Empirical banking – ENSAE/PSE M2
Section 4: Global banks and the international transmission of shocks
Jean-Stéphane Mésonnier Banque de France
Global cross-border banking: a preview
[from BIS locational banking stats]
Global cross-border banking: a preview (2)
[from BIS locational banking stats]
High
concentration:
20 host
countries
account for
95% of total
cross-border
banking claims
Overview
1. Methodology:
– nonstandard SE issues in Diff-in-diff style papers – Demand vs supply with bank-firm data: Khwaja and
Mian, 2008
2. International transmission of bank capital and liquidity shocks
3. Liquidity management within global banking
organizations and the transmission of domestic monetary policy shocks at home and abroad
JS MESONNIER - Empirical Banking - Spring term 2016 4
References
• Khwaja, Asim, and Atif Mian, 2008, Tracing the impact of bank liquidity shocks:
Evidence from an emerging market, American Economic Review 98, 1413-1442.
• Peek, Joe, and Eric Rosengren, 1997, The international transmission of financial shocks: The case of Japan, American Economic Review 87, 495-505.
• Schnabl, Philipp, 2012, Financial globalization and the transmission of bank liquidity shocks: Evidence from an emerging market, Journal of Finance, 67 (3), 897-932.
• Cetorelli, Nicola, and Linda Goldberg, 2012, Liquidity management of US global banks: internal capital markets in the great recession, Journal of International Economics, 88, 299-311.
• Cetorelli, Nicola, and Linda Goldberg, 2012, Banking globalization, monetary transmission, and the lending channel. Journal of Finance, 67 (5), 1811-1843.
Non-standard SE issues in DinD style studies
• Reference: Angrist and Pischke, chap. 8
• Clustering and the Moulton (1986) factor in data with group structure – Solutions:
• Parametric using Moulton formula
• Liang-Zeger (1996) variance correction (STATA: cluster)
• WLS on group averages
• Serial correlation issues in panels / DinD models – Cf. Bertrand, Dufflo, Mullainathan (QJE, 2004)
– Most simple (and frequent) solution: time collapsing data
Khwaja and Mian 2008: Tracing the impact of bank liquidity shocks
• Important paper for its methodology: causal impact of bank-level shock on bank lending identified with multibank firms in bank-firm setup
• Issue: assessing the impact of a negative shock on bank liquidity (drop in deposits) for bank lending
– Exogenous shock linked to Pakistan’s nuclear tests in 1998: freeze on dollar deposits held by Pakistani banks, triggering a run with a lot of heterogeneity across banks
• Main results: impact mostly through quantities supplied
– Intensive margin: 1% fall in bank liquidity => 0,6% reduction in bank loan supply
– Extensive margin: 1% fall in bank liquidity => 0,12 pp reduction in
KM 2008: Pakistan’s nuclear tests and the dollar funding shock of local banks
• 1990s: banking reforms => increase of dollar denominated deposits by Pakistani banks
– NB: these dollar deposits in practice held by the central bank (SPB) against small fee => no exchange rate risk for depositors
– Large (endogenous) heterogeneity in dollar deposits share of local bank liabilities (in [0%-98%])
• 1998 Q2: nuclear tests in India and Pakistan => international sanctions
=> Government freezes dollar accounts to limit BoP crisis.
– Substantial losses for depositors => run on banks => liquidity shock proportional to ex ante share of dollar deposits
• Large currency depreciation (wealth effects) and recession: need to control for falling credit demand
KM 2008: The bank liquidity shock of 1998 in
Pakistan
KM 2008: The bank liquidity shock of 1998 in
Pakistan (2)
KM 2008: Data
• SBP’s credit information bureau:
– Exhaustive, quarterly information on bank loans to firms, July 1996- March 2000 => collapsed at bank-firm exposure level
• Annual bank balance sheet information for all 145 financial institutions – independent variable = dollar deposits / deposits, controls
• Time collapse: average loan amount pre crisis (1996Q3-1998Q1) vs average post crisis (1998 Q3-2000 Q1)
– Avoids problems due to autocorrelation of residuals, cf. Bertand et al. (2004)
• Final sample: 42 banks allowed to open dollar demand deposits,
KM 2008: Empirical methodology (1)
• OLS regression is likely to be biased by unobserved demand:
• Unbiased estimator involves firm FE for multibank firms:
KM 2008: shocked banks were good banks…
KM 2008: results: non-parametric preview
KM 2008: robustness issues
• The within-firm approach controls for generic credit demand
• What if loan-demand is specific?
