Income inequality has been increasing for the past 20 years. A substantial part of the increase in income differences can be explained by changes in the return to education. In dollar terms, 1973 college graduates earned 45 percent more than high school graduates; by 1994 they earned 65 percent more, based on real average hourly wages for college and high school graduates (Baumol and Blinder, 1997). The increasing income disparities between groups of differing educational attainment raises concern that access to postsecondary education (PSE) may not be as widespread as desired.
President Clinton urged for the goal of universal college access in his 1997 State of the Union address, “We must make the thirteenth and fourteenth years of education—at least two years of college—just as universal in America by the 21st century as a high school education is today, and we must open the doors of college to all Americans. ” Using data from the National Education Longitudinal Study of 1988 (NELS) and the National Postsecondary Student Aid Study (NPSAS), this study examines access to postsecondary education by individuals in different income and test score groups.
While many studies have found a statistically significant effect of income on college enrollment,1 less attention has been paid to the effect of family income after controlling for student achievement. This study specifically addresses this issue. We also explore differences in the decision of whether or not to attend PSE or in the type of PSE attended. We are interested in whether students are substituting less expensive alternatives (such as public or 2-year institutions) for high cost institutions, or whether they are not attending PSE at all.
However, we do not examine selectivity of institutions attended. Another goal of this study is to determine which factors, including high school experiences, are especially important in determining college enrollment patterns. Hossler and Maple (1993) find that information on individual background factors allows them to predict, with a high degree of accuracy, which ninth-graders will go to college. The emphasis in our study is on how 1 See, for example, Leslie and Brinkman (1987), Savoca (1990), Schwartz (1986), and Mortenson and Wu (1990). SECTION I. INTRODUCTION 1 MATHTECH, INC.
early indicators, such as expectations and course-taking behavior in the eighth grade, are related to college attendance six years later. 2 Last, we explore whether financial aid availability is a critical factor in determining PSE attendance. The combined effects of shifting federal support from grants to loans, and college tuition increasing at a rate faster than inflation are expected to have a large impact on enrollment patterns for low income youth. This report examines knowledge of and attitudes toward financial aid, and the relationship between such factors and PSE attendance.
We also examine the effect of financial aid receipt on PSE attendance. In summary, the main research questions addressed in this report are: 1. 2. 3. 4. What percentage of students attend PSE, and what types of PSE do they attend? How are income and test score related to who goes to college? What factors, including high school experiences, are especially important in determining college enrollment patterns? Is financial aid availability a critical factor for determining PSE attendance? The rest of the report proceeds as follows.
Section II describes the literature on individual and institutional factors that affect PSE attendance. Section III provides an overview of the data used in this report. It describes the NELS data, the NPSAS data, samples and weights used in the study, and correction of standard errors for sampling techniques. Section IV examines who goes to college. The section highlights the main answers to the first two research questions posed above, in a univariate or multivariate framework. Section V examines factors related to PSE attendance.
Section VI explores the importance of financial aid, including knowledge of financial aid, financial aid applications, and the relationship between being offered financial aid and PSE attendance. Last, we include a bibliography of cited references. The executive summary (at the beginning of the report) highlights our findings and provides policy implications. An NCES study, not yet released, has focused on the “pipeline to higher education” using the NELS data (NCES, 1997). SECTION I. INTRODUCTION 2 2 MATHTECH, INC. One subset of analysis for this study is the group of low income, high test score students.
Low income, high test score students may have the potential to benefit greatly from PSE attendance and, therefore, we want to identify factors or constraints, particularly financial ones, that might limit the students’ educational opportunities beyond high school. SECTION I. INTRODUCTION 3 MATHTECH, INC. II. LITERATURE REVIEW Much of the research on college enrollment patterns is founded upon the “human capital” model Gary Becker advanced. According to this theory, one decides to enroll in college as an investment in future earning power.
Individuals calculate the value of attending college by comparing costs (direct and indirect) with expected income gains, and they make the decision that will maximize their utility over the long term. To understand enrollment behavior according to this model, it is necessary to look at such factors as tuition levels, student financial aid, average wages for high school graduates, and the difference in lifetime earnings between high school and college graduates. Economists and others agree, however, that non-monetary factors also play a major part in the college enrollment decision.
