Reorganization versus Liquidation: Default Incentives for Student Loans * Anamaria Felicia Ionescu † April 2005, Revisited June 2007 Department of Economics, Colgate University , 13 Oak Dr, Hamilton, NY, 13346, USA. Abstract I consider the implications of a change in the bankruptcy rule from liquidation to reorganization that made student loans nondischargeable. I develop a life-cycle econ- omy that allows for heterogeneity in assets, earnings post-college, and debt students encountered in college to study repayment incentives created by alternative bankruptcy regimes across different groups of college graduates. The model explains quantitatively and qualitatively characteristics of defaulters under reorganization. I find that the policy change induced a significant decline in default rates. Financially constrained borrowers will choose to default under liquidation, whereas under reorganization bor- rowers default for strategic reasons. Findings suggests substantial welfare gains for bottom earnings quartile relative to the higher earnings quartiles when dischargeabil- ity is allowed. Current work extends this set-up by endogenizing college investment and studies the effects of the policy change on college enrollment and human capital accumulation. JEL classification: D91; G33; I22 Keywords: Bankruptcy Reform; Student loans dischargeability; Default incentives * A special thanks to Kartik Athreya, Dean Corbae, S. Chattarjee, B. Ravikumar, Peter Rupert, Nicole Simpson, Chris Sleet, Gustavo Ventura, Galina Vereschgagina, and Steve Williamson for useful discussions. I would also like to thank participants at Macro Midwest Meetings at Washington University, Midwest Economics Meetings, seminar series at University of Iowa, Cleveland Fed, Colgate University, and Liberal Arts Colleges Series. † Tel.: (315)228-7955; fax: (315)228-7033; E-mail address: fi[email protected]1
30
Embed
Reorganization versus Liquidation: Default Incentives · PDF fileReorganization versus Liquidation: ... the case reorganization and repayment ... and Tertilt incorporates both bankruptcy
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Reorganization versus Liquidation: Default Incentives
for Student Loans∗
Anamaria Felicia Ionescu†
April 2005, Revisited June 2007
Department of Economics, Colgate University , 13 Oak Dr, Hamilton, NY, 13346, USA.
Abstract
I consider the implications of a change in the bankruptcy rule from liquidation toreorganization that made student loans nondischargeable. I develop a life-cycle econ-omy that allows for heterogeneity in assets, earnings post-college, and debt studentsencountered in college to study repayment incentives created by alternative bankruptcyregimes across different groups of college graduates. The model explains quantitativelyand qualitatively characteristics of defaulters under reorganization. I find that thepolicy change induced a significant decline in default rates. Financially constrainedborrowers will choose to default under liquidation, whereas under reorganization bor-rowers default for strategic reasons. Findings suggests substantial welfare gains forbottom earnings quartile relative to the higher earnings quartiles when dischargeabil-ity is allowed. Current work extends this set-up by endogenizing college investmentand studies the effects of the policy change on college enrollment and human capitalaccumulation.
∗A special thanks to Kartik Athreya, Dean Corbae, S. Chattarjee, B. Ravikumar, Peter Rupert, NicoleSimpson, Chris Sleet, Gustavo Ventura, Galina Vereschgagina, and Steve Williamson for useful discussions.I would also like to thank participants at Macro Midwest Meetings at Washington University, MidwestEconomics Meetings, seminar series at University of Iowa, Cleveland Fed, Colgate University, and LiberalArts Colleges Series.
The rapid rise in personal bankruptcy filing rates in the last decade with a historic high of
7% in 1997 (of U.S. adult population) centered attention on the nation’s bankruptcy rule.
The literature is voluminous with the main focus on studying incentives created by various
bankruptcy laws on filing behavior within unsecured credit. In the same period, default rates
for student loans averaged 12% with the highest rate of 22.4% in 1990 (a 2-year basis cohort
default rate). The total amount of outstanding debt reached $25 billion in 2001. Little
attention, however, has been given to analyzing bankruptcy rules under the student loan
market. But evidence about how much borrowers and lenders respond to the incentives in
bankruptcy laws would help policy makers as they work to redesign it. This paper studies
repayment incentives in an environment that mimics the student loan market characteristics
and considers effects of changes in the bankruptcy rule across different groups of college
graduates.
