Printer Friendly

A crisis in student loans? How changes in the characteristics of borrowers and in the institutions they attended contributed to rising loan defaults.

ABSTRACT This paper examines the rise in student loan default and delinquency. It draws on a unique set of administrative data on federal student borrowing matched to earnings records from de-identified tax records. Most of the increase in default is associated with borrowers at for-profit schools, 2-year institutions, and certain other nonselective institutions. Historically, students at these institutions have constituted a small share of all student borrowers. These nontraditional borrowers have largely come from lower-income families, attended institutions with relatively weak educational outcomes, faced poor labor market outcomes after leaving school, and defaulted at high rates. In contrast, default rates have remained low among borrowers who attended most 4-year public and nonprofit private institutions and among graduate school borrowers--who collectively represent the vast majority of the federal loan portfolio--despite the severe recession and these borrowers' relatively high loan balances. The higher earnings, low rates of unemployment, and greater family resources of this latter category of borrowers appear to have helped them avoid adverse loan outcomes even during times of hardship. Decomposition analysis indicates that changes in the characteristics of borrowers and the institutions they attended are associated with much of the doubling in default rates between 2000 and 2011, with changes in the type of schools attended, debt burdens, and labor market outcomes explaining the largest share.

**********

Between 2000 and 2014, the total volume of outstanding federal student debt nearly quadrupled to surpass $1.1 trillion, the number of student loan borrowers more than doubled to reach 42 million, and default rates among recent student loan borrowers rose to their highest levels in 20 years. This increase in debt and default and more widespread concern about the effects of student loan debt on young Americans' lives has contributed to a belief that there is a crisis in student loans. Using new administrative data sources, we examine recent changes in the market for federal student loans with a particular focus on the sources of rising default rates, the roles played by educational institutions, and the labor market outcomes of borrowers.

These data show that to the extent that there is a crisis, it is concentrated among borrowers who attended for-profit schools and, to a lesser extent, 2-year institutions and certain other nonselective institutions. We refer to these borrowers as "nontraditional" because, as students, they tend to be older, often enroll less than full time, and are living independently of their parents, and also because historically there were relatively few for-profit students and because 2-year students rarely borrowed. As a result, in 2000 these borrowers represented a small share of all federal student loan borrowers and an even smaller share of loan balances.

However, during and soon after the recession, the number of nontraditional borrowers grew to represent almost half of all new borrowers. They experienced poor labor market outcomes, had few family resources, and owed high debt burdens relative to their earnings. Their default rates skyrocketed. Of all the students who left school, started to repay federal loans in 2011, and had fallen into default by 2013, about 70 percent were nontraditional borrowers. (1)

In contrast, the majority of undergraduate and graduate borrowers from 4-year public and private (nonprofit) institutions, or "traditional borrowers," have experienced strong labor market outcomes and low rates of default, despite having the largest loan balances and facing the severe headwinds of the recent recession. While the number of traditional borrowers also increased rapidly over time, recent borrowers' family backgrounds and labor market outcomes are not much different from their peers' in earlier years, especially for graduate students and undergraduates at relatively selective institutions. In fact, traditional borrowers earned more, on average, in 2013 than their peers had in 2002. While graduates in the late 2000s were hit harder than other cohorts by the recession, the unemployment rate of traditional borrowers who left school and started repaying their loans in 2011 was 7.7 percent in 2013 compared to 6.6 percent for the comparable cohort of recent borrowers in 2002.

These results derive from a new database formed by the merger of administrative records on student loan burdens to earnings information from de-identified tax records. The data provide annual information on student characteristics, the institutions they attended, loan balances and loan status from 1970 to 2014, and labor market outcomes from 1999 to 2013 for a 4-percent sample of all federal student borrowers. The sample includes about 46 million annual observations on 4 million individual borrowers, assembled from hundreds of millions of individual records of loan transactions, aid applications, and earnings records. These data were assembled to improve budget estimates and inform policy regarding programs with both spending and tax components. They also provide unique advantages over prior survey and credit-panel data sets because they allow detailed examination of the role played in student loan defaults by institutions, labor market outcomes, and other potential contributing factors. (2)

These data show that the number of new nontraditional borrowers increased steadily since the mid-1990s, as enrollment in for-profit institutions returned to growth after having declined earlier in the 1990s and then surged during the recession and as the weak labor market boosted enrollment and increased borrowing rates, particularly among 2-year students. (3) Because of the relatively short enrollment durations of many of these new borrowers, the combination of new enrollment and rapid turnover resulted in a flood of nontraditional borrowers, disproportionate to their share of enrollment, who were out of school and into loan repayment after the recession. For instance, in 2011, while for-profit students made up only 9 percent of all postsecondary students (according to the National Center for Education Statistics) and 25 percent of all active federal borrowers, they represented more than 31 percent of borrowers leaving school and starting to repay federal loans that year. Combined with students from 2-year institutions, who represented an additional 16 percent of borrowers starting to repay loans that year, this meant that almost half of borrowers in their first years of repayment were nontraditional borrowers.

In addition to being more numerous and in their earliest years of loan repayment after the recession, recent nontraditional borrowers appear to be a particularly high-risk population. They tend to be older when they first enroll, to be from lower-income families, and to live in poorer neighborhoods. They are more likely to be first-generation borrowers. They attend programs they are less likely to complete and, after enrollment, are more likely to live in or near poverty and to experience weak labor market outcomes, outcomes that worsened disproportionately during the recession. And their loan burdens, though smaller on average both in absolute terms and relative to their earnings, have tended to increase faster over time.

All these factors contributed to high default rates among nontraditional borrowers. About 30 percent of nontraditional borrowers required to start repayment on loans in 2011 defaulted within three years, compared to 13 percent among traditional undergraduate borrowers and 3 percent among graduate borrowers. Many more appear to be struggling with their loans but have avoided default through protections such as forbearance, deferment, and income-based repayment programs, which allow borrowers to suspend or make reduced payments during times of hardship. Using decomposition analysis, we find that changes in observable characteristics--like the backgrounds of students, their labor market outcomes, and the schools they attend--can explain between half and two-thirds of this increase in default, with changes in the types of institutions attended alone explaining between one-quarter and one-half of the increase in default rates between 2000 and 2011. However, much of the increase in default rates, particularly among nontraditional borrowers, cannot be explained simply by factors like their family background or labor market outcomes, suggesting that factors we cannot observe, such as the quality of the education received, students' satisfaction with their institutions, and other financial or economic difficulties specific to nontraditional borrowers, may also be driving up default rates.

These high rates of default are unlikely to persist because of the recent normalization in enrollment patterns post-recession, increased scrutiny and policing of for-profit institutions, and other factors that have contributed to a decline in the number of nontraditional borrowers. From 2010 to 2014, the number of new borrowers fell by 44 percent at for-profit schools and by 19 percent at 2-year institutions. Because of the relatively long life cycle of a student loan, these changes will not be fully felt for several years. (4) In addition, rising enrollment in income-based repayment programs will help many borrowers experiencing economic hardships avoid default. The decline in the rate of new borrowers and an uptick in the number of borrowers paying off loans have already contributed to a sharp slowdown in the growth rate of borrowers and aggregate debt. In 2014, for instance, the number of borrowers increased by about 1 million, down from an average annual increase of 2 million from 2009 to 2012.

