Policy Research Working Paper 10266 Global Bank Lending under Climate Policy Asli Demirguc-Kunt Alvaro Pedraza Fredy Pulga Claudia Ruiz-Ortega Development Economics Development Research Group December 2022 Policy Research Working Paper 10266 Abstract What is the response of bank foreign subsidiaries to climate evidence that banks with low environmental scores exit policy in their host countries? This paper finds that global in response to climate initiatives. The findings show that banks with high environmental performance increase their strengthening climate policy might be a win-win strategy presence in countries after local authorities strengthen their for policymakers—in addition to addressing carbon emis- climate-related actions. Through their foreign subsidiaries, sion reduction, climate-related initiatives also appear to these banks expand their credit by 4.6 percent following attract foreign capital from lenders with strong preferences an increase of one-standard deviation in the host country’s for green assets. climate policy index. Importantly, the paper does not find This paper is a product of the Development Research Group, Development Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at ademirguckunt@gmail.com, apedrazamorales@worldbank.org, fredy.pulga@unisabana.edu.co, and cruizortega@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Global Bank Lending under Climate Policy Asli Demirguc-Kunt1 Alvaro Pedraza2 Fredy Pulga3 Claudia Ruiz-Ortega4 JEL classification: G21, G28, D62, Q54 Keywords: Global banks, climate change, environmental performance 1 ademirguckunt@gmail.com, Non-Resident Fellow, Center for Global Development 2 apedrazamorales@worldbank.org, World Bank 3 fredy.pulga@unisabana.edu.co, Universidad de la Sabana 4 cruizortega@worldbank.org (corresponding author), World Bank. 1818 H Street NW, MSN MC3-633. Washington DC 20433. 202-473-8798. 1. Introduction A central objective of climate policy is the reduction of carbon emissions, either by promoting renewable energies and increasing energy efficiency across sectors, or by imposing restrictions on activities contributing to greenhouse gas (GHG) emissions. However, heterogeneity in the stringency of climate policy among governments can lead to a reallocation of factors across countries, altering the effectiveness of policy actions. On the one hand, multinational firms might shift production to countries with weaker environmental standards to circumvent costly regulations (Hanna, 2010; Chung, 2014; Cai, et al., 2016). Alternatively, fiscal support for green technologies and other government policies that promote the transition to a less-carbon intensive economy might attract foreign capital; strong climate policies might be appealing to multinationals with more efficient and cleaner technologies than their domestic counterparts (Brucal, et al., 2017). In this context, banks, the largest source of external finance for most firms, can play a major role. By adjusting their lending, banks may undermine efforts to combat climate change if they move capital towards firms operating in countries with weaker climate standards, a regulatory arbitrage to circumvent climate policies (Houston & Lin, 2012; Benincasa, et al., 2021). Conversely, banks may support climate strategies if they expand their lending into countries implementing climate policies by financing innovation and greener activities. Understanding if the financial sector, and in particular banks, act as a conduit in the implementation of climate policies remains an open question. In this paper, we examine how global banks –those operating physical branches and subsidiaries in foreign countries– adjust their credit following changes in climate policy. These financial institutions typically expand abroad to follow their international clients and to seek new profit opportunities in host countries. They represent a large share of the banking sector in the countries where they operate,5 and move global liquidity into or away from local economies with important real effects (Bruno & Hauswald, 2014; Giannetti & Ongena, 2012; Poelhekke, 2015). We identify adjustments in the credit supply of global banks in two ways. First, we exploit the cross-section of banks operating in a given country-year and compare changes in the lending of foreign banks to those of their domestic peers as climate policy strengthens. Second, we exploit 5 In 2013, foreign banks accounted for 43 and 36 percent of the total banking assets in developing and high-income countries respectively (Claessens & Van Horen, 2015). 2 variation in the lending of a global bank at a given period across countries. More precisely, using data on foreign bank subsidiaries, we analyze if banks operating in multiple countries change their credit in jurisdictions that advance their climate agenda. To the best of our knowledge, this is the first paper that analyzes the lending response of foreign bank subsidiaries to domestic policies tackling climate change. We find that after authorities in the host country strengthen their climate-related actions, there is a marginal increase in the lending portfolio of banks. This small average effect masks some important heterogeneity in the cross section. In particular, subsidiaries of global banks with the highest scores in environmental performance respond to host-country climate policy by increasing their total credit in that market –credit grows by 4.6% following an increase of one-standard deviation in the country climate policy index. Consistent with this finding, total bank employment at that location increases. Importantly, we