Geopolitics Meets Monetary Policy: Decoding Their Impact on Cross-Border Bank Lending

Board of Governors of the Federal Reserve System International Finance Discussion Papers ISSN 1073-2500 (Print) ISSN 2767-4509 (Online) Number 1403 February 2025 Geopolitics Meets Monetary Policy: Decoding Their Impact on Cross-Border Bank Lending Swapan-Kumar Pradhan, Viktors Stebunovs, El˝od Tak´ats, and Judit Temesvary Please cite this paper as: Pradhan, Swapan-Kumar, Viktors Stebunovs, El˝od Tak´ats, and Judit Temesvary (2025). “Geopolitics Meets Monetary Policy: Decoding Their Impact on Cross-Border Bank Lending,” International Finance Discussion Papers 1403. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/IFDP.2025.1403. NOTE: International Finance Discussion Papers (IFDPs) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the International Finance Discussion Papers Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers. Recent IFDPs are available on the Web at www.federalreserve.gov/pubs/ifdp/. This paper can be downloaded without charge from the Social Science Research Network electronic library at www.ssrn.com.

Geopolitics Meets Monetary Policy: Decoding Their Impact on Cross-Border Bank Lending* Swapan-Kumar Pradhan Viktors Stebunovs Bank for International Settlements Federal Reserve Board Basel, CH Washington, DC, USA Swapan-Kumar.Pradhan@bis.org Viktors.stebunovs@frb.gov Előd Takáts Judit Temesvary ** Bank for International Settlements Federal Reserve Board Basel, CH Washington, DC, USA elod.takats@bis.org Judit.temesvary@frb.gov January 2025 Abstract: We use bilateral cross-border bank claims by nationality to assess the effects of geopolitics on cross-border bank flows. We show that a rise in geopolitical tensions between countries — disagreements in UN voting, broad sanctions, or sentiments captured by geopolitical risk indices — significantly dampens cross-border bank lending. Elevated geopolitical tensions also amplify the international transmission of monetary policies of major central banks, especially when geopolitical tensions coincide with monetary policy tightening. Overall, our results suggest that geopolitics is roughly as important as monetary policy in driving cross-border lending. Keywords: Monetary policy; Geopolitical tensions; Cross-border claims; Diff-in-diff estimations JEL Codes: E52; F34; F42; F51; F53; G21 * The views expressed in this paper are solely those of the authors and shall not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve System or of the Bank for International Settlements (BIS). We thank Jessica Ye for excellent assistance. We are grateful for comments from Dario Caldara and Goetz von Peter, from colleagues at the BIS and the Federal Reserve Board, and from seminar participants at the BIS, Hamilton College, the Hungarian Economic Association and the International Monetary Fund. The BIS confidential data was obtained by Swapan-Kumar Pradhan and Előd Takáts under the purview of their association with the BIS. The remaining co-authors, Viktors Stebunovs and Judit Temesvary, did not have any unauthorized access to this data while working on this paper/project. ** Corresponding author. 1

1 Introduction Geopolitical risks and tensions have soared over the past decades: we have witnessed the proliferation of geopolitical fragmentation and even wars. These geopolitical tensions threaten economic activity as they drive uncertainty higher and divert trade and investments along geopolitical fault lines. The realization of geopolitical risks, such as sanctions or wars, further weighs on macroeconomic outcomes across the world. Notwithstanding, the effects of geopolitics in shaping capital flows, in particular bank flows, have been little studied so far. Indeed, how large is the impact of these geopolitical effects on cross-border bank lending? Do they strengthen or weaken the impact of monetary policy of major central banks on cross-border bank lending? We study these questions by focusing on three measures of geopolitical tensions and risks: (1) UN voting disagreement between country pairs, captured by an ideal point distance following the Bailey et al. (2017) methodology, which we consider a measure of materialized geopolitical tensions; (2) trade, financial, military, and other bilateral sanctions, which serve as another measure of materialized geopolitical tensions; and (3) a potential precursor of geopolitical fragmentation and broad sanctions: geopolitical risk in lender and borrower countries, captured by Caldara and Iacoviello (2022)’s geopolitical risk indices (GPRs). We find that geopolitics affects cross-border bank flows in an economically and statistically significant way. The rise in geopolitical tensions directly dampens cross-border bank lending and also amplifies the international transmission of monetary policy. Both the direct effects and the interaction effects with monetary policy are stronger for materialized geopolitical tensions (i.e. bilateral UN voting disagreement and bilateral sanctions) than for unrealized geopolitical tensions (as measured by the difference in GPRs of country pairs or by GPRs of borrower countries). Specifically, we show that UN voting disagreement has the largest effect, followed by sanctions. 2

To provide context, we also estimate the international transmission of monetary policy of major central banks, identified in Takats and Temesvary (2020). These monetary policy effects provide a benchmark for geopolitical effects: the results suggest that geopolitics is as significant as monetary policy in driving cross-border bank lending. We investigate the joint effects of geopolitical tensions and monetary policy based on the bank lending channel (Kashyap and Stein, 2000). The bank lending channel posits that a rise in interest rates, and the subsequent tightening in liquidity conditions affect constrained banks more. The intensification of geopolitical tensions could further affect constrained banks more, as they might be perceived to be even riskier in the new environment - and as such, these banks might find acquiring additional liquidity more costly. Hence, constrained banks could cut their lending even more when geopolitical tensions and monetary tightening coincide. Our empirical results support the bank lending channel-based theory: geopolitical tensions amplify the international transmission of monetary policy and the interaction is particularly strong when a rise in geopolitical tensions coincide with monetary policy tightening. We show that the interaction effect of monetary policy and geopolitics explains nearly as much of the variation in bilateral lending flows as monetary policy alone does – and is particularly potent in the context of rising interest rates and worsening geopolitical tensions. The interaction effects are again stronger for materialized geopolitical tensions than for unrealized tensions. Our unique identification strategy relies on the currency dimension of the international bank lending channel: monetary policy of a currency issuer will affect cross-border flows in that currency even when neither the lender banking system nor the borrowers’ country uses the currency as its own. In other words, we look at cross-border bank lending flows between thirdcountry pairs. As an example, we look at how U.S. monetary policy interacts with geopolitical 3

tensions between the U.K. and Russia in driving U.K. banks’ dollar lending to borrowers in Russia. We posit that monetary policies of reserve currency issuers are independent of geopolitical tensions among third-party countries. In our example, U.S. monetary policy is independent of the geopolitical tensions between the U.K. and Russia. Therefore, our approach avoids confounding monetary policy and geopolitical tensions.1 Our identification strategy is afforded by detailed data on the network of cross-border bank claims of lending banking systems on bank and non-bank borrowers in individual foreign countries by currency denomination (USD, EUR, JPY, GBP and CHF).2 These data are only accessible at the BIS. We combine the bank flow data with (1) country pair-specific quarterly measures of geopolitical tensions and risk; and (2) with shadow policy interest rate measures for USD, EUR, JPY, GBP and CHF from Krippner (2024). Our findings are robust to extensive robustness checks. The results hold across lending to both financial and non-financial borrowers; across borrowers in advanced and emerging economies; and when accounting for cross-currency monetary policy effects and common trends in geopolitical risk. 1 To further strengthen our identification, we exclude each reserve currency issuer country’s banking system’s lending in their own currency. As an example, we exclude U.S. banks’ lending in U.S. dollars which could have confounding effects with U.S. monetary policy. We also control for source country monetary policy and currency valuation. Finally, as fiscal policy has notable effects on monetary policy transmission (Pradhan et al, 2024), we account for fiscal policy effects via inclusion of fiscal controls (in levels and interactions) and extensive fixed effects. 2 We use granular data from the Stage 1 and Stage 2 enhancements to the international locational banking statistics by nationality (LBSN) of the BIS. Our data is characterized as “unrestricted” – by definition, including all confidential observations that reporting countries provided for use only by the BIS. Stage 1 enhancements include a breakdown of counterparties by country and local currency positions by bank nationality, starting from 2012:Q2, also covering counterparty sector breakdowns such as banks, interoffice, central banks, unrelated banks, and aggregated nonbanks. Stage 2 enhancements, introduced in 2013:Q4, add a subsector breakdown for the nonbank sector, distinguishing between non-bank financial institutions and non-financial sectors, with further details, on an encouraged basis, for corporates, governments, and households. 4

Our results are policy relevant. For policy makers in reserve currency-issuing countries, understanding the effects of geopolitical tensions on monetary policy transmission can help gauge changes in global liquidity conditions in their currency. For policy makers in the source countries of lending banks, understanding the effects of geopolitical tensions can help gauge cross-border bank lending activities of their banks and thus, domestic credit conditions. For policy makers in borrowers’ countries, understanding the effects of geopolitical tensions can help gauge credit supply via cross-border bank lending to their country, to better manage periods of volatile bank flows. The paper proceeds as follows. In Section 2, we review our contributions in the context of the related literature. In Sections 3 and 4, we describe the data and methods. In Sections 5 and 6, we detail results, discuss implications, and offer robustness checks. We conclude in Sections 7. 2 Literature review and hypothesis development We develop our hypotheses in the context of two strands of the literature: 1) papers on the bank lending channel and its international extension; and 2) studies of the effects of various factors, including geopolitical risk and tensions, on international financial capital flows. We also draw on concepts, hypotheses, and data from other literature strands. The concept of the bank lending channel of monetary policy in the domestic context originates from Kashyap and Stein (2000). The bank lending channel posits that a rise in monetary policy rates increases the cost of borrowing for banks across the board; however, balance sheetconstrained banks (e.g. those with lower liquidity or capital) see a larger cost increase, due to being perceived as riskier by investors in financial markets. As a result, these banks cut their lending 5

more than their unconstrained peers. Subsequently, papers on the international impact of domestic monetary policy have identified cross-border bank lending as a spillover channel (Cetorelli and Goldberg, 2012; Forbes and Warnock, 2012; Bruno and Shin, 2015a; 2015b; Temesvary et al., 2018). Focusing on the bank lending channel, Takats and Temesvary (2020) identify the currency dimension of the international bank lending channel (CDIBL): a rise in interest rates associated with a reserve currency reduces cross-border lending in that currency across the globe, even among counterparties that do not use that currency as their own. More broadly, studying lending in various currencies, several papers have shown that the monetary policy of a currency issuer can also transmit into lending in that currency in foreign countries via various channels (Ongena et al., 2021; Avdjiev and Takats, 2019). Based on the CDIBL, our Hypothesis 1 posits that a tightening in the monetary policy associated with a reserve currency of lending leads to subsequently lower bilateral cross-border lending flows in that currency. These effects can be particularly strong for banking systems exposed to heightened geopolitical risk. These banks, due to the heightened uncertainty arising from geopolitical escalation, can see a disproportional rise in funding costs in global financial markets, causing them to adjust their lending flows more. Therefore, we expect the negative lending effects of monetary policy to be stronger among country pairs with higher geopolitical tensions or risk. The second strand of literature that we build upon focuses on the impact of factors other than monetary policy on cross-border lending. While a large body of literature has studied source and borrowers’ country-specific drivers of cross-border bank lending (De Haas and van Lelyveld, 2014; Rose and Wieladek, 2014; Cetorelli and Goldberg, 2012; Giannetti and Laeven, 2012; De Haas and van Horen, 2012; Buch et al., 2014; Cerutti et al., 2014; Cerutti et al., 2015), papers that 6