Sociologists’ models of status attainment have suggested a number of background variables that join with economic factors to influence college plans. These include both personal traits (e. g. , academic ability) and interpersonal factors, such as the level of encouragement a student receives from parents and teachers. Hossler and Maple (1993) suggest that individual decisions on enrollment can be broken down into three stages: predisposition, search, and choice. According to their research, students who will ultimately attend college can be differentiated from those who will not as early as the ninth grade.
Within the econometric and sociological models outlined above, the factors affecting enrollment in college can be divided into two general types: those specific to individual students, such as academic achievement and parental education levels, and those specific to educational or vocational alternatives, such as college tuition, financial aid, and unemployment levels. Students’ enrollment decisions can be viewed as jointly determined by their individual characteristics and the institutional or societal conditions that prevail. We first review individual traits that affect college enrollment, and then institutional determinants.
SECTION II. LITERATURE REVIEW 4 MATHTECH, INC. A. INDIVIDUAL FACTORS THAT AFFECT COLLEGE ENROLLMENT Several studies have used data from the National Longitudinal Study of the High School Class of 1972 (NLS72), the National Longitudinal Survey of Labor Market Experience, Youth Cohort (NLSY), and the High School and Beyond Survey (HSB) to examine the factors affecting college enrollments. Manski and Wise (1983), Rouse (1994), and a number of others have used the variables included in these data sets to estimate multinomial logit models of enrollment decisions.
Among the researchers, there seems to be considerable agreement regarding the individual traits that help to determine enrollment. These traits are discussed below. Manski and Wise (1983) presented a key point, namely that the enrollment process begins with the student’s decision to apply to college. This is much more important than the decisions made by college admissions personnel, since most would-be college students are likely to be admitted to some postsecondary institution of average quality.
Jackson (1988) reports that in 1972, more than 97 percent of college applicants were admitted to at least one of their top three choices. The factors of greatest interest, then, are those that cause the student to seek to enroll. Both Manski and Wise (1983) and Rouse (1994) find that individual traits such as achievement levels, high school class rank, and parental education levels are of primary importance in determining the likelihood of a student’s applying to college.
They state that higher family income levels increase the probability of application as well, but to a lesser extent. Manski and Wise also cite a “ ‘peer’ or high school quality effect,” such that the larger the share of a high school senior’s classmates who attend 4-year schools, the more likely he or she will be to do the same. A recent NCES report (1997) describes the relationship among six risk factors (such as changing schools two or more times) and PSE attendance rates. St. John and Noell (1989) and St. John (1990) draw similar conclusions from the NLS72 and HSB data sets.
St. John and Noell state that certain “social background variables” appear to make college enrollment more likely. These include higher test scores, higher grades, higher SECTION II. LITERATURE REVIEW 5 MATHTECH, INC. maternal education levels,3 and family income, as cited by Manski and Wise and Rouse. Other key variables include participation in an academic track during high school and “high postsecondary aspirations,” as measured by students’ reporting of the highest level of schooling they expect to achieve.
Hossler and Maple (1993) find that parental education levels have a stronger effect on enrollment plans than student ability or income level. Other background factors that researchers have found to be significant include the level of parental encouragement (Hossler, Braxton, and Coopersmith, 1989) and students’ own expectations about the college decision (Borus and Carpenter, 1984). Jackson (1988) concludes that test scores, grades, taking part in a college preparatory program, and attending a school with many college-going peers are the student attributes most important for college enrollment.
Kohn, Manski, and Mundel (1976) report that parents’ education level has a positive effect on a student’s likelihood of enrollment, but state that this effect decreases as family income rises. A number of researchers have examined the effects of family income levels on college enrollment. Manski (1992:16) concludes that there are “persistent patterns of stratification of college enrollments by income. ” Both Manski (1992) and Kane (1995) present census data for multi-year periods that show, for ascending income levels, a steadily increasing percentage of 18to 19-year-old dependent family members enrolled in college.
Using the same data source, Clotfelter (1991) and Mortenson and Wu (1990) cite positive income effects for the 18- to 24year-old group as well. Hauser (1993) finds large family income effects on college enrollment for White and Hispanic families, but he finds no such effects for Black families. 3 St. John and Noell do not include paternal education levels as a variable in their study. Manski and Wise and Rouse consider maternal and paternal education levels as separate variables, but present their conclusions in terms of parental education levels as a whole.