The Federal Student Loan Program (FSLP) has grown significantly in the recent years
with 7 million people currently borrowing under the program.1 One in twenty of them
defaults on his loan payments. High default rates in the late 1980s have led legislators
to introduce a series of policy reforms that gradually made student loans nondischargeable
under Chapter 13 in the Bankruptcy Code. Rather than a disposal of the assets through
liquidation sale under Chapter 7, the reorganization chapter gives the debtor the opportunity
to restructure his assets and liabilities. He needs to reorganize and start repaying his loans.
Dischargeability was initially restricted in 1990 to a 7-year first payment basis or undue
hardship basis, the former feature being eliminated by Higher Education Amendments to the
Bankruptcy Code in 1998. A couple of questions arise immediately: How are the repayment
incentives affected by the change in the bankruptcy rule? What are the implications for
1In the fiscal year 2001, 54$ billion were borrowed under the FSLP. In the same year the unsecured debtamounted to $692 billion.
2
welfare and default rates? I am concerned in particular with redistributional effects. Finally,
what are the consequences for the decision to invest in college?
In order to address the proposed issues, I develop a heterogeneous life-cycle economy that
builds up on my previous work on college enrollment, borrowing and repayment under the
existing FSLP. My previous research focuses on college enrollment, borrowing and repayment
in an environment that accounts for various repayment schemes available under the FSLP, but
abstracts from accounting for different bankruptcy arrangements, the focus of the current
study. In the current paper, I allow for dischargeability on student loans and study the
relationship between debt, earnings and repayment incentives under both liquidation and
reorganization rules. The option to discharge one’s debt provides partial insurance against
bad luck such as job loss, but drives up interest rates making life-cycle smoothing more
difficult. I introduce two sources of uncertainty in this economy: earnings and interest
rate on loans (under the program the interest rate is based on the 91 day Treasury Bill
rate and it fluctuates with the market). The agent can self-insure against these shocks by
accumulating assets. I allow for heterogeneity in earnings, asset level post-college and debt
agents encountered in college to study repayment incentives across different groups of college
graduates.
The novelty of this work is that it simulates bankruptcy characteristics of the student
loan market, which are very different than those of the standard credit markets. Student
loans are not secured by any tangible asset, so there might be some similarities with the
unsecured debt market, but unlike those types of loans (credit cards), guaranteed student
loans are uniquely risky - eligibility conditions being very different. Loans are financial need
based, not credit ratings based and are subsidized by the government. Agents are eligible to
borrow up to the full college cost minus the expected family contribution. When repaying
college loans, borrowers face a menu of repayment schedules. More importantly, the interest
rate does not reflect the risk that some borrowers might exercise the option to default as in
3
the standard credit market, hence the difficulty in my treatment of capturing that particular
risk. The feedback of any bankruptcy law into the interest rates is exactly how the default
is paid for. I endogenize the bankruptcy decision, crucial to the proposed welfare analysis.
I consider penalties on defaulters similar to those implemented in the actual program, that
might bear part of the default risk. To conduct the proposed policy experiments, I first
calibrate the model to match key properties of distributions for college debt, assets and
life-cycle earnings for college graduates.
The model explains quantitatively and qualitatively characteristics of defaulters for the
reorganization period, as delivered by Baccalaureate and Beyond (B&B 93/97) data set for
college graduates in 1992/1993: borrowers with lower earning levels and higher debt levels
are more likely to choose default over repayment. I use the model to run a counter-factual
experiment that allows for the possibility to discharge one’s debt. My results suggests that
the change in the bankruptcy rule from liquidation to reorganization induced a decline in
default rates by 11.66%. In the case of liquidation financially constrained people will choose
to default, whereas under reorganization people default for strategic reasons rather than
financial constraints. The model implies a welfare improvement under liquidation relative to
the case reorganization and repayment is required with substantial welfare gains for bottom
earnings quartile relative to the higher earnings quartiles. Current work extends this set-up
by endogenizing college investment and human capital accumulation and studies the effects
of the change in the bankruptcy rule on college enrollment and human capital investment.