examine the role of geopolitical risk and tensions in banks’ cross-border lending decisions are still relatively scarce. For example, Catalan et al. (2024) analyze the effects of geopolitical tensions on capital flows in a gravity model and show that rising geopolitical tensions lead to a decline and diversion of investment. Of lesser relevance for us, Goldberg and Hannaoui (2024) and Ferbermayr et al. (2020) study how geopolitical tensions and financial sanctions, respectively, affect the share of U.S. dollars in foreign official reserves. Niepmann and Shen (2024) show that when geopolitical risk increases, domestic lending by U.S. banks is negatively affected. Other strands of the literature provide more ground for hypothesis development. For example, in the international trade literature, Bosone and Stamato (2024) show that geopolitical fragmentation weighs on international trade in manufactured goods. Febermayr et al. (2020) introduce a comprehensive global sanctions database. Syropoulos et al. (2024) update this database and document a dramatic increase in the number of sanctions over the 2019-22 period. The authors also apply a gravity model and find that bilateral trade sanctions significantly limit international trade. Afesorgbor (2019) studies the differential effects of threatened vs. imposed sanctions. In the macroeconomic literature, Caldara and Iacoviello (2022) develop a seminal news-based measure of geopolitical risks and show that such risks cause declines in employment and economy-wide and firm-level investment. Wang et al. (2019) show a negative relationship between geopolitical risk and firm-level investment, too. On the finance side, Alfonso et al. (2024) find that geopolitical tensions contribute to the rise of European countries’ sovereign risk and that this relationship is more pronounced during turbulent times. Yilmazkuday (2024) shows that an adverse shock to global geopolitical risk reduces stock prices in the year following the shock in a number of countries, and that this stock price response depends on the country’s involvement in the geopolitical event. 7

Building on the above literature, in our Hypothesis 2, we posit a negative relationship between our measures of geopolitical tensions/risk and cross-border bank flows. The conjectured negative relationship between geopolitical tensions/risk and bank lending flows is consistent with Catalan et al. (2024)’s findings for UN voting disagreement. Furthermore, based on the two streams of literature described above, we conjecture in our Hypothesis 3 that the negative connection between increasing geopolitical tensions/risk and crossborder bank lending is particularly strong in tightening monetary policy environments. This hypothesis is novel but is also intuitive: borrower economies face a double whammy as escalating geopolitical tensions boost uncertainty. Tighter financial conditions make it harder to cope with this increased uncertainty, as contractionary monetary policy aggravates the cost of acquiring liquidity. Therefore, following a monetary policy tightening, banks cut back lending to borrowers in countries affected by geopolitical tensions especially hard. 3 Data 3.1 Dependent variable: Cross-border bank flows We use granular bilateral data from the BIS international banking statistics by nationality (LBSN) (see, Takats and Temesvary, 2020; 2021). This dataset includes restricted (only for sharing among reporting countries) as well as confidential observations (that reporting countries provide only for use by the BIS.3 It also offers a breakdown of counterparties by country and local currency positions by bank nationality, starting from 2012:Q2. The dataset covers counterparty sector breakdowns such as banks, interoffice, central banks, unrelated banks, and aggregated nonbanks. 3 However, since late 2015, the BIS releases some of these data, but with a limited scope for confidentiality reasons (see Avdjiev et al., 2015). A significant share of the reported bilateral data remains restricted and confidential. 8

Beginning 2013:Q4, the data include a subsector breakdown for the nonbank sector, distinguishing between non-bank financial institutions and non-financial sectors. Our use of nationality-based data rather than residence-based data is suitable as we assess that the strongest geopolitical effects on a bank occur at the level of the decision-making unit, i.e. the banking conglomerate as a whole. As an example, a bank will react to a sanction imposed by its headquarter jurisdiction more strongly than to a similar sanction imposed in the jurisdiction of one of its subsidiaries. In other words, we are interested in how geopolitical developments in the home country of the parent bank, on a consolidated basis, affect its lending decisions vis-à-vis borrowers’ countries.4 We focus on major lenders among advanced economies that include the U.S. and European bank lending systems. Our lending sample consists of bilateral cross-border exposures of these lending banking systems to borrowers in over 180 countries during the 2012:Q2–2023:Q4 period. As described above, for each lending banking system and country of borrowers, our dataset is broken down by currency denomination and borrower sector. We focus on the top five currencies of global lending (USD, EUR, JPY, GBP, and CHF) and the two main target sectors of borrowers (banks and non-banks). We also separate out the interoffice sub-category from the “banks” target sector and delineate non-bank financial institutions (NBFIs) from the non-financial sector (NFS) in the “non-banks” target sector. This dataset is unique as it simultaneously provides an overlay of the four dimensions that we need to answer our research questions: (A) the currency composition of cross-border claims; (B) the residence of borrowers, (C) the sector breakdown of borrowers, and (D) the nationality of lending banking systems. Dimension (A), currency composition, allows us to map the relevant 4 Other papers in the banking literature also posit that lenders conduct risk management at the highest level of consolidation—for example, see Lee et al. (2022). 9

networks and flows in each currency, that is, to map bilateral claims in USD, EUR, JPY, GBP, and CHF and their evolution over time, purged of valuation effects. Dimension (B), the residence of borrowers, enables us to account for the (borrowers’) country-specific drivers of cross-border bank lending. As such, we can even apply borrowers’ country*time fixed effects in most of our estimations to account for changes in credit demand. Dimension (C), the sector of borrowers, allows us to identify effects across sectors, an important feature as the bank and non-bank sectors can have notably different economic relevance. Dimension (D), the lender’s nationality, enables us to identify the headquarter, i.e. the highest-level banking entity in the corporate chain, of the lending banking systems. This allows us to identify the decision-making unit (Fender and McGuire, 2010; Cecchetti et al., 2010; Committee on the Global Financial System, 2011) and to control for the possible confounding effects of financial centers. While we do not focus on the role of fiscal policies, we still control for fiscal effects as the literature shows they are important determinants of cross-border bank flows (Pradhan et al., 2024). Our sample set of source banking systems is defined by the availability of consistent data coverage for fiscal statistics. We concentrate on the 16 advanced-economy lending banking systems in the Eurostat database and add the United States – therefore, our home (source) countries encompass the two largest currency areas, the USD and the EUR. The included set of source countries (European Union countries; Nordic countries; and the United States) make up over 50 percent of total cross-border bank claims (54 percent of claims on banks and 56 percent of claims on nonbanks).5 In our estimations, we exclude claims that are denominated in the banking system’s own 5 While other proprietary sources contain imputed quarterly values for a small set of additional countries, their calculations rely on assumptions and source-specific methods; therefore, we choose to focus solely on the set of countries included in the Eurostat coverage and the United States, to ensure data consistency. In additional regressions (available by request) we repeat our estimations on the full sample and find consistent results. 10

currency (for instance, we exclude euro area banks’ EUR claims, due to policy endogeneity concerns). The currency composition of claims in our sample is closely comparable to the composition observed in the full set of countries. As Graph 1.C shows, among the five currencies on which we focus in our sample, the USD and EUR are clearly dominant, with shares of 49 percent and 28 percent, respectively at end-2022 (comparable respective shares in the full data are 51 percent and 36 percent). The other three currencies in our sample have notably lower shares: the GBP, JPY, and CHF make up 12 percent, 8 percent, and 3 percent, respectively, at end-2022 (Graph 1.C). In terms of borrowers’ sectors, lending to banks and non-banks make up 55 percent and 45 percent of claims in our sample, respectively, at end-2022 (the sectors have about equal shares in the full data). Since 2012, the share of claims on banks has declined and the share of claims on non-banks has increased in both our sample and the full data. Graphs 1.A and 1.B show the currency breakdown of claims by target sector over time. We define bilateral cross-border lending flows (the main outcome/dependent variable of interest) as the quarterly percent change in bilateral cross-border bank claims from a source banking system to borrowers in a given country, denominated in one of the five reserve currencies. Importantly, we adjust flows for the effects of exchange rate changes as follows: before we calculate the quarterly percent changes in bilateral claims, we convert the (reported) dollar value of claims back to the original currency amount, using the contemporaneous exchange rate between the USD and the original currency of lending. There is substantial variation in quarterly (exchange rate-adjusted) cross-border lending flows. The average quarterly bilateral flow (in quarterly percentage change) is -0.13 percent and 11

has a standard deviation of 55 percent (Table 1). Across countries, the average flows vary over time as well, ranging from -5 percent to 5 percent at times. 3.2 Changes in monetary policy For part of our sample period, unconventional/balance sheet-focused monetary policy actions by the Federal Reserve, the European Central Bank, the Bank of Japan, the Bank of England, and the Swiss National Bank drove policy rates to zero or into negative territory. Therefore, to measure monetary policy changes associated with these five currencies, we cannot simply use changes in the headline policy interest rates. Given that the interaction and transmission of monetary policy effects can be very different during unconventional monetary policy regimes (Takats and Temesvary, 2020) and fiscal policy regimes (Hofmann et al., 2021; Wang, 2018), it is important to capture liquidity conditions accurately even when the policy interest rate is at the zero lower bound. Therefore, as is now standard in the related banking literature (Buch et al. 2019, Temesvary et al., 2018; Lhuissier et al., 2019, among others), we use shadow interest rates to measure changes in financial market liquidity conditions related to monetary policy actions during periods of binding effective lower bounds. We employ shadow rates constructed by Krippner (2024) which are available consistently across the five major reserve currencies over our full sample period. In robustness checks, we employ the Wu-Xia shadow rates (Wu and Xia, 2016) as alternative measures; however, these shadow rates are available for only a subset of the currencies that we examine. As the short-term shadow rates are not subject to the zero lower bound (ZLB), they can capture expansionary monetary policy actions by turning negative (Graph 2). By construction, the 12

shadow rates are nearly identical to the policy rates during conventional (non-ZLB) periods, and negative in times of binding ZLB. All five shadow rates fell below zero during the period when monetary conditions continued to ease, and the nominal policy interest rates hit the zero lower bound. During our sample period, the average short-term shadow rate was -0.69 percent; in contrast, the average central bank policy rate for the major reserve currencies was 1.32 percent. We measure changes in the monetary policy stance as quarterly changes (from one quarter to the next, in percentage points) in the currency-specific shadow interest rates. Across currencies, monetary policy was characterized by a slightly contractionary stance in our sample (albeit among broadly ample liquidity conditions), with average quarterly increases of 12 basis points, ranging from -1.8 to 2.6 percents in the sample (Table 1). 3.3 Measures of geopolitical tensions and risks 3.3.1 UN voting disagreement (IPD) To measure political disagreement between country pairs, we use Bailey et al. (2017)’s estimated absolute distances between the “ideal points” of country pairs. The political science literature defines an ideal point as a (latent) ideological position of an actor on a political spectrum, estimated from discrete choice models. The absolute distance between a pair of ideal points is then a natural measure of political disagreement between two actors. Turning to Bailey et al. (2017), they propose a novel dynamic ordinal spatial model to estimate ideal points for countries on a single dimension that reflects country positions toward the U.S.-led liberal order based on United Nations General Assembly (UNGA) votes. Their approach is particularly appealing because it controls for the content of the UNGA’s voting agenda and thus it does a better job at separating signal from noise in identifying foreign policy shifts than earlier approaches (for example, the S- 13

