SLIDE 1
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
Housing Finance, Boom-Bust Episodes, and the Macroeconomy Carlos - - PowerPoint PPT Presentation
Housing Finance, Boom-Bust Episodes, and the Macroeconomy Carlos - - PowerPoint PPT Presentation
I NTRODUCTION T HE M ODEL C ALIBRATION R ESULTS Conclusions Housing Finance, Boom-Bust Episodes, and the Macroeconomy Carlos Garriga Federal Reserve Bank of St. Louis Aaron Hedlund University of Missouri RBNZ Housing Conference 2017 The
SLIDE 2
SLIDE 3
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
HOUSING MARKETS AND CREDIT IN THE U.S.
◮ During the Great Depression, the housing market collapsed as homeowners could not rollover their short-term low LTV loans due to the failure of local banks. ◮ The introduction of high LTV long-term mortgages and nationwide credit were important drivers of the postwar housing boom (Chambers, Garriga, and Schlagenhauf 2012). ◮ 2000s boom was fueled by a credit expansion and declining cost of borrowing.
2000 2005 2010 2015
Real House Prices
1 1.1 1.2 1.3 1.4 1.5 2000 2005 2010 2015
Real Mortgage Debt
1 1.2 1.4 1.6 1.8 2 2000 2005 2010 2015
Real Risk-Free Rate
- 2
- 1
1 2 3 4 2000 2005 2010 2015
Real 30-Year FRM Rate
1 2 3 4 5 6
SLIDE 4
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
QUESTIONS AND METHODOLOGY
Question: How do credit and mortgage arrangements impact the dynamics of housing and the macroeconomy? ◮ Construct a quantitative macro model with extensive and intensive margins for housing and mortgage borrowing. ◮ Analyze the role of credit vs. income-productivity applied to the 2000s housing boom-bust. ◮ Assess the aggregate and cross-sectional implications of mortgage structure. ◮ Evaluate the impact of macroprudential policies targeted at mortgage design.
SLIDE 5
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
TAKEAWAYS
◮ Anatomy of the boom: credit booms have a bigger impact
- n housing and consumption than does income growth.
◮ Asymmetric macro effects: consumption responds more strongly to house price movements during the bust than during the boom. ◮ Mortgage Structure
◮ Refinancing and Equity Extraction: the ability to refinance and extract equity substantially magnifies housing booms and the spillovers to consumption. ◮ Mortgage Duration: long-term debt mitigates rollover risk and has a substantial impact on foreclosure, ownership, and consumption dynamics.
◮ Macroprudential Regulations: Tighter LTV constraints blunt the boom and mitigate the bust. PTI constraints are not as effective at reducing endogenous fragility, but they increase homeownership during the boom.
SLIDE 6
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
MODEL SUMMARY
Households ◮ Preferences E0 ∞
t=0 βtu(ct, ch,t). Own h ∈ H with ch = h or
consume apartment space ch = a ∈ [0, a], a ≤ h. Labor efficiency e · s with cdf F(e) and transitions πs(s′|s). Technology ◮ Consumption Yc = zcNc. New housing Yh = Fh(L, Sh, Nh). Apartments produced using linear, reversible technology. Long-Term, Defaultable Mortgage Debt ◮ Extensive vs. intensive margins of credit. FRMs vs. ARMs. ◮ Long-term contracts with default, prepayment, and the ability to extract equity. Decentralized Housing Market ◮ Search frictions endogenize housing illiquidity.
