Global Banks and International Shock Transmission Evidence from

Global Banks and International Shock Transmission: Evidence from The Crisis Nicola Cetorelli Linda Goldberg 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

A collapse of bank loans to emerging markets dominated the drop in capital flows

Channels of international transmission through global banks Large global bank Domestic parent balance sheet Liquid assets Deposits Loans Other Funds Domestic loans External borrowing Cross-border loans Capital Large build up of $ (long-term) assets financed with shortterm $ funding. 4

Channels of international transmission through global banks Large global bank Domestic parent balance sheet Foreign affiliate balance sheet Liquid assets Deposits Foreign liquid Deposits assets Loans Other Funds Loans Domestic loans External borrowing Foreign local loans Cross-border loans Internal borrowing Other Funds Internal lending Capital 5

Channels of international transmission vary by type of bank Domestic bank in EM country Liquid assets Deposits Loans Other funds International interbank funds (cross-border borrowing) Capital 6

A collapse of bank loans to emerging markets dominated the drop in capital flows

Identification challenges n n 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 n 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. n IMF data on domestic lending in EM countries. 9

Identification challenges n n 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 2007 q 2. 10

n Still problem of demand simultaneity. OLS estimate likely to be biased n Unobservable demand component n 11

n Use Fixed Effect specification: n Identification from the comparison of lending growth to the same EM country by banks from different source countries. n Identification strategy as in Kwhaja and Mian (AER 2008) 12

n n n 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

Bank Lending to Emerging Markets Fixed Effects Regressions Cross-Border Local Claims (1) (2) (3) (4) -0. 316** -0. 380*** -0. 271* -0. 269 -- -0. 037 -- -0. 927** -- -0. 309 -- 3. 070*** # obs. 1029 245 R 2 0. 250 0. 253 0. 397 0. 421 Vulnerability Vienna Initiative Countries Vulnerability · Vienna Fixed effect coefficients not reported. *** p < 0. 01, ** p < 0. 05, * p < 0. 10

Loan Supply Contractions from Source to EM Destinations Cross Border Lending Local Lending Pre-Crisis Bilateral Quarterly Average ($ millions) Germany (low vulnerability) Spain (mid vulnerability) United Kingdom (high vulnerability) Post-Pre % Change Pre-Crisis Bilateral Quarterly Average ($ millions) Post-Pre % Change 9, 233 -8. 02% 5, 136 -6. 88% 1, 454 -18. 26 % 14, 417 -15. 66% 3, 644 -28. 44% 8, 547 -24. 39%

Channels of international transmission vary by type of bank Domestic bank in EM country Liquid assets Deposits Loans Other funds International interbank funds (cross-border borrowing) Capital 18

Domestic Lending Supply Growth Shock

Takeaways • 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.

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 (Mc. Guire and Von Peter, 2009) n n n Proxies of dollar funding risk: Ideally one would want measure of maturity mismatch. Data allows to construct upper bounds. V 1: OMA_L+ NB_L + OTH B_L + FX SWAPS (if negative) V 2: OMA_L+ NB_L + NET OTH B_L + FX SWAPS (if negative) V 3: q q n 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

Low ex ante vulnerability High ex ante vulnerabillity Belgium Germany Denmark Finland Ireland Italy Luxembourg Portugal Sweden Australia Canada Switzerland Spain France Great Britain Japan Netherlands United States 25

Identification of Transmission via Loan Supply • D 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 D , whether defined over cross-border loans or local claims. • But, OLS regression (1) is biased due to unobservable demand component, hj. 26

Identification of Transmission via Loan Supply 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. g coefficients compare lending between banks hit severely and banks not hit severely to the same country. 27

Additional econometric observations n While OLS estimates are biased, they can be used [with (2)] to extract information on loan demand shocks. q q n 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 domesticallyowned banks by country, we isolate loan supply effects from these effects for comparison with loan supply D by foreign owned banks. 28

29

Cross-border lending growth to emerging markets 30

Local claims lending growth to emerging markets 31

0. 6 Cross-border lending pre and post crisis By ex ante high or low vulneraibility 0. 4 Average Log Lending 0. 2 0 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 Low vulnerability High vulnerability -0. 2 -0. 4 -0. 6 -0. 8 32

33

Regional differences and the Vienna initiative 34

35

Domestic bank lending growth in emerging markets 36

37