Most of the studies reviewed here do not distinguish between mother’s and father’s education levels. One exception is the study by Kohn, Manski and Mundel (1976). This study estimates a model using subsamples of the SCOPE survey from two different states. While one group shows that the father’s education level has a greater effect on the likelihood of college attendance than does the mother’s, the other group shows the mother’s education level as having a greater effect. SECTION II. LITERATURE REVIEW 6 MATHTECH, INC. B. INSTITUTIONAL FACTORS THAT AFFECT COLLEGE ENROLLMENT.
In addition to the factors that operate at an individual level, researchers have found a variety of institutional factors, or factors pertaining to educational and vocational alternatives, that affect college enrollment levels. Manski and Wise (1983) include among these factors tuition level, “quality of school” (as measured by the average combined SAT score of incoming freshmen), and the availability of government and institutional financial aid.
Rouse (1994) examines the factor of proximity by estimating changes in enrollments that would result from decreasing the average distance to the nearest 2-year college. She also considers the effects of tuition levels and financial aid availability, as well unemployment rates, which serve as a measure of competing opportunities available to high school seniors.
Tuition levels are another institutional factor with a significant effect on college enrollment. Leslie and Brinkman (1987) review 25 studies on this subject, and find a general consensus that a $100 increase in tuition nationwide, in 1982–1983 academic year dollars, would result in a 6 percent decline in the college participation rate for the 18- to 24-year-old group.
Savoca (1990) makes the point that high tuition levels may lessen postsecondary enrollments in the aggregate by discouraging some students from ever applying to college. The effects of tuition levels are moderated in many cases by the effects of financial aid. McPherson and Schapiro (1991) state that the variable of interest should be net cost, or tuition less financial aid. At the initial stages of the enrollment decision, however, students often lack information on their eligibility for financial aid and the amount of aid they would be likely to receive.
Researchers have differing views regarding the effects of financial aid on enrollment at different types of institutions. Reyes (1994) finds that increases in financial aid positively affect both 2-year and 4-year college enrollment rates, based on information from the NLSY and HSB. Manski and Wise (1983), using the NLS72, conclude that financial aid affects students’ decisions to attend 2-year institutions, as opposed to not going to college at all. However, this study finds that enrollments at 4-year schools have little sensitivity to the availability of financial aid.
Manski and Wise do not consider the effect of financial aid on the student’s choice between a 2-year and a 4-year institution. SECTION II. LITERATURE REVIEW 7 MATHTECH, INC. Other researchers have compared the effects of decreasing tuition with the effects of increasing financial aid. Manski and Wise (1983) find that for those attending 2-year schools, an additional dollar of financial aid would be worth more than a one dollar reduction in tuition. St. John (1990:172) also finds that “college applicants…
[are] more responsive to changes in student aid than to changes in tuition,” except for those in the upper income group. Kane (1995), however, argues that while financial aid increases may be more equitable because they are means tested, they are not as effective as decreases in tuition. This is a consequence of the complexity of the financial aid application process and the unwillingness of low income families to borrow to finance a college education. When studying the effect of tuition and financial aid on PSE enrollment, the group to be especially concerned about is low income students.
Leslie and Brinkman (1987) and Savoca (1990) find that tuition levels affect enrollment decisions for low income students much more than for middle and upper income groups. By the same token, the availability of financial aid is a much more crucial factor for those at lower income levels. Orfield (1992) notes that the maximum Pell grant is less than one-fifth of the tuition at an elite university. Such a gap between aid and costs, he contends, may steer many low income students toward lower cost schools. Hearn’s 1991 study supports this hypothesis.
He finds that when academic ability, achievement, and other factors are controlled for, lower income students are especially likely to choose institutions of lower selectivity. Schwartz (1985) finds that low income students are affected differently by publicly provided financial aid and aid supplied by institutions. He states that public grants tend to promote greater equity among income groups in college enrollment. Private grants, however, are often awarded on the basis of academic ability, and they tend to favor students who could afford to go to college without them.
