second chance: life without student debt · 2020-06-20 · student debt with 800,000 private...

Motivation Setting Data Main Results Conclusions

Second Chance: Life without Student Debt

Marco Di Maggio1, Ankit Kalda2 and Vincent W. Yao3

1Harvard Business School & NBER

2Indiana University

3Georgia State University

Oct 2019

Di Maggio, Kalda and Yao Student Debt Relief 1 / 32


Student Debt Delinquencies on the Rise

14 Di Maggio, Kalda and Yao Student Debt Relief 2 / 32


Policy Makers Worry about Student Debt Defaults

The newly appointed Chairman of the Federal Reserve stated

“You do stand to see longer-term negative effects on people who can’t pay off their student loans. It hurts their credit rating, it impacts the entire half of their economic life, as this goes on and as student loans continue to grow and become larger and larger, then it absolutely could hold back growth.”



Policy Debate


Three main findings

Discharged borrowers reduce their credit balances, by both reducingtheir demand for credit and limiting the use of the existing account.They are also significantly less likely to default on other accounts, filefor bankruptcy and experience medical defaults.These borrowers’ geographical mobility increases, as well as, theirprobability to change job and ultimately their income.

In contrast to most existing literature, we show the importance ofrelaxing long-term constraints.


This Paper

We use plausibly exogenous variation to quantify the effects of student debt forgiveness on credit and labor market outcomes for distressed borrowers (don’t quantify costs).


Discharged borrowers reduce their credit balances, by both reducingtheir demand for credit and limiting the use of the existing account.They are also significantly less likely to default on other accounts, filefor bankruptcy and experience medical defaults.These borrowers’ geographical mobility increases, as well as, theirprobability to change job and ultimately their income.



This Paper


Three main findings


They are also significantly less likely to default on other accounts, filefor bankruptcy and experience medical defaults.These borrowers’ geographical mobility increases, as well as, theirprobability to change job and ultimately their income.



This Paper


Three main findings

Discharged borrowers reduce their credit balances, by both reducing their demand for credit and limiting the use of the existing account.


These borrowers’ geographical mobility increases, as well as, theirprobability to change job and ultimately their income.



This Paper


Three main findings

Discharged borrowers reduce their credit balances, by both reducing their demand for credit and limiting the use of the existing account. They are also significantly less likely to default on other accounts, file for bankruptcy and experience medical defaults.




This Paper


Three main findings

Discharged borrowers reduce their credit balances, by both reducing their demand for credit and limiting the use of the existing account. They are also significantly less likely to default on other accounts, file for bankruptcy and experience medical defaults. These borrowers’ geographical mobility increases, as well as, their probability to change job and ultimately their income.



This Paper


Three main findings

Discharged borrowers reduce their credit balances, by both reducing their demand for credit and limiting the use of the existing account. They are also significantly less likely to default on other accounts, file for bankruptcy and experience medical defaults. These borrowers’ geographical mobility increases, as well as, their probability to change job and ultimately their income.

In contrast to most existing literature, we show the importance of relaxing long-term constraints.



Student Loan Market: Background

Student loans are basically split into federal loans and private student loans.

The rate for federal loans is fixed and set by Congress.

In almost all cases, federal student loans have better terms than the heavily advertised and expensive private student loans.

Many people with private student loans, like those who took on subprime mortgages, end up shouldering debt that they never earn enough to repay.


We use “Clerical Errors” made by the trust in this process that wipedoff loans for students whose paperwork were not in order/went missing

Trust lost a series of collection lawsuits against the borrowers theywere trying to collect from because they failed to prove that theyowned the debt in the first place.

We hand-collect a unique dataset with information about theselawsuits, including borrower identities and addresses among otherthings, and merge this with credit bureau data to examine acomprehensive set of outcomes.


Setting: Random Debt Discharge

National Collegiate Student Loan Trust is the largest holder of private student debt with 800,000 private student loans totaling $12 billion

Purchases student loans from lenders, securitizes them and allows investors to invest in bonds


Trust lost a series of collection lawsuits against the borrowers theywere trying to collect from because they failed to prove that theyowned the debt in the first place.






We use “Clerical Errors” made by the trust in this process that wiped off loans for students whose paperwork were not in order/went missing








Trust lost a series of collection lawsuits against the borrowers they were trying to collect from because they failed to prove that they owned the debt in the first place.







Trust lost a series of collection lawsuits against the borrowers they were trying to collect from because they failed to prove that they owned the debt in the first place.

