Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Investigating the association of student choices of major on college student loan default: a propensity-scored hierarchical linear model
(USC Thesis Other)
Investigating the association of student choices of major on college student loan default: a propensity-scored hierarchical linear model
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
Running Head: CHOICE OF MAJOR & STUDENT LOAN DEFAULT
Investigating the association of student choices of major with college student loan
default: A propensity-scored hierarchical linear model
Shirley Parry
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY (URBAN EDUCATION POLICY)
May 2017
Dissertation Committee Members
Dr. William G. Tierney, Chair
Dr. Tatiana Melguizo
Dr. John Slaughter
CHOICE OF MAJOR & STUDENT LOAN DEFAULT i
Table of Contents
TABLE OF CONTENTS .............................................................................................................. i
LIST OF TABLES ....................................................................................................................... iv
LIST OF FIGURES ...................................................................................................................... v
CHAPTER 1 – INTRODUCTION TO THE STUDY ............................................................... 1
BACKGROUND .............................................................................................................................. 3
Massification of Higher Education and Rising Tuition Costs ................................................ 3
Higher-education Funding ...................................................................................................... 5
Student Debt, Repayment and Default .................................................................................... 9
Loan Default and the Economy ............................................................................................. 11
STATEMENT OF THE PROBLEM ................................................................................................... 13
PURPOSE OF THE STUDY ............................................................................................................. 17
CHAPTER 2 – LITERATURE REVIEW ................................................................................ 19
STUDENT CHARACTERISTICS ..................................................................................................... 19
Race/Ethnicity and Socioeconomic Status ............................................................................ 19
Age ........................................................................................................................................ 20
Gender ................................................................................................................................... 21
Academic Preparation and Attainment ................................................................................. 21
Choice of Major .................................................................................................................... 22
Student Knowledge of Loan Debt ......................................................................................... 24
Debt Load.............................................................................................................................. 25
Transfer and Program Completion ....................................................................................... 25
Employment after Graduation .............................................................................................. 26
INSTITUTIONAL CHARACTERISTICS ............................................................................................ 27
GAP IN THE LITERATURE ............................................................................................................ 29
CHAPTER 3 – METHODOLOGY/DATA COLLECTION/ANALYSIS ............................. 31
RESEARCH QUESTIONS .............................................................................................................. 31
POPULATION AND SAMPLE ......................................................................................................... 31
HIERARCHICAL LINEAR MODELING ........................................................................................... 33
PROPENSITY SCORE MATCHING ................................................................................................. 34
THE MODEL: QUESTION 1 .......................................................................................................... 36
Propensity Score Estimation ................................................................................................. 36
Hierarchical Linear Modeling (FEDLOAN as DV) ............................................................. 36
THE MODEL: QUESTIONS 2 AND 3 .............................................................................................. 38
Hierarchical Linear Modeling (LOANST09 as DV) ............................................................. 38
THE MODEL: QUESTION 4 .......................................................................................................... 39
Hierarchical Linear Modeling (JOBST09 as DV) ................................................................ 40
LIMITATIONS .............................................................................................................................. 41
Majors ................................................................................................................................... 41
CHOICE OF MAJOR & STUDENT LOAN DEFAULT ii
Loan Default ......................................................................................................................... 42
BPS Database ....................................................................................................................... 43
CHAPTER 4 – DATA ANALYSIS ........................................................................................... 44
DESCRIPTIVE STATISTICS ........................................................................................................... 45
Institutions and Majors ......................................................................................................... 45
Gender and Age .................................................................................................................... 46
Race/Ethnicity ....................................................................................................................... 47
Cumulative Federal Loans, Repayment Status, and Employment Status by Institution Type
............................................................................................................................................... 50
Cumulative Federal Loans, Repayment Status, and Employment Status by Major .............. 51
Summary ............................................................................................................................... 54
STATISTICAL TESTS ................................................................................................................... 55
Continuous Variable (FEDLOAN) and Categorical Variable (Institution Type – FP09) .... 55
Continuous Variable (FEDLOAN) and Categorical Variable (Major – MAJ09B) .............. 56
Chi-square Tests of Significance for Loan Repayment and Employment Status at Different
Types of HEIs and in Different Majors ................................................................................. 60
Summary ............................................................................................................................... 62
REGRESSION ANALYSIS ............................................................................................................. 64
Association between Students’ Ability to Repay Federal Student Loans and Major ............ 64
Association Between Students’ Ability to Repay Federal Student Loans and the Type of
Higher-education Institution Attended ................................................................................. 67
Association between Students’ Employment Status by Major .............................................. 67
Association between Students’ Employment and the Type of Higher-education Institution. 72
Association between Students’ Cumulated Federal Loan Amount and Major ..................... 73
Association between Students’ Cumulated Federal Loan Amount and Type of Higher-
education Institution ............................................................................................................. 77
Intraclass Correlation ........................................................................................................... 78
CONCLUSION .............................................................................................................................. 79
Loan Repayment Status ......................................................................................................... 79
Employment Status ................................................................................................................ 80
Cumulative Federal Student Loan Amount ........................................................................... 80
CHAPTER 5 – CONCLUSIONS AND RECOMMENDATIONS ......................................... 82
LITERATURE REVIEW AND METHODOLOGY ............................................................................... 83
RESEARCH STUDY QUESTIONS AND FINDINGS ........................................................................... 84
Question 1: To What Extent Do Total Cumulated Student Loans Differ Across Fields of
Study (Both Career/Vocational and Degree Programs)? ..................................................... 84
Question 2. To What Extent Do Student Loan Default Rates Differ Across Fields of Study
(Both Career/Vocational and Degree Programs)? ............................................................... 93
Question 3. Controlling for Field of Study, to What Extent Do Institutional Factors Explain
Differences in the Likelihood of Default? ............................................................................. 95
CHOICE OF MAJOR & STUDENT LOAN DEFAULT iii
Question 4: To What Extent Do Different Fields of Study Affect Graduates’ Employment,
and How Does This Differ Among Different Types of Institutions? ..................................... 98
LIMITATIONS ............................................................................................................................ 100
Majors ................................................................................................................................. 100
Loan Default ....................................................................................................................... 100
BPS Database ..................................................................................................................... 101
SUMMARY ................................................................................................................................ 102
To What Extent Do Total Cumulated Student Loans Differ Across Fields of Study (Both
Career/Vocational and Degree Programs)? ...................................................................... 102
To What Extent Do Student Loan Default Rates Differ Across Fields of Study (Both
Career/Vocational and Degree Programs)? ...................................................................... 102
Controlling for Field of Study, to What Extent Do Institutional Factors Explain Differences
in the Likelihood of Default? .............................................................................................. 103
To What Extent Do Different Fields of Study Affect Graduates’ Employment, and How Does
This Differ Among Different Types of Institutions? ............................................................ 103
CONCLUSIONS .......................................................................................................................... 104
RECOMMENDATIONS FOR FUTURE RESEARCH ......................................................................... 106
REFERENCES .......................................................................................................................... 110
CHOICE OF MAJOR & STUDENT LOAN DEFAULT iv
List of Tables
Table 1. Comparison of Undergraduate Enrollment by Institution Type 2010 and 2014…….......4
Table 2. Summary of Federal Aid Disbursed to Students by Program…………………………...6
Table 3. Ten-Year Trend in Student Aid and Nonfederal Loans per Full-Time Equivalent (FTE)
Student Used to Finance Postsecondary Education Expenses, in 2011 Dollars, 2002–03 to
2012–13…………………………………………………………………………………...7
Table 4. Breakdown of Majors by Higher-Education Institution Types………………………...46
Table 5. Distribution of Traditional vs. Non-traditional Students..……………………………...47
Table 6. Racial/Ethnic Demographics of the BPS 2004–2009 Data.……………………………48
Table 7. Race/Ethnicity of Enrolled Students by Institution Type………………………………48
Table 8. Academic Performance and Parental Education of Students by Institution……………50
Table 9. Cumulative Federal Student Loan, Loan Repayment Status, and Employment Status by
Institution………………………………………………………………………………...52
Table 10. Federal Student Loan Amount, Loan Repayment and Employment Status by Major..53
Table 11. ANOVA Test of Significance for Cumulative Federal Student Loan and Institution
Type……………………………………………………………………………………...57
Table 12. Pairwise Comparisons of Means with Equal Variances Over Different Institution
Types……………………………………………………………………………………..57
Table 13. ANOVA Test of Significance for Cumulative Federal Student Loan and Major...…..58
Table 14. Pairwise Comparisons of Means with Equal Variances………………………………59
Table 15. Chi-square Test of Hypothesis: Loan Repayment and Employment Status and
Institution Type…………………………………………………………………………..61
Table 16. Chi-square Test of Hypothesis: Loan Repayment and Employment Status and
Major……………………………………………………………………………………..63
Table 17. Regression Estimates Expressed in Odds Ratio of Likelihood of Default by Major
Compared to Undeclared………………………..……………………………………….65
Table 18. Regression Estimates Expressed in Odds Ratio of Inability to Repay by
Major……………………………………………………………………………………..66
Table 19. Regression Estimates Expressed in Odds Ratio of Inability to Repay Loans by
Institution………………………………………………………………………………...69
Table 20. Regression Estimates Expressed in Odds Ratio of Employment Status by Major
Compared to Undeclared………………………………………………………………...70
Table 21. Regression Estimates Expressed in Odds Ratio of Employment Status by Major……71
Table 22. Regression Estimates Expressed in Odds Ratio of Employment Status by
institution………………………………………………………………………………...72
Table 23. Regression Estimates of Cumulative Federal Student Loan by Major………………..76
Table 24. Regression Estimates of Cumulative Federal Student Loan by Institution…………...78
Table 25. Intraclass Correlations Between Outcomes and Majors and Institutions……………..79
Table 26. Regression Estimates of Cumulative Federal Student Loan Amount by Major…..…..92
CHOICE OF MAJOR & STUDENT LOAN DEFAULT v
List of Figures
Figure 1. Age at first year enrolled distribution of BPS 2004-2009 population…………………47
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 1
Chapter 1 – Introduction to the Study
How much is a degree worth? Is a degree in the humanities worth as much as a degree in
the sciences? How much are students borrowing to fund their studies? Can they pay back their
loans? How much does a degree cost? Should a degree with a higher lifetime-earnings potential
cost more? The idea that investment in human capital increases future earnings for the individual
(Becker, 1961) is widely accepted and embraced. People who attain higher levels of education
(bachelor’s degrees as compared to high-school diplomas, for example) consistently earn more,
with higher lifetime earnings than those who do not pursue postsecondary education (Carnevale,
Rose, & Chea, 2011). Investment in human capital also boosts the economy (Becker, Murphy, &
Tamura, 1993). Since the economic development of a country depends on the technological
skills and knowledge of its workforce, an investment in human capital by the government
facilitates and sustains economic growth.
Student enrollment in higher education has increased annually since the Higher
Education Act of 1965, and it is currently at its highest level since the National Center for
Education Statistics began collecting data, both in terms of absolute numbers and percentage of
college-age (18–24) population (NCES, 2016). At the same time, due to factors such as campus
expansion and state budget cuts to higher education, these students also face the pressure of
financing college. As college tuition increases outpace wage and household income gains,
students must turn to federal, state, and private grants and loans to finance their higher education.
In 2014, 70% of college seniors carried student loans, with the average amount of $29,000
(“Student debt and the class of 2014,” 2015). The current total loan amount stands at over $1.3
trillion (Federal Reserve, 2016a), which is the highest it has ever been. Total student loans in the
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 2
U.S. are higher than auto loans ($1.1 trillion) and credit card debt ($729 million) (Federal
Reserve, 2016b; Issa, n.d.). Although state grants, subsidies, as well as federal grant programs
are available, they do not nearly cover most students’ tuition and room and board expenses.
As costs have increased relentlessly over the past two decades, students have relied on
federal loans to fund their studies. Since the passage of the Higher Education Act (HEA) in 1965,
the government has provided tuition assistance to students from low- and some middle-income
families through grants and loans. Under the provision of Title IV of the HEA and its subsequent
reauthorizations, the federal government offers students access to both subsidized and
unsubsidized loans, both of which must be repaid after graduation. With more students enrolling
in college every year, the total volume of loans in the federal loan program has grown. The loan
amount per student has also increased as tuition has risen. As long as these loans are serviced and
repaid, the loan amount per student does not pose a problem for students or for the economy.
However, in recent years we have witnessed a worrying trend, as the student loan default rate has
risen. The three-year cohort default rate (CDR)—the share of an institution’s borrowers with
delinquent federal loans—grew from 13.4% in 2009 to 14.7% in 2010, with the for-profit higher-
education sector registering the highest rate of default at 21.8%, followed by public higher-
education institutions at 13%, and private non-profit colleges and universities at 8.2%. This
number has since decreased. The CDR in 2013 stood at an average of 11.3% (Department of
Education, 2016). The cohort at proprietary (for-profit) institutions continued to show the highest
default rate (15%) among the three types of higher-education institutions (HEIs), although the
absolute percentage had come down from its peak in 2010. The CDR at public HEIs fell to 11%
and private non-profit institutions saw their CDR decline to 7%.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 3
To address the problems posed by a rising student loan debt burden and the increasing
loan default rate accompanying it, there must be a clear understanding of why students default, to
what extent they default, and whether or not their choices of major and choices of education
institutions are associated with their tendency to default. The proposed study uses a propensity-
scored hierarchical linear model to segregate loan default data by program type. This serves the
purpose of providing a more detailed picture of the increasingly significant student loan and
default issue at hand.
Background
Massification of Higher Education and Rising Tuition Costs
It has been a goal for many federal administrations to expand higher-education
opportunities for high-school graduates, going as far back as the G.I. Bill after World War II.
This commitment was cemented through the aforementioned HEA of 1965, which made federal
funds available for student financial aid. As the number of students seeking higher education
grew, so too did the budget for Title IV funding (as federal funding for higher education is
generally known) under the HEA. Previously reserved for the privileged few, higher education
has become a rite of passage for a large percentage of high-school graduates. Whereas only half
of high-school graduates entered college in 1970, this number has grown to 70% in 2010 (U.S.
Census Bureau, 2012). College enrollment (both full-time and part-time) increased from 13.7
million in 1990 to 21.4 million in 2010 (College Board, 2013a). Graduation rates have also risen.
Among those 25 years or older in the U.S., only 11% were college graduates in 1970, but by
2010, 30% had college degrees (U.S. Census Bureau, 2012).
Students in the United States have three main choices for higher education: public
research universities, 4-year comprehensive undergraduate universities and community colleges
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 4
(such as those in the University of California, California State Universities, and California
Community Colleges systems), private non-profit colleges and universities (such as Harvard
University and University of Southern California), and for-profit colleges and universities (such
as University of Phoenix and other, smaller proprietary schools). In 2014, public institutions
educated an overwhelming majority (75%) of the 17.7 million undergraduate students in the
U.S., while private non-profit colleges and universities enrolled 16% of that total (Ginder, Kelly-
Reid, & Mann, 2015). In the last two decades, however, for-profit colleges and universities
(FPCUs) have experienced substantial growth, increasing their share in the higher-education
sector from 2% in 1987 to over 9% in 2010 (Bennett, Lucchesi, & Vedder, 2010). In the 2010–
2011 academic year, 12% of all postsecondary student enrollment was at FPCUs (National
Conference of State Legislatures, 2012). Due to recent negative publicity from federal and state
investigations into FPCUs’ aggressive recruitment tactics, false marketing promises, and federal
aid fraud, student enrollment at FPCUs has fallen back to 9% of the total undergraduate
population (Fain, 2015; Smith, 2015). Table 1 shows the distribution of student enrollment by
institution type.
Table 1
Comparison of Undergraduate Enrollment by Institution Type 2010 and 2014
2010 % of total 2014 % of total
Total 21,600,000 100 20,700,000 100
Public 15,200,000 70 14,800,000 71
Private non-
profit 3,900,000 18 4,000,000 19
For-profit 2,500,000 12 1,900,000 9
Source: National Center for Education Statistics
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 5
Higher-education Funding
The increase in higher-education enrollment can be attributed to an overall increase in the
college-age population, but even more so it is the product of an increase in non-traditional
college students. Common defining characteristics of non-traditional students are age (over 24
years old), diverse backgrounds (race and gender), living off campus, working full- or part-time,
and being enrolled in non-degree occupational programs (NCES, n.d.). These students are also
likely to come from lower-income households and to need federal or state grants and loans to
finance their higher education.
To facilitate wider participation in postsecondary education, the federal government
made provisions to offer aid dollars in the form of need-based grants and loans. Funding for
public universities depends largely on state budget appropriations. Title IV student aid funding
consists of grants, loans, and work-study funds. Title IV grants, which do not have to be repaid,
are given to students who demonstrate the greatest financial need (examples include Pell Grants
and Federal Supplemental Educational Opportunity Grants). Federal loans, such as Perkins and
Direct/Indirect Subsidized Stafford Loans, are low-interest-rate loans that generally must be
repaid in full within a certain period of time after graduation. Work-study programs allow
students to pay for their educational expenses in part by working part-time on campus. From
academic year 2001–2002 to 2013–2014, the total amount of federal financial aid awarded to
students under Title IV of the HEA jumped from $72.3 billion to an estimated $133.8 billion, a
12-year increase of 85% (NASFAA, 2014). Table 2 shows a summary of federal aid disbursed to
students by program in 2014.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 6
Table 2
Summary of Federal Aid Disbursed to Students by Program in 2014
Programs (in millions of dollars) % of total
Loan programs
William D. Ford Federal Direct Loan Program 99,355
Federal Perkins Loan Program 1,167
Subtotal Loan Programs 100,522 75
Grant Programs
Federal Pell Grant Program 31,554
Federal Supplemental Educational Opportunity
Grant Program 694
The Teacher Education Assistance for College
and Higher Education Grant Program 97
Other Grant Program/ Rounding 0
Subtotal Grant Programs 32,345 24
Work-Study Programs
Federal Work-Study Program 928
Rounding 1
Subtotal Work-Study Program 929 1
Grand Total 133,796 100
Source: Federal Student Aid.
The proportion of each of these types of funding has fluctuated over the years, but loans
constitute an increasing percentage from year to year. This has taken place as the government has
made Title IV financial aid available to more and more middle-class students, whose household
incomes are too high to be eligible for grants. A typical low-income student in 1972–1973, for
example, would receive 75% in grants and 25% in loans to cover college expenses. By 2012–
2013, a typical student on average borrowed about as much as he or she receives in grants
(College Board, 2013a).
In 2012–2013, $68 billion (60%) of undergraduate student aid distributed consisted of
federal loans, while $45 billion (39%) consisted of federal grants (only $0.9 billion [1%]
consisted of federal work-study; College Board, 2013b). This presented quite a reversal from
1992–1993, when grants comprised 65% of total aid, and the loan proportion was only 33%.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 7
Table 3
Ten-Year Trend in Student Aid and Nonfederal Loans per Full-Time Equivalent (FTE) Student Used to Finance Postsecondary Education Expenses,
in 2011 Dollars, 2002–03 to 2012–13
02–03 03–04 04–05 05–06 06–07 07–08 08–09 09–10 10–11 11–12 12–13
Nonfederal Student Loans $839 $1,047 $1,328 $1,556 $1,745 $1,829 $848 $573 $500 $504 $553
Education Tax Benefits $534 $560 $561 $559 $544 $526 $764 $1,108 $1,237 $1,256 $1,274
Federal Parent Loans (PLUS)
& Grad PLUS Loans
$494 $603 $674 $716 $845 $849 $857 $984 $1,136 $1,173 $1,083
Federal Loans – Unsubsidized $1,726 $1,896 $2,000 $2,064 $2,014 $2,158 $2,882 $3,143 $3,066 $2,975 $3,483
Federal Loans – Subsidized $1,983 $2,132 $2,181 $2,137 $2,069 $2,292 $2,355 $2,569 $2,635 $2,576 $1,740
Private and Employer Grants $714 $745 $780 $825 $863 $908 $886 $839 $861 $885 $916
Institutional $1,794 $1,917 $1,982 $2,085 $2,168 $2,215 $2,212 $2,350 $2,505 $2,642 $2,789
Federal Pell Grants $1,182 $1,230 $1,204 $1,110 $1,060 $1,156 $1,304 $2,024 $2,311 $2,109 $2,027
State Grants $588 $580 $605 $598 $627 $630 $599 $602 $593 $590 $612
Federal Campus-Based
Programs
$324 $329 $313 $293 $278 $246 $192 $170 $168 $167 $161
Other Federal Programs $361 $399 $429 $428 $462 $464 $478 $769 $885 $819 $894
Total $10,540 $11,438 $12,056 $12,371 $12,675 $13,273 $13,377 $15,132 $15,896 $15,696 $15,533
SOURCES: National Center for Education Statistics (NCES), unpublished IPEDS enrollment data.
This table was prepared in October 2013.
Table reproduced from CollegeBoard.org, http://trends.collegeboard.org/student-aid/figures-tables/ten-year-trend-student-aid-and-nonfederal-
loans-fte
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 8
Table 3 shows the trend in student aid and loans over the decade from the 2002–2003 to
2012–2013 for each full-time equivalent (FTE) student. Whereas the average amount of Federal
Pell Grants has increased from approximately $1,200 to $2,000, federal loans have risen more,
from $4,200 to $6,400 (College Board, 2013a). Students today are borrowing more than before
to finance their education.
To accommodate and compete for students, private non-profit and public HEIs have built
bigger campuses, employed more faculty and administrative staff, and offered more classes and
services. This has led to rises in tuition costs, which escalated during the 1980s and 1990s to
reflect the burgeoning costs of operations, and have persisted since. Public universities also have
had to grapple with continuous cuts in state funding for higher education; one solution has been
to increase student tuition and fees to cover state budget shortfalls (Urahn & Conroy, 2015).
Tuition at FPCUs has also risen in part due to overall higher costs of college, but also in part to
capitalize on the rising trend in tuition costs; to maximize profits, FPCUs charge what the market
can bear. As a result, tuition cost increases have outstripped consumer prices by a margin of 3 to
1 between 1985 and 2011 (Rampell, 2012).
Tuition and fees at higher education institutions have risen at average rates between an
inflated-adjusted 2% (for public in-state HEIs) to 4% (for private non-profit HEIs) every year
over the past two decades (College Board, 2013b; NCES, 2011a). While the annual average
increase may seem small, tuition increases have outstripped household income growth, making a
college education more and more expensive (Davidson, 2015). Despite increases in federal and
state funding for financial aid, such provisions often cannot cover the total cost of a college
education. During the period from 2000–2001 to 2010–2011, prices for undergraduate tuition,
room, and board at public institutions rose 42%, and prices at private non-profit institutions rose
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 9
31%. For 2010–2011, annual prices for undergraduate tuition, room, and board were estimated to
be $13,564 at public institutions, $36,252 at private not-for-profit institutions, and $23,495 at
private for-profit institutions (NCES, 2011b). Students on average receive from the government
$8,400 per year ($2,000 in grants and $6,400 in loans) in financial aid, but the average cost of a
year of college at a public four-year university is over $13,000; this leaves $4,600 to be paid out
of pocket by the student.
When federal grants and loans do not cover the total price of college, students try to
negotiate institutional discounts on tuition. They also resort to borrowing from private lenders,
such as banks, where interest rates may be higher and repayment terms more stringent. The
private loan option was not a popular or necessary one until 1996–1997, when only 6% of all
loans ($45 billion) were private. Over the next 11 years, dependence on private loans grew to
reach a peak of 25% of total loans in 2007–2008. It has since decreased to 8%, as the
government has increased the amount of federal unsubsidized loans available to students. In
2012–2013, 50% of all loans were federal unsubsidized loans, 25% were federal subsidized
loans, and the remainder was more or less evenly divided among non-federal (private), parent
PLUS (parent loans for undergraduate students), and graduate PLUS loans (College Board,
2013b).
Student Debt, Repayment and Default
The increasing level of student loan debt brings with it the challenge of adequately
serving loan repayments and, failing that, the problem of loan defaults. The public media have
recently brought much attention to the fact that the total amount of student loans outstanding in
the U.S. crossed the $1 trillion threshold in 2012 (Laing, 2012; Block, 2012; Goodale, 2012),
reaching $1.2 trillion in 2013 (Weismann, 2013). This figure is substantially more than
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 10
outstanding credit card loans ($704 billion) or auto loans ($734 billion) (Lam, 2012; Federal
Reserve Bank of New York, 2012). This total loan amount increases as more students enter
college, and therefore more students borrow. The rise in total loan amounts is further amplified
because of the higher average loan amount per student. The average loan amount borrowed per
full-time equivalent (FTE) student was $1,116 in 1972–1973, and it reached a peak of $6,900 in
2009–2010 before coming down to $6,400 in 2012–2013 (College Board, 2013a).
Students are borrowing more because their family household incomes have not kept up
with the rise in college tuition. From 1987 to 2008, the percent increase in real income by
quintile for the first four quintiles was 5.8%, 4.9%, 8.1%, and 14.3%, respectively. Real net price
at public two-year institutions, on the other hand, outpaced real income growth by 11.0%, real
net price at public four-year institutions grew by 37.4%, and real net price at private four-year
institutions grew by 36.6% (Martin & Gillen, 2011). This indicates that, regardless of the actual
household income in dollars, the share of college tuition as a percentage of household income
has increased.