Clotfelter (1991) expresses the same concern about the effects of institutional aid. Manski and Wise (1983) note that even public aid is not always awarded where the need is greatest. They state that in 1979, 59 percent of Basic Educational Opportunity Grants were awarded to students who would probably have gone to college in the absence of such aid. Table 1 summarizes the data sources used in the studies mentioned here. SECTION II. LITERATURE REVIEW 8 MATHTECH, INC.
Table 1 MAIN DATA SOURCES FOR WORKS CITED IN LITERATURE REVIEW STUDY Borus, Michael E.and Carpenter, Susan A. , “Factors Associated with College Attendance of High-School Seniors” (1984) Clotfelter, Charles T. , “Demand for Undergraduate Education” (1991) Hauser, Robert M. , “Trends in College Entry among Whites, Blacks, and Hispanics” (1993) Hearn, James C. , “Academic and Nonacademic Influences on the College Destinations of 1980 High School Graduates” (1991) Hossler, Don, Braxton, John and Coopersmith, Georgia, “Understanding College Choice” (1989).
Hossler, Don and Maple, Sue, “Being Undecided about Postsecondary Education” (1993) Jackson, Gregory A., “Did College Choice Change during the Seventies? ” (1988) Kane, Thomas, “Rising Public College Tuition and College Entry: How Well Do Public Subsidies Promote Access to College? ” (1995) Kohn, Meir G. , Manski, Charles F. , and Mundel, David S. , “An Empirical Investigation of Factors which Influence College-going Behavior” (1976) Leslie, Larry L. , and Brinkman, Paul T. , “Student Price Response in Higher Education” (1987) Manski, Charles F. , and Wise, David A. , College Choice in America (1983) Manski, Charles F. , “Income and Higher Education” (1992)
McPherson, Michael S., and Schapiro, Morton Owen, “Does Student Aid Affect College Enrollment? New Evidence on a Persistent Controversy” (1991) Mortenson, Thomas G. , and Wu, Zhijun, “High School Graduation and College Participation of Young Adults by Family Income Backgrounds 1970 to 1989” (1990) National Center for Education Statistics. “Confronting the Odds: Students At Risk and the Pipeline to Higher Education” (1997). MAIN DATA SOURCES 1979 and 1980 National Longitudinal Surveys of Labor Market Experience, Youth Cohort (NLSY) Review of studies done by others, with data from Current Population Survey (CPS) and High.
School and Beyond (HSB) CPS HSB, Higher Education Research Institute (HERI) data Review of studies done by others Cluster sample of 5,000 Indiana ninth graders National Longitudinal Study of the High School Class of 1972 (NLS72), HSB NLSY, HSB, CPS School to College: Opportunities for Postsecondary Education (SCOPE) Survey Meta-analysis of studies done by others NLS72 NLS72, HSB, CPS Cooperative Institutional Research Program (CIRP) data, CPS HSB, Current Population Report, CPS NELS SECTION II. LITERATURE REVIEW.
9 MATHTECH, INC. STUDY Orfield, Gary, “Money, Equity, and College Access” (1992) Reyes, Suzanne, “The College Enrollment Decision: The Role of the Guaranteed Student Loan” (1994) Rouse, Cecilia Elena, “What to Do after High School: The Two-Year versus Four-Year College Enrollment Decision” (1994) St. John, Edward P. , and Noell, Jay, “The Effects of Student Financial Aid on Access to Higher Education: An Analysis of Progress with Special Consideration of Minority Enrollment” (1989) St. John, Edward P., “Price Response in Enrollment Decisions:
An Analysis of the High School and Beyond Sophomore Cohort” (1990) Savoca, Elizabeth, “Another Look at the Demand for Higher Education: Measuring the Price Sensitivity of the Decision to Apply to College” (1990) Schwartz, J. Brad, “Student Financial Aid and the College Enrollment Decision: The Effects of Public and Private Grants and Interest Subsidies” (1985) Schwartz, J. Brad, “Wealth Neutrality in Higher Education: The Effects of Student Grants” (1986) MAIN DATA SOURCES Review of history of federal student financial aid programs NLSY, HSB NLSY, HSB, CPS.