SLIDE 7
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
HOUSEHOLD PORTFOLIO CHOICE
New buyers/homeowners who refinance:
VR
- wn(y, (rm, m), h, s, 0) =
max
m′,b′,c≥0 u(c, h) + βE
- (Wown + Rsell)(y′, (rm, m′), h, s′, 0)
- subject to
c + γp(h) + qbb′ + m ≤ y + q0
m((rm, m′), b′, h, s)m′
q0
m((rm, m′), b′, h, s)m′ ≤ ϑp(h)
y′ = we′s′ + b′
Homeowners who make a regular payment:
VC
- wn(y, (rm, m), h, s, 0) = max
l,b′,c≥0 u(c, h) + βE
- (Wown + Rsell)(y′, (rm, m′), h, s′, 0)
- subject to
c + γp(h) + qbb′ + l ≤ y l ≥ rm 1 + rm m m′ = (m − l)(1 + rm) y′ = we′s′ + b′
SLIDE 8
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
MORTGAGE FINANCING
◮ Key features: endogenous default premia, prepayment, equity extraction through costly refinancing. ◮ For ARMs, rm = rm adjusts every period. ◮ Mortgage prices satisfy the recursive relationship:
q0
m((rm, m′), b′, h, s)m′ =
1 (1 + ζ)(1 + rm) E
sell + repay
- ηs(θs(p′
s, h))m′ + no sale (do not try/fail)
- [1 − ηs(θs(p′
s, h))]
× d′ ϕ min
- JREO(h), m′
- default + repossession
+ (1 − ϕ)
- no repossession
(1 + ζ)q0
m((rm, m′), b′′, h, s′)m′
- continuation value with m′
+(1 − d′) m′1[Refi] + 1[No Refi] l
- payment
+ (1 + ζ)q0
m((rm, m′′), b′′, h, s′)m′′
- continuation with m′′ = (m′ − l)(1 + rm)
SLIDE 9
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
CALIBRATION
◮ Calibrate the economy to the late 1990s. ◮ Important to match the LTV distribution (especially the right tail), homeownership rate, and foreclosures.
Description Parameter Value Target Model Source/Reason Homeownership Rate a 2.005 67.0% 67.2% Census Housing Wealth (Owners) ω 0.8177 2.49 2.49 1998 SCF Median Borrower LTV 62.90% 65.51% 1998 SCF Borrowers with LTV ≥ 70% 40.00% 43.43% 1998 SCF Borrowers with LTV ≥ 80% β 0.9657 25.0% 24.2% 1998 SCF Borrowers with LTV ≥ 90% 14.50% 11.27% 1998 SCF Borrowers with LTV ≥ 95% 9.20% 7.97% 1998 SCF Median Owner Liq. Assets/Earn 0.16 0.15 1998 SCF Foreclosure Starts (Annual) γs 0.6550 1.60% 1.87% Nat’l Delinquency Survey Full Calibration Details
SLIDE 10
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
ANATOMY OF THE HOUSING BOOM
Time (years)
1 2 3 4 5
House Prices
1 1.1 1.2 1.3 1.4 1.5 Credit + Productivity Productivity Only
Time (years)
1 2 3 4 5
Ownership Rate (%)
64 65 66 67 68 69 Credit + Productivity Productivity Only
Time (years)
1 2 3 4 5
Consumption
1 1.1 1.2 1.3 1.4 1.5 Credit + Productivity Productivity Only
◮ Credit booms are much larger than productivity booms. ◮ Consumption increases much more during a credit boom, partly because of housing as an ATM. ◮ The credit boom has a minimal impact on the ownership rate because of the equilibrium increase in house prices.
SLIDE 11
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
CREDIT AND THE “NEW NARRATIVE”
Low Income Middle Income High Income Average Borrower LTV Pre-Boom 59.3% 61.3% 70.3% Productivity Only 56.4% 58.9% 57.1% Productivity + Credit 60.9% 65.8% 69.3% ∆Credit +4.5% +6.9% +12.2% High-LTV Share∗ Pre-Boom 13.9% 14.6% 36.3% Productivity + Credit 16.7% 22.7% 31.1% Consumption Change Productivity Only 4.8% 4.2% 1.3% Productivity + Credit 6.0% 11.7% 13.3% ∆Credit +1.2% +7.5% +12.0%
◮ Broad-based increase in borrowing and consumption from the credit expansion. ◮ Ownership shifts toward larger houses (13% move up with productivity alone vs. 22% when credit also expands).
