Talking 'Bout My Generation: Defining “First-Generation College Students” in Higher Education Research
by Robert K. Toutkoushian, Robert A. Stollberg & Kelly A. Slaton - 2018
Background/Context: There have been numerous studies conducted in the higher education literature to determine whether parental education is related to the academic plans and success of their children. Within this literature, particular emphasis is often given to children who are “first-generation college students.” However, researchers and policy makers have not reached agreement on what constitutes a first-generation college student and whether the definition affects the findings from their studies.
Purpose/Objective/Research Question/Focus of Study: In this study, we examined whether the way in which first-generation college status was defined affected its association with the likelihood of a student going to college. We used data from the Educational Longitudinal Study of 2002 (ELS:02), which is a nationally representative longitudinal sample of 10th-grade students in 2002 who were followed up in 2004, 2006, and 2012.
Research Design: We used binary and multinomial logistic regression analysis to examine how first-generation college status, as well as other personal, family, and school characteristics, were associated with whether a student took a college entrance exam, applied to college, and enrolled in college. For this study, we constructed eight different definitions of a first-generation college student. The definitions varied with regard to the level of education needed for a parent to be considered “college educated” and the number of parents meeting the education criteria.
Conclusions/Recommendations: Our results showed that the connection between first-generation college status and these three outcomes varied depending on how first-generation college status was defined. In general, we found larger deficits for first-generation college students when neither parent was college educated and when college educated was defined as earning a bachelor’s degree or higher. First-generation college students faced the largest deficits for enrolling in college, and smaller (but often significant) deficits for taking a college entrance exam and applying to college. The results imply that researchers should be very specific about how they are defining first-generation college status and should determine whether their findings are sensitive to how the variable was defined.
Higher education policy makers and stakeholders have long been interested in finding ways to help more students go to college. For some, it is hoped that higher societal levels of educational attainment will lead to a wide range of financial and social benefits for all (McMahon, 2009). Others approach the topic from an equity perspective, in that education provides a mechanism for individuals to possibly reduce inequalities between selected groups of individuals and mitigate the effects of family background on a persons well-being (Berger, 2000). To help inform these policies, researchers have focused considerable attention on how students make decisions about whether to go to college after high school. Often these studies concentrate on factors that can be adjusted by policy makers, such as financial aid, to entice more students to attend college (Dynarski & Kreisman, 2013; Scott-Clayton, 2015; Toutkoushian & Hillman, 2012).
There is general agreement that there is a strong association between the choices that students make regarding college and the educational attainment of their parents (e.g., Pascarella & Terenzini, 2005; Perna & Titus, 2005). Unsurprisingly, students who come from families with highly educated parents are themselves more likely to be predisposed to go to college and to actually enroll. On the other hand, students who are the first in their families to go to collegefirst-generation college students (FGCS)may face large educational, financial, social, and cultural challenges in both going to and succeeding in college (Cardoza, 2016).
Even though many researchers have used measures of parental education in their studies, we do not yet fully understand how parental education shapes the decisions of students. The lack of understanding is due in part to differences in how researchers represent parental education in their studies. This problem is particularly troublesome when researchers explicitly focus on FGCS because there are many ways that this construct has been and could be defined. In general, a first-generation college student is someone who is the first member of his or her family to go to college. Because data are rarely available on the education levels for everyone in a students family lineage, first-generation status is usually based solely on parental education. It is not clear, however, whether the term parent should apply only to biological parents or be broadened to include stepparents, foster parents, and adoptive parents. The distinction is important because those serving as parental figures play an integral role in the college choice process of their children.
In addition to considering who qualifies as a parent, different levels of parental education are used to define FGCS. Whether a parent has gone to college can vary by completion status (e.g., enrolling vs. earning a degree) as well as the type of institution attended (two-year or four-year). Definitions of first-generation status may also vary by the number of parents who have postsecondary experience. Some researchers, including Nunez and Cuccaro-Alamin (1998) and Choy (2001), identified someone as a FGCS if neither parent enrolled in any postsecondary institution. Other researchers (Collier & Morgan, 2008; Pike & Kuh, 2005) only referred to someone as a FGCS if neither parent earned a bachelors degree.
Often the definition of FGCS used by researchers depends on the way in which information on parental education was collected. For example, a survey of ninth-graders in the state of Indiana asked students, Has one or both of your parents or guardians graduated from a four-year college or university? (Toutkoushian, Hossler, DesJardins, McCall, & Canche, 2015). In this survey, guardians are counted as parents, and college experience is defined as graduating from a four-year institution. Even when more detailed survey questions are used, surveying students about their parents education can yield missing and inaccurate data because respondents may not always know the precise education levels of each of their parents.
Understanding how best to define first-generation college status is important because there are numerous programs throughout the United States aimed at helping FGCS matriculate and succeed in postsecondary education. For instance, The Bill and Melinda Gates Foundation (2012), Complete College America (2014), the Lumina Foundation, and the federal government have called for significant increases in the proportion of citizens with some form of postsecondary credential. Likewise, a number of states are tying institutional funding to the numbers of students earning postsecondary degrees (Burke, 2005). Because FGCS are less likely to both enroll in and graduate from college, policy makers often focus on programs for FGCS to help achieve broad college completion goals for the nation. The connection between parental education and postsecondary decisions will be particularly relevant as the demographic composition of the U.S. population continues to shift toward racial/ethnic groups such as Hispanics, whose parental education levels on average have been relatively low (Vargas & Conlon, 2011).1 It is thus necessary to define who among these students will be counted as FGCS because their aspirations and college decisions may be different from their non-first-generation peers, and having a standard definition for first-generation status will be important for understanding both how best to serve these students and the impact that postsecondary education is having on a national scale.