This paper complements two directions of study in the literature: bankruptcy and higher
education policies. The first line of research has focused on personal bankruptcy laws and
their implications for filing rates with significant contributions by Athreya, and Chattergee,
Corbae, Nagajim, and Rios-Rull. These studies innovate by explicitly modeling a menu
of credit levels and interest rates offered by credit suppliers with the focus on default un-
4
der Chapter 7 within the unsecured credit market.2 A recent paper by Livshits, MacGee,
and Tertilt incorporates both bankruptcy regimes, contrasting liquidation within U.S. to
reorganization in Germany in a life-cycle model with incomplete markets calibrated to the
two economies. The literature has mostly abstracted from studying bankruptcy under both
regimes within U.S. Higher education government policies and their effects have been ex-
tensively studied as well with most of the interest directed toward subsidies for financially-
constrained students.3 The literature has generally ignored the analysis of default behavior
under the student loan market. An exception is the work by Lochner and Monge who look at
the interaction between borrowing constraints, default, and investment in human capital in
an environment based on the U.S. Guaranteed Student Loan (GSL) program. They develop
a model to explain empirical findings regarding the characteristics of defaulters. As opposed
to their paper, I incorporate the analysis of incentives created by the reform that changed
the bankruptcy rule and provide both a qualitative and a quantitative assessment of policy
implications on repayment behavior, and default rates.
To my knowledge, this is the first paper to study the effects of both liquidation and
reorganization bankruptcy regimes on default incentives for student loans and welfare. The
paper is organized as follows: Section 2 provides background on the bankruptcy rules; next
two sections describe the model and the calibration procedure; I present results in Section 5
and conclude in the last section.
2 Bankruptcy and The Cost of Default
This section provides some background information on the details of the bankruptcy rules
under the FSLP. The bankruptcy in the model is closely related to Chapter 7, “The Liqui-
2Other relevant research include papers by Athreya and Simpson, Li and Sartre, and White.3Important contributions are papers by Becker, Caucutt and Kumar, Cameron and Heckman and Carneiro
and Heckman, and Keane and Wolpin.
5
dation Chapter” and Chapter 13, ”Adjustment of the Debts of an Individual With Regular
Income”, one of the “Reorganization Chapters” under the Bankruptcy Code. The most sig-
nificant distinction between them is in regard to the administration of the estate. Rather
than a disposal of the assets through liquidation sale, the purpose of the reorganization chap-
ters is to preserve and protect the integrity of assets from the claims of creditors, so as to
permit the debtor an opportunity to reorganize and restructure his assets and liabilities and
to become economically viable. Table 1 presents a summary of main differences between the
Purpose Disposal of the assets Protection of the assets integrityDischargeability Allowed Not allowed
Cost Wage garnishment Wage garnishmentExclusion from credit markets Debt increase
Seizure of tax refunds Seizure of tax refundsLoss of consolidation rights
Benefit Loans discharged Bad credit report erased
Students who participated in the loan program before 1990 could file for bankruptcy under
Chapter 7 without any restrictions and could discharge on their loans. Dischargeability was
initially limited in 1990 by the Student Loan Default Prevention Initiative Act to one of
the two cases: (1) if the first payment on the debt became due more than 7 years before
the filing of the bankruptcy petition or (2) if it would cause undue hardship on the debtor.
The Higher Education Amendments to the Bankruptcy Code in 1998 eliminated the 7-year
discharge basis, keeping ”undue hardship” as the only basis for obtaining a discharge. In
this paper, I refer to this period as “dischargeability” or “liquidation”. After the reforms,
students cannot discharge on their loans anymore. They file for bankruptcy under Chapter
13 and enter a repayment plan. The indebted defaulter is required to reduce consumption
to finance at least partial repayment of his obligations. Availability of discharge is limited,
6
so I will refer to this period as “nondischargeability” or “reorganization”.4
The design of the current program is such that students need to repay on their loans six
month after graduation. The rate on education loans is set by the government, based on
the 91-day Treasury-bill rate, and it fluctuates with the market. Borrowers start repaying
under the standard plan that assumes fluctuating payments. If they do not make any pay-
ments within 270 days, they are considered in default, unless an agreement with the lender
is reached. Default status is reported to credit bureaus. Penalties on the defaulter include:
garnishment of their wage, seizure of federal tax refunds, possible hold on transcripts, in-
eligibility for future student loans, bad credit reports that exclude them from other credit
markets. The main differences between the consequences to default under the two regimes
are that (1) in the case of liquidation, the loan is discharged and no repayment is required
but in the case of reorganization, a repayment plan is implemented; (2) Under liquidation,
defaulters are excluded from credit markets, whereas under reorganization bad credit reports
are erased and credit market participation is not restricted; (3) The wage garnishment is in-
terrupted under reorganization once the defaulter enters repayment. His debt can increase
by as much as 25%. The defaulter loses his right to consolidate after default, so paying un-
der no-consolidation status is the only available option the borrower enters repayment next
period. Institutions with high default rates are also penalized, but I focus on the individual
decision, so I abstract from those.