One reason traditional student loan borrowers have avoided default is that they experience favorable labor market outcomes, with low rates of nonemployment (even in the recession) and relatively high earnings; moreover, they are more likely to come from higher-income families in the first place. Most traditional borrowers have not accumulated large balances. While average debt burdens have increased and some borrowers have accumulated very large balances (4 percent of borrowers had balances over $100,000 and 14 percent had balances over $50,000 in 2014), most borrowers with large balances are graduate students, parents, and "independent" undergraduate borrowers often from for-profit schools. Indeed, one consequence of these patterns is that borrowers in the top 20 percent of the income distribution owe more than one-third of outstanding student loan debt.

Beyond examining the sources of the rise in default and delinquency, these data also inform a broader debate regarding the implications of rising student indebtedness. One concern is that rising rates of default reflect excessive borrowing and overextended finances, which could impair students' abilities to finance first homes and to live independently of their families, or could constrain their occupational choices, reducing rates of homeownership and marriage, or their entrepreneurial risk taking. (5) Our results suggest a potentially different interpretation for many of the observed relationships between rising student borrowing and worsening outcomes: a shift in the composition of borrowers toward higher-risk or more disadvantaged individuals. Just as these shifts contribute to higher default rates, they may also contribute to lower rates of homeownership or to constrained occupational choices. In fact, increases in default rates due to compositional shifts could overshadow relatively beneficial investments in higher education, which may be less worrisome or even desirable (Akers and Chingos 2014; Avery and Turner 2012; Dynarski and Kreisman 2013; Sun and Yannelis 2016). Indeed, for most borrowers (and the majority of the student loan portfolio) the educational investments financed with their loans are associated with favorable economic outcomes, and most borrowers appear able to manage their debt even during recessionary periods.

Ultimately, an important question for understanding the welfare consequences of student aid programs and for developing new policies to improve their effectiveness is whether attending college was worthwhile for borrowers, even those with high rates of default. On average, education is among the most productive of investments individuals can make because the benefits of higher earnings and better well-being accumulate over a lifetime (Greenstone and Looney 2011). But the relatively weak labor market performance, high default rates, and increasing debt burdens of many borrowers raise concerns that not all students are better off. One specific area of concern is that the costs of education and the debts of borrowers have increased relative to labor market returns. Understanding why costs and debt burdens are rising, and parsing out the relative contributions of the recession's effects on enrollment, households' savings and ability to borrow, contraction in public support for education, rising costs of attendance, and other factors, would help identify whether steep increases are associated with changes in the return on students' educational investments.

Similarly, on the other side of the cost-benefit ledger, the benefits of loan-financed education depend on the quality of education provided and the labor market return specific to those investments. That requires knowing not just how borrowers are doing today, but the difficult-to-measure counterfactual of how they would have fared if they had not attended a particular institution. Understanding the differences in the costs and returns to different institutions--and how to encourage higher-return investments--is therefore a key predicate for improving federal loan programs. This appears to be especially true for nontraditional borrowers, for whom educational opportunities appear to vary more in cost and quality, and for whom their educational and financing choices appear to have much larger implications for their longer-term well-being.

The remainder of this paper is organized as follows. In section I we provide background on the structure of federal student loan programs. In section II we discuss the data sources used in the paper. In section III we analyze the factors associated with the increase in student loan debt and discuss the rise of nontraditional borrowers and the implications for borrowing and for borrowers' default and labor market outcomes. In section IV we provide an analysis of the characteristics of nontraditional borrowers and their backgrounds. In section V we provide information on borrowers' labor market outcomes. In section VI we analyze the debt burdens of borrowers over time. In section VII we focus on a key outcome--loan repayment--and conduct regression and decomposition analysis of factors associated with the rise in student loan default; we then discuss the flows of borrowers during and after the Great Recession as well as the potential implications for future repayment. In section VIII we conclude and provide suggestions for further research.

I. Background: The Structure of Federal Student Loan Programs

The analysis in this paper focuses on federal student lending programs, which were first established in 1958 to provide low-cost loans to students and were subsequently expanded several times, notably under the Higher Education Act of 1965. These federal student lending programs accounted for the nation's largest source of nonmortgage household debt in 2014. (6) The aim of these student loan programs was to alleviate credit constraints for borrowers, who internalize many of the benefits of education. (7)

In 1966, President Lyndon Johnson articulated the purpose of the student loan program in these words:

Under this new loan program, families will finance college education for their children in the same way that they finance the purchase of a home: through long-term, federally guaranteed private loans. For millions of families, the financial burden of college education will now be lifted; new opportunities will open for American students. (Johnson 1966)

The vast majority of student loans in the United States are federally guaranteed or direct loans made by the Department of Education. (8) The main federal lending program today is the Federal Direct Loan program, which was created by the Higher Education Amendments of 1992. Since 2010, the Direct Loan program has accounted for all federal student loans. Under this program, postsecondary institutions originate loans under federal lending rules, and loan servicing is handled by the Department of Education through private servicing contractors.

Direct Loans, which can be made both to undergraduate and graduate students, come in four types: Unsubsidized Stafford, Subsidized Stafford, PLUS, and consolidation loans. Unsubsidized, PLUS, and consolidation loans are available to all borrowers attending eligible institutions, while Subsidized loans are available based on a financial needs test. One main difference between Unsubsidized and Subsidized loans is that interest does not accrue for borrowers of Subsidized loans while they are in school. PLUS loans are available to the parents of dependent undergraduate students (Parent PLUS) and to graduate and professional students. Independent undergraduate students are not eligible for PLUS loans, but are allowed to borrow additional Stafford loans up to higher maximums. Consolidation loans allow students to combine all of their federal loans into one loan to simplify payments. Loan limits are set by legislation, and loans can be used only to meet education expenses like tuition and other costs of attendance. Fees for Direct Loans were raised slightly following the 2013 budget sequestration; borrowers are charged an origination fee of 1 percent for Stafford Loans and 4 percent for PLUS loans.

Table 1 provides an overview of these federal borrowing programs. The table presents the total loan balance in each fiscal year (over $1.1 trillion in 2014) and the total number of borrowers (roughly 42.8 million in 2014); it also disaggregates those figures into the shares of loans and borrower types each year. Over time, graduate loans and Parent PLUS loans have increased as a share of federal lending. In 1994, about 68 percent of the portfolio was undergraduate loans, and by 2014 that rate had declined to about 59 percent of the portfolio. In terms of numbers of borrowers, the same growth in graduate and parent loans is apparent. However, much of the increase in graduate debt is held by a rising share of students taking out both graduate and undergraduate loans. The persistence of borrowing at the undergraduate and graduate levels, and the increases in graduate and parent loans (whose loan amounts are limited only by costs of attendance), prove to be important reasons why aggregate and per-student loan amounts increase over time.

