Cross-Border Monetary Policy Spillovers: Evidence from the IBRN Matthieu Bussière (Banque de France), Robert Hills (Bank of England) and Alejandro Jara (Banco Central de Chile)

April 19 2017 Joint IMF-IBRN Conference on the Transmission of Macroprudential and Monetary Policies Across Borders The views are those of the authors and not necessarily reflective of those of the Banque de France, the Bank of England or the Banco Central de Chile. Introduction

 Cross-border monetary policy spillovers figure prominently in policy debates (G20, BIS and OECD meetings, IMF WEO, etc).

 Large academic literature focusing on misc. aspects (reaction of asset prices, exchange rates, bank lending, macro variables, etc). To pick a few: Gertler and Karadi, 2015; Gilchrist et al. 2015, Monnet, 2014, Coibion 2012; Obstfeld et al, 2010; Gorodnichenko et al. 2016; Swanson and Williams, 2014; Jiménez et. al, 2012; Wright, 2011; Obstfeld, 2015, Altunbas et al. 2012; Hanson and Stein, 2012. Christensen and Rudebusch, 2012; Bredin and Stevenson, 2010; Angeloni et al., 2015; Behn et al. 2015; Duygan- Bump et al. 2013; Cetorelli and Goldberg, 2012a,b; Ioannidou et al, 2015; Berndt and Yeltevkin, 2015; Ireland, 2015; Barakchian and Crowe, 2013; Bekaert et al., 2013; Massa and Zhang, 2013; Goldberg, 2013 Project Overview -1

This IBRN project focuses on the reaction of lending of global banks to monetary policy decisions of EA, US, UK, and JP. We use for this the unique bank-level datasets of the individual country teams of the IBRN (17 countries + BIS and ECB)  this enables cross-country comparisons and checking robustness of the results across studies.  Inward transmission: how changes in foreign monetary policy affect the balance sheet of hosted affiliates of foreign banks and domestic banks (ref. Morais, Peydro and Ruiz, 2015 for Mexico)  Outward transmission: how domestic banks adjust their foreign lending to changes in domestic monetary policy, including via their affiliates located in other countries (ref. Correa and Murry, 2009, Cetorelli and Goldberg, 2012, Temesvary, Ongena and Owen, 2015) Project Overview - 2 Monetary policy can affect lending through two main channels: 1. Bank lending channel. MP affects the liability side of banks’ balance sheets, i.e. if monetary policy tightens, banks experience a drop in deposits. If banks are not able to mobilize alternative funding, i.e. via wholesale funding markets, they have to adjust the asset side of their balance sheet.  Interaction with deposit funding share or liquidity ratio.  Other constraints include the level of capitalization of the banks.  This channel is less likely to work during times of abundant liquidity (ZLB periods when central banks inject high amounts of liquidity through UMP). Project Overview - 2 Monetary policy can affect lending through two main channels: 2. Portfolio channel / balance sheet channel. MP affects the asset side of banks’ balance sheets. MP changes the supply of assets available in financial markets, i.e. when monetary policy tightens (central bank sells sovereign bonds), the price of sovereign bonds drops and lending becomes relatively less attractive. Banks react to these changes by adjusting the assets side of their balance sheet. MP also affects the riskiness of assets, i.e. in when MP tightens, the creditworthiness of firms and households weakens. Banks therefore reduce lending.  Interaction: lending exposure to domestic borrowers and the composition of assets.  This channel is above all at work during UMP/QE. Project Overview - 2 Pre-view of the empirical results (NB work in progress) Outward Inward transmission transmission Foreign-owned All resident banks All resident banks resident branches Bank Bank Bank Country Portfolio Portfolio Portfolio Lending Lending Lending Austria   Canada   Chile   France       Ireland   Korea     Russia      United Kingdom     United States   BIS ECB  Total 101 92 15 18 47 44 Project Overview -2 Monetary policy variable: we used a range of instruments in FD: - Actual short-term policy rates (e.g. FFR). - A separate QE variable (CB balance sheet / GDP). - shadow policy rates (Krippner, 2015; also Wu and Xia, 2014-16). Note that measures may differ in levels but are similar once expressed in growth rates, which matters here. - A FFR orthogonalized to remove the systematic component from GDP growth and inflation (can be interpreted as Taylor-like residual). The identification relies on heterogeneity at the bank level. Note that the “right” MP indicator is specific to the transmission channel (bank lending channel versus portfolio and balance sheet channel). Also, use dummy variables for structural breaks at ZLB. Project Overview -2

Dependent variables: • Type of instrument. Here we focus on bank lending. • Counterparty sector:  For the inward exercise: domestic loans to the private non- bank sector  For the outward exercise: local lending by affiliates to the non-bank private sector Project Overview -3 Transmission channels: the identification consists in obtaining a bank specific shock Project Overview -3 The main regressions take the following form:

