Global Banks and International Shock Transmission: Evidence from The CrisisNicola Cetorelli Linda GoldbergFederal Reserve Bank NY Federal Reserve Bank NY
NBER
The views expressed in this paper are those of the individual authors and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System. 1
Capital flows to emerging markets plummeted during the crisis
-100
0
100
200
300
400
500
-100
0
100
200
300
400
500
1998 2000 2002 2004 2006 2008
Private Capital Flows to Emerging Markets By Region
Emerging Asia
Emerging Europe
Latin America
USDbn USDbn
A collapse of bank loans to emerging markets dominated the drop in capital flows
-150
-50
50
150
250
350
450
550
-150
-50
50
150
250
350
450
550
1998 2000 2002 2004 2006 2008
Private Capital Flows to Emerging Markets
FDI
Portfolio Equity
USDbn USDbn
Bank Loans
Net Debt Securities
Channels of international transmission through global banks
4
DepositsLiquid assets
Large global bank
Domestic parent balance sheet
Loans Other Funds
Capital
External borrowingDomestic loans
Cross-border loans
Large build up of $ (long-term) assets financed with short-term $ funding.
Channels of international transmission through global banks
5
DepositsLiquid assets
Large global bank
Foreign liquid assets
Domestic parent balance sheet
Foreign affiliatebalance sheet
LoansDomestic loans
Cross-border loans
Other FundsExternal borrowing
Capital
Loans
Deposits
Other Funds
Capital
Internal borrowing
Foreign local loans
Internal lending
Channels of international transmission vary by type of bank
6
Deposits
Capital
Liquid assets
Loans
Domestic bank in EM country
Other funds
International interbank funds(cross-border borrowing)
A collapse of bank loans to emerging markets dominated the drop in capital flows
-150
-50
50
150
250
350
450
550
-150
-50
50
150
250
350
450
550
1998 2000 2002 2004 2006 2008
Private Capital Flows to Emerging Markets
FDI
Portfolio Equity
USDbn USDbn
Bank Loans
Net Debt Securities
Identification challenges
Not all suppliers of fund hit the same way by shock. Lending drop should be especially strong for those ex ante more exposed to $ funding shock
Lending supply or lending demand?
8
Identification challenges
Data: BIS international banking statistics. For each BIS reporting country, data on international
claims vis-à-vis EM countries.
17 source countries to 94 EM countries.
Both cross-border lending and local lending. IMF data on domestic lending in EM
countries.
9
Identification challenges
Need measure of ex-ante exposure to $ funding risk.
Built with confidential BIS data on $ assets and liabilities of banks in each source country. Used data up to 2007q2.
10
0 1 ( $ )ij i ijL Ex ante vulnerability
Still problem of demand simultaneity. OLS estimate likely to be biased
Unobservable demand component
11
0 1 ( $ )ij ji ijL Ex ante vulnerability
0 1 ( $ )ij i ijL Ex ante vulnerability
Use Fixed Effect specification:
Identification from the comparison of lending growth to the same EM country by banks from different source countries.
Identification strategy as in Kwhaja and Mian (AER 2008)
12
1 ( $ )i jj i ijL Ex ante vulnerability FE
While OLS estimates are biased, they can be used to extract information on loan demand shocks.
Turns out useful to evaluate significance of third channel (changes in lending supply by domestic banks)
From OLS-FE estimations on local lending of source country banks we can gauge importance of the demand bias
13
Number of Emerging Market Countries (of 94) in BIS Reporting Country Lending
International Claims (Cross-Border) Local Claims in Local Currency
Source Country Pre Crisis 2006Q3-2007q2
Post Crisis 2008q3-2009q2
Pre Crisis 2006Q3-2007q2
Post Crisis 2008q3-2009q2
United States 72 76 41 42 Japan 50 47 15 15 Australia 33 32 1 3 Belgium 72 71 11 14 Canada 63 65 21 22 Switzerland 80 79 23 23 Germany 82 81 19 23 Denmark 59 58 13 1 Spain 70 67 16 17 France 86 82 34 43 Great Britain 86 86 37 35 Ireland 46 43 2 1 Sweden 64 63 6 7 Portugal 52 49 5 5 Netherlands 79 78 29 29 Luxembourg 37 38 0 0 Italy 66 72 20 19
Cross-Border Local Claims
(1) (2) (3) (4)
Vulnerability -0.316** -0.380*** -0.271* -0.269Vienna Initiative Countries -- -0.037 -- -0.927**Vulnerability · Vienna -- -0.309 -- 3.070***
# obs. 1029 1029 245 245
R2 0.250 0.253 0.397 0.421
Fixed effect coefficients not reported. *** p < 0.01, ** p < 0.05, * p < 0.10
Bank Lending to Emerging MarketsFixed Effects Regressions
Cross-Border Local Claims
(1) (2) (3) (4)
Vulnerability -0.316** -0.380*** -0.271* -0.269Vienna Initiative Countries -- -0.037 -- -0.927**Vulnerability · Vienna -- -0.309 -- 3.070***
# obs. 1029 1029 245 245
R2 0.250 0.253 0.397 0.421
