Consumption and House Prices in the Great Recession: Model Meets Evidence

Consumption and House Prices in the Great Recession: Model Meets Evidence Greg Kaplan Kurt Mitman Gianluca Violante MFM 9-10 March, 2017

Outline 1. Overview 2. Model 3. Questions Q1: What shock(s) drove the boom-bust in p h? Q2: How does the fall in p h transmit to C? Q3: Could a debt-forgiveness policy have cushioned the bust? 4. Further evidence 5. Conclusions 6. Appendix 0 Kaplan, Moll and Violante (2017)

Three questions 1. What shock(s) drove the boom-bust in p h? Expectations about future growth in p h Credit conditions important for homeownership, leverage and foreclosure 1 Kaplan, Moll and Violante (2017)

Three questions 1. What shock(s) drove the boom-bust in p h? Expectations about future growth in p h Credit conditions important for homeownership, leverage and foreclosure 2. How does the fall in p H transmit to C? Mostly a wealth effect, not collateral effect 1 Kaplan, Moll and Violante (2017)

Methodology Model: aggregate shocks move equilibrium p h 2 Kaplan, Moll and Violante (2017)

Methodology Model: aggregate shocks move equilibrium p h Parameterize: match cross-sectional and lifecycle micro data 2 Kaplan, Moll and Violante (2017)

Methodology Model: aggregate shocks move equilibrium p h Parameterize: match cross-sectional and lifecycle micro data Simulate boom-bust Compare with aggregate time-series data House prices Consumption Rent-price ratio Home ownership Leverage Foreclosures Compare against micro data 2 Kaplan, Moll and Violante (2017)

Model Demographics OLG lifecycle economy with work & retirement Endowments Workers face uninsurable risk in individual earnings y Preferences Utility over nondurable c and housing services h Housing Finite number of house sizes h H Households can buy a unit of h at price p h, or rent it at rate ρ Linear transaction cost κ h (p h h) for sellers 3 Kaplan, Moll and Violante (2017)

Financial instruments Liquid saving (b > 0): one-period bond, exogenous interest rate r b (fixed) 4 Kaplan, Moll and Violante (2017)

Financial instruments Liquid saving (b > 0): one-period bond, exogenous interest rate r b (fixed) Mortgages (m): long-term, fixed-rate debt contract Price schedule q j (h, m, b, y) set by competitive banking sector Amortized over remaining lifetime at rate r b (1 + ι) Refinancing option available (cash-out) at cost κ m Max Loan-to-Value at origination only m λ m p h h Max Payment-to-Income at origination only π λ π y 4 Kaplan, Moll and Violante (2017)

Closing the model Final good sector Y = Z N w = Z Construction sector Labor + housing permits aggregate housing investments I(p h ) Rental sector Buys housing from sellers and rents them out, or vice-versa, sells rental units to home buyers Operating cost ψ per unit of housing owned and rented out Zero-profit condition yields equilibrium rental rate ρ Government Taxes workers (with mortgage interest deduction) and properties, sells land permits, and pays SS benefits to retirees 5 Kaplan, Moll and Violante (2017)

Aggregate shocks 1. Aggregate labor income: Z 2. Credit conditions: (i) credit limits (λ m, λ b, λ π ) (ii) intermediation wedge ι 6 Kaplan, Moll and Violante (2017)

Aggregate shocks 1. Aggregate labor income: Z 2. Credit conditions: (i) credit limits (λ m, λ b, λ π ) (ii) intermediation wedge ι 3. Beliefs / News about future housing demand: Three regimes for ϕ (share of housing services in u): (a) ϕ L : low housing share and unlikely transition to ϕ H (b) ϕ L : low housing share and likely transition to ϕ H (c) ϕ H : high housing share Boom-Bust: shift from (a) to (b), and back to (a) 6 Kaplan, Moll and Violante (2017)

Shock Processes 1. Aggr. labor income: NIPA wages & salaries per capita 2. Credit conditions: λ m : 95% 110%, λ b : 20% 30% λ π : 25% 60%, ι m : 100 BP 75 BP 3. Beliefs: Case-Shiller-Thompson & Burnside-Eichenbaum-Rebelo Realized path for shocks 1.04 Productivity, Z 1.1 Financial Deregulation, 6 m 0.07 Expected Price Growth 1.02 1.05 0.06 1 1.0 0.05 0.04 0.98 0.95 0.03 0.96 2000 2005 2010 2015 Year 0.9 2000 2005 2010 2015 Year 0.02 2000 2005 2010 2015 Year The shift in beliefs hits in 2001 and reverts back in 2007 7 Kaplan, Moll and Violante (2017)

Consumption and house price dynamics House Price Consumption 1.1 1.3 1.2 1.05 1.1 1 1 0.9 0.8 Model Data 2000 2005 2010 2015 Year 0.95 2000 2005 2010 2015 Year 8 Kaplan, Moll and Violante (2017)