In this study, we use longitudinal data from ELS to examine a cohort of approximately 7,300 tenth graders and determine in more detail whether the way in which researchers identify FGCS affects their postsecondary plans. We estimate a series of binomial and multinomial logistic regression models in which the dependent variables include whether a student planned on taking the SAT or ACT, applied to college, or enrolled in a two-year or four-year postsecondary institution. In particular, we focus on eight alternative definitions of first-generation college status depending on the level of parental education and the number of parents meeting these criteria, and whether our findings were affected by who is counted as a parent in the calculation.
Because there are varied reasons behind a students choice whether to go to college, it is important to consider framing decisions about college from multiple perspectives. Accordingly, we rely on the college choice model, human capital theory, and social and cultural capital theory to guide this study. Human, social, and cultural capital each contribute to and have direct impacts on the college choice process. Human capital theory (Becker, 1975) argues that, because earnings and educational attainment are related, families with highly educated parents on average have the financial resources necessary to pay for their childrens education. Similarly, students with college-educated parents may themselves perform better in school, which in turn contributes to their educational aspirations and enrollment. Individuals invest in themselves through acquiring new knowledge and skills in college. As these attributes accumulate, an individual builds human capital that increases his or her value in the workplace, eventually leading to higher compensation (Becker, 1975; Mincer, 1958; Schultz, 1961). As a result, human capital theory helps explain why students make the decision to attend college.
Social and cultural capital theories (Bourdieu, 1986; Lin, 2002) suggest that students with college-educated parents are more likely to learn about college from them and thus in turn may consider going to college themselves. Students with college-educated parents may also have more interactions with other adults who have gone to college, which could provide them with additional information and role models for forming their own educational aspirations (OConnor, Hammack, & Scott, 2010; Perna, 2006; Sandefur, Meier, & Campbell, 2006). The delineation between cultural and social capital is frequently blurred in the literature on FGCS as befitting interrelated concepts. Berger (2000), in acknowledging Bourdieu (1986) and McDonough (1997), defined cultural capital as a symbolic, rather than material, resource . . . it is a type of knowledge that members of the upper class value but is not taught in schools (p. 98). Social capital, however, consists of the extrafamilial networks accessible by students and parents (Bourdieu, 1986; Coleman, 1988), the cumulative cultural capital of which can be leveraged for assistance or gain in a transactional manner (Wells, 2008). Cultural and social capital theorists assert that those with greater cultural capital, or nonfinancial social assets, achieve greater success in life (Bourdieu, 1977). Despite the additional cultural capital that education affords, FGCS would theoretically, on average, lag behind their non-FGCS peers because of the initial deficit. When both parents have gone to college, they may also have more total social and cultural capital between them that can be used to help their children to go to college (Dumais & Ward, 2010; Prospero & Vohra-Gupta, 2007; Wells, Seifert, Padgett, Park, & Umbach, 2011). Work by Nichols and Islas (2016) shows that parents of FGCS tend to push their students through college with support, whereas the parents of non-first-generation students pull their students through college and share social capital that the parents of FGCS do not have. In similar work, Armstrong and Hamilton (2013) explained that parents social networks deeply impact the access that their children have to resources and people with firsthand knowledge of the college-going experience.
Other theoretical insights into the postsecondary decisions of students can be gained through the work of Don Hossler and colleagues (Hossler, Braxton, & Coopersmith, 1989; Hossler & Gallagher, 1987; Hossler, Schmit, & Vesper, 1999) and others (Paulsen, 1990; Perna, 2006) who have argued that the college choice process involves at least three major phases: predisposition, search, and choice. Although human, social, and cultural capital theories help inform each of these stages, they may have different impacts at each stage. For example, human capital theory may play a more prominent role in the choice stage than in earlier stages because it is at this juncture when students and their families have more precise information about costs and whether they can afford specific institutions in their choice sets.
There is a long history of examining the connection between parental education and childrens educational attainment. Not surprisingly, most of these studies have found strong positive associations between parental education and student aspirations and enrollments. Researchers have used several approaches to measure parental education over time. Some have represented parental education with continuous measures (years of education for mother and/or father), or dummy variables for highest level of education completed (HS, AA, BA, etc.) alone or combined with years of education to measure sheepskin effects (Bitzan, 2009). For example, Card (1993) focused on two continuous variables for the years of parental education, while other studies, like in Dubow, Boxer, and Huesman (2009), utilized a single composite variable with values for different degree levels attained by parents (such as 1 = high school, 2 = some college, 3 = associates degree, and so on). Although these approaches are parsimonious, they restrict each increment in parental education to have the same impact on students. A more flexible approach to measuring parental education is to use multiple dichotomous variables for different levels of parental education (e.g., Paulsen & St. John, 2002).
Others have considered whether children who are the first in their families to go to college make the same choices as students with college-educated parents. All three theoretical frameworks would predict that FGCS are at a disadvantage relative to their peers, and most studies have found that, in fact, FGCS are less likely than non-FGCS to go to college or take steps to prepare for college (Chen & Carroll, 2005; Choy, 2001; Ishitani, 2006; Lohfink & Paulsen, 2005; Padgett, Johnson, & Pascarella, 2012; Pascarella, Pierson, Wolniak, & Terenzini, 2004; Ward, Siegel, & Davenport, 2012).
However, the way in which FGCS status is determined varies across studies. As noted by ACT (2013), The term first-generation student is defined differently by a number of organizations, often differing in the extent of exposure to postsecondary education (e.g., enrolled, attended, or completed) as experienced by disparate combinations of parent/guardian arrangements (e.g., highest extent of exposure for one parent/guardian or both parents/guardians) (p. 17). As documented by University Institutional Research and Reporting (2012), for example, the National Center for Education Statistics (NCES) has identified FGCS as those whose parents have not attended college (Office of Educational Research and Improvement, 1998), and the College Board has labeled students as first generation when their parents have not earned a two-year degree (Dougherty et al., 2007). Likewise, with parental education being a crucial component of FGCS, it is important to define what is meant by parent. Although some studies specify that parents are only the biological mother and father of the student (Astin, 1964; Wells et al., 2011), and others construct survey questions related to parents/guardians (Saenz, 2007), most do not specify precisely what is meant by this term.