Prior to the Federal Direct Loan program, the Federal Family Education Loan (FFEL) program also disbursed federally guaranteed loans through private lenders following lending rules for federally guaranteed loans. The main difference between the programs was financing through private capital or direct federal funds, and students saw few differences in lending rules. (9) The Perkins loan program provides additional loans to low-income borrowers with exceptional financial need. Perkins loans make up a small share of federal student loan programs, and the analysis in this paper excludes them.

Interest rates are set by Congress and were identical for Stafford borrowers under the Direct Loan and FFEL programs, but can vary for graduate and undergraduate borrowers. Historically, undergraduate Stafford loan interest rates have been both fixed and variable, depending on the year, and rates have varied between 8.25 percent (1999) and 3.4 percent (2004). In the 2010-11 academic year, interest rates were 6.8 percent; in 2015, they dropped to 4.25 percent.

For most loans, after leaving school, repayment begins after a six-month grace period. Once repayment begins, payment can be stopped through either deferment or forbearance. Loans can go into deferment if a borrower re-enrolls in school, becomes unemployed (for up to 3 years), faces economic hardship, or joins the military or the Peace Corps. Forbearance also allows borrowers to defer loans for up to one year if they are ill, face financial hardship, or perform national service. Interest typically continues to accrue while loans are in deferral or forbearance. (10)

The standard repayment plan for student loans is a 10-year plan. Extended repayment plans of up to 25 years are also available to many borrowers with large balances. In addition, income-based and income-contingent repayment options are available to many borrowers with low incomes and high relative debt burdens. Historically, take-up of income-driven repayment plans has been low, although it has been rising in recent years. Bruce Chapman (1997, 2006) provides a discussion of many of the theoretical issues related to income-contingent repayment plans as well as an overview of income-contingent repayment plans in an international context.

Under the Income-Based Repayment and Pay As You Earn plans, borrowers pay the lesser of 10-15 percent of their income or their payment under a 10-year plan. Under the Income-Contingent Repayment plan, borrowers pay the lesser of 20 percent of their discretionary income or what they would pay under a fixed repayment plan. In many cases, borrowers can pay more under an income-contingent plan than they would under the standard plan, but in some cases loan balances can be discharged after 25 years in repayment. Eligibility for these programs has varied historically depending on the type of loan, time of entry into borrowing, entry into repayment, and debt-to-income ratios. In 2014, the Pay As You Earn plan was made available to all borrowers regardless of entry into borrowing. (11)

Students who miss a payment are considered delinquent, and servicers are required to report delinquency to credit bureaus within 90 days. Loans are in default if delinquent for more than 270 days. Unlike other consumer loans, student loans are nearly impossible to discharge in bankruptcy. If a borrower goes into default, she loses eligibility for deferment, forbearance, and alternative repayment plans, and the loan is assigned to a collection agency. The borrower is then liable for late fees, collection costs, and accruing interest. The government is obligated to collect defaulted loan amounts using wage garnishment of up to 15 percent of the borrower's wages, and through the Treasury Offset Program, which withholds any tax refunds and certain other payments, like a portion of Social Security retirement or disability benefits. Accounting for collection costs, the Department of Education (2014) estimates that recovery rates using these methods were approximately 75 to 85 percent in 2014, compared to recovery rates of 70 percent for mortgage loans (Downs and Xu 2015). Historically, recovery rates have been lower; Deborah Lucas and Damien Moore (2010) estimate recovery rates of around 50 percent in the early 2000s.

II. New Administrative Data Sources

The estimates presented in this paper derive from a random 4-percent sample of federal student loan borrowers assembled from components of the National Student Loan Data System (NSLDS), which is the primary data system used to administer the federal loan programs described above. This data system maintains the information needed to run the loan programs, including the repayment system; assessing eligibility for loans using information from financial aid applications; disbursing loans to institutions based on the students' academic level; tracking when students withdraw or graduate to determine when they must begin repayment and if and when they enter deferment, forbearance, or alternative repayment plans; and providing the financial accounting of loan balances, interest accrual, transactions, and other changes in loan status. Hundreds of individual pieces of information contained in multiple databases, drawn from hundreds of millions of individual records of aid applications, loan transactions, and status updates are distilled into about 46 million annual observations on 4 million borrowers.

The panel, which follows the same borrowers over time starting from when they first take out a federal student loan, is based on data originally constructed by the Department of Education's Budget Service Division for use in budget projections. These files include information on student characteristics derived from each Free Application for Federal Student Aid (FAFSA) filed by students; information on each loan disbursed by Federal Student Aid (FSA), including the loan balance, its status, and changes in status over time; the institution the loan was disbursed to; and information on Pell Grants received. Information on each borrower obtained from the FAFSA is generally available only for loans originated after fiscal year 1995. However, most of the basic loan information (such as loan amounts and dates of origination, repayment, and default, and institution of study) is available from all sample borrowers starting in fiscal year 1969. (12) The sample is representative of more than 99 percent of federal loans and borrowers. However, while we include Parent PLUS loans in our tabulations of borrowing amounts, when examining the experiences of borrowers as they complete school, enter the labor market, and begin repaying their loans, we focus exclusively on student borrowers and exclude outcomes (and economic status) of parent borrowers.

These data are merged to a panel of administrative earnings and income records that span the (calendar year) period from 1999 to 2014 (data for 2014 are incomplete and preliminary). The primary data of interest are the earnings and total incomes of borrowers. Individual earnings are derived from information reports from employers (W-2s) and from self-employment earnings reported on the Schedule C of individual tax returns. Total income is the sum of all income sources reported by taxpayers; if the taxpayer is married and filing a joint return, this includes any income and earnings of the spouse. In addition, information on filing status and the number of dependent children and federal poverty levels is used to construct indicators of poverty. (13)

II.A. Sample Construction

To examine the dynamics of borrowing in the loan market, we focus on the flows of borrowers as they enter borrowing (when they originate their first federal loans) and when they enter repayment (when they start repaying their loans) and look at the relationship of those flows to the overall stock of federal loans and to aggregate student-loan outcomes. We treat the year a loan entered into repayment as a primary focus because that year is typically the first time a student exits school, enters the labor market, is required to make payments, and first becomes liable for delinquency and default. Specifically, we define "entering repayment" as the time when a borrower's last loan enters into repayment, that is, when all of a borrower's loans are in repayment. We define repayment cohorts based on the fiscal year each borrower entered repayment. These definitions closely approximate the aggregate measures of debt, default, and average loan burdens produced by the Department of Education (2014).

We define entrants as first-time borrowers, assign them to entry cohorts based on the fiscal year their first loans were originated, and use information on the students from the first loan-related FAFSA filed and the institution that originated the loan. This provides a consistent measure of new originations and borrower characteristics when borrowers first enter the loan system. Defining flows based on first-time and last-time borrowing obscures the fact that an educational career sometimes involves multiple spells of borrowing (re-entry) as students take time off, change institutions, or go to graduate school. In practice, however, this convention has little effect on our analysis of borrowers, because loan outcomes are strongly correlated within borrowers (that is, when borrowers default they default on all loans) and because most variation in loan outcomes occurs after the borrower leaves school and permanently enters the labor force. In addition, the fact that spells of borrowing may overlap at the end of the sample period introduces censoring effects. For example, some borrowers entering repayment in 2014 will subsequently return to graduate school.