do not find any evidence that global banks with low environmental scores exit countries implementing climate policies. That is, there is no evidence that their total credit shrinks in their host countries nor that the banks reduce their labor force in those locations. Similarly, domestic banks, regardless of their environmental score are mostly irresponsive to climate actions by local authorities. Our findings suggest that after the implementation of climate policy, foreign banks with preferences towards green assets tend to increase their lending in that country, leading to a net increase in foreign capital. We use the Climate Policy component of the Climate Change Performance Index as the measure of the strength of policy actions in 58 countries. Published annually since 2007 by Germanwatch e.V., the Climate Policy is a survey of over 500 climate and energy experts from civil societies assessing the resolve of policymakers to address climate change. For instance, the survey aims to capture whether (and to what extent) a country’s authorities are taking concrete actions to reduce GHG emissions and to promote renewable energies (Burck, et al., 2022).6 The Climate Policy measure (CPM) has a key advantage for our analysis. Rather than capturing outcomes in a country, such as actual GHG emission reduction which might take time to materialize, the CPM is indicative of the experts’ perception of new policies, and the 6 Some of these policies include tax incentives for renewable energy, support for technology innovation, and initiatives that induce climate resilience. The CP also aims to capture strategies to reduce forest degradation and other protections to ecosystem biodiversity, and national peat land protection. 3 implementation of previous policy frameworks. Hence, the CPM can have considerable variation within a country and over time. We combine the Climate Policy measure with Fitch Fundamentals, a comprehensive dataset reporting the balance sheets of most commercial banks operating in each country. We obtain the Environmental scores of publicly listed banks (E-scores) from the ESG components available in Refinitiv. We assign to all the subsidiaries of a global bank the same score as the parent institution. With this data, we examine differences across banks in their lending adjustments following changes in the climate policy of a given country. More precisely, we investigate differences between global vs. domestic banks, and across banks with varying environmental performance. One concern about a naïve model where credit growth is regressed on a climate policy index is that there can be country level characteristics correlated with both the climate policy and with bank lending. For instance, an improvement in economic conditions can lead to both the introduction of new climate initiatives and to an increase in bank lending, especially from global banks. To mitigate such concerns, we include country level characteristics that are known to correlate with bank credit, such as local economic conditions and characteristics of the legal environment (De Haas & Van Lelyveld, 2010). Further, we saturate our model with country-year fixed effects to control for unobserved time-varying factors in each country that might affect the overall supply of credit. Our identification effectively compares, within a country, the response across banks. We interpret the finding that only global banks with high environmental scores increase their lending after the introduction of climate policies as evidence of a green-preference channel. Using the sample of global banks, we present an alternative empirical exercise where we compare the credit growth across foreign subsidiaries of the same bank. In this exercise, we control for a series of home-host country pair characteristics that are associated with foreign bank credit (e.g., geographic, and cultural proximity as in Mian, 2006), and include parent bank-year fixed effects. Here, we are comparing for a global bank and year, whether there are shifts in the credit supply across countries. We confirm our finding that global banks with high E-scores increase their credit in host countries that strengthen their climate policy. At the same time, we do not find 4 any evidence that global banks with low environmental performance reduce their supply of credit to countries strengthening their climate actions. Another concern with our empirical exercise is that rather than measuring banks’ preferences for green assets, the environmental score could be capturing other underlying characteristics unrelated to climate objectives. Moreover, because foreign banks are particularly sensitive to the legal environment in the host country (Quian & Strahan, 2007), if climate policies signal an improvement in the institutional framework, banks with high environmental standards might increase their credit for reasons not linked to climate actions. We address these issues in two ways. First, we show that the environmental score is indeed correlated with variables that are related to green outcomes. For instance, we show that global banks headquartered in countries with higher GHG emissions tend to have lower environmental scores, even after controlling for a wide battery of bank observable characteristics. Second, we examine the role of corporate governance in the banks’ response to climate actions, since these might be related to risk taking and to the legal features of the host country (Anginer, et al., 2018). We classify banks by their Refinitiv Governance score and estimate horse-race regression models with both the environmental