The Effect of Central Bank Liquidity Injections on Bank Credit Supply

Transcription

1 The Effect of Central Bank Liquidity Injections on Bank Credit Supply Luisa Carpinelli Bank of Italy Matteo Crosignani Federal Reserve Board AFA Meetings Banks and Central Banks Session Chicago, 8 January 2017 Disclaimer: The views expressed in this presentation are solely the responsibility of the authors and should not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve System or of anyone else associated with the Federal Reserve System, the Bank of Italy or of the Eurosystem.

2 Lender of Last Resort and Bank Credit Supply Central banks acting as LOLR during recent crises - Fed (TAF), ECB (LTRO), BoE (FLS) - Goal: restore bank credit supply following a negative shock Theory - Banks are fragile because of liquidity transformation - Banks hit by a negative shock reduce credit supply - LOLR liquidity provision restores bank credit supply Empirical evaluation is difficult - Recent episodes - Data limitation - Even with data, not obvious how to use

3 This Paper Tests the LOLR theory - ECB December 2011 liquidity provision - Official goal of supporting bank lending The intervention - The 3-Year Long Term Refinancing Operation - Provision of collateralized loans to banks - Largest liquidity provision in history (e1 trillion) Effect on Italian bank credit supply - Italian banks hit by a dry-up before the intervention - We combine the national credit registry with bank security-level holdings

4 Contribution 1) Central bank liquidity is effective in restoring bank credit supply following a wholesale funding dry-up - Banks hit by the dry-up : (i) reduce credit supply during the dry-up (ii) restore credit supply after the intervention - Firms benefit from the intervention What s new? LOLR effect on bank credit supply 2) LOLR liquidity encourages reaching-for-yield - Banks not hit by the dry-up use LOLR liquidity to buy risky securities What s new? Transmission varies in the cross-section

5 SETTING AND DATA

6 Rising Sovereign Credit Risk (CDS Spreads) Dec2005 Dec2007 Dec2009 Dec2011 Dec2013 Italy Spain

7 Dec 2011: ECB Lendse1 trillion to Eurozone Banks 3-Year Long Term Refinancing Operation (LTRO) - Largest LOLR intervention in history - Turning point of the crisis - Italian and Spanish banks largest users (2/3 total uptake) Simple design - 3-year maturity collateralized cash loans - Banks can choose how much to obtain in two allotments - Need to pledge collateral (government bonds, ABS,...) Italy as a laboratory

8 #1: Italian Banks Hit by a Dry-Up

9 #2: Data on the Entire Intermediation Chain ECB loans to banks - Bank-level borrowing at ECB Bank Characteristics - Detailed composition of funding - Security-level holdings and collateral use Bank Loans to Firms - All outstanding loans >e30,000 (Credit Registry) - Term loans, credit lines, trade credit Firm Characteristics - Large subset (55%) of firms

10 EFFECT ON BANK CREDIT SUPPLY

11 Two Empirical Challenges 1) Borrowers are not randomly assigned to banks - Stock of credit is an equilibrium quantity (demand, supply) Compare credit growth from different banks within firms (Khwaja and Mian (2008)) 2) Banks choose how much to borrow from the central bank - Use of bank-level LTRO uptakes as a source of variation would probably capture other bank characteristics Use bank reliance on the foreign wholesale market as a measure of exposure to the dry-up and the intervention

12 Exposure to the Dry-Up Exposure j = ForeignWholesaleFunding j,jun11 Assets j,jun11

13 Exposure to the Dry-Up Exposure j = ForeignWholesaleFunding j,jun11 Assets j,jun % of loans belong to banks with exposure> 5%

14 Exposed Banks are Large and Highly Levered Unit Exposed Non-Exposed Total Assets bne Tier 1 Ratio Units Leverage Units ROA % Assets Government Bonds % Assets Securities % Assets Credit to Firms % Assets Household Deposits % Assets

15 Exposed Banks are Large and Highly Levered Unit Exposed Non-Exposed Total Assets bne Tier 1 Ratio Units Leverage Units ROA % Assets Government Bonds % Assets Securities % Assets Credit to Firms % Assets Household Deposits % Assets Need to control for bank characteristics

16 Two Questions 1) Does the dry-up affect bank credit supply? - Compare normal and dry-up periods 2) Does the liquidity provision affect bank credit supply? - Compare dry-up and intervention periods

17 Baseline Specification Log(Credit ijt ) = α+β 1 Exposure j I Run,Int + β 2 Exposure j I Int + µ it + γ ij + φ X jt + ǫ ijt - Observations at the(i, j, t) firm-bank-period level

18 Baseline Specification Log(Credit ijt ) = α+β 1 Exposure j I Run,Int + β 2 Exposure j I Int + µ it + γ ij + φ X jt + ǫ ijt - Observations at the(i, j, t) firm-bank-period level - LHS: credit granted on credit lines, term loans, trade credit