Consumption and house price dynamics House Price Consumption 1.1 1.3 1.2 1.05 1.1 1 1 0.9 0.8 Benchmark Belief Only 2000 2005 2010 2015 Year 0.95 2000 2005 2010 2015 Year 8 Kaplan, Moll and Violante (2017)

Consumption and house price dynamics House Price Consumption 1.1 1.3 1.2 1.05 1.1 1 1 0.9 0.8 Benchmark Belief Only Income Only 2000 2005 2010 2015 Year 0.95 2000 2005 2010 2015 Year 8 Kaplan, Moll and Violante (2017)

Consumption and house price dynamics House Price Consumption 1.1 1.3 1.2 1.05 1.1 1 0.9 0.8 Benchmark Belief Only Income Only Credit Only 2000 2005 2010 2015 Year 1 0.95 2000 2005 2010 2015 Year 8 Kaplan, Moll and Violante (2017)

Beliefs vs actual change in preferences House Price Consumption 1.1 1.3 1.05 1.2 1.1 1 1 0.9 0.8 Benchmark Demand Only 2000 2005 2010 2015 Year 0.95 0.9 2000 2005 2010 2015 Year Preference shock: similar rise in p h, but C falls! 9 Kaplan, Moll and Violante (2017)

Dynamics of rent-price ratio 1.1 1 0.9 0.8 0.7 Benchmark 2000 2005 2010 2015 Year 10 Kaplan, Moll and Violante (2017)

Dynamics of rent-price ratio 1.1 1 0.9 0.8 0.7 Benchmark Belief Only Income Only Credit Only 2000 2005 2010 2015 Year ( ) 1 δh τ h ρ = ψ + p h 1 + r b E ph [p h] Belief about future appreciation shared by investment company 10 Kaplan, Moll and Violante (2017)

Dynamics of home ownership 1.1 Benchmark 1.05 1 0.95 2000 2005 2010 2015 Year 11 Kaplan, Moll and Violante (2017)

Dynamics of home ownership 1.1 1.05 Benchmark Belief Only Income Only Credit Only 1 0.95 2000 2005 2010 2015 Year Loosening of credit limits drives rise in home-ownership Households constrained in tenure choice, not in housing choice 11 Kaplan, Moll and Violante (2017)

Dynamics of leverage and foreclosure 1.8 Leverage 0.04 1.6 0.03 1.4 0.02 1.2 1 0.01 Foreclosure rate Benchmark 0.8 2000 2005 2010 2015 Year 0 2000 2005 2010 2015 Year 12 Kaplan, Moll and Violante (2017)

Dynamics of leverage and foreclosure 1.8 1.6 1.4 1.2 1 Leverage 0.04 0.03 0.02 0.01 Foreclosure rate Benchmark Belief Only Income Only Credit Only 0.8 2000 2005 2010 2015 Year 0 2000 2005 2010 2015 Year Credit loosening is key for constant leverage pre-boom Interaction between beliefs and credit important for foreclosure 12 Kaplan, Moll and Violante (2017)

Why credit shock does not affect p h Max LTV/PTI ratios affect housing demand if renters (extensive margin) or home-owners (intensive margin) are constrained in housing choice (not tenure choice) 1. BOOM: Rental market relaxes these constraints 2. BUST: Long-term mortgage debt relaxes these constraints Are we missing the credit supply aspect of the shock, i.e. cheap credit flowing to low-quality borrowers? No: endogenous relaxation in lending standards in response to belief-driven boom 13 Kaplan, Moll and Violante (2017)

Cheaper credit for low-quality borrowers Endogenous Borrowing Rate (1/q m -1) 0.14 0.12 No shocks Credit Only All Shocks 0.1 0.08 0.06 0.04 0.7 0.8 0.9 1 1.1 Leverage Lenders also expect prices to rise and default rates to fall 14 Kaplan, Moll and Violante (2017)

Deleveraging or wealth effect in the bust? -.3 -.2 -.1 0.1 Change in Log Consumption Renters Owners 0.05.1.15 Debt as a Fraction of Total Wealth -.3 -.2 -.1 0.1 Change in Log Consumption Renters Owners 0.1.2.3.4 Housing Share of Total Wealth Deleveraging: WEAK Wealth effect: STRONG 15 Kaplan, Moll and Violante (2017)

Counterfactual principal reduction program All homeowners with LTV >95%: forgive excess debt 16 Kaplan, Moll and Violante (2017)