II.B. Variable Construction

Most variables used in our analysis are straightforward; characteristics of borrowers, like family income, age, and gender, are taken directly from the FAFSA. Data on the neighborhoods of borrowers--local unemployment rates, poverty rates, median household income, and percent black, white, and Hispanic--are derived by matching the ZIP code provided on the first FAFSA with ZIP code-level statistics from the 2000 Decennial Census. Loan information, such as disbursements and balances, are the sum of all Direct and FFEL undergraduate and graduate loans or Parent PLUS loans at the end of the fiscal year for each borrower. We use these loan types and the reported academic level of borrowers to differentiate undergraduate from graduate borrowers. All dollar amounts are in real 2014 dollars (adjusted using the personal consumption expenditure deflator) unless noted otherwise.

In practice, over the course of an educational career students may attend multiple institutions and take out both undergraduate and graduate loans. Our general approach is to classify students based on their characteristics, the institution they attended, and their level of school when they borrow for the first time. Hence, borrowers may be classified as attending a 2-year school even though some may ultimately complete a 4-year degree elsewhere, and borrowers who start their education (and student loan borrowing) at 4-year institutions may ultimately go on to graduate or professional schools. One implication of this choice is that when we examine the eventual loan burden of a student starting to repay her loans, her loan burden may include a combination of undergraduate and graduate loans even though she started off as an undergraduate borrower.

In practice, this assumption has little effect on our conclusions because changes in enrollment between sectors are relatively rare. For example, a borrower attending a 4-year institution is generally likely to complete his education there. As a result, there is little difference in outcome measures like default rates by institution type whether they are based on first institution or last institution attended.

However, for certain loan measures, such as the number of borrowers or the amounts borrowed by institution type, whether to measure based on first or last institution attended has a greater effect, particularly for nontraditional borrowers. In particular, some individuals who started at 2-year institutions went on to attend 4-year institutions or for-profit institutions where they accumulated larger debts, and some students who started as traditional borrowers later returned to school and last attended a for-profit school, especially during the recession. Qualitatively, the enrollment and borrowing patterns are quite similar whether we use first or last institution attended, but the levels can be somewhat different. (14)

One important advantage of these data over other sources is the availability of information on the institutions that students borrowed to attend, including the specific schools, the controls (public, private, or for-profit), and types (2-year or 4-year). To illustrate the role of institutions, we present much of our analysis based on the type of institution attended. In particular, we use a common index of selectivity from Barron's Educational Series (2008) to segment students based on the control, type, and selectivity of the institutions they attended, and their level of study into six broad groups: for-profit institutions; 2-year public and private institutions (the vast majority of which are community colleges); nonselective 4-year public and private institutions (schools that Barron's reports as admitting more than 85 percent of applicants); somewhat selective institutions (which admit 75 to 85 percent of applicants); selective institutions (which admit fewer than 75 percent of applicants); and graduate-only borrowers (borrowers whose first and only loans were graduate loans). (15)

Our primary indicator of student loan distress is the 3-year cohort default rate measured as the fraction of borrowers entering repayment in a fiscal year who are in default on a federal loan within 3 years (1,095 days) from the date the loan entered repayment. We introduce several other indicators of student loan burdens or delinquency, including debt-service-to-earnings ratios and rates of negative amortization, which we define as the fraction of student loan borrowers who owe more on their loans at a specified time after they entered repayment.

These estimates of the aggregate loan volume, number of borrowers, and the cohort default rate closely mirror the official measures produced by the Department of Education. Figure 1 shows default rates over time in our sample, compared with aggregate statistics released by the Department of Education. (16) The replicated default rate closely matches the pattern of published statistics for most of the overlapping period. However, our replication clearly differs slightly for several reasons that relate to our sample construction and to the construction of the official default rate. First, our sample is a person-by-year sample, which means any student appears only once. In official statistics, a borrower who attends multiple institutions may be included multiple times. Second, we focus on the last time a borrower enters repayment, when default risks are likely to be higher. Finally, our sample includes all institutions, all undergraduate and graduate loans, and all (nonparent) borrowers, which is a broader array of programs and students than is counted in the official rate. For instance, some students may be excluded from the official rate if their institution demonstrated they met certain criteria for exclusion. Prior to 1995, direct loans were excluded from the official default rate, and borrowers from recently closed institutions appear to have been excluded, which may have affected estimates during periods when the number of participating institutions was declining sharply. For the aforementioned reasons, our estimates are likely to be somewhat higher than the official rate, particularly in earlier years.

[FIGURE 1 OMITTED]

III. The Rise of Nontraditional Borrowing and Its Consequences

A primary focus of our analysis is on the divergent outcomes of what we call nontraditional and traditional borrowers. In this section, we describe how nontraditional borrowing increased in recent years, compare this increase in borrowing to increases in enrollment measured in other data sources, examine how the increase in nontraditional borrowing affected the composition of the federal loan portfolio, and illustrate how these changes in borrowing patterns over the recession resulted in a sharp increase in the number of nontraditional borrowers out of school and into the labor market as the recession waned.

III.A. New Borrowers at For-Profit Schools, 2-Year Institutions, and Other Nonselective Institutions

As discussed earlier, we define traditional borrowers as those attending 4-year public and private institutions because they represent one case of a "typical" college student: They start college in their late teens, soon after completing high school, are dependent on their parents (or assumed to be) for aid purposes, pursue 4-year degrees and, frequently, head on to graduate study. The median age of first-time undergraduate borrowers at these schools is 19 (26 for graduate borrowers) and more than 80 percent of undergraduates at relatively selective institutions are dependents for purposes of financial aid.

In addition, these borrowers are traditional in that historically they represented a large share of federal borrowers and loan amounts. In 1999, borrowers at 4-year public and private institutions and graduate-only borrowers represented about 70 percent of new borrowers, about two-thirds of all federal student loan borrowers (the stock), and almost 80 percent of aggregate student loans outstanding. One reason for the outsized influence of 4-year public and private institutions is that these institutions, particularly the most selective private institutions and graduate professional schools, were relatively more expensive, hence students there had a greater need to borrow to attend.

Nontraditional borrowers, in contrast, constituted only a small share of federal student loan borrowers and an even smaller share of the aggregate student loan portfolio. For instance, full-time undergraduate students at for-profit schools were a relatively small share of all new full-time students in 2000 (10 percent) and an even smaller share of total full-time plus part-time enrollment (3 percent). While almost all for-profit students took out federal loans to study, their relatively small share of enrollment meant that they still represented only 20 percent of new borrowers, 14 percent of active borrowers, and 13 percent of outstanding federal loans. Hence, they were a relatively small share of both students and student-loan dollars, though a disproportionate (but still modest) share of borrowers.