Previous studies measuring FGCS status have done so using various methods. First-generation status could be based on whether neither parent graduated from college, neither parent attended college, both parents did not attend/graduate from college, or one specific parent (e.g., mother) has done so. The most common measure of FGCS status in early studies was when neither parent attended any form of postsecondary education (e.g., Astin, 1964; Billson & Terry, 1982). More recently, the majority of studies tend to define first-generation status as one or both parents having not graduated from college (Inkelas, Daver, Vogt, & Leonard, 2007; Ishitani, 2003, 2006; Padgett et al., 2012; Prospero & Vohra-Gupta, 2007; Wells et al., 2011). Likewise, there is still no consensus among researchers on whether college is defined as a two- or four-year institution.
Researchers often rely on surveys to collect data related to first-generation status, but survey data are not always accurate. Students may not know their parents highest level of education or may give false information, which could make the data inaccurate. Also, survey instruments rarely define what is meant by parent, so there is often not a distinction between the various types of families represented in the survey, which limits the conclusions that can be made with the data.
DATA AND METHODOLOGY
For this study, we relied on data from the Education Longitudinal Study (ELS) of 2002. ELS is a nationally representative sample of more than 16,000 tenth-grade students who were surveyed in 2002 and followed up with in 2004, 2006, and 2012. The high response rates to the follow-up surveys (~89%) ensured that there was a sufficient number of students in the sample to conduct the statistical analyses for our particular study. To focus attention on the role of parental education on students, we restricted our sample to 10th-grade students who were living full time with two parents (defined as either biological, step, adopted, or foster parents), with parents reporting their educational attainment. After deleting cases with missing data on the dependent variables, the final weighted sample used in our study consisted of approximately 7,300 students (all sample sizes are rounded per NCES requirements).
There are a number of reasons that ELS is an ideal data set for the purpose of our study. First, the data are nationally representative, and therefore, the findings can be applied to the larger set of students in the United States. Second, the survey collected information from students at the predisposition, search, and choice stages of their college careers. Thus, the data enable us to determine how parental education was related to the initial postsecondary aspirations of students as well as whether they subsequently enrolled in college. Third, ELS collected information about parental education from the parents and not the students, providing a more reliable measure of parental educational attainment. Finally, the ELS survey instrument asked each parental respondent to identify their relationship to the student and their level of education. The available options were biological parent, adoptive parent, stepparent, foster parent, grandparent, partner, other relative, or other guardian.
We now provide information on the dependent and explanatory variables used in our study. More details on how we constructed the variables for this study can be found in the Appendix.
We focused on four dependent variables. The first was whether a student in 10th grade indicated that he or she had taken or planned on taking the SAT or ACT during high school (SAT). In most states, the act of taking either entrance exam was a good early indicator of a students predisposition for going to college.2 The second dependent variable was whether the student applied to at least one postsecondary institution (Apply). The third dependent variable that we considered was whether a student enrolled in a two-year or four-year postsecondary institution (Enroll). Finally, the last dependent variable (Enroll024) separated the college enrollment decision into three categories: enrolled in a two-year institution, enrolled in a four-year institution, or did not enroll in college.
Variables for First-Generation College Status
We constructed eight different measures to identify FGCS (P). The variations in these measures depended on two factors: the level of parental education required for a student to be counted as a FGCS, and the number of parents who met the education criteria. We considered four different levels of educational attainment that could be used to define FGCS: (1) parent(s) have at most a high school degree; (2) parent(s) have at most started (but not completed) an associates degree; (3) parent(s) have at most completed an associates degree; and (4) parent(s) have at most completed an associates degree or started (but not completed) a bachelors degree. For each of these four categories, we created three variables to represent the number of parents who met the designated education criteria.
Other Explanatory Variables
The remaining explanatory variables fell into four general categories. The first group of variables (D) represented student-level characteristics, such as their gender, race, and ethnicity. Student characteristics also included the students grade point average in Grade 9 and the students score on standardized tests in mathematics and reading collected by ELS.3 The second group (F) denoted family measures that may have affected a students demand for higher education, such as family income (5 levels), number of dependents, whether they lived with both biological parents, and number of siblings. The third group of explanatory variables (S) represented high school characteristics, including enrollments in Grade 10, geographic region, whether located in an urban, suburban, or rural area, whether public, percent students on free lunch, and percent students taking AP courses. Finally, the last set of explanatory variables (I) included measures of parental involvement with their childrens education to capture social and cultural capital effects on students. We created dummy variables for this last category from questions in ELS:02 pertaining to whether students often or sometimes had discussions with parents about (1) high school courses, (2) high school activities, (3) high school grades, (4) taking the SAT or ACT, and (5) going to college. To retain more students in the sample, we created dummy variables for students with missing data on gender, race, ethnicity, family income, number of siblings and dependents, percent students on free lunch or in AP courses, and parental involvement questions.
For each of the dependent variables, we estimated models in which we focused on the relationship between FGCS and the outcome variable after taking into account these other control variables. Because each of the first two dependent variables was binary, we used logistic regression analysis to estimate the key parameters of the model. For the model in which the dependent variable was the type of institution attended, we used multinomial logistic regression. In all models, we converted the logistic coefficients to marginal effects, weighted the data to take into account the two-stage stratified sampling design used by NCES, and clustered the standard errors at the school level to adjust for the nonindependence of students in the same school.