and governance scores. Our estimates on the credit growth of banks with high environmental standards are mostly unchanged, but importantly, corporate governance does not seem to be driving the banks’ response to climate policies. Overall, these findings provide credence to our interpretation that global banks with strong preferences for green assets expand their lending in countries that strengthen their climate policies. Our paper contributes to a growing literature on climate change and finance.7 For example, it has been well documented that investors ask for a premium to hold assets from firms with high exposure to climate-change risks (Atanasova & Schwartz, 2019; Delis, et al., 2019; Bolton & Kacperczyk, 2021). However, the literature on the implications of climate policies on bank lending has been rather scant. Reghezza, et al. (2022) show that bank reallocate lending away from polluting firms after the Paris Agreement and Miguel et al. (2022) show that banks limit their supply of credit to climate-change exposed firms after the introduction of new capital requirements in Brazil. 7 See Giglio et al. (2021) for a review of this literature. 5 Closely related to our analysis, Benincasa et al. (2021) find that global banks react to higher climate policy stringency in their home country by increasing their cross-border lending through greater participation in syndicated loans. Contrary to our results, these authors document banks’ behavior in the syndicated loan market that is consistent with a regulatory arbitrage channel – whereby banks from countries with strict regulations engage in cross-border activities in jurisdictions with weaker regulations.8 Rather than looking at the syndicated loan market, which is mainly concentrated in the largest corporations in a country, our work focuses on bank lending by subsidiaries of international banks, and their response to climate policy in their host country. Since foreign subsidiaries represent a large share of the lending portfolio of global banks (Cetorelli & Goldberg, 2012), it is important to understand how credit responds to climate mitigation strategies in this context. Additionaly, because subsidiaries are regulated by the authorities at the host country and rely on local retail funding, they might behave similarly to domestic banks (Aldasoro, et al., 2022). Finally, strong climate policies are not neccesarily restrictive of bank activity. For instance, tax incentives for renewable energy and fiscal support for technology innovation might encourage lenders to direct funding to these sectors rather than drive away capital. In a recent paper, Kacperczyk & Peydró (2022) use syndicated loans to show that after a bank announces its commitment to carbon neutrality, it decreases (increases) its lending to firms with more (less) scope-1 emissions.9 The authors interpret the reallocation of credit towards activities with low GHG emissions as evidence of banks’ preferences toward green assets. Our findings are complementary and shed novel light on a key aspect of the green preference channel. Specifically, we show that global banks with high environmental performance increase their exposure to countries that strenghten their climate policy actions. Our results suggest that climate policy might be a win-win strategy for policy makers by directly reducing carbon emissions and also attracting capital from global lenders with strong preferences for green assets. Our paper is also related to the growing literature that examines the exposure of financial institutions to climate change through two different risk drivers: the physical impact of climate 8 There is evidence that banks circumvent tight restrictions on lending activity and more capital requirements by transferring funds to markets with fewer regulations. See for example, Houston & Lin (2012), Ongena et al. (2013), Aiyar et al. (2014), and Karolyi & Taboada (2015). 9 Direct GHG emissions that occur from sources that are controlled or owned by the firm. 6 change and the policy risk of the transition to a less carbon-intensive economy –through the exposure to firms with business models not aligned with a low-carbon setting (Krueger et al., 2020; Seltzer et al., 2020; BIS, 2021; Duniz et al., 2021). Since regulations to fight climate change could erode the value of banks’ credit exposure or the corresponding collateral, supervisory authorities have introduced prudential regulation to account for the impact of climate-related risks. The combination of green policies (e.g., those that promote renewable energies, increase in energy efficiency, and reduce greenhouse gas emissions) with prudential measures implies that banks operate in a complex regulatory framework with substantial trade-offs. We contribute to this literature by providing evidence on how global banks take actions to increment or reduce their presence in countries with more stringent climate policies. 