- Email: [email protected]
Student loan and credit risk in Korea
Economics Letters 135 (2015) 121–125
Contents lists available at ScienceDirect
Economics Letters journal homepage: www.elsevier.com/locate/ecolet
Student loan and credit risk in Korea✩ Byung-Suk Han, Hyoung-Goo Kang, Sang-Gyung Jun ∗ Hanyang University, Business School, South Korea
highlights • • • •
We analyze factors affecting default of student loans in Korea, using nationwide data. Default of student loans is a function of gender, major, loan amount, etc., being consistent with prior studies. The effect of age and marital status in our analysis is not consistent with prior studies. We first document new variables affecting student loan defaults such as grace period and repayment period.
article
info
Article history: Received 13 July 2015 Received in revised form 17 August 2015 Accepted 20 August 2015 Available online 28 August 2015
abstract This study first analyzes factors affecting default of student loans in Korea, using nationwide data. Default of student loans is a function of gender, major, loan balance, etc., being consistent with prior studies. However the effect of age and marital status in our analysis is not consistent with prior studies. We first document new variables affecting student loan defaults such as grace period and repayment period. © 2015 Elsevier B.V. All rights reserved.
JEL classification: G210 Keywords: Korea Student loan Default Consumer finance Credit risk
1. Introduction The delinquency rate of US student loan has exceeded 11.2% (Federal Reserve Bank of New York, 2013). The risk of student loan default is regarded as a potential factor that may possibly trigger the next economic crisis (Stiglitz, 2013). Indeed, the 2008 financial crisis started when the housing-loan delinquency rate exceeded 10% (Palacios and Kelly, 2014). In addition, while total consumer debt has increased by 61% over the last 10 years, the size of private and federal student loan has jumped 362% to 1.1 trillion USD. This reaches 10% of total consumer debt as of 2013 (Shah, 2014). Hence, student loans are a serious issue in US. Korea faces similar problems. While the annual growth rate of consumer debt was 6% in 2013, the growth rate of education
✩ The authors would like to thank an anonymous referee for helpful comments. This study was supported by the research fund of Hanyang University (HY-2014). ∗ Correspondence to: 222 Wangsimni-ro, Seongdong-gu, Seoul, 133-791, South Korea. Tel.: +82 2 2220 2883. E-mail address: [email protected] (S.-G. Jun).
http://dx.doi.org/10.1016/j.econlet.2015.08.018 0165-1765/© 2015 Elsevier B.V. All rights reserved.
debt was 12%. Importantly, 33% of the educational debt, amounting to about 10 billion USD, was borrowed by college students (The Bank of Korea, 2014); the amount of college student loan has increased by 2.5 billion USD each year from its inception in 2009. Furthermore, loans are expected to grow to 26 billion USD by the mid-2020s. In addition, the number of loan defaulters surged by 32% to 40,419 in 2012 (Korea Student Aid Foundation, 2013). Conclusively, the loan balance and defaults keep ballooning, foreshadowing significant financial and social problems in Korea. Such trends and risks partially result from various policy shifts and pressures. Education has always been an important and strategic issue in Korea. College tuition has increased considerably while the government has little room to support education costs due to budget constraints. More and more students depend on student loans for their education. The government is facing the risk of much loss from student loans in the current system. It cannot politically and socially afford large scale defaults in student loans. Hence, the growing credit risk in student loans is a complex problem for students, themselves, and for the society as a whole. We examine factors affecting student loan defaults in Korea. We extend existing literature by adding new variables about loan
122
B.-S. Han et al. / Economics Letters 135 (2015) 121–125
characteristics and academic attributes. While prior studies have the limitation of using narrow data of a certain university or region, we use a unique nationwide student loan data. The dataset of Korean student loans also provides an international comparison, giving the opportunity to extend existing literature primarily focusing on US situation.
into five categories: (1) demographic characteristics: gender, age, marriage, income, etc.; (2) academic characteristics: major, grade, student classification,2 etc.; (3) loan characteristics: loan balance, grace period, loan accounts, etc.; (4) financial-support characteristics: interest subsidies and scholarships; (5) institution characteristics: type of institution, province of institution, etc. This categorization extends Flint (1997).
2. Literature review 4. Results Existing literature on student-loan default focuses on the USA where the student loan is an important issue for higher education (Gross et al., 2009). Since the 1980s, the USA has tilted to using student loans from using scholarships in order to expand the opportunities for higher education. Since then, academic studies have examined several key aspects of student loans. There are several consensuses in the literature. Men are more likely to default than women (Podgursky et al., 2002). The older the borrower, the higher the probability of default (Herr and Burt, 2005). Studentloan default increases with the number of family members to take care of and being divorced or widowed (Volkwein and Szelest, 1995). The higher the income, the lower the default (Wilms et al., 1987). The lower the grade, the higher the default (Christman, 2000). Students majoring in humanities and social science experience higher default mainly attributable to their lower future earnings (Schwartz and Finnie, 2002). Debt balance increases default rate (Choy and Li, 2006). Our empirical analysis will include those traditional variables and examine whether such consensus holds. In addition, we will examine how loan characteristics affect default that existing literature has disregarded. Our analysis is also the first about the Korean student loan. There are also debates. In Flint (1997), scholarships do not affect default; however in Greene (1989), the more students receives scholarships, the lower the default rate; in Steiner and Teszler (2005), financial support tends to yield a lower default rate. In Woo (2002), two-year community college students default more than four-year university students do, which Knapp and Seaks (1992) disagree with. Such controversial findings in existing studies may be attributable to their data limitation because existing studies use limited data specific to certain regions and universities. We fill this gap in the literature using nationwide data without any sample selection bias. 3. Empirical specification and data We use the Korean Direct Loan data of 1 161 478 accounts between the second-semester of 2009 and the second-semester of 2012. The Korean Direct Loan means mortgage style loans where monthly repayments are fixed for a set period.1 All data are from KOSAF (Korea Student Aid Foundation), a government agency established to provide student loans. Since default from an account eventually affects all accounts of its borrower, our unit of analysis is the borrower. Thus, consolidating the accounts into borrowers, we have 626 575 students. We use the logit model including almost all variables from prior literature, while excluding some variables irrelevant to Korea, such as race. Our dependent variable is default. Default occurs when a borrower is overdue more than six months. This is the practical procedure and definition used by KOSAF and the Korean government. We examine an extensive list of independent variables including loan characteristics. The independent variables are grouped
1 In Korea, two types of loan programs are offered; (1) Direct Loan, (2) Income Contingent Loan, and if student applicants are from the upper 20% income level, they should apply for Direct Loan.