19 Baseline Specification Log(Credit ijt ) = α+β 1 Exposure j I Run,Int + β 2 Exposure j I Int + µ it + γ ij + φ X jt + ǫ ijt - Observations at the(i, j, t) firm-bank-period level - LHS: credit granted on credit lines, term loans, trade credit - RHS: Exposure interacted with: - I Run,Int equal to one in the intervention and dry-up periods - I Int equal to one in the intervention period

20 Baseline Specification Log(Credit ijt ) = α+β 1 Exposure j I Run,Int + β 2 Exposure j I Int + µ it + γ ij + φ X jt + ǫ ijt - Observations at the(i, j, t) firm-bank-period level - LHS: credit granted on credit lines, term loans, trade credit - RHS: Exposure interacted with: - I Run,Int equal to one in the intervention and dry-up periods - I Int equal to one in the intervention period - Fixed Effects: bank-firm (γ ij ) and firm-time (µ it )

21 Baseline Specification Log(Credit ijt ) = α+β 1 Exposure j I Run,Int + β 2 Exposure j I Int + µ it + γ ij + φ X jt + ǫ ijt - Observations at the(i, j, t) firm-bank-period level - LHS: credit granted on credit lines, term loans, trade credit - RHS: Exposure interacted with: - I Run,Int equal to one in the intervention and dry-up periods - I Int equal to one in the intervention period - Fixed Effects: bank-firm (γ ij ) and firm-time (µ it ) - Bank balance sheet controls X jt

22 Effect on Bank Credit Supply Log(Credit ijt ) = α+β 1 Exposure j I Run,Int + β 2 Exposure j I Int + µ it + γ ij + φ X jt + ǫ ijt

23 Effect on Bank Credit Supply Log(Credit ijt ) = α+β 1 Exposure j I Run,Int + β 2 Exposure j I Int + µ it + γ ij + φ X jt + ǫ ijt Credit Exposure j I Int *** *** *** *** (0.045) (0.037) (0.037) (0.040) (0.031) Exposure j I Int 0.247*** 0.251*** 0.245*** 0.172*** 0.115** (0.061) (0.044) (0.043) (0.043) (0.053) Time FE Bank FE Firm-Time FE Relationship Controls Bank-Firm FE Bal. Sheet Controls N 2,322,142 2,322,142 2,322,142 2,171,749 2,171,749 R

24 Effect on Bank Credit Supply Log(Credit ijt ) = α+β 1 Exposure j I Run,Int + β 2 Exposure j I Int + µ it + γ ij + φ X jt + ǫ ijt Credit Exposure j I Int *** *** *** *** *** (0.045) (0.037) (0.037) (0.040) (0.031) Exposure j I Int 0.247*** 0.251*** 0.245*** 0.172*** (0.061) (0.044) (0.043) (0.043) (0.053) Time FE Bank FE Firm-Time FE Relationship Controls Bank-Firm FE Bal. Sheet Controls N 2,322,142 2,322,142 2,322,142 2,171,749 2,171,749 R

25 Four Additional Results 1) Is the effect sizable? - With no LTRO, credit to firms 5.6% instead of observed 3.6% during Dec 2011-June Dry-up causes a 3.7% contraction, LTRO a+2% increase 2) Which banks drive the dynamics of credit supply? - High leverage banks (Kashyap and Stein (1995)) 3) To which firms banks reduce and restore credit supply the most? - Risky firms, especially during the restoration of credit supply (Jimenez et al. (2014)) 4) Are firms affected? - Firms unable to fully switch lenders during the dry-up - Firms expand total borrowing after the intervention

26 TRANSMISSION CHANNEL

27 All Banks Take Advantage of Cheap LOLR Liquidity The LTRO successfully attracts many banks - ECB liquidity is cheap compared to the private market - Our sample banks borrowe181.5 billion - Median uptake is 9.7% of total assets Little heterogeneity in banks uptakes of the ECB liquidity - Dry-up is uncorrelated with liquidity uptakes Bank exposure to the dry-up is not a valid instrument for the LTRO uptakes. The effect on bank credit supply is unrelated to the central bank liquidity provision.

28 Reconciling our Findings We exploit a regulatory intervention by the Italian government to reconcile our findings Banks need to pledge collateral to access LOLR - Sovereign bonds, ABS, government guaranteed securities Italian Government Guarantee Program - Decree Law passed shortly after the LTRO announcement - Treasury can guarantee otherwise ineligible securities - Treasury collects a 1% fee - Banks created collateral worth $103 billion - Government guarantees backs half Italian LTRO uptakes

29 Banks Hit by the Dry-Up Use the Govt Guarantee Uptake j = α+ βexposure j,jun11 + µx j,jun11 + ǫ j

30 Banks Hit by the Dry-Up Use the Govt Guarantee Uptake j = α+ βexposure j,jun11 + µx j,jun11 + ǫ j TotalUptake Uptake GovtCollateral Exposure Jun ** (0.197) (0.101) LEV Jun *** (0.284) (0.146) ROA Jun ** (0.041) (0.021) T1R Jun *** * (0.220) (0.113) NPL Jun (0.247) (0.127) Large Jun * (4.312) (2.215) N R