The more worrisome situation, however, is the rise in the number of students failing to
repay their loans, both in terms of the total number as well as the percentage of students who
default. A measure of this is the cohort default rate (CDR), defined as the percentage of an
institution’s student borrowers whose federal loan is more than 270 days delinquent. The two-
year CDR of all higher-education institutions rose from 7% in 2008 to 9.1% in 2010, and the
three-year CDR rose from 13.4% in 2009 to 14.7% in 2010 (U.S. Department of Education,
2012). The three-year CDR is more telling, because it indicates that more graduates were unable
to keep up with their loan repayments past the first two years. Breaking it down by institution
sector, the three-year CDR rose from 11% to 13% for public HEIs and from 7.5% to 8.2% for
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 11
private HEIs, while the CDR was highest in FPCUs, at 21.8% (even though it has already
decreased slightly from 22.7%) during the two years when the three-year CDR is available.
Although FPCUs constitute a small portion of the total higher education pie, they
nevertheless serve more than 2.2 million students (NCES, n.d.). Many of these students are older
than traditional students, are from less-educated families with lower incomes (Tierney &
Hentschke, 2007), and are demographically more likely to default on their student loans
regardless of the type of institution that they attend (Gross, Cekic, Hossler, & Hillman, 2009).
93% of students at accredited for-profit two-year and less-than-two-year institutions receive
some form of federal aid or loan (NCES, 2011a), and those institutions have a three-year CDR of
almost 22%—the highest default rate amongst all types of HEIs in the U.S. (U.S. Department of
Education, 2011a). Moreover, the growth of the CDR is particularly pronounced at FPCUs, with
the three-year CDR rising significantly from 19% in 2007–2008 to 22% in 2009–2010 (FinAid,
n.d.; Department of Education, n.d.).
Loan Default and the Economy
Loans are a central means for the government to provide financing for higher education
in the U.S. Therefore, an adequate understanding of the impact of existing loan policies is
important for future regulations to maximize access and equity for students, especially for those
who are disadvantaged and underserved. Loan default has ramifications not only for the
individuals who are directly affected, but also for higher-education institutions, for future
students, for taxpayers, and for the economy as a whole.
For individual borrowers, the Bankruptcy Abuse Prevention and Consumer Protection
Act of 2005 states that no qualified education loans can be discharged (FinAid, n.d.), which
means that such loans must be repaid either through the government garnishing directly from
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 12
wages, through deductions from federal benefits payments, or through other legal action.
Moreover, delinquent loans incur collection costs and destroy the borrower’s credit rating (U.S.
Department of Education, n.d.). Universities and colleges with default rates that either exceed 25%
for three consecutive years or reach 40% in one year are subject to sanctions that may include
losing eligibility for future Title IV (federal student loan) funding (Department of Education,
2011).
Of the total student loans outstanding in 2011, about 90% are federally administered, and
the rest are from private-sector lenders (House Committee on Financial Services, 2012). Federal
student loans that are not repaid become the responsibility of the taxpayer (Field, 2010).
Defaulted loans cost the U.S. federal government billions of dollars that could otherwise have
been put to use to fund future grants or other education priorities (Sallie Mae, n.d.). Finally,
future students may suffer the indirect consequences of rising default rates in the form of stricter
repayment rules set by the government, heavier non-repayment penalties, and perhaps a
shrinking total federal student loan budget.
Student loan defaults also have an unfavorable ripple effect on the economy (Pianin,
2012; Valenti, Edelman, & Van Ostern, 2013). Students who are about to enter college may
think twice about incurring loans for a higher education that they may not be able to afford to
pay back. This has the potential to lower the number of educated skilled workers in the labor
force, which may in turn weaken the competitiveness of the U.S. economy. Moreover, people
struggling to pay off their student debt have much less discretionary income to save toward big
purchases and investments, such as cars and houses. For a consumer-driven economy such as the
U.S., lower consumer spending equals lower gross domestic product and slower growth.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 13
At present, the annual increase in federal and state grant aid is insufficient to cover rising
tuition costs. Hence seven out of ten college students borrow from federal student loans in order
to meet all the expenses necessary to earn a degree. As long as graduates can find employment
that pays well enough for them to repay their loans in a timely manner, however, the benefits of
borrowing for a college degree far outweighs the cost. The problem occurs when student loans
are taken out for diplomas or degrees that do not add economic value. When earnings cannot
cover the loan repayment amounts, students fall behind in their payments and may eventually
default on their loans.
Statement of the Problem
What are the reasons for the high CDR at for-profit colleges and universities? Do the
students themselves have characteristics that somehow put them more at risk of defaulting? Or is
it because students at these institutions borrow more than what they can realistically pay back
after graduation? If so, are graduates from some majors more likely to default than others? Do
institutional factors explain any of the differences in default for students studying the same major
after controlling for student characteristics?
Many critics have spoken out strongly against FPCUs and raised questions about their
financial and educational practices. Critics point to the fact that 86% of revenues at FPCUs
derived from federal loans (Senate Committee, 2012) — as much as $32 billion dollars
nationwide—while more than half of their enrolled students dropped out before earning a
certificate or degree. Almost one in four students defaulted on their federal loans. The Senate
Committee on Health, Education, Labor and Pensions (HELP) requested the Government
Accountability Office (GAO) to conduct two undercover studies between 2010 and 2011 to
expose unethical behavior in recruitment and instruction. In two years, six Congressional
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 14
hearings were held, which generated six detailed reports that were widely covered in the media
(Senate Committee, 2012; GAO, 2010, 2011). All of these reports put FPCUs in an unfavorable
light. The reports stated that although FPCUs played an important part in higher education, they
failed to safeguard student success by not putting in place the necessary support services to
ensure graduation, job placement, and loan repayment. Federal funds were instead used
extensively for non-academic purposes, such as recruitment and marketing. The pressure to
enroll increasing numbers of students encouraged misrepresentation and fraudulent claims during
the recruitment process, and the pressure to graduate these students led to grade inflation. The
reports also raised doubts about the quality of the programs offered due to the high drop-out rate
as well as the inability of their graduates to secure gainful employment, even after course
completion—both factors that led to a higher incidence of students defaulting on loans.
In response to these charges, industry representatives have claimed that the CDRs of
FPCUs are on average no worse than most (and much better than some) other higher-education
institutions when the demographics of the students who are being served are considered. They
argue that a disproportionately large percentage of students attending FPCUs are students who
have many of the “at-risk” characteristics summarized by Horton (2015), which make them run a
high risk of failure and non-completion. Industry insiders allege that significant resources have
been invested in helping graduates manage their debts by providing them with financial literacy
instruction and also by tracking these students down to encourage them to address their debt
situations (Kirkham, 2012). Expense disclosures from the annual reports of publicly traded
FPCUs nevertheless reveal that more money is spent on marketing and executive compensation
than salaries for lecturers and counselors (Angulo, 2016).
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 15
Enrollment data from the National Center of Education Statistics (NCES) does show that
students enrolled in FPCUs are more likely to be non-traditional students (i.e. aged 25 years or
older) and more likely to belong to minority groups (NCES, 2015). These are two important
background characteristics that impact college persistence and success (Horton, 2015).
Race/ethnicity and graduation/completion rates are the two strongest predictors of loan
repayment and student loan default rates (Gross et al., 2009). Enrollment data from the
Department of Education shows that, over the years, FPCUs do enroll more minorities than their
traditional public and private non-profit counterparts, and this may contribute to the differences
in default rates (Woo, 2002a, 2002b; Steiner & Teszler, 2005). Studies have also found that
institution sector and level are associated with default rates (Deming, Goldin, & Katz, 2011;
Belfield, 2013; Hillman, 2014). What has not been studied in detail, however, is how the choice
of field of study (major), conditional on the institution sector, is associated with loan default
rates or the student’s ability to repay loans.
Both proponents and critics of FPCUs have cited evidence to support their conflicting
claims, yet little research has been done to address the contradiction. Although students at
FPCUs have greater student debt burdens and default rates than students from traditional HEIs,
the story is not so straightforward.
The institutional factor focused on in this study is institutional type or sector (public, for-
profit, or private non-profit). Recent studies using the Beginning Postsecondary Students survey
(BPS 04/09) database that includes a nationally representative sample of postsecondary students
and institutions find statistically significant association between default and institutional
characteristics (Hillman, 2014; Belfield, 2015).
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 16
Policymakers originally envisioned that federal student loans would remove economic
barriers for students for whom loan availability increases the likelihood of college enrollment
(Baum & Schwartz, 2013). It was expected that this educational opportunity would in turn open
up economic and social opportunities that would not otherwise be available to them. From 2007–
2012, seven out of 10 undergraduates borrowed across all institutions, and at FPCUs that number
jumped to 88%
(TICAS, 2014).
Indeed, federal loans have helped tens of millions of students over the years to achieve
the American dream of prosperity and success via higher education. In recent years, however, the
numbers suggest that borrowers may be saddled with high student loans that are becoming more
difficult to repay. 37 million Americans had student loans outstanding in 2012 (Brown,
Haughwout, Lee, Mabuta, & van der Klaaw, 2012). With the average outstanding loan per
borrower reaching $23,300 and a still-recovering domestic economy limiting job availability
(ADP, 2012; U.S. Department of Commerce, n.d.), rising student loan default is a social as much
as an economic problem. Not only does it leave billions of dollars out of circulation, but
borrowers’ standard and quality of living decline due to bad credit, less discretionary income,
and increased psychological stress. This is ironic given that the original intent of the Higher
Education Act was to make equal opportunity the “centerpiece of public policy toward higher
education” for needy students (Gladieux, 1995; Chen, 2008). Instead, today student loan debt
and default are contributing to the widest U.S. wealth gap since the 1920s (Trumbull, 2012;
Saez, 2012); this is exactly the opposite of what was intended when Title IV federal student
loans were first made available.
The rising student loan debt burden, and the increasing loan default rate accompanying it,
need to be addressed by studying why students default, to what extent do they default, and if it is
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 17
their choice of major combined with their choice of education institution that is associated with
their decision to default. Since previous studies show mixed results with regard to the
significance of institutional variables affecting loan default—although more recent studies have
found institutional variables to be a factor—additional research will further clarify this debate.
This topic is timely because the rapid growth in for-profit higher education has only occurred in
the past 20 years. Loan default literature from previous years drew on data collected in the early
1990s that did not contain sufficient numbers of FPCUs, thus posing limitations on the external
validity in explaining differences in that sector. Given the current changing political climate,
deregulation in the for-profit higher education sector is likely, and there may be a rapid increase
in the number of for-profit institutions again. Since FPCUs are now a vital and growing player in
the U.S. higher-education system, that gap in the literature needs to be filled.
Purpose of the Study
This study is, to my knowledge, the first to segregate loan default data by program type,
and it serves the purpose of providing a more detailed picture of the increasingly significant
student loan and default issue at hand. The study delves deeper beyond the aggregate student
loan default landscape by investigating whether default rates differ for students pursuing
different majors. Additionally, this research ascertains whether students who graduated with the
same major are more or less likely to default depending on whether their degrees are from for-
profit rather than public or private non-profit HEIs.
The study aims to answer the following questions:
1. To what extent do total cumulated student loans differ across fields of study (both
career/vocational and degree programs)?
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 18
2. To what extent do student loan default rates differ across fields of study (both
career/vocational and degree programs)?
3. Controlling for field of study, to what extent do institutional factors explain
differences in the likelihood of default?
4. To what extent do different fields of study affect graduates’ employment, and
how does this differ among different types of institutions?
The findings from this study will identify the specific programs (majors) in which
graduates have a higher likelihood of defaulting on student loans. By examining default rate on
the level of program of study, this research will provide a more focused perspective that has not
yet been explored. If the likelihood of student loan default does vary from major to major
regardless of institution sector, then this implies that there are majors for which the costs of loans
outweigh the benefits of a postsecondary education. This result would offer additional
information that students can use to make better choices prior to enrollment. They can perhaps
choose alternative means of financing their education, or decide to enroll in a different program
or major altogether. If differences are found in the degree to which students are likely to default
on their loans at different institutional types—while controlling for major—then students may
want to choose to attend the institutions that set them up with better prospects of loan repayment.
Federal policymakers and administrators at HEIs can build on these findings by launching
further investigations into the reasons behind differentials in default rate, finding ways to
improve loan repayment outcomes either through regulations or internal audits.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 19
Chapter 2 – Literature Review
This chapter is a review of current literature on student and institutional characteristics
influencing students’ ability to repay loans. It is divided into two sections: student characteristics
(including demographic characteristics, academic preparation and attainment, choice of major,
student knowledge of loan debt, debt load, transfer and program completion, and employment
after graduation) and institutional characteristics. It concludes with a discussion of the gap in
literature which is the focus of this dissertation.
Student Characteristics
Race/Ethnicity and Socioeconomic Status
Studies have shown that ethnicity is one of the best predictors of student loan default.
African-American and Hispanic/Latino students are more likely to default than other ethnicities
(Woo, 2002a; Woo, 2002b). In a meta-study that aggregated data from student loan literature
prior to 2006, race/ethnicity was found to be one of the main variables predicting student loan
default (Gross, Cekic, Hossler, & Hillman, 2009). More recent studies have cast doubt on these
longstanding findings. One study found that students from lower and middle socioeconomic
backgrounds had a higher risk of defaulting on student loans than those from families with
higher socioeconomic status, regardless of race (Houle, 2013). However, Goldrick-Rab, Kelchen,
and Houle (2014) cautioned that it is not always possible to speak of class and race separately;
students of lower SES are also often African-American and Hispanic minorities.
Thus, racial and/or ethnic characteristics point to other associated factors that predict the
likelihood of default, especially socioeconomic status. Students from low-income families
accumulate more debt than those in higher brackets (Steiner & Teszler, 2005; Houle, 2013), and
the likelihood of default decreases as family income increases (Woo, 2002a). First-generation
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 20
college students are more likely to default than students with college-educated parents, who have
higher average earnings and thus higher socioeconomic status than parents with a high-school-
level education (Choy & Li, 2006). Although Choy and Li also found that a rising amount of
debt increased the likelihood of default, the latest studies have shown that default is highest
amongst students who borrow less than $10,000 (Dynarski, 2015; Looney & Yannelis, 2015).
To highlight the complexity of the issue, Chen and DesJardins (2010) found that Pell
Grants and subsidized loans lowered the probability of low-income students dropping out of
college, but the current provision of federal student aid does not meet the needs of minority
students. Although federal student grants are awarded based on household income, African-
American and Hispanic/Latino students are less well served than White students because they
earn less. 50% of African-American families earn less than $35,000 per year, compared to 41%
of Hispanic/Latino families and 31% of White families in 2015 (Proctor, Semega, & Kollar,
2016). More than half of African-American students (56%) and Hispanic/Latino students (58%)
had unmet needs after all aid was considered, compared with 40% of White students (Long &
Riley, as quoted in Chen & DesJardins, 2010). The study suggested that even though minorities
may have received financial aid, they did not receive a sufficient amount to meet their needs,
leading to an increased tendency to drop out. Since degree completion is strongly associated with
lower student loan default rates (Woo, 2002a, 2002b; Dynarski, 2015), failure to complete a
degree (i.e., dropping out) increases the likelihood of default.
Age
Age is also an important factor to consider with regards to default. The majority of
studies have reported that as age increases, the likelihood of default rises as well (Podgursky,
Ehlert, Monroe, Watson, & Wittstruck, 2002; Woo, 2002a, 2002b; Steiner & Teszler, 2005),
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 21
although one study found evidence to the contrary (Steiner & Teszler, 2003). It may be that older
students have families and dependents to support. The more dependents a student has, the more
likely he or she is to default (Woo, 2002a). Similarly, marital status is also a determinant. Older
students with more years of life experience are more likely than younger students who enroll in
colleges directly from high school to be married and have children. On the other hand, students
who have access to family support, regardless of age, are less likely to default than those without
family assistance (Woo, 2002a).
For-profit colleges and universities (FPCUs) educate a disproportionate number of non-
traditionally aged students compared to public and private non-profit institutions. More than half
(58%) of students enrolled in FPCUs in 2007–2008 were older than the traditional college age
(18–24 years old), compared to 30% in public and 25% in private non-profit HEIs. Older
individuals tend to have family responsibilities and dependents, which is a factor contributing to
the likelihood of default.
Gender
Gender is a less definitive predictor of loan default. Although men are more likely to
default on loans than women (Podgursky et al., 2002; Woo, 2002a), women take longer to repay
loans (Choy & Li, 2006). The gender pay gap may contribute to women having more difficulty
in repaying loans quickly (Corbett & Hill, 2012). However, other studies have found the
difference in loan default/repayment to be statistically insignificant (Harrast, 2004).
Academic Preparation and Attainment
If a student borrows for college but drops out without graduating, he or she is expected to
have lower lifetime earnings than a college graduate (Carnevale, Rose, & Cheah, 2011). Lower
earnings make it more difficult for these students to repay their loans, which may lead to
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 22
delinquency and default. Any measures that improve college persistence and completion will
probably decrease student loan default rates (Steiner & Teszler, 2003).
Completing a higher-education program with a degree or certificate is another one of the
strongest predictors of loan repayment (Woo, 2002a; Gross et al., 2009). The link between
completion of an academic program and default also extends back to high school, as students
who enroll in a postsecondary program with a GED or after dropping out of high school are more
likely to default than those with a high-school diploma (Huelsman, 2015).
Since completing a degree or certificate program is positively associated with loan
repayment, the factors that are associated with successful degree completion also predict better
outcomes for loan default. The better a student’s academic rank in high school, GPA, and
standardized test scores, the less likely he or she is to default on loans (Podgursky et al., 2002;
Woo, 2002a; Steiner & Teszler, 2003).
Choice of Major
Studies have found that earnings after graduation related to field of study affect personal
income and one’s ability to repay loans (Steiner & Teszler, 2005). When it comes to post-
graduation employment prospects and earnings, not all college majors are equal (Carnevale,
Strohl, & Melton, 2011; Carnevale, Jayasundera, & Hanson, 2012). Petroleum engineering,
pharmaceutical sciences and computer science have the highest median earnings of $120,000,
$105,000, and $98,000, respectively. Counseling psychology is the lowest-paying bachelor’s
degree major, with a median of $29,000, followed by early-childhood education, with median
earnings of $36,000, and theology/religious vocations and human services and community
organization, which both have median earnings of $38,000. The cost of a degree is also
differentiated by the level of education (e.g., certificate, associate, or bachelor’s degree) as well
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 23
as by major. It is reasonable to expect that a four-year education culminating in a B.A. in early-
childhood education costs roughly the same as a four-year B.Sc. degree in petroleum science
from the same higher-education institution. Assuming that students in both majors borrow the
same amount of money to finance their education, their repayment amount is the same. But, the
graduate with the B.A. degree will find more hardship repaying the loan, because servicing the
loan will take up a larger proportion of his or her disposable income.
The educational costs of the same degree (e.g., a B.Sc. in engineering) can also vary
depending on the type of institution in which the student enrolls. The 2016–2017 annual tuition
for an undergraduate student in California, for example, would be approximately $47,500 at
University of Southern California (USC), a private non-profit, $13,000 for any University of
California (UC) campus, approximately $6,500 at a California State University campus, $1,100
at a community college, and $22,000 at a for-profit university
1
. This wide range in tuition costs
can be somewhat mitigated not only by federal, state and institutional grants, but also by taking
into account the time it takes to complete an undergraduate degree, as well as the retention rate
(the percentage of students who continue after the first year) at each institution. USC has a
retention rate of 97% and a six-year graduation rate of 92% (USC, 2016). The UC system has an
average of 95% retention, and the six-year graduation rate is 84%; thus, more students take
longer to complete their degree (University of California, 2016). 80% of California State
University students continue past their first year, and only 51% graduate within six years
(CalState, 2015). Although the earnings differential among institution types will not be as drastic
as graduating with different majors, there will still be an upper and lower limit. The wage
1
The average annual cost of a for-profit university is derived from the program costs of $88,000 for 139 credits at
DeVry University and divided over four years.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 24
differential in the case of an engineering major can be as much as $49,000 between the median
earnings of the 25
th
and 75
th
percentile (Carnevale, Strohl, & Melton, 2011). However, although
there is evidence that graduates at for-profits earn less than those from traditional higher-
education institutions (HEIs) on average (Deming et al., 2011), there is no study to address
whether the differences are localized within specific majors.
Belfield (2013) estimated loan status (loan repayment rate) by looking at the largest CIP
(classification of instructional programs) code offering at less-than-two-year and two-year for-
profit colleges. Among the seven fields of study analyzed, only the beauty and cosmetology
sector had a statistically different loan balance per student full-time equivalent (FTE). Students at
two-year for-profit colleges carried double the loan balance of those who enrolled at less-than-
two-year for-profit institutions. Nursing programs had the slowest repayment rates at both levels
of FPCUs. However, Belfield’s model used only one program of study for each for-profit college
type, and his study focused mainly on loan repayment rather than on default rates.
Student Knowledge of Loan Debt
Gross et al. (2009) identified a factor that had been relatively unexplored in previous
studies, but had been increasingly embraced by policymakers in recent years. This is students’
knowledge about and attitudes toward debt and repayment. Existing studies have presented
mixed results. One study’s findings showed that ignorance about the borrowing process was
related to default (Christman, 2000). Loan counseling has been found to be associated with lower
rates of default (Podgursky et al., 2002; Steiner & Teszler, 2005), but a recent study has found
that institutional expenditures on academic and student support are not significantly associated
with decreased rates of student loan defaults (Ishitani & McKitrick, 2016). This, along with the
increasing debt load experienced by postsecondary graduates in the past decade, has prompted
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 25
the federal government to mandate all colleges and universities to make available online the real
costs of education in as clear and as transparent a manner as possible.
Debt Load
Although Avery and Turner (2012) argued that there was little evidence that the average
burden of loan repayment had increased in recent years, they did highlight that borrowing
behavior for students at FPCUs was higher in all levels of credit attainment than at traditional
HEIs. It seems logical to consider that the higher the debt load a student has, the more likely he
or she is to default, due to the burden of repaying it. However, findings have been mixed in this
regard. While one finding identified a positive relationship between a student’s total debt and
default (Choy & Li, 2006), another reported an inverse relationship—that is, a student with a
higher level of debt was less likely to default than another with a lower level of debt. Hillman
(2014) confirmed a non-linear (U-shaped) relationship between debt amount and default rate in
his model, finding that default rates were higher at either end of the borrowing spectrum. The
tendency of low debt amounts to be associated with higher default rates may be explained by
considering students who drop out of a postsecondary program without a degree, but with some
loans that they cannot repay. The correlation between high total debt and high default rate can be
more intuitively explained by the situation of those who stay longer to complete their degrees,
but accrue annual loans that add up to an unserviceable total.
Transfer and Program Completion
Studies on the relationship between student transfer and loan default have been mixed.
While one study has shown that students who receive transfer credits are less likely to default
(Gross et al., 2009), another study presented contradictory results. Woo (2002a, 2002b) found
that students who attained bachelor’s degrees at one institution were less likely to default than
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 26
those who transferred, and Herr and Butt (2005) also found that students were more likely to
default when they transferred credits.
Gross et al. (2009) identified college completion as the strongest predictor of loan
repayment (and lower default rates). Adelman (2006) identified five factors that consistently
contributed to minority students’ bachelor’s degree completion, focusing on interventions that
were most responsive to change by external parties with little to modest effort: (1) 20 or more
credits in the first year of enrollment; (2) elimination of no-penalty withdrawals and no-credit
repeats; (3) use of summer terms; (4) no delay of entry (after high school); and (5) adequate
high-school academic curriculum. Because of the strong relationship between completion and
successful loan repayment, the above factors also should predict loan default. Results from past
studies support this assumption.
Students who enroll in more credit hours, those who complete their attempted courses,
and those who are continuously enrolled are less likely to default (Harrast, 2004; Steiner &
Teszler, 2005). However, the longer the student is enrolled, the chances of default increase
(Podgursky et al., 2002), probably due to the fact that more loans are borrowed each additional
year, adding to the total amount to be repaid.
Employment after Graduation
Graduates from FPCUs on average earn less than those from traditional two- and four-
year public and private non-profit institutions (Deming et al., 2011). There are statistically
significant positive effects of a certificate or degree from public and private non-profit
universities on earnings outcomes, but not from FPCUs (Lang & Weinstein, 2012). There is also
a higher probability of FPCU graduates being unemployed for longer after leaving school.
Students who attended FPCUs and later defaulted cited unemployment as the cause for default in
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 27
higher proportion than defaulters who attended other types of institutions (Dynarski, as quoted in
Gross et al., 2009).
Institutional Characteristics
There are obvious differences in the educational and debt outcomes of students
graduating from different types of HEIs in the U.S. Available data show that the average six-year
graduation rate at public HEIs was 57% in 2011 (National Center for Education Statistics, n.d.),
and the average student in 2008 incurred $20,600 in loans (Project on Student Debt, 2010).