Descriptive statistics are presented in Table A.1 of Appendix. Out of a total of 626 757 students, the number of defaults amounts to 35 274, implying that the default rate is around 5.63%. The major findings, as presented in Table 1, are as follows. About demographic characteristics, men are more likely to default. However, women tend to apply for loans more than men do, contrasting with Mortenson (1989) in that women use less debt due to their risk aversion. Possibly, male students can access family wealth or parttime blue-color working opportunities better. The younger the borrower, the higher the default. This result also contrasts with the finding of the previous studies (Herr and Burt, 2005). This difference may come from the unique feature of the Korean student loan system, i.e. (1) the qualifying age of a Direct Loan borrower increased after the inception of an alternative, income-contingent loan (ICL) in 2010; (2) students from high-income families apply for a Direct Loan, because they are not allowed to apply for an ICL. Consequently, Direct Loan holders are comprised of older students from high income level families. Therefore, the negative relationship between age and default can be found. The married are less likely to default than the single. This also differs from previous studies (Volkwein and Szelest, 1995). Our interpretation is that, in Korea today, being married when young is a luxury given the cost of marriage ceremony, marriage infrastructure (e.g. housing and home appliances) and household management. Those who can afford such luxury will default less. As for income, the higher the income, the less likely the default. Academic characteristics of students affect the default rate of student loan. Students majoring in humanities, art and physical education have a higher default rate than other majors, everything else being equal. These academic fields may imply lower employment rate and wage, increasing default risks. Interestingly, students in higher year of college default more. This result implies the importance of job prospect in affecting defaults of student loans. If a student stays longer in school, he delays entering the job market. The bleaker the job market prospect, the longer the postponement in entering the job market. Furthermore, staying too long in school and not entering the job market on time can deteriorate the quality of human capital. The higher is the student’s scholastic grade, the lower the default. Defaults occur more for freshmen, readmission students and transfer students. If a freshman fails to meet the minimum criteria for income-contingent loan (ICL), she can only apply for Direct Loan. Readmission students tend to restart their study after they fail in the previous programs due to low grades. Thus, our results suggest that academic record can signal the characteristics as diligence, academic capability and credit risk. Transfer students default more than others; this conflicts Woo (2002) in which the more schools a student attends, the lower the default. As for loan characteristics about which we are the first to document in the literature, the longer grace period and repayment period, the less likely the default. Thus, rescheduling debts maturity and grace period can increase the debt capacity of
2 Student classification variable includes being Readmission student or Transfer student, etc.
B.-S. Han et al. / Economics Letters 135 (2015) 121–125
123
Table 1 Logit regression result. This table shows the results from the Logit model. The dependent variable takes one in case of default, and zero otherwise. Default is defined as the case where a borrower is overdue more than six months. This is the practical procedure and definition used by KOSAF and the Korean government. The independent variables are grouped into five categories: (1) demographic characteristics: gender, age, marriage, income, etc.; (2) academic characteristics: major, grade, student classification, etc.; (3) loan characteristics: loan balance, maturity, grace period, loan accounts, etc.; (4) financial-support characteristics: interest subsidies and scholarships; (5) institution characteristics: type of institution, province of institution, etc. Significance levels 0.1, 0.05, and 0.01 are denoted by *, **, and *** respectively. Group
Demographic characteristics
Variable
Reference
Dummy
Gender Age
Women –
Marriage
Single
Income
–
Men – Married Widowed Divorced –
Major
Social science
Scholastic grade Year of college
– –
Student classification
Enrolled
Loan balance(LN) Grace period Repayment period Loan accounts
– – – –
Academic characteristics
Loan characteristics
Financial-support characteristics
Institution characteristics
Intercept AIC SC −2 Log L R-square
– – – –
Interest subsidy
No subsidy
Scholarship
Not received
Status
National
Province
Seoul
Type
4 year
–
– 255 857 256 390 255 763 0.025
