Title: Screening for Moral Hazard and Adverse Selection: Evidence from the Home Equity Market
1Screening for Moral Hazard and Adverse
Selection Evidence from the Home Equity Market
- Sumit Agarwal, Federal Reserve Bank
- of Chicago
- Brent W. Ambrose, Penn State University
- Souphala Chomsisengphet, OCC
- Chunlin Liu, Univ. of Nevada-Reno
2Theoretical Motivation
- Stiglitz and Weiss (1981)
- Despite the use of interest rate or collateral to
screen borrowers, lenders still face imperfect
information and are not able to entirely
distinguish borrower risks. - Overall expected loan profitability declines even
when loan rate increases - High-risk applicants will accept the higher
interest rate while low-risk applicants will exit
the applicant pool. - Adverse selection problem ? credit rationing
- Bester (1985)
- Menu of contracts containing combinations of
interest rate collateral - Borrowers contract selection reveals their risk
level ex ante - High-risk borrowers select lower collateral
requirement (higher rates) - Low-risk borrowers select higher collateral
requirement (lower rates) - Impact of adverse selection on credit rationing
is then eliminated
3Theoretical Motivation
- Definitions
- Adverse selection is an ex ante event that occurs
when potential borrowers respond to credit
solicitations offered by banks. - Riskier borrowers respond to credit offerings at
higher interest rates and/or lower collateral
requirements - Moral hazard usually refers to the incentives (or
lack thereof) for borrowers to expend effort to
fulfill their contractual obligations.
4Our Objectives
- Research Questions
- Part 1
- Do borrowers self-select loan contracts designed
to reveal information about their risk level
(Bester, 1985)? - Conditional on the borrowers contract choice,
does adverse selection still exist (Stiglitz and
Weiss, 1981)? - Part 2
- Do lender efforts to mitigate adverse selection
and moral hazard problems effectively reduce
default risks ex post? - If so, by how much?
5Home Equity Credit Market
- Home equity represents a large (and growing)
segment of the consumer credit market. - Market Size (2005) 702 billion
- Typical Home Equity Menu
- Risk-based pricing according to loan-to-value
- Less than 80 LTV
- 80 to 90 LTV
- Greater than 90 LTV
- Thus, ideal setting for examining adverse
selection and moral hazard.
6(No Transcript)
7Data
- Home equity contract originations from a large
financial institution - 108,117 consumers applying for home equity
contract from lenders standardized menu (March -
December 2002) - 8 Northeastern states MA, ME, CT, NH, NJ, NY,
PA, RI - Observe
- Borrowers initial contract choice
- Lenders primary screening (accept, reject, or
additional screening) - Lenders counteroffer
- Borrowers response to counteroffer
- Borrowers repayment behavior (origination -
March 2005) - Other observable information
- Borrowers credit quality and purpose for the
loan - Demographics income, debts, age, occupation
8Data
9Empirical Analysis
101.1 Contract Choice
- Three contract choices ? borrower risk sorting
mechanism - LTV ? 80 ? pledging at least 20 cents per dollar
loan (j1) - 80 lt LTV lt 90 ? pledging 20-10 cents per dollar
loan (j2) - LTV ? 90 ? pledging 10 cents or less per dollar
loan (j3) - Test whether riskier borrowers (lower credit
quality) tend to self-select a higher risk
contract (offer less collateral) - W borrower credit quality
- X control variables (demographics, prop type,
loan purpose, etc...)
111.1 Contract Choice Table 3
- Independent Variables
- Borrower Characteristics
- Borrower risk (FICO and FICO2)
- Log(Income)
- Log (Borrower Age)
- Log (House Tenure)
- Debt-to-income ratio
- Contract Characteristics
- First or Second Lien position indicator
- Line or Loan indicator
- Use of funds indicator
- (refinance, consumption, home improvement)
- First mortgage indicator
- Second home indicator
- Condo indicator
- Employment Control Variables
- Employment tenure Log(Years on the Job)
- Type of employment
- self-employed, retired, home-maker
121.1. Contract Choice Table 3
- Less credit-worthy borrowers (lower FICO) are
more likely to apply for higher LTV home equity
products (pledging less collateral per dollar). - For example,
- Relative to a borrower with a score of 800, a
borrower with FICO score of 700 is 18.4 more
likely to select an 80-90 LTV contract than one
with LTV ? 80. - Relative to a borrower with a score of 800, a
borrower with FICO score of 700 is 19.6 more
likely to apply for a LTV gt 90 than one with LTV
? 80. - Consistent with predictions by Bester (1985).
131.1 Contract Choice
- Conclusion
- We find evidence that borrowers do select
contracts that reveal information about their
risk level.
141.2 Lender response (Table 5)
- If lender systematically screens for adverse
selection and moral hazard, then we should
observe a positive correlation between the
likelihood of additional screening and collateral
offered (LTV), holding all else constant. - Multinomial logit model
- The likelihood of a lender rejecting an applicant
or subjecting an applicant to additional
screening based on LTV, borrower risk
characteristics, loan characteristics, and other
control variables. - Base case loans that were accepted out-right
(without additional screening)
151.2 Lender response (Table 5)
- Lender more likely to conduct additional
screening or reject contracts with lt 20 cents per
dollar of collateral than those with gt 20 cents
per dollar of collateral. - For example,
- LTV gt 90 contract is 18.4 more likely to be
rejected (15.8 more likely to be screened again)
than LTV 80 contract. - 90 ? LTV gt 80 contract is 8.7 more likely to be
rejected (12 more likely to be screened again )
than LTV 80 contract. - 80-90 LTV contract lender more likely to conduct
additional screening than reject. - LTV gt 90 contract lender more likely to reject
than conduct additional screening.
161.2 Lender Response
- Conclusion
- Evidence that lender followed standard
underwriting protocol.
171.3 Test for Adverse Selection
- Test for the presence of adverse selection
conditional on the borrowers choice of contract
type - Examine the loan performance of the 62,251
borrowers whose applications were accepted
outright (without additional screening). - Competing-Hazard Model of Default Prepayment
- The time to prepayment, Tp, and time to default,
Td, are random variables that have continuous
probability distributions, f(tj), where tj is a
realization of Tj (jp,d). - The joint survivor function conditional on
time-varying covariates - where gjn(r,H,X) ? time-varying function of the
relevant interest rates, property values, loan
characteristics, borrower characteristics - Z ? macro-economic factors,
- ?p and ?d ? unobservable heterogeneity factors
181.3 Test for Adverse Selection
- If adverse selection based on unobserved risk
characteristics is present, then we should find a
significant relationship between initial LTV and
ex post default. - If adverse selection is not present, then we
should observe no systematic relationship between
initial LTV and default risk.
191.3 Competing Risks Model (Table 6)
- Independent Variables
- Borrower Characteristics
- Borrower risk (FICO and FICO2)
- Log(Income)
- Log (Borrower Age)
- Log (House Tenure)
- Debt-to-income ratio
- Contract Characteristics
- Lender LTV
- First or Second Lien position indicator
- Line or Loan indicator
- Use of funds indicator
- (refinance, consumption, home improvement)
- First mortgage indicator
- Second home indicator
- Condo indicator
- Auto pay
- Time-varying Option Characteristics
- Current LTV (CLTV and CLTV2)
201.3 Evidence of Adverse Selection (Table 6)
- Observable risk characteristics
- 100 point ? FICO ? default risks ? 43 (prepay ?
15) - Rate refinancing ? 3.7 less likely to default
(2.8 more likely to prepay) - No first mortgage ? 6.8 less likely to default
(3.1 less likely to prepay) - One percentage point higher DTI ? 2.1 more
likely to default (2.2 more likely to prepay) - ? current LTV (e.g., 1 house price depreciation)
? 4 more likely to default (1 less likely to
prepay) than borrowers whose current LTV ? (i.e.,
house price appreciation)
211.3 Evidence of Adverse Selection (Table 6)
- After controlling for the observable risk
characteristics, borrowers with higher initial
LTV contract (pledging less collateral per dollar
loan) are more likely to default. - Relative to borrowers with LTV 80, those with
80 lt LTV lt 90 are 2.2 more likely to default
(4.5 less likely to prepay) - Those with LTV ? 90 are 5.6 more likely to
default (6.6 less likely to prepay)
221.3 Evidence of Adverse Selection
- Conclusion
- Evidence consistent with the presence of adverse
selection on unobservables in the home equity
lending market (Stiglitz Weiss, 1981). - Evidence also consistent with findings of adverse
selection in the credit card market (Ausubel,
1999).