SLIDE 12
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
THE BOOM, BUST, AND RECOVERY
Time (years)
2 4 6 8 10
House Prices
0.9 1 1.1 1.2 1.3 1.4
Time (years)
2 4 6 8 10
Average Time on Market
10 20 30 40 50
Time (years)
2 4 6 8 10
Annual Foreclosure Rate
0.01 0.02 0.03 0.04
Time (years)
2 4 6 8 10
Ownership Rate
0.64 0.65 0.66 0.67 0.68
Time (years)
2 4 6 8 10
Consumption
0.9 0.95 1 1.05 1.1
Time (years)
2 4 6 8 10
Median Borrower Leverage
0.5 0.6 0.7 0.8 0.9 1
∆Pricesboom ∆Cboom Ownboom ∆Pricesbust ∆Cbust Ownbust Model +44.6% +12.2% 68.1% −24.5% −18.5% 64.3% Data +41.9% +5.1% 69.2% −25.9% −15.0% 64.2%
◮ Downside labor market uncertainty and tighter credit are key to generating the bust.
SLIDE 13
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
ASYMMETRIC BALANCE SHEET EFFECTS
Time (years)
1 2 3 4
House Prices (Boom)
1 1.1 1.2 1.3 1.4 1.5 Boom Fixed ph
Time (years)
1 2 3 4
Consumption (Boom)
1 1.05 1.1 1.15 1.2 Boom Fixed ph
Time (years)
1 2
C-ph Elasticity (Boom)
0.05 0.1 0.15 0.2 0.25 0.3
Time (years)
5 6 7 8 9
House Prices (Bust)
0.5 0.6 0.7 0.8 0.9 1 Bust Fixed ph
Time (years)
5 6 7 8 9
Consumption (Bust)
0.8 0.85 0.9 0.95 1 Bust Fixed ph
Time (years)
5 6 7
C-ph Elasticity (Bust)
0.05 0.1 0.15 0.2 0.25 0.3
◮ Consumption dynamics are driven largely by the ability to withdraw equity from the household balance sheet. ◮ Debt overhang in the bust makes balance sheet effects much larger than during the boom.
SLIDE 14
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
ASYMMETRIC HOMEOWNERSHIP DYNAMICS
◮ The credit channel is highly nonlinear because of the foreclosure double trigger. ◮ Also, fluctuations in housing liquidity are asymmetric. ◮ Models without an extensive margin (own vs. rent, default
- vs. pay) miss this channel.
Time (years)
1 2 3 4 5
Ownership Rate (Boom)
0.66 0.68 0.7 0.72 0.74 0.76 0.78 Boom Boom; Fixed ph
Time (years)
5 6 7 8 9 10
Ownership Rate (Bust)
0.56 0.58 0.6 0.62 0.64 0.66 0.68 Bust Bust; Fixed ph
SLIDE 15
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
EQUITY EXTRACTION AND REFINANCING
Time (years)
1 2 3 4
House Prices
1 1.05 1.1 1.15 1.2 1.25 1.3 1.35 1.4 1.45 1.5 FRM FRM No Refi
Time (years)
1 2 3 4
Consumption
1 1.02 1.04 1.06 1.08 1.1 1.12 1.14 1.16 1.18 1.2 FRM FRM No Refi
Time (years)
1 2 3 4
Consumption Change (%)
2 4 6 8 10 12 14 16 18 20 FRM, 0 < LTV < 50 FRM No Refi, 0 < LTV < 50
Time (years)
1 2 3 4
Consumption Change (%)
2 4 6 8 10 12 14 16 18 20 FRM, LTV > 80 FRM No Refi, LTV > 80
◮ Without the ability to refinance, the house price boom is 40% smaller and exhibits less overshooting. ◮ When houses can’t be used as ATMs, the spillover to consumption is smaller and more gradual. ◮ The difference in consumption dynamics is most stark for highly leveraged owners.