students significantly. This is an important policy implication. Next, the more the debt balance, the more the default. The more scholarship, the less default. These are intuitive results. We have interesting findings on the effect of financial supports designed to reduce the burden of borrowing students. We find that interest subsidies have different effects on default depending on their types. There are three types of interest subsidy programs: (1) Full-interest subsidy for students whose family income belongs to the lowest 20 percentile, (2) Type I subsidy entitling 4% point reduction of interest rate for students whose family income is in between the lowest 30 and 50 percentile, and (3) Type II subsidy for 1.5% point reduction of interest rate for students whose family income is in between 60 and 70 percentile. Students with a full-interest subsidy or a Type I subsidy default less than those without interest subsidies. However, students
Engineering Education Art & physical education Medical science & pharmacy Humanities Natural science – – Freshmen Readmission Transfer
Coefficient 0.133
−0.005 −0.170 0.162 0.076 −0.080
−0.161 0.030 0.313 −0.475 0.108 −0.121 −0.017 0.024 0.543 0.723 0.430 0.105
−0.002 −0.003 0.043
SE 0.013∗∗∗ 0.002∗∗ 0.049∗∗∗ 0.248 0.097 0.002∗∗∗ 0.017∗∗∗ 0.036 0.017∗∗∗ 0.034∗∗∗ 0.020∗∗∗ 0.020∗∗∗ 0.000∗∗∗ 0.008∗∗∗ 0.019∗∗∗ 0.067∗∗∗ 0.059∗∗∗ 0.005∗∗∗ 0.000∗∗∗ 0.000∗∗∗ 0.006∗∗∗
Full subsidy Type I subsidy Type II subsidy Received
−0.460 −0.064 −1.019
0.021∗∗∗ 0.020∗∗∗ 0.020∗∗∗ 0.031∗∗∗
Municipal Private Gangwon Gyunggi Gyungnam Gyungbuk Gwangju Daegu Daejeon Busan Sejong Ulsan Inchon Jeonnam Jeonbuk Jeju Chungnam Chungbuk 2 year 3 year 5 year
0.323 0.210 0.564 0.135 0.483 0.421 0.548 0.176 0.448 0.344 0.145 0.086 0.037 0.575 0.520 0.679 0.366 0.518 0.654 0.450 −0.270
0.062∗∗∗ 0.023∗∗∗ 0.030∗∗∗ 0.020∗∗∗ 0.032∗∗∗ 0.025∗∗∗ 0.033∗∗∗ 0.032∗∗∗ 0.029∗∗∗ 0.024∗∗∗ 0.059∗∗ 0.069 0.038 0.038∗∗∗ 0.030∗∗∗ 0.058∗∗∗ 0.025∗∗∗ 0.030∗∗∗ 0.016∗∗∗ 0.022∗∗∗ 0.070∗∗∗
–
−2.843
0.097∗∗∗
Somers’ D Gamma Tau-a C
0.382 0.382 0.041 0.691
0.178
who receive a Type II subsidy default more than those without subsidies. The sticker rate of Direct Loan is 3.9% in 2012, which implies that the students with Type I subsidies practically pay no interest. Therefore, the result shows that financial subsidies for the-lowest-income students are not easy to yield desirable effect of reducing loan defaults: Only financial subsidies ensuring zero interest could reduce default, ceteris paribus. For institution characteristics, we consider college type, province and status. The USA and Korean governments penalize colleges with high default rate. For example, such institutions cannot participate in government student aid programs. Therefore, institution variables have considerable effect on the default rate of student loans, as institutions have incentives to be concerned about the credit risks. We have found that students from private
124
B.-S. Han et al. / Economics Letters 135 (2015) 121–125
Table A.1 Variables
Total applicants
Demographic characteristics Gender Men Women Total Age Marriage Divorced Widowed Single Married Total Income Academic characteristics Major Art & Physical education Social science Natural science Humanities Engineering Education Medical science & Pharmacy Total Scholastic grade Year of college Student Freshmen classification Readmission Transfer Enrolled Total Loan characteristics Loan balance Graced period Repayment period Loan accounts
Defaulter
Number
%
309 704 316 871 626 575
49.43 50.57 100
Mean
23.83 2 127 280 605 497 18 671 626 575
SD
%
18 188 17 086 35 274
5.87 5.39 5.63
151 19 34 135 969 35 274
7.10 6.79 5.64 5.19 5.63
4.49
0.34 0.04 96.64 2.98 100 5.71
Number
3.12
91 431
14.59
7 187
7.86
150 483 86 721 83 943 161 966 20 637 31 394
24.02 13.84 13.40 25.85 3.29 5.01
8 217 4 676 4 524 8 427 959 1 284
5.46 5.39 5.39 5.20 4.65 4.09
626 575
100
35 274
5.63
84.10 2.62
10.83 1.18
49 332
7.87
5 155
10.45
2 682 5 627 568 934 626 575
0.43 0.90 90.80 100
263 315 29 541 35 274
9.81 5.60 5.19 5.63
USD10 Months Months Number
565.31 43.26 105.06 1.86
416.91 26.82 60.14 1.18
Financial support characteristics Interest subsidy No subsidy Type II subsidy Type I subsidy Full subsidy Total Scholarship Not received Received Total
345 699 61 370 81 947 137 559 626 575 560 546 66 029 626 575
55.17 9.79 13.08 21.95 100 89.46 10.54 100
21 685 3 639 4 300 5 650 35 274 34 003 1 271 35 274
6.27 5.93 5.25 4.11 5.63 6.07 1.92 0.00
Institution characteristics Status Municipal Private National Total Province Jeju Gangwon Sejong Jeonnam Gyungnam Chungnam Chungbuk Gyungbuk Ulsan Jeonbuk Daejeon Inchon Busan Gyunggi Kwangju Daegu Seoul Total Type 2 year 3 year 4 year 5 year Total
4 754 551 703 70 118 626 575 4 353 14 095 20 250 27 058 21 990 25 659 25 972 24 874 44 301 55 041 25 399 132 276 51 854 6 295 19 220 5 104 122 834 626 575 129 220 55 065 433 692 8 598 626 575
0.76 88.05 11.19 100 0.69 2.25 3.23 4.32 3.51 4.10 4.15 3.97 7.07 8.78 4.05 21.11 8.28 1.00 3.07 0.81 19.60 100 20.62 8.79 69.22 1.37 100
342 32 307 2 625 35 274 384 979 1 353 1 781 1 447 1 688 1 705 1 614 2 693 3 325 1 516 7 675 2 859 334 900 235 4 786 35 274 12 417 3 485 19 152 220 35 274
7.19 5.86 3.74 5.63 8.82 6.95 6.68 6.58 6.58 6.58 6.56 6.49 6.08 6.04 5.97 5.80 5.51 5.31 4.68 4.60 3.90 5.63 9.61 6.33 4.42 2.56 5.63
Mean
SD
. 23.40
4.41
. 5.39
3.08
. 80.27 2.29
13.38 1.14
526.12 39.44 93.52 1.73
365.83 27.29 59.77 0.97
.