The Department of Education and student loan guaranty agencies are authorized to take
(”garnish”) a limited portion of the wages of a student loan debtor who is in default. Due to
recent changes in the law, the Department may garnish up to 10% of the defaulter’s wages.
Wage garnishment could be avoided if it would result in an extreme financial hardship for
the defaulter. I will adjust the garnishment relative to the minimum income.
4As a practical matter, it is very difficult to demonstrate undue hardship unless the defaulter is physicallyunable to work.
7
The IRS can intercept any income tax refund the defaulter may be entitled to until the
student loans are paid in full. This is one of the most popular methods of collecting on
defaulted loans, and the Department of Education annually collects hundreds of millions
of dollars this way. Each tax year, the guaranty agency holding the defaulter’s loans must
review the account to verify that the defaulter has not made any loan payments within the
previous 90 days. In this case, the agency notifies the IRS that the loans are in default
and if the defaulter is entitled to a tax refund, the IRS proposes to keep all or some of the
tax refund. This punishment can be avoided, however, if the borrower has repaid the loan,
is making payments under a negotiated repayment agreement or the loans were discharged
in bankruptcy. Thus, I will not model this punishment given either dischargeability or
immediate repayment in my set-up.
3 Model
This model focuses on the repayment incentives across different groups of college graduates
under the FSLP in an environment that mimics the student loan market characteristics
for both liquidation and reorganization periods. The set-up is a life cycle economy with
heterogeneous agents that differ in three dimensions. Time is discrete and indexed by j.
There are two sources of uncertainty in this economy: earnings and interest rate on loans.
Agents maximize the present value of utility over the life-cycle:
maxJ∑
j=1
βj−1u(cj),
where u(.) is strictly concave and increasing, and β is the discount factor. The per-period
utility function is CRRA, u(cj) =cj
1−σ
1−σ, with σ as the coefficient of risk aversion.
In this set-up, college graduates start with the debt accumulated in college, some initial
8
asset and earnings which are contingent on their ability and human capital stock. They need
to repay on their student loans.5 Education investment is risky: the interest rate on loans is
uncertain. Agents optimally choose the repayment status for college loans after they learn
out the realized interest rate and income shock. Agents may save at the riskless interest rate.
3.1 Agent’s Problem
The agent maximizes the present value of utility over the lifetime by saving using the risk
free asset and choosing the repayment status for his college loan. The problem is given by:
maxxj ,pj ,dj
[E
∑Jj=1 βj−1 c1−σ
j
1−σ
]
s.t. cj ≤ zjyj − pj(dj) + xj(1 + rf )− xj+1
dj+1 = (dj − pj)(1 + rj), pj ∈ P, xj+1 > 0.
x j Agents derive utility from consuming each period. They supply one unit of time, the
productivity of which is given by zjyj with zj the stochastic component and yj the age specific
conditional mean of labor productivity, given agent’s initial earnings post-college, y1. The
idiosyncratic shocks, zj to earnings each period evolve according to a Markov process with
support Z=[z, z], where z represents a bad shock and z represents good productivity shock.
The Markov process is characterized by the transition function Qz and is assumed to be the
same for all households. Their current savings are xj+1 and bankruptcy status is reflected
by the payment they have chosen pj(dj) ∈ P = [pnc, 0] with pnc the variable payment and 0
payment in the case they declare bankruptcy.
The problem is formulated in a dynamic programming framework. In the last period of
5I abstract from including the college enrollment and human capital decisions for now in order to focuson the repayment behavior for both bankruptcy regimes.
9
life, however, the agent consumes his savings to allow for heterogeneity in savings at the end
of the life-cycle. The value function after the last period of life is set to VJ+1(x, d, r, z) = u(x).
