Credit Regimes and the Seeds of Crisis

Transcription

1 Credit Regimes and the Seeds of Crisis Nelson Lind University of California, San Diego February 14, 2017

2 Figure 1-3. Nonprime Lending Soared in the 2000s Share of Subprime and Alt-A Mortgage Originations 40 Share of First-Lien Mortgage Originations (Percent) Subprime Alt-A Source: Inside Source: Mortgage Inside Mortgage Finance, 2009 Finance, Mortgage 2009 Market Mortgage Statistical Market Annual. Statistical Annual, Belsky and Richardson (2010).

3 Why the Non-Prime Mortgage Boom? This Paper: Theory of Rational Non-Prime Lending and Housing Cycles

4 Theory of Rational Non-Prime Lending and Housing Cycles Asymmetric information about riskiness of borrower income. Three endogenous market segments: Pooling no information produced Screening costly income verification Separating high down payments to separate types

5 Preview of Results 1 Falling income risk ùñ more pooling, less screening. 2 Switch from screening to pooling regime ùñ housing boom. Change in contract offered to marginal home buyer As in data, reduced down payments, interest rates, and documentation Likelihood depends on housing supply, screening technology, policy 3 Return to screening regime ùñ foreclosure crisis.

6 Contributions 1 Theory of non-prime lending and housing cycles Kiyotaki and Moore (1997), Bernanke et al. (1999), Iacoviello (2005), Dell Ariccia and Marquez (2006), Gorton and Ordoñez (2014) 2 Regime changes drive turning points, not shocks. Reduce shocks: Justiniano et al. (2015), Kaplan et al. (2016) Non-linear solution method: Lind (2015) 3 Mechanism consistent with and may reconcile micro-evidence Mian and Sufi (2015), Ferreira and Gyourko (2015), Adelino et al. (2016), Albanesi et al. (2016), Justiniano et al. (2016) 4 Framework for policy/counterfactual analysis

7 Road Map 1 Mortgage Market Equilibrium 2 Joint Mortgage and Housing Market Equilibrium 3 Equilibrium Dynamics 4 Cross-sectional Predictions

8 A Model of the Mortgage Market Two periods: t 0, 1 Borrowers, i P r0, 1s: Utility: Two types: upc i,0, C i,1 q C i,0 ` βc i,1. τ i # 0 if safe 1 if risky. Risky borrowers have no income with chance: φ PrY i 0 τ i 1s. Competitive Lenders: Free entry, risk neutral, exogenous cost of funds of 1 ` r ă 1{β Screening technology: learn borrower type with cost κ

9 Information X i, borrower i observables Credit history, employment status, etc. Lender s inference (Bayes rule): Prτ i 1 X i xs f X px τ i 1qPrτ i 1s f X pxq ρpxq. Among borrowers with ρpx i q ρ, a fraction ρ are risky. Summary statistic (risk index): ρ i ρpx i q.

10 Housing Identical homes Must own or rent: P period 0 price, R rent Idiosyncratic housing risk: Z i period 1 home value iid F A1. For hpzq F 1 pzq 1 F pzq, zhpzq is strictly increasing and continuous in z. A2. The scaled hazard rate gets sufficiently large: 1 p1 ` rqβ lim zhpzq ą zñ8 p1 φqλ.

11 Foreclosure Technology 45 Payoff L Lender Borrower L 1 λ Ex-post home value: Z Foreclosure payoffs given debt of L i : Lender: Borrower: mintl i, p1 λqz i u maxtp1 λqz i L i, 0u Fraction λ of home value lost in foreclosure

12 Mortgage Contracts No state-contingent contracts A contract specifies: 1 a period 0 transfer from the lender to borrower of T. 2 a period 1 payment of L by the borrower to the lender. 3 if screening occurs. Foreclosure threat incentivizes repayment.

13 Timing Period 0 Period 1 Borrowers Apply to a contract Home values and borrower incomes realized Default or pay Lenders Offers given ρ i ρpx i q Accept or reject applications Initiate foreclosure

14 Equilibrium There are many equilibria. Focus on stable equilibria surviving standard selection criteria: Kohlberg and Mertens (1986), Cho and Kreps (1987), Cho and Sobel (1990) Proposition (Existence and Uniqueness) Let A1 and A2 hold. Suppose that borrowers apply for screening contracts whenever indifferent. Then, for almost every ρ P r0, 1s, there exists a stable Nash equilibrium that is unique up to the identity of lenders making offers. In this equilibrium: 1 Lenders make zero profits. 2 If safe borrowers use a contract, then it maximizes their valuation.