Specifically, we first estimated the following five equations:
where P0 = 1 if neither parent is college educated; P1 = 1 if only one parent is college educated; D, F, S, and I are sets of variables as defined earlier; SAT = 1 if student has taken or planned on taking the SAT or ACT as of Grade 10; Apply = 1 if student applied to any college; Enroll = 1 if student enrolled in any college; are parameters to be estimated; and = random error term. A student would then be defined as a FGCS when neither parent was college educated (i.e., P0 = 1) or when at least one parent was not college educated (i.e., P0 = 1 or P1 = 1). These equations allowed us to see whether the difference in college-going rates of FGCS versus non-FGCS was reduced once we controlled for parental involvement in their childrens education, and whether the effect of parental education depends on how many parents were college educated. We also estimated conditional logit models for the decision to apply to college (Equation 3), where the sample was restricted to those students who took the SAT, and for the decision to enroll in college (Equation 5), where the sample was restricted to only those who applied to college. The conditional logit models enabled us to determine whether FGCS status still had an effect on these stages of higher education demand even when we looked only at those students who were predisposed to go to college.
Finally, for the last dependent variable pertaining to type of institution in which the student enrolled, we used multinomial logistic regression analysis to estimate the parameters of the model.
The purpose of this model was to see if FGCS had a differential effect on a students decision to enroll in a two-year or a four-year institution, as opposed to enrolling in any postsecondary institution.
In Table 1 we provide descriptive statistics for the variables in our study. Focusing on the dependent variables, we found that about three quarters of the students in our sample indicated that they had taken or were planning to take the SAT or ACT exam at some time during high school. We observed that the vast majority of students (86%) applied to at least one postsecondary institution and that 82% enrolled in a postsecondary institution, with about one third of these students selecting a two-year institution.
Table 1. Descriptive Statistics for Variables Used in the Study
Notes: Sample size is approximately 7,300 (rounded per NCES requirements). Complete variable definitions are contained in the Appendix.
In Figure 1, we show the breakdown of students by four different parental education levels and the number of parents in each education level. For example, the first set of bars on the left-hand side of the figure reveal that 53% of students had two parents with at least some college education (defined as any postsecondary education), 25% had only one parent with some college education, and 22% had no parents with college education. In contrast, on the far right-hand side of the figure, we show that 23% of students had two parents with a bachelors degree, another 23% of students had exactly one parent with a bachelors degree, and 54% had no parent with a bachelors degree. Accordingly, when FGCS was defined as neither parent being college educated, between 22% and 54% of students would qualify. If we instead defined FGCS to include those with at least one non-college-educated parent, then the percentage of FGCS varied from 47% to 77% (summing the light gray and white bars in Figure 1). The fact that the size of the FGCS population varies considerably by definition is important for policy makers, educators, and governments to know as they develop support programs for FGCS.
Figure 1. Breakdown of Students by Parental Education
Notes: Sample size is approximately 7,300 (rounded per NCES requirements). Light gray bars: Neither parent is college educated. White bars: Only one parent is college educated. Dark gray bars: Both parents are college educated. Any College: Parent(s) enrolled in any postsecondary education. AA Degree: Parent(s) earned an associates degree or attended a four-year institution. Some BA: Parent(s) enrolled in a four-year institution. BA Degree: Parent(s) earned a bachelors degree.
To begin our focus on the relationship between parental education and their childs plans for college, we provide breakdowns of our dependent variables by parental education in Table 2. To do this in a parsimonious fashion, we only considered the subset of students for whom both parents had the same level of education (roughly 60% of our sample). Overall, the data show that as parental education increased, students were more likely to plan on taking the SAT or ACT, apply to college, and enroll in college. The gaps were particularly large for students whose parents had not gone beyond a high school education. The last two rows illustrate that the type of institution attended also varied considerably by parental education. Students whose parents had less than a bachelors degree were equally likely to enroll in either a two- or four-year institution, whereas students with parents who had a bachelors or graduate degree were much more likely to select a four-year institution.
Table 2. College Predisposition, Application, and Enrollment by Parental Education
Notes: Data only include students whose parents have the same level of education. Sample size is approximately 3,700 (rounded per NCES requirements). High school: Highest degree completed is a high school diploma or equivalent. Associates degree: Both parents report starting or completing an associates degree as their highest level of education. Bachelors degree: Both parents report starting or completing a bachelors degree as their highest level of education. Graduate degree: Both parents report starting or completing a graduate degree. Means are statistically different across parental education categories at the 1% level.
Next, we examined how the means for the dependent variables in our study differed between FGCS and non-FGCS using each of the eight alternative definitions. The results are shown in Table 3. We focused on the same four categories of college-educated parent, ranging from having any college experience to earning at least a bachelors degree. Within each category, we calculated the means for students broken down by the number of college-educated parents. For example, the first row shows that for students with two parents who have some college experience, 85% of them took the SAT or ACT, 93% of them applied to college, and 92% enrolled in college. The two possible definitions of FGCS for each category correspond to the last two rows within each category. The table lets us show how the means for FGCS compared with students with two college-educated parents, how the means differed by definition of college-educated parent, and how students with exactly one college-educated parent compared on average to those with two or no college-educated parents.
Table 3. Dependent Variables Means by First-Generation College Status
Notes: Sample size is approximately 7,300. Values in the table show the percentage of students in each category broken down by the number of college-educated parents in their family. Any College: Parent(s) enrolled in any postsecondary institution. AA Degree: Parent(s) earned an associates degree or attended a four-year institution. Some BA: Parent(s) enrolled in a four-year institution. BA Degree: Parent(s) earned a bachelors degree.
The first three columns of means show that as the definition of college-educated parent became more restrictive, the averages of the dependent variables for all groups increased. Similarly, we observed a shift in the means toward enrolling in a four-year institution as the definition of a college-educated parent increased. The differences in the means between students with two college-educated parents and those with no (or fewer than two) college-educated parents varied by definition as well, though there was no uniform pattern to these differences.
In Table 4, we focus on how the first-generation status of students, as well as the other explanatory variables, related to a students plans for taking the SAT or ACT during high school, applying to college, and enrolling in college. We only present the complete findings for the case in which a college-educated parent is someone who has any postsecondary education; however, we found very similar results for the nonparental education variables when we used alternative measures of college-educated parents.