Public HEIs also had an average three-year CDR of 13% as of 2013 (U.S. Department of
Education, 2013). In the same period, private non-profit HEIs had an average graduation rate of
65%, average loans per student of $27,650, and a three-year CDR of 8.2%. For-profit HEIs had
the lowest graduation rate of the three (42%) and the highest loan amount per student ($33,050),
as well as the highest three-year CDR (21.8%). Since there is such a disparity of student loan
default figures amongst these three institutional types, it is reasonable to hypothesize that
institutional type is associated with loan default. Examining default rates by institutional level,
Kantrowitz (2010) finds that default rates are highest at less-than-two-year institutions (25.7%),
followed by two-year colleges (10.7%), and then four-year colleges and universities (6.6%),
using data from Beginning Postsecondary Students survey (BPS 96/01).
Studies that have used ordinary least-squares regression have found that there is no
statistically significant difference in default rates among varying types of institutions (Knapp &
Seaks, 1992; Volkwein et al., 1998), after taking into account student background characteristics
and institutional resources. Their results suggested that student characteristics explained most of
the variation in loan default, and that institutional characteristics had insignificant effects. It is
well documented that for-profit HEIs educate a larger portion of minorities, disadvantaged
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 28
students, and older non-traditional students as compared to public and private non-profit
institutions (Knapp, Kelly-Reid, & Ginder, 2010; Deming et al., 2011), and these characteristics
strongly predict a higher likelihood of student default. However, each of these studies has its
own limitations with regards to data. Both Knapp and Seaks (1992) and Woo (2002a; 2002b)
utilized data from one state only. Woo used data from California and Knapp and Seaks used data
from Pennsylvania; therefore their findings are not representative of the national population.
Steiner and Tezsler’s (2005) data were even narrower, as they studied one higher-education
institution only (in Texas), and Harrast (2004) also based his study on only one university. Their
data, therefore, can only explain student loan default patterns within a specific population, not
necessarily extending to the general population. Volkwein and colleagues’ studies (1995, 1998)
employed data from three sources: the National Postsecondary Student Aid Study, the Integrated
Postsecondary Education Data System, and College Board surveys. Although their findings are
widely cited, their studies were completed over 15 years ago and do not reflect the reality of
national college-going trends and the behavior of students attending higher-education institutions
today.
More recent studies using the nationally representative Beginning Postsecondary
Students (BPS) longitudinal survey data (for years 2004–2009) have provided new results that
are contradictory to the earlier findings. Deming et al. (2011), Belfield (2013), and Hillman
(2014), all using the BPS 04/09 database, have found that, even controlling for student
characteristics, students at FPCUs still have much greater student debt burdens and loan default
rates than those at traditional HEIs across institutional type (for-profits, public, and private non-
profits). These studies also based their calculations on more sophisticated statistical methods,
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 29
such as hierarchical linear modeling and propensity score matching, which reduced the standard
errors of the results.
Belfield (2013) used stepwise regression to explore the default rates and loan repayment
rates across institution sectors and levels. Belfield reported that, while loan balances and default
rates were highest at FPCUs (including both two-year and four-year institutions), there was no
statistically significant difference in default rates after accounting for student characteristics,
financing options, and institutional practices. However, large loan repayment rate differences
remain, with loan repayment rates at FPCUs significantly lower than at public and private non-
profit HEIs. On the other hand, such institutional effects are present and statistically significant
in studies that use propensity score matching (Deming et al., 2011) and hierarchical linear
modeling (Hillman, 2014).
Gaps in the Literature
Studies have exhaustively documented student characteristics that increase the likelihood
of loan defaults. Institutional factors, however, have not been given much emphasis until recently.
Due to the limitations of the databases used in many previous studies, researchers did not have
sufficient information to study institutional factors. The most widely cited articles on loan default,
as aggregated in Gross et al.’s (2009) meta-study, used national databases that did not include
non-traditional students. This in effect excluded data from many for-profit institutions, because
non-traditional students tend to enroll in FPCUs due to more open enrollment policies and
straightforward admission processes.
Since 2010, more studies focusing on for-profit institutions have begun to appear in
journals because of the availability of national databases such the BPS 04/09 survey. By
including non-traditional and transfer students, the BPS survey is able to capture a sample
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 30
population that contains students who enroll in FPCUs. All studies using these wider data sets
find institutional effects to be to some extent associated with loan default or loan repayment
timelines. Although the field of study (major) is included in the statistical models, it is used
mainly as a control factor. Since there is a wide wage differential among and between majors
(Carnevale, Strohl, & Melton, 2011), it is reasonable to ask whether these correlations remain
after controlling for student characteristics. Or, does the value of a given major (i.e., to find
employment and earn higher wages) vary from one institution sector to another? This is a gap in
the existing literature and I now turn to how I will study the issue in Chapter 3.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 31
Chapter 3 – Methodology/Data Collection/Analysis
Research Questions
This study seeks to understand loan default behavior among students who borrow federal
loans. It does this by delving deeper beyond the aggregate student loan default landscape. By
investigating which particular majors and career programs are associated with a higher likelihood
of default across institution sectors (for-profit, public, and private non-profit), the study aims to
answer the following questions:
1. To what extent do total cumulated student loans differ across fields of study (both
career/vocational and degree programs)?
2. To what extent do student loan default rates differ across fields of study (both
career/vocational and degree programs)?
3. Controlling for field of study, to what extent do institutional factors explain
differences in the likelihood of default?
4. To what extent do different fields of study affect graduates’ employment, and
how does this differ among different types of institutions?
Three-level hierarchical linear modeling is used to address these questions. In addition, a
propensity score matching algorithm is performed with relevant covariates, before running the
calculations with Stata 14 statistical software. The variables used are based on findings in
existing literature as discussed in Chapter 2.
Population and Sample
This study uses the Beginning Postsecondary Students (BPS) Longitudinal Study as its data
source. BPS is a sample survey of first-time beginning postsecondary students conducted by the
National Center for Education Statistics (NCES). It includes both traditional and non-traditional
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 32
students. The survey draws its cohorts from the National Postsecondary Student Aid Study
(NPSAS), which collects financial data on nationally representative cross-sectional samples of
postsecondary students. NPSAS provides the base-year data for first-time beginning students,
and BPS follows them through their higher-education years and into the workforce. The cohort
used here is the third BPS cohort (2004–2009; hereafter BPS 04/09), selected from NPSAS:04
(i.e., base year 2004), and it includes students who began their postsecondary education in 2004.
The first follow-up survey by BPS was in 2006, and the second was in 2009.
Aside from data regarding comprehensive student and institutional characteristics, BPS
04/09 also includes information on students’ major fields of study, reasons for school selection,
employment, income, and occupations. Specifically, it includes data for students who entered
postsecondary programs after a break from high school—who by definition will be older
students—and those who received a GED or another type of high-school completion. Because
these student are served to a greater extent by for-profit colleges and universities (FPCUs), there
is a greater representation of those institutions in this survey.
BPS is nationally representative in that it includes students in nearly all sectors and levels
of higher-education institutions (HEIs) located in the 50 states. As such, the findings of this
study reflect the population of postsecondary students in the U.S. BPS 04/09 consists of 18,640
first-time beginning students from 1,360 institutions. During the second BPS follow-up in 2009,
a total of 15,160 of these students gave full or partial interviews, largely due to the inability to
locate, nonresponses, and exclusions. The large sample size and the high response rate (81.3%)
reduce errors resulting from bias. Weighting is used to adjust for unit nonresponse. This study
will use weighting provided in the database.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 33
Item nonresponse rate is low. Nineteen of 385 items in BPS 04/09 have total nonresponse
rates over 5%. Missing data for all variables on the data file were imputed by NCES prior to the
full database being made available to the public.
Hierarchical Linear Modeling
Hierarchical Linear Modeling (HLM) is used to address the research questions. There are
three dependent variables: cumulative federal loans (FEDLOAN), loans status (LOANST09), and
employment status (JOBST09). FEDLOAN is a continuous variable, and the analysis uses the
multilevel mixed effects model. Since the dependent variable LOANST09 is a categorical
variable with 4 categories (0 = no loans, 1 = loans paid in full or cancelled, 2 = loans in
repayment, 3 = loans deferred/in forbearance/in default, and 4 = not in repayment), and JOBST09
is a dichotomous variable (Yes, No), the analyses of these variables use the multinomial logit
intercept- and slope-as-outcome model (Raudenbush & Bryk, 2002; Rabe-Hesketh & Skrondal,
2012).
The reasons for using HLM to study the program (major) and institutional effects on loan
repayment status are twofold. First, students enroll in different types of HEIs for reasons such as
socioeconomic situation, race or ethnicity, and educational preparation. The institutions that they
choose may indicate that students in the same institutions are more similar compared to those in
others. Further, students within a particular institution share similar experiences, which may lead
to increased homogeneity over time. Testing hypotheses about the relationship between variables
at two or more levels is one of the primary uses of HLM (Bryk & Raudenbush, as quoted in
Perna & Titus, 2005).
Secondly, HLM is appropriate for statistical reasons because a failure to account
statistically for different units of analysis (i.e., student, major, and institution) can lead to
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 34
aggregation bias, miscalculation of standard errors, and heterogeneity of regression (Bryk &
Raudenbush, as quoted in Perna & Titus, 2005). HLM is superior to ordinary least squares
(OLS/General Linear Models) because HLM theoretically produces appropriate error terms that
control for potential dependency due to nesting effects, while OLS does not (Newman, Newman,
& Salzman, 2010).
The odds-ratio of a logit regression represents the change in the odds of a student
defaulting on federal loans in a field of study relative to the undeclared major, which will be the
reference category in this study. An odds-ratio greater than 1 represents an increase in the
likelihood of loan repayment for those in a particular major or type of institution, whereas an
odds-ratio less than 1 represents a decrease in the likelihood of repayment. This study uses a
three-level hierarchical structure to explore the data: students (level 1) in majors (level 2) across
HEIs (level 3). The multi-level model examines the relative importance of institution sector,
institution level, and declared major as influences on students’ likelihood to default on loans.
Propensity Score Matching
Some researchers will argue that institutional characteristics do not factor into the
likelihood of student loan default because students self-select into different kinds of institutions.
The unobserved differences, they argue, must therefore derive from the students’ own
background characteristics. One of the research questions in this study seeks to isolate the effect
of graduating from a for-profit HEI on the likelihood of loan default. Although an OLS
regression may show a statistically significant relationship between the dependent and
independent variables, confounding variables can contribute to omitted-variable bias and affect
the results. It may be that a student who enrolls in a FPCU is more likely to default because he or
she has poorer planning skills (e.g., he or she did not take the courses or tests required for
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 35
admission to public HEIs, and later as a borrower he or she did not budget well for repayment).
Moreover, the choices that students make in selecting majors or schools may lead to selection
bias, thus leading the study to overestimate the impact of institutional factors (Murnane &
Willett, 2011).
Since it is impossible to randomly assign students to majors and institutions, the
propensity score matching method is added to adjust for bias. It aims to mimic randomization by
statistically equating subsets of units on all observed covariates to control for differences, thus
making the groups of students receiving treatment (enrollment at FPCU) and not receiving
treatment (enrollment at public and private non-profit HEIs) more comparable (Rosenbaum &
Rubin, 1983). Ideally, matching should include all student characteristics that previous studies
have identified to affect student loan default; this enables the study to find the best possible
matches. For this study, I will use the level 1 (student) covariates available in BPS 04/09 that
have been shown to be predictors of default in existing literature. I will employ the nearest-
neighbor matching algorithm in the statistical software, Stata 14, to get a matched equation. The
strategy is to examine situations in which students, based on observed covariates, have a similar
“propensity” of receiving treatment, yet one of them received treatment and the other did not.
This identifies a matched sample, which consists of the observations in the treatment group and
the observations in the control group. After matching the data in this way, I will run the HLM
model as detailed below.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 36
The Model: Question 1
Propensity Score Estimation
Score
i
treat
= α
0
+ α
1
major
i1
+ α
2
Male
i2
+ α
3
Nontrad
i3
+ α
4
Race
i4
+ α
5
ParentEd
i5
+ α
6
HSGPA
i6
+
α
7
TeSATder
i7
+ α
8
ColGPA
i8
+ e
ijk
Hierarchical Linear Modeling (FEDLOAN as DV)
Level 1 (student):
y
ij
~ Bernoulli (φ
ij
)
logit (φ
ij
) = LOANST09
ij
FEDLOAN
ijk
= π
0jk
+ π
1jk
major
ijk
+ π
2jk
Male
ijk
+ π
3jk
Nontrad
ijk
+ π
4jk
Race
ijk
+ π
5jk
ParentEd
ijk
+
π
6j
HSGPA
ijk
+ π
7j
TeSATder
ijk
+ π
8jk
ColGPA
ijk
+ e
ijk
Level 2 (major):
π
0jk
= β
00k
+ β
01k
Major
0jk
+ r
0jk
π
1jk
= β
10k
π
2jk
= β
20k
π
3jk
= β
30k
π
4jk
= β
40k
π
5jk
= β
50k
π
6jk
= β
60k
π
7jk
= β
70k
π
8jk
= β
80k
Level 3 (institution):
β
00k
= γ
000
+ γ
001
FP
00k
+ γ
002
Locale
00k
+ γ
003
FP
00k
*Major
0jk
+ Score
i
treat
+ u
00k
β
10k
= γ
100
β
20k
= γ
200
β
30k
= γ
300
β
40k
= γ
400
β
50k
= γ
500
β
60k
= γ
600
β
70k
= γ
700
β
80k
= γ
800
β
01k
= γ
010
Combined model with treatment effects estimation:
FEDLOAN
ijk
= Score
i
treat
+ γ
000
+ γ
001
FP
00k
+ γ
002
Locale
00k
+ γ
003
FP
00k
*Major
0jk
+ γ
010
Major
0jk
+
γ
100
major
ijk
+ γ
200
Male
ijk
+ γ
300
Nontrad
ijk
+ γ
400
Race
ijk
+ γ
500
ParentEd
ijk
+ γ
600
HSGPA
ijk
+
γ
700
TeSATder
ijk
+ γ
800
ColGPA
ijk
+ e
ijk
+ r
0jk
+ u
00k
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 37
where i denotes student, j denotes major, and k denotes institution.
The dependent variable, FEDLOAN, is a continuous variable equal to the loan amount
borrowed for college. Gender is a dummy variable, with Male being the reference category.
Nontrad is an age variable (whether the student is a traditional (<25 years old) or nontraditional
student (≥25 years old]; NCES, n.d.). Race is a set of seven dummy variables, with each set
representing White, Black or African American, Hispanic or Latino, Asian, Native Hawaiian or
Pacific Islander, American Indian or Alaska Native, and Two or more races/Other. A value of (1)
denotes that the student is of a particular race, and (0) indicates that the student is not. Parent
education (ParentEd) is a categorical variable that indicates if a student’s parents were educated
at a high-school level or lower (0), which is the reference category, if the parents had some but
less than four years of college (1), or if the parents had a bachelor’s degree or above (2). HSGPA
is a categorical variable that describes the student’s grades from high school; levels include (1)
D- to D, (2) D to C-, (3) C- to C, (4) C to B-, (5) B- to B, (6) B to A-, and (7) A- to A. TeSATder
is a continuous variable that describes the student’s SAT scores from 400 to 1600. Finally,
ColGPA is a categorical variable that describes the student’s grades in college following the
same scale as HSGPA.
Level 2 (major) contains the explanatory variable Major. This is a categorical variable for
which the reference is undeclared (0), with values for humanities (1), social sciences/behavioral
sciences (2), math/science (3), computer/engineering (4), education (5), business/management
(6), health (7), and vocational/technical fields (8).
Level 3 (institutional) characteristics also contain explanatory variables. This set of
variables includes a category variable (FP) indicating whether the HEI is public (0, the reference
category), private non-profit (1), or for-profit (2). Locale is a categorical variable distinguishing
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 38
between town (0, the reference category), city (1), and rural (2). The treatment effects estimation
(Score
i
treat
) is also added into the equation in this level.
Bivariate analysis will describe the relationship between field of study, institution sector,
and failure to repay loans. The regression results from the combined model will answer the first
research question: To what extent do total cumulated student loans differ across fields of study
(both career/vocational and degree programs)?
The Model: Questions 2 and 3
Hierarchical Linear Modeling (LOANST09 as DV)
Level 1 (student):
y
ij
~ Bernoulli (φ
ij
)
logit (φ
ij
) = LOANST09
ij
LOANST09
ijk
= π
0jk
+ π
1jk
major
ijk
+ π
2jk
Male
ijk
+ π
3jk
Nontrad
ijk
+ π
4jk
Race
ijk
+ π
5jk
ParentEd
ijk
+
π
6jk
HSGPA
ijk
+ π
7jk
TeSATder
ijk
+ π
8jk
ColGPA
ijk
+ e
ijk
Level 2 (major):
π
0jk
= β
00k
+ β
01k
Major
0jk
+ r
0jk
π
1jk
= β
10k
π
2jk
= β
20k
π
3jk
= β
30k
π
4jk
= β
40k
π
5jk
= β
50k
π
6jk
= β
60k
π
7jk
= β
70k
π
8jk
= β
80k
Level 3 (institution):
β
00k
= γ
000
+ γ
001
FP
00k
+ γ
002
Locale
00k
+ γ
003
FP
00k
*Major
0jk
+ u
00k
β
10k
= γ
100
β
20k
= γ
200
β
30k
= γ
300
β
40k
= γ
400
β
50k
= γ
500
β
60k
= γ
600
β
70k
= γ
700
β
80k
= γ
800
β
01k
= γ
010
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 39
Combined model with treatment effects estimation:
LOANST09
ijk
= γ
000
+ γ
001
FP
00k
+ γ
002
Locale
00k
+ γ
003
FP
00k
*Major
0jk
+ γ
010
Major
0jk
+ γ
100
major
ijk
+ γ
200
Male
ijk
+ γ
300
Nontrad
ijk
+ γ
400
Race
ijk
+ γ
500
ParentEd
ijk
+ γ
600
HSGPA
ijk
+
γ
700
TeSATder
ijk
+ γ
800
ColGPA
ijk
+ e
ijk
+ r
0jk
+ u
00k
where i denotes student, j denotes major, and k denotes institution.
To answer the second research question (To what extent do student loan default rates
differ across fields of study?), I will analyze the coefficients of Major across all the categories
and compare them to the reference category (undeclared). The dependent variable, LOANST09 is
a categorical variable with 4 categories (0 = no loans, 1 = loans paid in full or cancelled, 2 =
loans in repayment, 3 = loans deferred/in forbearance/in default, and 4 = not in repayment).
Level 1 contains the control variables that are student characteristics, including primarily dummy
or categorical variables. A coefficient <1 suggests that a student pursuing a particular major is
less likely to default than if he or she were enrolled as an undeclared major studying business
(reference category). A coefficient >1 suggests that the likelihood of default will be higher.
To answer the third research question (Controlling for field of study, to what extent do
institutional factors explain differences in the likelihood of default?), I will analyze the
coefficients of the level-3 explanatory variables and their interactions with the level-2 variable
(Major). Compared to the reference category of each variable, a coefficient <1 suggests that the
likelihood of a student defaulting is lower than the reference category. A coefficient >1 suggests
the likelihood of defaulting will be higher than the reference category.
The Model: Question 4
Finally, to answer the last research question (To what extent do different fields of study
affect graduates’ employment, and how does this differ among different types of institutions?), I
will construct another HLM model with the student’s employment status (JOBST09) as the
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 40
dependent variable, and I will use the same propensity score estimation. Since the JOBST09
variable is a dichotomous variable (Yes or No), I will run a multinomial logit regression.
Hierarchical Linear Modeling (JOBST09 as DV)
Level 1 (student):
JOBST09
ijk
= π
0jk
+ π
1jk
Male
ijk
+ π
2jk
Nontrad
ijk
+ π
3jk
ParentEd
ijk
+ π
4jk
Race
ijk
+ π
5jk
LoanAmt
ijk
+
π
6j
ParentEd
ijk
+ π
7j
HSGPA
ijk
+ π
8jk
ColGPA
ijk
+ e
ijk
Level 2 (major):
π
0jk
= β
00k
+ β
01k
Major
0jk
+ r
0jk
π
1jk
= β
10k
π
2jk
= β
20k
π
3jk
= β
30k
π
4jk
= β
40k
π
5jk
= β
50k
π
6jk
= β
60k
π
7jk
= β
70k
π
8jk
= β
80k
Level 3 (institution):
β
00k
= γ
000
+ γ
001
FP
00k
+ γ
002
Locale
00k
+ γ
003
FP
00k
*Major
0jk
+ u
00k
β
10k
= γ
100
β
20k
= γ
200
β
30k
= γ
300
β
40k
= γ
400
β
50k
= γ
500
β
60k
= γ
600
β
70k
= γ
700
β
80k
= γ
800
β
01k
= γ
010
Combined model with treatment effects estimation:
JOBST09
ijk
= γ
000
+ γ
001
FP
00k
+ γ
002
Locale
00k
+ γ
003
FP
00k
*Major
0jk
+ γ
010
Major
0jk
+ γ
100
major
ijk
+
γ
200
Male
ijk
+ γ
300
Nontrad
ijk
+ γ
400
Race
ijk
+ γ
500
ParentEd
ijk
+ γ
600
HSGPA
ijk
+
γ
700
TeSATder
ijk
+ γ
800
ColGPA
ijk
+ e
ijk
+ r
0jk
+ u
00k
where i denotes student, j denotes major, and k denotes institution.
The coefficients for the level-3 (institutional) explanatory variables will be interpreted as
odds ratios of the various categories in the variable compared to the reference category of that
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 41
variable. For example, the coefficient for for-profit institutions in the FP variable indicates the
odds ratio for a student to find employment graduating from a FPCU, compared to another
student who graduates from a public institution (reference). Since the dependent variable is a
categorical variable, I employ a multinomial logit random intercept model using Stata 14
software to test how likely each program and institution sector is to contribute to default rates,
with other covariates as control.
Limitations
Majors
Although the sample size of BPS 04/09 included 16,684 students, there were some
majors in which too few students were in default or had deferment/forbearance status on their
loans to allow the study to draw conclusions. For example, in physical sciences, there were nine
students in deferment/forbearance and zero students in default, and in computer/information
sciences, there were 18 students in deferment/forbearance and six students in default. Such small
samples existed in math, engineering, life sciences, and education, as well. For this reason, I
reorganized the 13 majors identified by BPS into nine categories.
Unfortunately, there were many respondents who did not report their majors (categorized
as missing data), including 60% of students enrolled in for-profit institutions. Since students at
for-profits made up only 13% of the sample (2,103 out of 16,684 respondents), the remaining
numbers for each major reported were very small. To produce meaningful results, I collapsed
some of the majors into categories (such as math/science, vocational/technical,
computer/engineering, and business/management), which yielded nine fields of study.
Math/science, social/behavioral sciences, and education majors at FPCUs still had relatively few
students; without sufficient numbers, results for these majors at FPCUs lack statistical power.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 42
This is, however, not an issue for data on public and private not-for-profit HEIs, where each
major was well represented.
Loan Default
The problem of sample size within majors applied also for default status. I would not
have been able to run any meaningful regressions based on default numbers in single or low-
double digits. For the loan default variable, numbers for all three types of institutions were small.
To address this problem, I combined students who deferred, who were in forbearance, and who
were in default into one variable, which I named “inability to repay.” In deferment, the payment
of principal and interest was delayed for up to three years. For forbearance, principal and interest
was delayed or reduced for up to 12 months. In both cases, students had signaled the inability or
unwillingness to repay loans. Although some students, especially those in public and private
non-private HEIs, would be in deferment because they were continuing onto graduate school, the
BPS data did not make that distinction. A portion of these deferments and forbearances might
have started out short term, but almost 40% of these cases ended up in default (Miller, 2015). In
all three categories (deferment, forbearance, and default), students were unable to pay down their
loans (Delisle & McCann, 2014).
Due to the small numbers, I also kept the institutional categories to three types: for-profit,
private non-profit, and public HEIs. Although some health and vocational/technical programs
were two-year programs or less—and there might be different loan default outcomes for these
majors in different program lengths—further subdividing the institutions into four-year, two-
year, and less-than-two-year programs would yield nine institutional types, with very small
numbers in each cell. The analysis then would not have enough statistical power to generate
meaningful results.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 43
BPS Database
The Beginning Postsecondary Students 2004/2009 database was the most current
database available for this study. As noted above, data from BPS 04/09 are derived from the
National Postsecondary Student Aid Study, a nationwide study conducted by the National Center
for Education Statistics. Even backed by the federal government, the data on student loans from
these sources remained incomplete (Dynarski, 2014). Moreover, the dataset offered by BPS
04/09 is somewhat dated, as it collected information from students who first began college in
2003/2004. The data were collected during the Great Recession years, with low employment
opportunities and deflated salaries and wages. Additionally, subsequent to the release of this
data, there was much focus on the practices of FPCUs (Deming, Golden, & Katz, 2012; Cellini
& Chaudhary, 2014; Cellini & Goldin, 2014), which led to tighter federal regulation and scrutiny
in that sector. Databases with a more current cohort might generate different regression results,
although the next round of BPS data will not be available until after 2017.
