Data Some Loans Are More Equal Than Others: Mortgage Duration Model Third-Party Originations and Defaults Pricing in the Subprime Mortgage Industry Home Page Title Page Scott D. Grimshaw Grant R. McQueen Barrett A. Slade Brigham Young University ◭◭ ◮◮ William P. Alexander ◭ ◮ Capital One Page 1 of 15 Interface 2003 Go Back Full Screen Close Published 2002 in Real Estate Economics, 30, pp. 667-697 Quit
Outline • Subprime Industry Data • Difference Between Mortgage Duration Model Retail and Third Party Originated Loans Pricing Home Page • Borrower Repayment Behavior Title Page • Mortgage Duration Model ◭◭ ◮◮ • More Controls or Price the Risk? ◭ ◮ Page 2 of 15 Go Back Full Screen Close Quit
1. Data Data Subprime Industry Background Mortgage Duration Model • Loans to Individuals with Pricing Troubled or Thin Credit Home Page • Loans are 2 to 7% Higher than Title Page Conforming Rates ◭◭ ◮◮ • 1995 Originations $18 billion ◭ ◮ 1997 Originations $66 billion Page 3 of 15 • Loan Portfolios Often Funded via Securitization Go Back • Subprime Industry Problems Full Screen Close Quit
Agency Costs • Lenders Obtain Loans From Several Sources Data • Retail: Mortgage Duration Model Pricing Consumer Deals Directly With Company that Home Page Intends to Hold or Securitize Loan and Acts as Title Page Underwriter ◭◭ ◮◮ • Third Party Originators (TPOs): ◭ ◮ More Concerned About Volume, Which Generates Fees from Borrower and/or Page 4 of 15 Premiums from Lender Go Back • Key Distinction: Full Screen Retail Bear Prepayment and Default Risk Close TPOs Don’t Quit
• LaCour-Little & Chun (1999) show TPOs have Incentives to ‘Churn’ Borrower Data • TPOs have Incentive to Game Underwriting Mortgage Duration Model Process, Resulting in Higher Defaults Pricing • Passive Gaming: Insufficient Underwriting Home Page Title Page • Active Gaming: Exaggerate Credit Worthiness or Home Value ◭◭ ◮◮ • Is This Fraud? ◭ ◮ Page 5 of 15 – First Payment Defaults – Even Loans Vigorously Underwritten Go Back May Default Due to Economic Downturns, Full Screen Unemployment, Housing Deflation Close Quit
Competing Risks • Borrower Repayment Behavior Over Life of Loan Data • Each Month, Borrower Chooses to: Mortgage Duration Model Pricing – Make Monthly Payment Home Page – Pay Outstanding Balance Title Page – Default ◭◭ ◮◮ • Is This Fraud? ◭ ◮ Ability-to-Pay Views Home Ownership as a Consumption Good and Borrowers Default Page 6 of 15 When They Can No Longer Make Payments Go Back Full Screen Close Quit
• Or Rational Investing? Home Ownership is an Investment Where Data Mortgage Contains: Mortgage Duration Model – Call Option Pricing Borrower Prepays and Calls in the Old Loan Home Page When Market Interest Rate Falls Below Title Page Contract Rate ◭◭ ◮◮ – Put Option ◭ ◮ Borrower Defaults and Puts the House to the Page 7 of 15 Lien Holder When Home Value Drops Below Loan Value Go Back Full Screen Close Quit
2. Mortgage Duration Model Data • Sample 23,200 Loans Originated or Acquired by Mortgage Duration Model a National Subprime Mortgage-Lending Firm Pricing • Individual First-Lien Fixed-Rate Loans Secured Home Page by Residential Real Estate Originated Between Title Page 1 Jan 1996 and 31 Dec 1998 ◭◭ ◮◮ • Follow Loans through Sep 2000 ◭ ◮ For any given loan, observe t = min( t d , t p , c ) and Page 8 of 15 the type of termination where Go Back t d = mortgage duration until default Full Screen t p = mortgage duration until prepayment Close c = observed length due to censoring Quit