SLIDE 16
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
FIXED-RATE VS. ADJUSTABLE RATE MORTGAGES
Time (years)
4 6 8 10 12
Ownership Rate
0.63 0.64 0.65 0.66 0.67 0.68 0.69 FRM ARM
Time (years)
4 6 8 10 12
Annual Foreclosure Rate
0.02 0.04 0.06 0.08 0.1 FRM ARM
Time (years)
4 6 8 10 12
Median Borrower Leverage
0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 FRM ARM
◮ FRMs and ARMs exhibit the same house price boom as long as refinancing is possible. Without refinancing, the ARM boom is 30% and the FRM boom is 27%. ◮ The ARM economy is more sensitive to interest rate movements during the bust and recovery.
SLIDE 17
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
FIXED-RATE VS. ADJUSTABLE RATE MORTGAGES
Time (years)
5 5.5 6 6.5 7
Consumption Change (%)
- 30
- 25
- 20
- 15
- 10
- 5
Housing Bust
FRM, Renters ARM, Renters
Time (years)
5 5.5 6 6.5 7
Consumption Change (%)
- 30
- 25
- 20
- 15
- 10
- 5
Housing Bust
FRM, Owners ARM, Owners
Time (years)
5 5.5 6 6.5 7
Consumption Change (%)
- 30
- 25
- 20
- 15
- 10
- 5
Housing Bust
FRM, 0 < LTV < 50 ARM, 0 < LTV < 50
Time (years)
5 5.5 6 6.5 7
Consumption Change (%)
- 30
- 25
- 20
- 15
- 10
- 5
Housing Bust
FRM, LTV > 80 ARM, LTV > 80
Time (years)
7 8 9 10 11
Consumption Change (%)
5 10 15 20
Recovery
FRM, Renters ARM, Renters
Time (years)
7 8 9 10 11
Consumption Change (%)
5 10 15 20
Recovery
FRM, Owners ARM, Owners
Time (years)
7 8 9 10 11
Consumption Change (%)
5 10 15 20
Recovery
FRM, 0 < LTV < 50 ARM, 0 < LTV < 50
Time (years)
7 8 9 10 11
Consumption Change (%)
5 10 15 20
Recovery
FRM, LTV > 80 ARM, LTV > 80
◮ Consumption is more sensitive to interest rates in the ARM economy, particularly among highly leveraged owners.
SLIDE 18
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
MORTGAGE DURATION AND ROLLOVER RISK
Time (years)
4 6 8 10 12
House Prices
0.8 0.9 1 1.1 1.2 1.3 FRM 1-Period Loan
Time (years)
4 6 8 10 12
Sales Rate
0.5 1 1.5 2 2.5 3 FRM 1-Period Loan
Time (years)
4 6 8 10 12
Ownership Rate
0.5 0.55 0.6 0.65 0.7 FRM 1-Period Loan
Time (years)
4 6 8 10 12
Average Time on Market
10 20 30 40 50 60 FRM 1-Period Loan
Time (years)
4 6 8 10 12
Annual Foreclosure Rate
0.2 0.4 0.6 0.8 1 FRM 1-Period Loan
Time (years)
4 6 8 10 12
Consumption
0.8 0.85 0.9 0.95 1 1.05 1.1 FRM 1-Period Loan
Time (years)
4 6 8 10 12
Median Borrower Leverage
0.6 0.8 1 1.2 1.4 FRM 1-Period Loan
Time (years)
4 6 8 10 12
Outstanding Debt
1 1.2 1.4 1.6 1.8 2 FRM 1-Period Loan
◮ Mortgage duration has almost no impact on housing dynamics during the boom. ◮ During the bust, house prices are unaffected but
- wnership, foreclosures, and consumption respond more
severely.