We hand-collect a unique dataset with information about these lawsuits, including borrower identities and addresses among other things, and merge this with credit bureau data to examine a comprehensive set of outcomes.



Credit Bureau Data

Using identities and addresses from the court filings data, we requested Equifax to merge it to their anonymous data on individuals that has two components to it.

Credit Data: covers 260 million borrowers and over 1.8 billion single loans, and is updated monthly.

Includes loan level information on account type, originations, balances, credit limits, monthly payments and performance.

Income Data: covers 30 million individuals and has information on income, employer, industry of employment, job titles among other variables.



Distribution of Legal Settlements

Calculated by authors based on the sample used in the paper.



Empirical Methodology

Difference-in-differences framework where individuals involved in the failed collection lawsuits (whose debt is discharged) constitute our treatment group.

For every treated individual, we find a matched control group comprising all individuals who reside in the same ZIP code, are of same age, carry similar student loan amounts, and crucially, who defaulted on their student loans as well (but their debt is not discharged).

In robustness checks, we also provide evidence that the results hold when we take advantage of the different discharge timing within the sample of treated individuals.



Empirical Methodology

Formally, the main specification is the following:

Outcomei ,j ,t = α + β × (Treatedi × Postt ) + µi + γi×τ + εi ,j ,t (1)

We include individual and County by event-month fixed effects.

The Post dummy is purposely capturing several months (36) after the discharge because for some of our outcomes we would expect a lagged reaction.


(sd of 11,600)

Dependent Var Student Loan Student Loan Credit ScoreAccounts Balance

(1) (2) (3)

DebtRelief × Post -0.65*** -5319.04*** 7.03***(0.05) (148.94) (0.98)

Individual FE Yes Yes YesCounty x Event-Month FE Yes Yes Yes

Observations 6,010,381 6,010,381 6,010,381R2 0.8 0.84 0.58

Credit Score increases for borrowers experiencing discharge


Student Debt Relief Validation

Debt relief is associated with an average decline of student loan balance of $6,855


Dependent Var Student Loan Student Loan Credit ScoreAccounts Balance

(1) (2) (3)

DebtRelief × Post -0.65*** -5319.04*** 7.03***(0.05) (148.94) (0.98)

Individual FE Yes Yes YesCounty x Event-Month FE Yes Yes Yes

Observations 6,010,381 6,010,381 6,010,381R2 0.8 0.84 0.58




Debt relief is associated with an average decline of student loan balance of $6,855 (sd of 11,600)






Dependent Var Student Loan Student Loan Credit Score Accounts Balance

(1) (2) (3)

DebtRelief × Post -0.65*** -5319.04*** 7.03*** (0.05) (148.94) (0.98)

Individual FE Yes Yes Yes County x Event-Month FE Yes Yes Yes

Observations 6,010,381 6,010,381 6,010,381 R2 0.8 0.84 0.58



Main Results

Debt Behavior

Delinquency and Bankruptcy outcomes

Mobility and Income



Debt Balances

Dependent Var Total Balance Credit Card Auto Mortgage (Ex. Stud) Balance Balance Balance

(1) (2) (3) (4)

DebtRelief × Post -4,303.21*** -369.44*** -226.81*** -888.24*** (652.21) (28.99) (69.58) (163.55)

Individual FE Yes Yes Yes Yes County x Event-Month FE Yes Yes Yes Yes

Observations 6,010,381 6,010,381 6,010,381 6,010,381 R2 0.8 0.83 0.77 0.86

Average debt (ex student loans) declines by 26% relative to the mean before the discharge

Decline occurs across different components of debt


http:4,303.21


Total Debt: Dynamics

●

● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●

●● ● ● ●

●

●

●●● ●

●● ● ● ●

●●●

●●● ●

●●●● ●

●● ●

−10

000

−50

000

Distance from Debt Discharge (Months)

Tota

l Deb

t (E

x S

tude

nt L

oans

)

−20 −15 −10 −5 0 5 10 15 20



How borrowers reduce debt: Credit Cards & Home Loans

Dependent Var Account Utilization Payment Account Origination Payment Opening Opening Amount

(1) (2) (3) (4) (5) (6)

DebtRelief × Post -0.002** -0.018*** 12.58*** -0.001** -9,402.83** 38.98** (0.001) (0.004) (1.99) (0.0004) (3,799.04) (13.10)

Individual FE Yes Yes Yes Yes Yes Yes County × Event-Month FE Yes Yes Yes Yes Yes Yes

Observations 6,010,381 6,010,381 1,299,622 6,010,381 2,042,908 1,291,613 R2 0.095 0.607 0.58 0.36 0.95 0.89

Account openings and origination amount (conditional on openings) decline while payments increase

Similar results for auto loans



How borrowers reduce debt: Credit Utilization

●●● ●

●

● ● ●

● ●

●● ● ●

● ● ● ● ● ●● ● ●

●

●

● ●

● ● ●●●●●●● ●

●●

●● ● ●

●

●

●

●

●●

−0.

