1
How did the crisis in international funding markets affect bank lending? Balance sheet evidence from the UK
Shekhar AiyarInternational Monetary Fund and Bank of England.
2
Motivation
How did problems originating in one asset class in one country propagate internationally and spark the Great Recession?
A standard stylized explanation relies on the globalization of the banking system and has two parts.
1.Stress in the US banking system (and others with direct exposure to US mortgages / structured products) spread internationally through bank funding markets.
2.This shock to banks’ external funding was transmitted domestically through a reduction in credit supply.
This paper seeks to identify step 2 above for the UK.
3
Related Literature
• Older literature on impact of non-monetary shocks on bank lending: Bernanke (1983); Peek and Rosengren (1997)
• Khwaja and Mian (2008). Impact of external funding shock on domestic lending for Pakistani banks. Use 1998 nuclear test to identify equation.
• Schnabl (2011). Liquidity shock to global banks from Russian default in 1998 led to fall in lending to Peruvian banks.
• Cetorelli and Goldberg (2010). Liquidity shocks in developed country banking systems led to contraction in loan supply to EMEs.
• Amiti and Weinstein (2009), Chor and Manova (2009). Impact of bank health on trade volumes (via trade finance).
4
•Between end-Q1 2008 and end-Q3 2009, the external liabilities of UK-resident banks fell by 22%.
•The previous largest 6-quarter fall was 9%, during the ERM crisis.
0
2000
4000
6000
8000
10000
01/12/1977
01/08/1980
01/04/1983
29/11/1985
29/07/1988
29/03/1991
29/11/1993
29/07/1996
29/03/1999
29/11/2001
29/07/2004
29/03/2007
29/11/2009
US$
Bill
ion
Quarter
Chart 1: An unprecedented shock to banks' external funding
UK-resident banks' external liabilities
UK-resident banks' external liabilities adjusted for exchange rate changes
5
Empirical strategy: data• As a global financial center, UK has a large and heterogenous
sample of resident banks, HQ’d in several countries.
• Study uses detailed regulatory balance sheet data reported to BOE at quarterly frequency (CL, CE, BT, BE forms).
• Focus on period during which shock to external funding occurred (March 2008 through September 2009).
• Bank mergers dealt with by creating a synthetic merged series.
• Domestic lending constructed as sum of lending to Households, Businesses, Other Banks and OFIs. Each subseries constructed by summing relevant items in BT / BE reporting forms.
6
Empirical strategy: estimation
We want to estimate
where i indexes banks∆DLi = change in (log) domestic lending
∆XLi = change in (log) external liabilities
where j indexes sector {Households, Businesses, Other Banks, OFIs}; denotes bank i’s claims on sector j as a ratio of its total domestic claims; and denotes the change in lending by all banks to sector j.
Jj
jiji TBLsDEM
iiiii ZDEMXLDL '21
ijs
jTBL
7
Identification
• If is exogenous, no problem.
• But this needs to be established, not assumed.
• Potentially OLS is biased and inconsistent (omitted variables / reverse causality).
• Hence implement IV, using 3 instruments for
iXL
iXL
8
First instrument• Share of repos in bank’s external liabilities at t=0.
• Measures ex-ante reliance on wholesale funding.
• Plenty of evidence that repo markets became very stressed, with a substantial increase in repo haircuts (Gorton and Metrick 2009). Bear Stearns collapse in March 2008 precipitated by loss of secured short-term funding, despite adequate capital (Acharya and Gale 2009).
• Funding shock was transmitted in part through repo market (Raddatz 2010).
• So higher reliance on external repo funding at the beginning of the shock period should be associated with bigger funding shock.
• Equally, external reliance on repo should not impact future domestic lending except through the funding shock.
9
Second instrument
• Share of external liabilities to foreign affiliates at t=0.
• Measures the degree to which liabilities are “within firm”.
• In response to monetary shocks “internal capital markets” are activated to insulate against shock. Cetorelli and Goldberg (2010).
• In financial crises foreign subsidiaries rely on liquidity from parent banks: De Haas and Lelyveld (2010).
10
Third instrument
• Banking system stress in country of ownership.
• Use data on nationality of bank HQ to divide all nations into 9 regions (USA, UK, EUR, CHE, CAN, JAP, NJA, OTH).
• For each region, use change in LIBOR-OIS spread as a measure of banking system stress.
• Wherever possible, use region-specific LIBOR equivalent (EURIBOR for EUR, SIBOR for SIN, BBSR for AUS etc.)
• Taylor and Williams (2008) emphasize that LIBOR-OIS spread measures counterparty risk not liquidity risk.