31 Banks Hit by the Dry-Up Use the Govt Guarantee Uptake j = α+ βexposure j,jun11 + µx j,jun11 + ǫ j TotalUptake Uptake GovtCollateral Exposure Jun (0.197) (0.101) LEV Jun *** (0.284) (0.146) ROA Jun ** (0.041) (0.021) T1R Jun *** * (0.220) (0.113) NPL Jun (0.247) (0.127) Large Jun * (4.312) (2.215) N R

32 Transmission of Central Bank Liquidity Log(Credit ijt ) = α+ βuptake j I Int + µ it + γ ij + φ X jt + ǫ ijt

33 Transmission of Central Bank Liquidity Log(Credit ijt ) = α+ βuptake j I Int + µ it + γ ij + φ X jt + ǫ ijt Log(Credit) TotalUptake I Int (0.144) Uptake GovtCollateral I Int 0.249** (0.122) Uptake StandardCollateral I Int * (0.142) Bank-Firm FE Firm-Time FE Bal. Sheet Controls N 1,381,420 1,381,420 1,381,420 R

34 Transmission of Central Bank Liquidity Log(Credit ijt ) = α+ βuptake j I Int + µ it + γ ij + φ X jt + ǫ ijt Log(Credit) TotalUptake I Int (0.144) Uptake GovtCollateral I Int Uptake StandardCollateral I Int (0.122) (0.142) Bank-Firm FE Firm-Time FE Bal. Sheet Controls N 1,381,420 1,381,420 1,381,420 R

35 One Open Question: How Do Less Exposed Banks Use the Central Bank Liquidity? Analyze holdings of government bonds - Holdings of government bonds double in Dec 2011-June Large literature on the sovereign-banks doom loop

36 Banks Reach for Yield Using Government Bonds Non Exposed Banks Exposed Banks Share of Total Assets 2011m6 2011m m6

37 Transmission to Government Bond Holdings Govt j,t = α+γ j + η t + βexposure j,jun11 I Int + Γ j,t + ǫ jt

38 Transmission to Government Bond Holdings Govt j,t = α+γ j + η t + βexposure j,jun11 I Int + Γ j,t + ǫ jt Govt Govt Domestic Govt GIIPS Govt GIPS Govt nongiips Exposure Jun11 I Int ** ** ** (0.244) (0.243) (0.243) (0.243) (0.270) LEV Jun11 I Int (0.253) (0.255) (0.255) (0.253) (0.277) ROA Jun11 I Int (0.053) (0.052) (0.052) (0.052) (0.047) T1R Jun11 I Int 0.171** 0.165** 0.166** 0.171** 0.127** (0.069) (0.068) (0.068) (0.069) (0.059) NPL Jun11 I Int (0.129) (0.128) (0.128) (0.129) (0.130) Large Jun11 I Int (2.533) (2.382) (2.383) (2.535) (2.751) Bank FE Time FE N R

39 Transmission to Government Bond Holdings Govt j,t = α+γ j + η t + βexposure j,jun11 I Int + Γ j,t + ǫ jt Govt Govt Domestic Govt GIIPS Govt GIPS Govt nongiips Exposure Jun11 I Int ** (0.244) (0.243) (0.243) (0.243) (0.270) LEV Jun11 I Int (0.253) (0.255) (0.255) (0.253) (0.277) ROA Jun11 I Int (0.053) (0.052) (0.052) (0.052) (0.047) T1R Jun11 I Int 0.171** 0.165** 0.166** 0.171** 0.127** (0.069) (0.068) (0.068) (0.069) (0.059) NPL Jun11 I Int (0.129) (0.128) (0.128) (0.129) (0.130) Large Jun11 I Int (2.533) (2.382) (2.383) (2.535) (2.751) Bank FE Time FE N R

40 Transmission to Government Bond Holdings Govt j,t = α+γ j + η t + βexposure j,jun11 I Int + Γ j,t + ǫ jt Govt Govt Domestic Govt GIIPS Govt GIPS Govt nongiips Exposure Jun11 I Int ** ** ** ** (0.244) (0.243) (0.243) (0.243) (0.270) LEV Jun11 I Int (0.253) (0.255) (0.255) (0.253) (0.277) ROA Jun11 I Int (0.053) (0.052) (0.052) (0.052) (0.047) T1R Jun11 I Int 0.171** 0.165** 0.166** 0.171** 0.127** (0.069) (0.068) (0.068) (0.069) (0.059) NPL Jun11 I Int (0.129) (0.128) (0.128) (0.129) (0.130) Large Jun11 I Int (2.533) (2.382) (2.383) (2.535) (2.751) Bank FE Time FE N R

41 Final Thoughts Central banks should require good quality collateral - Banks hit by the run are collateral constrained - Italian government guarantee: fiscal side of the intervention Banks that need the LOLR the most cannot access it Loan Maturity - ECB was providing short-term liquidity before Dec Banks were liquid, but facing large rollover risk LOLR loan maturity matters for bank rollover risk LORL incentivizes reaching for yield Trade-off: information Vs. incentivize reaching for yield