The problem is solved backwardly and the Bellman equation is given by
Vj(y, x, d, r, z) = maxx′,p,d′
(zyj + (1 + rf )x− x′ − p(d))1−σ
1− σ
+βEVj+1(y′, x′, d′, r′, z′)
s.t. d′ = (d− p)(1 + r), p ∈ P, x′ > 0
with expectation over both income and rate shocks. Solutions to this problem are given by
optimal decision rules: x∗j(y1,x, r, z) and p∗j(y,1x, d, r, z), which describes the optimal choice
of asset carried to the next period and the optimal repayment choice as a function of age,
j, initial earnings post-college, y1, assets, x, and college debt, d, when the realized state is
r and z The value function, V1(y1, x1, d1), gives the maximum present value of utility from
state y1 when initial assets are x1 and debt level is d1. Corresponding to the bankruptcy
status, there are two value functions: the value for the case in which the borrower is repaying
and the value in the case the borrower declares bankruptcy.
Case 1: Repayment status
The agent has not defaulted yet, but he might choose to do so in the current period.
Optimal repayment implies maximizing over the two value functions, V R, which represents
maintaining the repayment status and V AD, which means default occurred in the current
period.
V Rj (y, x, d, r, z) = max
l,h′,x′,p,d′
(zyj + (1 + rf )x− x′ − p(d))1−σ
1− σ+ βE max[
V Rj+1(y
′, x′, d′, r′, z′), V ADj+1 (y′, x′, d′, r′, z′)
]s.t. d′ = (d− p)(1 + r), p ∈ P
10
With the appropriate parameters and the estimated Markov processes for loan rates and
income shocks, I solve for the optimal repayment choice, p∗j(y1, x, d), ∀j = 1, 2..., J.
Case 2: Default
The value functions for the default status are given separately for each bankruptcy rule.
• Dischargeability/Liquidation
Once the borrower decides to declare bankruptcy, there is no repayment in the period default
occurs and any period thereafter. The consequences to default are modeled to mimic those
in the data: a wage garnishment and exclusion from credit markets. In my model, this
corresponds to a garnishment of a fraction ρL of the earnings and the inability to borrow
and save in the risk free market within three periods after default.6 V DL represents the value
function for the period default occurs and the following two periods
V DLj (y, x, d, r, z) =
[(zyj(1− ρL) + (1 + rf )x)1−σ
1− σ+ βEV DL
j+1(y′, z′)
]
s.t. d′ = 0
and V ADL represents the value function for periods after default, when liquidation occurred
and penalties were imposed.
V ADLj (y, x, z) = max
x′>0
[(zyj(1 + rf )x− x′)1−σ
1− σ+ βEV ADL
j+1 (y′, x′, z′)
]
• Non-Dischargeability/Reorganization
6Prohibiting saving is meant to capture the seizure of assets in a Chapter 7 bankruptcy.
11
In the case of bankruptcy under reorganization, the consequences to default include a wage
garnishment the period default occurs, reflected in the model by ρR ∈ (0, 1), and an increase
in the debt level by µ the agent enters repayment the next period. Once he enters repayment,
it is assumed that he will never default again. The bad credit reports are erased and he is
allowed to borrow/save in the bond market without any restrictions. V DR represents the
value function for the period default occurs
V DRj (y, x, d, r, z) = max
x′
[(zyj(1− ρR) + (1 + rf )x− x′)1−σ
1− σ+ βEV ADR
j+1 (y′, x′, d′, r′, z′)
]
s.t. d′ = d(1 + µ)(1 + r),x′ > 0
and V ADR represents the value function for periods after default, when reorganization is
required.
V ADRj (y, x, d, r, z) = max
x′
[(zyj + (1 + rf )x− x′ − pnc(d))1−σ
1− σ+ βEV ADR
j+1 (y′, x′, d′, r′, z′)
]s.t. d′ = (d− pnc)(1 + r),x’>0
4 Calibration
In order to conduct policy experiments, I first calibrate the above life-cycle model. The
calibration process involves the following steps: First, I assume parameter values for which
literature provides evidence and I use data from the Department of Education to set the
policy parameters. Second, I calibrate the loan rates process, using the time series for 3-
month Treasury Bills for 1980-1996. The third step involves calibrating the distribution
12
of initial asset holdings, the distribution of education debt and market entry earnings for
college graduates. The initial asset distribution is calibrated using youth assets for college
graduates using the Survey of Consumer Finance (SCF) 1983. The distribution of education
debt and market entry earnings for college graduates are calibrated using the Baccalaureate
and beyond (B&B) data set. The calibration of the earnings process uses the PSID family
files 1969 to 2001. The model period equals one year.