While 2-year borrowers--primarily community college students--were also a small share of federal borrowers and loan amounts, the reason was quite different. Community college students rarely borrowed and when they did, they borrowed relatively small amounts. In 2000, community college students were 27 percent of new fall enrollment and 43 percent of total postsecondary undergraduate enrollment--more than double the enrollment share of private nonprofit institutions and almost 15 times the reported enrollment of the for-profit sector. However, according to the Department of Education, in the 2000-01 school year, only 15 percent of new 2-year public students borrowed. (17) As a result, they accounted for only 9 percent of active undergraduate federal borrowers in 2000 and, because their average loan amounts were relatively small, they accounted for only about 4 percent of undergraduate federal loan originations that year and 8 percent of all outstanding federal loans.

More generally, nontraditional borrowers are more likely to be older at entry; their median age at first borrowing is 24 at for-profit schools and 23 at 2-year institutions. They are more likely to be independent for financial aid purposes, which means they are subject to higher borrowing limits and less likely to draw on the support of their parents. While the characteristics of the students themselves are therefore relatively similar at for-profit and 2-year schools, the share that leave with loans and the average loan burden is much lower among community college students.

Of course, categories based solely on an institution's ownership or control and the predominant type of degree awarded miss some heterogeneity within and across groups. Nonselective 4-year public and private institutions are a particularly heterogeneous group and include many institutions whose students have background characteristics, borrowing rates, and loan and labor market outcomes more similar to students at for-profit institutions than to those at more selective 4-year institutions. They may also be unconventional in other dimensions, such as having a predominate focus on online education. The estimates we provide of the outcomes of students from nonselective 4-year public and private institutions therefore reflect an average of potentially disparate outcomes within that borrowing population.

To illustrate the changes in borrowing over time by type of institution, figure 2 (top panel) provides more perspective on the rise (and decline) of borrowing using estimates of the number of first-time borrowers at each type of institution each fiscal year. This figure shows the steady growth of the for-profit sector over the last 15 years and, especially, the surge in enrollment during the recession. In 2009 and 2010, borrowers at for-profit schools represented roughly 30 percent of new borrowers (about 980,000 new borrowers each year), and new borrowers at 2-year schools represented roughly 16 percent of new borrowers (and 19 percent at its peak in 2012 of 580,000 borrowers). Between 2006 and their respective peaks, the number of new borrowers at 2-year schools jumped by 71 percent, and the number at for-profit schools by 60 percent. While new borrowing also increased at public and private 4-year institutions during the recession, the increases from 2006 to their relative peaks were much smaller: 31 percent at nonselective schools (2011 peak), 12 percent at the most selective schools (2011), and 24 percent among graduate-only borrowers (2009). From 2009 to 2011, almost half of all new federal borrowers (45 percent) were students at either for-profit or 2-year schools.

[FIGURE 2 OMITTED]

III.B. Comparison of Enrollment and Borrowing in NCES and NSLDS

The increase in and importance of nontraditional borrowers is less visible in official statistics that report the level of total or new enrollment. Many nontraditional borrowers appear to enroll in part-time or certificate programs or enroll outside the usual academic calendar, which starts in the autumn, and therefore they appear to fall outside the definitions used in official enrollment statistics. In addition, because borrowers at 2-year institutions and for-profit institutions enroll for much shorter durations than those at other 4-year institutions, the annual level of enrollment undercounts the number of individual borrowers flowing through for-profit and community colleges.

Table 2 provides a basic comparison of the enrollment data from the National Center for Education Statistics (NCES), which is the primary public source for measuring college enrollment, with NSLDS data. The table compares the NCES measures of total fall undergraduate enrollment (including full-time and part-time students) to estimates of the number of active undergraduate borrowers and the amount they borrowed from NSLDS in 2000 and 2011. The measures are disaggregated by school type available in the NCES: public 2-year, public 4-year, private nonprofit, and for-profit. The first two columns show total fall enrollment at degree-granting institutions, followed by the number of total active borrowers taking out loans that year. The last two columns show federal undergraduate loan originations.

As table 2 shows, total fall undergraduate enrollment rose from 13.2 million in 2000 to 18.1 million in 2011, a 37 percent increase. During the same interval enrollment at for-profit institutions increased by more than 300 percent, and as a share of all students enrollment at for-profits increased from 3 to 9 percent. (18) Enrollment at 2-year public schools increased roughly 25 percent, though as a share of all students it declined from 43 to 39 percent. Prior to and during the recession, many new students enrolled in school, but those students disproportionately enrolled at for-profit institutions and at lower-than-average rates at community colleges and private nonprofit schools. These enrollment patterns contributed to increases in the number of borrowers, not only because they increased the number of eligible students but also because the composition of students shifted toward for-profit schools (where the ratio of borrowers to NCES-reported students is 1.4 to 1) and away from 2-year and private nonprofits, where fewer students tended to borrow.

In addition, the rate of borrowing among NCES-reported enrolled students increased, particularly at 2-year public institutions where borrowing rose from 7 percent of students to 20 percent. Rates of borrowing per student increased by about 20 percent at 4-year public and private institutions (from 41 to 49 percent of students at 4-year public institutions and from 52 to 63 percent at private institutions), and declined slightly at for-profits. Because enrollment at for-profits increased as well as the borrowing rates at 2-year public schools, students at these two types of schools increased as a share of all active borrowers, rising from 9 to 17 percent and from 14 to 25 percent, respectively, between 2000 and 2011. Other things equal, had enrollment increased more evenly across institution types--for instance, had more students attended 2-year schools rather than for-profit schools--and had the increase in the share of community college students who borrowed been smaller, then the number and distribution of student loan borrowers would have been quite different, with a larger share of 4-year public and private borrowers and fewer nontraditional borrowers.

The fact that there are almost 40 percent more for-profit borrowers (from NSLDS data) than reported for-profit students (from NCES data) is one indication that the NCES enrollment figures understate the number of nontraditional borrowers (see the first two columns of the lower panel of table 2). Several factors are likely to explain this discrepancy, including borrowers in nondegree programs and borrowers enrolled outside the traditional academic cycle. Increases in such students may also be one reason why the number of active borrowers rose relative to NCES measures of enrolled students at 2-year public institutions. Another reason why enrollment statistics understate the number and growth of nontraditional borrowers is that the duration of enrollment differs across institution types. A short-duration program serves more students over a given period of time because its student body turns over more quickly. As a result, for a given level of enrollment at 2-year and for-profit schools, there were disproportionately more new borrowers being produced. (19) Hence, increases in the level of enrollment at for-profit institutions and in borrowing rates at 2-year public schools had outsized impacts on the number and composition of student loan borrowers.