For inward transmission:

The dependent variable in this case is the log change of lending to the private non-bank sector by bank b at time t Project Overview -3 The main regressions take the following form:

For outward transmission:

The dependent variable in this case is local lending by the affiliate of the domestic bank b in country j at time t or an alternative variable as shown in table 1 above Project Overview - 4

Found Significant Differential Spillovers in Some Specification at the 10% Level:

Outward Inward transmission transmission Foreign-owned resident All resident banks All resident banks branches Monetary Policy Bank Bank Bank Portfolio Portfolio Portfolio Choice Lending Lending Lending   Any         Short Rate X ZLB         Short Rate+QE         Taylor-like Residual         ΔSSR X ZLB       The international transmission of monetary policy: evidence from the UK Robert Hills

IBRN/IMF Conference: The Transmission of Macroprudential and Monetary Policies Across Borders: 19 April 2017 Session 2: Research panel on international banking and monetary policy spillovers with Dennis Reinhardt and Rhiannon Sowerbutts Characteristics of the UK financial system

• Major global financial hub • Large presence of foreign banks • Direct cross-border links are high and varied • By currency: USD accounts for highest proportion of both cross-border lending and borrowing

• Quarterly data, 2000 Q4 to 2015 Q4 • 253 UK-resident banks as at 2015 Q4 • o/w 57 UK-owned, 120 foreign-owned branches, 76 foreign- owned subs Transmission channels

• A tightening of monetary policy abroad can, in principle, impact on lending to the UK real or financial sectors positively or negatively, depending on the relative strength of:

• Portfolio channel (banks increase lending abroad) • Bank funding channel (banks decrease lending abroad)

• Our prior for the UK: – high dependence on cross-border wholesale funding implies that the bank funding channel might be sizable – but UK banks receive funding from multiple sources, so might be easier to replace Lending to UK non-financial sectors

(1) (2) Net cross-border Cross-border lending borrowing from Ctry to Ctry (% of Assets) (% of Assets)

ΣΔMP US_t to t-3(* Channel) 0.121* -0.00707 0.0527 0.888 ΣΔMP EA_t to t-3(* Channel) 0.0193 0.00799 0.670 0.836 ΣΔMP JP_t to t-3(* Channel) 0.169** 0.0142 0.0242 0.891

Sum of Coefficients on all foreign ΔMP*Channel 0.310*** 0.0151 0.00240 0.905

• Banks with a higher share of their claims in the US experience a more positive change in their bank lending growth to UK real sectors • Banks that are more dependent on funding from the US show a smaller (though insignificant) increase in lending to the UK real economy

16 Lending to UK financial sectors

(1) (2) (3) Net cross-border Cross-border lending ...from Ctry Banks borrowing from Ctry to Ctry (% of Assets) (% of Assets) (% of Assets)

ΣΔMP US_t to t-3(* Channel) -0.0270 -0.0203 -0.141* 0.641 0.690 0.0914 ΣΔMP EA_t to t-3(* Channel) 0.00603 -0.00641 -0.0191 0.898 0.906 0.674 ΣΔMP JP_t to t-3(* Channel) -0.291*** 0.0470 -0.210 0.00261 0.742 0.117

Sum of Coefficients on all foreign ΔMP*Channel -0.312** 0.0203 -0.371** 0.0102 0.905 0.0239

• Banks more dependent on net interbank funding from the US tend to reduce their interbank lending in the UK by more than other banks, once the US tightens monetary policy

17 The UK as an entrepôt?

18 The UK as an entrepôt?

(1) (2) (3) (4) (5) (6) LHS: Lending to ex US. MP: US LHS: Lending to ex EA. MP: EA LHS: Lending to ex JP. MP: JP Channel: Gross Assets Net funding Gross Assets Net funding Gross Assets Net funding

ΣΔMP CTRY _t to t-3 * Channel CTRY -0.00268 0.00502 0.0578** -0.0158 -0.0644 -0.215 0.960 0.925 0.0460 0.625 0.401 0.625 Channel CTRY 0.0653 -0.00312 0.0876*** -0.0854*** 0.163* -0.0529 (0.0413) (0.0389) (0.0257) (0.0199) (0.0856) (0.0501)

Time fixed effects Yes Yes Yes Yes Yes Yes Bank fixed effects Yes Yes Yes Yes Yes Yes Observations 11,707 11,446 12,035 11,743 8,032 7,970 R-squared 0.067 0.070 0.047 0.049 0.064 0.063 Number of banks 335 330 357 350 222 221

• Tentative evidence that banks exposed to €A transmit ECB monetary policy to the rest of the world in line with the portfolio channel