Detailed findings and analysis are presented in Chapter 4.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 44
Chapter 4 – Data Analysis
Data analysis using Beginning Postsecondary Students 2004–2009 (BPS 04/09) database
aims to study whether a student’s major is associated with the his or her ability to repay federal
student loans, how institutional factors explain the likelihood of default, and whether
employment status after leaving college is associated with the majors and the type of higher-
education institutions (HEIs) chosen.
I used propensity-scored hierarchical linear modeling and multinomial logistic
regressions to answer the following research questions:
1. To what extent do total cumulated student loans differ across fields of study (both
career/vocational and degree programs)?
2. To what extent do student loan default rates differ across fields of study (both
career/vocational and degree programs)?
3. Controlling for field of study, to what extent do institutional factors explain
differences in the likelihood of default?
4. To what extent do different fields of study affect graduates’ employment, and
how does this differ among different types of institutions?
The results were expected to show that cumulated student loans do differ across fields of
study with the amount dependent upon various student characteristics. Similarly, I expected the
likelihood of a student’s inability to pay to be higher in majors in which students take out higher
loan amounts.
Controlling for student characteristics, students at for-profit colleges and universities
(FPCUs) were expected to be more likely than students at other types of HEIs to be unable to
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 45
repay their loans across of the amount they borrowed and their major.
It was also expected that the field of study would play a small but statistically significant
role in the amount borrowed and the likelihood of default. Graduates of different majors have
varying earnings outcomes, and therefore they were also expected to have varying levels of
success as they seek employment. I expected to find that students in STEM majors had a higher
likelihood of securing employment. However, I expected that students in similar majors across
institutions would have different employment outcomes, and that students from FPCUs would
find it harder than other students to find jobs.
Descriptive Statistics
Institutions and Majors
The BPS 04/09 database includes 16,684 students enrolled in a variety of HEIs. There are
four-year, two-year, and less-than-two-year institutions that are run as public, private non-profit,
and private for-profit entities. 64% of students in the cohort attended public institutions, while
23% went to private non-profit colleges and universities and 13% enrolled in private for-profit
entities in 2009. Table 4 shows the breakdown of the major areas of study by institution types.
4,707 (41%) students failed to include a major in their survey responses, and therefore they are
not included in the model. Of those who declared a major, the most popular choices were
business/management (12%) and vocational/technical programs (10%), while education was the
least popular (5%).
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 46
Table 4
Breakdown of Majors by Higher-Education Institution Types
Major when last enrolled
(as of 2009)
Institution type last attended(as of 2009)
Private
Public
Private
Total
for-profit non-profit
Undeclared 79 908 71 1058
Humanities 79 829 549 1457
Social/Behavioral 21 778 582 1381
Math/Science
a
7 602 429 1038
Computer/Engineering
b
97 624 225 946
Education 6 630 253 889
Business/Management 162 1251 586 1999
Health 211 1085 252 1548
Vocational/Technical
c
175 1132 354 1661
Unknown
d
1266 2865 576 4707
Total 2103 10704 3877 16684
Notes:
a
Math and science majors are collapsed into one category.
b
Computer and engineering are collapsed into one category.
c
Vocational/technical and other professional fields are collapsed into one category.
d
Unknown indicates missing values from the database, and these students are not included in
subsequent modeling.
Gender and Age
The sample consisted of 9,809 females (58%) and 6,875 males (41%). They ranged in
age from 15 to 79 during their first year enrolled. Figure 1 shows the age distribution. For the
purpose of this research, students’ ages are categorized into traditional college age (<25) and
non-traditional college age (≥25). It is clear that most of the students (86%) first enrolled around
20 years old, as evidenced by the spike in density around 20 years of age. The ages of the non-
traditional students, on the other hand, span somewhat evenly between 25 to 45 years and then
gradually trail off. Table 5 shows the frequency distribution of the two categories.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 47
Table 5
Distribution of Traditional vs. Non-traditional Aged Students
Age first year
enrolled Total Percent Public Percent
Private
Non-
Profit Percent
For-
Profit Percent
<25 14,267 86 9175 86 3634 94 1458 69
≥25 2,417 14 1529 14 243 6 645 31
Total 16,684 100 10704 100 3877 100 2103 100
Race/Ethnicity
Most of the students surveyed were White (63%), with Black/African-American and
Hispanic/Latino students each representing approximately 13% of the population. The rest of the
population consisted of Asian students (5%) and those who identified as other (1%) or as two or
more races (3%; see Table 6). Sorting these data by institution type shows, however, that the
overall distribution of students across race/ethnicity is not replicated at each type of HEI. While
public institutions showed a distribution similar to that of the overall population, more White
students (70%) and fewer Black/African-American students (9%) and Hispanic/Latino students
(10%) made up the student body of private not-for-profit 75% more Black/African-American
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 48
students (23% vs. 12%) and Hispanic/Latino students (also 23% vs. 12%) than do public
institutions (Table 7).
Table 6
Racial/Ethnic Demographics of the BPS 2004–2009 Data
Race/Ethnicity Frequency Percent Cumulative
White 10,738 64.36 64.36
Black or African American 2,200 13.19 77.55
Hispanic or Latino 2,112 12.66 90.21
Asian 773 4.63 94.84
American Indian or Alaska Native 117 0.7 95.54
Native Hawaiian or Other Pacific Islander 49 0.29 95.83
Other 229 1.37 97.21
Two or more races 466 2.79 100.00
Total 16,684 100
Table 7
Race/Ethnicity of Enrolled Students by Institution Type
Private for-profit Public Private not-for-profit
Race/Ethnicity Frequency Percent Frequency Percent Frequency Percent
White 936 44.51 7,075 66.10 2,727 70.34
Black or African American 494 23.49 1,362 12.72 344 8.87
Hispanic or Latino 487 23.16 1,231 11.50 394 10.16
Asian 63 3.00 501 4.68 209 5.39
American Indian or Alaska Native 18 0.86 70 0.65 29 0.75
Native Hawaiian or Other Pacific
Islander 9 0.43 30 0.28 10 0.26
Other 27 1.28 146 1.36 56 1.44
Two or more races 69 3.28 289 2.70 108 2.79
Total 2,103 100.00 10,704 100.00 3,877 100.00
Academic Performance and Parental Education
Compared to students enrolled in public and private non-profit institutions, parents of
students at for-profit institutions had a lower educational level. 54% of parents of students at
FPCUs had a high-school level of education or below, compared to 33% for public colleges and
universities and 20% for private not-for-profits. This pattern is inverted when comparing the
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 49
proportion of students with a parent who has a bachelor’s degree or above: only 21% of students
at FPCUs have a parent with this level of attainment, compared to 41% for public HEIs and 59%
for private non-profit institutions. This suggests that parents of students enrolled in FPCUs were
less well-educated than their counterparts for other types of institutions.
As for the students themselves, students in private not-for-profit and public institutions
had achieved better high-school grade point averages (GPAs) than those enrolled in private for-
profits. 83% of students in public colleges and universities and 94% of students in private non-
profit HEIs had high-school GPAs of grade B- or above, whereas only 69% of students in
FPCUs had achieved those grades. This suggests that students enrolled in FPCUs were not as
well-prepared for higher education as those who attended other types of institutions. However,
once they were enrolled, more than 70% of students at FPCUs achieved As and Bs. This rate,
although still lagging, was more comparable to that of students in public (78%) and private non-
profit institutions (88%). Table 8 presents the academic performance and parental education of
students by institution.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 50
Table 8
Academic Performance and Parental Education of Students by Institution
Private for-profit Public Private not-for-profit
Frequency Percent Frequency Percent Frequency Percent
Parents’ highest level
of education attained
High school or below 1,089 53.99 3,429 32.74 777 20.29
Vocational/<2 years of college 263 13.04 1,327 12.67 370 9.66
Associate/<4 years of college 247 12.25 1,430 13.66 423 11.05
Bachelor’s degree 259 12.84 2,325 22.20 997 26.04
Master’s/Professional or
above 159 7.88 1,961 18.73 1262 32.96
Total 2,017 100.00 10,472 100.00 3,877 100.00
Student high-school GPA
D- to D 2 0.17 19 0.23 2 0.06
D to C- 21 1.76 49 0.58 11 0.32
C- to C 83 6.95 227 2.69 39 1.14
C to B- 264 22.09 1,037 12.29 176 5.14
B- to B 199 16.65 1,261 14.94 308 9.00
B to A- 455 38.07 3,025 35.83 1,013 29.59
A- to A 171 14.31 2,823 33.44 1,875 54.75
Total 1,195 100.00 8,441 100 3,424 100
Student estimated GPA 2009
Mostly As 216 25.81 1,359 17.34 708 21.45
As and Bs 318 37.99 2,972 37.91 1476 44.71
Mostly Bs 140 16.74 1,827 23.31 711 21.54
Bs and Cs 124 14.81 1,288 16.43 341 10.33
Mostly Cs 16 1.91 252 3.21 45 1.37
Cs and Ds 10 1.19 69 0.88 4 0.12
Mostly Ds or below 1 0.12 30 0.38 2 0.06
School does not award grades 12 1.43 42 0.54 14 0.42
Total 837 100.00 7,839 100.00 3,301 100.00
Cumulative Federal Loans, Repayment Status, and Employment Status by Institution Type
At FPCUs, only 19% of students attended without federal student loans. Most of them
(81%) took out federal student loans, with 66% borrowing $15,000 or less. Although this loan
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 51
amount is relatively small, 21% of the students who took out loans were unable to repay their
loans and were either in deferment (where repayment is temporarily delayed for up to three
years), forbearance (where repayment is stopped or reduced for 12 months), or default.
This was in contrast to students in other types of institutions. At public colleges and
universities, 51% had no federal loans, and 32% borrowed $15,000 or less. The percentage of
students who were unable to repay was 9%. At private not-for-profit institutions, more students
took out loans (about 65%) and in greater amounts. 29% borrowed up to $15,000, and 32%
borrowed between $15,001 and $30,000. However, despite these higher loan amounts, only 8%
of the loans were in deferment, forbearance, or default.
Looking at the employment picture, while most students across all institution types were
employed as of 2009, students graduating from public and private not-for-profit institutions had
similar employment rates (81% and 84%, respectively). In contrast, graduates from private for-
profit institutions had a lower employment rate (71%). These statistics are presented in Table 9.
Cumulative Federal Loans, Repayment Status, and Employment Status by Major
When the loan data are sorted by majors, it is evident that about one third of the students
in each major did not take out any federal loans. The exception to this tendency was those who
were undeclared (50% of whom had no federal loans) and those whose majors were missing in
the data (53% of whom had no loans). Relatively evenly across all majors, between 56% and
60% of students borrowed some federal loan amount up to $30,000. The employment numbers
show that between 81% and 90% of students in each declared major were employed by the end
of the study. The undeclared and those who did not declare a major had lower employment rates
(around 75%).
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 52
However, the ability to repay federal loans varied more markedly across majors.
Computer science/engineering (6%), math/science (8%), and education majors (8%) had the
lowest percentages of graduates who were unable to repay and were either in deferment,
forbearance, or default. Undeclared majors and majors in health-related programs had the highest
percentage (both at 13%) of students who demonstrate the inability to pay. About 9% of
graduates in social/behavioral sciences and business/management, and around 10% of graduates
in humanities and vocational/technical programs were unable to repay. Table 10 summarizes the
federal loan amount owed, repayment status, and employment status in detail.
Table 9
Cumulative Federal Student Loan, Loan Repayment Status, and Employment Status by Institution
Private for-profit Public Private not-for-profit
Frequency Percent Frequency Percent Frequency Percent
Federal loan amount owed
none 404 19.21 5,423 50.66 1,327 34.23
up to $15,000 1,389 66.05 3,519 32.88 1,124 28.99
$15,001 to $30,000 243 11.55 1,513 14.13 1,241 32.01
$30,001 to $45,000 62 2.95 211 1.97 163 4.20
$45,001 & above 5 0.24 38 0.36 22 0.57
Total 2,103 100.00 10,704 100.00 3,877 100.00
Federal student loan repayment
status in 2009
No federal loans 404 19.21 5,423.00 50.66 1,327 34.23
Loans paid in full or cancelled 204 9.70 566.00 5.29 261 6.73
In repayment 898 42.70 2,858.00 26.70 1,630 42.04
Inability to pay: deferment,
forbearance, or default 441 21.00 957.00 9.00 298 8.00
Not in repayment 156 7.42 900.00 8.41 361 9.31
Total 2,103 100.03 10,704.00 100.06 3,877 100.31
Employment Status
Not currently employed 494 28.70 1,403 19.00 430 16.22
Currently employed 1,227 71.30 5,983 81.00 2,221 83.78
Total 1,721 100.00 7,386 100.00 2,651 100.00
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 53
Table 10
Federal Student Loan Amount, Loan Repayment and Employment Status by Major
Major when last enrolled, 2009
Undeclared Humanities Social/Behavioral Math/Science Comp./Eng. Education Business/Mgmt. Health Vocational/Tech. Unknown
Federal loan
amount owed Freq Pct Freq Pct Freq Pct Freq Pct Freq Pct Freq Pct Freq Pct Freq Pct Freq Pct Freq Pct
none 533 50.38 566 38.85 536 38.81 388 37.38 368 38.90 323 36.33 759 37.97 544 35.14 639 38.47 2,498 53.07
up to $15,000 437 41.30 449 30.82 377 27.3 296 28.52 300 31.71 273 30.71 616 30.82 672 43.41 543 32.69 2,069 43.96
$15,001 to $30,000 79 7.47 383 26.28 411 29.76 299 28.81 242 25.58 248 27.90 530 26.51 277 17.89 389 23.42 139 2.95
$30,001 to $45,000 6 0.57 50 3.43 53 3.84 50 4.82 30 3.17 39 4.39 81 4.05 47 3.04 79 4.76 1 0.02
$45,001 & above 3 0.28 9 0.62 4 0.29 5 0.48 6 0.63 6 0.67 13 0.65 8 0.52 11 0.66 0 0.00
Total 1058 100 1457 100 1381 100 1038 100 946 100 889 100 1999 100 1548 100 1661 100 4,707 100
Federal student loan
repayment status,
2009
No federal loans 533 50.38 566 38.85 536 38.81 388 37.38 368 38.9 323 36.33 759 37.97 544 35.14 639 38.47 2,498 53.07
Loans paid in full
or cancelled 68 6.43 75 5.15 61 4.42 60 5.78 64 6.8 39 4.39 100 5.00 83 5.36 71 4.27 410 8.71
In repayment 244 23.06 514 35.28 489 35.41 351 33.82 343 36.3 340 38.25 740 37.02 529 34.17 624 37.57 1,212 25.75
Inability to pay
a
136 12.85 149 10.23 128 9.27 78 7.51 55 5.8 75 8.43 182 9.10 202 13.05 163 9.82 528 11.22
Not in repayment 77 7.28 153 10.50 167 12.09 161 15.51 116 12.3 112 12.60 218 10.91 190 12.27 164 9.87 59 1.25
Total 1058 100 1457 100 1381 100 1038 100 946 100 889 100 1999 100 1548 100 1661 100 4,707 100
Employment status,
2009
Not currently
employed 195 25.62 175 18.98 127 15.58 78 15.45 66 11.93 49 9.88 183 14.75 123 15.17 169 16.58 1,162 25.07
Currently employed 566 74.38 747 81.02 688 84.42 427 84.55 487 88.07 447 90.12 1058 85.25 688 84.83 850 83.42 3,473 74.93
Total 761 100 922 100 815 100 505 100 553 100 496 100 1241 100 811 100 1019 100 4,635 100
Note:
a
Inability to pay includes loans in deferment, forbearance, and default.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 54
Summary
The BPS 04/09 database consists of information from 16,684 students who first entered
postsecondary institutions in 2004. 64% of them were at public HEIs, 23% at private non-profits,
and 13% at private for-profit institutions. There were more females (58%) than males (42%).
Their ages ranged between 15 and 79, with the majority (86%) falling among the traditional 18–
24-year-old college-going age; 14% of the sample population were non-traditional students. 63%
of the students identified as White, 13% as Black/African American, 13% as Hispanic/Latino,
5% as Asian, and 3% as two or more races.
Private for-profit colleges and universities (FPCUs) accepted more Black/African-
American and Hispanic/Latino students than public institutions, and these students came from
lower socioeconomic backgrounds. 54% of parents of the students at FPCUs had an education
level of high school or lower, compared to 33% for public institutions and 20% for private non-
profit HEIs. Only 69% of students at FPCUs had high-school GPAs of B- or above. 94% of
students attending public and private non-profit colleges and universities had high-school GPAs
of B- or above.
81% of students at for-profits borrowed federal student loans. Of those, 61% borrowed
$15,000 or less, and 21% of those students were unable to repay loans and either went into
deferment, forbearance, or default. 49% of students at public HEIs borrowed. 32% borrowed
$15,000 or less, and 9% of public HEI students who borrowed had trouble repaying. At private
non-profit colleges and universities, 20% in the database borrowed; 29% borrowed at $15,000 or
less, but a larger proportion borrowed larger amounts (32% borrowed between $15,001 and
$30,000). 8% of those at private non-profit HEIs were unable to repay their loans. Employment
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 55
rate was lowest among graduates from FPCUs (71%). 81% of graduates from public institutions
and 84% of those from private non-profit institutions secured employment.
In eight of the nine majors identified in this study, between 56% and 60% of students
borrowed federal student loans. Approximately 50% of the students with undeclared majors took
out any federal loans. In each major, 81% to 90% of the students found employment. The
exceptions were students with undeclared majors; only 71% of that population were employed.
Undeclared and health-care majors had the highest proportion of students who were unable to
repay their federal loans (13%). Math/science and education majors had the lowest percentage of
students unable to repay (6% to 8%).
Statistical Tests
To measure how well the main independent variables are related to the dependent
variable, I used two types of tests of significance.
Continuous Variable (FEDLOAN) and Categorical Variable (Institution Type – FP09)
Table 11 shows the results of one-way analysis of variance (ANOVA) test of significance
for institution types. The test yields a p value of less than 0.01 (p<.01), which suggests that there
was a difference in cumulative federal student loan borrowed through 2009 between students
enrolled in different types of higher-education institutions. The null hypothesis that there was no
significant difference is therefore rejected.
Additionally, pairwise comparisons of means with equal variances (Table 12) show that
there were statistically significant differences (p<.01) in federal loan amounts borrowed among
all three types of institutions. Students at public HEIs borrowed on average $1,591 (p<.01) less
than those at private for-profit institutions. Students at private not-for-profit HEIs borrowed
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 56
$3,119 (p<.01) more than students at FPCUs, and more than $4,710 (p<.01) more than students
at public universities and colleges.
Continuous Variable (FEDLOAN) and Categorical Variable (Major – MAJ09B)
Table 13 shows the results of one-way analysis of variance (ANOVA) test of significance
for majors. The test yields a p value of less than 0.01 (p<0.01), which suggests that there was a
difference in cumulative federal student loan borrowed through 2009 between students enrolled
in different majors. The null hypothesis that there was no significant difference is therefore
rejected. Additionally, pairwise comparisons of means with equal variances (Table 14) show that
there were statistically significant differences in federal loan amounts borrowed by students in all
nine majors (p<.01). Out of the 35 pairs of comparisons between majors, 19 had statistically
significant differences in the amount of federal loan borrowed (p<.01). Those majors are
highlighted in bold in Table 14 below.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 57
Table 11
ANOVA Test of Significance for Cumulative Federal Student Loan and Institution Type
Summary of cumulative federal student
loan amount borrowed through 2009
Mean Std. Dev. Frequency
Private for-profit 7528.61 8224.67 2,103
Public 5937.16 8894.01 10,704
Private not-for-profit 10647.36 10792.53 3,877
Total 7232.31 9493.75 16,684
Source SS df MS F prob > F
Between groups 6.34E+10 2 3.17E+10 366.88 0.00
Within groups 1.44E+12 16681 86344047.8
Total 1.50E+12 16683 90131291.2
Bartlett’s test for equal variances: x
2
(2) = 290.0487 Prob> x
2
= 0.000
Table 12
Pairwise Comparisons of Means with Equal Variances Over Different Institution Types
Unadjusted Unadjusted
Cumulative Loan Total Contrast Std. Err. t P>t [95% Conf. Interval]
FP09
Public vs. Private for-profit −1591.45 221.64 −7.18 0.00 −2025.89 −1157.01
Private not-for-profit vs.
Private for-profit 3118.74 251.65 12.39 0.00 2625.48 3612.01
Private not-for-profit vs. Public 4710.19 174.18 27.04 0.00 4368.79 5051.60
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 58
Table 13
ANOVA Test of Significance for Cumulative Federal Student Loan and Major
Major when last
enrolled
Summary of cumulative federal loan amount
borrowed through 2009
Mean Std. Dev. Frequency
Undeclared 4312.15 6881.65 1,058
Humanities 9294.55 10435.04 1,457
Social 9879.66 10552.58 1,381
Math/Science 9979.47 10743.94 1,038
Computer Science/
Engineering 9093.20 10350.25 946
Education 10163.90 11045.92 889
Business/
Management 9535.87 10786.57 1,999
Health 8079.96 9634.12 1,548
Vocational/
Technical 9187.59 10854.05 1,661
Total 8898.34 10385.24 11,977
Analysis of Variance
Source SS Df MS F Prob>F
Between groups 2.85E+10 8 3.56E+09 33.72 0.00
Within groups 1.26E+12 11968 105546290
Total 1.29E+12 11976 107853282
Bartlett’s test for equal variances: x
2
(8) = 313.6606 Prob> x
2
= 0.000
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 59
Table 14
Pairwise Comparisons of Means with Equal Variances
over: MAJ09B
Unadjusted Unadjusted
Cumulative loan amount Contrast Std. Err. t P>t [95% Conf. Interval]
MAJ09B
Humanities vs. undeclared 4982.39 414.97 12.01 0.00 4168.98 5795.81
Social/behavioral science vs. undeclared 5567.51 419.75 13.26 0.00 4744.73 6390.28
Math/science vs. undeclared 5667.31 448.82 12.63 0.00 4787.55 6547.08
Computer science/engineering vs.
undeclared
4781.05 459.71 10.40 0.00 3879.95 5682.15
Education vs. undeclared 5851.75 467.42 12.52 0.00 4935.52 6767.98
Business/management vs. undeclared 5223.72 390.59 13.37 0.00 4458.10 5989.34
Health vs. undeclared 3767.80 409.81 9.19 0.00 2964.51 4571.09
Vocational/technical vs. undeclared 4875.44 404.11 12.06 0.00 4083.32 5667.56
Social/behavioral science vs. humanities 585.11 385.83 1.52 0.13 -171.19 1341.41
Math/science vs. humanities 684.92 417.28 1.64 0.10 -133.02 1502.86
Computer/engineering vs. humanities -201.34 428.97 -0.47 0.64 -1042.19 639.50
Education vs. humanities 869.36 437.22 1.99 0.05 12.32 1726.39
Business/management vs. humanities 241.32 353.89 0.68 0.50 -452.36 935.01
Health vs. humanities -1214.59 375.00 -3.24 0.00 -1949.65 -479.53
Vocational/technical vs. humanities -106.95 368.76 -0.29 0.77 -829.78 615.88
Math/science vs. social/behavioral science 99.81 422.03 0.24 0.81 -727.44 927.05
Computer science/engineering vs. social/
behavioral science
-786.46 433.59 -1.81 0.07 -1636.36 63.45
Education vs. social/behavioral science 284.24 441.76 0.64 0.52 -581.68 1150.17
Business/management vs. social/behavioral
science
-343.79 359.48 -0.96 0.34 -1048.43 360.85
Health vs. social/behavioral science -1799.70 380.28 -4.73 0.00 -2545.11 -1054.30
Vocational/technical vs. social/behavioral
science
-692.07 374.13 -1.85 0.06 -1425.42 41.28
Computer/engineering vs. math/science -886.26 461.79 -1.92 0.06 -1791.45 18.93
Education vs. math/science 184.44 469.48 0.39 0.69 -735.81 1104.69
Business/management vs. math/science -443.59 393.04 -1.13 0.26 -1214.02 326.83
Health vs. math/science -1899.51 412.15 -4.61 0.00 -2707.38 -1091.64
Vocational/technical vs. math/science -791.87 406.48 -1.95 0.05 -1588.64 4.90
Education vs. computer/engineering 1070.70 479.89 2.23 0.03 130.04 2011.37
Business management vs. computer
science/engineering
442.67 405.43 1.09 0.28 -352.03 1237.37
Health vs. computer science/engineering -1013.25 423.97 -2.39 0.02 -1844.30 -182.19
Vocational/technical vs. computer
science/engineering
94.39 418.47 0.23 0.82 -725.87 914.65
Business/management vs. Education -628.03 414.15 -1.52 0.13 -1439.84 183.78
Health vs. education -2083.95 432.33 -4.82 0.00 -2931.38 -1236.51
Vocational/technical vs. education -976.31 426.93 -2.29 0.02 -1813.16 -139.46
Health vs. business management -1455.91 347.82 -4.19 0.00 -2137.71 -774.12
Vocational/technical vs. business/mgmt. -348.28 341.09 -1.02 0.31 -1016.87 320.32
Vocational/technical vs. health 1107.637 362.941 3.05 0.00 396.2136 1819.06
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 60
Chi-square Tests of Significance for Loan Repayment and Employment Status at Different
Types of HEIs and in Different Majors
Since repayment status (LOANST09), employment status (JOBST09), major (MAJ09B),
and type of institution (FP09) are all categorical variables, I use the chi-square test of
independence to test the following null hypotheses: that
1. there is no difference in loan repayment status for students who graduated from
different types of institutions;
2. there is no difference in employment status for students who graduated from
different types of institutions;
3. there is no difference in loan repayment status for students who majored in different
areas of study; and
4. there is no difference in employment status for students majoring in different areas
of study.
