggle to find secure employment or find themselves underemployed following graduation. Underemployment has been an especially troubling trend for students who graduated around the time of the Great Recession—over 40% of recent college graduates report being employed in a role that does not require a college degree (Federal Reserve Bank of New York, 2024).

2012 Latent Classes

Four years after graduation, in 2012, graduates’ socioeconomic status grouped into similar clusters with both a socioeconomically secure and insecure group readily identifiable, with most graduates falling into the secure group. However, the unemployed group was no longer present, and a new group of SES stable was identified. Graduates in the secure group (see blue triangles in Figure 2) had response probabilities that were split across levels of student debt as a percent of income, with a moderate likelihood of responding that they had either substantial levels (2%–12%) or no debt; a high probability of not owning a home; a high probability of responding they were making $22,800 to $37,000 a year2 and a low probability of responding that they were making less than $22,800; a high likelihood of having a job with benefits; the highest probability of responding they had been unemployed less than one percent of the time since graduation; a high likelihood they were satisfied with their job security; and a high likelihood of having two jobs. While this group did have lower salaries and a range of student debt levels, they presented non-monetary markers of socioeconomic status, including benefits, satisfaction with job security, and less unemployed time. As the largest percentage of respondents, the emergence of the satisfied group suggests that students may have prioritized positions with job security over higher levels of income.

Compared to the SES secure groups, the SES stable group (denoted by the green squares in Figure 2) had probabilities indicating a higher socioeconomic status level across most measures. The SES secure group had the highest probability of paying less than 4% of their income in student loans and the highest likelihood of paying nothing. Additionally, they had the highest probability of responding that they owned their own home, made over $53,000 a year, had a job that provided benefits, were satisfied with their job security, and were unemployed less than one percent of the time since graduation. However, they also still had a high probability of working two jobs. A potential explanation for this finding is that for the secure group, this second job could be used by graduates to bolster economic security or allow them the resources to purchase a home. Overall, these probabilities suggest the graduates had reached a relatively stable socioeconomic status, to the point where almost half were able to purchase their own home.

At the other end of the spectrum, the SES insecure group, denoted by the red circles in Figure 2, was much less likely to achieve higher levels on the socioeconomic measure included in this analysis. Like the satisfied group, the insecure group was polarized across the level of student debt, with a moderate probability of responding that they had high (12%+) and no student debt. Graduates in this group also had a high likelihood of responding that they did not own a home, were making less than $22,800 a year, and a low likelihood of responding that they were in a job that provided benefits. They had a moderate likelihood of responding that they had been unemployed less than one percent of the time, were not satisfied with their job, and had only one job. These probabilities suggest that these students may be in a more challenging socioeconomic position than their SES satisfied or SES secure peers and less satisfied with their status.

Relationship With Covariates 2009

There were also significant differences in the likelihood of graduates belonging to different classes based on major and family income. In 2009, graduates who majored in computer science and engineering, compared to business, were more likely to be in the SES secure group versus the insecure group. In contrast, graduates who majored in biology, general studies, social sciences, humanities, education, and other majors were more likely to be in the insecure group versus the secure group compared to business graduates. Additionally, computer science, engineering, and biology graduates were more likely to be in the unemployed group versus the insecure group than business graduates. Education graduates were less likely to be in the unemployed group than the insecure group, compared to business graduates. Socioeconomic status did not influence group membership in 2009 (see Table 6).

Note. Data are from the U.S. Department of Education, National Center for Education Statistics, 2008/12 Baccalaureate and Beyond Longitudinal Study (B&B:08/12). Business majors are the reference group for major comparison. SES = socioeconomic status.

*p < .05, **p < .01, ***p < .001.

Similar relationships also emerge for the 2012 model. Graduates who majored in computer science and engineering were more likely to be in the stable versus the insecure group than business school graduates. In contrast, graduates who majored in biology, general studies, social sciences, humanities, education, and other majors were more likely to be in the insecure versus the stable group compared to business graduates. Graduates who majored in engineering, biology, social sciences, and humanities were also less likely to be in the secure group versus the insecure group. In 2012, family income increased the likelihood of belonging to the stable group versus the insecure group (see Table 7).

Note. Data are from the U.S. Department of Education, National Center for Education Statistics, 2008/12 Baccalaureate and Beyond Longitudinal Study (B&B:08/12). Business majors are the reference group for major comparison. SES = socioeconomic status.

*p < .05, **p < .01, ***p < .001.

