Overborrowing, Financial Crises and Macro-prudential Policy Javier Bianchi University of Wisconsin Enrique G. Mendoza University of Maryland & NBER
The case for macro-prudential policies Credit booms are often followed by deep recessions, asset price crashes, and banking crises Credit booms occurred with 2.8% frequency in 1960-2010, and about 1/3 featured banking crises (Mendoza & Terrones (12)) in this sense the 2008-09 global crisis had a typical pattern. Macro-prudential policy (MPP) has a clear goal: to prevent overborrowing at the macro level by affecting agents behavior ex ante but specifics of MPP design are less clear Theoretical models motivate quantity/price strategies, but quantitative models of the effects of MPP are scarce
Two key quantitative questions Can a micro-level financial friction cause significant systemic (or macro) overborrowing? Can it explain financial crises or affect business cycles? Sound MPP starts with a good model of crises Similar question as in the broad literature on financial frictions Is macroprudential policy effective to prevent overborrowing and financial crises? What are its main features? How does it affect incidence and magnitude of financial crises? What are its effects on asset pricing behavior (excess returns, Sharpe ratios, price of risk)?
What we do in this paper Answer the questions using an equilibrium asset pricing model with a collateral constraint. Compare decentralized eq. (DE) with a conditionally efficient social planner (SP) subject to IDENTICAL credit possibilities. The credit constraint plays two key roles: 1. Triggers Fisher's debt-deflation feedback mechanism, which causes deep recessions via financial amplification 2. Introduces a pecuniary externality via price of collateral assets (in good times agents do not internalize that lower leverage weakens Fisherian deflation in bad times ) A planner that reduces debt ex ante improves welfare.
Related literature Pecuniary credit market externalities: Participation constraints: Jeske (06), Wright (06), Lustig (00) using Kehoe & Levine (93) Collateral constraints: Caballero & Krishnamurthy (01), Lorenzoni (08), Korinek (09), Stein (10), survey by Galati and Moessner (11)... Quantitative studies: Bianchi (09), Nikolov (09), Jeanne & Korinek (10), Benigno et al. (10)... Amplification effects of financial frictions: Bernanke & Gertler (89), Kiyotaki & Moore (97), Bernanke, Gertler & Gilchrist (99), Aiyagari & Gertler (99), Kocherlakota (00), Mendoza & Smith (06), Mendoza (10).
Main findings 1. DE and SP yield similar average debt and leverage 2. but crises are larger and more frequent in DE Probability of financial crises increases by a factor of 3. Asset prices fall 17 ppts more (24% v. 7% for SP). Credit and consumption fall about 10 ppts more Overall cyclical variability is also higher 3. Mean excess return and Sharpe ratio rise by factors of 6 and 10, and market price of risk increases 81%. 4. SP s allocations implementable with state-contingent taxes on debt (1% on average, positively corr. with leverage) and on dividends (-0.4% on average)
Main elements of the model Inter-period non-state-contingent debt for self insurance & intra-period debt for working capital (WK) Collateral constraint limits total debt to fraction of market value of physical assets (in fixed supply) Production with labor and physical assets WK has zero financing cost but requires collateral Standard TFP shocks only (crises with realistic features result from endogenous amplification) GHH preferences remove wealth effect on labor supply
Representative firm-household problem in the decentralized economy Maximize: s.t. budget constraint and collateral constraint
Decentralized equilibrium conditions
Excess asset returns: Asset pricing conditions Forward solution for asset prices:
Taking as given Social Planner's problem
Pecuniary credit externality DE s private marginal utility cost of borrowing: SP s social marginal utility cost of borrowing: where and
Optimal macro-prudential policy Decentralize planner s eq. with state contingent taxes Tax on debt implements SP s bond decision rule: Tax on dividends makes asset prices equivalent:
Calibration
Decision rules for bonds in low TFP state l.r. prob: DE 4% SP 2% l.r. prob: DE 70% SP 69% l.r. prob: DE 27% SP 29%
Equilibrium land prices in low TFP state
Debt dynamics: amplification effects b t+1 b t+1 =b t b t
Long-run distribution of leverage ratio
Comparison of financial crisis events Simulate DE and SP economies for 100,000 periods. Define crisis events: binding credit constraint with fall in credit of more than 1sd. Construct 5-year windows centered in crisis periods. Compute median shocks in [t-2, t-1, t, t+1, t+2] and median initial debt at t-2. Simulate DE and SP given initial debt and sequence of shocks, and compare also against fixed price case (constant collateral price)
Differences in asset pricing properties Collateral constraint causes sharp drop in asset demand, and leads to higher and Sharpe ratio rises because rises more than (DE overcompensates risk-taking) The market price of risk rises Endogenous fat tails in distribution of asset returns
Asset pricing moments
Endogenous fat tails in CDF of returns
Sensitivity analysis: key parameters Collateral coefficient (loan-value ratio): Affects collateral s response to given price effects Negatively related to probability of a binding constraint. but effects are nonlinear (Fisherian deflation removed at =0 and also for large, nonbinding s) Risk aversion coefficient: Affects price elasticity of asset demand Disutility from binding constraints Frisch elasticity of labor supply: Higher elasticity produces larger output drops Smaller wage effects on collateral constraint
Sensitivity Analysis
Sensitivity analysis
Conclusions and future work Collateral constraints introduce systemic pecuniary externality that increase magnitude and incidence of financial crises, mean excess returns, volatility of returns and Sharpe ratios Optimal MPP taxes on debt and dividends neutralize externality, but implementation is difficult: State-contingent policies that require detailed information on debt and leverage of a large set of economic agents Taxing dividends during crises politically difficult, but selective implementation reduces welfare MPP has to adapt to fin. innovation and differences in information/beliefs (Bianchi, Boz & Mendoza (2012))
U.S. household debt relative to the value of residential land
Decentralized equilibrium in recursive form solve: Back