III.C. Implications for the Stock of Borrowers and Debt

The inflow of these new borrowers caused both the stock of outstanding debt and the borrowers to increase in size and change in composition. Table 3 shows the number of outstanding student loan borrowers from 1985 to 2014 by the type of institution they first attended. In 2000, borrowers from for-profit and 2-year institutions accounted for less than 35 percent of all borrowers; by 2014, the number of borrowers had more than doubled at for-profit schools and 2-year institutions, rising by 114 and 167 percent, respectively. The number of borrowers at selective public and private institutions increased by much less, about 69 percent. Almost 45 percent of the increase in the number of borrowers from 2000 to 2014 (when the number of borrowers increased by roughly 21.7 million) were borrowers who started at for-profit institutions (26 percent) and 2-year institutions (18 percent). (20) By 2014, almost 40 percent of all federal borrowers were nontraditional borrowers. Had the number of nontraditional borrowers instead increased at the same rate as the number of traditional borrowers over this period, there would have been roughly 3.5 million (21 percent) fewer nontraditional borrowers. Moreover, the data on enrollment rates and borrowing-per-student in table 2 suggest that if a greater share of rising enrollment had occurred at 2-year public institutions or at 4-year public and private institutions, where borrowing rates are relatively low, rather than at for-profits, or had borrowing among community college students increased at rates closer to the rates at 4-year public institutions, then there would have been even fewer nontraditional borrowers.

Similarly, the increase in the number of nontraditional borrowers is an important contributor to the increase in overall debt and to the share of debt owed by students from for-profit and 2-year institutions. Table 4 shows aggregate federal student loan debt by institution type first attended. Between 2000 and 2014, the amount of debt owed by borrowers who first attended a for-profit institution more than quadrupled from $40 billion to $189 billion, and nearly quintupled from $24 billion to $117 billion among borrowers who had first attended a 2-year public institution. The share of outstanding loan balances attributable to for-profit school students increased from 13 to 17 percent over that period, and from 8 to 10 percent among 2-year college students. (21) Loan volumes also increased rapidly at more selective institutions, although the rate of increase was slower. The share of loans owed by borrowers from 4-year public and private institutions and graduate-only borrowers correspondingly fell by almost 7 percentage points from 79 to 73 percent. Had the number of nontraditional borrowers merely increased at the same rate as traditional borrowers, and had per-student borrowing increased by the same amount, then total debt owed by nontraditional borrowers would have been $94 billion (31 percent) lower.

A more concrete picture of how changes in nontraditional borrowing have shaped the student loan market is evident in table 5, which presents estimates of the cumulative debts of students according to the institutions they last attended in 2000 and 2014 for the 25 institutions whose students owed (collectively) the most in federal student loan debt. The debt of the students associated with each institution in each of the years includes the cumulative federal loan liabilities of borrowers, including undergraduate, graduate, and parent loans, plus any accrued interest. In addition, the analysis assigns each borrower's cumulative debts--including any debts incurred at previous institutions--to the last institution of borrowing. (22) In 2000, with the exception of the University of Phoenix, all the institutions were either 4-year public or private nonprofit institutions, often state flagship universities and institutions with large graduate programs. In 2014, 8 of the 10 institutions whose students owed the most aggregate debt and 13 of the top 25 were for-profit institutions, and one private nonprofit institution was largely an online program. (23) At certain institutions, a majority of the debt was accumulated by graduate and professional students with high average balances, such as at Walden University, Nova Southeastern University, New York University, and the University of Southern California (at these institutions a larger share of the total debt owed by students was accumulated at prior institutions). At other institutions, almost all of the debt was undergraduate debt, such as at the University of Phoenix, Strayer University, Kaplan University, Ashford University, and ITT Technical Institute.

The final two columns show, respectively, the fraction of students that defaulted and the fraction of their initial balances repaid by 2014 for the 2009 cohort. First, there is substantial heterogeneity across institution types in terms of default rates and balances repaid, with students at some for-profits experiencing 5-year default rates approaching 50 percent. Second, while the share of balances repaid and default rates are highly correlated, students from some schools maintain low default rates despite not paying down their debts. This could be due to deferment, forbearance, the use of income-based repayment plans, or other plans that allow borrowers to suspend or reduce their payments without risk of default.

III.D. Post-Recession Exodus: Rapid Increase in Repayment Flows

The lower panel of figure 2 shows the subsequent flow of borrowers into repayment by institution type. The large increase in borrowing at the onset of the recession subsequently turned into a mass exodus of borrowers into repayment as the recession waned. The exodus was magnified not just by the enrollment patterns during the recession--the fact that many new borrowers sheltered from the labor market by enrolling--but by the durations of their enrollment. Many borrowers from 4-year schools extended their enrollment by staying in school somewhat longer, delaying entry into repayment. The surge in borrowing at the start of the recession was driven by relatively short-duration programs (such as 1-year and 2-year certificates and degrees) or short-duration enrollments as students rapidly dropped out. The confluence of these factors resulted in a spike of borrowers into repayment.

We will return to these dynamics later, but one observation worth noting here is that over the last several years of the period there were simply many more borrowers entering into repayment relative to the number of new borrowers. About 63 percent of borrowers and about 62 percent of the aggregate value of loans were in repayment in 2013. This was up from 53 percent of borrowers and 60 percent of the value of loans in 2007. Hence, a much larger share of borrowers (who hold a much larger share of total loan dollars) are feeling the burden of paying their loans today than in previous years. Moreover, many more borrowers are in the earliest years of repayment, a time when the loan burdens (relative to earnings) are highest and when default rates peak. Even absent any other changes in the loan market, this increase in the number and share of new entrants should be expected to result in high absolute numbers of students in default or struggling in their first years.

In addition to the sheer volume of borrowers entering repayment, the composition of borrowers and the institutions they attended changed substantially. To illustrate the magnitude of these changes, figure 3 compares the number of borrowers entering repayment in 2000 and 2011. (24) The number of (undergraduate and graduate) borrowers entering repayment from for-profit institutions increased rapidly, from about 237,000 in 2000 (18 percent of borrowers) to 930,000 in 2011 (31 percent of borrowers). At 2-year institutions, the number of borrowers entering repayment increased from about 150,000 in 2000 (12 percent) to about 470,000 in 2011 (16 percent). Hence, in 2011 borrowers from for-profit and 2-year schools represented roughly 47 percent of federal student loan borrowers entering repayment. After 2011, the number of for-profit borrowers entering repayment remained above 900,000 through 2014 and the number of 2-year borrowers continued to rise, hitting about 740,000 in 2014.

[FIGURE 3 OMITTED]

IV. Who Are These New Borrowers? Characteristics and Educational Outcomes of Nontraditional Borrowers

The changes in who borrowed and where they borrowed have important implications for the composition and credit quality of the pool of borrowers, their educational outcomes, the amount of debt borrowers accrued, and their economic well-being after enrollment. Compared to other borrowers, the students who borrowed to attend for-profit and 2-year institutions were from more disadvantaged backgrounds (based on their family income) and were older, more independent, and, especially during the recession, more likely to have struggled in the labor market. Nontraditional borrowers tended to attend institutions with relatively poor completion rates, and many appear to have failed to complete the programs they started. These latter factors, in particular, are associated with relatively poor labor market and loan outcomes.

IV.A. Demographics and Family Background

The panels of figure 4 summarize the characteristics and educational outcomes of borrowers entering into repayment in 2011 to provide an understanding of how changes in enrollment and borrowing patterns affected these borrowers' overall characteristics. In each panel, the figure presents a cross-section of the characteristics or educational outcomes of borrowers who begin to borrow at different types of institutions as undergraduates. (25) While their characteristics have changed within institutions over time, the cross-sectional differences are persistent and therefore give a fairly good indication of how shifts in the share of borrowers across types of institutions are likely to change the characteristics of the borrowing pool.