4.1 Parameters
The parameter values are given in table 2 .
Table 2: Parameter ValuesParameter Name Value Target/Source
β Discount factor 0.96 Avg real interest rate=4%σ Coef of risk aversion 2rf Risk free rate 0.04 Avg real interest rateJ Model periods 34 Real life age 24-57T Loan duration 10 Dept. of Educationρ Wage garnishment upon default 0.1 Dept. of Educationµ Debt increase upon default - Ch 13 0.05 Dept. of Ed. - default ratee Minimum earnings upon default $4117 Dept. of Education
The discount factor is 1/1.04 to match the risk free rate of 4%, and the coefficient of risk
aversion chosen is standard in the literature. The agent’s lifetime is taken to be 34 model
periods, which corresponds to a real life age of 24 to 57. The penalties upon default are picked
according to data from the Department of Education. These involve a wage garnishment
upon default, ρL = ρR = 0.10, and the minimum wage that would trigger financial hardship
on the part of the defaulter. The Department of Education and student loan guaranty
agencies are authorized to garnish up 10% of the defaulter’s wages. The garnishment is not
imposed, however, if it means that the weekly income would be less than 30 times the federal
minimum wage. Based on the current minimum wage of $5.15, this means that a minimum
13
of $154.40 (30 x $5.15) of the weekly wages are protected from garnishment. In my model,
this would be translated in an annual minimum income of $4117 in 1984 constant dollars.
The debt increase is picked to target the default rate for college graduates 92/93 in the B&B
data sample from the NPSAS data set (Department of Education).
4.2 Loan rate process
The rate follows a stochastic process, given by a 2 by 2 transition matrix Π(R′, R) on {R, R},
calibrated to match the Treasury Bills rate in the 90s. To estimate the stochastic process
for loan rates, I use the time series for 3-month Treasury Bills for 1980-1996, adjusted for
inflation. I fit the time series with the AR(1) process: Rt = µ(1−ρ)+ρRt−1+ε, ε ∼ N (0, σ2).
The estimates of the two moments are given by ρ = 0.9038 and σ = 0.7788. I aggregate
this to annual data; the autocorrelation is given by 0.297 and the unconditional standard
deviation by 1.817. I have approximated this process as a two-state Markov chain. The
support is R ∈ {1.038, 1.075}. The transition matrix is
0.65 0.35
0.35 0.65
.
4.3 Distribution of college debt and initial assets and earnings
post college
In the distribution of asset, education debt, and earnings post-college, I use the SCF and
the B&B data set respectively for college graduates 1992/1993. Measure of assets includes
paper assets, current value of home, value of other properties, value of all vehicles, value of
business. Paper assets are given by the sum of financial assets, cash value of life insurance,
loans owed, gas leases, value of land contracts and thrift accounts. Debt includes only loans
granted under the FSLP and guaranteed by the government. Other sources of financial aid
such as family loans or private loans not guaranteed are excluded. The summary statistics
14
are given in table 3 in constant dollars 1984 and the densities of initial assets and college
debt in figures 1 and 2. The correlation between earnings at the market entry level and debt
post-college is taken from the B&B data set and set to 0.63.
Table 3: Asset and Debt Post-collegeSummary Statistics* Mean Standard Deviation
Asset holdings 10434 19304Debt post college 7369 5044
Earnings post college 15604 7052* This is in 1984 constant dollars.
Figure 1: Asset Density
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
x 104
0
1
2
x 10−4
Constatnt Dollars 82−84
Figure 2: Debt Density
0 0.5 1 1.5 2 2.5 3
x 104
0
0.2
0.4
0.6
0.8
1
1.2
1.4x 10
−4
Constant dollars 82−84
15
Figure 3: Age Profile of Earnings
4.4 Earnings Process
The earnings process is given by the product zjyj where zj is the stochastic component. The
deterministic age profile of labor productivity post-college {yj}Jj=1 is taken from my previous
research where the age profile of earnings is computed based on PSID family files for college
graduates, given individual characteristics: ability and human capital stock. Figures 3 show
life cycle real earnings given the market entry level of earnings post college.