2. Data To examine whether foreign banks’ subsidiaries adjust their lending portfolio in response to climate policies of their host countries, we combine data on: (i) climate policy stringency, (ii) banks’ balance sheets and environmental standards, and (iii) host and home country characteristics. In this section, we provide detailed description of the sample construction and main variables for the analysis. Climate policy. Published annually since 2007, the Climate Change Performance Index (CCPI) tracks countries’ efforts to combat climate change. As an independent monitoring tool, it aims to enhance transparency in international climate politics and enables comparison of climate protection efforts and progress made by individual countries. A module within the CCPI is the Climate Policy measure (CPM), an annual survey among energy policy experts from non- governmental organizations, universities and think tanks within the countries that are evaluated, rating the climate-related measures from their governments. The policies evaluated include initiatives to promote renewable energies, the increase in energy efficiency and other measures to reduce greenhouse gas emissions.10 Within each policy area, experts evaluate both the strength and the level of implementation of the respective policy framework, and rank countries within a range of 0 (the lowest score) to 20 (the highest score). In line with the Paris Agreement, experts also 10 Besides the climate policy component, other categories in the index are the GHG emissions, renewable energy, and energy use. Since the methodology of these categories was modified in 2017, we exclusively focus on the climate policy component, which allows us to compare the progress of countries in terms of climate policy regulation from 2007 to 2020. 7 evaluate the ambition level and the efforts of each country to reduce national emissions, the so- called Nationally Determined Contributions (NDCs). Our sample includes 58 countries with information on the CPM between 2007 and 2020.11 During our sample period, the average CPM for both developed and developing countries is around 10, but since this measure captures government attitudes towards climate actions, there is large variation across countries and even within each country over time (Panels A and B of Figure 1). In 2007 for instance, while the climate policy component scores of developed countries in North America were close to zero, countries in Europe had an average score of about 14. Also, the increase in the climate policy score for North America between 2007 and 2016 was followed by a large contraction, reflecting the exit of the U.S. from the Paris Agreement. Bank-level data. We obtain banks’ yearly financial statements from Fitch Fundamentals. The data set covers both private and publicly traded financial institutions operating across 200 countries and comprises a full set of balance sheet, profitability, and employment information. We restrict the data to banks operating in the 58 countries for which the CCPI is collected and exclude financial institutions with no information on their total assets, common equity, and gross loans. We complement the bank-level data with Factset Revere Geographic Revenue (GeoRev). GeoRev provides annual information of listed companies’ revenue by country, which we use to identify global banks and their subsidiaries. For each country and year in the Fitch data set, we find global banks and their subsidiaries by merging banks by their names in GeoRev. We define global banks as banks that: 1) operate in the same country of their headquarters but report at least 1 percent of their revenue generated in a different country; 2) operate in a different country as their headquarters’ country. Of the 502 banks matched in GeoRev and Fitch, we identify 173 that derived at least 1 percent of their total revenue from their international operations across the 58 countries in our sample. For example, according to GeoRev, Banco Santander, domiciled in Spain, generated 24% of its total revenue from Brazil, 17% from the United States, 11% from the United Kingdom, 8% from Mexico, and 5% from Chile during 2020. Of the 173 cross-border banks in the 11 Of the 58 countries in the CCPI data set, only three do not have information for the entire 2007-2020 period. These countries are Chile, which was added to the CCPI in 2020, and Iceland and Singapore, whose data is available until 2017. 8 sample, 122 report revenues in subsidiaries located in developed countries, compared to 99 in developing countries. For the group of publicly traded banks in the sample, we use the Environmental score (E- score) from the ESG component reported in Refinitiv. This score is benchmarked by industry and captures the relative performance of each company based on their own corporate records. For the environmental pillar, it measures three main themes: emission, innovation, and resource use. Overall, the score captures the percentile ranking of a company relative to its peers, whether it committed to environmental standards such as reduction of CO2 emissions, protection of biodiversity and capital measure capacity to reduce environmental costs, and capability to promote sustainability. The environmental scores are available for 211 banks. Among these, 90 are domestic banks and 121 are global banks. Due to the bounded nature of this measure, we use a logistic transformation, = (1− ). For the transformed variable, the median bank has an environmental score of 1.11, and banks at the 25th and 75th percentiles reporting scores of -0.36 and 2.08, respectively. Country-level data: In some specifications, we exploit variation on the lending of global banks across their subsidiaries. We include in our dataset information on the economic development of countries where global banks operate (i.e., lagged log GDP per capita) as well as geographical and cultural closeness between the countries of operation and origin of global banks (i.e., distance between countries and whether home and host countries share common language), as these characteristics have an important role in cross-border lending (Qian and Strahan, 2007; Giannetti and Yafeh, 2012). The complete summary statistics of our data are displayed in Table 1, with the variable definitions listed in Appendix Table A1. Our final sample is restricted to countries with available CPM data and for banks with E-scores.12 In Panel A, we report the statistics at the country-year level. Of the countries in our sample, 39 percent are developing economies (see Appendix Table A2 for the country list). In our final sample, the median country has 5 publicly listed banks, and 12 While our focus is on global banks, constraining the sample to lenders with reported E-scores largely reduces the sample of domestic banks. In robustness exercises, we compare the behavior of foreign banks to the universe of domestic banks in each country (16,373 in total) and confirm our findings. 