Amplifying slumps, taming booms: the bitter consequences of high - PowerPoint PPT Presentation

Amplifying slumps, taming booms: the bitter consequences of high indebtedness Cyril Couaillier, Valerio Scalone Sciences Po Banque de France June 11, 2019 1 / 25

This paper ◮ Assess the impact of the credit cycle affect on the propagation of monetary and credit shocks ◮ Allow for different impact on positive and negative shocks (i.e. for state-dependent sign asymmetries) ◮ Identify large state dependent asymmetries: contractionnary shocks are amplified when the credit cycle is high 2 / 25

Motivation ◮ In the aftermath of the GFC, indebtedness has been pointed as an important amplification factor of the downturn (Jord` a et al. [2013], Mian et al. [2017]); ◮ When agents are on their debt limit, shocks affecting debt limits can force borrowers to deleverage and further reduce spending: ◮ State effect: shocks are amplified when vulnerability is high (financial accelerator: Bernanke et al. [1996], Kiyotaki and Moore [1997], Occasionally Binding Constraint: Guerrieri and Iacoviello [2016], Maffezzoli and Monacelli [2015]) ◮ Sign effect: shocks with negative effect on debt limit hit more the economy in absolute terms than shocks of the same size but with opposite sign ◮ Several credit to output ratio transformation display good properties as early warning indicators of financial crises 3 / 25

Literature 1 Asymmetry matters to study amplification of shocks and financial vulnerability: Aikman et al. [2016, 2017], Aladangady [2014], Alpanda and Zubairy [2017], Barnichon and Matthes [2016], Bauer and Granziera [2016], Cloyne et al. [2016], Harding and Klein [2018], Hofmann and Peersman [2017], Jord` a et al. [2016], Juselius et al. [2016] 2 Non-linear dynamics related to indebtedness: Barnichon et al. [2016], Carriero et al. [2018], Jord` a et al. [2013, 2015, 2016], Mian et al. [2017] 3 Credit growth is a good early warning indicator of financial crises and their costs Bridges et al. [2017], Jord` a et al. [2011] 4 / 25

The strategy ◮ Time series analysis with non-linear terms detecting state and sign effects; ◮ Use of an interaction variable (3y difference in Credit to GDP Ratio) to capture continuous build-up of vulnerability; Credit/GDP ◮ Smooth Local Projections (SLP) method by Barnichon and Brownlees [2018]) to allow for non-linearities and limit noise in the estimation ◮ Two identification strategies ◮ Instrument variable (SVAR-LP) with High Frequency Identification of monetary shocks from Miranda-Agrippino and Ricco [2018]) ◮ Sign restrictions to jointly recover monetary policy and credit supply shocks ◮ Estimation on quarterly US data (1983-2018) 5 / 25

The empirical model For each horizon h = 0 ... H , the setting is: Y t + h = α h + Σ p ℓ =1 L h ,ℓ X t − ℓ + Λ h X ✶ , ¯ X t − 1 + Φ h k t − 1 X t − 1 (1) + Ψ h k t − 1 X ✶ , ¯ X t − 1 + u h , t with Y t vector of endogenous variables, Z t = ( Y t , Z t ) ′ vector of regressors variables, k t scalar interaction variable, � X 1 , t − 1 ✶ X 1 , t − 1 < ¯ � X ✶ , ¯ X 1 X , ¯ t − 1 = ... X vector of cutoff values, u h , t vector X n , t − 1 ✶ Xn , t − 1 < ¯ Xn of errors at horizon h . L h ,ℓ Λ h Φ h Ψ h the coefficient matrices at horizon h . 6 / 25

Data ◮ Sample for US: 1983Q1-2018Q2 ◮ Endogenous variables ◮ Real GDP growth; ◮ Inflation; ◮ Shadow short term rate (Wu and Xia [2016]), robustness with 1-y gov. bond rate; ◮ Lending rates for HH (30-y mortgage rate) and NFC (Moody’s BAA Corporate bond yield); ◮ Total credit (loans and debt securities) to HH and NFC, net flow over stock; ◮ Ratio of credit to households over credit to Non-Financial Corporations ◮ Exogenous interaction variable: Credit to GDP ratio 3-y variation, in pp; Credit/GDP 7 / 25

Benchmark specification and identification strategy ◮ Process with one lag; ◮ 20 quarters horizon; ◮ Cut-off values for asymmetry: historical mean; ◮ First identification: identification with High Frequency (Miranda-Agrippino and Ricco [2018]) ◮ Robustness and credit shock identification with sign restrictions 8 / 25