Table 4. Marginal Effects of FGCS on SAT/ACT Taking, and Applying to and Enrolling in College
Notes: Sample size is approximately 7,300 for Models 1, 2, and 4; approximately 5,500 for Model 3; and approximately 6,300 for Model 5. Figures in the table represent average marginal effects. Standard errors are shown in parentheses below marginal effects and are clustered at the school level. Data are weighted using survey weights for participation in 10th grade. Reference category for parental education is both parents attended a postsecondary institution. Reference category for race is White. Reference category for income is $50K to $100K. Models also include variables for missing income, ethnicity, siblings, dependents, school FLP, school AP, and race.
+p < .10. *p < .05. **p < .01. ***p < .001.
Beginning with the two main variables of interest for our study, students with no college-educated parents were 4.4% less likely than students with two college-educated parents (reference group) to plan on taking the SAT or ACT after taking into account selected personal, family, and school characteristics, as well as measures of parental involvement in their education. Similarly, these same students were 9% less likely to apply to any college and 9.6% less likely to enroll in any college, holding constant the same control measures. Accordingly, two thirds to three quarters of the mean differences in these dependent variables were explained by the other factors in the models, and the remaining gaps were statistically significant at the 1% level or higher. When we only looked at those students who indicated that they had taken or will take the SAT or ACT (column 3), students with no college-educated parents were still 7.2% less likely to apply to college. Likewise, out of the subset of students who applied to college, students with no college-educated parents were 4.5% less likely to enroll in college. With regard to students with exactly one college-educated parent, they were as likely as those with two college-educated parents to take the SAT or ACT but were about 4% less likely to apply to college, and 4% to 6% less likely to enroll in college. Therefore, these students were still at a disadvantage when compared with students with two college-educated parents, but their disadvantages were not as great as for those with no college-educated parents.
Among the key results for the variables not related to parental education are the following: Females were more likely than males to plan on taking the SAT/ACT, apply to college, and enroll in college. We found that Black and Asian students were more likely than White students to prepare for and enroll in college. Not surprisingly, there were strong positive relationships between a students aptitude and academic performance and their likelihood of taking the SAT, applying to college, and enrolling in college. Family income was found to have a positive effect on a students probability of applying to and enrolling in college but did not have a similar effect on taking the SAT or ACT. And students living with both biological parents were 3% to 4% more likely to apply to and enroll in college, although the effect on enrollment became insignificant in the conditional logit model for only college applicants.
Turning to school-level variables, we observed that students in public schools were less likely than their counterparts in private schools to apply to and enroll in college, and students in the Mid-Atlantic and East North Central states were more likely than those in the Pacific states to apply to and enroll in college. Finally, of the five measures of parental involvement in their childs education that we considered, the largest association was found for students who indicated that they discussed college with their parents.
In Table 5, we summarize the findings for the parental education variables after estimating Equations 1, 2, and 4 (as shown in Table 4) for each alternative definition of college-educated parents. We report the estimated coefficients for parental education broken down by the number of college-educated parents for each definition. Recall that a student could be classified as a FGCS when neither parent is college educated, or when at least one parent is not college educated. Comparing the coefficient estimates in the top four rows with the bottom four rows, it can be seen that the estimated disadvantage for FGCS varied by the way in which it was defined. For the outcome of applying to college, for example, FGCS were between 4.8% and 9.0% less likely than students with two college-educated parents to apply. Likewise, the negative effect for FGCS enrolling in college ranged from a low of 7.9% to a high of 14.3%. With regard to whether a student takes the SAT or ACT, not only were FGCS less likely to do so, but the difference was not statistically significant for three of the eight estimates. Students with exactly one college-educated parent were as likely as non-FGCS to take the SAT or ACT, less likely to apply to college in two of the four definitions, and less likely to enroll in college across all four definitions. In each case, the deficits faced by students with one college-educated parent were smaller than those with no college-educated parents regardless of how a college-educated parent was defined. Reading across the parental education definitions, however, we could discern no general pattern in the magnitude of the FGCS deficiencies.
Table 5. Effects of Definitions of First-Generation College Student on Selected Outcomes
Notes: Marginal effects for the first-generation college student variables. Standard errors are in parentheses and are clustered at the school level. Data are weighted using survey weights for participation in 10th grade. Each model also controls for gender, race, GPA, standardized test score, number of siblings and dependents, family status and income, Grade 10 enrollments, school attributes, and measures of parental involvement in childs education. a Parents include biological, step, adopted, and foster parents (n ~ 7,300). b Education level of a college-educated parent: Any College: Parent(s) attended any postsecondary institution; AA Degree: Parent(s) earned an associates degree or attended a four-year institution; Some BA: Parent(s) enrolled in a four-year institution (does not include enrolling in or completing an associates degree); BA Degree: Parent(s) earned a bachelors degree. +p < .10. *p < .05. **p < .01. ***p < .001.
In Table 6, we present the results from the multinomial logistic model in which the dependent variable was whether a student either enrolled in a 2-year college, enrolled in a 4-year college, or did not enroll in college. As in Table 4, we only show the complete regression results for the case in which college-going parents are defined as those with any postsecondary experience.
Table 6. Marginal Effects of FGCS on Type of College Attended
Notes: Sample size is approximately 7,300. Figures in the table represent average marginal effects. Standard errors are shown in parentheses below marginal effects and are clustered at the school level. Data are weighted using survey weights for participation in 10th grade. Reference category for parental education is both parents enrolled in a postsecondary institution. Reference category for race is white. Reference category for income is $50K to $100K. Models also include variables for missing income, ethnicity, siblings, dependents, school FLP, school AP, and race. +p < .10. *p < .05. **p < .01. ***p < .001.