04−

0.02

0.00

0.02


Cre

dit C

ard

Util

izat

ion

−20 −15 −10 −5 0 5 10 15 20



Credit Demand

Dependent Var Total Multi–Inquiry Inquiries Indicator

(1) (2)

DebtRelief × Post -0.24*** -0.02*** (0.050) (0.005)

Individual FE Yes Yes County x Event-Month FE Yes Yes

Observations 6,010,381 6,010,381 R2 0.56 0.45

Demand less credit as reflected by decline in inquiries.



Main Results

Debt Behavior Debt declines as borrowers demans less credit and make higher payments

Delinquency and Bankruptcy outcomes

Mobility and Income



Delinquency Extensive Margin

Dependent Var All DLQ Credit Card Auto Mortgage Accounts DLQ DLQ DLQ (Ex. Stud) Accounts Accounts Accounts

(1) (2) (3) (4)

DebtRelief × Post -0.11*** -0.10*** -0.01** -0.01*** (0.020) (0.020) (0.004) (0.003)

Individual FE Yes Yes Yes Yes County x Event-Month FE Yes Yes Yes Yes

Observations 6,010,381 6,010,381 6,010,381 6,010,381 R2 0.74 0.74 0.7 0.76

Borrowers are 11% less likely to become delinquent on other loans.

Less likely to file for bankruptcy, be foreclosed upon and default on medical bills



Delinquency: Dynamics

●

● ●●●●

●● ● ● ●

●● ● ●

●●● ●

●●● ●

●

●

●●●

●●● ●

● ● ● ● ● ● ●●●

●●●

● ● ● ● ●

−0.

2−

0.1

0.0

0.1


All

Del

inqu

ent A

ccou

nts

(Ex

Stu

dent

Loa

ns)

−20 −15 −10 −5 0 5 10 15 20



Main Results

Debt Behavior Debt declines as borrowers lower credit demand and increase payments

Delinquency and Bankruptcy outcomes Less likely to become delinquent and file for bankruptcy

Mobility and Income



Mobility & Income

Dependent Var Mobility Job Change Moving to Moving to Income ($) Different Higher Paying Industry Industry

(1) (2) (3) (4) (5)

DebtRelief × Post 0.005*** 0.004** 0.003** 0.01** 79.72*** (0.001) (0.002) (0.001) (0.004) (31.330)

Individual FE Yes Yes Yes Yes Yes County x Event-Month FE Yes Yes Yes Yes Yes

Observations 6,010,381 967,411 967,219 245,114 471,547 R2 0.31 0.19 0.17 0.34 0.57

Higher geographic and job mobility

Earn $3,000 more over three years following discharge



Income: Dynamics

●

●●●

● ●

● ●

●

●

●●●

● ● ●●

●

●

● ●

●●●

● ●●

●● ●

● ●●

● ●

●

● ● ● ●●

●

●

●

●

●

●●

●

−20

00

200

400


Inco

me

($)

−20 −15 −10 −5 0 5 10 15 20



Main Results

Debt Behavior Debt declines as borrowers lower credit demand and increase payments

Delinquency and Bankruptcy outcomes Less likely to become delinquent and file for bankruptcy

Mobility and Income Earn $3,000 higher total income over three years following discharge



Plausible Mechanisms

Liquidity constraints Most borrowers might’ve stopped paying off their loan - debt relief less likely to impact disposable income Control group might still be subject to wage garnishment - there might be differential impact on disposable income But we find similar results for sample of only treated borrowers Less likely to be the driving mechanism

Credit Score Changes Controlling for changes in credit scores yields similar/stronger estimates Less likely to be important in our setting



Plausible Mechanisms

Delinquency Flag Removal Delinquency flag on credit file may limit employment opportunity set if employers check for it Flag removal on debt discharge may allow borrowers to access labor market more freely

Debt Overhang High levels of debt for delinquent borrowers may reduce their incentives to look for better opportunities Relieving debt will lower this friction Find stronger results for higher levels of debt relief - mobility and income results concentrated among borrowers experiencing higher than median levels of debt relief (where avg relief is 12k)



Other Results and Robustness

Durable consumption as measured by car purchases increases

Similar results using sample of only treated individuals, i.e. using different discharge timing within the sample of treated individuals as the source of variation.