The top-left panel shows that borrowers at more selective institutions tend to come from relatively more affluent backgrounds, with the median family income of dependent borrowers at the most selective institutions being about $80,000, as compared with about $48,000 for borrowers at nonselective 4-year schools and 2-year schools and $30,000 for those at for-profit institutions. These disparities widened modestly over the preceding decade. Because higher family income is positively correlated with labor market outcomes and negatively with default, these differences are one reason why outcomes vary across groups. It is well known that for-profit students tend to come from lower-income backgrounds (Deming, Goldin, and Katz 2012; Cellini 2009), and the observed results are consistent with more vulnerable borrowers from lower-income backgrounds increasingly borrowing to enroll in these institutions.

As the top-right and middle-left panels show, nontraditional borrowers are likely to be older and independent for financial aid purposes. For the 2011 cohort, the median age at entry was 24 for for-profit schools and 23 for 2-year institutions, compared to 19 for students at 4-year institutions; for graduate-only borrowers, the median age at entry was 26. More than 90 percent of borrowers at the most selective 4-year schools were dependent borrowers as compared with 70 percent at nonselective institutions, 50 percent at 2-year institutions, and 37 percent at for-profit schools. This pattern is important for two reasons. First, older, independent borrowers may have less ability to draw on their families for support during times of hardship. Second, independent borrowers have substantially higher loan limits, which allow them to take out more loans each year and to accumulate a higher total amount.

The middle-right and bottom-left panels illustrate two other dimensions of disadvantage. The middle-right panel shows the fraction of borrowers that are reported to be first-generation college students based on the information in their financial aid forms. About 57 percent of the 2011 repayment cohort who had attended for-profit schools were first-generation postsecondary students, compared to 51 percent of those at 2-year schools, 43 percent at nonselective 4-year institutions, and 25 percent at the most selective institutions. Similarly, students from 2-year institutions and for-profit schools were more likely to live in areas with a higher fraction of households living in poverty and with a higher minority population (based on the 2000 Census).

[FIGURE 4 OMITTED]

IV.B. Educational Outcomes

Finally, the bottom-right panel in figure 4 shows that nontraditional borrowers and those from nonselective 4-year institutions appear less likely to graduate from their programs of study than borrowers from most 4-year public and private institutions, based on information reported by the institution to the NSLDS. (26) For instance, at 4-year public institutions, about 71 percent of borrowers completed a 4-year degree, as had 83 percent of borrowers at 4-year private institutions, as reported by the institutions to NSLDS. Among borrowers entering repayment in 2011 who had started at for-profit institutions, about 28 percent had completed a 4-year degree and 41 percent a 2-year degree. (Among borrowers who had started at 4-year for-profit institutions, about 49 percent had completed a 4-year degree and 6 percent a 2-year degree.) At 2-year public institutions, about 39 percent of borrowers were reported to have completed a 2-year degree and 18 percent a 4-year degree. (27) In other words, the institution of first enrollment also had important implications for whether these borrowers were likely to complete a degree and whether they would earn a 2-year or 4-year degree. As we show later, completion rates are strongly associated with subsequent student loan default.

In all, the rise of nontraditional borrowing shifted the composition to borrowers more likely to struggle with their loan burdens--toward older, mid-career borrowers; borrowers from more disadvantaged family backgrounds and poorer neighborhoods; and toward programs many were less likely to complete.

V. Labor Market Outcomes of Borrowers

Other key differences between traditional and nontraditional borrowers are their divergent labor market outcomes and the differential impact the recession has had on each group. Drawing on earnings records from tax data, we examine the labor market outcomes of these borrowers after they have entered repayment. For each repayment cohort, the labor market outcomes of borrowers differed based on the institutions they attended, with traditional borrowers earning substantially more. In addition, the earnings and employment rates of nontraditional borrowers declined much more over time, particularly during the recession.

Focusing first on the outcomes of all borrowers in repayment, including those who may have entered repayment many years earlier, shows surprising strength in the outcomes of student loan borrowers. Between 2002 and 2013, the median earnings of all employed traditional borrowers in repayment actually increased. For graduate borrowers, the increase was from $61,000 to $63,100; for borrowers from the most selective 4-year institutions, from $47,300 to $48,000. During the same period, median earnings declined among nontraditional borrowers. The declines differed across sectors: from $24,800 to $23,200 for for-profit borrowers and from $30,100 to $25,900 for 2-year borrowers. Unemployment rates edged up slightly (by about 1 percentage point) among all groups. (28)

Focusing more narrowly on borrowers in their earliest years of repayment, we find that while the pattern of relative outcomes is similar, all groups fared worse. The top panel of figure 5 shows the unemployment rate of borrowers by institution type and compares the experiences of the 2000 and 2011 cohorts two years after entering repayment. For the 2000 cohort, unemployment rates among traditional borrowers were low, ranging from 6.3 percent for borrowers from selective 4-year schools to 6.5 percent for graduate borrowers and 10 percent for borrowers from nonselective 4-year schools. For the same cohort, the unemployment rate for 2-year borrowers was 12.2 percent and 13.2 percent for for-profit borrowers. Hence, even prior to the recent recession, there were large differences in employment across borrowers by institution type.

During the recession, unemployment rates rose substantially for nontraditional borrowers, but they rose much less among other borrowers. Among for-profit borrowers the rate jumped to 20.6 percent and among 2-year borrowers to 16.9 percent. For relatively selective 4-year borrowers, the rate increased from 6.3 to 7.2 percent, and for graduate borrowers it increased from 6.5 to 7.1 percent. In other words, even among students leaving school in 2010 and 2011, near the peak of the recession, there was almost no change in the rate of employment among most traditional borrowers. While the insulating effects of a college degree are apparent in the aggregate unemployment statistics, it is clear that those effects also applied to even most young college borrowers in the years immediately after enrollment.

[FIGURE 5 OMITTED]

For those who did find work, a similar pattern applies to their earnings. The bottom panel of figure 5 presents the median earnings of borrowers with earnings of at least $ 1,000 by the institution type they first attended for the 2000 and 2011 repayment cohorts. In both cohorts, graduate-only borrowers and borrowers from more selective 4-year institutions earned substantially more than other borrowers. For the 2011 cohort, for instance, the median graduate-only borrower earned about $56,100 and the median borrower from a selective undergraduate institution earned about $42,300. In contrast, the median for-profit borrower who worked earned about $20,900 and the median borrower from a 2-year institution about $23,900. The median borrower from nonselective 4-year institutions earned about $29,100.

This pattern reflects two things. First, it reflects long-standing differences in earnings levels across borrowers from different institutions; even in 2000, a borrower from a 4-year selective school earned roughly 66 percent more than a borrower from a for-profit school. Second, it reflects the disproportionate blow to the labor market outcomes of borrowers from less-selective institutions, for-profit schools, and 2-year colleges. Between the 2000 and 2011 cohorts, the median earnings of borrowers declined by 24 percent among for-profit borrowers (two years after entering repayment), 23 percent among 2-year borrowers, and 14 percent among nonselective 4-year borrowers, but only 7 percent among borrowers from the most selective institutions, and 6 percent among graduate borrowers.

