The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Do Changes in Mortgage Credit Constraints Explain the Housing Boom and Bust? Andra Ghent University of Wisconsin-Madison Tipping Points III Symposium Washington, DC October 12, 2018 Introduction Empirical Facts Model Parameterization Results Conclusions Introduction • Significant changes in mortgage credit and home ownership over past two decades • Causes of 2000-2007 housing boom still not completely understood • Effects of regulatory tightening in the residential mortgage market during bust unclear • Declines in home ownership hard to disentangle from changes in preferences for home ownership, changes in household formation, etc... 1 / 29 Introduction Empirical Facts Model Parameterization Results Conclusions This Paper 1. Summarize trends in home ownership and mortgage debt over past two decades 2. Present life cycle model that explores effect of relaxing and tightening mortgage credit constraint on home ownership and mortgage debt level • focus on change in maximum Loan-to-Value (LTV) household can take on to buy home Take homes: 1. Relaxation of LTV constraint cannot explain 2000-2007 boom period data • in data no increase in age-adjusted US home ownership during boom period 2. Tightening of LTV constraint can explain some of the decline in US home ownership in the bust period 2 / 29 US Home Ownership Rate 1994 - 2017 Aggregate and by Income Category Source: U.S. Census Bureau, Current Population Survey / Housing Vacancy Survey, July 26, 2018. Introduction Empirical Facts Model Parameterization Results Conclusions Aging of US Population Source: Age and Sex Composition: 2010. 2010 Census Briefs, U.S. Census Bureau. ⇒ Need to look at home ownership rates within age categories! 4 / 29 US Home Ownership by Age Category Source: U.S. Census Bureau, Current Population Survey / Housing Vacancy Survey, July 26, 2018. Introduction Empirical Facts Model Parameterization Results Conclusions Summary of Home Ownership Patterns 1. Increase in aggregate and age-adjusted home ownership rates 1994-2001 2. Slight decrease in age-adjusted home ownership rates 2001-2007 3. Significant decrease in aggregate and age-adjusted home ownership rates 2007-2017 6 / 29 US Real Mortgage Debt and Home Prices, 1994-2017 Sources: Federal Reserve Financial Accounts of the United States, FHFA, FRED, and BLS. Introduction Empirical Facts Model Parameterization Results Conclusions Can Rising Home Prices Explain Explosion in Debt Levels? Hypothetical Change in Mortgage Debt for Home Price Growth of 40% Assume: No increase in down payment size or income Home Price LTV Mortgage Debt Down Payment 2000 $ 300,000 80% $ 240,000 $ 60,000 2007 $ 420,000 86% $ 360,000 $ 60,000 Growth 40% 50% So, at most, ability to afford same home with rising prices can explain half the increase in mortgage debt Adelino, Schoar, and Severino (2018) actually find no change in CLTV ratios at origination so actual increase caused by decreasing affordability is likely much smaller 8 / 29 Growth in Mortgage Debt in Boom Broad-Based Source: Foote, Loewenstein, and Willen (2016). Growth in Nonprime Securities Issuance of MBS in $B Source: SIFMA. But average income for borrowers of mortgage in PLMBS pools >100,00$ (Ghent, Hernández-Murillo, and Owyang (2015)) • Subprime / alt-A was a middle-class phenomenon Take Aways 1. Housing boom period (2000-2007) saw no growth in home ownership rate other than through demographic change 2. More than doubling of stock of mortgage debt during boom • Explosion of Nonprime PLMBS market • Must be due to growth along the intensive margin given home ownership patterns • Likely a significant role for home equity extraction after mortgage origination during boom 3. Significant decline in non-demographic related homeownership in bust Introduction Empirical Facts Model Parameterization Results Conclusions Model Overview • Life cycle / OLG endowment economy • Households choose: • • • • • • tenure house size (if owners) mortgage type (if