Overall, the results reveal that students with no college-educated parents were 9% less likely than students with two college-educated parents to enroll in a four-year institution. Similarly, students with one college-educated parent faced a 5.8% deficit in likelihood of enrollment at a four-year institution. However, we found no differences among these three groups of students in their likelihood of enrolling in a two-year institution. Accordingly, the negative effects that we reported for FGCS in Table 4 for enrolling in any college were concentrated in four-year institutions. We also found that students with higher academic performance, living with both biological parents, having higher family income, and who reported having more discussions with their parents about school activities and college were more likely to enroll in four-year institutions. Curiously, the data revealed large differences in the enrollment patterns by race, with Asian students being 10% more likely and Black students being 20% more likely than White students with similar attributes to enroll in four-year institutions.
Finally, in Table 7, we report the marginal effects for alternative definitions of the parental education variables in the multinomial logistic regression models. Relative to students with two college-educated parents, students with one or no college-educated parents were significantly less likely to enroll in a four-year institution regardless of how college-educated parent was defined. The marginal effects, however, fluctuated substantially across the various models, ranging from a low of about 9% to a high of more than 15% for those with no college-educated parents. As before, students with one college-educated parent were less likely than students with no college-educated parent to enroll in a four-year institution.
Table 7. Effect of First-Generation College Student on Type of Institution Enrolling
Notes: Dependent variable: Enrollment by type of institution. Marginal effects are shown for the first-generation college student variables. Standard errors are in parentheses and are clustered at the school level. Data are weighted using survey weights for participation in 10th grade. Each model also controls for gender, race, GPA, standardized test score, number of siblings and dependents, family status and income, Grade 10 enrollments, school attributes, and measures of parental involvement in childs education. a Parents include biological, step, adopted, and foster parents (n ~ 7,300); b Education level of a college-educated parent: Any College = Parent(s) attended any postsecondary institution; AA Degree = Parent(s) earned an associates degree or enrolled in a 4-year institution; Some BA = Parent(s) enrolled in a 4-year institution (does not include enrolling in or completing an associates degree); BA Degree = Parent(s) earned a bachelors degree. + p <.10, * p<.05, ** p<.01, *** p<.001.
SUMMARY AND DISCUSSION
Policy makers and higher education researchers have long been interested in examining the experiences and outcomes of students who are traditionally underrepresented in colleges and universities. Particular attention has been paid to students who are the first in their families to go to college; however, there is no general agreement as to who qualifies as a first-generation college student. In this study, we used data from a nationally representative sample of students to determine whether the way in which researchers define first-generation college status affects its association with the postsecondary aspirations and actions of students. We considered eight alternative definitions of FGCS and found that the percentage of students identified as first-generation varied substantially across definitions. Thus, how many students are counted as a first-generation college student can be greatly affected by the particulars of how one defines it. This result alone has important implications for postsecondary institutions and government agencies to consider as they design programs to help FGCS succeed in academia, because the cost of implementing such programs will naturally vary with the size of the targeted group. Our results suggest that broadening the definition of college-educated parent from any college to bachelors degree may result in a doubling or tripling of the number of students who would be eligible to use such assistance programs.
Turning to the statistical models, we found that the marginal effect of FGCS on the outcomes examined varied depending on how FGCS was defined. No general pattern emerged with regard to differences in the magnitudes of the marginal effects by the level of education used to define a college-educated parent. The models consistently showed that students with no college-educated parents faced larger college deficits than those with one college-educated parent, and both groups were less likely to apply to and enroll in college than those with two college-educated parents. Accordingly, focusing assistance efforts for only those students with no college-educated parents would overlook a number of other students who still face large hurdles in going to college.
Across models, we found that two thirds to three quarters of the average differences in outcomes between students with one or more non-college-educated parent and those with two college-educated parents were attributed to the control variables that we considered. Not surprisingly, student aptitude and academic performance had strong connections to college predisposition and enrollment. Students who were the first in their families to go to college were in general less likely to apply to and enroll in college after controlling for student performance, family income, school characteristics, and parental involvement in their education. In addition, the conditional logit models showed that FGCS faced deficits even among the subset of students who had taken the SAT or ACT or applied to college. These findings suggest that the disadvantage for FGCS is inadequately explained by the measures of financial resources and social/cultural capital with regard to education present in the models. Accordingly, human, social, and cultural capital play a part in, but do not fully explain, the differences in postsecondary decisions for FGCS and non-FGCS.
Taken together, then, what do our results say about how policy makers and researchers should define first-generation college status? If the researcher is interested in simply determining whether FGCS and non-FGCS have similar predispositions and likelihoods of going to college, then in most instances, it does not appear to matter how a college-going parent is defined. On the other hand, if the size of the deficit between FGCS and non-FGCS students is of importance, then our findings showed that the way in which a college-educated parent is defined matters. Likewise, students with one college-educated parent are also in need of support programs to help them apply to and enroll in college at rates similar to their peers who have two college-educated parents.
A remaining important question for researchers is: Why are first-generation college studentsregardless of how they are measuredless likely than their peers to want to go to college and to eventually enroll, and how does this relate to the way in which FGCS status should be measured? The act of having enrolled in college at any level may help parents understand and transmit to their children what is needed to pursue a college education, and thus simply going to college may be an important defining line between students. Or perhaps it is only when parents earn a college degree that they understand the financial and nonfinancial benefits of college and impart them to their children. In each of these instances, the effect of parental education on children is driven by the social capital and information that parents convey to their children that then translate into observed behavior. Alternatively, it could be that parents with college experience have inherent advantages that are used to help enable their children to go to college. The fact that gaps in predisposition, search, and enrollment remain after controlling for the available measures of financial resources and parental involvement suggests that there is more driving the differences than would be suggested by human, social, and cultural capital theories. The results may also be due to the fact that the financial and parental support variables we used in our study are but crude approximations of the true constructs of importance.