Nontraditional borrowers entering the labor market in 2011 therefore faced a particularly severe outlook, with almost 20 percent of them unemployed and with the earnings of those who were working down more than 20 percent relative to their peers in earlier years. Hence, while most federal borrowers in repayment on their outstanding student loans had experienced relatively little change in earnings and employment over the course of the recession (or at least, had roughly the same earnings and employment rates as their peers in earlier years), the most recent cohorts of students faced particularly unfavorable outcomes.

VI. Debt Burdens

The previous sections illustrate that there were many more nontraditional borrowers during and after the recession, that they are a particularly disadvantaged and high-risk group, and that they face relatively poor labor market outcomes when finishing school, particularly the most recent cohorts of borrowers exiting school during and after the recession. This section examines how much they owed in federal loans and their debt service burdens.

VI.A. Increases in Borrowing per Student

Nontraditional borrowers accrued relatively less debt than other borrowers. As table 6 shows, among borrowers entering repayment in 2011, the median borrower at for-profit institutions entered with about $ 10,500 in debt and the median borrower at 2-year institutions with about $9,600, compared to median borrower debts of $17,600 among those at nonselective 4-year public and private institutions and $23,000 among those at the most selective institutions. Nevertheless, the increases in per-borrower debt have been much larger among nontraditional borrowers and borrowers from nonselective schools.

While the pattern of increases in average balances is largely the same across institutions, the magnitude of the increase is much larger among nontraditional borrowers because a rising share of them accumulated substantial loan burdens. For instance, among borrowers who started at for-profit institutions, the increase in average balances was 51 percent between 2000 and 2011 (from $10,700 to $16,200). This compares during the same period with increases of 32 percent for those at 2-year institutions (from $13,000 to $17,100) and 42 percent for those at the most selective 4-year institutions (from $27,500 to $39,100). (These balances reflect the cumulative balance over a student's career, which may include starting to borrow at a 2-year institution but subsequently borrowing at a 4-year institution, or borrowing initially at a 4-year institution and later in graduate school.) Increases in loan limits appear to have accelerated the accumulation of federal debt burdens in recent years. For instance, there is a clear increase in borrowing starting after the 2007-08 academic year, when loan limits were raised by changes made under the Higher Education Reconciliation Act of 2005.

VI.B. Repayment Burdens Relative to Earnings

The combination of higher loan amounts and worsening labor market outcomes has increased the burden on borrowers. To examine this burden, we produce estimates of debt-service-to-earnings (DE) ratios. To provide consistent measures of debt service, we assume the standard 10-year repayment plan (a 10-year amortizing loan) and use the (weighted average) interest rate on each student's loans in the year of repayment to estimate the annual payments. (29)

Table 7 provides estimates of the median DE ratio (the median debt service payment divided by median earnings of employed borrowers) for borrowers two years after entering repayment, by institution type. For the cohort entering repayment in 2011, the overall DE ratio was approximately 7.1 percent, almost two percentage points above the ratio of 5.3 percent in 2000. DE ratios have edged up within all groups since 2000, from 3.9 to 6.9 percent among for-profit borrowers, from 3.3 to 5.5 percent among 2-year borrowers, and from 8.1 to 9.9 percent among graduate borrowers. Borrowers from selective institutions experienced the smallest increase of 0.9 percentage point. It is important to note that these ratios compare debt burdens to earnings, but not to ability to pay. Hence, the median annual debt service payment we calculate for the 2011 cohort of for-profit borrowers ($2,200) or 2-year borrowers ($2,300) may be a much larger share of their disposable income than the annual payments for borrowers from relatively more selective 4-year institutions ($5,400) or graduate-only borrowers ($8,700).

Borrowers from 2-year colleges have historically had the lowest DE ratios--around 3 percent prior to the 2007 cohort--although, as in other sectors, their DE ratios have risen as well. While the median DE ratio for 2-year borrowers increased from 3.3 percent for the 2006 cohort to 5.5 percent for the 2011 cohort, the ratio remained below that for other institution types. Borrowers at for-profit institutions historically had relatively low DE ratios, but saw the largest increases during the prior decade, rising by 3 percentage points from 2000 to 2011. DE ratios tend to be higher early in repayment, which is also when the majority of defaults tend to occur, and fall gradually over time as borrowers' nominal earnings rise during repayment. It is worth noting that the sectors driving the vast majority of defaults--the for-profit and two-year sectors--do not have higher DE ratios than other sectors. In fact, graduate borrowers have the highest DE ratios but the lowest default rates.

VI.C. The Distribution of Debt by Income Level

Figure 6 illustrates the relationship between the amount of debt a borrower has accrued and his or her earnings. The upper panel shows nonparametric estimates of the density of earnings. (30) The lightest line shows the earnings density for borrowers with less than $25,000 in debt, and the moderately dark line shows the earnings density for borrowers with between $25,000 and $75,000 in debt. The darkest line shows the earnings density for borrowers with a large amount of borrowing--more than $75,000, which corresponds roughly to the 95th percentile of outstanding debt. The sample is restricted to individuals between the ages of 25 and 34 and individuals with a positive loan amount. As a basis for comparison, the dashed line shows the density of all tax filers between the ages of 25 and 34. (31)

[FIGURE 6 OMITTED]

The lower panel shows average income in $10,000 borrowing bins in 2010. The figure shows that the larger the student debt balance, the more the student tends to earn. This relationship is intuitive--students with larger debts tend to have been enrolled longer, achieved higher levels of educational attainment, pursued higher levels of postsecondary education (such as a bachelor's or graduate degree instead of a certificate), and have attended 4-year institutions where borrowing amounts are greatest, which tend to be the more selective 4-year institutions. For these reasons, borrowers with more debt tend to earn much more.

Figure 6 also shows that borrowers tend to have higher earnings than non-borrowers. Individuals between the ages of 25 and 34 with no student debt earn $37,545 on average, while individuals with student debt earn $43,224. In addition, larger debt amounts are strongly correlated with higher earnings, with mean incomes of $51,555 for borrowers with more than $25,000 in debt and $40,612 for borrowers with less than $25,000 in debt. Borrowers with student debt above $75,000 on average earn slightly more than $60,000, and there are significantly more individuals with debt above $75,000 than those with less debt who earn more than $100,000. The online appendix provides additional information on debt balances and the distribution of debt by borrower incomes as well as their family income, including information regarding the characteristics of borrowers with especially large balances.
COPYRIGHT 2015 Brookings Institution
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2015 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:p. 1-45
Author:Looney, Adam; Yannelis, Constantine
Publication:Brookings Papers on Economic Activity
Article Type:Report
Date:Sep 22, 2015
Words:11233
Previous Article:Editors' introduction.
Next Article:A crisis in student loans? How changes in the characteristics of borrowers and in the institutions they attended contributed to rising loan defaults.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters