Uncertainty and International Banking Claudia M. Buch (Deutsche Bundesbank) Manuel Buchholz (Halle Institute for Economic Research) Lena Tonzer (EUI, Halle Institute for Economic Research) IMF/DNB Conference, Amsterdam June 2014 The views expressed reflect those of the authors and not necessarily those of the Deutsche Bundesbank.
Motivation Since the outbreak of the financial crisis, banks have withdrawn from international markets and have become more reluctant to supply loans. Uncertainty rises in recessions and has a negative impact on short-run hiring and investment in the manufacturing sector (Bloom 2014). However, evidence on the relationship between increased levels of uncertainty in the banking system and banks lending activities is scarce (Valencia 2013). Similar to a nonfinancial firm that makes an investment, it might be beneficial for a bank to postpone the loan decision in the presence of uncertainty. Thus, we are interested in uncertainty in banking and banks response in (inter)national loan supply.
Research questions What have been patterns of uncertainty in banking during the crisis? We construct a measure for uncertainty based on bank-level data (Bloom et al. 2012). It is based on the cross-sectional dispersion of bank-level shocks to total assets, short-term funding, productivity, and profitability. How does uncertainty affect banks lending behavior? We exploit our measure of uncertainty in banking to analyze the effect of uncertainty on loan supply. We follow Cornett et al. (2011) who analyze the impact of funding shocks on banks investment patterns depending on their balance sheet strength and liquidity management. We are interested in whether foreign-owned banks react differently to domestic uncertainty than domestically-owned banks. We analyze if cross-country differences in uncertainty affect cross-border bank lending.
Uncertainty in banking
Definition & measurement During times of uncertainty future outcomes become less predictable. E.g. banks become worse in forecasting their business outcomes (assets, profits etc.). A larger dispersion of shocks at the bank-level can be interpreted as higher uncertainty in banking. To construct the measure of uncertainty in banking, we proceed in two steps. First, we extract the unexplained components of the variables on which our measure of uncertainty in banking is based. Second, we compute the dispersion per country and year across these unexplained parts to obtain a cross-sectional measure of uncertainty in banking (Bloom et al. 2013). We look at shocks to total assets, short-term funding, bank productivity, and profitability (RoA).
Definition & measurement 1. The shocks for the four variables are derived from the following model: X X ) = log( X ) = α + α + ε log( i ) log( i 1 where log( X ) is the growth rate of bank i's assets/ short-term funding/ productivity or i profitability (RoA, in levels) at time t in country j. We include bank and country-year fixed effects: and If we assume that banks forecast according to this model, we capture the individual bank forecast error in each year by the residual of the regression model. i α i i α i 2. We compute the cross-sectional dispersion as the standard deviation (SD): BLU = SD[ i Hence, our bank-level uncertainty measure for country j at time t is based on the dispersion of the unpredicted part of the regression model. A higher dispersion of shocks means that the distribution of individual forecast errors gets wider, which reflects higher uncertainty in a country s banking sector. ε ]
Sample & data Our sample is based on banks in 48 countries which belong to the OECD, EU, and/or G20. The sample period spans the years 1998-2012. In the bank-level analysis, we use data from Bankscope. The explanatory variables control for banks' balance sheet strength and liquidity management (Cornett et al. 2011). We compare our measures of uncertainty in banking to alternative measures: Financial sector: bank stock return volatility, stock market volatility (Datastream). Real sector: GDP growth, firm return dispersion (IMF, Bloom 2014). Other: disagreements in forecasts, economic policy uncertainty index (Bloom 2014, Baker et al. 2013). Banks international activities: We exploit information on banks foreign ownership status (Claessens and Van Horen 2014). We use aggregate bilateral cross-border lending of countries banking systems (Bank for International Settlements).
Descriptive statistics
Uncertainty in banking increased during the financial crisis.
Uncertainty in banking fluctuates less than alternative measures of uncertainty.
Regression analysis
How does uncertainty affect bank lending? Loans Assets i i 1 = v i + v t + α loggdpdef + α loggdp + α X + α BLU + α BLU * X + ε 1 2 3 i 1 4 5 i 1 i Dependent variable: Difference in loan volume relative to total assets in t-1. Uncertainty in banking: BLU (bank-level uncertainty) measured by the crosssectional dispersion in shocks to bank-specific variables. Identify the response of banks to uncertainty depending on their balance sheet strength and liquidity management: BLU * X i 1 Control variables : Balance sheet strength and liquidity management. X i 1 Banks' liquid assets in total assets, bank capitalization, the share of customer deposits in total assets, banks' balance sheet size, and committed loan obligations. Bank/ year or country-year fixed effects to control for time-invariant bank characteristics/ effects common to all banks. Standard errors clustered by bank.