.
.
B.-S. Han et al. / Economics Letters 135 (2015) 121–125
and municipal institutions default more than those from national institutions. The higher tuition of private schools3 and student quality can produce this result. Students attending universities in Seoul default less than those from other regions. Student quality and job market prospect in Seoul can explain this result. Students from two-year and three-year colleges default more than those from four-year universities, while five-year university students default less than all other groups. Interestingly, the employment rate for two-year or three-year college graduates is higher, but their average wage is lower than four-year or five-year university graduates. Therefore, wage affects credit risk more than does employment. Such results suggest a policy implication: it makes sense to categorize institutions and to offer them customized incentives in dealing with their student credit risks. 5. Conclusion Existing studies on student loan have used only data from a certain university or region. Firstly, using nationwide unique data, we analyze credit risks in student loans. We also examine new variables such as loan characteristics to extend prior literature. Some of our results are different from the literature. In addition, this study is the first analysis on the Korean student loan system. Our study addresses important topics given the explosion and risk implied in student loans. This study has a limitation of excluding data of incomecontingent loan growing fast in Korea and in many other countries like UK, Australia, etc. This subject is the area that should be examined in the subsequent studies. Appendix. Descriptive statistics See Table A.1.
3 As of 2013, average tuition of private universities is about 1.8 times higher than national universities.
125
References Choy, S.P., Li, X., 2006. Dealing with debt: 1992–93 bachelor’s degree recipients 10 years later. Postsecondary education descriptive analysis Report. NCES 2006156. National Center for Education Statistics. Christman, D.E., 2000. Multiple realities: Characteristics of loan defaulters at a twoyear public institution. Community Coll. Rev. 27, 16–32. Federal Reserve Bank of New York. 2013. Quarterly report on household debt and credit. Available at: http://www.newyorkfed.org (accessed 15.06.15). Flint, T.A., 1997. Predicting student loan defaults. J. High. Educ. 322–354. Greene, L.L., 1989. An economic analysis of student loan default. Educ. Eval. Policy Anal. 11, 61–68. Gross, J.P., Cekic, O., Hossler, D., Hillman, N., 2009. What matters in student loan default: A review of the research literature. J. Stud. Financ. Aid 39, 19–29. Herr, E., Burt, L., 2005. Predicting student loan default for the University of Texas at Austin. J. Stud. Financ. Aid 35, 2. Knapp, L.G., Seaks, T.G., 1992. An analysis of the probability of default on federally guaranteed student loans. Rev. Econ. Stat. 404–411. Korea Student Aid Foundation, 2013. KOSAF Statistical Yearbook 2012. Mortenson, T.G., 1989. Attitudes toward educational loans. J. Stud. Financ. Aid 19, 4. Palacios, M., Kelly, A.P., 2014. A better way to finance that college degree. Available at: http://www.wsj.com (accessed 15.06.15). Podgursky, M., Ehlert, M., Monroe, R., Watson, D., Wittstruck, J., 2002. Student loan defaults and enrollment persistence. J. Stud. Financ. Aid 32, 27–42. Schwartz, S., Finnie, R., 2002. Student loans in Canada: An analysis of borrowing and repayment. Econ. Educ. Rev. 21, 497–512. Shah, N., 2014. US household debt increases. Available at: http://www.wsj.com (accessed 15.06.15). Steiner, M., Teszler, N., 2005. Multivariate Analysis of Student Loan Defaulters at Texas A&M University, TG (Texas Guaranteed Student Loan Corporation). Stiglitz, J.E., 2013. Student debt and the crushing of the American dream. Available at: http://www.nytimes.com (accessed 15.06.15). The Bank of Korea. 2014. Financial Stability Report 2014. Volkwein, J.F., Szelest, B.P., 1995. Individual and campus characteristics associated with student loan default. Res. High. Educ. 36, 41–72. Wilms, W.W., Moore, R.W., Bolus, R.E., 1987. Whose fault is default? A study of the impact of student characteristics and institutional practices on guaranteed student loan default rates in California. Educ. Eval. Policy Anal. 9, 41–54. Woo, J.M.H., 2002. Clearing accounts: The causes of student loan default, EdFund.
Contents lists available at ScienceDirect
Economics Letters journal homepage: www.elsevier.com/locate/ecolet
Student loan and credit risk in Korea✩ Byung-Suk Han, Hyoung-Goo Kang, Sang-Gyung Jun ∗ Hanyang University, Business School, South Korea
highlights • • • •
We analyze factors affecting default of student loans in Korea, using nationwide data. Default of student loans is a function of gender, major, loan amount, etc., being consistent with prior studies. The effect of age and marital status in our analysis is not consistent with prior studies. We first document new variables affecting student loan defaults such as grace period and repayment period.
article
info
Article history: Received 13 July 2015 Received in revised form 17 August 2015 Accepted 20 August 2015 Available online 28 August 2015
abstract This study first analyzes factors affecting default of student loans in Korea, using nationwide data. Default of student loans is a function of gender, major, loan balance, etc., being consistent with prior studies. However the effect of age and marital status in our analysis is not consistent with prior studies. We first document new variables affecting student loan defaults such as grace period and repayment period. © 2015 Elsevier B.V. All rights reserved.
JEL classification: G210 Keywords: Korea Student loan Default Consumer finance Credit risk
1. Introduction The delinquency rate of US student loan has exceeded 11.2% (Federal Reserve Bank of New York, 2013). The risk of student loan default is regarded as a potential factor that may possibly trigger the next economic crisis (Stiglitz, 2013). Indeed, the 2008 financial crisis started when the housing-loan delinquency rate exceeded 10% (Palacios and Kelly, 2014). In addition, while total consumer debt has increased by 61% over the last 10 years, the size of private and federal student loan has jumped 362% to 1.1 trillion USD. This reaches 10% of total consumer debt as of 2013 (Shah, 2014). Hence, student loans are a serious issue in US. Korea faces similar problems. While the annual growth rate of consumer debt was 6% in 2013, the growth rate of education
✩ The authors would like to thank an anonymous referee for helpful comments. This study was supported by the research fund of Hanyang University (HY-2014). ∗ Correspondence to: 222 Wangsimni-ro, Seongdong-gu, Seoul, 133-791, South Korea. Tel.: +82 2 2220 2883. E-mail address: [email protected] (S.-G. Jun).
http://dx.doi.org/10.1016/j.econlet.2015.08.018 0165-1765/© 2015 Elsevier B.V. All rights reserved.
debt was 12%. Importantly, 33% of the educational debt, amounting to about 10 billion USD, was borrowed by college students (The Bank of Korea, 2014); the amount of college student loan has increased by 2.5 billion USD each year from its inception in 2009. Furthermore, loans are expected to grow to 26 billion USD by the mid-2020s. In addition, the number of loan defaulters surged by 32% to 40,419 in 2012 (Korea Student Aid Foundation, 2013). Conclusively, the loan balance and defaults keep ballooning, foreshadowing significant financial and social problems in Korea. Such trends and risks partially result from various policy shifts and pressures. Education has always been an important and strategic issue in Korea. College tuition has increased considerably while the government has little room to support education costs due to budget constraints. More and more students depend on student loans for their education. The government is facing the risk of much loss from student loans in the current system. It cannot politically and socially afford large scale defaults in student loans. Hence, the growing credit risk in student loans is a complex problem for students, themselves, and for the society as a whole. We examine factors affecting student loan defaults in Korea. We extend existing literature by adding new variables about loan
122
B.-S. Han et al. / Economics Letters 135 (2015) 121–125
characteristics and academic attributes. While prior studies have the limitation of using narrow data of a certain university or region, we use a unique nationwide student loan data. The dataset of Korean student loans also provides an international comparison, giving the opportunity to extend existing literature primarily focusing on US situation.
into five categories: (1) demographic characteristics: gender, age, marriage, income, etc.; (2) academic characteristics: major, grade, student classification,2 etc.; (3) loan characteristics: loan balance, grace period, loan accounts, etc.; (4) financial-support characteristics: interest subsidies and scholarships; (5) institution characteristics: type of institution, province of institution, etc. This categorization extends Flint (1997).
2. Literature review 4. Results Existing literature on student-loan default focuses on the USA where the student loan is an important issue for higher education (Gross et al., 2009). Since the 1980s, the USA has tilted to using student loans from using scholarships in order to expand the opportunities for higher education. Since then, academic studies have examined several key aspects of student loans. There are several consensuses in the literature. Men are more likely to default than women (Podgursky et al., 2002). The older the borrower, the higher the probability of default (Herr and Burt, 2005). Studentloan default increases with the number of family members to take care of and being divorced or widowed (Volkwein and Szelest, 1995). The higher the income, the lower the default (Wilms et al., 1987). The lower the grade, the higher the default (Christman, 2000). Students majoring in humanities and social science experience higher default mainly attributable to their lower future earnings (Schwartz and Finnie, 2002). Debt balance increases default rate (Choy and Li, 2006). Our empirical analysis will include those traditional variables and examine whether such consensus holds. In addition, we will examine how loan characteristics affect default that existing literature has disregarded. Our analysis is also the first about the Korean student loan. There are also debates. In Flint (1997), scholarships do not affect default; however in Greene (1989), the more students receives scholarships, the lower the default rate; in Steiner and Teszler (2005), financial support tends to yield a lower default rate. In Woo (2002), two-year community college students default more than four-year university students do, which Knapp and Seaks (1992) disagree with. Such controversial findings in existing studies may be attributable to their data limitation because existing studies use limited data specific to certain regions and universities. We fill this gap in the literature using nationwide data without any sample selection bias. 3. Empirical specification and data We use the Korean Direct Loan data of 1 161 478 accounts between the second-semester of 2009 and the second-semester of 2012. The Korean Direct Loan means mortgage style loans where monthly repayments are fixed for a set period.1 All data are from KOSAF (Korea Student Aid Foundation), a government agency established to provide student loans. Since default from an account eventually affects all accounts of its borrower, our unit of analysis is the borrower. Thus, consolidating the accounts into borrowers, we have 626 575 students. We use the logit model including almost all variables from prior literature, while excluding some variables irrelevant to Korea, such as race. Our dependent variable is default. Default occurs when a borrower is overdue more than six months. This is the practical procedure and definition used by KOSAF and the Korean government. We examine an extensive list of independent variables including loan characteristics. The independent variables are grouped
1 In Korea, two types of loan programs are offered; (1) Direct Loan, (2) Income Contingent Loan, and if student applicants are from the upper 20% income level, they should apply for Direct Loan.