NLS72, HSB HSB NLS72 HSB, CPS HSB, CPS SECTION II. LITERATURE REVIEW 10 MATHTECH, INC. III. DATA A. NELS DATA While a number of studies have used data from the National Longitudinal Survey, Youth Cohort (NLSY), the National Longitudinal Study of the High School Class of 1972 (NLS72), and the High School and Beyond Survey (HSB) to examine the factors affecting college enrollments, this work effort is among the first to use NELS to analyze these types of issues. In 1988, NELS initially surveyed over 24,000 public and private school eighth graders throughout the United States.
The nationally representative eighth grade cohort was tested in four subjects (mathematics, reading, science, and social studies). Two teachers of each student (representing two of the four subjects) were also surveyed, as was an administrator from each school. On average, each of the 1,052 participating schools was represented by 24 students and five teachers. Parents were also surveyed, providing researchers with detailed information on family background variables.
Since 1988, the initial eighth grade cohort has been re-surveyed three times (and has been “freshened” with new sample members). The first follow-up of NELS (spring, 1990), included the same components as the base year study, with the exception of the parent survey, which was not implemented in the 1990 round. It also included a component on early dropouts (those who left school between the end of eighth grade and the end of 10th grade). The second follow-up (spring, 1992), repeated all components of the first follow-up study and also included a parent questionnaire.
However, this time only one teacher of each student (either a mathematics or a science teacher) was asked to complete a teacher questionnaire. High school transcript data were also collected for these students. A subsample of the NELS:88 second follow-up sample was again followed-up in the spring of 1994, when most sample members had been out of high school for 2 years. In all, 14,915 students were surveyed, most through computer-assisted telephone interviewing.
Major content areas for the third follow-up questionnaire were: education histories; work experience histories; work-related training; family formation; opinions and other experiences; occurrence or SECTION III. DATA 11 MATHTECH, INC. non-occurrence of significant life events; and income. Data collection for this wave began on February 4, and ended on August 13, 1994. At the time the data were collected, most of the respondents were 2 years out of high school. Table 2 summarizes the components of the different waves of the surveys.
Table 2 OVERVIEW OF NELS NELS Components Grades included Cohort Base Year Spring term 1988 grade 8 students: questionnaire, tests questionnaire questionnaire two teachers per student (taken from reading, mathematics, science, or social studies) First Follow-up Spring term 1990 modal grade = sophomore students, dropouts: questionnaire, tests none questionnaire two teachers per student (taken from reading, mathematics, science, or social studies) Second Follow-up Spring term 1992 modal grade = senior students, dropouts: questionnaire, tests, H. S. transcripts questionnaire questionnaire one teacher per student (taken from mathematics or science).
Third Follow-up Spring 1994 H. S. + 2 years all individuals: questionnaire none none none Parents Principals Teachers B. NPSAS DATA Because the NELS database does not contain detailed information on financial aid, the National Postsecondary Student Aid Study (NPSAS) database is used to supplement our study with additional financial aid information. This database is used to predict financial aid for the respondents in NELS based on demographic and other characteristics that are available in both databases.
NPSAS is constructed specifically to provide information on financing of postsecondary education, so it is a good candidate for this use. This database surveys a nationally representative sample of undergraduate, graduate, and first-professional students attending less than 2-year, 2-year, 4-year, and doctoral granting institutions. Both students who receive and those who do not receive financial aid are surveyed. SECTION III. DATA 12 MATHTECH, INC. The 1993 NPSAS study collected information on more than 78,000 undergraduate and graduate students at about 1,100 institutions.
To be eligible, students must have been enrolled between May 1, 1992 and April 30, 1993 at a postsecondary institution in the United States or Puerto Rico. The students had to be enrolled in courses for credit, and in a program of 3 months or longer. Also eligible for inclusion were students who received a bachelor’s degree between July 1, 1992 and June 30, 1993. Students who were enrolled in a GED program or who were also enrolled in high school were not included. C. SAMPLE AND WEIGHTS Of the 14,915 respondents in the third NELS follow-up, 13,120 are represented in all four waves of the NELS data.
The remaining 1,795 respondents are either first follow-up “freshened” students,4 second follow-up freshened students,5 base-year ineligibles,6 or base-year eligible students who declined to participate in one or more of the survey waves, but who did participate in the third survey wave. The breakdown of these 1,795 respondents is as follows: 501 first follow-up freshened students, 102 second follow-up freshened students, 271 base-year ineligibles, and 921 base-year eligibles with missing survey waves.