SLIDE 19
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
MORTGAGE DURATION AND ROLLOVER RISK
Time (years) 5 6 7 8 9 Consumption Change (%)
- 30
- 25
- 20
- 15
- 10
- 5
FRM, Renters 1-Period Loan, Renters
Time (years) 5 6 7 8 9 Consumption Change (%)
- 30
- 25
- 20
- 15
- 10
- 5
FRM, Owners 1-Period Loan, Owners
Time (years) 5 6 7 8 9 Consumption Change (%)
- 30
- 25
- 20
- 15
- 10
- 5
FRM, 0 < LTV < 50 1-Period Loan, 0 < LTV < 50
Time (years) 5 6 7 8 9 Consumption Change (%)
- 60
- 50
- 40
- 30
- 20
- 10
FRM, LTV > 80 1-Period Loan, LTV > 80
◮ For homeowners with equity, there is little rollover risk during the bust. ◮ Highly leveraged owners experience a consumption disaster with short-term debt.
SLIDE 20
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
MACROPRUDENTIAL POLICY: LTV CAPS
Time (years)
2 4 6 8 10
House Prices
0.9 1 1.1 1.2 1.3 1.4 1.5 FRM FRM Tight LTV
Time (years)
2 4 6 8 10
Sales Rate
0.5 1 1.5 2 2.5 FRM FRM Tight LTV
Time (years)
2 4 6 8 10
Ownership Rate
0.64 0.65 0.66 0.67 0.68 0.69 FRM FRM Tight LTV
Time (years)
2 4 6 8 10
Average Time on Market
10 20 30 40 50 FRM FRM Tight LTV
Time (years)
2 4 6 8 10
Annual Foreclosure Rate
0.005 0.01 0.015 0.02 0.025 0.03 0.035 FRM FRM Tight LTV
Time (years)
2 4 6 8 10
Consumption
0.85 0.9 0.95 1 1.05 1.1 FRM FRM Tight LTV
Time (years)
2 4 6 8 10
Median Borrower Leverage
0.5 0.6 0.7 0.8 0.9 1 FRM FRM Tight LTV
Time (years)
2 4 6 8 10
Outstanding Debt
1 1.1 1.2 1.3 1.4 1.5 1.6 FRM FRM Tight LTV
◮ LTV caps substantially shrink the size of housing booms without crowding out homeownership. ◮ They also decrease the endogenous fragility of the economy and reduce the size of busts.
SLIDE 21
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
MACROPRUDENTIAL POLICY: PTI CAPS
Time (years)
2 4 6 8 10
House Prices
0.9 1 1.1 1.2 1.3 1.4 1.5 FRM FRM PTI
Time (years)
2 4 6 8 10
Sales Rate
0.5 1 1.5 2 2.5 FRM FRM PTI
Time (years)
2 4 6 8 10
Ownership Rate
0.64 0.65 0.66 0.67 0.68 0.69 0.7 FRM FRM PTI
Time (years)
2 4 6 8 10
Average Time on Market
10 20 30 40 50 FRM FRM PTI
Time (years)
2 4 6 8 10
Annual Foreclosure Rate
0.005 0.01 0.015 0.02 0.025 0.03 0.035 FRM FRM PTI
Time (years)
2 4 6 8 10
Consumption
0.85 0.9 0.95 1 1.05 1.1 FRM FRM PTI
Time (years)
2 4 6 8 10
Median Borrower Leverage
0.5 0.6 0.7 0.8 0.9 1 FRM FRM PTI
Time (years)
2 4 6 8 10
Outstanding Debt
1 1.1 1.2 1.3 1.4 1.5 1.6 FRM FRM PTI
◮ PTI caps are less effective at reducing house price booms but have an even larger positive impact on ownership. ◮ However, PTI limits do not reduce endogenous fragility and can exacerbate house price declines during a bust.