In the parametrization of the stochastic idiosyncratic labor productivity process I follow
Storesletten, Telmer, and Yaron (2004) who build a panel from the Panel Study of Income
Dynamics (PSID) to estimate the idiosyncratic component of labor earnings. They use
annual data from PSID from 1968 to 1991 for wage earnings and report separate values
for different skill levels. This estimation is suitable for my model. With uij = ln(zij) the
16
stochastic part of the labor income process for household i at time j, the estimated model is:
uij = yij + εij
yij = ρzi,j−1 + νij
where εij N(0, σ2ε ) and νij N(0, σ2
ν) are innovation processes. The variables yij and εij are
realized at each period over the life cycle and are referred to as persistent and transitory ’life-
cycle shocks”, respectively. The reported values are ρ = 0.935, σ2ε = 0.017, and σ2
ν = 0.061. I
have approximated this process as a two state Markov Chain, normalizing the average value
for the idiosyncratic shock to be 1. The resulting support is the set Z = {0.286, 1.714} with
the transition probability matrix:
0.9455 0.0045
0.0455 0.9455
.
5 Results
I first describe data findings regarding repayment incentives under reorganization and the
predictions of the model for this bankruptcy regime; in Section 5.2 I present the results of
the counterfactual experiment that changes the bankruptcy rule to allow for dischargeability
and I analyze the relationship between college debt, post-graduate earnings, and default
under both bankruptcy rules. Finally, in section 5.3, I present welfare implications of the
two regimes.
5.1 Repayment Incentives under Reorganization
Table 4 describes data findings for default rates across different groups of college graduates
who re-pay on their education loans facing chapter 13 bankruptcy. Data descriptions are
17
Table 4: Default By Characteristics - Data - ReorganizationVariable Default Rate Sample Size
based on NPSAS samples. I use B&B for college graduates 1992/1993. The data set is
nationally representative and it is comprised in students, parents and institutions. School
information is transcript based and student information is based on interviews. The survey
has followed a random sample of 11,000 individuals who received their baccalaureate degree
during the 1992-1993 academic year through 1997. There is an initial survey at graduation
time and two follow up interviews in 1994 and 1997. I restrict my attention to the graduates
that borrowed for undergraduate education under the FSLP and graduated from college in
the period 1992/07 - 1993/06. I do not take into account students who went to graduate
school because those continuing to graduate school are eligible for deferments in their loan
repayment. The default rate is defined as the number of respondents who reported that they
had defaulted since graduation, divided by the total number of respondents.
Interesting empirical patterns include:7
7Other two features, that I do not present here and I do not model include:1. Default rates vary across students with different undergraduate majors, but those differences largely
18
Table 5: Default Rates - ReorganizationVariable Model Data
All 5.6 5.6Average annual earningsQuartile 1 (< $10000) 8.74 8.36
1. Default rates are declining in earnings, both conditioned and unconditioned on college
debt.
2. Default rates are increasing in education debt levels both conditioned and uncondi-
tioned on earnings post-college.
3. Default rates are U-shaped in SAT/ACT scores, even after controlling for post-college-
earnings and education debt.
Figures 4 show default rates against earnings levels for all four groups of debt post-college
and against debt levels for all four groups of earnings post-college respectively.
Table 5 and figures 5 present the model counterparts for default rates across different
groups of characteristics. The model replicates the first two facts quantitatively and qualita-
tively. Borrowers with lower earning levels are more likely to choose default over repayment.
Given the income contingent punishment upon default, the consequences for borrowers within
lower income groups will be less severe. Furthermore, if borrowers qualify for the minimum
level of earnings, wage garnishment is not imposed. Borrowers with higher debt levels will
be more inclined to opt for default, given higher debt burdens relative to their income. Note
disappear after controlling for actual post-school earnings and education debt.2. Blacks and Hispanics default at significantly higher rates than whites and Asians, even after controlling
for actual post-school earnings and education debt.
19
Figure 4: Default Rates - Data - Reorganization
Note: Default rates are given for 1997, four years after graduation against earnings post-college at the market entrance level and against debt accumulated in college.
20
Figure 5: Default Rates - Model - Reorganization
Note: Default rates are computed four years after graduation against earnings post-collegeat the market entrance level and against debt accumulated in college.