Counterfactual principal reduction program All homeowners with LTV >95%: forgive excess debt House Price 1.1 Consumption 1.2 1 0.8 1.6 1.4 1.2 1 0.8 Bench. Mod. 2000 2005 2010 2015 Year Leverage 2000 2005 2010 2015 Year 1.05 1 0.95 0.03 0.02 0.01 0 2000 2005 2010 2015 Year Foreclosure rate 2000 2005 2010 2015 Year Beneficiaries account for small share of C 16 Kaplan, Moll and Violante (2017)

Credit growth Mian-Sufi: credit growth concentrated in low-income groups Foote et al.: no, equally distributed across income groups 17 Kaplan, Moll and Violante (2017)

Credit growth Mian-Sufi: credit growth concentrated in low-income groups Foote et al.: no, equally distributed across income groups Share of Debt 0.1.2.3.4.5 Shares of Mortgage Debt 2001 2007 1 2 3 4 5 Income Quintile of Household Low-income hh switch from rent to buy, high-income hh upsize 17 Kaplan, Moll and Violante (2017)

Moretgage origination Mian-Sufi: mortgage origin. concentrated in subprime groups Adelino et al.: no, equally distributed across groups 18 Kaplan, Moll and Violante (2017)

Moretgage origination Mian-Sufi: mortgage origin. concentrated in subprime groups Adelino et al.: no, equally distributed across groups Share of Debt 0.2.4.6.8 Shares of Originated Mortgage Debt 2001 2007 Above Median Below Median Default Risk Young hh switch from rent to buy, older hh upsize 18 Kaplan, Moll and Violante (2017)

Foreclosures Mian-Sufi: foreclosures concentrated in subprime groups Albanesi et al.: no, proportionally rising more for other groups 19 Kaplan, Moll and Violante (2017)

Foreclosures Mian-Sufi: foreclosures concentrated in subprime groups Albanesi et al.: no, proportionally rising more for other groups Share of Foreclosures 0.2.4.6.8 1 2004 2006 2008 2010 2012 Year IncomeQuintile=1 IncomeQuintile=2 IncomeQuintile=3 IncomeQuintile=4 IncomeQuintile=5 Everyone levers up, including middle-income households 19 Kaplan, Moll and Violante (2017)

What did we learn from the model? 1. Shift in expected house appreciation key to boom-bust in p H 2. This explanation is consistent with recent micro evidence 3. Endogenous relaxation of credit conditions from change in beliefs 4. Credit important for home-ownership, leverage, foreclosures, but not p H 5. p h transmits to C through wealth effects 6. Principal reduction program would not have mitigated drop in C 20 Kaplan, Moll and Violante (2017)

Change in home ownership by age Log-change (relative to mean) Log-change (relative to mean) 0.15 Boom 0.05 Bust 0.1 0 0.05 Data Model -0.05 Data Model 0-0.1-0.05 30 40 50 60 Age -0.15 30 40 50 60 Age It s the young who go in/out of housing market 22 Kaplan, Moll and Violante (2017)

Shock to Interest Rate House Price ND Consumption 1.12 1.3 1.1 1.08 1.2 1.06 1.1 1.04 1.02 1 1 0.9 0.98 0.96 0.8 2000 2005 2010 2015 Years 2000 2005 2010 2015 Years 23 Kaplan, Moll and Violante (2017)

Consumption response by age during Bust Annual Consumption Growth by Age Bust (%) -1.3-1.4-1.5-1.6-1.7-1.8-1.9-2 25 30 35 40 45 50 55 60 65 70 75 80 Age c in the baseline - c in the Income-only counterfactual 24 Kaplan, Moll and Violante (2017)

Parameterization strategy Parameter values disciplined by facts from household-level micro-data Distributional stats: mortgages, housing wealth, renters, and consumption Moment Empirical value Model Value Fraction homeowners w/ mortgage 0.66 0.57 Aggr. mortgage debt / housing value 0.42 0.36 P10 LTV ratio for mortgagors 0.15 0.28 P90 LTV ratio for mortgagors 0.92 0.76 Aggr. home-ownership rate 0.66 0.65 P10 Housing NW / total NW for owners 0.11 0.12 P90 Housing NW / total NW for owners 0.95 0.98 Avg.-size owned house / rented 1.5 1.4 Avg. earnings owners / renters 2.05 2.02 BPP consumption insurance coef 0.36 0.43 25 Kaplan, Moll and Violante (2017)

Solution and simulation Equilibrium computed with a version of Krusell-Smith (1998) Forecasting rule used by households in their problem: log p h = a 0 (Z, Z ) + a 1 (Z, Z ) log p h Aggregate consistency: in equilibrium, forecasting rule is also law of motion for prices Note: ρ computable from zero-profit condition, given p h and E[p h ] 26 Kaplan, Moll and Violante (2017)