Joint Survival Function S ( t d , t p | X d ( t d ) , X p ( t p ) , θ d , θ p , β d , β p ) Data = P[ T d > t d , T p > t p | X d ( t d ) , X p ( t p ) , θ d , θ p , β d , β p ] Mortgage Duration Model t d � Pricing � exp( α dn + β ′ = exp − θ d d X d ( t d )) Home Page n =1 Title Page t p � exp( α pn + β ′ − θ p p X p ( t p )) ◭◭ ◮◮ n =1 ◭ ◮ where Page 9 of 15 X d ( t d ) and X p ( t p ) Possibly Time-Varying Covariates Go Back β d and β p Corresponding Parameters Full Screen θ d and θ p Allow Correlated Risks Close Quit
Unobserved Frailty • Two Groups With Different Willingness to Data Exercise Option: Ruthless and Woodhead Mortgage Duration Model • Frailty: For Ruthless, ( θ dR , θ pR ) Pricing and For Woodhead, ( θ dW , θ pW ) Home Page Title Page • θ dW < θ dR ⇒ Facing Similar Mortgage Situations, the Ruthless Group is More Likely to ◭◭ ◮◮ Exercise Default Option than Woodhead Group ◭ ◮ • Mixture Model with Page 10 of 15 γ R = P[ Ruthless ] γ W = P[ Woodhead ] Go Back Maximum Likelihood Estimation of Full Screen β d , β p , θ dR , θ pR , θ dW , θ pW , γ R , γ W Close Quit
Default: Covariate Hazard Ratio Data TPO 1.2776 Mortgage Duration Model Put 2.1561 Pricing Put if Put > 1 3.7716 Home Page Call 0.9220 A- Grade 0.7320 Title Page B+ Grade 1.0000 ◭◭ ◮◮ B Grade 1.3558 ◭ ◮ B- or Lower Grade 2.0534 Page 11 of 15 30 Year Term 1.4134 Loan Amount ($1000) 0.9987 Go Back %∆ Employment 0.9647 Full Screen θ dW =0.0002 and ˆ ˆ θ dR =0.0033 Close γ R =0.62 and ˆ ˆ γ W =0.38 Quit
Prepayment: Covariate Hazard Ratio Data TPO 0.9538 Mortgage Duration Model Put 0.4193 Pricing Call 1.0694 Home Page Call if Call > 0 1.1539 Title Page A- Grade 0.9475 B+ Grade 1.0000 ◭◭ ◮◮ B Grade 1.0613 ◭ ◮ B- or Lower Grade 1.2498 Page 12 of 15 30 Year Term 1.2536 Loan Amount ($1000) 1.0022 Go Back θ dW =0.0017 and ˆ ˆ θ dR =0.0207 Full Screen γ R =0.62 and ˆ ˆ γ W =0.38 Close Quit
3. Pricing Data • TPO Loan is 27.76% More Likely to Default Mortgage Duration Model Than a Similar Retail Loan Pricing • Create More Controls or Write an Home Page Incentive-Efficient and Enforceable Contract Title Page with TPO Requires Directly Connecting the TPO Actions to Loan Default ◭◭ ◮◮ • In Efficient Markets, Investors Price Risk ◭ ◮ Page 13 of 15 • As Managers Learn About TPO Default Risk, Expect Them to Price Default Risk by Charging Go Back Higher Interest Rates to TPO Generated Loans Full Screen Close Quit
Subprime Risk Premium: Difference in Loan Contract Rate and 10 Year Treasury Data Explanatory Variable Coefficient TPOE=1 if TPO Originated 1 Jan 1996 - 28 Feb 1997 -2.769 Mortgage Duration Model TPOL=1 if TPO Originated After 1 Mar 1997 49.459 Pricing A- Grade -8.322 B+ Grade 0.0000 Home Page B Grade 94.096 B- or Lower Grade 163.295 Title Page LTV for A- & B+ Grade 0.204 30 Year Term 1.942 ◭◭ ◮◮ Loan Amount ($1000) -0.001 ◭ ◮ Test for TPOE=TPOL, F =525.8, p -value < 0.0001 Page 14 of 15 As Subprime Industry Matured and TPO Risk Go Back Discovered, A 50-Basis-Point Premium Was Full Screen Charged to Compensate for Extra Risk Close Quit
Summary • Are All Loans Created Equal? Data • No: Loans That Appear Equal (Similar Loan Mortgage Duration Model Grade, Term, Amount, Option Incentives) Are Pricing Not Equal in their Subsequent Performance Home Page Title Page • Yes: Market Became Efficient With Respect to Agency Risk, So Investors are Indifferent ◭◭ ◮◮ Between Retail and TPO Loans ◭ ◮ Page 15 of 15 Go Back Full Screen Close Quit
Recommend
More recommend