Students and Families

Higher education is often presented as a linear pathway to successful socioeconomic outcomes, leading students and families to invest significant tuition dollars (as well as room and board) in pursuit of that outcome (Pope & Fermin, 2003; Roth, 2019). Many students also select majors based on the perceived economic returns of those majors (Carnevale et al., 2021; Eide et al., 2016; Thomas & Zhang, 2005), leading to increased enrollment and emphasis on disciplines such as STEM and business (Dutt-Ballerstadt, 2019). Our results indicate that students who majored in engineering and computer science were more likely to be in the SES secure group in 2009 and the stable group in 2012, indicating that these majors may provide a certain level of not just monetary benefits but also security and socioeconomic well-being. However, students may choose to pursue majors and occupations such as education or humanities because of a sense of purpose or fulfillment associated with such careers (Melguizo & Wolniak, 2012), despite knowing that they may earn less than if they were to go into STEM or business.

An additional nuance to the findings around computer science and engineering is that while graduates were more likely to be in the secure group in 2009, they were also more likely to be in the unemployed group compared to the insecure group. These results suggest a dichotomy for graduates of these majors. Those who were able to find employment achieved relative socioeconomic security; however, they were almost equally likely to be unemployed. Some of this could be attributed to the fact that these students graduated at the height of the great recession—a time when many graduates struggled to find employment (NCES, 2012)—or they could indicate that some graduates generally struggle to find employment within these fields immediately after graduation (NCES, 2012). However, five years after graduation in 2012, graduates from computer science and engineering were more likely to be in the stable group, which presented both high likelihoods of economic security and satisfaction with their job security. A possible explanation for these results could be that five years after graduation, only graduates who are secure within their careers in STEM remain in them since attrition is an issue within STEM careers (Jelks & Crain, 2020).

Administrators and Policymakers

For higher education administrators, our results reinforce existing literature indicating that graduating from higher education is not a guaranteed ticket to socioeconomic success and that graduates may take diverse trajectories following graduation (Arum & Roska, 2014; Vedder et al., 2013). While some students do obtain relative economic stability following graduation, others may struggle. Within these groups, the individual variables provide a more nuanced picture of graduates’ lives. Higher education administrators often tout the benefits of higher education regarding increased access to benefits, less time unemployed, and health outcomes (Clayton & Torpey-Saboe, 2021; Ma et al., 2019; Ma & Pender, 2023). However, our results indicate this is only the case for some students.

Additionally, there were some critical differences based on graduates major. Engineering and computer science students were more likely than business students to have higher levels of socioeconomic outcomes both one and four years after graduation, indicating that these majors may provide more security both in the short- and long-term. More concerning is that many of the other majors resulted in students being more insecure regarding socioeconomic status. These results should not be taken as an indication that students should not major in humanities or social sciences but instead should direct administrators to majors where students may need more support in terms of launching their careers. While business and STEM fields tend to have more structured internship pathways that lead to full-time employment following graduation (Moss-Pech, 2021), students in other majors may find it more challenging to acquire valuable internships. We encourage administrators to continue to invest in resources to support career resources for these majors, as well as work with employers in fields associated with these majors to create more robust internships. Of additional note is that these graduates obtained their bachelor’s degree at the height of the Great Recession, during which many individuals struggled to acquire jobs (NCES, 2012).

As policymakers continue to push for increased accountability for higher education institutions in terms of student outcomes (Kelchen, 2018), using measurements, such as what we have presented in this study, that encompass more than just income becomes more important. When policymakers present income as the primary measure of postsecondary success, they omit the myriad of other benefits associated with attending higher education. A primary focus on income disadvantages disciplines that while vital to society (e.g., education, social services) do not pay as well as other disciplines. While our results indicate that certain majors such as education, social sciences, and humanities are less likely to lead to socioeconomic security than business majors, there was only one measure that assessed how graduates evaluated their jobs (job security satisfaction). Considering other aspects of graduates’ feelings about their jobs in future analyses could further elucidate the non-monetary value associated with these majors, as well as the value to society of such career paths that are not necessarily the highest paying.

Future Research

While the goal of this paper was not to provide the measure of socioeconomic status that all researchers should use, we hope that by presenting the capabilities and potential of LCA in measuring socioeconomic success that other researchers will also be interested in using such methods to build more nuanced conceptualizations. The benefits of this approach may be further bolstered by comparisons to more common econometric models to compare the outcomes. We believe such methods can expand our understanding of not just socioeconomic success but a tool for examining both individual and institutional impacts on this outcome. We hope that other researchers will take these tools and continue to build on and expand upon our conceptualization as they examine postsecondary success.