Finally, our study offers some recommendations for how to collect data on parental education. When asking questions of students about their parents, researchers should be very clear about definitions, such as who should be considered a parent and what is meant by college-educated parent. The best advice for researchers who are developing survey instruments to collect information on parental education is to include a larger menu of alternatives for students and parents. In this way, researchers can easily test the sensitivity of their findings to the specific way in which first-generation college status is defined in their work. It also has the potential to allow researchers to examine how the effects of FGCS status change when particular factors are controlled for in statistical models.
Research is supported by funding from the National Science Foundation and the Association for Institutional Research grant no. RG14-5499. Earlier versions of this article were presented at the meetings of the Association for Institutional Research (Denver, CO, May 2015), the Association for the Study of Higher Education (Denver, CO, November 2015), and the Southern Economic Association (New Orleans, LA, November 2015). We would like to thank Stephen Quaye, Manu Raghav, Jason Lee, and participants in these meetings for their helpful comments on prior drafts of the article.
1. According to the U.S. Census Bureau, in 2014, the percentage of adults age 25 and over with some college experience was 63% for non-Hispanic Whites and 37% for Hispanics (see Table 3 at http://www.census.gov/hhes/ socdemo/education/data/cps/2014/tables.html). Population projections from the U.S. Census Bureau show that the U.S. population that is Hispanic is projected to rise from 18% in 2015 to 29% in 2060 (see Table 10 at http://www.census.gov/population/projections/data/national/2014/summarytables.html).
2. Several states required all students in high school to take either the ACT or SAT, and thus, taking a college entrance exam in these states does not necessarily provide information about the students interest in going to college.
3. The specific scores were based on mathematics and reading tests administered by the Program for International Student Assessment (PISA) in 2003. The PISA scores were linked to ELS:02 through a concordance procedure.
ACT. (2013). The condition of college & career readiness 2013First-generation students. Retrieved from http://www.act.org/newsroom/data/2013/states/pdf/FirstGeneration.pdf
Armstrong, E., & Hamilton, L. (2013). Paying for the party: How college maintains inequality. Cambridge, MA: Harvard University Press.
Astin, A. (1964). Personal and environmental factors associated with college dropouts among high aptitude students. Journal of Educational Psychology, 55, 219227.
Becker, G. (1975). Human capital: A theoretical and empirical analysis, with special reference to education. Chicago, IL: University of Chicago Press.
Berger, J. (2000). Optimizing capital, social reproduction, and undergraduate persistence. In J. M. Braxton (Ed.), Reworking the student deaprture puzzle (pp. 195224). Nashville, TN: Vanderbilt University Press.
Bill and Melinda Gates Foundation. (2012). New grants to reimagine financial aid & help increase postsecondary access, success and completion. Retrieved from http://www.gatesfoundation.org/media-center/press-releases/2012/09/postsecondary-financial-aid-grants-announcement
Billson, J., & Terry, M. (1982). In search of the silken purse: Factors in attrition among first-generation students. College and University, 58, 5775.
Bitzan, J. (2009). Do sheepskin effects help explain racial earnings differences? Economics of Education Review, 28, 759766.
Bourdieu, P. (1977). Cultural reproduction and social reproduction. In J. Karabel & A. Halsey (Eds.), Power and ideology in education (pp. 487511). Oxford, England: Oxford University Press.
Bourdieu, P. (1986). The forms of capital. In J. Richardson (Ed.), Handbook of theory and research for the sociology of education (pp. 241258). New York, NY: Greenwood.
Burke, J. (2005). Reinventing accountability: From bureaucratic rules to performance results. In J. Burke (Ed.), Achieving accountability in higher education (pp. 216245). San Francisco, CA: Wiley.
Card, D. (1993). Using geographic variation in college proximity to estimate the return to schooling (NBER Working Paper No. 4483). Cambridge, MA: National Bureau of Economic Research.
Cardoza, K. (2016, January 20). First-generation college students are not succeeding in college, and money isnt the problem. The Washington Post. Retrieved from https://www.washingtonpost.com/posteverything/wp/2016/01/20/first-generation-college-students-are-not-succeeding-in-college-and-money-isnt-the-problem/
Chen, X., & Carroll, C. (2005). First-generation students in postsecondary education: A look at their college transcripts. Postsecondary education descriptive analysis report (NCES 2005-171). Washington, DC: National Center for Education Statistics.
Choy, S. (2001). Students whose parents did not go to college: Postsecondary access, persistence, and attainment. The Condition of Education 2001, 1843.
Coleman, J. (1988). Social capital in the creation of human capital. American Journal of Sociology, 94, S95S120.
Collier, P., & Morgan, D. (2008). Is that paper really due today?: Differences in first-generation and traditional college students understandings of faculty expectations. Higher Education, 55, 425446.
Complete College America. (2014). Alliance of the states: Leading the college completion movement. Retrieved from http://completecollege.org/the-alliance-of-states/
Dougherty, P., Gernand, R., Pollock, C., & Wiley, A. (2007, June). First-generation students in the 2006 SAT cohort. Paper presented at the annual forum of the Association for Institutional Research, Kansas City, MO.
Dubow, E., Boxer, P., & Huesmann, L. (2009). Long-term effects of parents education on childrens educational and occupational success: Mediation by family interactions, child aggression, and teenage aspirations. Merrill-Palmer Quarterly, 55, 224249.
Dumais, S., & Ward, A. (2010). Cultural capital and first-generation college success. Poetics, 38, 245265.
Dynarski, S., & Kreisman, D. (2013). Loans for educational opportunity: Making borrowing work for todays students (Discussion Paper 2013-05). The Hamilton Project. Retrieved from http://www.brookings.edu/~/media/research/files/papers/2013/10/21-student-loans-dynarski/thp_dynarskidiscpaper_final.pdf
Hossler, D., Braxton, J., & Coopersmith, G. (1989). Understanding student college choice. In J. Smart (Ed.), Higher education: Handbook of theory and research (Vol. 5, pp. 231288). New York, NY: Agathon Press.