Fixed effect coefficients not reported. *** p < 0.01, ** p < 0.05, * p < 0.10
Bank Lending to Emerging MarketsFixed Effects Regressions
Loan Supply Contractions from Source to EM Destinations
Cross Border Lending Local Lending
Pre-Crisis Bilateral Quarterly Average
($ millions)
Post-Pre % Change
Pre-Crisis Bilateral Quarterly Average
($ millions)
Post-Pre % Change
Germany(low vulnerability)
9,233 -8.02% 5,136 -6.88%
Spain(mid vulnerability)
1,454 -18.26 % 14,417 -15.66%
United Kingdom(high vulnerability)
3,644 -28.44% 8,547 -24.39%
Channels of international transmission vary by type of bank
18
Deposits
Capital
Liquid assets
Loans
Domestic bank in EM country
Other funds
International interbank funds(cross-border borrowing)
Domestic Lending Supply Growth Shock
VARIABLES (1) (2) (3) (4) High V2 share in -0.370*** -0.348** -0.453*** cross border (H) (0.135) (0.135) (0.160) Cross border 0.431* 0.311 -0.174 share in funding (X) (0.243) (0.238) (0.467) (H)·(X) 0.818 (0.678) Constant 0.654*** 0.370*** 0.610*** 0.677*** (0.093) (0.044) (0.098) (0.112) Observations 58 58 58 58 R-squared 0.118 0.050 0.145 0.167
The dependent variable is a measure of domestic bank lending growth pre-post crisis for each emerging market country. Lending in the “pre” crisis period is defined as the time average between 2006q2 and 2007q2. Lending in the “post” crisis period is defined as the time average between 2008q3 and 2009q2. Share of cross-border lending is the ratio of total cross-border lending in a country to total domestic lending. High V2 share in cross border is the share of cross border funding obtained from source countries with V2 values above the median value across source countries. Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 .
• Significant Drop in Lending Supply to Emerging Markets
• Shock transmitted both directly - cross-border lending channel - and indirectly - through internal capital markets channel of banks managing liquidity globally.
• Economic magnitude of transmission channels is large.
Takeaways
• Policy interventions to support balance sheet of developed countries’ banks (Vienna Initiative) alleviated local claims transmission.
• Evidence for third channel as well: domestic EM banks especially reliant on cross-border funding from ex-ante highly vulnerable banks were most affected.
• General “openness” not a factor.
Takeaways
22
Gross ST US dollar funding risks*
international claims
* Liabilities to official monetary authorities + International liabilities to non-banks + Local liabilities to US residents booked by US offices + Net Liabilities to banks + cross-currency FX swap (if negative)
Germany Spain United Kingdom
$ billion 865 247 1,524
% intl. claims (25.4) (57.8) (90.0)
Ex ante dollar vulnerability (McGuire and Von Peter, 2009) Proxies of dollar funding risk: Ideally one would want measure of maturity mismatch.
Data allows to construct upper bounds.
V1: OMA_L+ NB_L + OTH B_L + FX SWAPS (if negative) V2: OMA_L+ NB_L + NET OTH B_L + FX SWAPS (if negative) V3:
OMA_L+ NB_L (if negative) + NET OTH B_L + FX SWAPS (if negative) OMA_L+ NET OTH B_L + FX SWAPS (if negative)
We scale each measure by total international country claims.
24
25
Low ex ante vulnerability
Belgium Germany Denmark FinlandIreland Italy Luxembourg Portugal Sweden
High ex ante vulnerabillity
Australia Canada Switzerland Spain France Great Britain Japan Netherlands United States
Identification of Transmission via Loan Supply
26
• measured pre- versus post- crisis
• Use measure of ex ante dollar vulnerability to pick up relative size of source country/bank funding shock
• Larger funding shock associated with larger lending , whether defined over cross-border loans or local claims.
• But, OLS regression (1) is biased due to unobservable demand component, j.
0 1 ( $ 1 )ij i j ijL Ex ante vulnerability
Identification of Transmission via Loan Supply
27
Solution: obtain identification of loan supply effect by comparing lending by banks hit differently by the shock but lending to the same country
• any change in demand should be common across lenders, and would not affect the comparison.
The FE specification in (2) captures the j specific demand shock with the vector of FE variables FEj.
coefficients compare lending between banks hit severely and banks not hit severely to the same country.
12 ij i j j ijL D FE
Additional econometric observations While OLS estimates are biased, they can be used
[with (2)] to extract information on loan demand shocks. By construction, the residuals from the OLS regressions from
(1) should reflect a noise component plus the idiosyncratic demand component for each country of destination.