15 Equilibrium Foreclosure and Default Lenders use foreclosure following any default. Borrowers strategically default if Z i ă L i. Additionally, risky borrowers default when Y i 0 (with chance φ). Ex-post default rate if L i L, ρ i ρ, and Z i Z: DpL, ρ; Zq 1tZ ă Lu ` 1tZ ě Luρφ

16 Anticipated Payoffs Lender expected profit: ΠpL, ρq L lomon Face value T ` 1 ΠpL, ρq 1 ` r ı E DpL, ρ; Z i q looooooooooooomooooooooooooon maxtl p1 λqz i, 0u Default losses Borrower expected utility: UpL, τq T P ` βupl, τq, τ EZ i L looomooon Expected equity # 0 if safe 1 if risky ı ` E DpL, τ; Z i q maxtl loooooooooomoooooooooon Z i, λz i u Net default payoff

17 Equilibrium Contracts

18 Equilibrium Contracts Contract Surplus: V Risk Index: ρ share risky observables

19 Equilibrium Contracts V sep 1 pp Rq max L 1`r ΠpL, 0q ` βupl, 0q Contract Surplus: V 1 s.t. 1`r ΠpL, 0q ` βupl, 1q ď P R picq Risk Index: ρ

20 Equilibrium Contracts V pool 1 pρq max L 1`r ΠpL, ρq ` βupl, 0q Contract Surplus: V Risk Index: ρ

21 Equilibrium Contracts V scrn 1 pρq max L 1`r ΠpL, 0q ` βupl, 0q κ 1 ρ Contract Surplus: V Risk Index: ρ

22 Equilibrium Contracts Contract Surplus: V V scrn Risk Index: ρ

23 Equilibrium Contracts Contract Surplus: V V scrn V pool Risk Index: ρ

24 Equilibrium Contracts Contract Surplus: V V scrn V sep V pool Risk Index: ρ

25 Equilibrium Contracts: High Income Risk Pooling Screening Separating Contract Surplus: V V scrn V sep V pool ρ L ρ Risk Index: ρ

26 Equilibrium Contracts: Low Income Risk Pooling Screening Separating Contract Surplus: V V scrn V sep V pool ρ L ρ Risk Index: ρ

27 Summary Ó income risk ùñ less screening, more pooling. More pooling ùñ more low-documentation loans Documentation fell during the boom: Jiang et al. (2014). But, equilibrium depends on home prices: V sep pp Rq.

28 Joint Housing and Mortgage Market Equilibrium V sep pp Rq max L s.t. 1 ΠpL, 0q ` βupl, 0q 1 ` r 1 ΠpL, 0q ` βupl, 1q ď P R picq 1 ` r Size of separating segment linked to home price through IC constraint. Equilibrium in the mortgage market ùñ housing demand curve.

29 Safe Borrower Surplus: V Risk Index: ρ Net Home Price: P R Homeownership Rate: H

30 V sep V safe ρ Risk Index: ρ P R V Homeownership Rate: H

31 V sep V safe ρ Risk Index: ρ V safe P R V H d N safe Homeownership Rate: H

32 V sep ρ Risk Index: ρ V safe V P R N safe H d Homeownership Rate: H

33 V sep ρ L Risk Index: ρ ρ V safe V P R N safe H d H pool Homeownership Rate: H

34 V sep ρ L Risk Index: ρ ρ V safe V V risky P R N safe H pool Homeownership Rate: H

35 V sep 0 ρ L Pooling Regime Risk Index: ρ ρ Screening Regime P 0 R H s Homeownership Rate: H

36 V sep 0 ρ L Risk Index: ρ Pooling Regime ρ Screening Regime P 0 R H s Homeownership Rate: H

37 V sep 0 ρ L Risk Index: ρ Pooling Regime ρ Screening Regime P 1 R P 0 R H s Homeownership Rate: H

38 V sep 1 V sep 0 ρ Risk Index: ρ Pooling Regime Screening Regime P 1 R P 0 R H s Homeownership Rate: H

39 Summary Ó income risk can trigger pooling regime when marginal home buyer gets pooled with safe borrowers Simultaneous Boom in home price Relaxation of down-payment requirements Elimination of screening (income documentation) requirements Automated Underwriting