score approach, which at times fails at observational validity). The Bailey et al. (2017) measure has been widely used in the political science literature, and it is becoming increasingly prevalent in the international economics literature as well (Catalan et al., 2024; Goldberg and Hannaoui, 2024). IPDs vary broadly across country pairs and over time, as shown by the descriptive statistics in Table 1. 3.3.2 Bilateral sanctions We use measures of bilateral sanctions (including total, financial, military, travel, trade, and other sanctions) from Felbermayr et al. (2020) and Syropoulos et al. (2024). The Global Sanctions Database tallies bilateral and multilateral sanctions globally, across three dimensions: type, political objective, and extent of success. Through 2016, the use of sanctions had increased, as sanctions became more diverse, and the share of trade sanctions fell. The period from 2019 to 2022 brought a notable rise in total sanctions due to new impositions by the United States, with the biggest increase in 2021. Financial and trade sanctions became increasingly prevalent over the past decade, the latter driven by the sharp increase in trade sanctions on Russia. European countries are the most frequent imposers of sanctions. 3.3.3 Geopolitical Risk Index (GPR) We use Caldara and Iacoviello (2022)’ GPRs by country to quantify geopolitical risks. Their index is a news-based measure of adverse geopolitical events and associated risks, with higher geopolitical risk foreshadowing lower investment and employment and implying higher disaster probability and larger downside risks. Quarterly changes in the sample GPRs vary from - 5.95 to 7.17, with a standard deviation of 0.52 (Table 1). 14

3.4 Fiscal controls As discussed above, our data coverage is defined by the consistent availability of data for an important control variable: the fiscal stance of source countries (Pradhan et al., 2024). Our fiscal measure is defined as quarterly changes in (source) country government debt-to-GDP ratios (in percentage points) from the Eurostat statistical database and from FRED. Across countries and over time, in our sample government debt-to-GDP ratios stood at 88 percent; but with substantial variation, ranging from 20 percent to over 200 percent. The quarterly change in debt-to-GDP ratios also ranged widely, from a decline of 10 percentage points to an increase of 26 percentage points across all countries and time periods (Table 1). 3.5 Exchange rate adjustments and other controls We take several steps to control for valuation effects arising from the data’s feature that the claims are reported after conversion to U.S. dollars. As discussed above, we calculate the quarterly bilateral flows only after converting the claims back to the original currency amount at contemporaneous exchange rates. In addition, we include quarterly changes in the exchange rate between the USD and the currency of lending as a control. On average and across currencies, the USD appreciated slightly against the other reserve currencies. Furthermore, we include changes in the bilateral exchange rate between the currencies of the source country and borrowers’ country among our controls, as such valuation changes can have important confounding effects on the strength of transmission and policy interactions (Leith and Wren-Lewis, 2008), including the possibility that foreign assets becoming cheaper due to a domestic currency appreciation might be driving lending outflows. Across currencies and on 15

average, we saw an appreciation of the source country currency relative to the currency of the borrowers’ country during our sample period. Lastly, we add quarterly changes in the central bank policy rate of the source country of banking systems, as controls. During the sample period, on average, the central bank policy rates of the source lending systems increased by 58 basis points per quarter (Table 1). 4 Estimation methodology A challenge in our estimations is the endogeneity of monetary policy to geopolitical risk and developments, as we describe above. In order to address this endogeneity, when we investigate the interactions of monetary policy effects with the impact of changes in bilateral geopolitical risk or tensions, we need to focus on a monetary policy that is not connected to and is not affected by the geopolitical situation between source bank lending systems and borrowers’ countries. Therefore, we focus on the monetary policy of the issuer of the reserve currency and not that of the country of the lending banking system, as in Takats and Temesvary (2020; 2021) or Pradhan et al. (2024). Our main dependent variable of interest is quarterly changes in bilateral cross-border claims. This variable, Δclaims is the quarterly change in the natural logarithm of bilateral claims between the source lending banking system and the borrowers’ country, denominated in one of the five reserve currencies. Our main explanatory variables are (1) the change in the source and borrowers’ country-specific (bilateral) geopolitical measure (GeoPol), as defined in Section 3 above, and (2) the change in the monetary policy stance (monetary) associated with the major currencies of lending (USD, EUR, JPY, GBP, and CHF) as measured by the Krippner (2024) shadow interest rates. The identification assumption is that the monetary policy of the currency 16

issuer is not connected to and is not affected by the geopolitical situations between source bank lending system and borrowers’ countries. We consistently add four lags of the dependent variable to the set of regressors to address possible time persistence. To avoid using observations where common factors influence both monetary policy and bank lending, we exclude own currency lending from all our estimations (as, for example, domestic economic developments can drive both U.S. monetary policy and U.S. banks’ USD lending decisions). We also exclude “same country” lending (in the terminology of Takats and Temesvary, 2020) – lending relationships in which foreign subsidiaries of global banks lend back to their home country. The reason to exclude such same country lending is that it may be driven by liquidity management considerations unrelated to geopolitical tensions. 4.1 Baseline estimations Our benchmark estimation examines bank lending flows as a function of changes in 𝛥𝛥𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 geopolitical measures between bank lending system i and borrowers’ country j ( ), as 𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖 well as a function of the monetary policy by currency issuer c ( ). We formulate Equation (1) as: 𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝑖𝑖𝑖𝑖 (1) 4 𝛥𝛥𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = ∑𝑘𝑘=1(𝛾𝛾1𝑘𝑘𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝑖𝑖𝑖𝑖−𝑘𝑘 + 𝛾𝛾2𝑘𝑘𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖−𝑘𝑘 + + 𝛾𝛾3𝑘𝑘𝛥𝛥𝑐𝑐𝛥𝛥_𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝑖𝑖𝑖𝑖−𝑘𝑘 + 𝛾𝛾4𝑘𝑘𝐶𝐶𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝑐𝑐𝑖𝑖/𝑖𝑖/𝑖𝑖/𝑖𝑖−𝑘𝑘)+𝐹𝐹𝐹𝐹𝑖𝑖/𝑖𝑖/𝑖𝑖/𝑖𝑖 + 𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 Our set of control variables in include (a) monetary policy changes 𝐶𝐶𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝑐𝑐𝑖𝑖/𝑖𝑖/𝑖𝑖/𝑖𝑖−𝑘𝑘 associated with the source lending system i ( ); (b) valuation effects between the USD and the currency of lending c ( )𝛥𝛥; (𝑐𝑐c𝛥𝛥)_ v𝛥𝛥a𝑐𝑐l𝛥𝛥u𝛥𝛥a 𝑖𝑖 t 𝑖𝑖 ion effects between the currency of source 𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐ℎ_𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝑖𝑖,𝑖𝑖 17

lending system i and that of borrowers’ country j ( ); and (d) changes in the debt-to- 𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐ℎ_𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝑖𝑖𝑖𝑖𝑖𝑖 GDP ratio in source lending system i ( ). 𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥/𝛥𝛥𝛥𝛥𝛥𝛥𝑖𝑖𝑖𝑖 The set includes various combinations of source country i, borrowers’ country j, and 𝐹𝐹𝐹𝐹𝑖𝑖/𝑖𝑖/𝑖𝑖/𝑖𝑖 currency c fixed effects, as well as subsets of source country*borrowers’ country, borrowers’ country*time, and borrowers’ country*currency*time fixed effects. The inclusion of changes in bilateral geopolitical measures that are contemporaneous to changes in reserve currency monetary policy helps to further mitigate endogeneity concerns. We predict the cumulative effects of interest rate changes and bilateral geopolitical changes to be both negative: < 0 and < 0. 4 4 ∑𝑘𝑘=1𝛾𝛾1𝑘𝑘 ∑𝑘𝑘=1𝛾𝛾2𝑘𝑘 In Equation (2), we also include the interaction of our two key explanatory variables: , as follows: 𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝑖𝑖𝑖𝑖 ∗ 𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖 (2) 4 𝛥𝛥𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = ∑𝑘𝑘=1(𝛿𝛿1𝑘𝑘𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝑖𝑖𝑖𝑖−𝑘𝑘 ∗𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖−𝑘𝑘 +𝛿𝛿2𝑘𝑘𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝑖𝑖𝑖𝑖−𝑘𝑘 + + 𝛿𝛿3𝑘𝑘𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖−𝑘𝑘 +𝛿𝛿4𝑘𝑘𝐶𝐶𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝑐𝑐𝑖𝑖/𝑖𝑖/𝑖𝑖/𝑖𝑖−𝑘𝑘)+𝐹𝐹𝐹𝐹𝑖𝑖/𝑖𝑖/𝑖𝑖/𝑖𝑖 + 𝜂𝜂𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 Based on our hypotheses outlined above, we expect to find a negative sum of coefficients on the interaction terms: < 0. 4 ∑𝑘𝑘=1𝛿𝛿1𝑘𝑘 4.2 Estimations by borrowers’ sector Next, we examine how the effect of geopolitical measures and their interactions with monetary policy effects depend on the target sector of lending. We write Equation (3) as: (3) 𝑆𝑆 4 𝛥𝛥𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = ∑𝑘𝑘=1(𝜙𝜙1𝑘𝑘𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝑖𝑖𝑖𝑖−𝑘𝑘 ∗𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖−𝑘𝑘 +𝜙𝜙2𝑘𝑘𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝑖𝑖𝑖𝑖−𝑘𝑘 + υ 𝑆𝑆 + 𝜙𝜙3𝑘𝑘𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖−𝑘𝑘 +𝜙𝜙4𝑘𝑘 𝐶𝐶𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝑐𝑐𝑖𝑖/𝑖𝑖/𝑖𝑖/𝑖𝑖−𝑘𝑘)+𝐹𝐹𝐹𝐹𝑖𝑖/𝑖𝑖/𝑖𝑖/𝑖𝑖 + 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 18

where the superscript S denotes the target sector of lending. As discussed in the hypothesis development above, we expect that monetary policy transmission and geopolitical tensions’ effects vary in importance across target sectors. For instance, bilateral sanctions may affect lending to the non-financial sector more than loans to banks and non-bank financial institutions. 4.3 Country-specific geopolitical risk measures In some estimations, we examine the roles of source and borrowers’ country-specific geopolitical risks separately, rather than the role of bilateral measures. Accordingly, we estimate Equation (4) as follows: (4) 4 𝛥𝛥𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = ∑𝑘𝑘=1(𝛽𝛽1𝑘𝑘𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝑖𝑖𝑖𝑖−𝑘𝑘 ∗𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝑖𝑖/𝑖𝑖𝑖𝑖−𝑘𝑘 +𝛽𝛽2𝑘𝑘𝛥𝛥𝑐𝑐𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝛥𝛥ν𝛥𝛥𝑖𝑖𝑖𝑖−𝑘𝑘 + + 𝛽𝛽3𝑘𝑘𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝑖𝑖/𝑖𝑖𝑖𝑖−𝑘𝑘 +𝛽𝛽4𝑘𝑘𝐶𝐶𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝑐𝑐𝑖𝑖/𝑖𝑖/𝑖𝑖/𝑖𝑖−𝑘𝑘)+𝐹𝐹𝐹𝐹𝑖𝑖/𝑖𝑖/𝑖𝑖/𝑖𝑖 + 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 where is now specific to source country i or borrowers’ country j, rather than a 𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝛥𝑐𝑐𝑖𝑖/𝑖𝑖𝑖𝑖−𝑘𝑘 bilateral measure. 4.4 Saturating the model with increasingly stringent fixed effects As we build up our estimation models, we include increasingly stringent sets of fixed effects to strengthen identification. Specifically, we add various combinations of source country i, borrowers’ country j, and currency c fixed effects, as well as subsets of source country*borrowers’ country, borrowers’ country*time, or borrowers’ country*currency*time fixed effects: • In Model 1, we include time fixed effects for each quarter ( ) to control for unobserved global factors. Our inclusion of time fixed effects con𝐹𝐹tr𝐹𝐹ol 𝑖𝑖 s for time-varying global 19