Descriptive statistics are presented in Table A.1 of Appendix. Out of a total of 626 757 students, the number of defaults amounts to 35 274, implying that the default rate is around 5.63%. The major findings, as presented in Table 1, are as follows. About demographic characteristics, men are more likely to default. However, women tend to apply for loans more than men do, contrasting with Mortenson (1989) in that women use less debt due to their risk aversion. Possibly, male students can access family wealth or parttime blue-color working opportunities better. The younger the borrower, the higher the default. This result also contrasts with the finding of the previous studies (Herr and Burt, 2005). This difference may come from the unique feature of the Korean student loan system, i.e. (1) the qualifying age of a Direct Loan borrower increased after the inception of an alternative, income-contingent loan (ICL) in 2010; (2) students from high-income families apply for a Direct Loan, because they are not allowed to apply for an ICL. Consequently, Direct Loan holders are comprised of older students from high income level families. Therefore, the negative relationship between age and default can be found. The married are less likely to default than the single. This also differs from previous studies (Volkwein and Szelest, 1995). Our interpretation is that, in Korea today, being married when young is a luxury given the cost of marriage ceremony, marriage infrastructure (e.g. housing and home appliances) and household management. Those who can afford such luxury will default less. As for income, the higher the income, the less likely the default. Academic characteristics of students affect the default rate of student loan. Students majoring in humanities, art and physical education have a higher default rate than other majors, everything else being equal. These academic fields may imply lower employment rate and wage, increasing default risks. Interestingly, students in higher year of college default more. This result implies the importance of job prospect in affecting defaults of student loans. If a student stays longer in school, he delays entering the job market. The bleaker the job market prospect, the longer the postponement in entering the job market. Furthermore, staying too long in school and not entering the job market on time can deteriorate the quality of human capital. The higher is the student’s scholastic grade, the lower the default. Defaults occur more for freshmen, readmission students and transfer students. If a freshman fails to meet the minimum criteria for income-contingent loan (ICL), she can only apply for Direct Loan. Readmission students tend to restart their study after they fail in the previous programs due to low grades. Thus, our results suggest that academic record can signal the characteristics as diligence, academic capability and credit risk. Transfer students default more than others; this conflicts Woo (2002) in which the more schools a student attends, the lower the default. As for loan characteristics about which we are the first to document in the literature, the longer grace period and repayment period, the less likely the default. Thus, rescheduling debts maturity and grace period can increase the debt capacity of
2 Student classification variable includes being Readmission student or Transfer student, etc.
B.-S. Han et al. / Economics Letters 135 (2015) 121–125
123
Table 1 Logit regression result. This table shows the results from the Logit model. The dependent variable takes one in case of default, and zero otherwise. Default is defined as the case where a borrower is overdue more than six months. This is the practical procedure and definition used by KOSAF and the Korean government. The independent variables are grouped into five categories: (1) demographic characteristics: gender, age, marriage, income, etc.; (2) academic characteristics: major, grade, student classification, etc.; (3) loan characteristics: loan balance, maturity, grace period, loan accounts, etc.; (4) financial-support characteristics: interest subsidies and scholarships; (5) institution characteristics: type of institution, province of institution, etc. Significance levels 0.1, 0.05, and 0.01 are denoted by *, **, and *** respectively. Group
Demographic characteristics
Variable
Reference
Dummy
Gender Age
Women –
Marriage
Single
Income
–
Men – Married Widowed Divorced –
Major
Social science
Scholastic grade Year of college
– –
Student classification
Enrolled
Loan balance(LN) Grace period Repayment period Loan accounts
– – – –
Academic characteristics
Loan characteristics
Financial-support characteristics
Institution characteristics
Intercept AIC SC −2 Log L R-square
– – – –
Interest subsidy
No subsidy
Scholarship
Not received
Status
National
Province
Seoul
Type
4 year
–
– 255 857 256 390 255 763 0.025