40 Equilibrium Dynamics Extend to infinite horizon setting Sale value of home proportional to future home price Three risk levels: ρit P t0, ρ L, ρ H u Construction firms build new homes. Absentee landlords supply apartments (elastic rental supply). Solve with regime-switching perturbation, Lind (2015)

41 Simulation: Responses to Fall in Income Risk Risky Chance of No Income (%) Credit Regime Screen Pool Parameterization % Home Price Non-Prime Share (%) % Homeownership Foreclosure Rate (%)

42 Summary Turning points of cycle due to regime changes, not shocks. Mean reversion ùñ credit-fueled boom ends in a foreclosure crisis Reduce the number of shocks: Justiniano et al. (2015), Kaplan et al. (2016) Conclusion

43 Cross-sectional Predictions Debt Allocation Mian and Sufi (2015), Albanesi et al. (2016) Mortgage Rates Justiniano et al. (2016)

44 Simulation: Debt Allocation % in Debt by Borrower Type % in Debt by Market Segment Safe Risky Zero Risk Low Risk High Risk Consistent with and may reconcile micro-evidence. Detail Mian and Sufi (2015), Albanesi et al. (2016)

45 Simulation: Mortgage Rates Mortgage APR by Borrower Type Mortgage APR by Market Segment Safe Risky Zero Risk Low Risk High Risk Explanation for the mortgage rate conundrum. Detail Justiniano et al. (2016)

46 Summary Consistent with and may reconcile evidence on debt growth. An explanation for the mortgage rate conundrum. Also, consistent with rise of foreclosures across borrowers. Graphs Ferreira and Gyourko (2015), Albanesi et al. (2016)

47 Conclusion Theory of rational non-prime lending and housing cycles Fall in income risk ùñ switch to pooling ùñ housing boom. Mean reversion ùñ return to screening ùñ foreclosure crisis. Endogenous regimes drive turning points, not shocks. Non-linear solution method: Lind (2015) Mechanism consistent with and may reconcile micro-evidence Framework to assess policy Tradeoff between distortions and reducing crises Dodd-Frank Optimal policy? Extend to include costly recession due to crisis.

48 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements The Role of Computerization Adoption of automated underwriting between 1995 and Application of artificial intelligence and statistical decision making 60%-70% automation by Straka (2000) Key implication: fall in application time cost. Gates et al. (2002)

49 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements Manual vs. Automated Underwriting: High Income Risk Pooling Screening V auto Contract Surplus: V V sep V manual V pool ρ manual ρåuto Risk Index: ρ

50 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements Manual vs. Automated Underwriting: Low Income Risk Pooling Screening V auto Contract Surplus: V V sep V pool V manual ρ manual ρåuto Risk Index: ρ

51 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements V safe Net Home Price: P R V risky low V risky med H d high H d med H d low V risky high Homeownership Rate: H Figure: Housing demand with manual and automated screening as income risk varies. Solid lines correspond to demand with automated screening while dashed lines show demand with manual screening.

52 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements Summary Manual underwriting ùñ high application time cost Risky borrowers do not apply for screening contracts. Pooling segment less sensitive to income risk. Automated underwriting ùñ low application time cost Risky borrowers apply for screening contracts. Pooling segment more sensitive ùñ non-prime booms more likely. Return

53 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements Parameterization Return ln φ t p1 ρ φ 1 ρφ 2 qµφ ` ρ φ 1 ln φ t 1 ` ρ φ 2 ln φ t 2 ` σ φ ε φ t Z it log normal with mean 1 δ and s.d. σ z lp p1 αq{α t H new t Parameter Value Target/Source λ 0.27 Campbell et al. (2011) δ % annual depreciation (Piazzesi and Schneider (2016)) σ z annual s.d. of home prices (Piazzesi and Schneider (2016)) β.95 Iacoviello (2005) r lnp1{.99q Iacoviello (2005) R 1 normalize rental rate to 1 exppµ φ q 11.1% 6% s.s. spread on risky contract κ % s.s. spread on screening contract in segment L ρ L 8.35% 0.75% s.s. spread on pooling contract in segment L ρ H 47.4% 1% s.s. spread on screening contract in segment H n L, n H 29.3%, 70.1% 1% s.s. share pooling and 1% s.s. share risky borrowers l 3.52 ˆ % s.s. homeownership rate 1 α α 1.5 median supply elasticity in Saiz (2010) ρ φ ρ φ 2.63 σ φ.05