To take advantage of the longitudinal nature of the NELS data and to be consistent across models and issues in the report, we focus our work on the sample of 13,120 respondents represented in all four waves of the NELS data. Consequently, the weight used in our analyses, (“F3PNLWT”) applies to sample members who completed questionnaires in all four rounds of NELS:88. As a result, the longitudinal analyses that we conduct, and the estimates that are produced in this study can only be used to make projections to the population of spring 1988 eighth graders.
In the descriptive tables, all percentages are weighted using F3PNLWT, including the analyses with the high school transcript data. Those who were tenth graders in 1990 but were not in the base-year sampling frame, either because they were not in the country or because they were not in the eighth grade in the spring term of 1988. Those who were 12th graders in 1992 but were not in either the base year or first follow-up sampling frames, either because they were not in the country or because they were not in the eighth (10th) grade in the spring term of 1988 (1990).
6 5 4 Students excluded in 1988 due to linguistic, mental, or physical obstacles to participation. 13 SECTION III. DATA MATHTECH, INC. This sample includes dropouts, since the purpose of this study is to examine the overall question of what characteristics of eighth graders in 1988 are related to PSE attendance. We focus on early indicators, such as educational expectations and course-taking behavior in the eighth grade, and not on the “pipeline” of high school experiences that a dropout would lack access to.
However, the dropouts were not asked the same set of survey questions as the other respondents, and, therefore, some of the analysis does not include dropouts. For each of our tables or figures, we note whether or not the dropouts are included in the analysis. D. CORRECTED STANDARD ERRORS Because NELS data are collected through a multi-stage sampling scheme, calculation of standard errors through standard methods can understate these errors. The sampling technique used in NELS is a selection of schools, and then within schools, a selection of students.
With this sampling method, the observations of different students may not be independent from one another. Stata™, the statistical software used for analysis in this report, corrects the standard errors for these sampling techniques. Except for multinomial logit models, for which this correction is not available, survey correction techniques are used, and we note whenever the corrections are not used. However, we have found that such corrections do not have a large effect on our results, and therefore, we present all results with confidence. E. VARIABLE DEFINITIONS.
The appendix to this study contains definitions of the key variables used in our analysis. For each key variable, we describe how we constructed the variable and we list the names of the NELS variables used in the construction. SECTION III. DATA 14 MATHTECH, INC. IV. WHO GOES TO COLLEGE? A. WHAT PROPORTION OF STUDENTS ATTEND COLLEGE, AND WHAT TYPE OF COLLEGE DID THEY ATTEND? We begin our analysis by examining the demographics of postsecondary school choice and discussing our main findings regarding college attendance rates and types of postsecondary education (PSE) attended.
As shown in Table 3, a majority of 1988 eighth graders attend some type of PSE by 1994. Overall, 62. 7 percent of the respondents attend PSE. (Note that in all of the tables in this report, all percentages are weighted. ) Students are most likely to attend a 4-year public or a less than 4-year public school. Approximately 24 percent of the students attend each of these types of schools. Next most common are 4-year private schools. Just over 11 percent of the respondents attend 4-year private schools.
Only 4 percent of the respondents attend less than 4-year private schools. Thirty-seven percent of the respondents do not attend any type of PSE. Women are slightly more likely than men to attend PSE. While 60 percent of men attend PSE, 65 percent of women attend. Women are more likely than men to attend 4-year private schools and less than 4-year private schools. Native Americans, Blacks, and Hispanics are least likely to attend PSE and Asians and Pacific Islanders are most likely to attend PSE.
Hispanics are most likely to attend less than 4year private schools. Students whose parents have higher education levels are much more likely to attend PSE. While only 33 percent of students whose parents have less than a high school education attend PSE, 90 percent of students whose parents have an advanced degree attend PSE. SECTION IV. WHO GOES TO COLLEGE? 15 MATHTECH, INC. Table 3 DEMOGRAPHICS OF POSTSECONDARY SCHOOL CHOICE1 No PSE 4-Year Public 4-Year Private.