19 Next steps: joint paper with HKMA

• Working jointly with HKMA, comparing results

• Why UK and HK? Both are important financial centres – Global funding/lending hubs with diverse financial linkages – Wide variety of foreign banks, different organisational forms – Role of FX-denominated lending International Bank Lending Channel of Monetary Policy: Evidence from Chile Alejandro Jara

IBRN/IMF Conference The Transmission of Macroprudential and Monetary Policies Across Borders 19 April 2017

Disclaimer: the opinions expressed in this presentation are of my own and do not necessary represent the view of the Central Bank of Chile Some key characteristics of the Chilean banking system

• A relatively simple banking system’s balance sheet – Dominated by commercial banks. – Lending accounts for more than 70% of total assets. – A relatively simple funding structure • Demand and time deposits account for more than 50% o total funding. • Institutional investors are a very relevant source of funding (deposits and bonds holdings). • Low reliance on cross-border borrowing. – Significant presence of foreign banks, although as subsidiaries – Banks’ direct cross-border exposure is concentrated in the US and the Eurozone Preliminary findings

• In our study we … – Focus on internationally exposed banks that are relevant to the domestic lending market – Only big and medium banks (12 banks) – Benchmark regressions focus on US and EA monetary policy

• Finding #1: Direct effects of a tightening in foreign monetary policy reduces bank lending domestically – Robust to different type of credit and measures of MP shock Finding #1: A negative direct effect of foreign monetary policy into domestic bank lending growth Preliminary findings

• Finding #2: (contrary to expected) the funding channel through cross-border exposure is positive – A tightening of monetary policy overseas increases lending growth rates (except for retail consumption loans) – These results are somehow robust to different measures of cross- border banking exposure (gross versus net) and whether the exposure is aggregate or through intra-group exposures – We argue that this might be due to the role played by the institutional investors • It is well documented (IMF 2013) that under certain conditions gross flows might be upset by the retrenchment of foreign assets • In Chile, pension funds might be behind this dynamic as we (partially) show below Finding #2: Cross-border funding exposure channel is positive, contrary to what is expected Role of institutional investors as a channel of MP Preliminary findings

• Finding #3: portfolio rebalancing channel either weak or not significant – Different results depending on the type of loans and the measure of monetary policy shock considered Portfolio rebalancing channel is not significant Preliminary findings

• Finding #4: Banks’ characteristic channels can be summarized as follows: – Higher exposure to short-term funding increases the negative effect of a tightening in foreign monetary policy – Direct (negative) effect is less significant for larger banks – No significant effect through the capital adequacy ratio Next steps

• Working paper with the Chilean experience (co-authored with David Moreno) • For journal submission, Chilean team is working with Korea and Poland. – These three countries are similar in size, open economies, with complementary regional exposures. – Joint paper focused on (i) portfolio composition and (ii) the risk- taking channel. • Other issues and robustness exercises – Deepen the role of the institutional investors channel. – Look at impact on banks’ interest rates and other non-price terms of lending Characteristics of the Chilean banking system Institutional investors are a relevant source of funding though deposits and bonds’ holdings

Households Firms Mutual Funds Pension Funds Banks Others Domestic bonds External bonds Foreign loans Subordinated bonds 100

80

60

40

20

0 08 09 10 11 12 13 14 15 The structure of Chilean banks Relevant jurisdictions for foreign exposure Additional slides Project Overview -1 Project Overview -3 Balance sheet characteristics: . log of total real assets, i.e. assets deflated by GDP deflator ( , ) . percentage of banking organization’s regulatory Tier 1 risk-based capital 𝐿𝐿𝐿𝐿𝐿𝐿 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑏𝑏 𝑡𝑡−1 to asset ratio , . percentage of a bank’s portfolio of assets that is liquid 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇푇 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑏𝑏 𝑡𝑡−1 , . percentage of banking organization’s net intragroup funding, which is 𝑏𝑏 𝑡𝑡−1 defined𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 as𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 the𝑎𝑎𝑎𝑎 liabilities𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 minus the assets of the Head Office with the rest of the banking group, scaled by total assets ( , ) . percentage of the banking organization’s balance sheet financed with 𝑁𝑁𝑁𝑁𝑁𝑁 𝐼𝐼𝐼𝐼 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑏𝑏 𝑡𝑡−1 core deposits ,

𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑏𝑏 𝑡𝑡−1 Project Overview - 4

Found Significant Differential Spillovers in Some Specification at the 10% Level:

Inward Transmission via Resident Banks

Bank Lending Portfolio

US EA JP UK US EA JP UK # Significant 43 21 23 25 37 27 22 22 # Studied 97 65 55 75 92 73 54 60 % Significant 44% 32% 42% 33% 40% 37% 41% 37%