Identification Loans Assets i i 1 = v i + v t + α loggdpdef + α loggdp + α X + α BLU + α BLU * X + ε 1 2 3 i 1 4 5 i 1 i Supply vs. demand: We exploit heterogeneity across banks with respect to balance sheet strength and liquidity management. Reverse causality: We examine the effect of uncertainty measured in the whole banking sector of one country on an individual bank s loan supply. Omitted variables: We control for unobservable fixed-effects and alternative measures of uncertainty.
A higher level of uncertainty affects banks loan supply negatively. Dep. var.: ΔLoans t / assets t-1 Total assets Short-term funding Productivity RoA ΔLog GDP deflator t 0.341*** 0.352*** 0.317*** 0.322*** (0.059) (0.059) (0.059) (0.059) ΔLog real GDP t 0.618*** 0.658*** 0.563*** 0.623*** (0.083) (0.088) (0.084) (0.083) Bank-level uncertainty (BLU) t -4.309*** -0.355-2.510*** -1.335*** (0.674) (0.504) (0.512) (0.339) Liquid assets/assets t-1 x BLU t 0.036* 0.033** 0.060*** 0.007 (0.019) (0.016) (0.022) (0.014) Capital/assets t-1 x BLU t 0.086** -0.132*** 0.001 0.085*** (0.041) (0.036) (0.039) (0.028) Deposits/assets t-1 x BLU t -0.001-0.016-0.006-0.027* (0.014) (0.012) (0.019) (0.015) Log total assets t-1 x BLU t 0.720*** -0.248* 0.270 0.130 (0.175) (0.146) (0.185) (0.127) Comm. loans/(comm. loans + assets) t-1 x BLU t -0.045-0.118*** -0.020 0.023 (0.033) (0.036) (0.040) (0.021) Bank fixed effects yes yes yes yes Country fixed effects yes yes yes yes Year fixed effects yes yes yes yes Control variables yes yes yes yes Observations 10,282 10,282 10,164 10,282 R-squared 0.216 0.208 0.208 0.212 Number of banks 2,355 2,355 2,323 2,355
The effect of uncertainty on bank lending depends on banks liquidity ratio.
The effect of uncertainty on bank lending depends on banks capital ratio.
Do foreign-owned banks react differently? Loans Assets α BLU 5 i i 1 = v i + v t + α loggdpdef * Fown(0 /1) + α BLU 1 6 kt + ε ki + α loggdp 2 + α X 3 i 1 + Fown(0 /1) + α BLU 4 + Impact of internationalization on bank loan supply well documented. De Haas and van Lelyveld (2014), De Haas and van Horen (2013), Ongena et al. (2013), Giannetti and Laeven (2012), Cetorelli and Goldberg (2011). Foreign-owned banks might react differently to domestic uncertainty. Identification through ownership status of a bank: Fown(0 /1) Thus, we interact domestic uncertainty with the ownership status (1:=foreign owned): BLU * Fown(0 /1) And control for uncertainty in the country of the foreign owner: BLU kt
No conclusive evidence that foreign-owned banks react differently than domestically-owned banks. Dep. var.: ΔLoans t / assets t-1 Total assets Short-term funding Productivity RoA ΔLog GDP deflator t 0.323*** 0.341*** 0.349*** 0.338*** (0.060) (0.062) (0.061) (0.060) ΔLog real GDP t 0.867*** 0.854*** 0.807*** 0.787*** (0.113) (0.116) (0.118) (0.113) Fown(0/1) 2.401 3.692 2.288 0.711 (4.271) (3.683) (3.439) (3.618) Bank-level uncertainty (BLU) -1.591** 0.243-1.083** -1.781*** (0.658) (0.410) (0.503) (0.629) Fown(0/1) x BLU -0.686-0.022 1.651** 0.502 (0.908) (0.628) (0.720) (0.964) Bank-level uncertainty (BLU) kt -0.961-1.495* -0.781** -0.080 (0.741) (0.772) (0.323) (0.418) Bank fixed effects yes yes yes yes Country fixed effects yes yes yes yes Year fixed effects yes yes yes yes Control variables yes yes yes yes Observations 2,576 2,566 2,464 2,573 R-squared 0.281 0.268 0.266 0.274 Number of banks 636 636 611 636
Cross-border lending reacts to differential in home and host country uncertainty in banking. Dep. var.: Log (Cross-border assets jkt /(GDP *GDP kt )) Total assets Short-term funding Productivity RoA ΔLog GDP Deflator - ΔLog GDP Deflator kt 0.028** -0.003 0.036*** 0.014 (0.012) (0.005) (0.006) (0.011) ΔLog real GDP kt - ΔLog real GDP 0.084*** 0.031*** -0.028** 0.018 (0.018) (0.011) (0.013) (0.015) BLU - BLU kt 0.016-0.125*** -0.066** -0.123*** (0.033) (0.037) (0.030) (0.041) Observations 9,088 8,944 7,473 9,044 R-squared 0.41 0.41 0.47 0.41 Number of country pairs 817 817 772 817 Country-pair FE Yes Yes Yes Yes Country-year FE Yes Yes Yes Yes
Concluding remarks We develop a measure of uncertainty in banking based on bank-level data. Higher uncertainty in banking affects banks loan supply negatively. This effect is heterogeneous across banks: Lending by banks which have higher liquidity buffers and are better capitalized tend to be less affected by domestic uncertainty. Foreign-owned banks do not seem to respond differently than domestically-owned banks. Cross-border bank lending reacts to differentials in home and host country uncertainty in banking. Our results point to the importance of sufficient bank liquidity and capital to stabilize lending during times of uncertainty.