students significantly. This is an important policy implication. Next, the more the debt balance, the more the default. The more scholarship, the less default. These are intuitive results. We have interesting findings on the effect of financial supports designed to reduce the burden of borrowing students. We find that interest subsidies have different effects on default depending on their types. There are three types of interest subsidy programs: (1) Full-interest subsidy for students whose family income belongs to the lowest 20 percentile, (2) Type I subsidy entitling 4% point reduction of interest rate for students whose family income is in between the lowest 30 and 50 percentile, and (3) Type II subsidy for 1.5% point reduction of interest rate for students whose family income is in between 60 and 70 percentile. Students with a full-interest subsidy or a Type I subsidy default less than those without interest subsidies. However, students
Engineering Education Art & physical education Medical science & pharmacy Humanities Natural science – – Freshmen Readmission Transfer
Coefficient 0.133
−0.005 −0.170 0.162 0.076 −0.080
−0.161 0.030 0.313 −0.475 0.108 −0.121 −0.017 0.024 0.543 0.723 0.430 0.105
−0.002 −0.003 0.043
SE 0.013∗∗∗ 0.002∗∗ 0.049∗∗∗ 0.248 0.097 0.002∗∗∗ 0.017∗∗∗ 0.036 0.017∗∗∗ 0.034∗∗∗ 0.020∗∗∗ 0.020∗∗∗ 0.000∗∗∗ 0.008∗∗∗ 0.019∗∗∗ 0.067∗∗∗ 0.059∗∗∗ 0.005∗∗∗ 0.000∗∗∗ 0.000∗∗∗ 0.006∗∗∗
Full subsidy Type I subsidy Type II subsidy Received
−0.460 −0.064 −1.019
0.021∗∗∗ 0.020∗∗∗ 0.020∗∗∗ 0.031∗∗∗
Municipal Private Gangwon Gyunggi Gyungnam Gyungbuk Gwangju Daegu Daejeon Busan Sejong Ulsan Inchon Jeonnam Jeonbuk Jeju Chungnam Chungbuk 2 year 3 year 5 year
0.323 0.210 0.564 0.135 0.483 0.421 0.548 0.176 0.448 0.344 0.145 0.086 0.037 0.575 0.520 0.679 0.366 0.518 0.654 0.450 −0.270
0.062∗∗∗ 0.023∗∗∗ 0.030∗∗∗ 0.020∗∗∗ 0.032∗∗∗ 0.025∗∗∗ 0.033∗∗∗ 0.032∗∗∗ 0.029∗∗∗ 0.024∗∗∗ 0.059∗∗ 0.069 0.038 0.038∗∗∗ 0.030∗∗∗ 0.058∗∗∗ 0.025∗∗∗ 0.030∗∗∗ 0.016∗∗∗ 0.022∗∗∗ 0.070∗∗∗
–
−2.843
0.097∗∗∗
Somers’ D Gamma Tau-a C
0.382 0.382 0.041 0.691
0.178
who receive a Type II subsidy default more than those without subsidies. The sticker rate of Direct Loan is 3.9% in 2012, which implies that the students with Type I subsidies practically pay no interest. Therefore, the result shows that financial subsidies for the-lowest-income students are not easy to yield desirable effect of reducing loan defaults: Only financial subsidies ensuring zero interest could reduce default, ceteris paribus. For institution characteristics, we consider college type, province and status. The USA and Korean governments penalize colleges with high default rate. For example, such institutions cannot participate in government student aid programs. Therefore, institution variables have considerable effect on the default rate of student loans, as institutions have incentives to be concerned about the credit risks. We have found that students from private
124
B.-S. Han et al. / Economics Letters 135 (2015) 121–125
Table A.1 Variables
Total applicants
Demographic characteristics Gender Men Women Total Age Marriage Divorced Widowed Single Married Total Income Academic characteristics Major Art & Physical education Social science Natural science Humanities Engineering Education Medical science & Pharmacy Total Scholastic grade Year of college Student Freshmen classification Readmission Transfer Enrolled Total Loan characteristics Loan balance Graced period Repayment period Loan accounts
Defaulter
Number
%
309 704 316 871 626 575
49.43 50.57 100
Mean
23.83 2 127 280 605 497 18 671 626 575
SD
%
18 188 17 086 35 274
5.87 5.39 5.63
151 19 34 135 969 35 274
7.10 6.79 5.64 5.19 5.63
4.49
0.34 0.04 96.64 2.98 100 5.71
Number
3.12
91 431
14.59
7 187
7.86
150 483 86 721 83 943 161 966 20 637 31 394
24.02 13.84 13.40 25.85 3.29 5.01
8 217 4 676 4 524 8 427 959 1 284
5.46 5.39 5.39 5.20 4.65 4.09
626 575
100
35 274
5.63
84.10 2.62
10.83 1.18
49 332
7.87
5 155
10.45
2 682 5 627 568 934 626 575
0.43 0.90 90.80 100
263 315 29 541 35 274
9.81 5.60 5.19 5.63
USD10 Months Months Number
565.31 43.26 105.06 1.86
416.91 26.82 60.14 1.18
Financial support characteristics Interest subsidy No subsidy Type II subsidy Type I subsidy Full subsidy Total Scholarship Not received Received Total
345 699 61 370 81 947 137 559 626 575 560 546 66 029 626 575
55.17 9.79 13.08 21.95 100 89.46 10.54 100
21 685 3 639 4 300 5 650 35 274 34 003 1 271 35 274
6.27 5.93 5.25 4.11 5.63 6.07 1.92 0.00
Institution characteristics Status Municipal Private National Total Province Jeju Gangwon Sejong Jeonnam Gyungnam Chungnam Chungbuk Gyungbuk Ulsan Jeonbuk Daejeon Inchon Busan Gyunggi Kwangju Daegu Seoul Total Type 2 year 3 year 4 year 5 year Total
4 754 551 703 70 118 626 575 4 353 14 095 20 250 27 058 21 990 25 659 25 972 24 874 44 301 55 041 25 399 132 276 51 854 6 295 19 220 5 104 122 834 626 575 129 220 55 065 433 692 8 598 626 575
0.76 88.05 11.19 100 0.69 2.25 3.23 4.32 3.51 4.10 4.15 3.97 7.07 8.78 4.05 21.11 8.28 1.00 3.07 0.81 19.60 100 20.62 8.79 69.22 1.37 100
342 32 307 2 625 35 274 384 979 1 353 1 781 1 447 1 688 1 705 1 614 2 693 3 325 1 516 7 675 2 859 334 900 235 4 786 35 274 12 417 3 485 19 152 220 35 274
7.19 5.86 3.74 5.63 8.82 6.95 6.68 6.58 6.58 6.58 6.56 6.49 6.08 6.04 5.97 5.80 5.51 5.31 4.68 4.60 3.90 5.63 9.61 6.33 4.42 2.56 5.63
Mean
SD
. 23.40
4.41
. 5.39
3.08
. 80.27 2.29
13.38 1.14
526.12 39.44 93.52 1.73
365.83 27.29 59.77 0.97
.