9 there are 3 banks operating in countries at the 25th percentile, and 11 banks in countries at the 75th percentile. Panel B of Table 1 presents the summary statistics calculated at the bank-year level in our sample, where subsidiaries of global banks are counted as separate entities, and their information is captured at the country of operation. Banks in our sample have on average 8.5 billion dollars in assets, 4.2 billion dollars in deposits, and yearly credit growth of 9 percent. The number of banks’ employees is also growing over time, although at a smaller pace than credit, at 2 percent per year. In terms of geographical penetration, the average bank operates in 7 countries where climate policy data is collected (between 1 and 11 countries for the 25th and 75th percentile). In Figure 2 we document the differences between listed domestic and global banks in the sample along four observable characteristics. As shown, banks with presence in foreign markets have higher environmental scores but tend to be smaller than banks with only domestic focus. Among listed banks, the distribution of the equity to asset ratio and yearly portfolio growth seems to be similar between these two groups. 3. Methodology Our objective is to measure whether subsidiaries of global banks expand or contract their lending in response to climate policies in the host country. Furthermore, we examine if changes in the climate policy rating have differential effects on the credit supply of foreign bank subsidiaries relative to domestic banks. To do so, we estimate the following bank-country equation: ∆log ,, = 0 + 1 ,−1 ∗ ,, + 2 ,, + 3 ,,−1 + , + ,, (1) The dependent variable is credit growth, ∆log ,, ; the change in the logarithm of the value of loans of each bank b operating in country c between years t-1 and t. GLOBAL is a dummy variable equal to one if bank b is a subsidiary of a foreign bank in year t and zero otherwise. The climate policy measure, ,−1, captures the strength of the climate actions in the host country of the subsidiary. The key parameter of interest in (1) is the estimated coefficient on ,−1 ∗ ,, , denoted by 1 . We control for bank-level characteristics, ,,−1 , which include bank size (log of total assets and log of total deposits) and bank’s common equity ratio (common equity over total assets). Equation (1) includes country-year fixed effects to control for time-varying factors in each country that might affect the overall supply of credit. Our strategy 10 effectively compares, within a country, the response in total credit from domestic and global banks to changes in climate policy. We estimate robust standard errors by double clustering at the bank- year level to account for serial correlation between each lender over time. While equation (1) is useful to compare the behavior of domestic vs. foreign banks, there might be confounding factors that affect the supply of cross-border lending. For instance, economic conditions in the home country of the foreign bank and other market conditions in the countries where the global bank operates could affect the supply of credit in each host country. If the credit growth of a subsidiary is managed by the corporate headquarters (De Haas & Van Lelyveld, 2010), conditional on home and host country conditions, the relative growth of the lending portfolio across locations (i.e., each host country) should be a good proxy for the targets set by the global bank. In other words, our dependent variable is expected to capture how global banks manage their subsidiaries and distribute credit. As an alternative specification, we study the behavior within a global bank in the countries where it operates. To be precise, we estimate the following equation for the group of global banks in the sample: ∆ log ,, = 0 + 1 ,−1 + 2 , + 3 ,,−1 + , + ,, (2) where , are country-level controls that have been shown to affect cross-border credit supply (De Haas & Van Lelyveld, 2010; Karolyi & Toboada, 2015). These include cultural shared aspects between lenders and borrowers (common spoken language), geographic distance between the headquarters and the subsidiary, and the host country’s demographic and macroeconomic characteristics. The key aspect in equation (2) is that we control for bank-year fixed effects. In this setting, we are effectively comparing for a global bank and year, whether there are shifts in the credit supply in response to changes in the climate policy stringency of host countries. 