The results are presented in Tables 15 and 16. Table 15 shows that the probabilities that
loan repayment status and employment status are independent of the type of institution in which
a student enrolls are both less than .05 (p<.05). This means that there is a relationship between
loan repayment status and the type of higher-education institution attended. There is also a
relationship between employment status and the type of higher-education institution attended.
The null hypothesis (1), that there is no difference in the loan repayment status for
students who graduated from different types of institutions, is rejected. The null hypothesis (2),
that there is no difference in the employment status for students who graduated from different
types of institutions, is also rejected.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 61
Table 15
Chi-square Test of Hypothesis: Loan Repayment and Employment Status and Institution Type
Key
frequency
x
2
contribution
Federal student loan Type of higher education institution
repayment status in 2009 Private for-profit Public Private non-profit Total
No federal loans 404 5,423 1,327 7,154
274.8 151.2 67.7 493.7
Loans paid off or 204 566 261 1,031
cancelled 42.2 13.8 1.9 57.9
In repayment 898 2,858 1,630 5,386
70.7 103.3 114.4 288.4
In deferment, forbearance 441 957 298 1,696
or default 241.5 15.8 23.4 280.7
Not in repayment 156 900 361 1,417
2.9 0.1 3.1 6.0
Total 2,103 10,704 3,877 16,684
632 284.2 210.5 1126.8
Pearson x
2
(8) = 1.1e+03 Pr = 0.000
Employment status
as of 2009
Not currently employed 494 1,403 430 2,327
69.1 2.4 17.1 88.5
Currently employed 1,227 5,983 2,221 9,431
17.0 0.6 4.2 21.8
Total 1,721 7,386 2,651 11,758
86.1 2.9 21.3 110.4
Pearson x
2
(2) = 110.3700 Pr = 0.000
Table 16 shows that the probabilities that loan repayment status and employment status
are independent of the specific major from which a student graduates are both less than .05
(p<.05). This means that there is a relationship between loan repayment status and the major a
student chooses. There is also a relationship between employment status and the major a student
chooses. The null hypothesis (3), that there is no difference in the loan repayment status for
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 62
students who graduated from different majors, is rejected. The null hypothesis (4), that there is
no difference in the employment status for students who graduated from different majors, is also
rejected.
Summary
ANOVA and Chi-square tests are used to estimate how well the independent variables
are related to the dependent variables. The results of one-way analysis of variance (ANOVA) test
of significance between cumulative loan amount (FEDLOAN) and institution type (FP09) yields
a p value of less than 0.01 (p<.01), which suggests that there is a difference in cumulative federal
student loan borrowed through 2009 between students enrolled in different types of higher-
education institutions. The null hypothesis that there is no significant difference is therefore
rejected.
The results of one-way analysis of variance (ANOVA) test of significance for FEDLOAN
and field of study (MAJ09B) yields a p value of less than 0.01 (p<0.01), which suggests that
there is a difference in cumulative federal student loan amount borrowed through 2009 between
students enrolled in different majors. The null hypothesis that there is no significant difference is
therefore rejected.
Results of the chi-square tests of significance show that students’ loan repayment statuses
differ among different types of institutions and across different majors. Results also show that
there is variation among students’ employment status depending on both the major they chose
and the type of HEI (public, private non-profit, or private for-profit) from which they graduated.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 63
Table 16
Chi-square Test of Hypothesis: Loan Repayment and Employment Status and Major
Key
frequency
x
2
contribution
Major when last enrolled 2009 (condensed)
Federal student loan
repayment status
in 2009
Undeclared Humanities
Social/
Behav.
sciences
Math/
Science
Computer
Science/
Engineering
Education
Business/
Mgmt.
Health
Vocational/
Technical
Total
No federal loans 533 566 536 388 368 323 759 544 639 4,645
36.0 0.0 0.0 0.6 0.0 1.5 0.4 5.5 0.1 44.1
Loans paid off or 68 75 61 60 64 39 100 83 71 621
cancelled 3.1 0 1.6 0.7 4.6 1.1 0.1 0.1 2.7 14
In repayment 244 514 489 351 343 340 740 529 624 4,174
42.2 0.1 0.1 0.3 0.5 2.9 2.7 0.2 3.5 52.6
In deferment,
forbearance, 136 149 128 78 55 75 182 202 163 1,168
or default 10.4 0.3 0.3 5.3 15 1.6 0.9 17.3 0 51.2
Not in repayment 77 153 167 161 116 112 218 190 164 1,358
15.4 0.9 0.7 15.9 0.7 1.2 0.3 1.2 3.1 39.5
Total 1,058 1,457 1,381 1,038 946 889 1,999 1,548 1,661 11,977
107.2 1.3 2.7 22.9 20.9 8.3 4.4 24.3 9.4 201.4
Pearson x
2
(32) = 201.4121 Pr = 0.000
Employment status
in 2009
Not currently 195 175 127 78 66 49 183 123 169 1165
employed 40.0 3.9 0.3 0.3 6.6 12.7 2.0 0.7 0.0 66.4
Currently 566 747 688 427 487 447 1,058 688 850 5,958
employed 7.8 0.8 0.1 0.0 1.3 2.5 0.4 0.1 0.0 13.0
Total 761 922 815 505 553 496 1,241 811 1,019 7,123
47.8 4.6 0.4 0.3 7.9 15.2 2.3 0.8 0.0 79.4
Pearson x
2
(8) = 79.4268 Pr = 0.000
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 64
Regression Analysis
Association between Students’ Ability to Repay Federal Student Loans and Major
I use hierarchical linear modeling because the students are nested in majors, which in turn
are nested in institutions. Additionally, the dependent variable and many predictors have binary
outcomes, and some are multi-categorical variables. Therefore, I use multinomial logit
regression. The analysis employs a combined mixed-effects multinomial logit regression.
Table 17 shows both the model with the array of majors and the full model with
demographic predictors added in. Compared to students with undeclared majors, the odds of
students borrowing federal loans and being unable to repay were 1.5 times (p<.01) higher in
most cases. For students in education and health in particular, the odds of being unable to pay
were highest among all majors, at 1.7 (p<.01) and 1.8 times (p<.01), respectively, compared to
students who were undeclared. Looking at the demographics, older students were 6% less likely
than younger students to be unable to repay loans. The odds for deferment/forbearance/default
were also lower for students who lived in rural areas, as well as for those whose parents had a
higher level of education. Black students were 2.4 times (p<.01) more likely to have loans that
they were unable to repay. Somewhat unexpectedly, the odds of a student accumulating debt that
she or he could not repay was 1.1 times higher if her or his grade point average was higher, and
odds were even (1.0 times) if the student’s SAT score was higher.
Looking at each of the majors individually without comparison, the only major that is
statistically significant (p<.01) is the undeclared major, in which the odds of the student having
trouble repaying is 32% lower than if the student had chosen another major. Table 18 shows this
result. Demographic regression estimates remain largely unchanged.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 65
Table 17
Regression Estimates Expressed in Odds Ratio of Likelihood of Default by Major Compared to
Undeclared
Likelihood of Default (Odds Ratio)
Major
(Comparison = undeclared)
Model with
majors only
Full model with
demographic variables
Humanities 1.45 *** 1.49 ***
Social/Behavioral Science 1.49 *** 1.49 ***
Math/Science 1.60 *** 1.72 ***
Computer/Engineering 1.45 *** 1.77 ***
Education 1.77 *** 1.74 ***
Business/Management 1.51 *** 1.42 ***
Health 1.70 *** 1.46 ***
Vocational/Technical 1.48 *** 1.46 ***
Demographic Characteristics
Age
0.94 **
Male
0.97
Locale
0.81 ***
Parents’ education
0.81 ***
High-school GPA
1.02
GPA at college 2009
1.11 ***
SAT score
1.00 ***
Race – White
1.15
Race – Asian
0.88
Race – Black or African
American
2.40 ***
Race – Hispanic or Latino
0.88
Race – Native Hawaiian
or Pacific Islander
0.90
Race – American Indian
or Alaska Native
1.05
Race – Other
1.08
Intercept 1.61
8.57 ***
Note: *p<.10, **p<.05, ***p<.01
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 66
Table 18
Regression Estimates Expressed in Odds Ratio of Inability to Repay
by Major
Inability to Repay
(Odds Ratio)
Major
Undeclared
0.68 ***
Humanities
1.00
Social/Behavioral Science 1.00
Math/Science
1.15
Computer/Engineering
1.21
Education
1.20
Business/Management
0.96
Health
1.03
Vocational/Technical
0.99
Demographic Characteristics
Age
0.94 **
Male
0.97
Locale
0.81 ***
Parents’ education
0.81 ***
High-school GPA
1.02
GPA at college 2009
1.11 ***
SAT score
1.00 ***
Race – White
1.15
Race – Asian
0.88
Race – Black or African
American
2.40 ***
Race – Hispanic or Latino
0.88
Race – Native Hawaiian
or Pacific Islander
0.90
Race – American Indian
or Alaska Native
1.04
Race – Other
1.08
Intercept
12.31 ***
Note: *p<.10, **p<.05, ***p<.01
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 67
Association Between Students’ Ability to Repay Federal Student Loans and the Type of
Higher-education Institution Attended
Table 19 shows that students who graduated from FPCUs had 1.8 times (p<.01) higher
odds of graduating with loan debt that they were unable to or had trouble repaying, whereas the
odds for students graduating from public institutions were 47% (p<.01) lower. In terms of
demographics across all institutions, the odds for the inability to repay loans was lower (3%,
p<.05) when the student was male, when the student lived in a rural area (20%, p<.01), and
when the parents of the student had a higher level of education (19%, p<.01). The odds were 2.4
times (p<.01) higher for Black students to borrow and be unable to repay federal loans.
Association between Students’ Employment Status by Major
Table 20 shows the regression estimates of the likelihood of employment by major (with
undeclared major as the comparison), first with the majors themselves and secondly with
demographic predictors added. The odds of students finding employment were statistically
significantly higher in all majors compared to the undeclared major, with graduates in education
almost four times (p<.01) as likely to get a job. Computer science/engineering majors had 2.5
times (p<.01) higher odds, and healthcare related majors had two times ((p<.01) higher odds of
job security. Graduates of humanities had the lowest odds of securing a job (1.4 times, p<.01),
but these odds were still higher than for undeclared majors.
Adding in the demographic predictors lowers the odds ratio for graduates of the various
majors to procure jobs, but the trend remains the same. Graduates of all majors—other than
humanities, for which the estimate is now statistically insignificant—had higher odds of finding
jobs than graduates of undeclared majors. Additionally, regression estimates for demographic
predictors suggest that older graduates had 9% (p<.10) lower odds, and Hispanic/Latino students
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 68
had 27% lower odds (p<.10) of securing employment. Interestingly, the odds of finding a job
were lower for students with a higher college GPA (0.93, p<.05), whereas the odds of finding a
job were higher for those with a higher high school GPA (1.08, p<0.10).
Looking at the regression estimates of each major individually in Table 21, there are
fewer predictors that are statistically significant with a confidence level of p<.10 or less.
Graduates of undeclared and humanities majors had lower odds of employment (.88, p<.01, and
.68, p<.01, respectively), but the odds of education majors finding employment was 1.89 times
(p<.01) more likely. For demographic predictors, regression estimates are the same as Table 20
as discussed above.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 69
Table 19
Regression Estimates Expressed in Odds Ratio of Inability to Repay Loans by Institution
Loan Repayment Status
(Odds Ratio)
Institution
For-profit
1.80 ***
Public
0.53 ***
Private not-for-profit
1.05
Demographic Characteristics
Age
0.94
Male
0.97 **
Locale
0.80 ***
Parents’ education
0.81 ***
High-school GPA
1.02
GPA at college 2009
1.11 ***
SAT score
1.00
Race – White
1.16
Race – Asian
0.88
Race – Black or African American
2.41 ***
Race – Hispanic or Latino
0.87
Race – Native Hawaiian
or Pacific Islander
0.90
Race – American Indian
or Alaska Native
1.02
Race – Other
1.73
Intercept
12.64 ***
Note: *p<.10, **p<.05, ***p<.01
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 70
Table 20
Regression Estimates Expressed in Odds Ratio of Employment Status by Major Compared to Undeclared
Employment Status (Odds Ratio)
Model with
majors only
Full model with
demographic variables
Major
(comparison = undeclared)
Humanities 1.38 *** 1.21
Social/Behavioral Science 1.69 *** 1.56 ***
Math/Science 1.71 *** 1.54 **
Computer/Engineering 2.53 *** 2.57 ***
Education 3.90 *** 3.35 ***
Business/Management 1.89 *** 1.78 ***
Health 2.00 *** 1.82 ***
Vocational/Technical 1.74 *** 1.76 ***
Demographic Characteristics
Age
0.91 *
Male
0.93
Locale
1.01
Parents’ education
0.96
High-school GPA
1.08 *
GPA at college 2009
0.93 **
SAT score
1.00
Race – White
1.35
Race – Asian
0.83
Race – Black or African
American
0.90
Race – Hispanic or Latino
0.73 *
Race – Native Hawaiian
or Pacific Islander
1.28
Race – American Indian
or Alaska Native
1.10
Race – Other
1.45
Intercept 12.00 ** 13.59 **
Note: *p<.10, **p<.05, ***p<.01
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 71
Table 21
Regression Estimates Expressed in Odds Ratio of Employment Status by Major
Employment Status
(Odds Ratio)
Major
Undeclared
0.88 ***
Humanities
0.68 ***
Social/Behavioral Science 0.88
Math/Science
0.87
Computer/Engineering
1.45
Education
1.89 ***
Business/Management
1.01
Health
1.03
Vocational/Technical
0.88
Demographic Characteristics
Age
0.91 *
Male
0.94
Locale
0.00
Parents’ education
0.97
High-school GPA
1.08 *
GPA at college 2009
0.93 **
SAT score
1.00 **
Race – White
1.35
Race – Asian
0.84
Race – Black or African
American
0.90
Race – Hispanic or Latino
0.72 *
Race – Native Hawaiian
or Pacific Islander
1.28
Race – American Indian
or Alaska Native
1.10
Race – Other
1.46
Intercept
24.04 ***
Note: *p<.10, **p<.05, ***p<.01
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 72
Association between Students’ Employment and the Type of Higher-education Institution.
Table 22 shows that students graduating from for-profit institutions had lower odds of
securing employment (.56, p<.05) compared to those graduating from public and private not-for-
profit HEIs. They were nearly half as likely as their counterparts to have employment. In terms
of demographics, graduates who were older had slightly lower odds of finding a job than
younger graduates (.92, p<.10), and the odds of Hispanic/Latino graduates being employed were
28% (p<.10) lower than for those of all other races/ethnicities. The odds of having a job were
slightly higher (1.07, p<.10) for graduates with higher high-school GPAs, although they were
slightly lower (.92, p<.05) for graduates with higher college GPAs.
Table 22
Regression Estimates Expressed in Odds Ratio of Employment Status by Institution
Loan Repayment Status
(Odds Ratio)
Institution
For-profit
0.56 **
Public
0.93
Private not-for-profit
1.18
Demographic Characteristics
Age
0.92 *
Male
0.93
Locale
1.02
Parents’ education
0.96
High-school GPA
1.07 *
GPA at college 2009
0.92 **
SAT score
1.00
Race – White
1.33
Race – Asian
0.82
Race – Black or African
American
0.90
Race – Hispanic or Latino
0.72 *
Race – Native Hawaiian
or Pacific Islander
1.27
Race – American Indian
Alaska Native
1.09
Race – Other
1.44
Intercept
26.07 ***
Note: *p<.10, **p<.05, ***p<.01
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 73
Association between Students’ Cumulated Federal Loan Amount and Major
I use three-level hierarchical linear modeling because the students are nested in majors,
which in turn are nested in institutions. The dependent variable (FEDLOAN) is a continuous
variable, and, therefore, I am able to add propensity-scored weights into the model to better
match students with similar demographical details. Table 23 lists the results of the regressions.
Results with statistically significant estimates of p<.10 will be discussed.
Undeclared. Students with an undeclared major borrowed $4,998 (p<.01) less than the
average amount borrowed across all majors. Undeclared majors borrowed less if they lived in
rural areas (−$861.62, p<.01). For each higher level of their parents’ education, students
borrowed $930.78 less (p<.01). For each unit increase of high-school and college GPA, the loan
amount increased $487.10 (p<.01) and $596.87 (p<.01), respectively, while for each unit
increase in SAT score, borrowing decreased slightly (−$2.61, p<.01). Students identifying as
Asian and Hispanic/Latino borrowed less than average (−$1,903.68, p<.05, and −$791.89,
p<.10, respectively), and those identifying as Black/African American borrowed more
($4,350.98, p<.01).
Humanities. Humanities majors borrowed more if they were male ($791.48, p<.10) and
Black/African American ($4,668.25, p<.01). For every unit increase in high-school and college
GPA, the borrowing also increased ($620.65, p<.01, and $606.38, p<.01, respectively). Average
borrowing decreased for each higher level of parents’ education ($−1,185.81, p<.01), for each
unit increase in SAT score (−$2.98, p<.05), and for Hispanic/Latino students (−$1,589.16,
p<.10).
Social/behavioral sciences. Social and behavioral science majors borrowed $1,142.44
(p<.01) more than the average amount borrowed across all majors. Black students in this major
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 74
borrowed more ($4755.30, p<.01). For each unit increase in college GPA, borrowing also
increased ($663.40, p<.01). Students whose parents had higher levels of education borrowed less
(−$1225.98, p<.01). For each unit increase in SAT score, borrowing decreased as well (−$4.51,
p<.01).
Math/science. Black students in math and science majors borrowed more than the
average amount ($5,326.08, p<.01). High-school and college GPA predicted higher levels of
borrowing ($1,055.59, p<.01, and $644.57, p<.01, respectively, per unit increase in GPA). Math
majors living in rural locales borrowed less (−$1,593.36, p<.01). Parents’ level of education and
student SAT scores also predicted lower borrowing amounts (−$838.14, p<.01, and −$10.92,
p<.01, respectively).
Computer science/engineering. For each higher level of parents’ education, computer
and engineering majors borrowed less (−$1,276.19, p<.01). For each unit increase in SAT score,
they also borrowed less (−$4.65, p<.01). High-school and college GPA predicted higher levels
of borrowing ($1,096.31, p<.01, and $948.65, p<.01, respectively, per unit increase in GPA).
Black students in computer sciences and engineering borrowed substantially more ($6,302.58,
p<.01) than the average amount.
Education. Students majoring in education borrowed $2,128.69 (p<.01) more than the
average amount across all majors. High-school and college GPA predicted higher levels of
borrowing ($899.96, p<.01, and $636.62, p<.05, respectively, per unit increase in GPA).
Students in education identifying as Black/African American and American Indian/Alaska
Native also borrowed more ($5,021.85, p<.10, and $5,880.72, p<.05, respectively). Students
living in rural areas borrowed less (−$1,153.00, p<.05), as did those whose parents had higher
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 75
levels of education (−$1,061.29, p<.01, per unit increase in level of education). For each year
increase in student age, borrowing decreased (−$983.63, p<.10).
Business/management. Black/African-American students in business and management
studies borrowed more ($5,246.36, p<.01), and Hispanic/Latino students borrowed less
(−$1,741.23, p<.05). High-school and college GPA predicted higher levels of borrowing
($544.05, p<.01, and $550.51, p<.05, respectively, per unit increase in GPA). Business and
management students living in rural locales borrowed less (−$822.04, p<.01). Each unit increase
in SAT score also predicted less borrowing (−$3.45, p<.01).
Health fields. White and Asian students in health-related programs borrowed less than
the average amount borrowed across all majors (−$3,466.00, p<.01, and −$4,911.51, p<.01,
respectively). High-school and college GPA predicted higher levels of borrowing ($564.33,
p<.05, and $696.87, p<.01, respectively, per unit increase in GPA). Students living in rural areas
borrowed less (−$923.28, p<.10). Students whose parents had higher levels of education
borrowed less (−$779.41, p<.10), as well. Each unit increase in SAT score also predicted less
borrowing (−$3.99, p<.05).
Vocational/technical. White and Asian students in vocational or technical programs
borrowed less than the average amount borrowed across all majors (−$1,857.08, p<.05, and
−$4,121.33, p<.05, respectively). High-school and college GPA predicted higher levels of
borrowing ($540.79, p<.05, and $539.17, p<.01, respectively, per unit increase in GPA). Male
students borrowed on average $1,098.05 (p<.05) less than female students. Students living in
rural areas borrowed less (−$1,415.23, p<.01). Students whose parents had higher levels of
education borrowed less (−$866.77, p<.01).
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 76
Table 23
Regression Estimates of Cumulative Federal Student Loan by Major
Major
Undeclared Humanities
Social/
Behavioral
Sciences
Math/
Science
Computer/
Engineering Education
Business/
Management Health
Vocational/
Technical
Coefficient −4998.81 *** 583.23
1142.43 *** 1389.34
590.24
2128.69 *** −16.78
−1380.66
−260.73
Age 55.59
253.89
−69.26
327.06
206.88
−983.63 * −229.67
267.28
25.27
Male −242.29
791.48 * −627.27
−75.06
99.46
685.44
−146.80
−26.33
−1098.05 **
Locale −861.62 *** −563.78
−415.33
−1593.36
**
* −1046.45
−1153.00 ** −287.92
−923.28 * −1415.23 ***
Parents’ education −930.78 *** −1185.81 *** −1225.98 *** −838.14
**
* −1276.19 *** −1061.29 *** −822.04 *** −779.41 *** −866.77 ***
High-school GPA 487.10 *** 620.65 *** 374.76
1055.59
**
* 1096.31 *** 899.96 *** 544.05 *** 564.33 ** 540.79 **
GPA at college 2009 596.87 *** 606.38 *** 663.40 *** 644.57
**
* 948.65 *** 636.62 ** 550.51 *** 696.87 *** 539.17 ***
SAT score −2.16 *** −2.98 ** −4.51 *** −10.92
**
* −4.65 *** 0.60
−3.45 *** 3.99 ** 1.52
Race – White −45.97
−191.94
1211.00
430.72
1038.47
964.96
445.17
−3466.00 ** −1857.08 **
Race – Asian −1903.68 ** −1328.34
117.80
−1694.04
−826.36
1161.15
−1627.06
−4911.51 ** −4121.33 **
Race – Black or
African American 4350.98 *** 4668.25 *** 4755.30 *** 5326.08
**
* 6302.58 *** 5021.85 * 5246.36 *** 153.00
4143.54
Race – Hispanic or
Latino −791.89 * −1589.16 * −1070.72
−1305.82
23.82
−723.17
−1741.23 ** −1248.61
−552.31
Race – Native
Hawaiian or
Pacific Islander −843.59
−2360.26
−3697.52
−1915.13
309.33
488.94
483.75
−3380.73
695.14
Race – American
Indian or
Alaska Native 977.05
−686.14
1173.68
1565.00
3737.04
5880.72 ** 1616.66
516.56
37.85
Race – Other 113.46
582.78
1594.09
1826.24
247.18
547.30
667.50
−2774.58
−679.72
Intercept 9874.99 *** 6130.38
13461.66 ** 10467.87
4103.37
−8459.12
15271.06 *** 2873.38
8804.38
Note: *p<.10, **p<.05, ***p<.01
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 77
Association between Students’ Cumulated Federal Loan Amount and Type of Higher-
education Institution
Table 24 shows the regression results of cumulative federal student loan by institution.
Private for-profit HEIs. At for-profit colleges and universities, students identifying as
Black/African American and American Indian/Alaska Native borrowed more than the average
amount ($6,775.16, p<.01, and $4945.46, p<.10, respectively). For each year increase in age,
students borrowed $1,509.31 less (p<.10). Students with parents who had a higher level of
education also borrowed less (−$624.40, p<.05, for each unit increase in level of education).
Public HEIs. At public higher-education institutions, older students borrowed more
($224.72, p<.10, for each year increase in age). High-school and college GPA predicted higher
levels of borrowing ($317.10, p<.01, and $474.38, p<.01, respectively, per unit increase in
GPA). Males borrowed $379.64 (p<.10) less than females. Students living in rural areas
borrowed $692.11 (p<.01) less than students living in cities. Students with parents who had a
higher level of education also borrowed less (−$1,094.21, p<.01, for each unit increase in level
of education). For each unit increase in SAT score, borrowing decreased by $1.91 (p<.01). Asian
students borrowed less (−$2,739.00, p<.01), but Black/African-American and American-
Indian/Alaska-Native students borrowed more ($4,958.54, p<.01, and $1,453.82, p<.05,
respectively).