SVAR-LP for monetary shock So-called SVAR-LP strategy first proposed in Gertler and Karadi [2015] 1. Regress estimated errors u p t of policy rate on a time series of structural policy shocks to obtain the part of error due to u p those shocks (ˆ t ). We use the series developed by Miranda-Agrippino and Ricco [2018] built by using the market surprises and taking into account central banks’ private information and informational rigidity MA&R (2018) 2. Regress other reduced form errors u q t on the fitted policy u p errors ˆ t to recover the ratios of the impacts on the monetary policy shock on all the values 3. Determine impact to the monetary policy shocks through a variance covariance decomposition 9 / 25

Monetary shock - HFI Figure: Impulse responses to a monetary shock ` a la MA&R response: GDP response: Inflation response: GDP response: Inflation 0.5 1.0 0.5 0.5 0.5 0.0 0.0 0.0 0.0 −0.5 −0.5 −0.5 −1.0 −0.5 −1.0 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 response: Shadow policy rate response: Lending rate HH response: Shadow policy rate response: Lending rate HH 0.4 0.5 0.5 0.2 0.2 0.0 0.0 0.0 0.0 −0.5 −0.2 −0.2 −0.5 −1.0 −0.4 −0.4 quantile quantile 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0.1 0.1 response: Lending rate NFC response: Credit flow HH response: Lending rate NFC response: Credit flow HH 0.9 0.9 0.50 0.25 10 0 0.25 5 0.00 −10 0.00 0 −0.25 −0.25 −5 −0.50 −20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 response: Credit flow NFC response: Ratio of HH to NFC credit response: Credit flow NFC response: Ratio of HH to NFC credit 15 0.04 0.02 5 0.02 10 0.00 0 0.00 5 −0.02 −5 −0.02 0 −0.04 −5 −0.06 −10 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 Figure: Contractionary shock Figure: Expansionary shock The red (green) lines are the impulses at the 90th (10th) percentile of the credit cycle. Shaded areas represent the bootstrapped 90% confidence intervals. Cumulated IRF for GDP, inflation and credit flows 10 / 25

Sign restriction identification 1. Compute the eigendecomposition of the covariance matrix Γ so that Γ D Γ ′ = ˆ Ω 2. draw a set of independent normal vectors and take the resulting R of their their QR decomposition 3. Check whether Γ Q verifies some sign conditions 4. If not repeat 2-3 Sign restriction matrix inspired from Gambetti and Musso [2017] and Furlanetto et al. [2017]: GDP Inflation ST rate HH Rate NFC Rate HH Cr NFC Cr HH/NFC Cr AD + + + + + AS + - MP + + - HH CS + + + - + + NFC CS + + + - + - 11 / 25

Non-linear effects for monetary policy shock Strong asymmetric and state-dependent effects ◮ Only recessionary shocks have statistically significant effects on income and credit growth; ◮ Recessionary shock are amplified by the credit cycle: ◮ Strong negative and persistent effect on GDP when vulnerability is high ◮ Spread increases more when vulnerability is high ◮ Effect on credit statistically significant only when vulnerability is high ◮ Credit channel stronger on households than on firms (HH/NFC credit ratio decreases when vulnerability is high) 12 / 25

Monetary shock Figure: Impulse responses to a monetary shock response: GDP response: Inflation response: GDP response: Inflation 1.0 1.0 0.25 0.5 0.4 0.5 0.00 0.0 0.0 0.0 −0.25 −0.5 −0.5 −0.50 −0.4 −1.0 −1.0 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 response: Shadow policy rate response: Lending rate HH response: Shadow policy rate response: Lending rate HH 0.4 0.3 0.5 0.2 0.5 0.2 0.1 0.0 0.0 0.0 0.0 −0.5 −0.1 −0.5 −0.2 −0.2 −1.0 −0.3 quantile quantile 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0.1 0.1 response: Lending rate NFC response: Credit flow HH response: Lending rate NFC response: Credit flow HH 0.9 0.9 10 10 0.4 0.2 5 0.2 0.0 0 0 0.0 −0.2 −10 −0.2 −5 −0.4 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 response: Credit flow NFC response: Ratio of HH to NFC credit response: Credit flow NFC response: Ratio of HH to NFC credit 15 0.04 10 0.01 10 0.02 0.00 5 0.00 5 −0.01 0 −0.02 0 −0.02 −0.04 −5 −5 −0.03 −0.06 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 Figure: Contractionary shock Figure: Expansionary shock The red (green) lines are the impulses at the 90th (10th) percentile of the credit cycle. Shaded areas represent the bootstrapped 90% confidence intervals. Cumulated IRF for GDP, inflation and credit flows 13 / 25