4. Results 4.1 Foreign bank subsidiaries vs. domestic banks We first estimate the basic model, Equation (1), in which the credit growth of a bank depends on the lagged climate policy measure in a country, controlling for bank and country characteristics. The estimates are displayed in columns (1) and (2) of Table 2. The coefficient for the CPM 11 variable, which captures the average response across banks, suggests that banks in our sample increase their credit after the introduction of climate-related actions, that is, after the country’s authorities strengthen their climate policy. The magnitude of the coefficient implies that the average bank expands its credit by 1.12% following an increase of 1-standard deviation of the CPM (0.25 x 4.47 = 1.12). Notably, the coefficient for the interaction between climate policy and the global bank dummy is indistinguishable from zero. Hence, we cannot reject the null hypothesis that the average credit growth of global and domestic banks is the same after climate policy strengthens. It is possible that the aggregate results are masking important heterogenous behavior in the cross section. For example, if banks base their credit decision on their preferences for green assets (Kacperczyk & Peydró, 2022), their reaction to government-sponsored green initiatives might be more pronounced. We explore this channel by further classifying banks by their environmental performance and study whether the response to climate policy depends on the bank’s ex-ante environmental score. To be precise, we examine how banks with better/worse environmental scores adjust their lending portfolio in response to climate policy stringency, using the logistic transformation of the Refinitiv E-score, ,−1 . Banks with high environmental scores include those with established climate-related strategies and those that show strong commitment to higher environmental standards. With this strategy we aim to compare if there are any differences in credit allocation depending on the banks’ environmental policies. We estimate the following equation: ∆log ,, = 0 + 1 ,−1 + 2 ,−1 ∗ ,−1 + 3 ,, + 4 ,−1 ∗ ,, + 5 ,−1 ∗ ,, + 6 ,−1 ∗ ,, ∗ ,−1 + ,,−1 + , + ,, (3) The set of controls are the same as those introduced in equation (1). In particular, we estimate equation (3) with country-year fixed effects to control for time-varying factors in each country. Effectively, we are comparing within a host country and a year, the response in total credit between domestic and global banks, and by environmental performance (columns 3 and 4). We find that the response to changes in climate policy is concentrated among global banks, and more specifically, among those with high environmental standards. The positive and statistically significant coefficient for the triple interaction, ENV * GLOBAL * CPM, indicates that these banks increase their credit after the host country strengthens its climate policy. The magnitude of 12 the coefficient in column (4) implies that a global bank with E-score in the 75th percentile of the distribution expands its credit by 2.38% more than a bank in the 25th percentile. Domestic banks, on the other hand, do not appear to change their lending volumes following climate initiatives from local authorities. In Figure 2, we show that the distribution of environmental scores is different between domestic and foreign banks, with the latter mostly skewed towards higher scores. These differences might emerge from low coverage or even variations in the scoring methodology for banks with exclusive domestic focus. To deal with this issue, we replace the variable ENV by a dummy variable equal to 1 if a domestic (global) bank has an environmental score above the 50th- percentile of the distribution of domestic (global) banks in year t-1 and zero otherwise; that is, we condition the distribution of E-scores separately for domestic and global banks. The results, presented in columns (5) and (6), confirm our main finding: The response to changes in climate policy is largely driven by the actions of banks with high environmental standards. The magnitude of the coefficient for the triple interaction ENV * GLOBAL * CPM (column (6)), implies that global banks with E-scores above the median substantially expand their credit by 4.6% in response to a 1-standard deviation increase in the CPM of the host country (1.03 x 4.47 = 4.6). As an alternative measure for banks’ expansion in a market, we examine whether banks grow their staff in a particular country. That is, we replace the dependent variable in models (1) and (3) by a bank’s employment growth, ∆log ,, , which is the change in the logarithm of the number of employees for each bank per year. The results, presented in Table 3, further support the view that global banks with high environmental standards increase their focus in countries with strong climate policies. In particular, estimates in column 6 indicate that the expansion of bank staff in response to changes in a country’s climate policy is exclusive of global banks with high E- scores. More concretely, global banks with E-scores above the median grow their employment by 4% for each 1-standard deviation increase in the CPM of a country. Throughout the paper, we benchmark the response of global banks relative to domestic banks operating in the same country. One limitation to our analysis is that by focusing on banks with environmental scores, the number of domestic banks is largely reduced in the final sample. As a robustness test, we examine the changes in credit and employment growth without excluding domestic banks, regardless of the availability of environmental performance data. In this exercise, 13 we compare the behavior of global banks with high and low environmental performance relative to the universe of domestic banks operating in each country. The results show that domestic banks do not alter their lending volume or employment in response to climate policies introduced by local authorities (Appendix Table A3). In contrast, global banks with high E-scores respond to the strengthening of climate policies, captured by increases in the CPM, by expanding their presence in these markets. Figure 3 summarizes our findings. Each panel plots the estimated difference (and associated confidence interval) of credit and employment growth across banks in response to a 1- standard deviation increase in the CPM. The figure compares three groups of banks: (i) global banks with E-scores above vs. below the median, (ii) global banks above the median E-score vs. the universe of domestic banks, and (iii) global banks below the median E-score vs. all domestic banks. Across all panels, the evidence suggests that the strengthening of climate policies in a country, rather than driving capital away, is attracting foreign lenders with preferences for green assets. 