Solution and simulation Simulation of boom-bust: realized path for shocks 1.04 Productivity, Z 1.1 Financial Deregulation, 6 m 0.07 Expected Price Growth 1.02 1.05 0.06 1 1.0 0.05 0.04 0.98 0.95 0.03 0.96 2000 2005 2010 2015 Year 0.9 2000 2005 2010 2015 Year 0.02 2000 2005 2010 2015 Year 1. Productivity: aggregate earnings data 2. Credit conditions: max LTV: 85% - 100%, HELOC limit: 20% - 30%, origination costs: 1% - 0 3. Beliefs: expected house price growth from Case-Shiller survey 27 Kaplan, Moll and Violante (2017)

Household problem: Renter A non-homeowner can stay a renter or become an owner: V n (b j, z j ; Ω) = max {V r (b j, z j ; Ω), V o (b j, z j ; Ω)}, where Ω denotes the vector of aggregate states (Z, µ) Those who choose to rent solve: V r (b j, z j ; Ω) = s.t. max u j (c j, s j ) + βe zj,ω [V n (b j+1, z j+1 ; Ω )] c j,h j,b j+1 c j + ρ(ω)h j + q b b j+1 b j + y j T (y j, 0) b j+1 0 s j = h j H z j+1 = Υ(z j ) Ω = Γ (Ω) 28 Kaplan, Moll and Violante (2017)

Household problem: Buyer Those who choose to buy and become owners solve: [ V o (b j, z j ; Ω) = max u j (c j, s j ) + βe zj,ω V h (x j+1, z j+1 ; Ω ) ] c j,b j+1,h j+1,m j+1 s.t. c j + q b b j+1 + p h (Ω)h j+1 + κ m b j + y j T (y j, 0) + q m (x j+1, z j ; Ω)m j+1 m j+1 λ m p h (Ω)h j+1 b j+1 0 h j+1 H, s j = ωh j+1 z j+1 = Υ(z j ), Ω = Γ (Ω) where x j+1 := (b j+1, h j+1, m j+1 ) 29 Kaplan, Moll and Violante (2017)

Household problem: Homeowner V h (x j, z j ; Ω) = max Pay: V p (x j, z j ; Ω) Refinance: V f (x j, z j ; Ω) Sell: V n ( b j, z j ; Ω) Default: V d (b j, z j ; Ω) where x j := (b j, h j, m j ) Seller s liquid assets after transaction: b j = b j κ h p h (Ω)h j (1 + r m ) m j + (1 δ h τ h ) p h (Ω)h j 30 Kaplan, Moll and Violante (2017)

Household problem: Homeowner A household that makes its mortgage payment solves : [ V p (x j, z j ; Ω) = max u(c j, s j ) + βe zj,ω V h (x j+1, z j+1 ; Ω ) ] c j,b j+1,π m s.t. c j + q b (b)b j+1 + (δ h + τ h ) p h (Ω)h j + π m b j + y j T (y j, m j ) π m π m m j+1 = (1 + r m ) m j π m b j+1 λ b p h (Ω)h j+1 s j = ωh j, z j+1 = Υ(z j ), where x j := (b j, h j, m j ) h j+1 = h j Ω = Γ (Ω) Note: Collateral effect for owners only through HELOCs 31 Kaplan, Moll and Violante (2017)

Household problem: Default V d (b j, z j ; Ω) = s.t. max u(c j, s j ) ξ + βe zj,ω [V r (b j+1, z j+1 ; Ω )] c j,h j,b j+1 c j + ρ (Ω) h j + q b b j+1 b j + y j T (y j, 0) b j+1 0 s j = h j z j+1 = Υ(z j ), Ω = Γ (Ω) Disutility of default ξ Household must rent for a period Return 32 Kaplan, Moll and Violante (2017)

Mortgage pricing Zero-profit condition by type j, x = (b, h, m), z yields: qm(x j+1, z j ; Ω) = 1 {[ E (1 + rm) m zj,ω g n ] j+1 + gf j+1 (1 + rm) m j+1 j+1 + g j+1 d ( min 1 δ h d ) p h (Ω )h j+1, (1 + rm) m j+1 [ + 1 g j+1 n ] gf j+1 gd j+1 [πm(x j+2, z j+1 ; Ω ) + qm(x j+2, z j+1 ; Ω ]} )m j+2 g n : sell g f : refinance g d : default g n = g f = g d = 0 make mortgage payment 33 Kaplan, Moll and Violante (2017)

Rental company Rental company owns housing units and rents them out to hh It can buy/sell units frictionlessly on the housing market H ( ) 1 [ ρ H + E 1 + r b Ω J( H ; Ω ) ] J( H; Ω) = [ max ψ H p h H (1 δ h τ h ) H ] + Optimization implies the equilibrium rental rate: ( ) 1 δh τ h ρ = p h + ψ 1 + r b E Ω [p h (Ω )] Return 34 Kaplan, Moll and Violante (2017)