54 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements Debt Allocation Seemingly conflicting evidence from credit bureau data Mian and Sufi (2015) Albanesi et al. (2016)

55 ebt: MortgageBalances Debt Allocation Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements 1999 ranking overstates debt growth for first quartile (subprime) 8Q lagged Equifax Riskscore 1999 Equifax Riskscore Ratio to 2001Q1 (3QMA) Quartile 1 (Lowest) Quartile 2 Quartile 3 Quartile 4 (Highest) Ratio to 2001Q1 (3QMA) Quartile 1 (Lowest) Quartile 2 Quartile 3 Quartile 4 (Highest) Mortgage balances. Ratio to 2001Q1. Figure: Mortgage Source: Authors balancescalculations in FRBNYbased CCP/Equifax on FRBNYData. CCP/Equifax Ratio todata. 2001Q1. Source: Albanesi et al. (2016)

56 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements Connecting Theory and Evidence Mian and Sufi (2015) measure debt by time invariant factors (τ i ). Albanesi et al. (2016) measure debt by observable factors (ρ i ).

57 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements Equilibrium Debt Levels Pooling Screening Separating Pooling Separating Contract Surplus: V V scrn V sep V pool V scrn V sep V pool ρ L ρ ρ Contract Transfer: T Risk Index: ρ Risk Index: ρ Figure: Debt (period 0 transfers) in screening (left) and pooling (right) regimes.

58 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements Summary Distinction between types (τ i ) and lender observables (ρ i q. May reconcile the microeconomic evidence. Following switch from screening to pooling: Average debt of risky borrowers increases. Average debt of safe borrowers may increase or decrease. Lenders allocate more debt to observably low risk borrowers. Return

59 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements The Mortgage Rate Conundrum: Justiniano et al. (2016) Summer of 2003, labor market recovers for recession of Fed announces end of easing cycle (June 2003). Long-term treasuries begin to rise. Historically, mortgage rates rose with treasury yields. But this time, they did not.

60 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements Residual Variation in THEMortgage MORTGAGE RATERates CONUNDRUM Percent Jan-00 Jan-01 Jan-02 Jan-03 Jan-04 Jan-05 Jan-06 Jan (I) Purchases and refis (II) Purchases (III) Refis (IV) Fixed rate (V) Fixed rate, purchases Figure 4.1. Average residuals in the five specifications of table 2. Figure: Residuals from r it c ` ft 1 1 β ` x it 1 γ ` other controls ` ε it. Source: Justiniano et al. (2016)

61 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements Equilibrium Mortgage Rates Pooling Screening Separating Pooling Separating Contract Surplus: V V auto V sep V pool V auto V sep V pool ρ L ρ ρ Mortgage Rate: r L{T 1 Risk Index: ρ Risk Index: ρ Figure: Mortgage rates in screening (left) and pooling (right) regimes.

62 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements Summary Recovery of labor market (fall in income risk) explains Rise in treasury yields (endogenous policy response) Switch from screening to pooling regime Shift in credit regime implies Non-prime lending and housing boom Fall in mortgage rates relative to treasury yields Return

63 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements Simulation: Foreclosure Rates Return Foreclosure Rate by Borrower Type Foreclosure Rate by Market Segment Safe Risky Zero Risk Low Risk High Risk Consistent with rise in foreclosures across all borrowers/segments. Ferreira and Gyourko (2015), Albanesi et al. (2016)

64 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements Policy: Dodd-Frank Act of 2010 Ability-to-repay requirements Must verify borrower ability to repay Monetary damages for violations. Qualifying mortgages ùñ protection from ex-post damages Non-qualifying mortgages (NQM) Incentive for minimal documentation Model as a secondary screening option

65 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements Minimal Screening Technology Low cost, imperfect screening κnqm ă κ ζ Prfail screening riskys Acceptance rate: 1 ρ lomon safe passing ` p1 ζqρ looomooon risky passing 1 ζρ Value to safe borrowers: V NQM pρq max L where f pρq p1 ζqρ 1 ζρ 1 κnqm ΠpL, f pρqq ` βupl, 1q 1 ` r 1 ζρ ă ρ is the share risky after screening.