Similar results using sub-sample of individuals for which we observe income data

Robust to different specifications (e.g. balanced panel one year around treatment, double clustering etc)


They are also significantly less likely to default on their accounts,above and beyond their student loan accounts.

Significantly more likely to move, change job and experience anincrease in income.

Although our analysis shows significant benefits, we are not able toestimate the welfare consequences of these policies.


Concluding Remarks

Borrowers experiencing debt relief are significantly more likely to engage in deleveraging, by both reducing their demand for credit and limiting the use of the existing account.


Significantly more likely to move, change job and experience anincrease in income.



Concluding Remarks


They are also significantly less likely to default on their accounts, above and beyond their student loan accounts.




Concluding Remarks



Significantly more likely to move, change job and experience an increase in income.



Concluding Remarks



Significantly more likely to move, change job and experience an increase in income.

Although our analysis shows significant benefits, we are not able to estimate the welfare consequences of these policies.



Summary Statistics

Panel A: Statistics of the Sample

Variable Mean St. Dev. Min Median Max

Number of Accounts 6.599 4.704 0 5 25 Total Debt ($) 25,690.01 39,779.00 0 12,652 293,080 Total Debt (Ex Student Loans, $) 16,300.29 130,076.15 0 7,930 293,080 Number of Accounts (Ex Student Loans) 2.901 3.284 0 2 25 Credit Card Accounts 2.116 2.589 0 1 13 Auto Accounts 0.538 0.757 0 0 3 Mortgage Accounts 0.115 0.421 0 0 3 Credit Card Balance ($) 1,132.53 2,431.23 0 0 18,017 Auto Balance ($) 3,943.34 7,021.75 0 0 31,877 Mortgage Balance ($) 4,422.04 22,359.52 0 0 175,443 Credit Card Utilization 0.341 0.338 0 0.258 1 Auto Loan Origination Amount ($) 20,629.78 12,724.36 550 17,339 77,868 Mortgage Origination Amount ($) 214,839.02 186,797.58 22,900 154,777 507,750 All Delinquent Accounts (Ex Student Loans) 1.302 1.864 0 1 17 Total Past-Due Amount (Ex Student Loans, $) 2,213.92 4,891.69 0 907 54,455 Mobility (1/0) 0.035 0.183 0 0 1 Income ($) 2,376.71 1,636.62 830.21 2,531.19 9,588.85 Credit Score 535.25 74.44 300 530 836


http:9,588.85http:2,531.19http:1,636.62http:2,376.71http:4,891.69http:2,213.92http:186,797.58http:214,839.02http:12,724.36http:20,629.78http:22,359.52http:4,422.04http:7,021.75http:3,943.34http:2,431.23http:1,132.53http:130,076.15http:16,300.29http:39,779.00http:25,690.01


Summary Statistics

Panel B: Different Population and Samples

All All Delinquent Sample Borrowers Student Student Treated (1% CCP) Loan Loan Individuals

Population Population

Number of Accounts 11.23 11.26 8.90 9.29 Total Debt ($) 22,271.52 36,105.21 40,634.51 49,943.09 Credit Card Accounts 11.84 11.28 4.61 2.96 Auto Accounts 0.95 1.09 0.78 0.63 Mortgage Accounts 0.80 0.71 0.23 0.19 Credit Card Balance ($) 51.78 134.70 269.37 1829.39 Auto Balance($) 16,954.98 16,595.81 14,353.55 4,464.43 Mortgage Balance ($) 186,211.67 194,967.58 134,257.00 6,469.94 Credit Card Utilization 0.43 0.64 0.98 0.37 Delinquent Accounts 0.44 0.83 3.44 5.15 Total Past-Due Amount ($) 1,471.48 2,580.82 14,847.59 6,028.63 Age 49.32 37.79 39.52 34.75


http:6,028.63http:14,847.59http:2,580.82http:1,471.48http:6,469.94http:134,257.00http:194,967.58http:186,211.67http:4,464.43http:14,353.55http:16,595.81http:16,954.98http:49,943.09http:40,634.51http:36,105.21http:22,271.52

MotivationSettingDataMain ResultsConclusions

second chance: life without student debt · 2020-06-20 · student debt with 800,000 private...

Documents