SLIDE 22
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
CONCLUSIONS
◮ Credit booms are significantly larger than productivity booms. ◮ The transmission of house prices to consumption is stronger during busts when equity extraction becomes more difficult. ◮ More heavily-indebted economies are more fragile with respect to negative shocks. ◮ Mortgage design has significant aggregate and cross-sectional effects, particularly during busts.
SLIDE 23
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
THE MODEL: FRICTIONAL HOUSING MARKET
◮ Importance of endogenous housing liquidity explored in Garriga and Hedlund (2016). ◮ Sellers choose list price ps and sell w/prob ηs(θs(ps, h)). ◮ Dynamic sorting problem simplified by brokers ⇒ block recursivity: θs, θb do not depend directly on distribution Φ.
ηs(θs(ps, h; Φ)) =
- p(Φ)h−ps
κsh
- γs
1−γs
ηb(θb(pb, h; Φ)) =
- pb−p(Φ)h
κbh
- γb
1−γb
◮ Equilibrium determination of sufficient statistic p(Φ):
- h∗ηb(θb(p∗
b , h∗; p))dΦrent =
Yh(p)
new housing
+ SREO(p)
REO housing
+
- hηs(θs(p∗
s , h; p))dΦown
- sold by owner
Back to Seller’s Problem
SLIDE 24
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
CALIBRATION I
◮ Calibrate the economy to match the cross-section of leverage in 1998, plus other key housing statistics.
Description Parameter Value Source/Reason Independent Parameters Autocorrelation ρ 0.952 Storesletten et al (2004) SD of Persistent Shock σǫ 0.17 Storesletten et al (2004) SD of Transitory Shock σe 0.49 Storesletten et al (2004) IES ν 0.13 Flavin and Nakagawa (2008) Risk Aversion σ 2 Standard Structure Share αS 30% Favilukis et al. (2016) Land Share αL 33% Lincoln Inst Land Policy Holding Costs η 0.7% Moody’s Depreciation (Annual) δh 1.4% BEA Rent-Price Ratio (Annual) rh 5% Sommer et al. (2013) Risk-Free Rate (Annual) r 1.0% Federal Reserve Board Servicing Cost (Annual) φ 3.1% 3.2% Real Mortgage Rate Mortgage Origination Cost ζ 0.4% FHFA Maximum LTV ϑ 125% Fannie Mae
- Prob. of Repossession
ϕ 0.5 2008 OCC Mortgage Metrics Credit Flag Persistence λf 0.9500 Fannie Mae
SLIDE 25
INTRODUCTION THE MODEL CALIBRATION RESULTS Conclusions
CALIBRATION II
Description Parameter Value Target Model Source/Reason Jointly Determined Parameters Homeownership Rate a 2.005 67.0% 67.2% Census Starter House Value h1 2.4250 1.75 1.75 American Housing Survey Housing Wealth (Owners) ω 0.8177 2.49 2.49 1998 SCF Borrowers with LTV ≥ 80% β 0.9657 25.0% 24.2% 1998 SCF Months of Supply∗ ξ 0.0016 5.40 5.42 Nat’l Assoc of Realtors
- Avg. Buyer Search (Weeks)
γb 0.0940 10.00 9.95 Nat’l Assoc of Realtors Maximum Bid Premium κb 0.0171 2.5% 2.5% Gruber and Martin (2003) Maximum List Discount κs 0.1029 15% 15% RealtyTrac Foreclosure Discount χ 0.0980 21% 21% Pennington-Cross (2006) Foreclosure Starts (Annual) γs 0.6550 1.60% 1.87% Nat’l Delinquency Survey Model Fit Median Borrower LTV 62.90% 65.51% 1998 SCF Borrowers with LTV ≥ 70% 40.00% 43.43% 1998 SCF Borrowers with LTV ≥ 90% 14.50% 11.27% 1998 SCF Borrowers with LTV ≥ 95% 9.20% 7.97% 1998 SCF Median Owner Liq. Assets/Earn 0.16 0.15 1998 SCF Calibration Summary