Backup I: Shock vs. level dispersion
Backup II: Country-year fixed effects Dep. var.: ΔLoans t /assets t-1 Total assets Short-term funding Productivity RoA Liquid assets/assets t-1 x BLU t 0.082*** -0.004 0.098*** 0.025 (0.023) (0.015) (0.023) (0.017) Capital/assets t-1 x BLU t 0.118*** -0.101*** -0.064* 0.082*** (0.045) (0.039) (0.035) (0.030) Deposits/assets t-1 x BLU t -0.024 0.010 0.005-0.043** (0.021) (0.018) (0.025) (0.017) Log total assets t-1 x BLU t 0.479*** -0.456*** -0.372* 0.255** (0.166) (0.165) (0.216) (0.130) Comm. loans/(comm. loans + assets) t-1 x BLU t -0.066* -0.105*** -0.005-0.019 (0.035) (0.037) (0.045) (0.025) Bank fixed effects yes yes yes yes Country-year fixed effects yes yes yes yes Control variables yes yes yes yes Observations 10,282 10,282 10,164 10,282 R-squared 0.305 0.300 0.297 0.306 Number of banks 2,355 2,355 2,323 2,355
Backup III: Sample split small, medium, large banks small medium large Total assets dispersion -8.621** -3.300*** -0.979* (3.597) (0.961) (0.538) Short-term funding dispersion Productivity dispersion RoA dispersion 2.136-1.628* -1.895*** (2.632) (0.968) (0.516) 3.081-2.787*** -0.290 (2.719) (0.760) (0.532) -3.804** -1.353*** -1.118*** (1.576) (0.453) (0.356)
Backup IV: BLU vs. Stock market volatility Dep. var.: ΔLoans t /assets t-1 Total assets Short-term funding Productivity Stock market volatility (UNC t ) 0.279-0.332-0.222 0.058 (0.314) (0.299) (0.309) (0.314) BLU t -4.593*** -0.338-2.323*** -1.615*** (0.703) (0.533) (0.509) (0.343) Liquid assets/assets t-1 x BLU t 0.040** 0.029* 0.065*** 0.010 (0.020) (0.016) (0.022) (0.014) Capital/assets t-1 x BLU t 0.075* -0.141*** -0.001 0.088*** (0.042) (0.036) (0.039) (0.031) Deposits/assets t-1 x BLU t -0.008-0.024* -0.000-0.036** (0.016) (0.013) (0.020) (0.015) Log total assets t-1 x BLU t 0.835*** -0.261* 0.251 0.203 (0.188) (0.153) (0.184) (0.134) Comm. loans/(comm. loans + assets) t-1 x -0.053-0.124*** -0.020 0.021 BLU t (0.034) (0.036) (0.040) (0.024) Liquid assets/assets t-1 x UNC t 0.011 0.016 0.020* 0.010 (0.012) (0.013) (0.012) (0.012) Capital/assets t-1 x UNC t 0.018 0.027 0.020-0.006 (0.025) (0.024) (0.024) (0.027) Deposits/assets t-1 x UNC t 0.020** 0.021** 0.018** 0.036*** (0.009) (0.009) (0.009) (0.010) Log total assets t-1 x UNC t -0.138* -0.012-0.027-0.100 (0.073) (0.068) (0.067) (0.072) Comm. loans/(comm. loans + assets) t-1 x UNC t RoA 0.000-0.002-0.006-0.003 (0.019) (0.018) (0.019) (0.022)