4.2 Within global banks So far, we compared how total credit growth differs between foreign and domestic banks when climate policy stringency changes in a given country. Although we interpret our coefficient of interest as the effect of the host country’s climate policy stringency on foreign bank subsidiaries relative to domestic banks, a bank’s willingness to grant credit through its subsidiaries may be affected by other aspects, such as markets’ characteristics where the bank is operating or economic conditions in its home country. To further rule out potentially confounding factors, we examine the results from the angle of the global bank. Specifically, we estimate equation (2) where we control for bank-year fixed effects. In this setting, we are effectively comparing for a global bank and year, whether there are shifts in the credit supply in response to changes in the climate policy actions of host countries. In addition, we also control for country-level aspects that have been shown to affect cross-border credit supply (De Haas & Van Lelyveld, 2010; Karolyi & Toboada, 2015): (i) cultural shared aspects between lenders and borrowers (common spoken language), (ii) geographic distance between the headquarters and the subsidiary, and (iii) the host country’s demographic and macroeconomic characteristics. 14 Results for equation (2) are presented in Table 4. The estimated coefficient, 1 in column (1), while positive is not statistically significant. That is, we cannot reject the null hypothesis that on average, global banks keep their lending portfolios constant following changes in climate policy. In columns (2) and (3), we explore the heterogenous response of global banks with different environmental standards by including the interactions of the E-score with ,−1 . We further confirm the result that only banks with high environmental scores react to climate measures, by increasing their supply of credit precisely in countries where local authorities are strengthening their green agenda. Estimates in column 3 indicate that for a global bank in the median of the environmental score distribution, an increase of 1-standard deviation in the CPM in a host country results in a credit expansion of 0.7% ([-0.313+ 0.291*1.61]*4.47 = 0.7) and up to 1.6% for a bank in the 75th percentile of the distribution. Importantly, we do not find evidence that international banks with low environmental scores reduce their credit supply in host countries where authorities are strengthening their climate-related policies. We further confirm our findings when we estimate the growth in bank employment –global banks increase their labor force in subsidiaries where authorities improve their climate actions (columns 4 to 6 in Table 4). 5. Alternative channels and robustness tests In our analysis, we assume that the environmental factor reported by Refinitiv captures the corporate policies of banks related with environmental concerns and climate-change. If this is the case, the E-score should be a good proxy for the preference over green assets, and in fact, a large number of academic papers have used this measure under the same assumption. However, a major concern with our empirical exercise is that the environmental score, in addition to measuring green preferences, could be capturing other underlying characteristics unrelated to climate objectives. For example, risk preferences, or preferences over legal frameworks. To take a closer look at this issue, we consider a host of regressions in which the dependent variable is the environmental performance of banks, measured by the one-year ahead logistic transformation of the E-score. The regressors include a battery of bank observable characteristics, such as size and geographical focus. Because the corporate governance factor in the ESG measure is often correlated with the E-score, we include the G-score as a regressor. The results are presented in Table 5 (columns 1-3). Larger banks with global focus tend to have higher E-scores. As expected, the environmental performance of a bank is highly correlated 15 with the previous-year governance score. In columns (2) and (3) we include the per capita GHG emissions of the country where the bank is headquartered. As shown, banks from countries with higher GHG emissions tend to have lower environmental scores. On the contrary, when we estimate a similar model where the dependent variable is the one-year ahead logistic transformation of the G-score, GHG emissions are not related to corporate governance (columns 4-6). In this case, only bank size and whether a bank is global are correlated with corporate governance policies of the bank. The evidence thus suggests that the E-scores indeed capture specific properties that relate to green aspects of banks. Given the relationship between governance and environmental performance, we examine the role of the G-score in