23Empirical Analysis
- Part II Secondary Screening
242.1 Lenders Counteroffer
- Factors that affect the lenders decision to make
one of the two counteroffers after the secondary
screening. - Counteroffer to further mitigate moral hazard
- if lender lowers LTV (increasing collateral
required per dollar loan to induce borrower
effort) and/or switches the product from a home
equity loan to a home equity line. - Counteroffer to further mitigate adverse
selection - if lender increases LTV and/or switches the
product from a home equity line-of-credit to a
home equity loan (increasing the APR to induce
borrower type). - Estimate a logit model to assess the likelihood
of a lender making a counteroffer designed to
mitigate adverse selection.
252.1 Adverse Selection Counter (Table 8)
- Higher risk borrowers less likely to receive
adverse selection counter offer. - Relative to borrower with a score of 800,
borrower with a FICO score of 700 is 24.6 less
likely to receive a counteroffer designed to
mitigate adverse selection than one designed to
mitigate moral hazard. - Borrowers who overvalue their property value
(relative to the banks estimated value) - One percentage point ? in the lenders LTV ratio
over the borrowers LTV ratio increases by 3.1
the probability that the lender counteroffers
with a contract designed to mitigate adverse
selection.
262.1 Adverse Selection Counter
- Conclusion
- Lender does systematically screen borrowers for
adverse selection and moral hazard.
272.2 Borrower response to counteroffer
- 2 Logit models of borrower response the
likelihood of a borrower rejecting a moral
hazard or adverse selection counteroffer. - Does secondary screen reintroduce adverse
selection? - Do low credit risk applicants reject counteroffer?
282.2. Moral hazard counteroffer (Table 10a)
- Each one percentage point decrease in the
counteroffer interest rate relative to the
original interest rate decreases the likelihood
of a borrower rejecting the moral hazard
counteroffer by 2.4. - If lender estimates a 10 percentage point higher
LTV than borrower, then likelihood of borrower
rejecting moral hazard counter increases by
0.65. - Indicates that counter offer introduces
additional adverse selection.
292.2. Adverse Selection Counter (Table 10b)
- Each one point increase in the counteroffer
interest rate over the original interest rate
increases the likelihood of a borrower rejecting
the counteroffer designed to mitigate adverse
selection by 1. - Less risky borrowers (lower FICO scores) more
likely to reject counter offer. - Results confirm that lenders mitigation efforts
introduce additional adverse selection.
302.3 Effectiveness of counteroffer (Table 11)
- Estimate a competing-risks hazard model
- Test the effectiveness of the lenders adverse
selection and moral hazard mitigation efforts - Sample
- Include all loans accepted following both the
primary and secondary screening - 83,411 borrowers
- 2 dummy variables identify
- Moral hazard counteroffer
- Adverse selection counteroffer
312.3 Effectiveness of counteroffer (Table 11)
- Relative to loans that did not receive additional
screening, - the risk of default ex post declines by 12.2
percent for loans that the lender ex ante
required additional collateral and/or switched
the contract from a home equity loan to a home
equity line. - Relative to loans that did not receive additional
screening, - the risk of default ex post declines by 4.2
percent for loans where the lender ex ante
reduced the required collateral and/or switched
the contract from a credit line to a home equity
loan.
322.3 Effectiveness of counteroffer (Table 11)
- Considerable difference in the marginal impact
- suggests that the lenders effort to mitigate
moral hazard ex ante is more effective than the
effort to mitigate adverse selection in reducing
the risk of default risk ex post. - consistent with lender being relatively more
successful in inducing additional borrower effort
ex post.
33Main Conclusions -- 1
- Borrowers choice of credit contract does reveal
information about her risk level. - Less credit-worthy borrowers are more likely to
select a contract requiring less collateral - Even after controlling for observable risk
characteristics, lender continues to face adverse
selection problems due to unobservable
information.
34 Main Conclusions -- 2
- Lenders efforts ex ante to mitigate adverse
selection and moral hazard can be effective in
reducing credit losses ex post. - Secondary screening and counteroffer designed to
mitigate moral hazard reduce default risk ex post
by 12. - Additional screening and counteroffer to mitigate
adverse selection reduce default risk ex post by
4.
35Main Conclusions -- 3
- Mitigation efforts impose costs (higher
prepayment rates) - Moral hazard mitigation increase the risk of
prepayment by 11. - Adverse selection mitigation increase the risk of
prepayment by 2.9. - Direct impact on secondary market investors and
their ability to predict prepayment speeds on a
securitized portfolio.