shocks—for example, a global component of geopolitical tensions—that might confound each policy. We also add in fixed effects for each borrowers’ country j ( ) to capture any 𝐹𝐹𝐹𝐹𝑖𝑖 time-invariant level differences. • In Model 2, we include borrowers’ country* time fixed effects ( ) to control for any 𝐹𝐹𝐹𝐹𝑖𝑖𝑖𝑖 potential time-varying credit demand changes in the borrowers’ country. This fixed effect absorbs any macro-related changes at the level of the borrowers’ country j. Inclusion of these fixed effects expands the logic outlined in Khwaja and Mian (2008), where identification relies on a firm borrowing from different banks. In our analysis, borrowers in a country obtain credit from different source lending systems. This feature allows us to control for borrower-specific demand factors through fixed effects. • In Model 3, in addition to time fixed effects, we include a fixed effect for each source lending system–borrowers’ country pair ( , to capture any potential bias stemming 𝐹𝐹𝐹𝐹𝑖𝑖𝑖𝑖) from historical lending relationships. For instance, this controls for the time-invariant specifics of the U.S.-U.K. lending relationship. • In Model 4, we include a fixed effect for each source lending system–borrowers’ country –time combination ( . This formulation captures any potential bias stemming from 𝐹𝐹𝐹𝐹𝑖𝑖𝑖𝑖𝑖𝑖) historical lending relationships (subsuming Model 3). In addition, it accounts for unobservable shocks to credit demand (subsuming Model 2). Lastly, it accounts for any source lending system-specific time-variant factors (source*time fixed effects). • In Model 5, in addition to time fixed effects, we include a fixed effect for each borrowers’ country–currency–time combination ( , to capture any unobservable shocks to credit 𝐹𝐹𝐹𝐹𝑖𝑖𝑖𝑖𝑖𝑖) demand in borrowers’ country denominated in a particular currency. For example, this controls for the specifics of lending to borrowers in Malaysia in 2019:Q4 in U.S. dollars. 20

Of note, this is the most extensive specification possible, where identification comes solely from variation across source lending systems. As such, this formulation subsumes borrowers’ country*time fixed effects (Model 3). In addition, it incorporates currency*time fixed effects ( ) which address the potential concern that some unobserved changes across the main𝐹𝐹 r𝐹𝐹e 𝑖𝑖 s 𝑖𝑖 erve currencies may drive our result. This strict fixed effect also absorbs the effects of currency-specific monetary policy changes, and the individual currency and time fixed effects. 5 Results We present the main results in Tables 2 to 7. Tables 2, 3, and 4 show our benchmark results for UN voting disagreement, broad sanctions, and relative borrower GPR, respectively. In each table, there are two columns for each model described in Section 4 above: one without geopolitical-monetary interaction terms (corresponding to Equation (1)) and one including interaction terms (corresponding to Equation (2)). Moving from left to right in each table, each set of two columns includes increasingly stringent fixed effects. Table 5 summarizes the economic significance of these three benchmark tables’ results. Tables 6 and 7 (which correspond to the Equation (3) specifications) decompose cross-border lending flows by target sector, examining lending to nonbanks (Table 6) and to banks (Table 7) separately.6 5.1 Benchmark results: UN voting disagreement 6 The results presented in Tables 2 and 3 and described in Section 5 are based on quarterized annual values of the IPD and total sanctions measures, respectively. In alternative specifications (available by request), we repeat these estimations using interpolated quarterly values for these variables and find that the results are highly comparable. 21

Table 2 shows the cumulative direct effects of changes in UN voting disagreement (captured by IPDs), of changes in the monetary policy associated with the currency of lending, and of the interaction of these two variables. Confirming our international bank lending channel hypothesis (Hypothesis 1), we find evidence, as shown in the first row of the table, that an increase in the shadow interest rate associated with the currency of lending over a four-quarter period leads to subsequently lower cross-border lending flows in a currency, consistent with the liquidity-reducing effect of monetary policy tightening. Signifying the important differentiating role of geopolitical tensions in the effect of monetary policy, the direct negative monetary effects are particularly strong in the specifications with interaction terms.7 Focusing on these interaction specifications, the marginal effect of a 100basis point increase in the short-term shadow interest rate over four quarters on subsequent lending flows ranges from a lending decline of 2.73 percentage points (henceforth, pp; column 6, with borrowers’ country*time fixed effects) to 4.58 pp (in column 8, with the demanding source country*borrowers’ country*time fixed effects). The aim of our paper is to understand how the transmission strengths of changes in monetary policy and geopolitical tensions depend on one another. Our bank lending channel Hypothesis 1 above posits that a worsening of bilateral UN voting disagreement amplifies monetary policy transmission; this is in part due to worsening geopolitical tensions leading to heightened investor risk perception of constrained banking systems. The consistently negative and significant interaction effects in the second row of Table 2 show convincing evidence that 7 Recall that the CDIBL does not necessarily require such significant direct monetary policy effect; only a significant interaction of this direct effect with measures of constraints. The reason is that the identification of the CDIBL is based on differing policy responses across more vs. less constrained source bank lending systems. 22

worsening UN voting disagreement amplifies the transmission of monetary policy, consistent with Hypothesis 1. A material rise in geopolitical tensions, amounting, for instance, to a five-standard deviation (about 1/5th of a unit) rise in the IPD subsequently increases the negative effect of a 100 bp monetary policy tightening on lending flows by a magnitude ranging from 2.81 pp8 (in column 2) to 15.1 pp (in column 10). We evaluate marginal effects to quantify the economic significance of the interaction of the geopolitical and monetary policy effects. The first three columns of Table 5 show marginal effects corresponding to the three most complete specifications in Table 2 (that is, corresponding to columns 6, 8, and 10). For instance, in column 2 of Table 5, at small changes in bilateral IPD (at the 10th percentile of the IPD distribution), a 100 bp rise in the shadow interest rate lowers crossborder lending by 3.87 pp. The corresponding effect at significant worsening of bilateral IPD (at the 90th percentile) is a decline of 5.4 pp. We can also rely on Table 5 to examine the differentiating role of monetary policy in the lending effect of worsening UN voting disagreement. Our Hypothesis 2 posits that increasing UN voting disagreement has direct negative effects on cross-border lending flows, and Hypothesis 3 suggests that this relationship is especially strong in the context of tightening monetary policy. Indeed, row 4 across the first three columns of Table 5 shows that at the sample median change in the shadow interest rate (corresponding to a 6 basis point quarterly tightening), the lending impact of rising UN voting disagreement is a decline of about 3.6 pp.9 At stronger monetary policy tightening (at the 75th percentile, as shown in column 3, row 5, corresponding to a 39 bp rise in rates), a five-standard deviation rise in IPD leads to an 8.2 pp decline in lending flows.10 To the 8 Obtained by taking the coefficient value in Table 2, column 2 (-14.06) and multiplying by (1/5). 9 Obtained by taking the coefficient value in row 4 of Table 5, column 3 (-17.94) and multiplying by (1/5). 10 In fact, Table 2 reveals that in our more complete specifications (including the interaction effects), a rise in UN voting disagreement over a four-quarter period leads to subsequently lower cross-border lending 23

backdrop of significant monetary policy tightening (at the 90th percentile of shadow rate changes), a five-standard deviation rise in IPD leads to an 15.2 pp decline in lending flows, as shown in Table 5, column 3. The target sector-specific IPD estimations in Tables 6 and 7 (first two columns) suggest that IPD primarily affects lending and policy transmission to the (bank and non-bank) financial sector. For instance, columns 2 of Tables 6 and 7 show that a five-standard deviation rise in IPD leads to a 5.27 pp and a 13.55 pp stronger effect of a 100 bp monetary policy tightening on crossborder flows to NBFIs and to inter-office banks, respectively. 5.2 Benchmark results: Sanctions Table 3 shows the cumulative direct effects of changes in bilateral sanctions, of changes in the monetary policy associated with the currency of lending, and of the interaction of these two variables. Consistent with our Hypothesis 1, in our more complete specifications (with interaction effects), we find evidence, as shown in the first row of Table 3, that an increase in the shadow interest rate associated with the currency of lending over a four-quarter period leads to subsequently lower cross-border lending flows in a currency, consistent with the liquidity-reducing effect of monetary policy tightening. Signifying the important differentiating role of sanctions in the effect of monetary policy, the direct negative monetary effects are particularly strong in the specifications with interaction terms. As the first row shows, the effect of a 100-basis point increase flows in a currency even without changes in monetary policy. Focusing on the interaction specifications, coefficients in the third row show that the marginal lending effect of a five-standard deviation (about 1/5th of a unit) increase in the IDP ranges from a decline of 1.22 pp (column 2, with borrowers’ country and time fixed effects) to 2.97 pp (in column 6, with borrowers’ country*time fixed effects). The direct effects are even larger in specifications without interaction terms. 24

in the short-term shadow interest rate over four quarters on subsequent lending flows ranges from a decline of 10.15 pp (column 2, with borrowers’ country and time fixed effects) to 13.39 pp (in column 8, with the demanding source country*borrowers’ country*time fixed effects). These results strongly support Hypothesis 1. Important for the identification of the bank lending channel is the interaction between the effects of our monetary and geopolitical variables; this interaction helps us understand how the transmission strengths of changes in monetary policy and measures of geopolitical tensions depend on one another. Our Hypothesis 2 above posits that intensifying bilateral sanctions may amplify monetary policy transmission, partly owing to worsening sanctions fueling heightened investor risk perception of banking system constraints. The negative and significant interaction effects in the second row of Table 3 show evidence that worsening bilateral sanctions amplify the transmission of monetary policy, consistent with Hypothesis 2. A notable rise in total sanctions (corresponding to what we could expect to see in the context of a geopolitical event), such as a five-standard deviation increase (which, based on Table 1, is about one-half of a unit), subsequently raises the negative effect of a 100 bp monetary policy tightening on lending flows by a magnitude ranging from 7.8 pp (in column 2) to 49.76 pp (in column 8). Studying marginal effects is instructive in quantifying the economic significance of the interaction of geopolitical and monetary policy effects. The middle three columns of Table 5 show marginal effects corresponding to the three most complete specifications in Table 3 (that is, corresponding to columns 6, 8, and 10). For instance, in column 5 of Table 5, at small changes in bilateral sanctions (at the 10th percentile of the sanctions distribution, row 1), a 100 bp rise in the policy interest rate lowers cross-border lending by 12.16 pp. The corresponding effect at a significant rise in sanctions (at the 90th percentile, row 3) is a decline of almost 17 pp. 25