The residuals from the corresponding fixed effect estimation should only reflect the noise component.
In conjunction with data on lending by domestically-owned banks by country, we isolate loan supply effects from these effects for comparison with loan supply by foreign owned banks.
28
29
Table 4. Summary statistics Variable Obs Mean Std. Dev. Min Max
Pre-post cross-border lending growth 390 0.230 0.663 -2.376 2.943 To EM Europe 129 0.385 0.714 -1.507 2.943 To EM Latin America 128 0.221 0.630 -2.376 2.327 To EM Asia 130 0.089 0.620 -1.545 2.278 Pre-post local lending growth 185 0.331 0.866 -6.788 3.379 To EM Europe 66 0.559 0.674 -0.766 3.206 To EM Latin America 54 0.208 0.845 -2.100 3.379 To EM Asia 65 0.201 1.010 -6.788 1.244 Pre-post domestic lending growth 22 0.427 0.218 -0.021 0.811 Ex-ante dollar vulnerability Correlations V1 18 0.780 0.492 0.064 1.674 V2 0.992 18 0.611 0.434 0.051 1.455 V3 0.701 0.710 18 0.208 0.201 0.009 0.831
30
(1) (2) (3) (4) (5) (6) VARIABLES OLS OLS OLS FE FE FE
iD
proxy
V1 -0.185*** -0.176*** (0.0685) (0.0659) V2 -0.222*** -0.211*** (0.0777) (0.0747) V3 -0.263 -0.218
(0.182) (0.176) Constant 0.373*** 0.365*** 0.282*** (0.0626) (0.0577) (0.0493) Observations 390 390 390 390 390 390 R-squared 0.018 0.021 0.005 0.239 0.241 0.228
Cross-border lending growth to emerging markets
31
(1) (2) (3) (4) (5) (6) VARIABLES OLS OLS OLS FE FE FE V1 -0.248* -0.198 (0.133) (0.134)
iD V2 -0.314** -0.261*
proxy (0.152) (0.154) V3 -1.074** -0.984** (0.431) (0.427) Constant 0.555*** 0.556*** 0.530*** (0.136) (0.126) (0.102) Observations 185 185 185 185 185 185 R-squared 0.019 0.023 0.033 0.310 0.313 0.323
Local claims lending growth to emerging markets
32
33
34
Regional differences Vienna initiative VARIABLES Cross-border lending Local lending Cross-border lending Local lending (OLS) (OLS) (FE) (FE) V2 -0.509*** -0.339 -0.262*** -0.408** (0.134) (0.278) (0.0813) (0.193) V2 ∙ Latin America 0.362* 0.272 (0.189) (0.390) V2 ∙ Asia 0.505*** 0.0395 (0.185) (0.378) Latin America -0.377*** -0.501 (0.139) (0.311) Asia -0.596*** -0.323 (0.138) (0.301) Constant 0.686*** 0.761*** (0.0971) (0.196) Vienna countries -0.268 -1.317*** (0.254) (0.468) V2 ∙ Vienna 0.461 3.573*** (0.742) (1.285) Observations 390 185 390 185 R-squared 0.072 0.054 0.247 0.346
Regional differences and the Vienna initiative
35
36
Domestic bank lending growth in emerging markets
(1) (2) (3) (4) (5) (6) VARIABLES Raw data Corrected data Raw data Corrected data Raw data Corrected data V2 -0.161* -0.0656 -0.145* -0.0573 (0.0885) (0.144) (0.0851) (0.148) Share of cross-border lending
0.229* 0.119 0.207* 0.110 (0.126) (0.204) (0.121) (0.210)
Constant 0.508*** 0.462*** 0.328*** 0.377*** 0.410*** 0.409** (0.0626) (0.102) (0.0705) (0.114) (0.0828) (0.144) Observations 22 22 22 22 22 22 R-squared 0.142 0.010 0.141 0.017 0.256 0.024
37
Table 3 Growth and Decline in Lending, percent change by source and destination
Emerging Europe Emerging Asia Latin America Source Region
Type of claim
2006Q3-
2007Q2
2008Q3-
2009Q2
2006Q3-
2007Q2
2008Q3-
2009Q2
2006Q3-
2007Q2
2008Q3-
2009Q2
North America International 49.3 10.6 30.3 40.6 13.5 21.8
Local 36.1 -4.5 21.8 6.5 18.4 -0.5
Europe International 52.0 -9.8 27.0 -21.5 18.4 -15.6
Local 55.1 -10.6 55.1 -9.6 31.6 -4.1
Asia International 29.1 -16.1 21.1 -8.2 28.1 -3.3
Local 113.9 -39.7 12.7 -6.2 7.3 -15.6
Domestic Banks Lending 32.1 -21.3 16.2 14.6 25.0 -13.5