• Problem if shocks to this specific demand is correlated with bank shock – Ex.: firms that borrow from shocked banks demand export-related
loans in dollars + demand for Pakistani exports drops
– Then, positive coefficient does not reflect bank lending channel – Idea: « within-firm » should be « within-firm and loan type »
• Solution in KM 2008: interacted firm and loan-type FE
• Remains looming problem in such bank-firm regression analyses!
Peek and Rosengren 1997: The international
transmission of financial shocks: the case of Japan
• Issue: international spillover of deleveraging by global banks under solvency stress?
• Case study: capital shortfall of international Japanese banks after the Nikkei krach of 1989-1992 while Basel I entered into force
• Look at domestic lending by branches / subsidiaries of Japanese banks in the US
• Results: strong impact of the negative capital shock to Japanese parent banks
– 1pp decline in a bank’s Capital/RWA => 6% decline in its US branches’ loans
– Almost no impact on lending by US-based subsidiaries
PR 1997: Context
• Strong international expansion of Japanese banks in the 1980s – 1988: all top ten global banks were Japanese
– 1990: US branches and subsidiaries of Japan = 18% of US C&I loans
• Basel I (1988) allows Japanese banks to include « hidden reserves » (latent valuation gains on equity participations) in tier 2
• Japanese stock market krach = sharp depletion of banks’ tier 2 => need to deleverage to meet Basel I capital regulation
• Strong relationships with Japan-based customers => less costly to adjust assets overseas
– Focus of Japanese regulators on unconsolidated ratios => less pressure on subsidiaries’ leverage
PR 1997: Data
• Selection of 29 Japanese banks with significant operations
– 11 city banks, 3 LT credit banks, 5 trust banks, 10 large regional banks
• Japanese branches/subsidiaries in the US: Fed’s Call reports – Branches activity collapsed at bank group level
• Semi-annual reports of Japanese banking groups
• Period: 1989 S1-1995 S3
PR 1997: Empirical model
• NB: adjusted branches’ loan growth = relative to loan growth by non- Japanese lenders in same state (weighted by assets if branches in several states) => controls for local demand
• Controls: LAND (land price in Japan = firms’ collateral value), US
macro controls (state-specific employment, Japanese FDI flows in US), bank-specific controls (branches’ assets, NPL, loan portfolio
composition, bank type dummies)
• Similar regression for Japanese subsidiaries
PR 1997: Japanese bank affiliates shed non-
Japanese assets first
Schnabl 2012: The International Transmission of
Bank Liquidity Shocks: Evidence from an Emerging Market
• Do international banks transmit liquidity shocks across countries? What are the real effects?
• Identification challenges
– If systemic liquidity shock: not enough variance across banks?
– Disentangling credit supply from demand by firms – Do firms have funding substitutes?
• Setting: exploits 1998 Russian default crisis as negative liquidity shock to international banks + focus on real impact on firms in Peru
– Peruvian banks: locally vs globally funded, vs foreign affiliates – Peruvian credit register = bank-firm exposure data
Schnabl 2012: The 1998 Russian default as a natural experiment
• 1992: Russian implements large reforms => rise of private sector and financial markets, but more macro instability
• 1995: economic stabilization program => falling revenues + surging government debt
• 1997: Asian crisis => concerns about Russia => fixed-exchange rate regime under pressure => peg dropped in Aug. ‘08 => panic and default
• Crisis came as a surprise (IMF backing f Russia) => changing
expectations of global investors (banks) about EM economies implied lowered exposures
• Peru hit by: drop in bank-to-bank loans (20% of Peruvian bank liabilities precrisis from international banks) + lower demand for exports (Brazil,
Schnabl 2012: data
• Bank data on all Peruvian banks (from Supervisor: SBS) => 43 bank
s
– Three types depending on (1) foreign vs domestic ownership and (2) share of internationally funded liabilities (above/below median of 6.2%)
– Exposure ranking: (HIGH) Domestically owned with high international
funding > (INTERMEDIATE) Foreign-owned > (LOW) Domestically owned with low international funding
– Desc. Stats: 15 HIGH comparable with 13 INTERMEDIATE, not with 15 LOW
• Bank-to-bank loans from international to Peruvian banks (SBS)
• Bank-firm loan data (SBS’ credit register)
• Firm-level balance sheet data (Tax administration)
Schnabl 2012: International lenders to Peruvian
banks (top 10 Peruvian banks)
Schnabl 2012: Bank summary stats
Schnabl 2012: foreign parent banks and arm’s
length lenders of Peruvian banks equally hit
Schnabl 2012: arm’s length lenders transmit shock
Schnabl 2012: parent banks shield affiliates more
Schnabl 2012: formal analysis
• Stage 1: international => Peruvian banks, using BtB loans
– Note: Lender and borrower FE, time collapse (4 months before, 12 months after Russian default), bank-level clustering
Schnabl 2012: First stage, results
Schnabl 2012: formal analysis (2)
• Stage 2: Peruvian banks => firms, using BtF exposures (within firm approach)
– Note: Multibank firms only, borrower FE, time collapse (1 year before, 1 year after Russian default), bank-level clustering
Schnabl 2012: Second stage, results:
high exposure (=absorbed benchmark) implies less lending growthSchnabl 2012: Second stage, results (2): grouping
firms by type
Schnabl 2012: Second stage, results (3): alternative
shock measure
Schnabl 2012: Second stage, placebo test: 2 years
vs 1 year before Russian default
Schnabl 2012: formal analysis (3)
• Stage 3: Firms’ funding => firms’ performance, using firm data
– Note: no firm FE here => identification assumption : credit demand othogonal to firm exposure conditional on observables
– time collapse (1 year before, 1 year after Russian default), multibank firms (for comparison with above)
– Y = change in a firm’s borrowing, or in share of its defaulted loans, or dummy for survival in 2005 (linear proba).