Private not-for-profit HEIs. The average student at a private not-for-profit institution
borrowed $3,327.00 (p<.05) more than the average amount for the total population. Males
borrowed less than females (−$480.24, p<.10). High-school and college GPA predicted higher
levels of borrowing ($431.30, p<.05, and $831.20, p<.01, respectively, per unit increase in
GPA). Students with parents who had a higher level of education borrowed less (−$1,212.75,
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 78
p<.01, for each unit increase in level of education). Asian students borrowed less (−$2,971.32,
p<.01), but students identifying as Black or African American and American Indian or Alaska
Native borrowed more ($3,897.11, p<.01, and $3,544.20, p<.05, respectively).
Table 24
Regression Estimates of Cumulative Federal Student Loan by Institution
Type of institution
Private for-profit Public
Private
not-for-profit
Coefficient −2376.21
−2591.91
3327.00 **
Age −1509.31 * 224.72 * 122.62
Male 348.17
−379.64 * −480.24 *
Locale −769.72
−692.11 *** −524.58
Parents’ education −624.40 ** −1094.21 *** −1212.75 ***
High-school GPA 87.03
317.10 *** 431.30 **
GPA at college 2009 427.88
474.38 *** 831.20 ***
SAT score 1.42
−1.91 *** −4.25 ***
Race – White 2716.94
317.65
−1428.97
Race – Asian 811.19
−2739.00 *** −2971.32 ***
Race – Black or African
American 6775.16 *** 4958.54 *** 3897.11 ***
Race – Hispanic or
Latino 137.95
−671.18
−635.39
Race – Native Hawaiian
or Pacific Islander −2537.83
−1606.13
−3544.20 **
Race – American Indian
or Alaska Native 4945.46 * 1453.82 ** 527.79
Race – Other 3230.798
920.37
−1756.87
Intercept 1760.55
7879.14 *** 11651.36 ***
Note: *p<.10, **p<.05, ***p<.01
Intraclass Correlation
Table 25 shows the intraclass correlations for the three-level nested models tested. The
first is the level-three intraclass correlation at the institution level, and the second is the level-two
intraclass correlation at major-within-institution level. Results show that loan repayment and
employment status, as well as total cumulative federal loan amount, are only slightly correlated
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 79
within the same institutions and also within the same majors and institutions. Major random
effects are estimated between 3% and 5%, and major and institution random effects are between
5% and 10% of the total residual variance. Although the effects are small, there are nonetheless
statistically significant findings worth noting.
Table 25
Intraclass Correlations Between Outcomes and Majors and Institutions
Level Loan Repayment Status Employment Status Cumulative Federal Loan
Institution
0.09 0.03 0.03
Major | Institution
0.10 0.05 0.05
Conclusion
Since the students were nested in different majors within different types of higher-
education institutions, I used hierarchical linear modeling to estimate the effects of a student’s
choice of major and the choice of HEI on loan repayment status (the ability or inability to pay),
on employment status (employed or not employed), and on the total amount of federal loans
accumulated.
Loan Repayment Status
Students in health and education majors had the highest odds of being unable to repay
loans compared to the undeclared major, although for all majors, when compared to the
undeclared major, there were higher odds of loans being in deferment, forbearance, or default.
The odds of students being unable to repay loans were 2.4 times higher for Black students than
for students of any other race. The odds of a student being unable to repay loans were lower for
those whose parents had higher levels of education and for those living in rural areas. Older
students had lower odds of being in default than younger students, but students with higher GPA
and SAT scores had higher odds of defaulting than students with lower GPA and SAT scores.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 80
Students at for-profit HEIs had higher odds of having trouble repaying loans, whereas students at
public colleges and universities had lower odds. Other predictors were statistically insignificant.
Employment Status
Graduates of all majors had higher odds of securing employment than the comparison
group (undeclared). However, graduates in the education field were four times more likely to
find employment than undeclared majors. Graduates of computer science/engineering and
health-related majors were more than twice as likely as undeclared majors to have a job. The
major with the lowest odds of employment was humanities. Older students and Hispanic/Latino
students had a more difficult time securing a job: their odds were lower than younger students
and students of other races, respectively. Students with higher college GPA had slightly lower
odds of being employed, but students with higher high-school GPA had slightly higher odds.
Strikingly, students graduating from for-profit institutions were half as likely as those in other
types of HEIs to find employment. Other predictors were statistically insignificant.
Cumulative Federal Student Loan Amount
Students in undeclared majors borrowed less on average than students in other majors
(−$4,998.81, p<.01). Students in social and behavioral sciences majors, on the other hand,
borrow more ($1,142.43, p<.01). Age was only statistically significant, albeit weakly (p<.10), as
a predictor in the education major, decreasing the amount borrowed by $983.63 for each year
increase in age. Locale was a strong (p<.01) predictor of cumulative loan amount, and the
amount borrowed was lower if the student lived in a rural setting. Parents’ education was another
strong predictor of cumulative loan amount, and it decreased the amount borrowed for each unit
level increase. High-school and college GPA, both strong predictors, each increased the amount
borrowed for every unit increase in GPA scores. On the other hand, although SAT scores were a
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 81
strong predictor, they decreased amount borrowed by $10 or less per unit in all cases. African-
American and Black students consistently borrowed $4,000 to $6,000 more than students of
other races. White, Asian, and Hispanic/Latino students borrowed less. Identifying as American
Indian/Alaska Native was a strong predictor (p<.05) in only the education major, in which those
students borrowed $5,880.72 more than other races.
At private for-profit institutions, students identifying as Black/African American or
American Indian/Alaska Native borrowed more than the average amount. Students with more
highly educated parents borrowed less. At public and private not-for-profit HEIs, older students
borrowed more, while males and students living in rural areas borrowed less. Across all types of
institutions college and high-school GPAs predicted higher levels of borrowing, but SAT scores
and parents’ education predicted lower levels of borrowing. Additionally, Asian students
borrowed less, but Black and Native American/Alaska native students borrowed more.
In Chapter 5, I will discuss the findings further and identify implications going forward.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 82
Chapter 5 – Conclusions and Recommendations
In 2014, 70% of college seniors carried student loans, with the average amount owed
equal to $29,000 (“Student debt and the class of 2014,” 2015). The current total loan amount in
the U.S. stands at over $1.3 trillion (Federal Reserve, 2016a). To address the problems posed by
a rising student loan debt burden and the increasing loan default rate accompanying it, there must
be a clear understanding of why students default, to what extent they default, and whether or not
their choices of major and choices of education institutions are associated with their likelihood to
default.
This study segregated loan default data by program type to provide a more detailed
picture of the increasingly significant student loan default issue at hand. Using the Beginning
Postsecondary Students 2004–2009 database (BPS 04/09), I investigated whether students’
ability to pay differed for various majors across all higher-education institutions. Additionally, I
analyzed each major separately by institution type (private for-profit, private not-for-profit, and
public).
Since the students were studying different majors nested within different types of higher-
education institutions, I used propensity-scored hierarchical linear modeling to answer the
following research questions:
1. To what extent do total cumulated student loans differ across fields of study (both
career/vocational and degree programs)?
2. To what extent do student loan default rates differ across fields of study (both
career/vocational and degree programs)?
3. Controlling for field of study, to what extent do institutional factors explain
differences in the likelihood of default?
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 83
4. To what extent do different fields of study affect graduates’ employment, and
how does this differ among different types of institutions?
In this concluding chapter, I will summarize the current literature on the variables
affecting student loan amount and student loan default, as well as the study’s research
methodology. Next, I will discuss the implications of the findings from this research. I will then
explain the limitations of the research and follow up with conclusions. The final section contains
recommendations for future areas of research.
Literature Review and Methodology
The current literature on student loan default focuses on two main categories: student
characteristics and institutional factors. Student characteristics include such considerations as
age, gender, race/ethnicity, household income or parent’s educational level, student’s academic
preparation/attainment, knowledge of debt, debt load, program completion, and employment
after graduation. For the purpose of this study, the institutional factor considered was funding
structure (whether the institution was public, private non-profit, or private for-profit). There is
also substantial literature that examines the effect of the length of programs offered (less-than-
two-year, two-year, or four-year programs) on student loan outcomes, and I did not analyze those
factors in this study.
Student characteristics are well established as strong predictors of student loan default.
Early literature on student loan default found consistently that race/ethnicity and program
completion explained 20% and 26%, respectively, of loan default (Herr & Burt, as quoted in
Gross et al., 2009). Minorities, especially African Americans and students who dropped out of
college, were at higher risk of default than others. Those studies also found default likelihood to
be inversely correlated with parental level of education, students’ academic attainment, and
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 84
postgraduate employment and earnings. There was a weaker but still statistically significant
association between loan default and students’ age: older nontraditional students were more
likely to default than their 18–24-year-old counterparts.
More recent studies that include institutional characteristics in their models have found
that, even controlling for student characteristics, the type of higher-education institution (HEI)
did affect the likelihood of loan default. Default rates at for-profit colleges and universities
(FPCUs) were much higher than at public or private non-profit institutions (Deming, Goldin, &
Katz, 2012). Students at two-year institutions were also more likely to default than those at four-
year institutions (Campbell & Hillman, 2015).
My research used data from Beginning Postsecondary Students 2004–2009 (BPS 04/09),
which is a database derived from National Postsecondary Student Aid Study and compiled by the
National Center for Education Statistics. The data comprised 16,684 students who began college
in 2004. Since the data had student-level characteristics nested within major fields of study,
which in turn were nested in institutional factors, I ran the regressions using propensity-scored
hierarchical linear modeling to explore the relationship between cumulative loan amount and
students’ majors. I used multinomial logit regressions to estimate the likelihood of a student’s
inability to pay in relation to his or her employment status and choice of major.
Research Study Questions and Findings
Question 1: To What Extent Do Total Cumulated Student Loans Differ Across Fields of
Study (Both Career/Vocational and Degree Programs)?
Undeclared. Students with an undeclared major borrowed $4,998 (p<.01) less than the
average amount borrowed across all majors. Additionally, undeclared majors borrowed less if
they lived in rural areas (−$861.62, p<.01). For each higher level of their parents’ education,
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 85
students borrowed $930.78 less (p<.01). For each unit increase of high-school and college GPA,
the loan amount increased $487.10 (p<.01) and $596.87 (p<.01), respectively, while for each unit
increase in SAT score, borrowing decreased slightly (−$2.61, p<.01). Asian and Hispanic
students borrowed less than average (−$1,903.68, p<.05 and −$791.89, p<.10, respectively) and
Black students borrowed more ($4,350.98, p<.01).
It may seem odd that undeclared majors borrowed less. Since the data does not
differentiate between students who completed the degree and those did not continue, it may be
that undeclared majors borrowed less because they no longer attended classes. Undeclared
students in rural areas may borrow less because they attended local community colleges that
were less expensive, but that they dropped out before declaring a major and completing a degree.
Humanities. Humanities majors borrowed more than the average student if they were
male ($791.48, p<.10) and black ($4,668.25, p<.01). For every unit increase in high-school and
college GPA, the borrowing also increased ($620.65, p<.01, and $606.38, p<.01, respectively).
Average borrowing decreased for each higher level of parents’ education (−$1,185.81, p<.01),
for each unit increase in SAT score (−$2.98, p<.05), and for students who identified as
Hispanic/Latino (−$1,589.16, p<.10).
It is unclear why a male or African American/black student in humanities would borrow
more than the average student. Perhaps these students were transfer students and needed more
time to graduate, thereby incurring more loans. It could also be that classes in the humanities
were impacted due to budget constraints, requiring students to wait for classes to open in a later
semester, causing them to increase their loan borrowing.
Social/ behavioral sciences. Social and behavioral science majors borrowed $1,142.44
(p<.01) more than the average amount borrowed across all majors. Black students in this major
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 86
borrowed more ($4755.30, p<.01). For each unit increase in college GPA, borrowing also
increased ($663.40, p<.01). Students whose parents had higher levels of education borrowed less
(−$1225.98, p<.01). For each unit increase in SAT score, borrowing decreased as well (−$4.51,
p<.01).
The reason that social/ behavioral science majors borrowed more could be similar to the
one for humanities majors, and that is that classes were impacted and there were not enough
courses offered in order for students to finish on time. It is interesting that African American/
black students borrowed consistently more than other races/ ethnicities. This could be explained
by their being more likely to be economically disadvantaged, so they had to borrow more in
order to be able to attend college.
Math/ science. Black students in math and science majors borrowed more than the
average amount ($5,326.08, p<.01). High-school and college GPA predicted higher levels of
borrowing ($1,055.59, p<.01, and $644.57, p<.01, respectively, per unit increase in GPA). Math
majors living in rural locales borrowed less (−$1,593.36, p<.01). Parents’ level of education and
student SAT scores also predicted lower borrowing amounts (−$838.14, p<.01, and −$10.92,
p<.01, respectively).
Computer science/ engineering. For every higher level of parents’ education, computer
and engineering majors borrowed less (−$1,276.19, p<.01). For every unit increase in SAT
score, they also borrowed less (−$4.65, p<.01). High-school and college GPA predicted higher
levels of borrowing ($1,096.31, p<.01, and $948.65, p<.01, respectively, per unit increase in
GPA). Black students in computer sciences and engineering borrowed substantially more
($6,302.58, p<.01) than the average amount.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 87
In computer science/ engineering as well as in several other majors, students with higher
college and high school GPA borrowed more than students with lower grades. This could be
attributed to the higher academically achieving students choosing to go to selective or highly
selective, but also more expensive colleges and universities. The lower achieving students might
have opted for community colleges or shorter programs at for-profit institutions, which lowered
the total amount of students loans borrowed.
Education. Students majoring in education borrowed $2,128.69 (p<.01) more than the
average amount across all majors. High-school and college GPA predicted higher levels of
borrowing ($899.96, p<.01, and $636.62, p<.05, respectively, per unit increase in GPA). Black
and American Indian students in education also borrowed more ($5,021.85, p<.10, and
$5,880.72, p<.05, respectively). Students living in rural areas borrowed less (−$1,153.00,
p<.05), as did those whose parents had higher levels of education (−$1,061.29, p<.01, per unit
increase in level of education). For every year’s increase in student age, borrowing decreased by
(−$983.63, p<.10).
The finding that students in education borrowed more may seem counterintuitive.
However, education majors constitute a larger proportion of minority students than other majors.
These minority students may come from disadvantaged backgrounds and find the need to borrow
larger loan amounts in order to attend HEIs. In addition, schools of education typically have
fewer scholarships than other programs (e.g. STEM majors) and students may have to look for
other means (grants and student loans) to meet tuition costs and living expenses.
Business/ management. Black students in business and management studies borrowed
more ($5,246.36, p<.01), and Hispanic students borrowed less (−$1,741.23, p<.05). High-school
and college GPA predicted higher levels of borrowing ($544.05, p<.01, and $550.51, p<.05,
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 88
respectively, per unit increase in GPA). Business and management students living in rural locales
borrowed less (−$822.04, p<.01). Each unit increase in SAT score also predicted less borrowing
(−$3.45, p<.01).
Health. White and Asian students in health-related programs borrowed less than the
average amount borrowed across all majors (−$3,466.00, p<.01, and −$4,911.51, p<.01,
respectively). High-school and college GPA predicted higher levels of borrowing ($564.33,
p<.05, and $696.87, p<.01, respectively, per unit increase in GPA). Students living in rural areas
borrowed less (−$923.28, p<.10). Students whose parents had higher levels of education
borrowed less (–$779.41, p<.10). Each unit increase in SAT score also predicted less borrowing
(−$3.99, p<.05).
Vocational/ technical. White and Asian students in vocational and technical programs
borrowed less than the average amount borrowed across all majors (−$1,857.08, p<.05, and
−$4,121.33, p<.05, respectively). High-school and college GPA predicted higher levels of
borrowing ($540.79, p<.05, and $539.17, p<.01, respectively, per unit increase in GPA). Male
students borrowed on average $1,098.05 (p<.05) less than female students. Students living in
rural areas borrowed less, as well (−$1,415.23, p<.01). Students whose parents had higher levels
of education borrowed less (−$866.77, p<.01).
Analysis across majors. Table 26 displays all statistically significant results from the
fixed-effects model. The results show that cumulated loan amounts do differ depending on what
majors students chose to study. Students with undeclared majors borrowed less, while those in
social/behavioral sciences and education borrowed more. Age and gender were not determining
factors in all except one of the majors (humanities for gender, and education for age), and even
in those majors there was a weak correlation (p<0.1). Students’ locale was highly correlated in
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 89
five out of nine majors. Students living in rural settings borrowed less in all five majors than
their counterparts living in urban or suburban areas. In terms of academic preparation and
performance, both high-school and college grade point averages (GPA) were strongly correlated
in all nine fields of study, where students with higher GPAs borrowed more than those with
lower GPAs. SAT scores, on the other hand, had a small but statistically significant effect on
lowering student loan borrowing in seven majors. Parents’ level of education, which served as a
proxy for socioeconomic status, was also a significant factor associated with lower loan amounts
in all majors. The higher the level of parent education, the less the students borrowed.
Finally, race/ethnicity presented a mixed picture. Race/ethnicity was a statistically
significant factor in the amount of loans borrowed for White students in two majors (health and
vocational/technical), for Asian students in three majors (undeclared, health, and
vocational/technical), for Hispanic/Latino students in three majors (undeclared, humanities, and
business/management), and for students identifying as American Indian/Alaskan Native in one
major (education). Race/ethnicity predicted an increase in borrowing for Whites and Asians in
health majors and a decrease in borrowing for those groups in vocational/technical majors.
Students borrowed less than the average student if they identified as Hispanic/Latino in
undeclared, humanities, and business/management majors. Students borrowed more if they were
American Indian/Alaskan native (in education).
The numbers were statistically significant for African Americans/Blacks in the majority
(seven of nine) of fields of study. They borrowed consistently across most majors, and the
amount borrowed was more than for other races/ethnicities.
Discussion and implications. This study demonstrated that there was an association
between how much federal loans students borrowed and the majors that they chose to study, after
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 90
controlling for student demographic variables such as age, gender, race, and socioeconomic
status, as well as institutional factors. There was previously scant but conflicting extant literature
on student loan debt by major. While Hershbein, Harris, and Kearney (2014) found that student
loan debt was fairly similar across majors, Kantrowitz (2012) showed that students in
architecture, theology, and history were more likely to graduate with six-figure loan debt, while
students in computer science, mathematics, and healthcare were the least likely to do so. My
research supported Kantrowitz’s premise that the amount of federal loans students borrowed
depended on the major chosen. One possible reason for the different loan amounts across majors
could be varying program lengths and/or expenses for books and materials. More research is
needed in this area to draw a definite conclusion.
The finding that students in rural areas borrowed less than those in urban and suburban
areas might be explainable due to the lower cost of living in rural areas. Moreover, the higher
education options available in more rural or remote locales tend to be two-year community
colleges and less-selective four-year universities that typically charge lower tuition and fees
(Koricich, 2014). These more affordable conditions may enable students to borrow less to cover
their costs.
Borrowing was higher for students with higher high-school and college GPAs. These
results could be related to the institutions they choose to attend. High-achieving students tended
to attend highly selective public and private not-for-profit colleges and universities, which have
higher tuition as well as living expenses. It is therefore reasonable to assert that these students
were likely to borrow more federal loans than lower achieving students who attended less
selective, less expensive alternatives.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 91
It has been well documented that a higher parental level of education, which is a proxy
for socioeconomic status (SES), is correlated with lower loan debt accumulated for students
(Houle, 2013; Gross, Cekic, Hossler, & Hillman, 2009). Thus, my finding on this factor
supported existing literature in this area.
Although recent discourses on student loan debt have directed attention to students’ SES
rather than race/ethnicity (Houle, 2013; Looney & Yannelis, 2015), the consistent finding that
African-American/Black students borrowed more than students of other races/ethnicities cannot
be ignored (Baum & Steele, 2010; Baum & Johnson, 2015; Huelsman, 2015). Lower-income
students were predominantly Black/African-American and Hispanic/Latino (Simms, Fortuny, &
Henderson, 2009). If only SES were used to discuss student loan debt, then the result from this
study should have found that both Black/African-American students and Hispanic/Latino
students were equally likely to borrow more, or less, than those of other races. However, my
findings showed that Black/African-American students took out substantially bigger loans than
other races in 80% of the majors studied. On the other hand, Hispanic/Latino students borrowed
less than students of other races. A difference in borrowing behavior existed between these two
racial/ethnic groups, even after controlling for SES. Further research needs to be undertaken to
unpack these differences.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 92
Table 26
Regression Estimates of Cumulative Federal Student Loan Amount by Major
Major
Undeclared Humanities Social/
Behav.
Sciences
Math/
Science
Comp/
Engin.
Education Business/
Mgmt.
Health Vocat./
Technical
Coefficient −4998.81 *** 1142.43 *** 2128.69 ***
Age −983.63 *
Male 791.48 * −1098.05 **
Locale −861.62 *** −1593.36 *** −1153.00 ** −923.28 * −1415.23 ***
Parents’ education −930.78 *** −1185.81 *** −1225.98 *** −838.14 *** −1276.19 *** −1061.29 *** −822.04 *** −779.41 *** −866.77 ***
High-school GPA 487.10 *** 620.65 *** 1055.59 *** 1096.31 *** 899.96 *** 544.05 *** 564.33 ** 540.79 **
College GPA 2009 596.87 *** 606.38 *** 663.40 *** 644.57 *** 948.65 *** 636.62 ** 550.51 *** 696.87 *** 539.17 ***
SAT score −2.16 *** −2.98 ** −4.51 *** −10.92 *** −4.65 *** −3.45 *** 3.99 **
Race – White −3466.00 ** −1857.08 **
Race – Asian −1903.68 ** −4911.51 ** −4121.33 **
Race – Black 4350.98 *** 4668.25 *** 4755.30 *** 5326.08 *** 6302.58 *** 5021.85 * 5246.36 ***
Race – Hispanic −791.89 * −1589.16 * −1741.23 **
Race – Native
Hawaiian
5880.72 **
Race – American
Indian
Race – Other
Intercept 9874.99 *** 6130.38 13461.66 ** 10467.87 4103.37 −8459.12 15271.06 *** 2873.38 8804.38
Note: *p<.10, **p<.05,
***p<.01
Notes: Empty cells have statistically insignificant results.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 93
Question 2. To What Extent Do Student Loan Default Rates Differ Across Fields of Study
(Both Career/Vocational and Degree Programs)?
Using “undeclared” as the control major, the odds of students borrowing federal loans
and being unable to repay was 1.5 times (p<.01) higher in most fields of study. For the purpose
of this study, students were considered “unable to repay” if their loans were in deferment, in
forbearance, or in default. Of all majors, the highest odds of being unable to pay were for
students in education and health fields, at 1.7 (p<.01) and 1.8 times (p<.01), respectively,
compared to students who were undeclared. Looking at the demographics, the odds of older
students being unable to repay loans were 6% lower than for younger students. The odds for
deferment, forbearance, or default were also lower for students who lived in rural areas, as well
as for those whose parents had a higher level of education. The odds of a student accumulating
debt that he or she could not repay was 1.1 times higher if his or her grade point average was
higher, and odds were even (1.0 times) if a student’s SAT score was higher. Black/African-
American students were 2.4 times (p<.01) more likely to have loans that they were unable to
repay.
When regressing each major individually without comparison, the only major that was
statistically significant (p<.01) was the undeclared major, in which the odds of the student
having trouble repaying were 32% lower than if the student had chosen another major. This
finding may be because students who have reported their majors as undeclared might still be in
college, in which case they did not yet have to start repaying their loans. Alternatively, they
could also have discontinued taking classes, which meant that they have stopped borrowing more
loans. Earlier, the study found that undeclared majors borrowed less than students in other
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 94
majors (p.99), and it seemed logical to surmise that students with smaller amounts of loans
would have less trouble repaying their loans, as the results showed.
Discussion and implications. Although the study shows statistical difference in the
likelihood of being unable to repay loans amongst majors, the study did not find a clear
association between student loan default rate and the majors chosen. Intraclass correlation,
however, suggested that the type of institution and the chosen major together explained 10% of
student default rate, and that the field of study alone does explain 1% of the variance in default
data. Moreover, two other studies, Belfield (2011) and Steiner and Teszler (2005), have found
statistical variations in loan default rates amongst different majors, despite the fact that the
former was limited to two-year and less-than-two-year programs, and the latter was
geographically limited to Texas. Since there was some evidence to suggest a statistical
correlation between student major and default rate, research with more comprehensive data could
yield more definitive results.
The study found that the likelihood of loan default was strongly correlated with student
characteristics, including age, locale, academic performance (high-school and college GPA),
parents’ level of education, and race/ethnicity. Again, similar to the results between cumulated
loan amount and choice of major, Black/African American was the only racial identifier that had
a clear statistical correlation with default across all majors (controlling for SES [via parents’
level of education]). No other racial/ethnic identifier was statistically significant in estimating the
likelihood of default. This pointed to the possibility that there were reasons other than low
income levels that explained the 2.4 times increase in likelihood of this racial population to
default on federal student loans. Further qualitative studies can shed more light on this persistent
finding.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 95
Question 3. Controlling for Field of Study, to What Extent Do Institutional Factors Explain
Differences in the Likelihood of Default?