owners) whether to default consumption saving • Equilibrium mortgage rate for each mortgage type sets expected PV of mortgage equal to mortgage amount • Sources of risk: • home values (idiosyncratic) • income (idiosyncratic) 12 / 29 Introduction Empirical Facts Model Parameterization Results Conclusions Overview • Exogenous risk-free rate, r • Exogenous relative price of housing, q • Housing stock depreciates at a rate of δ every period • Home owners must pay δ every period in which they own to maintain the property • Financial intermediaries must pay a cost χ (percent of home value) to rehabilitate any home acquired through foreclosure 13 / 29 Households • Born at age 0 and live for at most J periods • start life with no assets and as renters • “Work” for the first JRET periods of life • Each period face a probability πj of dying • Bequest motive • Face stochastic income risk • income follows a Markov Process • If home owner, face stochastic home values Introduction Empirical Facts Model Parameterization Results Conclusions Households Tenure Choice • Each period, chooses whether to own or rent • If chooses to rent, no home size choice • rents a home of size h1 • If chooses to own, buys a home of size h2 (h2 > h1 ) or h3 (h3 > h2 ) • cannot buy a home of size h1 15 / 29 Introduction Empirical Facts Model Parameterization Results Conclusions Households Tenure Choice • Felicity depends on tenure • allow for the possibility that there is a utility premium from owning • Can transition in any period between owning and renting 16 / 29 Introduction Empirical Facts Model Parameterization Results Conclusions Households Home Values • Use same mechanism as Corbae and Quintin (2015) to capture home price volatility • Each period while an owner, there is a probability λ that the home will change in value • home of size h2 will stay size h2 with probability 1 − 2λ, will increase to size h3 with probability λ, and will decrease to size h1 with probability λ • home of size h3 will stay size h3 with probability 1 − λ and will decrease to size h2 with probability λ • owner-occupied home of size h1 will stay size h1 with probability 1 − λ and will increase to size h2 with probability λ • Rental homes do not change size (always size h1 ) 17 / 29 Introduction Empirical Facts Model Parameterization Results Conclusions Households Mortgage Choices • Two ways to finance home ownership: 1. Traditional Mortgages (TRADs): • require down payment of νTRAD % of the home value • term is T periods • payments are calculated such that the mortgage is fully amortizing over T periods • carry interest rate rTRAD 2. Low Down Payment (LDP) loans: • require down payment of just νLDP % of home value • term is T periods • payments are calculated such that the mortgage is fully amortizing over T periods • carry interest rate rLDP 18 / 29 Introduction Empirical Facts Model Parameterization Results Conclusions Households Mortgage Choices • No refinancing • keeps computation tractable • Can terminate the mortgage in any period by either • defaulting, or • prepaying • If defaults, loses the home and must rent for that period • Prepays by selling the home • pays selling cost ρ 19 / 29 Financial Intermediaries • Accepts household savings and makes mortgage loans • Earns the exogenously given rate r on savings • Pays a servicing cost, φ, on mortgages it holds • Holds a stock of housing capital which it can rent out at rate R per unit or sell to households as owner-occupied housing • Incurs the maintenance cost δ on its housing stock • Incurs a cost χqh of rehabilitating housing units it acquires through foreclosure • In equilibrium, it must make zero profits • =⇒ R = rq + δ Introduction Empirical Facts Model Parameterization Results Conclusions Equilibrium • Equilibrium mortgage interest rates, rTRAD and rLDP : • mortgage interest rate that makes the expected present value of the mortgage equal to the amount of the mortgage • lender discounts expected cash flows by r + φ • No closed form solution to this