.
.
B.-S. Han et al. / Economics Letters 135 (2015) 121–125
and municipal institutions default more than those from national institutions. The higher tuition of private schools3 and student quality can produce this result. Students attending universities in Seoul default less than those from other regions. Student quality and job market prospect in Seoul can explain this result. Students from two-year and three-year colleges default more than those from four-year universities, while five-year university students default less than all other groups. Interestingly, the employment rate for two-year or three-year college graduates is higher, but their average wage is lower than four-year or five-year university graduates. Therefore, wage affects credit risk more than does employment. Such results suggest a policy implication: it makes sense to categorize institutions and to offer them customized incentives in dealing with their student credit risks. 5. Conclusion Existing studies on student loan have used only data from a certain university or region. Firstly, using nationwide unique data, we analyze credit risks in student loans. We also examine new variables such as loan characteristics to extend prior literature. Some of our results are different from the literature. In addition, this study is the first analysis on the Korean student loan system. Our study addresses important topics given the explosion and risk implied in student loans. This study has a limitation of excluding data of incomecontingent loan growing fast in Korea and in many other countries like UK, Australia, etc. This subject is the area that should be examined in the subsequent studies. Appendix. Descriptive statistics See Table A.1.
3 As of 2013, average tuition of private universities is about 1.8 times higher than national universities.
125
References Choy, S.P., Li, X., 2006. Dealing with debt: 1992–93 bachelor’s degree recipients 10 years later. Postsecondary education descriptive analysis Report. NCES 2006156. National Center for Education Statistics. Christman, D.E., 2000. Multiple realities: Characteristics of loan defaulters at a twoyear public institution. Community Coll. Rev. 27, 16–32. Federal Reserve Bank of New York. 2013. Quarterly report on household debt and credit. Available at: http://www.newyorkfed.org (accessed 15.06.15). Flint, T.A., 1997. Predicting student loan defaults. J. High. Educ. 322–354. Greene, L.L., 1989. An economic analysis of student loan default. Educ. Eval. Policy Anal. 11, 61–68. Gross, J.P., Cekic, O., Hossler, D., Hillman, N., 2009. What matters in student loan default: A review of the research literature. J. Stud. Financ. Aid 39, 19–29. Herr, E., Burt, L., 2005. Predicting student loan default for the University of Texas at Austin. J. Stud. Financ. Aid 35, 2. Knapp, L.G., Seaks, T.G., 1992. An analysis of the probability of default on federally guaranteed student loans. Rev. Econ. Stat. 404–411. Korea Student Aid Foundation, 2013. KOSAF Statistical Yearbook 2012. Mortenson, T.G., 1989. Attitudes toward educational loans. J. Stud. Financ. Aid 19, 4. Palacios, M., Kelly, A.P., 2014. A better way to finance that college degree. Available at: http://www.wsj.com (accessed 15.06.15). Podgursky, M., Ehlert, M., Monroe, R., Watson, D., Wittstruck, J., 2002. Student loan defaults and enrollment persistence. J. Stud. Financ. Aid 32, 27–42. Schwartz, S., Finnie, R., 2002. Student loans in Canada: An analysis of borrowing and repayment. Econ. Educ. Rev. 21, 497–512. Shah, N., 2014. US household debt increases. Available at: http://www.wsj.com (accessed 15.06.15). Steiner, M., Teszler, N., 2005. Multivariate Analysis of Student Loan Defaulters at Texas A&M University, TG (Texas Guaranteed Student Loan Corporation). Stiglitz, J.E., 2013. Student debt and the crushing of the American dream. Available at: http://www.nytimes.com (accessed 15.06.15). The Bank of Korea. 2014. Financial Stability Report 2014. Volkwein, J.F., Szelest, B.P., 1995. Individual and campus characteristics associated with student loan default. Res. High. Educ. 36, 41–72. Wilms, W.W., Moore, R.W., Bolus, R.E., 1987. Whose fault is default? A study of the impact of student characteristics and institutional practices on guaranteed student loan default rates in California. Educ. Eval. Policy Anal. 9, 41–54. Woo, J.M.H., 2002. Clearing accounts: The causes of student loan default, EdFund.