Cumulative loan amount. At private for-profit institutions, students identifying as
Black/African American and American Indian/Alaska Native borrowed more than the average
amount ($6,775.16, p<.01, and $4945.46, p<.10, respectively). Older students borrowed less.
For each year increase in age, students borrowed $1,509.31 less (p<.10). Students with parents
who had a higher level of education also borrowed less (−$624.40, p<.05, for each unit increase
in level of education).
At public higher education institutions, older students borrowed more ($224.72, p<.10,
for each year increase in age). High-school and college GPA predicted higher levels of
borrowing ($317.10, p<.01, and $474.38, p<.01, respectively, per unit increase in GPA). Males
borrowed $379.64 (p<.10) less than females. Students living in rural areas borrowed $692.11
(p<.01) less than students living in cities and suburban locales. Students with parents who had a
higher level of education also borrowed less (−$1,094.21, p<.01, for each unit increase in level
of education). For each unit increase in SAT score, borrowing decreased by $1.91 (p<.01). Asian
students borrowed less (−$2,739.00, p<.01), but Black/African-American and American-
Indian/Alaska-Native students borrowed more ($4,958.54, p<.01, and $1,453.82, p<.05,
respectively).
The average student at private not-for-profit institutions borrowed $3,327.00 (p<.05) less
than the average amount. Males borrowed less than females (−$480.24, p<.10) at these
institutions. High-school and college GPA predicted higher levels of borrowing ($431.30, p<.05,
and $831.20, p<.01, respectively, per unit increase in GPA). Students with parents who had a
higher level of education borrowed less (−$1,212.75, p<.01, for each unit increase in level of
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 96
education). Asian students borrowed less (−$2,971.32, p<.01), but students identifying as
Black/African American and American Indian/Alaska Native borrowed more ($3,897.11, p<.01,
and $3,544.20, p<.05, respectively).
Ability to repay/ loan default. Students who graduated from for-profit institutions had
1.8 times (p<.01) higher odds of graduating with loan debt that they were unable to or had
trouble repaying than students in public or private non-profit institutions.. In terms of
demographics across all institutions, the odds for an inability to repay loans were lower (3%,
p<.05) when the student was male, when the student lived in a rural area (20%, p<.01), and
when the parents of the student had a higher level of education (19%, p<.01). The likelihood was
2.4 times (p<.01) higher for Blacks to be in debt and to be unable to repay federal loans.
Employment status. Students graduating from for-profit institutions had lower odds of
securing employment (.56, p<.05) compared to public and private not-for-profit higher education
institutions. They were nearly half as likely as their counterparts to have employment. In terms
of demographics, graduates who were older had slightly lower odds of finding a job than
younger graduates (.92, p<.10), and the odds of Hispanics being employed were 28% (p<.10)
less than for those identifying themselves as another race/ethnicity. The odds of having a job
were slightly higher (1.07, p<.10) for graduates with higher high-school GPA, and they were
slightly lower (.92, p<.05) for graduates with a higher college GPA.
Discussion and implication. There was a clear association between cumulated loan
amount and private non-profit HEIs, where students borrowed $3,327 more than the average
amount. No significant relationship emerged, however, with for-profit and public institutions.
This is likely the case because four-year private not-for-profit HEIs have the highest tuition and
fees amongst various types of HEIs (Baum & Steele, 2010; National Center for Education
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 97
Statistics, n.d.). As Table 6 in Chapter 4 showed, 32% of students in private not-for-profits
borrowed between $15,000 and $30,000, compared to 12% of students in FPCUs and 14% of
students in public HEIs. Private non-profit HEIs also enrolled more students whose family
income precluded them from receiving federal grants, but many of those students still often need
to take out student loans to fully fund their education (Pell Institute, 2015). The study’s finding
was in consensus with current literature in this regard.
Although two thirds of students who attended for-profit institutions borrowed less than
$15,000, this study found that these students were much more likely than students from private
not-for-profit and public HEIs to default on their loans. This finding supported other recent
research (Hillman, 2014; Belfield, 2012) and suggested that graduates from FPCUs had trouble
finding employment after graduation that paid well enough to enable them to repay loans in a
timely manner. An alternative explanation might be that these students dropped out of their
programs before successfully completing a degree; thus, without a postsecondary credential or
diploma, they subsequently could not find gainful employment to repay their loans. This problem
is compounded by my additional finding that students from for-profit HEIs had a lower
likelihood of securing employment compared to graduates from public or private non-profit
institutions. This made loan repayment for this particular segment of students even more
challenging.
This study did not separate two-year and four-year institutions within the three institution
types. Given that many public and private for-profit HEIs operate as two-year institutions, the
research might have provided clearer evidence of correlation if the data had been further
subdivided into less-than-two-year, two-year, and four-year clusters. This is because some
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 98
vocational/technical programs were two-year or less, and the graduation rates and employment
outcomes might be different among credential, diploma, and degree courses.
Question 4: To What Extent Do Different Fields of Study Affect Graduates’ Employment,
and How Does This Differ Among Different Types of Institutions?
The odds of students finding employment were statistically significantly higher in all
majors compared to the undeclared major, with graduates in education being almost four times
(p<.01) as likely to get a job. Computer/engineering majors had 2.5 times (p<.01) higher
chances of employment, and healthcare-related majors had two times (p<.01) higher odds of job
security. Graduates in the humanities had the lowest odds of securing a job (1.4 times, p<.01),
but this result was still higher than that of undeclared majors.
Adding in the demographic predictors lowered the odds ratio for graduates of the various
majors to procure jobs, but the trend was the same. Graduates of all majors—other than
humanities, in which the estimate was now statistically insignificant—had higher odds of finding
jobs than graduates of undeclared majors. Additionally, regression estimates for demographic
predictors suggest that older graduates had 9% (p<.10) lower odds and Hispanics had 27% lower
odds (p<.10) of securing employment. Interestingly, the odds of finding a job were lower with a
higher college GPA (0.93, p<.05) whereas the odds of finding a job were higher with a higher
high-school GPA (1.08, p<0.10).
Looking at the regression estimates of each major individually, there were fewer
predictors that were statistically significant with a confidence level of p<.10 or less. Graduates of
undeclared and humanities majors had lower odds of employment (.88, p<.01, and .68, p<.01,
respectively), but the odds of education majors finding employment were 1.89 times (p<.01)
higher.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 99
Discussion and implication. This study found statistical associations between major and
employment only in undeclared and humanities (lower odds of getting a job) and education
(higher odds of finding employment). Findings were somewhat surprising, because previous
studies have found that graduates from STEM (science, technology, engineering, and math) and
business fields were more in demand by employers (Adams, 2014) and were paid more
(Carnevale, Cheah, & Hanson, 2015) than other majors. In light of these earlier studies, one
would assume that there would be a strong correlation between math/science majors and
employment status. The findings that humanities graduates had lower odds of securing
employment, on the other hand, supported findings from other research (Carnevale, Cheah, &
Hanson, 2015).
One reason for this anomaly could be that the data available from BPS 04/09 were binary
(yes/no) responses to the question of whether or not graduates were employed. Since the
responses were self-reported and not verified by other wage-related data, this might pose
reliability error issues. It could also have been that the students who had graduated but were not
employed were pursuing advanced degrees in graduate school.
The results suggest that the type of institution at which students chose to attend college
was a more important determinant to financial security after graduation than what students chose
to study. This is the case in light of the findings that institutional characteristics affected
employment outcomes and that graduates of for-profit institutions, in particular, had
considerably lower odds of employment and higher odds of student loan default. Nevertheless,
since majors have been shown to have substantially disparate earnings outcomes both at entry-
level and over the courses of one’s lifetime (Carnevale, Strohl, & Melton, 2011; Hershbein,
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 100
Harris, & Kearney, 2014), choosing a major had limited but statistical significance in influencing
loan repayment likelihoods.
Limitations
Majors
Although the sample size of BPS 04/09 included 16,684 students, there were some
majors in which too few students were in default or had deferment/forbearance status on their
loans to allow the study to draw conclusions. For example, in physical sciences, there were nine
students in deferment/forbearance and zero students in default, and in computer/information
sciences, there were 18 students in deferment/forbearance and six students in default. Such small
samples existed in math, engineering, life sciences, and education, as well. For this reason, I
reorganized the 13 majors identified by BPS into nine categories.
Unfortunately, there were many respondents who did not report their majors (categorized
as missing data), including 60% of students enrolled in for-profit institutions. Since students at
for-profits made up only 13% of the sample (2,103 out of 16,684 respondents), the remaining
numbers for each major reported were very small. To produce meaningful results, I collapsed
some of the majors into categories (such as math/science, vocational/technical,
computer/engineering, and business/management), which yielded nine fields of study.
Math/science, social/behavioral sciences, and education majors at FPCUs still had relatively few
students; without sufficient numbers, results for these majors at FPCUs lack statistical power.
This is, however, not an issue for data on public and private not-for-profit HEIs, where each
major was well represented.
Loan Default
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 101
The problem of sample size within majors applied also for default status. I would not
have been able to run any meaningful regressions based on default numbers in single or low-
double digits. For the loan default variable, numbers for all three types of institutions were small.
To address this problem, I combined students who deferred, who were in forbearance, and who
were in default into one variable, which I named “inability to repay.” In deferment, the payment
of principal and interest was delayed for up to three years. For forbearance, principal and interest
was delayed or reduced for up to 12 months. In both cases, students had signaled the inability or
unwillingness to repay loans. Although some of these deferments and forbearances might have
started out short term, almost 40% of these cases ended up in default (Miller, 2015). In all three
categories (deferment, forbearance, and default), students were unable to pay down their loans
(Delisle & McCann, 2014).
Due to the small numbers, I also kept the institutional categories to three types: for-profit,
private non-profit, and public HEIs. Although some health and vocational/technical programs
were two-year programs or less—and there might be different loan default outcomes for these
majors in different program lengths—further subdividing the institutions into four-year, two-
year, and less-than-two-year programs would yield nine institutional types, with very small
numbers in each cell. The analysis then would not have enough statistical power to generate
meaningful results.
BPS Database
The BPS 04/09 database was the most current database available for this study. As noted
above, data from BPS 04/09 are derived from the National Postsecondary Student Aid Study, a
nationwide study conducted by the National Center for Education Statistics. Even backed by the
federal government, the data on student loans from these sources remained incomplete
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 102
(Dynarski, 2014). Moreover, the dataset offered by BPS 04/09 is somewhat dated, as it collected
information from students who first began college in 2003/2004. The data were collected during
the Great Recession years, with low employment opportunities and deflated salaries and wages.
Additionally, subsequent to the release of this data, there was much focus on the practices of
FPCUs (Deming, Golden, & Katz, 2012; Cellini & Chaudhary, 2014; Cellini & Goldin, 2014),
which led to tighter federal regulation and scrutiny in that sector. Databases with a more current
cohort might generate different regression results, although the next round of BPS data will not
be available until after 2017.
Summary
To What Extent Do Total Cumulated Student Loans Differ Across Fields of Study (Both
Career/Vocational and Degree Programs)?
This study demonstrated that there was an association between how much federal loans
students borrowed and the majors that they chose to study, after controlling for demographic
variables and institutional factors. African-American students borrowed more across all majors,
and their numbers were statistically significant in most. Hispanic students borrowed less, and
their numbers were statistically significant for undeclared majors, in humanities, and in
business/management programs. Higher parental levels of education were associated with lower
amounts of borrowing, but higher GPAs in high school and college were associated with higher
amounts of borrowing. Students in rural areas borrowed less.
To What Extent Do Student Loan Default Rates Differ Across Fields of Study (Both
Career/Vocational and Degree Programs)?
Although the study shows statistical difference in the likelihood of being unable to repay
loans amongst majors, the study did not find a clear association between student loan default rate
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 103
and the majors chosen. Among the nine groupings of fields of study, with the undeclared major
as the control group, students in all declared majors had higher odds of being unable to repay
loans than those in undeclared. However, given the relatively small numbers of students in
deferment, forbearance, or default in math and science fields and computer science/engineering
fields in this database, I was not able to produce any statistically significant regression estimates
for each major.
Controlling for Field of Study, to What Extent Do Institutional Factors Explain Differences
in the Likelihood of Default?
There was a clear association between cumulated loan amount and private non-profit
HEIs, where students borrowed $3,327 more than the average amount. However, no significant
relationships emerged for FPCUs or public institutions. Students at FPCUs were 1.8 times more
likely to be unable to repay loans, whereas students at public institutions had 47% less chance of
defaulting or otherwise delaying their repayment. African Americans were the only
race/ethnicity to demonstrate increased odds of defaulting (2.4 times higher), regardless of the
type of institutions. Regression estimates for all other races/ethnicities were statistically
insignificant.
To What Extent Do Different Fields of Study Affect Graduates’ Employment, and How
Does This Differ Among Different Types of Institutions?
This study found statistical association between major and employment only for those
with undeclared major, in humanities (lower odds of getting a job) and in education (higher odds
of finding employment). Students who had attended for-profit institutions were 44% less likely
to find employment than if they had graduated from a different type of institution, whereas
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 104
students from public institutions were 7% less likely to find employment, and graduates of
private non-profit HEIs were 18% more likely to find employment.
Conclusions
The study found that the students in undeclared majors borrowed less than the average
student loan amount, and that social/behavioral sciences and education majors borrowed more
than average. There was no statistical association between choice of major and the ability to
repay or default on loans. Students with undeclared majors and those in humanities were less
likely to secure employment, and graduates with education majors were more likely to find
employment.
The study also found that students who had attended for-profit colleges and universities
were more likely to default on their loans and less likely to find employment than those who had
attended public and private non-profit institutions. However, students at private non-profit HEIs
had the highest cumulated loan amount amongst the three types of institutions studied.
In terms of individual student characteristics, students whose parents had higher levels of
education borrowed less and were less likely to have federal student loans in deferment,
forbearance, or default. Similarly, students who lived in rural areas borrowed less and were less
likely to be unable to repay loans. Students with higher college GPA were more likely to
accumulate more student loans and more likely to be unable to pay. Black/African-American
students borrowed more student loans and had a higher likelihood of default than any other
ethnicity/race. Hispanic/Latino students were least likely to secure employment after college.
Although the study found inconclusive evidence of the relationship between college
major and the student’s ability to repay student loans, it did find that cumulative student loan
amounts varied between majors. Since college tuition is generally assessed on a per-unit basis,
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 105
regardless of major, the cost of a degree should be comparable for students studying different
subjects in the same institution. It warrants future research to identify why some students found
the need to borrow more in some majors but not in others. Nonetheless, since undergraduate
majors do not all have the same returns (Cellini, 2012; Cellini & Chaudhary, 2014; Carnevale,
Cheah, & Hanson, 2015), there exists a pressing need to continue to shine light on the costs of a
degree and whether it is worth taking out loans to fund it. The federal government has
implemented measures to help educate students to evaluate their financial situations against their
educational goals, with guidelines for financial literacy. I would argue that students should take
college-level financial literacy classes to learn to better manage their finances and repay loans
(including federal student loans) after obtaining a degree.
The study found an increased likelihood for Black/African-American students to borrow
more and to be unable to repay their loans, as well as for Hispanic/Latino students to be unable
to find a job after leaving college. In light of these findings, the federal government should
consider increasing the Pell Grants given out to low-income and underrepresented populations.
The average cost of a year of tuition is $9,410 for public institutions, $23,893 for four-year out-
of-state public HEIs, and $32,405 for private non-profit colleges and universities (“Average
Estimated Undergraduate Budgets, 2015–2016,” n.d.). The maximum Pell Grant available is
$5,730 (“Student Federal Grant Programs,” n.d.). Taking even a casual glance, one can see that
federal grants can cover barely one half of tuition at a public university and less than 20% of
tuition at a private non-profit institution. Although there are state grants and institutional grants
that will cover some of the shortfall, it is evident that these funds cannot fully cover the cost of a
college education. As a result, the average student graduates with $29,000 worth of student debt
(“Student Debt and the Class of 2014,” 2015). The federal government should consider policies
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 106
to increase the grant amount to cover at least the full tuition costs of attending college for those
demographics most at risk of not pursuing a college education or of dropping out part of the way
through due to financial hardship. An increase in grants would enable students to consider and
begin a college career, and it would help keep loan amounts at a more manageable level for them
once they have finished their education. Studies have also shown that graduates are happier at
their jobs if they don’t have to worry about repaying loans (Kantrowitz, 2012). Reducing the
amount of loans to repay after college would level the playing field so that graduates from
underprivileged backgrounds could start from a surer footing in their climb toward a more
financially secure future.
In the meantime, the government must manage the $1.3 trillion worth of student loans
currently owed. In order to help keep students from deferring due to inability to repay,
forbearing, and slowly drifting into default, the government has put into place more repayment
options, such as “pay-as-you-earn” income-based plans and extended 25-year payment plans.
Research is needed to analyze the effectiveness of these programs. Other additional measures
may include lowering the interest rate on loans, easing the eligibility for the pay-as-you-earn and
extended repayment plans, and raising the threshold of family income for grant eligibility so that
more students can access federal aid. The costs of default are high both to the individual student
borrower and to taxpayers, and it is well worth the effort to identify variables that affect student
loan default and to find solutions to address them.
Recommendations for Future Research
The study found that cumulated loan amount is associated with the choice of college
major. Will knowing the costs associated with each major influence a student’s decision to study
that particular subject? How can a student find out if the additional costs are worth it in terms of
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 107
earnings outcomes? It is beyond the scope of this study to determine the reasons and
consequences surrounding this result. Further research is recommended to answer these
questions more comprehensively.
The study also found that Black/African-American students consistently borrow more
and have a higher likelihood to be unable to repay their loans, regardless of what majors they
choose and which type of institutions in which they enroll. African Americans constitute 22% of
low-income families, but their median income is the lowest among all racial/ethnic groups in that
strata (Simms, Fortuny, & Henderson, 2009). Socioeconomic status cannot fully explain these
findings, however, because there is no clear evidence for those who identify as Hispanic/Latino,
who represent 30% of the lower-income groups, to borrow more or go into default with higher
odds than other races/ethnicities. In fact, Hispanic/Latino students tend to borrow less than the
average amount. A more in-depth and focused qualitative study may help to clarify some of these
nuances.
Results showed that the odds of students from for-profit higher education institutions
being unable to repay their loans in a timely manner was twice as likely as students in private
non-profits and more than 3 times as likely as students from public HEIs. Students at for-profits
were almost 50% less likely than students at public HEIs and more than 50% less likely than
those at private non-profits to successfully find employment. Even though recent figures showed
that the loan default rate for students at for-profit colleges and universities have declined from
19.1% in 2011 to 15.0% in 2013 (OSFAP, 2016), it continues to be highest amongst the different
types of higher education institutions.
Senate hearings have shed light on the aggressive recruitment tactics and misinformation
of employment prospects of for-profits, leading the better-informed public to shy away from
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 108
these institutions. Consequently, enrollment at 4-year for-profit HEIs has fallen 13.7% in the fall
semester of 2015, compared to the national student enrollment decline of 1.7% (National Student
Clearinghouse, 2015). To tackle the burden of college tuition and of repaying student loan debt,
the Student Aid and Fiscal Responsibility Act (2009) was signed into law in 2010 to increase
grant amounts and adjust the loan repayment scale to 10% of income (White House, 2010).
While there has been progress to monitor higher education institutions more closely to ensure
student success after graduation, and steps have been made towards student loan reform, recent
research has found that high loan default rates persist in for-profits (Swindell, Wright &
Gallegos, 2013). High loan default rates also exist in other areas of higher education as well,
namely in students who borrow less than $5,000, who attended community colleges and who
dropped out before completing their intended programs (Looney & Yannelis, 2015; Campbell &
Hillman, 2015).
With the new administration taking office in the coming months, policies that have been
put in place to make it easier for students to service their federal student loans may be at risk of
being rolled back. Additionally, while Senate has drawn attention to the unscrupulous practices
of the for-profit higher education sector, forcing some high profile campus closings in recent
years, the new administration may be more in favor of deregulation, which will encourage the
growth of the for-profit sector again. Finally, the reauthorization of the Higher Education Act is
likely to pass during the new presidency. The provisions within this Act will influence and shape
education policies in future years.
Student loan default in the US is an ongoing and increasingly urgent concern.
Researchers must continue to focus on this topic in order to keep it in the spotlight and to help
policymakers make informed decisions that affect future generations of income earners, whose
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 109
spending and saving habits will impact the health and direction of the economy for years to
come.
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 110
References
Adelman, C. (2006). The toolbox revisited: Paths to degree completion from high school to
college. Washington D.C.: U.S. Department of Education.
Angulo, A.J. (2016). Diploma mills: how for-profit colleges stiffed students, taxpayers and the
American dream. Baltimore: Johns Hopkins University Press.
Avery, C., & Turner, S. (2012). Student loans: Do college students borrow too much – or not
enough? Journal of Economic Perspectives, 26(1), 165-192.
Baum, S., & Steele, P. (2010). Who borrows most? Bachelor’s degree recipients with high levels
of student debt. Trends in Higher Education Series. New York, NY: The College Board.
Retrieved from https://trends.collegeboard.org/sites/default/files/trends-2010-who-
borrows-most-brief.pdf
Baum, S., & Schwartz, S. (2013). Student aid, student behavior, and educational attainment.
Retrieved from
http://gsehd.gwu.edu/sites/default/files/documents/PUBLISHED_Baum_Schwartz.pdf
Baum, S., & Johnson, M. C. (2015). Student debt: Who borrows most? What lies ahead? Trends
in Higher Education Series. New York, NY: The College Board. Retrieved from
http://www.urban.org/research/publication/student-debt-who-borrows-most-what-lies-
ahead
Becker, G. S. (1961). Investment in human capital: A theoretical analysis. The Journal of
Political Economy, 70(5), 4-49.
Becker, G. S. (1993). Human capital: A theoretical and empirical analysis with special reference
to education. 3
rd
edition. Chicago: The University of Chicago Press.
Becker, G. S., Murphy, K. M., & Tamura, R. (1993). Human capital, fertility and economic
growth. In G.S. Becker, Human capital: A theoretical and empirical analysis with special
reference to education. 3
rd
edition. Chicago: The University of Chicago Press.
Bennett, D. L., Lucchesi, A. R., & Vedder, R. K. (2010). For-profit higher education: Growth,
innovation and regulation. Washington, D.C.: Center for College Affordability and
Productivity.
Belfield, C. R. (2013). Student loans and repayment rates: The role of for-profit colleges.
Research in Higher Education, 54(1), 1-29.
Block, M. (2012, April 24). Student loan debt exceeds one trillion dollars. National Public Radio.
Retrieved from http://www.npr.org/2012/04/24/151305380/student-loan-debt-exceeds-
one-trillion-dollars
Brown, M., Haughwout, A., Lee, D., Mabuta, M., & van der Klaauw, W. (2012, March 5).
Grading student loans. Federal Reserve Bank of New York. Retrieved from
http://libertystreeteconomics.newyorkfed.org/2012/03/grading-student-loans.html
Brown, M., Haughwout, A., Lee, D., Scally, J., & van der Klaauw, W. (2015, February 19).
Looking at student loan defaults through a larger window. Federal Reserve Bank of New
York. Retrieved from
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 111
http://libertystreeteconomics.newyorkfed.org/2015/02/looking_at_student_loan_defaults_
through_a_larger_window.html
California Postsecondary Education Commission. (2006). Developing a statewide higher
education affordability policy (Commission Report 06-10). Sacramento, CA: Author.
California State University (2015). First-time full-time freshmen. 2005–2014 degree-seeking
FTF campus reports (CSRDE). Retrieved from
http://asd.calstate.edu/csrde/ftf/2014htm/sys.htm
Campbell, C., & Hillman, N. (2015). A closer look at the trillion: Borrowing, repayment, and
default at Iowa’s community colleges. The Association of Community College Trustees.
Washington, D.C.
Carnevale, A. J., Cheah, B., & Hanson, A. R. (2015). The economic value of college majors.
Washington, D.C.: The Center on Education and the Workforce. Retrieved from
https://www.luminafoundation.org/files/resources/economic-value-of-college-majors.pdf
Carnevale, A. J., Jayasundera, T., & Hanson, A. R. (2012). Five ways that pay along the way to
the B.A. Washington, D.C.: The Center on Education and the Workforce.
Carnevale, A. J., Rose, S. J., & Cheah, B. (2011). The college payoff: Education, occupation,
and lifetime earnings. Washington, D.C.: The Center on Education and the Workforce.
Retrieved from http://cew.georgetown.edu/collegepayoff/
Carnevale, A. J., Strohl, J., & Melton, M. (2011). For what it’s worth: The economic value of
college majors. Washington, D.C.: The Center on Education and the Workforce.
Cellini, S. R., & Chaudhary, L. (2014). The labor market returns to a for-profit college
education. Economics of Education Review, 43, 125-140.