11
-1
-0.5
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
03-Jan-06
03-Apr-06
03-Jul-06
03-Oct-06
03-Jan-07
03-Apr-07
03-Jul-07
03-Oct-07
03-Jan-08
03-Apr-08
03-Jul-08
03-Oct-08
03-Jan-09
03-Apr-09
03-Jul-09
03-Oct-09
LIBOR-OIS Spreads
USA UK EUR CHE JPY CAD SGP
12
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
USA UK EUR CHE JPY AUS CAD SGP HKG
LIBOR-OIS Spreads
Pre-sample period (2006 - Q1 2008) Q1 2008 - Q3 2009
13
1 2 3 4
Dependent variable: ∆DL 2SLS 2SLS 2SLS 2SLS
∆XL .55** 0.59** .65** .60**
0.27 0.25 0.28 0.28
DEMAND .035*** .032***
0.009 0.01
Size controls No No Yes Yes
N 141 141 141 141
Underidentification (H0: Not identified)
A-P chi-squared statistic 31.57 38.76 30.53 32.34p-value 0.00 0.00 0.00 0.00
Overidentifying restrictions (H0: Instruments uncorrelated with error process)
Sargan chi-squared statistic 0.35 0.17 0.12 0.071
p-value 0.84 0.92 0.94 0.96
Weak instruments (H0: Instruments are weak)
K-P Wald rank F-statistic 10.23 12.46 9.74 10.25
10% critical value (Stock and Yogo) 9.1 9.1 9.1 9.1
Table 1: Impact of change in external liabilities on change in domestic lending
14
1 2
Dependent variable: ∆DL 2SLS OLS
∆XL .60** .51***
0.28 0.09
DEMAND .032*** .034***
0.01 0.01
Size controls Yes Yes
N 141 141
R-squared 0.27
Exogeneity of explanatory variable (H0: Variable is exogenous)
Difference-in-Sargan statistic 0.14
p-value 0.71
Table 2: 2SLS and OLS
15
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1 2 3 4 5 6 7 8 9
Conditional Quantile Functions
95% quantile confidence interval
Quantile regression estimate
OLS estimate
1 2Dependant variable: ∆DL OLS Median Regression
∆XL .51*** .55***0.09 0.1
DEMAND .034*** .031***0.01 0.01
Size controls Yes Yes
N 141 141R-squared 0.27 0.21
Table 4: Median impact on change in domestic lending
16
1 2Dependant variable: ∆DL OLS OLS
∆XL .45*** .83***0.10 0.12
DEMAND .032*** .033***0.01 0.01
UOB 25.98***6.3
SUB -26.8***6.95
BRN -26.1***6.92
UOB*∆XL .38**0.17
SUB*∆XL -.52***0.17
BRN*∆XL -.32*0.19
Constant 3.02 29.6***4.56 6.66
Size controls Yes Yes
N 141 141R-squared 0.31 0.32
Table 5: The impact of bank type
17
1 2Dependant variable: ∆DL OLS OLS
∆XL .54*** .45**0.15 0.19
DEMAND .024** .024**0.011 0.012
Fraction of DL in FX (t=0) -21.5* -16.7912.77 11.59
(Fraction of DL in FX)*∆XL -0.8 -0.010.32 0.31
UOB 23.46***7.78
UOB*∆XL .41**0.22
Constant 12.19* 7.684.84 5.20
Size controls Yes Yes
N 141 141R-squared 0.29 0.32
Table 6: Lending in FX
18
1 2 3Dependant variable: ∆DL OLS OLS OLS
∆XL .56*** .49*** .39***0.15 0.10 0.12
DEMAND .033*** .033*** .031***0.01 0.01 0.01
Foreign assets / Total assets (t=0) -14.7210.74
(Foreign assets / Total assets)*∆XL -0.110.32
Foreign assets / Foreign liabilities (t=0) -6.56** -4.543.21 3.2
(Foreign assets / Foreign liabilities)*∆XL -0.003 0.040.06 0.06
UOB 24.18***6.44
UOB*∆XL .41**0.17
Constant 13.46** 11.98** 6.726.59 5.35 5.66
Size controls Yes Yes Yes
N 141 141 141R-squared 0.28 0.29 0.33
Table 7: Are foreign assets a significant buffer?
19
Households Businesses Other Banks OFIs1 2 3 4
Full sample ∆XL -63.18 (54.86) -552 (554) 1.16 (.59)* .51 (.28)*% of total lending 100 100 100 100N 122 134 139 130
Lending > £100 m∆XL -.27 (.36) .58 (.16)*** .41 (.22)* .69 (.24)***% of total lending 99.8 99.8 99.8 99.9N 27 91 105 73
Lending > £500 m∆XL .08 (.27) .47 (.15)*** .56 (.27)** .92 (.31)***% of total lending 99.6 98.4 99.2 98.8N 19 60 70 47
Lending > £1000 m∆XL .32 (.19) .45 (.19)** .87 (.26)*** 1.01 (.32)***% of total lending 99.1 96.5 98.6 96.9N 15 47 48 40
Sample
Change in lending to:
Table: Sectoral regressions
20
Change in lending to: 1 2 3 4 5 6 7 8 9
Businesses∆XL 0.53 .56** .66*** .62** .54*** .49*** .51*** .57*** .60**s.e. 0.54 0.22 0.22 0.24 0.17 0.1 0.09 0.18 0.29
Other Banks∆XL -0.02 0.31 .32* .51*** .39** .55*** .52** 0.58 0.14s.e. 0.25 0.19 0.17 0.19 0.18 0.16 0.23 0.43 1.28
OFIs∆XL -0.26 0.38 .54*** .84** .83* .93** .91** 1.08*** 1.11**s.e. 0.36 0.31 0.16 0.35 0.48 0.46 0.38 0.26 0.46
Decliles of conditional distribution
Table: Quantile regressions on components of domestic lending
21
Conclusions• Large, well-identified and robust impact of external funding shock on
domestic bank lending.
• Relationship holds across the distribution.
• Corroborates standard view that bank lending was significant in transmitting international financial crisis to real economy.
• Foreign-owned banks (branches and subs) cut back domestic lending more, irrespective of size of funding shock. UK-owned backs calibrated pullback in lending more closely to size of funding shock.
• Evidence for pullback in domestic lending to 3 out of 4 sectors: businesses, OFIs and other banks. No evidence of pullback in HH lending (maybe because of unwinding securitization).