The last three rows of Table 5, columns 4-6, show the differentiating role of monetary policy in the effect of intensifying bilateral sanctions on lending. Our Hypothesis 2 posits that worsening sanctions can have direct negative effects on cross-border lending flows, and Hypothesis 3 suggests that these effects are stronger in the context of tightening monetary policy. Indeed, as shown for instance by the last two rows of column 6 of Table 5, to the backdrop of tightening monetary policy (that is, at the 75th and 90th percentiles of interest rate changes, respectively), a marginal increase in sanctions leads to declines in lending. Although the magnitudes are small, these results lend support to our hypotheses. As expected, the target sector-specific estimations in Tables 6 and 7 (middle two columns) reveal that the overall lending effects of sanctions are strong for the non-financial sector (borrowers who may be most affected by direct sanctions effects) and banks (who in turn might finance sanctions-impaired borrowers). Importantly, thus far we have described the effects and interactions of sanctions of all types. In additional estimations, we examine the effects of changes in bilateral financial sanctions (Table A4) and bilateral trade sanctions (Table A5). These results show that the marginal effects of changes in financial and trade sanctions are equally significant, and larger in magnitudes, than the effects of changes in sanctions of all types documented in Table 3. 5.3 Benchmark results: Relative borrower GPR Table 4 shows the cumulative direct effects of changes in the difference of borrowers’ country GPR and source country GPR (henceforth, relative borrower GPR), of changes in the monetary policy associated with the currency of lending, and of the interaction of these two variables. 26

In our more complete specifications (with interaction effects), we find significant evidence, as shown in the first row of Table 4, that an increase in the shadow interest rate associated with the currency of lending over a four-quarter period leads to subsequently lower cross-border lending flows in a currency, consistent with the liquidity-reducing effect of monetary policy tightening. Consistent with the important differentiating role of relative borrower GPR in the transmission strength of monetary policy, the direct negative monetary effects are particularly strong in the specifications with interaction terms. Focusing on these interactive specifications, the marginal effect of a 100-basis point increase in the short-term shadow interest rate over four quarters on subsequent lending flows ranges from a decline of 3.33 pp (column 2, with borrowers’ country and time fixed effects) to 4.84 pp (in column 8, with the demanding source country*borrowers’ country*time fixed effects). Hypothesis 2 posits significant interactions of the effects of monetary policy and measures of geopolitical tensions. In this case, it implies that intensifying GPR in the borrowers’ country relative to the GPR of the source country may amplify monetary policy transmission, partly owing to relatively worse geopolitical risk fueling heightened investor risk perception. The negative and significant interaction effects in the second row of Table 4 show evidence that relatively worse borrower GPR amplifies the transmission of monetary policy. A five-standard deviation rise in relative borrower GPR (which, based on Table 1, is about 2.5 units) subsequently amplifies the negative effect of a 100 bp monetary policy tightening on lending flows by a magnitude ranging from 9.45 pp (in column 10) to 25.92 pp (in column 8).11 The last three columns of Table 5 show marginal effects corresponding to the three most complete specifications in Table 4 (that is, corresponding to columns 6, 8, and 10). For instance, 11 For instance, the -9.45 value is obtained as -3.78 (column 10, row 2) times (2.5). 27

as the first row in column 8 of Table 5 shows, at small changes in the borrowers’ country GPR relative to the source GPR (at the 10th percentile of the relative borrower GPR distribution), a 100 bp rise in the policy interest rate lowers cross-border lending by 2.85 pp. The corresponding effect at significantly worsening relative borrower GPR (at the 90th percentile, in row 3) is a decline of 6.81 pp. The last three rows of Table 5, columns 7-9, show the differentiating role of monetary policy in the lending effect of intensifying relative borrower GPR. Our Hypothesis 2 posits that worsening relative borrower GPR can have direct negative effects on cross-border lending flows, and Hypothesis 3 suggests that this is especially so in the context of tightening monetary policy. Indeed, as shown for instance by the last row of column 7 of Table 5, to the backdrop of tightening monetary policy (that is, at the 90th percentile of shadow rate changes), a marginal increase in relative borrower GPR leads to a lending decline of 35.6 pp. The target sector-specific estimations in the last two columns of Tables 6 and 7 reveal that the overall effect of worsening relative borrower GPR appears strongest in lending to the financial sector. For instance, row 2 in column 6 of Table 6 implies that a five-standard deviation (about 2.5 unit) rise in relative borrower GPR leads to a 15.56 pp stronger effect of a 100 bp monetary policy tightening on cross-border flows to NBFIs. Importantly, so far we have described results that look at a bilateral measure: changes in borrowers’ country GPR relative to changes in source country GPR. In additional estimations, we aim to unfold these findings further by examining the lending effects and monetary policy interactions of borrowers’ country GPR and source country GPR separately. In Table A6, odd columns show the effects of borrower GPR, and even columns show the effects of source GPR, for our most complete models. While the lending impact of rises in the GPR of both the source 28

lending system and of borrowers’ countries is material, Table A6 reveals that changes in the geopolitical risk of borrowers’ countries have generally stronger direct and interaction effects. 5.4 Results for different monetary policy-geopolitical tensions “states of the world” Hypothesis 3 outlined above is novel but is also quite intuitive: borrower economies face a double whammy as escalating geopolitical tensions boost uncertainty. Tighter financial conditions make it harder to cope with this increased uncertainty, as contractionary monetary policy aggravates the cost of acquiring liquidity. Therefore, following a monetary policy tightening, banks cut back lending to borrowers in countries affected by geopolitical tensions especially hard. Effects on cross-border lending could be particularly damaging as these two forces likely amplify each other. To explore this conjecture, we re-estimate our regression equation with stringent borrowers’ country*time fixed effects and including dummies that allow the regression coefficients to be specific to each of the following four “states of the world”: (1) tightening monetary policy and worsening geopolitical tensions; (2) tightening monetary policy and improving geopolitical tensions; (3) easing monetary policy and worsening geopolitical tensions; and (4) easing monetary policy and improving geopolitical tensions. We find that worsening geopolitical tensions significantly amplify the effects of tighter monetary policy; however, improving geopolitical tensions do not cushion the effects of contractionary monetary policy. Table 8 presents results corresponding to two states: doublewhammy (monetary policy tightens and geopolitical tensions worsen) and policy only-whammy (monetary policy tightens but geopolitical tensions deescalate).12 In columns 1, 3, and 5, we show 12 The full set of specifications for all states are available upon request. 29

estimation results for the double whammy state, while the even columns present results for the policy-only whammy state. Comparing the estimated interaction coefficients across states for each geopolitical measure, we see that worsening geopolitical tensions significantly amplify the effects of tighter monetary policy but improving geopolitical tensions do not cushion the contractionary monetary policy effects. The comparison is particularly straightforward when we measure geopolitical tensions via UN voting disagreement or total sanctions: the interaction terms in the double-whammy state are negative and statically significant, but they are statistically insignificant in the policy-only whammy state. When comparing the relative GPR results across columns, we examine the combined marginal effect of monetary policy in column 6. We see that, in a more benign geopolitical state cross-border bank flows may remain little changed, or even increase, despite monetary policy tightening. 5.5 Significance of geopolitical tensions Our results suggest that geopolitical tensions are economically significant drivers of international bank lending flows. A variance decomposition exercise reveals that geopolitics is at least as important a driver of cross-border bank lending flows as monetary policy. Focusing on the portion of variation not explained by our battery of fixed effects, we see that geopolitics directly explains 50 percent of the variation, monetary policy explains around 30 percent, and the geopoliticsmonetary policy interaction explains around 20 percent. 6 Alternative specifications and robustness checks 30

We run a set of additional specifications to ensure the robustness of our results. As we describe below, we explore the role of potential common trends across countries in geopolitical risk and tensions as well as the role of cross-currency effects of monetary policy. 6.1. Common drivers of geopolitical risk and tensions over time Materialized measures of bilateral geopolitical tensions, such as UN voting disagreement or sanctions, are generally specific to pairs of countries. However, in the aftermath of large shocks, immaterialized or prospective geopolitical risk (as measured by the GPR) may also have a global component. For instance, geopolitical risk around the world skyrocketed after Russia’s invasion of Ukraine. In other words, there might be common drivers of geopolitical risks (such as the beginning of the Russia-Ukraine war) that cause GPRs to move together across countries around stress events. We address this issue by running a set of regressions in which we “de-mean” all three of our geopolitical measures. For each measure, we calculate the quarter-specific average of measures (averaged over the cross section of countries) and subtract this average from the measure itself. This approach eliminates concerns about co-movement, as we are examining the effects of geopolitical changes above and beyond those observed commonly across countries. Table A1 shows our de-meaned results for IPD (columns 1-3), total sanctions (columns 4-6), and relative borrower GPR (columns 7-9). We find that our benchmark results are strongly robust to this alternative specification. 6.2. Cross-currency effects of monetary policy 31

Our inclusion of time fixed effects in all our estimations is powerful in addressing substitution effects, because the inclusion means comparing the effects of monetary policy changes relative to one another, at a given point in time. Even more so, our inclusion of variations of currency*time fixed effects in our most complete specifications fully controls for substitution effects from other monetary policies. However, in a set of alternative specifications, we directly account for the possibility that currency substitution patterns exist; that is, the effect of changes in the monetary policy associated with one reserve currency affecting lending flows in other reserve currencies. We run a set of regressions in which we use “relative interest rate changes” in place of interest rate changes, where the “relative interest rate change” is defined as a change in the shadow rate associated with the currency of lending, minus the (weighted) average change in the other four interest rates. This way, we examine the effects of “relative” monetary policy changes – that is, the effect of changes in the interest rate of one currency above and beyond the average change in the other reserve currency rates. We find that our results are strongly robust to this exercise for all three of our geopolitical measures, as shown in Table A2. 6.3. Cross-currency effects of monetary policy and common drivers of geopolitical tensions In Table A3, we combine the exercises described in the subsections above, as we estimate the individual and interactive effects of de-meaned geopolitical measures and relative interest rate changes. These estimations show that our benchmark results are strongly robust to these alternative specifications for all three of our geopolitical measures. 32