Chen, R., & DesJardins, S. L. (2010). Investigating the impact of financial aid on student dropout
risks: Racial and ethnic differences. The Journal of Higher Education, 81(2), 179-208.
Chen, R. (2008). Financial aid and student dropout in higher education: A heterogeneous
research approach. In J.C. Smart (Ed.). Higher Education: Handbook of Theory and
Research (v.23), pp. 209-239. Springer: New York.
Choy, S. P., & Li, X. (2006). Dealing with debt: 1992–93 bachelor’s degree recipients 10 years
later (NCES 2006-156). Washington, DC: US Department of Education, National Center
for Education Statistics.
Christman, D. E. (2000). Multiple realities: Characteristics of loan defaulters at a two-year public
institution. Community College Review, 27(4), 16-32.
Cochrane, D., & Reed, M. (2015). Student debt and the class of 2014. The Institute for College
Access and Success. Retrieved from
http://ticas.org/sites/default/files/pub_files/classof2014.pdf
College Board. (2013a). Trends in Student Aid 2013. Retrieved from
http://trends.collegeboard.org/home
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 112
College Board. (2013b). Trends in College Pricing 2013. Retrieved from
https://trends.collegeboard.org/college-pricing/figures-tables/tuition-fees-room-and-
board-over-time
College Board. (2015). Trends in College Pricing 2015. Retrieved from
https://trends.collegeboard.org/college-pricing/figures-tables/average-estimated-
undergraduate-budgets-2015-16
Corbett, C., & Hill, C. (2012). Graduating to a pay gap. The earnings of women and men one
year after college graduation. American Association of University Women. Washington,
D.C. Retrieved from http://www.aauw.org/files/2013/02/graduating-to-a-pay-gap-the-
earnings-of-women-and-men-one-year-after-college-graduation.pdf
Cunningham, A. F., & Kienzl, G. S. (2011). Delinquency: The untold story of student loan
borrowing. Washington: The Institute for Higher Education Policy. Retrieved from
http://files.eric.ed.gov/fulltext/ED517424.pdf
Davidson, A. (2015, September 8). Is college tuition really too high? New York Times. Retrieved
from http://www.nytimes.com/2015/09/13/magazine/is-college-tuition-too-
high.html?_r=1
Delisle, J., & McCann, C. (2014, September 14). Who’s not repaying student loans? More
people than you think. Forbes. Retrieved from
http://www.forbes.com/sites/jasondelisle/2014/09/26/whos-not-repaying-student-loans-
more-people-than-you-think/#1b51bf7c2fde
Deming, D. J., Goldin, C., & Katz, L. F. (2012). The for-profit postsecondary school sector:
Nimble critters or agile predators? A CAPSEE working paper. Center for Analysis of
Postsecondary Education and Employment. Retrieved from
http://eric.ed.gov.libproxy.usc.edu/PDFS/ED530124.pdf
Dynarski, S. M., & Scott-Clayton, J. E. (2006). The cost of complexity in federal student aid:
Lessons from optimal tax theory and behavioral economics. National Tax Journal, 59(2),
319-356.
Dynarski, S. (2014). An economist’s perspective on student loans in the United States. ES
Working Paper Series. Washington, D.C.: Brookings Institute. Retrieved from
https://www.brookings.edu/wp-
content/uploads/2016/06/economist_perspective_student_loans_dynarski.pdf
Dynarski, S. (2015, August 31). Why students with smallest debts have the larger problem. The
New York Times. Retrieved from http://www.nytimes.com/2015/09/01/upshot/why-
students-with-smallest-debts-need-the-greatest-help.html?_r=0
Federal Reserve Bank of New York (2012). Quarterly report on household debt and credit.
Author. Retrieved from
http://www.newyorkfed.org/research/national_economy/householdcredit/DistrictReport_
Q42011.pdf
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 113
Federal Reserve Bank of New York (2016, September). Consumer Credit – G.19. Author.
Retrieved from http://www.federalreserve.gov/releases/g19/current/
Field, K. (2010, July 11). Government vastly undercounts defaults. Chronicle of Higher
Education. Retrieved from http://chronicle.com/article/Many-More-Students-Are/66223
FinAid (n.d.). Student loans. Retrieved from
http://www.finaid.org/questions/bankruptcyexception.phtml
FinAid. (n.d.). Cohort default rates. Retrieved from
http://www.finaid.org/loans/cohortdefaultrates.phtml
Freedman, D. A., & Berk, R. A. (2008). Weighting regressions by propensity scores. Evaluation
Review, 32 (4), 392-409.
Goldrick-Rab, S., Kelchen, R., & Houle, J. (2014). The color of student debt: Implications of
federal loan program reforms for black students and historically black colleges and
universities. Wisconsin Hope Lab. Retrieved from
https://news.education.wisc.edu/docs/WebDispenser/news-connections-
pdf/thecolorofstudentdebt-draft.pdf?sfvrsn=4
Goodale, G. (2012, April 25). Student loans: as debts hit $1 trillion mark, protesters plan
Occupy-type events. Christian Science Monitor. Retrieved from
http://www.csmonitor.com/USA/Education/2012/0425/Student-loans-As-debts-hit-1-
trillion-mark-protesters-plan-Occupy-type-events.
Government Accountability Office. (2010). For-profit colleges. Undercover testing finds
colleges encouraged fraud and engaged in deceptive and questionable marketing
practices. Retrieved from http://www.harkin.senate.gov/documents/pdf/d10948t.pdf
Government Accountability Office. (2011). For-profit schools: Experiences of undercover
students enrolled in online classes at selected colleges.
Gross, J. P. K., Cekic, O., Hossler, D., & Hillman, N. (2009). What matters in student loan
default: A review of the research literature. Journal of Student Financial Aid, 39(1), 19-
29.
Haughwout, A., Lee, D., Scally, J., & van der Klaauw, W. (2015). Student loan borrowing and
repayment trends, 2015. Federal Bank of New York. Retrieved from
https://www.newyorkfed.org/medialibrary/media/newsevents/mediaadvisory/2015/Stude
nt-Loan-Press-Briefing-Presentation.pdf
Harrast, S. A. (2004). Undergraduate borrowing: A study of debtor students and their ability to
retire undergraduate loans. Journal of Student Financial Aid, 34(1), 21-37.
Hershbein, B., Harris, B., & Kearney, M. (2014). Major decisions: What graduates earn over
their lifetimes. Washington: The Hamilton Project. Retrieved from
https://test.knowledge.financial/assets/What%20Graduates%20Earn%20Over%20Their%
20Lifetimes.pdf
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 114
Horton, J. (2015). Identifying at-risk factors that affect college student success. International
Journal Of Process Education,7(1), 83-101.
Houle, J. N. (2013). Disparities in debt: Parents’ socioeconomic resources and young adult
student loan debt. Sociology of Education, 87(1), 53-69.
Huelsman, M. (2015). The Debt divide: The racial and class bias behind the “new normal” of
student borrowing. Demos.org. Retrieved from
http://www.demos.org/sites/default/files/publications/Mark-
Debt%20divide%20Final%20(SF).pdf
Ishitani, T. T., & McKitrick, S. A. (2016). Are student loan default rates linked to institutional
capacity? Journal of Student Financial Aid, 46(1), 17-37.
Issa, E. E. (2016). 2015 American household credit card debt study. Retrieved from
https://www.nerdwallet.com/blog/credit-card-data/average-credit-card-debt-household/
Kantrowitz, M. (2010). Default rate by institution level vs. degree program. FinAid.org.
Retrieved from http://www.finaid.org/educators/20100715institutionlevel.pdf
Kantrowitz, M. (2012). Who graduates college with six-figure student loan debt? FinAid.org.
Retrieved from http://www.finaid.org/educators/20120801sixfiguredebt.pdf
Kirkham, C. (2012, December 27). For-profit colleges manage student loan default rates, Senate
calls for investigation. Huffington Post. Retrieved from
http://www.huffingtonpost.com/2012/12/27/for-profit-colleges-student-loan-
default_n_2371688.html
Knapp, L. G., Kelly-Reid, J. E., & Ginder, S. A. (2012). Enrollment in Postsecondary
Institutions, Fall 2010; Financial Statistics, Fiscal Year 2010; and Graduation Rates,
Selected Cohorts, 2002–07. U.S. Department of Education. Retrieved from
http://nces.ed.gov/pubs2012/2012280.pdf
Koricich, A. (2014, April). The effects of rurality on college access and choice. Presented at the
annual conference of the American Education Research Association, Philadelphia, PA.
Laing, J. R. (2012, April 14). What a drag! At $1 trillion and climbing, the growing student-loan
debt could be a burden on economic growth for decades to come. Barron’s.
Lam, T. V. (2012, April 23). What America owes in student loans. National Public Radio.
Retrieved from http://www.npr.org/blogs/money/2012/04/18/150909686/what-america-
owes-in-student-loans
Lang, K., & Weinstein, R. (2012). Evaluating student outcomes at for-profit colleges. Working
paper 18201. Washington, D.C.: National Bureau of Economic Research. Retrieved from
http://www.nber.org/papers/w18201
Looney, A., & Yannelis, C. (2015). A crisis in student loans? How changes in the characteristics
of borrowers and in the institutions they attended contributed to rising loan defaults.
Brookings Papers on Economic Activity, Fall(2015), 1-89. Retrieved from
https://www.brookings.edu/bpea-articles/a-crisis-in-student-loans-how-changes-in-the-
characteristics-of-borrowers-and-in-the-institutions-they-attended-contributed-to-rising-
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 115
loan-defaults/
Martin, R. E., & Gillen, A. (2011). Measuring College Affordability. Retrieved from SSRN:
http://ssrn.com/abstract=1734914 or http://dx.doi.org/10.2139/ssrn.1734914
McMillion, R. (2004). Student loan default: Literature review. Austin, TX: Texas Guaranteed.
Mezza, A. A., & Sommer, K. (2015). A trillion dollar question: What predicts student loan
delinquencies? Federal Reserve Board. Washington, D.C. Retrieved from
https://www.federalreserve.gov/econresdata/feds/2015/files/2015098pap.pdf
Murnane, R. J., & Willett, J. B. (2011). Methods matter: Improving causal inference in
educational and social science research. New York: Oxford University Press.
National Association of Student Financial Aid Administrators (NASFAA). (2013). Student aid
profile: A review of 2012 federal programs. Retrieved from
http://www.nasfaa.org/national-profile/
National Center for Education Statistics. (n.d.). Definitions and data. Who is nontraditional?
Retrieved from http://nces.ed.gov/pubs/web/97578e.asp
National Center for Education Statistics. (n.d.). Fast facts. Graduation rates. Retrieved from
http://nces.ed.gov/fastfacts/display.asp?id=40
National Center for Education Statistics. (n.d.). Fast facts. Tuition costs of colleges and
universities. Retrieved from http://nces.ed.gov/fastfacts/display.asp?id=76
National Center for Education Statistics. (n.d.). Fast facts. Enrollment. Retrieved from
http://nces.ed.gov/fastfacts/display.asp?id=98
National Center for Education Statistics. (2011a). Table 349. Average undergraduate tuition and
fees and room and board rates charged for full-time students in degree-granting
institutions, by level and control of institution: 1964-65 through 2010-11. Retrieved from
http://nces.ed.gov/programs/digest/d11/tables/dt11_349.asp?referrer=report
National Center for Education Statistics (2014). Table 303.20 Total fall enrollment in all
postsecondary institutions participating in Title IV programs and annual percentage
change in enrollment, by degree-granting status and control of institution: 1995-2014.
Retrieved from
http://nces.ed.gov/programs/digest/d15/tables/dt15_303.20.asp?current=yes
National Center for Education Statistics (2016). Digest of Education Statistics, 2016.
Introduction. Retrieved from http://nces.ed.gov/programs/digest/d11/
National Student Clearinghouse (2015). Current term enrollment estimates. Fall 2015. Retrieved
from https://nscresearchcenter.org/currenttermenrollmentestimate-fall2015/
Newman, D., Newman, I., & Salzman, J. (2010). Comparing OLS and HLM models and the
questions they answer: Potential concerns for Type VI errors. Multiple Linear Regression
Viewpoints, 36(1), 1-15.
Office of Student Financial Assistance Program (2016). Official default rates for schools.
Retrieved from
http://www2.ed.gov/offices/OSFAP/defaultmanagement/schooltyperates.pdf
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 116
Perna, L. W., & Titus, M. A. (2005). The relationship between parental involvement and social
capital and college enrollment: An examination of racial/ethnic group differences.
Journal of Higher Education, 76 (5), 485-518.
Pianin, E. (2012, March 10). Student loans seen as potential “next debt bomb” for the economy.
The Washington Post. Retrieved from http://articles.washingtonpost.com/2012-03-
10/business/35449029_1_student-loan-private-education-loans-federally-subsidized-
loans
Podgursky, M., Ehlert, M., Monroe, R., Watson, D., & Wittstruck, J. (2002). Student loan
defaults and enrollment persistence. Journal of Student Financial Aid, 32(3), 27-42.
Project on Student Debt. (2010). Quick facts about student debt. Updated January 2010.
Retrieved from http://projectonstudentdebt.org/files/File/Debt_Facts_and_Sources.pdf
Proctor, B. D., Semega, J. L., & Kollar, M. A. (2016). Income and poverty in the United States:
2015. U.S. Department of Commerce Economics and Statistics Administration.
Washington, D.C. Retrieved from
http://www.census.gov/content/dam/Census/library/publications/2016/demo/p60-256.pdf
Rabe-Hesketh, S., & Skrondal, A. (2012). Multilevel and longitudinal modeling using Stata.
Volume II: Categorical responses, counts and survival. College Station, Texas: Stata
Press.
Rampell, C. (March 2, 2012). Why tuition has skyrocketed at state schools. Retrieved from
http://economix.blogs.nytimes.com/2012/03/02/why-tuition-has-skyrocketed-at-state-
schools/?_r=0
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data
analysis methods (2
nd
ed.). Thousand Oaks, CA: Sage Publications.
Rosenbaum, P. R., & Rubin, D. B. (1983). The central role of the propensity score in
observational studies for causal effects. Biometrika, 70(1), 41-55.
Saez, E. (2015). Striking it richer: The evolution of top incomes in the United States (updated
with 2013 estimates). Retrieved from http://eml.berkeley.edu/~saez/saez-UStopincomes-
2013.pdf
Sallie Mae. (n.d.). Preventing defaults. Retrieved, April 30, 2012, from
http://www020t.salliemae.com/NR/rdonlyres/DCD47E82-8014-4A47-93B4-
9FE58E887C77/11343/SallieMaePreventingDefaultsOverview1.pdf
Senate Committee on Health, Education, Labor & Pensions (2012). Senate Print 112-37,
Volumes 1-4. For profit higher education: The failure to safeguard the federal investment
and ensure student success. Retrieved from
http://www.gpo.gov/fdsys/browse/committeecong.action?collection=CPRT&committee=
health&chamber=senate&congressplus=112&ycord=0
Simms, M. C., Fortuny, K., & Henderson, E. (2009). Racial and ethnic disparities among low-
income families. The Urban Institute. Retrieved from
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 117
http://www.urban.org/sites/default/files/alfresco/publication-pdfs/411936-Racial-and-
Ethnic-Disparities-Among-Low-Income-Families.PDF
Steiner, M., & Teszler, N. (2005). Multivariate analysis of student loan defaulters at Texas A&M.
Austin, TX: TG Research and Analytical Services. Retrieved from
http://files.eric.ed.gov/fulltext/ED544417.pdf
Stinson, S. (2013, September 27). How financial literacy programs control student debt.
University Business. Retrieved from https://www.universitybusiness.com/article/how-
financial-literacy-programs-control-student-debt
Swimdell, B., Wright, G., & Gallegos, S. (2013). New data confirm troubling student loan
default problems. Retrieved from
http://ticas.org/sites/default/files/pub_files/CDR_2013_NR.pdf
Trumbull, M. (2012, February 14). America’s wealth gap: is it good, bad or irrelevant? Christian
Science Monitor. Retrieved from
http://www.csmonitor.com/USA/Politics/2012/0214/America-s-big-wealth-gap-Is-it-
good-bad-or-irrelevant
The Institute for College Access and Success. (2014). Quick facts about student debt. Retrieved
from http://ticas.org/sites/default/files/pub_files/Debt_Facts_and_Sources.pdf
The Pell Institute. (2015). Indicators of higher education equity in the United States. 45-year
trend report, 2015 revised edition. Retrieved from
http://www.pellinstitute.org/downloads/publications-
Indicators_of_Higher_Education_Equity_in_the_US_45_Year_Trend_Report.pdf
U.S. Census Bureau. (2012). Statistical Abstract of the United States: 2012 (131st Edition).
Washington, DC. Retrieved from http://www.census.gov/compendia/statab/
U.S. Department of Commerce Bureau of Economic Analysis. (n.d.). U.S. economy at a glance:
Perspective from the BEA accounts. Retrieved from
http://www.bea.gov/newsreleases/glance.htm
U.S. Department of Education. (2011, September 12). Default rates rise for federal student loans.
Retrieved from http://www.ed.gov/news/press-releases/default-rates-rise-federal-student-
loans
U.S. Department of Education. (2012, September 28). First official three-year student loan
default rates published. Department continues efforts to help students better manage their
debt. Retrieved from http://www.ed.gov/news/press-releases/first-official-three-year-
student-loan-default-rates-published
U.S. Department of Education. (n.d.). Facing loan default. Retrieved from
http://www2.ed.gov/offices/OSFAP/DCS/default.html
U.S. Department of Education. (n.d.). Three-year official cohort default rates by schools.
Retrieved from http://www2.ed.gov/offices/OSFAP/defaultmanagement/cdr.html
U.S. Department of Education. (September 30, 2013). Default rates continue to rise for federal
CHOICE OF MAJOR & STUDENT LOAN DEFAULT 118
student loans. Retrieved from http://www.ed.gov/news/press-releases/default-rates-
continue-rise-federal-student-loans
University of California. (2015). Accountability report 2015: Undergraduate student success.
Retrieved from http://accountability.universityofcalifornia.edu/2015/chapters/chapter-
3.html
University of Southern California. (2016). First-year student profile and admission information
2016-2017. Retrieved from https://about.usc.edu/files/2016/10/Freshman-Profile-
2016.pdf
Urahn, S. K., & Conroy, T. P. (2015). Federal and state funding of higher education: A changing
landscape. The Pew Charitable Trusts. Retrieved from
http://www.pewtrusts.org/~/media/assets/2015/06/federal_state_funding_higher_educatio
n_final.pdf
Valenti, J., Edelman, S., & Van Ostern, T. (April 10, 2013). Student-loan debt has a negative
rippling effect on the economy. Center for American Progress. Retrieved from
http://www.americanprogress.org/issues/higher-
education/news/2013/04/10/60173/student-loan-debt-has-a-rippling-negative-effect-on-
the-broader-economy/
Webber, K. L., & Rogers, S. L. (2014). Student loan default: Do characteristics of four-year
institution contribute to the puzzle? Journal of Financial Aid, 44(2), 99-124.
Weismann, J. (August 8, 2013). Our student loan system is broken, and these new statistics prove
it. The Atlantic. Retrieved from
http://www.theatlantic.com/business/archive/2013/08/our-student-loan-system-is-broken-
and-these-new-statistics-prove-it/278453/
White House (2010). Ensuring that student loans are affordable. Retrieved from
https://nscresearchcenter.org/currenttermenrollmentestimate-fall2015/
Woo, J. H. (2002a). Factors affecting the probability of default: Student loans in California.
Journal of Student Financial Aid, 32(2), 5-25.
Woo, J. H. (2002b). Clearing accounts: The causes of student loan default. Rancho Cordova,
CA: EdFund
Abstract (if available)
Abstract
Current student loans in the United States have reached 1.3 trillion dollars. Big loans, however, are not big problems unless they cannot be repaid, but numbers show that the default rate has been rising from single digits to 12.9% in 2011, particularly among low-income groups and college drop outs. Default rates are 50% higher in for-profit schools than at public universities. This study is about student loan defaults, specifically, whether a student’s choice of major is associated with a higher rate of student loan default. ❧ Previous research from the 1990’s showed that a student’s personal characteristics such as race, gender, high school academic achievement and social economic status are strong indicators of whether he or she defaults or not. More recent research, specifically with hierarchical linear modeling where linear regressions are performed within groups, have shown that default rate differ by institutions. Default rates are highest in for profits at around 21%, with public universities at 13% and private non-profit universities at 7%. ❧ I used multinomial logit regression and propensity scored hierarchical linear modeling on NCES’s Beginning Postsecondary Students 04/09 database to find out if one’s chosen major contributes in any way to a student defaulting on his or her loans. The study aims to answer the following questions: 1. To what extent do total cumulated student loans differ across fields of study (both career/vocational and degree programs)? 2. To what extent do student loan default rates differ across fields of study (both career/vocational and degree programs)? 3. Controlling for field of study, to what extent do institutional factors explain differences in the likelihood of default? 4. To what extent do different fields of study affect graduates’ employment, and how does this differ among different types of institutions? ❧ The study did not find any statistical significance between a student’s major and his/her inability to repay loans. However, there is an association between a student’s major and how much he or she borrows in total. Since college tuition is generally assessed on a per-unit basis, regardless of major, the cost of a degree should be comparable for students studying different subjects in the same institution. It warrants future research to identify why some students found the need to borrow more in some majors but not in others. ❧ Institutional factor plays a big role in determining a student’s likelihood to default. Students at for-profits are much more likely than students from private non-profit and public universities to default on their loans, even though the majority of them borrow less than $15,000. Students at private nonprofit universities borrow most, but they are the least likely of the three groups to default. With regards to the odds of employment, again, graduates from for-profit institutions have a lower employment rate compared to graduates from public and private non-profits. ❧ Across all three types of institutions, employment prospects are higher if a student graduates than if he/ she drops out. In this study, only Education majors have higher odds of finding employment. This was a little surprising because one would expect STEM majors to find work easily. One reason may be that these students have moved on to graduate school and are technically not employed. ❧ African Americans consistently borrowed more and are also more likely to default across all majors and institution types. If it is just a function of being low-income, then the Hispanic students would also be expected to borrow more and default more. But Hispanic students borrow less and the default rate is lower. More research needs to be done to unpack these differences. ❧ While this study did not find an association between one’s chosen major and the likelihood of default, the problem of default is real and needs to be addressed. Total loan amount continues to increase, and the total dollar amount of loan defaults is not decreasing and may even be rising. The costs of default are high both to the individual student borrower and to taxpayers, and it is well worth the effort to identify variables that affect student loan default and to find solutions to address them.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Institutional student loan cohort default rates by institution type
PDF
Decision-making and graduate school debt: a focus on master's degree students
PDF
Federal loan borrowing in community colleges: examining the decision making processes of non‐traditional community college students
PDF
Decision-making and graduate school debt: a focus on master's degree students
PDF
Decision-making and graduate school debt: a focus on master's degree students
PDF
The relationship between academic capitalism and student culture at two four-year higher education institutions
PDF
High-achieving yet underprepared: first generation youth and the challenge of college readiness
PDF
Three essays on the high school to community college STEM pathway
PDF
The intersection of grit and social capital: a mixed methods examination of successful first-generation college students
PDF
Oppression of remedial reading community college students and their academic success rates: student perspectives of the unquantified challenges faced
PDF
Undocumented college students: pursuing academic goals against the odds
PDF
Exploring how individual transition plans in a school district support special education student employment
PDF
A multi-perspective examination of developmental education: student progression, institutional assessment and placement policies, and statewide regulations
PDF
Evaluating the problem of college students graduating into underemployment and unemployment
PDF
Math and the making of college opportunity: persistent problems and possibilities for reform
PDF
Evaluating the effects of diversity courses and student diversity experiences on undergraduate students' democratic values at a private urban research institution
PDF
Conceptualizing college writing readiness for the 21st century: a tale of two classrooms
PDF
To what extent does being a former high school English learner predict success in college mathematics? Evidence of Latinx students’ duality as math achievers
PDF
Student mobility in policy and poverty context: two essays from Washington
PDF
Collaborative social networks in student affairs: an exploration of the outcomes and strategies associated with cross‐institutional collaboration
Asset Metadata
Creator
Parry, Shirley
(author)
Core Title
Investigating the association of student choices of major on college student loan default: a propensity-scored hierarchical linear model
School
Rossier School of Education
Degree
Doctor of Philosophy
Degree Program
Urban Education Policy
Publication Date
02/15/2017
Defense Date
01/18/2017
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
college major,employment status,for-profit,Higher education,inability to pay,OAI-PMH Harvest,student loan default
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Tierney, William G. (
committee chair
), Melguizo, Tatiana (
committee member
), Slaughter, John B. (
committee member
)
Creator Email
shirley.c.parry@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c40-337471
Unique identifier
UC11257922
Identifier
etd-ParryShirl-5007.pdf (filename),usctheses-c40-337471 (legacy record id)
Legacy Identifier
etd-ParryShirl-5007.pdf
Dmrecord
337471
Document Type
Dissertation
Rights
Parry, Shirley
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
college major
employment status
for-profit
inability to pay
student loan default