Schnabl 2012: Third stage:
Question: how to interpret low coeff of LOW in (1-2) ?
Cetorelli Goldberg 2012 JIE: Liquidity management of US global banks
• Global banks’ role as vehicle of international contagion highlighted
during subrime crisis and Great Recession (e.g. Acharya and Schnabl, 2010 IMF ER)
– Global banks manage liquidity at global level: active internal capital markets
• How does this work? Two views:
– Organizational pecking order: headquarters’ needs first
– Locational pecking order: global portfolio management (funding vs investment markets/countries)
• Strong policy implications
CG 2011a: peering inside US Global banks
• Use of new data from FFIEC: quarterly country exposures reports (FFIEC 09)
– Detailed {bank head}-{foreign affiliates (by location country)}
information
– 1) Bank i claims on residents in country j (cross-border vs local), 2) local liabilities of affiliates of i in j, 3) net internal liabilities of
affiliates in j vàv headquarters of bank i
= NetDueToijt (<0 if affiliates are net lenders)
• Sample of US banks (BHCs) with foreign offices, 2006-2010 – 50 banks, 2/3 US and 1/3 foreign-owned, unbalanced panel – Affiliates in 121 countries
CG 2012a: US bank affiliates worlwide, numbers
CG 2012a: US banks’ affiliates size
affiliates’ local liabilities + net inflow from parent
CG 2012a: Two funding shocks
• Identification requires shocks to US parent that are othogonal to overseas affiliates
• Idea: large US (and foreign banks) exposed to the ABCP freeze of 2007
• Two shocks:
– Negative in August 2007: subprime panic – Positive in December 2007: Fed’s TAF
• Instrument: banks’ exposure to the ABCP market in December 2006 – Source: Acharya and Schnabl (2010): matched ABCP conduits
rated by Moody’s with sponsoring bank organizations
CG 2012a: Analysis
• Two key variables characterizing local overseas markets:
• Dependant variable (pre vs post for each shock)
CG 2012a: Econometric analysis
• Note: specification allows to test for the two views: organizational vs locational pecking order
– Cf. value and signs of g0 and g3
• Includes coutry-, bank- and affilates-specific controls and interactions – Core-Funding and Core-Investment included in Xijt
CG 2012a: results support locational pecking order
CG 2012a: shock 2
Cetorelli and Goldberg 2012b: Banking globalization and monetary policy transmission
• Global banks manage internal liquidity at global scale.
– Can they insulate their domestic credit supply from the effects of monetary policy shocks?
– If so, do they transmit the shock elsewhere (to foreign affiliates)?
– Less domestic transmission but more international contagion of MP?
• Analyze response of US global banks to MP shocks
• Confirm that bank globalization matters for MP effectiveness
CG 2012b: Methodology
• New evidence on the lending channel:
– two-step approach adapted from Kashyap and Stein (2000) – First step: one regression for each quarter (cross-section)
– NB: concerns about endogeneity of liquidity ratio (Xit-1)
« alleviated » by instrumenting with residual of regression of X on NPL/L and C&IL/L
JS MESONNIER - Empirical Banking - Spring term 2016 53
CG 2012b: Methodology (2)
– Second step:
– NB: SE corrected for autocorrelation by Newey-West
CG 2012b: Methodology (3)
• Tests of internal funding activity
– Dependent variable: D(Net Due to Own foreign offices)it [NB: inflow into parent is signed positively = opposite sign to dependent variable in CG (2012a)]
CG 2012b: Methodology (4)
• Test of international transmission
CG 2012b: Data
• Call reports for all US chartered banks, 1980-2005, quarterly
– Large / small banks: above p95 / below p90 of asset size distribution – Global bank: if « foreign assets » > 0 (= lending from foreign offices)