6.4. Borrowers’ countries: OECD vs non-OECD Table A7 shows the results of estimations in which we delineate our sample by borrowers’ country OECD membership. We see that for IPD and relative borrower GPR, our main results hold across both OECD and non-OECD borrowers. In the case of sanctions, we find significance for non-OECD borrowers but not for OECD borrowers; as most sanctions are vis-à-vis non-OECD countries, we attribute this outcome to the very small number of observations in the OECD group. For better comparison of coefficient magnitudes across the two borrowers’ groups, it is instructive to standardize coefficients. A one standard deviation increase in IPD corresponds to an interaction coefficient of 39 for non-OECD countries, and a coefficient of 30 for OECD countries.13 Similarly, a one standard deviation increase in sanctions corresponds to an interaction coefficient of 19 for non-OECD countries. Lastly, a one standard deviation increase in relative borrower GPR corresponds to an interaction coefficient of around 7 for non-OECD countries, and a coefficient of around 14 for OECD countries. 7 Conclusion In this paper, we use a BIS dataset on bilateral cross-border bank claims by bank nationality to assess the effects of geopolitical tensions on cross-border bank flows denominated in reserve currencies, including the U.S. dollar. We show that a rise in geopolitical tensions—either materialized (captured by political disagreement across countries through UN voting or by sanctions) or unrealized (captured by geopolitical risk indices)—dampen such bank flows. We also 13 For OECD and non-OECD countries, respectively, the IPD standard deviations (s.d.) are 0.58 and 0.74; the sanctions s.d.’s are 0.57 and 1.59, and the relative GPR s.d.’s are 0.45 and 0.48. 33

show that geopolitical tensions amplify the international transmission strength of the monetary policies of major central banks. Furthermore, we show that cross-border bank lending declines especially hard when geopolitical tensions coincide with monetary policy tightening. We also find that geopolitical tensions are significant drivers of international bank flows, with lending effects that are comparable in magnitude to the bank lending channel of monetary policy. Our results are policy relevant. For policy makers in reserve currency-issuing countries, understanding the effects of geopolitical tensions on monetary policy transmission can help gauge changes in global liquidity conditions in their currency. For policy makers in the source countries of lending banks, understanding the effects of geopolitical tensions can help gauge cross-border bank lending activities of their banks and thus, domestic credit conditions. For policy makers in borrowers’ countries, understanding the effects of geopolitical tensions can help gauge credit supply via cross-border bank lending to their country. 34

References: Afesorgbor, S. K. (2019). “The impact of economic sanctions on international trade: How do threatened sanctions compare with imposed sanctions?”. European Journal of Political Economy, 56, 11-26. António, A., Alves, J., and Monteiro, S. (2024) “Beyond borders: Assessing the influence of Geopolitical tensions on sovereign risk dynamics”. European Journal of Political Economy, vol. 83, 102550. Avdjiev, S., McGuire, P., and Wooldridge, P. (2015). “Enhanced data to analyse international banking”. BIS Quarterly Review, September, 53-68. Avdjiev, S., and Takáts, E. (2019). “Monetary Policy Spillovers and Currency Networks in Cross- Border Bank Lending: Lessons from the 2013 Fed Taper Tantrum”. Review of Finance, vol. 23, iss. 5, 993–1029. Bailey, M. A., and Voeten, E. (2018). “A two-dimensional analysis of seventy years of United Nations voting”. Public Choice, 176, 33-55. Bailey, M.A., Strezhnev, A., and Voeten, E. (2017). “Estimating Dynamic State Preferences from United Nations Voting Data”. The Journal of Conflict Resolution, 61(2), 430–56. Bosone, C., and Stamato, G. (2024). “Beyond borders: how geopolitics is reshaping trade”. ECB working paper No. 2960, the URL: < https://www.ecb.europa.eu/pub/pdf/scpwps/ecb.wp2960~6c3cc5e5b0.en.pdf>. Buch, C. M., Bussierè, M., Goldberg, L., and Hills, R. (2019) “The international transmission of monetary policy”. Journal of International Money and Finance, vol. 91, 29-48. Buch, C. M., Eickmeier, S., and Prieto, E. (2014). “In search for yield? Survey-based evidence on bank risk taking”. Journal of Economic Dynamics and Control, 43, 12-30. Caldara, D., and Iacoviello, M. (2022). “Measuring geopolitical risk”. American Economic Review, 112(4), 1194-1225. Catalan, M., Fendoglu, S., and Tsuruga, T. (2024). “A Gravity Model of Geopolitics and Financial Fragmentation”. IMF Working Papers, 196. Cecchetti, S. G., Fender, I., and McGuire, P. (2010). “Toward a global risk map”. BIS Working Paper 309. Cerutti, E., and Claessens, S. (2017). “The great cross-border bank deleveraging: supply constraints and intra-group frictions”. Review of Finance, 21(1), 201-236. Cerutti, E., Hale, G., and Minoiu, C. (2015). “Financial crises and the composition of cross-border lending”. Journal of International Money and Finance, 52, 60-81. 35

Cetorelli, N., and Goldberg, L. S. (2012). “Banking globalization and monetary transmission”. The Journal of Finance, 67(5), 1811-1843. Committee on the Global Financial System (2011). “Global liquidity: concept, measurement and policy implications”. CGFS Papers 45. De Haas, R., and Van Lelyveld, I. (2014). “Multinational banks and the global financial crisis: Weathering the perfect storm?”. Journal of Money, Credit and Banking, 46(s1), 333-364. Felbermayr, G., Kirilakha, A., Syropoulos, C., Yalcin, E., and Yotov, Y.V. (2020). “The Global Sanctions Data Base”. European Economic Review, 129 (C). Fender, I., and McGuire, P. (2010). “Bank structure, funding risk and the transmission of shocks across countries: concepts and measurement”. BIS Quarterly Review, September. Giannetti, M., and Laeven, L. (2012). “The flight home effect: Evidence from the syndicated loan market during financial crises”. Journal of Financial Economics, 104(1), 23-43. Goldberg and Hannaoui, O. (2024). “Drivers of Dollar Share in Foreign Exchange Reserves”. Federal Reserve Bank of New York working paper. , L.S. Haas, R. D., and Horen, N. V. (2012). “International shock transmission after the Lehman Brothers collapse: Evidence from syndicated lending”. American Economic Review, 102(3), 231-237. Fernández-Villaverde, J., Mineyama, T., and Song, D. (2024). “Are We Fragmented Yet? Measuring Geopolitical Fragmentation and Its Causal Effect”. NBER Working Paper 32638. Lee, S. J., Liu L. Q. and Stebunovs, V. (2022). “Risk-taking spillovers of U.S. monetary policy in the global market for U.S. dollar corporate loans”. Journal of Banking & Finance, vol. 138, 105550. Niepmann, F., and Shen, L. “Geopolitical Risk and Global Banking”. Working paper. Pradhan, S.K., Takáts, E., and Temesvary, J. (2024). “How does fiscal policy affect the transmission of monetary policy into cross-border bank lending? Cross-country evidence”. BIS Working Papers, 1226. Rose, A.K., and Wieladek, T. (2014). “Financial protectionism? First evidence”. Journal of Finance, 69(5), 2127-2149. Syropoulos, C., Felbermayr, G., Kirilakha, A., Yalcin, E., and Yotov, Y.V. (2024). “The global sanctions data base–Release 3: COVID-19, Russia, and multilateral sanctions”. Review of International Economics, 32(s1), 12-48. Takáts, E., and Temesvary, J. (2020). “The currency dimension of the bank lending channel in international monetary transmission”. Journal of International Economics, 125, 103309. Takáts, E., and Temesvary, J. (2021). “How does the interaction of macroprudential and monetary policies affect cross-border bank lending?”. Journal of International Economics, 132, 103521. 36

Von Peter, Goetz. (2024). “Geopolitics and international finance: knowns and unknowns,” mimeo. Wang, X., Wu, Y., and Xu, W. (2019). “Geopolitical risk and investment”. Journal of Money, Credit and Banking, doi.org/10.1111/jmcb.13110. Yilmazkuday, H. (2024). “Geopolitical risk and stock prices”. European Journal of Political Economy, 83, 102553. 37

Currency share of cross-border claims1 In % of total in all currencies Graph 1 A. Bank sector B . Nonbank sector C. All sectors 1 Relates to total of 17 bank nationalities in the sample with currency positions denominated in USD, EUR, JPY, CHF and GBP. Excludes crossborder claims in home currency (ie EUR-denominated claims by euro area banks are excluded). Sources: BIS locational banking statistics (by nationality); authors’ calculations. 1

Central bank policy rate and Krippner shadow short rate by currency In percentage Graph 2 A. USD B. EUR C. JPY D. GBP E. CHF Sources: Leo Krippner (2020) ; IMF ; BIS 2

Table 1: Summary statistics Variable Description N Mean p50 SD Min Max Source Bilateral cross-border Defined as (ln(xbcunweighted) - ln(l.xbcunweighted))*100, where lending flows xbcunweighted is the currency denominated amount of cross-border 289,689 0.00 -0.13 63.40 -837.09 797.09 BIS (unweighted) claims of a bank nationality on a given counterparty country in time t Bilateral cross-border lending flows Winsorized value of Bilateral cross-border lending flows (unweighted) 289,689 0.07 -0.13 55.39 -229.87 241.33 BIS (winsorized, unweighted) Bilateral total claims Percentage share in bilateral outstanding claims of bank nationality all 313,357 0.01 0.00 0.05 0.00 2.24 BIS share currencies and all sectors (used as weights) Source government debt Geneneral government debt of banks' parent country, in % of GDP 313,357 88.21 96.90 35.84 19.60 210.30 Eurostat; FRED to GDP ratio Change in source government debt to GDP Dfference in source government debt to GDP ratio from t-1 to t 289,689 0.05 -0.40 2.73 -10.00 25.97 Eurostat; FRED ratio Change in source Quarterly change in GeoPolitical Risk Index of bank's parent country, geopolitical risk index 257,965 0.01 0.00 0.46 -4.02 4.42 Caldara & Iacoviello (2021) current basis (GPR) Quarterly change in GeoPolitical Risk Index of borrower country, current Change in borrower GPR 155,996 0.01 0.00 0.40 -5.78 7.20 Caldara & Iacoviello (2021) basis Change in relative Quarterly change in GeoPolitical Risk Index of (borrower minus source) 117,671 0.00 0.00 0.52 -5.95 7.17 Caldara & Iacoviello (2021) borrower GPR country, current basis Quaterly change in aggregate sanction indicators (1 /0) comprising Syropoulos, Felbermayr, Change in total arms, military, trade, financial, travel, and other sanctions (annual 50,587 0.01 0.00 0.10 -0.75 0.50 Kirilakha, Yalcin, and Yotov sanctions figures are quarterized) (2024) Syropoulos, Felbermayr, Change in financial Quaterly change in financial sanction indicator (1 /0) (annual figures are 50,587 0.00 0.00 0.04 -0.25 0.25 Kirilakha, Yalcin, and Yotov sanctions quarterized) (2024) Syropoulos, Felbermayr, Change in trade Quaterly change in trade sanction indicator (1 /0) (annual figures are 50,587 0.00 0.00 0.03 -0.25 0.25 Kirilakha, Yalcin, and Yotov sanctions quarterized) (2024) Change in ideal point Bailey, Strezhnev, and Quaterly change in Ideal Point Distance (annual figures are quarterized) 275,090 0.00 0.00 0.04 -0.42 0.29 distance Voeten (2017) Change in shadow Difference in currency-specific short-term shadow interest rate from t-1 289,689 0.12 0.06 0.65 -1.76 2.58 Krippner (2024) interest rate to 1 Change in source Difference in source central bank policy rate from t-1 to 1 313,357 0.58 0.05 1.24 -0.75 5.38 Krippner (2024) central bank policy rate Change in USD to Difference in USD to currency rate from t-1 to 1 313,357 -0.29 0.00 3.45 -14.44 9.25 BIS EUR/JPY/GBP/CHF rate Change in bilateral (Source, borrower) Difference in bilateral FX rate from t-1 to t 289,689 4.28 0.00 148.72 -1,374.83 13,924.38 BIS 3 exchange rate