problem • Solve numerically: • inner loop solves household optimization problem for each value of state variable • outer loop for mortgage interest rates 21 / 29 Introduction Empirical Facts Model Parameterization Results Conclusions Parameterization Demographics • Period corresponds to 3 years • Household ‘born’ at age 25 • Household lives to at most 85 chronological years of age (J = 20) • Household ‘retires’ at age 64 (JRET = 13) • Survival rates taken from Arias et al. (2008) 22 / 29 Parameterization Income • Assume that the income process during working years follows an AR(1) process: yt = ρyt −1 + γ1 aget + γ2 aget2 + ε t (1) where ε t has variance σε2 • Estimate (1) using triennial PSID data on earnings from 1967 to 1992 • Approximate (1) with a three state Markov chain using the approach of Tauchen and Hussey (1991) • After retirement, labor income is set to 60% of income in the last working year following Cocco, Gomes, and Maenhout (2005) and Yao and Zhang (2005) Introduction Empirical Facts Model Parameterization Results Conclusions Parameterization Preferences: Felicity Function • Felicity function follows u (c, h, H ) = ψ ln c + (1 − ψ) ln h • Set ψ to 0.76 implying that renters spend 24% of their consumption expenditure on housing (Davis and Ortalo-Magné, 2011) 24 / 29 Introduction Empirical Facts Model Parameterization Results Conclusions Parameterization Housing Costs • Set χ, foreclosure rehabilitation costs, to 0.25 (consistent with Campbell, Giglio, and Pathak, 2011) • Set T , mortgage term, to 10 such that mortgages have 30 year terms • Set vTRAD = 0.2 such that TRADs require a 20% down payment • Set r , risk-free rate, to 0.12 • Set ρ, selling-costs, to 8% as in Cocco (2005) 25 / 29 Introduction Empirical Facts Model Parameterization Results Conclusions Parameterization Housing Costs • Use the following parameters to calibrate the model to certain characteristics in the data • • • • • λ: probability of an idiosyncratic house price shock house sizes, h1 , h2 , and h3 mortgage servicing cost, φ q: relative price of housing δ: per period depreciation rate on housing 26 / 29 Steady State Equilibria Moment Home Ownership Low Income Mid Income High Income Under 35 35-44 45-54 55-65 65+ Loan-to-Income Share LDPs TRAD Mtg Rate NDP Mtg Rate Avg. 30-year Mtg Rate Foreclosure Rate 2001 68.0% Data 2007 67.8% 2016 63.7% 42% 68% 76% 81% 81% 182% 41% 67% 75% 80% 80% 241% 35% 59% 70% 75% 80% 230% 4.97% 4.34% 1.65% no LDP 71.3% 30% 83% 92% 31% 70% 81% 86% 76% 208% 5.27% 5.27% Model νLDP = 0.1 71.5% 30% 83% 92% 31% 71% 81% 86% 76% 207% 2.2% 5.27% 5.93% 5.28% νLDP = 0.0 72.5% 30% 83% 98% 37% 70% 81% 86% 76% 199% 12.0% 5.27% 7.08% 5.33% 1.29% 1.31% 1.41% Notes: 1) Data sources are US Census CPS / Housing Vacancy Survey, Federal Reserve Consumer Finance Survey, and Federal Reserve Bank of St. Louis. Squaring the Model with the Data Relaxing LTV constraint raises home ownership rate for young, high-income households • HHs that cannot come up with a down payment but want to smooth consumption • Reduces average debt ratios slightly If relaxation of LTV constraint caused the boom, we would have seen an increase in the home ownership rate, especially young HHs Model is consistent with tightening of LTV constraint during bust causing a decline in home ownership • Consistent with empirical evidence of Duca and Rosenthal (1994) and Gete and Reher (forthcoming) regarding effect of credit constraints on home ownership Introduction Empirical Facts Model Parameterization Results Conclusions Empirical Facts: 1. No increase in age-adjusted home ownership rate over boom 2. Doubling of real residential mortgage debt during boom 3. Significant increase in home ownership rate in years leading up to the boom 4. Significant decline in home ownership rate during bust Model of tenure choice predicts that main effect of relaxation of LTV constraint is an increase in home ownership Main beneficiaries of relaxation of LTV constraints in model are high-income young households 29 / 29