Hossler, D., & Gallagher, K. (1987). Studying student college choice: A three-phase model and the implications for policymakers. College and University, 62, 207221.
Hossler, D., Schmit, J., & Vesper, N. (1999). Going to college: How social, economic, and educational factors influence the decisions students make. Baltimore, MD: Johns Hopkins University Press.
Inkelas, K., Daver, Z., Vogt, K., & Leonard, J. (2007). Livinglearning programs and first-generation college students academic and social transition to college. Research in Higher Education, 48, 403434.
Ishitani, T. (2003). A longitudinal approach to assessing attrition behavior among first-generation students: Time-varying effects of pre-college characteristics. Research in Higher Education, 44, 433449.
Ishitani, T. (2006). Studying attrition and degree completion behavior among first-generation college students in the United States. Journal of Higher Education, 77, 861885.
Lin, N. (2002). Social capital: A theory of social structure and action (Vol. 19): Cambridge, England: Cambridge University Press.
Lohfink, M., & Paulsen, M. (2005). Comparing the determinants of persistence for first-generation and continuing-generation students. Journal of College Student Development, 46, 409428.
McDonough, P. (1997). Choosing colleges: How social class and schools structure opportunity. Albany: State University of New York Press.
McMahon, W. (2009). Higher learning, greater good: The private and social benefits of higher education. Baltimore, MD: Johns Hopkins University Press.
Mincer, J. (1958). Investment in human capital and personal income distribution. Journal of Political Economy, 66, 281302.
Nichols, L., & Islas, Á. (2016). Pushing and pulling emerging adults through college: College generational status and the influence of parents and others in the first year. Journal of Adolescent Research, 31, 5995.
Nunez, A., & Cuccaro-Alamin, S. (1998). First-generation students: Undergraduates whose parents never enrolled in postsecondary education (NCES 98-082). Washington, DC: National Center for Education Statistics, U.S. Government Printing Office.
OConnor, N., Hammack, F., & Scott, M. (2010). Social capital, financial knowledge, and Hispanic student college choices. Research in Higher Education, 51, 195219.
Office of Educational Research and Improvement. (1998). First-generation students: Undergraduates whose parents never enrolled in postsecondary education. Statistical Analysis Report, U.S. Department of Education, NCES 98-082.
Padgett, R., Johnson, M., & Pascarella, E. (2012). First-generation undergraduate students and the impacts of the first year of college: Additional evidence. Journal of College Student Development, 53, 243266.
Pascarella, E., Pierson, C., Wolniak, G., & Terenzini, P. (2004). First generation college students: Additional evidence on college experience and outcomes. Journal of Higher Education, 75, 249284.
Pascarella, E., & Terenzini, P. (2005). How college affects students (Vol. 2). San Francisco, CA: Jossey-Bass.
Paulsen, M. (1990). College choice: Understanding student enrollment behavior (ASHE-ERIC Higher Education Report, No. 6). Washington, DC: School of Education and Human Development, The George Washington University.
Paulsen, M., & St. John, E. (2002). Social class and college costs: Examining the financial nexus between college choice and persistence. Journal of Higher Education, 73, 189236.
Perna, L. (2006). Studying college access and choice: A proposed conceptual model. In J. Smart (Ed.), Higher education: Handbook of theory and research (Vol. 21, pp. 99157). Dordrecht, The Netherlands: Springer.
Perna, L., & Titus, M. (2005). The relationship between parental involvement as social capital and college enrollment: An examination of racial/ethnic group differences. Journal of Higher Education, 76, 485518.
Pike, G., & Kuh, G. (2005). First-and second-generation college students: A comparison of their engagement and intellectual development. Journal of Higher Education, 76, 276300.
Prospero, M., & Vohra-Gupta, S. (2007). First generation college students: Motivation, integration, and academic achievement. Community College Journal of Research and Practice, 31, 963975.
Saenz, V. (2007). First in my family: A profile of first-generation college students at four-year institutions since 1971. Los Angeles, CA: Higher Education Research Institute.
Sandefur, G., Meier, A., & Campbell, M. (2006). Family resources, social capital, and college attendance. Social Science Research, 35, 525553.
Schultz, T. (1961). Investment in human capital. American Economic Review, 51, 117.
Scott-Clayton, J. (2015). The role of financial aid in promoting college access and success: Research evidence and proposals for reform. Journal of Student Financial Aid, 45, 722.
Toutkoushian, R., & Hillman, N. (2012). The impact of state appropriations and grants on access to higher education and outmigration. Review of Higher Education, 36, 5190.
Toutkoushian, R., Hossler, D., DesJardins, S., McCall, B., & Canche, M. (2015). The effect of participating in Indianas Twenty-first Century Scholars program on college enrollments. Review of Higher Education, 39, 5995.
University Institutional Research and Reporting. (2012). IR first generation indicator. Bloomington, IN: University Institutional Research and Reporting. Retrieved from https://uirr.iu.edu/doc/resources/ir-data-guides/IR_FirstGeneration_Indicator.doc
Vargas, B., & Conlon, J. (2011). Are we ready for the approaching demographic tsunami?
College and University, 86, 6365.
Ward, L., Siegel, M., & Davenport, Z. (2012). First-generation college students: Understanding and improving the experience from recruitment to commencement. San Francisco, CA: Wiley.
Wells, R. (2008). The effects of social and cultural capital on student persistence: Are community colleges more meritocratic? Community College Review, 36, 2546.
Wells, R., Seifert, T., Padgett, R., Park, S., & Umbach, P. (2011). Why do more women than men want to earn a four-year degree? Exploring the effects of gender, social origin, and social capital on educational expectations. Journal of Higher Education, 82, 132.