Table 2: Effects of changes in Ideal Point Distance and in monetary policy on cross-border lending flows Model [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] Dependent variable: Bilateral cross-border lending flows Σ∆ Shadow Interest Rate {t-1 to t-4} -1.729*** -3.006*** -1.896*** -3.428*** -1.866*** -2.734*** -2.318*** -4.576*** np np [0.280] [0.320] [0.278] [0.321] [0.324] [0.360] [0.338] [0.391] np np Σ [∆Shadow Interest Rate * ∆ Ideal Point Distance] {t-1 to t-4} -14.060** -27.805*** -37.635*** -57.503*** -75.412*** [6.615] [6.713] [8.501] [14.944] [9.469] Σ∆ Ideal Point Distance {t-1 to t-4} -8.655*** -6.096* -2.392 5.872 -26.400*** -14.862*** np np -36.564*** -11.889** [2.886] [3.614] [3.040] [3.770] [3.656] [4.528] np np [3.742] [4.771] Observations 168,391 168,391 168,391 168,391 168,391 168,391 168,391 168,391 168,391 168,391 Time FE Yes Yes Yes Yes -- -- -- -- -- -- Borrower FE Yes Yes -- -- -- -- -- -- -- -- Borrower*Time FE No No No No Yes Yes -- -- -- -- Source*Borrower FE No No Yes Yes No No -- -- No No Source*Borrower*Time FE No No No No No No Yes Yes No No Borrower*Currency* Time FE No No No No No No No No Yes Yes Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 "np" indicates that the variable is subsumed by the included set of fixed effects. 4

Table 3: Effects of changes in total sanctions and in monetary policy on cross-border lending flows Model [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] Dependent variable: Bilateral cross-border lending flows Σ∆ Shadow Interest Rate {t-1 to t-4} -6.861*** -10.152*** -7.115*** -11.018*** -7.998*** -11.798*** -8.259*** -13.388*** np np [0.747] [0.950] [0.745] [0.953] [0.843] [1.010] [0.942] [1.159] np np Σ [∆Shadow Interest Rate * ∆ Total Sanctions] {t-1 to t- 4} -15.608*** -28.717*** -7.511 -99.510*** -6.776 [4.893] [5.193] [8.492] [22.800] [8.530] Σ∆ Total Sanctions {t- 1 to t-4} 5.123** 10.002*** 6.005** 16.361*** 0.685 2.085 np np 0.576 0.779 [2.497] [2.994] [2.582] [3.254] [3.512] [4.066] np np [3.403] [4.007] Observations 38,807 38,807 38,807 38,807 38,807 38,807 38,807 38,807 38,807 38,807 Time FE Yes Yes Yes Yes -- -- -- -- -- -- Borrower FE Yes Yes -- -- -- -- -- -- -- -- Borrower*Time FE No No No No Yes Yes -- -- -- -- Source*Borrower FE No No Yes Yes No No -- -- No No Source*Borrower* Time FE No No No No No No Yes Yes No No Borrower*Currency* Time FE No No No No No No No No Yes Yes Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 "np" indicates that the variable is subsumed by the included set of fixed effects. 5

Table 4: Effects of changes in relative borrower geopolitical risk and in monetary policy on cross-border lending flows Model [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] Dependent variable: Bilateral cross-border lending flows Σ∆ Shadow Interest Rate {t- 1 to t-4} -1.964*** -3.326*** -2.176*** -3.838*** -2.345*** -3.265*** -2.411*** -4.835*** np np [0.399] [0.461] [0.396] [0.460] [0.460] [0.515] [0.445] [0.514] np np Σ [∆Shadow Interest Rate * ∆ Relative borrower GPR] {t-1 to t-4} -5.911*** -5.950*** -8.620*** -10.366*** -3.781*** [0.748] [0.747] [0.954] [1.151] [1.175] Σ∆ Relative borrower GPR {t-1 to t-4} -1.500*** 2.757*** -1.548*** 2.754*** 1.966** 7.831*** np np -0.585 1.277 [0.567] [0.783] [0.564] [0.778] [0.917] [1.128] np np [0.907] [1.217] Observations 79,585 79,585 79,585 79,585 79,585 79,585 79,585 79,585 79,585 79,585 Time FE Yes Yes Yes Yes -- -- -- -- -- -- Borrower FE Yes Yes -- -- -- -- -- -- -- -- Borrower*Time FE No No No No Yes Yes -- -- -- -- Source*Borrower FE No No Yes Yes No No -- -- No No Source*Borrower*Time FE No No No No No No Yes Yes No No Borrower*Currency*Time FE No No No No No No No No Yes Yes Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 "np" indicates that the variable is subsumed by the included set of fixed effects. 6

Table 5: Effects of changes in geopolitical measures and in monetary policy on cross-border lending flows: Marginal effects from Tables 2, 3, and 4 Model [1] [2] [3] [4] [5] [6] [7] [8] [9] Dependent variable: Bilateral cross-border lending flows Geopolitical measure: Ideal point distance Total sanctions Relative borrower GPR MP effect at 10th ptile of geopolitical measure change -2.27 -3.87 np -11.70 -12.16 -- -1.62 -2.85 np MP effect at median geopolitical measure change -2.77 -4.63 np -11.79 -13.38 np -3.30 -4.88 np MP effect at 90th ptile of geopolitical measure change -3.25 -5.36 np -12.06 -16.98 np -4.91 -6.81 np geopolitical measure change effect at median MP change -17.86 np -17.94 1.47 np 0.23 7.13 np 0.97 geopolitical measure change effect at 75th ptile of MP change -29.37 np -41.02 -0.83 np -1.84 4.49 np -0.19 geopolitical measure change effect at 90th ptile of MP change -46.88 np -76.13 -4.32 np -5.00 -14.23 np -1.95 Borrower*Time FE Yes -- -- Yes -- -- Yes -- -- Source*Borrower FE No -- No No -- No No -- No Source*Borrower*Time FE No Yes No No Yes No No Yes No Borrower*Currency*Time FE No No Yes No No Yes No No Yes 7 "np" indicates that the variable is subsumed by the included set of fixed effects.

Table 6: Breakdown by target sector: Lending to Non-banks; Effects of changes in geopolitical measures and in monetary policy on cross-border lending flows Model [1] [2] [3] [4] [5] [6] Dependent variable: Bilateral cross-border lending flows to non-banks Geopolitical measure: Ideal point distance Total sanctions Relative borrower GPR Non- Non- Non- Sector of Borrowers NBFIs NBFIs NBFIs financials financials financials Σ∆ Shadow Interest Rate {t-1 to t-4} -1.340*** -6.594*** -12.071*** -11.772*** -0.588 -6.521*** [0.492] [0.982] [1.316] [3.300] [0.717] [1.239] Σ [∆Shadow Interest Rate * ∆ Geopolitical measure] {t-1 to t-4} -13.556** 75.291*** -12.071*** 38.497*** -1.466 -6.225*** [5.526] [15.922] [1.316] [13.908] [1.116] [1.612] Σ∆ Geopolitical measure {t-1 to t-4} 14.902*** -25.528** -0.175 26.562*** -0.501 -8.659*** [4.016] [10.663] [3.385] [9.354] [1.103] [1.774] Observations 82,033 28,689 19,434 4,812 35,649 17,337 Source*Borrower FE Yes Yes Yes Yes Yes Yes Borrower*Currency*Time FE No No No No No No Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 "np" indicates that the variable is subsumed by the included set of fixed effects. The non-bank sector comprises of non-bank financial institutions and the non-financial sector. 8

Table 7: Breakdown by target sector: Lending to Banks; Effects of changes in geopolitical measures and in monetary policy on cross-border lending flows Model [1] [2] [3] [4] [5] [6] Dependent variable: Bilateral cross-border lending flows to banks Geopolitical measure: Ideal point distance Total sanctions Relative borrower GPR Of which: Of which: Of which: Sector of Borrowers Total banks Total banks Total banks Inter-office Inter-office Inter-office Σ∆ Shadow Interest Rate {t-1 to t-4} -2.881*** -7.006*** -12.341*** -25.337*** -3.304*** -7.018*** [0.570] [1.088] [1.767] [3.983] [0.759] [1.316] Σ [∆Shadow Interest Rate * ∆ Geopolitical measure] {t-1 to t-4} -58.483*** -67.729*** -76.415*** -62.506 -5.760*** -5.375** [8.686] [17.179] [15.219] [41.503] [1.271] [2.223] Σ∆ Geopolitical measure {t-1 to t-4} 6.223 33.575*** 36.095*** 33.727 3.118** 4.969** [5.595] [10.927] [8.658] [24.812] [1.320] [2.352] Observations 63,306 23,659 13,948 4,331 35,095 16,207 Source*Borrower FE Yes Yes Yes Yes Yes Yes Borrower*Currency*Time FE No No No No No No Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 "np" indicates that the variable is subsumed by the included set of fixed effects. The bank sector comprises of affiliates [inter-office] and non-affiliated banks. 9

Table 8: Breakdown by monetary policy-geopolitical tensions "states": Effects of changes in geopolitical measures and in monetary policy on cross-border lending flows Model [1] [2] [3] [4] [5] [6] Dependent variable: Bilateral cross-border lending flows Geopolitical measure: Ideal point distance Total sanctions Relative borrower GPR Tightening Tightening Tightening Tightening Tightening Tightening MP and MP and MP and MP and MP and MP and Regime worsening easing GP worsening easing GP worsening easing GP GP tensions tensions GP tensions tensions GP tensions tensions Σ∆ Shadow Interest Rate {t-1 to t-4} -2.698*** -3.954*** -11.030*** 5.685 -5.958*** 0.552 [0.595] [0.626] [0.959] [16.079] [0.778] [0.716] Σ [∆Shadow Interest Rate * ∆ Geopolitical measure] {t-1 to t-4} -149.721*** 158.053 -46.926*** 22.551 -4.271*** -22.104*** [13.066] [180.030] [8.386] [74.261] [1.633] [2.249] Σ∆ Geopolitical measure {t-1 to t-4} 83.662*** -160.319*** 23.237*** -1.593 -1.90 29.194*** [9.155] [12.777] [4.581] [19.311] [1.953] [2.548] Observations 168,849 168,849 38,807 38,807 79,585 79,585 Time FE -- -- -- -- -- -- Borrower FE -- -- -- -- -- -- Borrower*Time FE Yes Yes Yes Yes Yes Yes Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 10