66 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements Inefficient NQM Screening A3. For each ρ P r0, 1s, V NQM pρq ă maxtv pool pρq, V scrn pρqu. NQM only used after introduction of ATR requirements

67 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements Ineffective NQM Acts Like Pooling Contract Surplus: V V scrn V NQM Ineffective NQM High Income Risk V pool V scrn Ineffective NQM Low Income Risk V pool V NQM Contract Surplus: V V NQM V scrn Effective NQM High Income Risk V pool V NQM Effective NQM Low Income Risk V scrn V pool

68 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements Effective NQM Dampens Sensitivity to Income Risk NQM Screening Separating NQM Separating Contract Surplus: V V scrn Ineffective NQM High Income Risk V sep V NQM V scrn Ineffective NQM Low Income Risk V sep V NQM ρ L ρ ρ NQM Screening Separating NQM Screening Separating Contract Surplus: V V scrn V NQM Effective NQM High Income Risk V sep V scrn Effective NQM Low Income Risk V sep V NQM ρ L ρ ρ L ρ

69 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements Summary Minimum screening requirements Dampen sensitivity of mortgage market to income risk Therefore dampen sensitivity of housing demand And reduce likelihood of boom-bust cycles Welfare reducing in this model Extend to incorporate cost of crisis (recession) ùñ policy tradeoff. Return

70 Automation Parameterization Debt Allocation Mortgage Rates Foreclosure Rates Screening Requirements M. Adelino, A. Schoar, and F. Severino. Loan originations and defaults in the mortgage crisis: The role of the middle class. Review of Financial Studies, page hhw018, S. Albanesi, G. De Giorgi, J. Nosal, and M. Ploenzke. Credit growth and the financial crisis: A new narrative. Technical report, Mimeo, Ohio State University, E. Belsky and N. Richardson. Understanding the boom and bust in nonprime mortgage lending. Technical report, Harvard Joint Center for Housing Studies, B. S. Bernanke, M. Gertler, and S. Gilchrist. The financial accelerator in a quantitative business cycle framework. Handbook of macroeconomics, 1: , J. Y. Campbell, S. Giglio, and P. Pathak. Forced sales and house prices. The American Economic Review, 101(5): , I.-K. Cho and D. M. Kreps. Signaling games and stable equilibria. The Quarterly Journal of Economics, pages , I.-K. Cho and J. Sobel. Strategic stability and uniqueness in signaling games. Journal of Economic Theory, 50(2): , G. Dell Ariccia and R. Marquez. Lending booms and lending standards. The Journal of Finance, 61(5): , F. Ferreira and J. Gyourko. A new look at the us foreclosure crisis: Panel data evidence of prime and subprime borrowers from 1997 to Technical report, National Bureau of Economic Research, S. W. Gates, V. G. Perry, and P. M. Zorn. Automated underwriting in mortgage lending: good news for the underserved? Housing Policy Debate, 13(2): , G. Gorton and G. Ordoñez. Collateral crises. The American Economic Review, 104(2): , M. Iacoviello. House prices, borrowing constraints, and monetary policy in the business cycle. American economic review, pages , W. Jiang, A. A. Nelson, and E. Vytlacil. Liar s loan? effects of origination channel and information falsification on mortgage delinquency. Review of Economics and Statistics, 96(1):1 18, A. Justiniano, G. E. Primiceri, and A. Tambalotti. Credit supply and the housing boom. Technical report, A. Justiniano, G. E. Primiceri, and A. Tambalotti. The mortgage rate conundrum. Technical report, G. Kaplan, K. Mitman, and G. Violante. Consumption and house prices in the great recession. Technical report, N. Kiyotaki and J. Moore. Credit cycles. The Journal of Political Economy, 105(2): , E. Kohlberg and J.-F. Mertens. On the strategic stability of equilibria. Econometrica, 54(5): , N. Lind. Regime-switching perturbation for non-linear equilibrium models. Working Paper, University of California, San Diego, A. Mian and A. Sufi. Household debt and defaults from 2000 to 2010: Facts from credit bureau data. Technical report, National Bureau of Economic Research, M. Piazzesi and M. Schneider. Housing and macroeconomics. Handbook of Macroeconomics, A. Saiz. The geographic determinants of housing supply. The Quarterly Journal of Economics, 125(3): , J. W. Straka. A shift in the mortgage landscape: The 1990s move to automated credit evaluations. Journal of Housing research, 11(2):207, 2000.