Table A1: Accounting for common global trends: Effects of changes in de-meaned geopolitical measures and in monetary policy on cross-border lending flows Model [1] [2] [3] [4] [5] [6] [7] [8] [9] Dependent variable: Bilateral cross-border lending flows Geopolitical measure: Ideal point distance Total sanctions Relative borrower GPR Σ∆ Shadow Interest Rate {t-1 to t-4} -2.904*** -4.677*** np -11.820*** -13.933*** np -3.218*** -4.766*** np [0.361] [0.393] np [1.009] [1.162] np [0.515] [0.514] np Σ [∆Shadow Interest Rate * ∆ De-meaned Geopolitical measure] {t-1 to t-4} -16.636*** -55.041*** -48.245*** -9.518 -110.218*** -6.776 -8.891*** -10.746*** -3.781*** [5.973] [11.364] [6.624] [8.490] [22.747] [8.530] [0.970] [1.191] [1.175] Σ∆ De-meaned Geopolitical measure {t-1 to t-4} -4.35 np -10.915** 2.69 np 0.779 7.973*** np 1.277 [4.126] np [4.308] [4.067] np [4.007] [1.130] np [1.217] Observations 168,849 168,849 168,849 38,807 38,807 38,807 79,585 79,585 79,585 Borrower*Time FE Yes -- -- Yes -- -- Yes -- -- Source*Borrower FE No -- No No -- No No -- No Source*Borrower*Time FE No Yes Yes No Yes Yes No Yes Yes Borrower*Currency*Time FE No No Yes No No Yes No No Yes Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 "np" indicates that the variable is subsumed by the included set of fixed effects. De-meaned geopolitical measure is the contemporaneous value of each geopolitical measure [as shown in column headings] minus the cross-sectional weighted average of the measure in that given quarter. 11

Table A2: Accounting for cross-currency effects of monetary policy: Effects of changes in geopolitical measures and in relative monetary policy on cross-border lending flows Model [1] [2] [3] [4] [5] [6] [7] [8] [9] Dependent variable: Bilateral cross-border lending flows Geopolitical measure: Ideal point distance Total sanctions Relative borrower GPR Σ∆ Relative Shadow Interest Rate {t-1 to t-4} -1.645*** -2.933*** np -7.583*** -8.752*** np -1.982*** -3.125*** np [0.256] [0.279] np [0.691] [0.799] np [0.365] [0.368] np Σ [∆Relative Shadow Interest Rate * ∆ Geopolitical measure] {t-1 to t-4} -1.645*** -42.108*** -2.255 -33.231*** -64.707*** -14.377 -9.744*** -8.327*** -5.785*** [0.256] [8.274] [10.005] [9.438] [15.384] [11.129] [0.837] [0.865] [1.180] Σ∆ Geopolitical measure {t-1 to t-4} -12.192*** np -29.907*** 1.586 79.764 0.854 3.183*** np -0.609 [3.399] np [3.501] [3.527] [318.815] [3.451] [0.927] np [0.933] Observations 168,849 168,849 168,849 38,807 38,807 38,807 79,585 79,585 79,585 Borrower*Time FE Yes -- -- Yes -- -- Yes -- -- Source*Borrower FE No -- No No -- No No -- No Source*Borrower*Time FE No Yes No No Yes No No Yes No Borrower*Currency* Time FE No No Yes No No Yes No No Yes Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 "np" indicates that the variable is subsumed by the included set of fixed effects. Relative shadow interest rate is the shadow rate of the lending currency minus the weighted average shadow rate across the other currencies. 12

Table A3: Accounting for cross-currency effects of monetary policy and common global trends: Effects of changes in demeaned geopolitical measures and in relative monetary policy on cross-border lending flows Model [1] [2] [3] [4] [5] [6] [7] [8] [9] Geopolitical measure: Ideal point distance Total sanctions Relative borrower GPR Dependent variable: Bilateral cross-border lending flows Σ∆ Relative Shadow Interest Rate {t-1 to t-4} -1.710*** -3.034*** np -8.265*** -9.663*** np -1.982*** -3.125*** np [0.258] [0.281] np [0.763] [0.880] np [0.365] [0.368] np Σ [∆Relative Shadow Interest Rate * ∆ Geopolitical measure] {t-1 to t-4} -24.916*** -41.902*** -2.255 -- -- -- -9.744*** -8.327*** -5.785*** [7.268] [8.227] [10.005] -- -- -- [0.837] [0.865] [1.180] Σ∆ Geopolitical measure {t-1 to t-4} -12.334*** np -29.907*** 0.229 np -0.22 3.183*** np -0.609 [3.398] np [3.501] [3.509] np [3.408] [0.927] np [0.933] Observations 168,849 168,849 168,849 38,807 38,807 38,807 79,589 79,589 79,589 Borrower*Time FE Yes -- -- Yes -- -- Yes -- -- Source*Borrower FE No -- No No -- No No -- No Source*Borrower*Time FE No Yes No No Yes No No Yes No Borrower*Currency* Time FE No No Yes No No Yes No No Yes Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 "np" indicates that the variable is subsumed by the included set of fixed effects. Relative shadow rate is the shadow rate of the lending currency minus the weighted average shadow rate across the other four currencies. 13 De-meaned geopolitical measure is the contemporaneous value of each geopolitical measure [in column headings] minus the cross-sectional weighted average of the measure in that given quarter.

Table A4: Effects of changes in financial sanctions and in monetary policy on cross-border lending flows Model [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] Dependent variable: Bilateral cross-border lending flows Σ∆ Shadow Interest Rate {t-1 to t-4} -6.861*** -10.340*** -7.115*** -11.237*** -7.999*** -11.942*** -8.260*** -13.697*** np np [0.747] [0.952] [0.745] [0.955] [0.843] [1.009] [0.942] [1.159] np np Σ [∆Shadow Interest Rate * ∆ Financial Sanctions] {t-1 to t-4} -10.340*** -42.240*** -11.942*** -309.633*** -99.888*** [0.952] [8.655] [1.009] [45.041] [28.979] Σ∆ Financial Sanctions {t-1 to t-4} 11.581** 23.660*** 11.312** 26.634*** 6.203 57.795*** np np 0.512 44.972*** [5.205] [6.175] [5.255] [6.225] [8.743] [14.245] np np [8.524] [16.080] Observations 38,807 38,807 38,807 38,807 38,807 38,807 38,807 38,807 38,807 38,807 Time FE Yes Yes Yes Yes -- -- -- -- -- -- Borrower FE Yes Yes -- -- -- -- -- -- -- -- Borrower*Time FE No No No No Yes Yes -- -- -- -- Source*Borrower FE No No Yes Yes No No -- -- No No Source*Borrower* Time FE No No No No No No Yes Yes No No Borrower*Currency* Time FE No No No No No No No No Yes Yes Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 "np" indicates that the variable is subsumed by the included set of fixed effects. 14

Table A5: Effects of changes in trade sanctions and in monetary policy on cross-border lending flows Model [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] Dependent variable: Bilateral cross-border lending flows Σ∆ Shadow Interest Rate {t-1 to t-4} -6.826*** -9.952*** -7.091*** -10.731*** -8.055*** -11.789*** -8.262*** -13.205*** np np [0.747] [0.949] [0.745] [0.952] [0.843] [1.010] [0.942] [1.160] np np Σ [∆Shadow Interest Rate * ∆ Trade Sanctions] {t-1 to t-4} -67.309*** -99.097*** -40.469* -406.747*** -21.29 [13.543] [14.615] [22.026] [93.236] [21.388] Σ∆ Trade Sanctions {t- 1 to t-4} 10.053* 32.648*** 18.385*** 58.435*** 11.198 18.863** np np 18.017** 21.586** [6.000] [8.124] [6.383] [9.311] [8.434] [9.376] np np [8.162] [9.096] Observations 38,807 38,807 38,807 38,807 38,807 38,807 38,807 38,807 38,807 38,807 Time FE Yes Yes Yes Yes -- -- -- -- -- -- Borrower FE Yes Yes -- -- -- -- -- -- -- -- Borrower*Time FE No No No No Yes Yes -- -- -- -- Source*Borrower FE No No Yes Yes No No -- -- No No Source*Borrower* Time FE No No No No No No Yes Yes No No Borrower*Currency* Time FE No No No No No No No No Yes Yes Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 "np" indicates that the variable is subsumed by the included set of fixed effects. 15

Table A6: Effects of changes in borrower and source geopolitical risk and in monetary policy on cross-border lending flows Model [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] Borrower Borrower Borrower Borrower Borrower Geopolitical measure: Source GPR Source GPR Source GPR Source GPR Source GPR GPR GPR GPR GPR GPR Dependent variable: Bilateral cross-border lending flows Σ∆ Shadow Interest Rate {t-1 to t-4} -3.083*** -3.639*** -3.595*** -4.150*** -2.642*** -4.351*** -4.136*** -5.692*** np np [0.417] [0.340] [0.417] [0.340] [0.466] [0.384] [0.476] [0.409] np np Σ [∆Shadow Interest Rate * ∆ GPR] {t-1 to t-4} -3.083*** -5.053*** -21.270*** -5.429*** -27.260*** -1.559** -23.361*** -1.648* np 1.604* [0.417] [0.609] [1.055] [0.614] [1.494] [0.689] [1.467] [0.865] np [0.882] Σ∆ GPR {t-1 to t-4} 13.368*** 2.085*** 13.803*** 2.198*** np -1.119 np np np -0.155 [1.113] [0.764] [1.104] [0.761] np [0.847] np np np [0.914] Observations 95,833 158,799 95,833 158,799 95,833 158,799 95,833 158,799 95,833 158,799 Time FE Yes Yes Yes Yes -- -- -- -- -- -- Borrower FE Yes Yes -- -- -- -- -- -- -- -- Borrower*Time FE No No o No Yes Yes -- -- -- -- Source*Borrower FE No No Yes Yes No No -- -- No No Source*Borrower*Time FE No No No No No No Yes Yes No No Borrower*Currency* Time FE No No No No No No No No Yes Yes Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 "np" indicates that the variable is subsumed by the included set of fixed effects. 16

Table A7: Breakdown by borrowers' country: OECD vs non-OECD countries; Effects of changes in geopolitical measures and in monetary policy on cross-border lending flows Model [1] [2] [3] [4] [5] [6] Dependent variable: Bilateral cross-border lending flows to non-banks Geopolitical measure: Ideal point distance Total sanctions Relative borrower GPR Country of Borrowers Non-OECD OECD Non-OECD OECD Non-OECD OECD Σ∆ Shadow Interest Rate {t-1 to t-4} -7.151*** -2.973*** -15.595*** -6.284*** -12.894*** -3.325*** [0.616] [0.471] [1.196] [2.079] [1.179] [0.560] Σ [∆Shadow Interest Rate * ∆ Geopolitical measure] {t-1 to t-4} -29.379*** -16.537* -29.614*** 80.983 -3.428* -6.440*** [6.544] [11.843] [4.862] [97.893] [2.018] [0.916] Σ∆ Geopolitical measure {t-1 to t-4} 9.258** 10.650* 17.296*** -70.003 0.855 2.766*** [3.768] [6.468] [3.021] [83.622] [1.985] [0.958] Observations 96,910 71,481 33,843 4,964 30,013 49,572 Time FE Yes Yes Yes Yes Yes Yes Borrower FE -- -- -- -- -- -- Source*Borrower FE Yes Yes Yes Yes Yes Yes Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 "np" indicates that the variable is subsumed by the included set of fixed effects. The non-bank sector comprises of non-bank financial institutions and the non-financial sector. 17