banks’ response to climate actions. Since foreign banks are sensitive to the legal environment in the host country (Quian & Strahan, 2007), and if climate policies are used as a signal of an improvement in the institutional framework, banks classified with high environmental standards might increase their credit for reasons not linked to climate actions. To the extent that the corporate governance of a bank is related to its risk taking and to its preferences for legal features of the host country (Anginer, et al., 2018), we could differentiate whether climate policies affect credit growth through green preferences or through other institutional conditions. To do this, we include in equation (3) the governance score and its interactions with the climate policy measure and the global bank dummy. We then estimate horse-race regression models (Table 6). Our estimates on the expansion of credit and employment of global banks with high environmental standards are mostly unchanged. In addition, the corporate governance factor does not seem to be driving the banks’ response to climate policies. Overall, these findings are in line with our interpretation that global banks with strong preferences for green assets expand their lending in countries that strengthen their climate policies. Finally, we explore other potential effects from climate policies on bank lending. For instance, fiscal support for green technology innovations might encourage risk taking among firms and lenders. At the same time, climate policies might ease the funding conditions for banks, especially those with higher environmental standards. To take a closer look at these issues, we examine if climate-related actions by domestic authorities are associated with changes in the risk profile of bank loans and to total bank deposits. More precisely, we estimate equation (3) using three separate dependent variables: the yearly percentage change in (i) loan loss provisions, (ii) the 16 share of non-performing loans, and (iii) deposits (results in Table A.4). Our evidence indicates that while global banks with high E-scores expand their presence in countries following an increase in the CPM, their provisions and share of non-performing loans remain constant. We also find that deposits remain mostly unchanged after climate policy actions, even among global bank subsidiaries with high environmental performance. Overall, the reallocation of credit towards jurisdictions with stronger climate policy does not appear to arise as a mechanical result from greater available funds, through increased deposits. Rather, our evidence is mostly consistent with the view that banks with high environmental standards respond to climate policy actions by increasing their lending in that location. Such credit expansion, in turn, does not appear to yield riskier loan portfolios for banks since nonperforming loans and loan loss provisions remain unchanged. 6. Conclusions This paper studies a growing yet understudied aspect of global banking, namely, the response from foreign banks to climate change policy actions. We ask whether global banks adjust their credit growth in foreign subsidiaries when regulatory authorities of host countries strengthen their climate policies. Using a sample of 120 global banks and an index of climate policy in 58 countries between 2007 and 2020, we find that the response to policy measures is largely driven by the ‘green profile’ of banks. Subsidiaries of foreign banks with high environmental standards (i.e., those with highest scores in emission, innovation, and resource use), increase their credit and overall presence following the implementation of climate-related actions in host countries. Importantly, ‘brown’ banks, those with low environmental scores, are mostly unresponsive to the host country climate policy. In the paper, we do not distinguish between types of climate policies. For example, we cannot analyze whether policies target technology subsidies, carbon pricing, or the introduction of performance standards. Instead, we capture the strength of the overall policy framework in a country through the Germanwatch e.V. climate policy measure. Future work in this area should focus on the response from banks to different climate change policies. In addition, there may be important credit reallocation across sectors and firms that is omitted in our analysis. The extent to which banks adjust their portfolios along these dimensions might have important implications for the effectiveness of climate mitigation strategies. Also, while the link between international 17 institutions and local environmental performance remains a controversial issue, there is evidence that foreign corporations transfer environmentally friendly technologies and practices to their foreign owned plants (Brucal, et al., 2017). It is possible that complementarities between green lending from global banks and green foreign direct investment might further amplify the impact of climate policies. Whether such complementarities are present, and the extent of their role, remain open questions. In conclusion, the paper has important policy implications. Our results highlight a selection mechanism whereby government commitments to address climate change also attract foreign banks with strong preferences for green assets, rather than spur capital flight. Therefore, climate policies appear to be a win-win strategy for policy makers, improving the environment by directly addressing carbon emission reduction, and also for attracting foreign finance. 18 References Aiyar, S., Calomiris, C. & Wieladek, T., 2014. Does macro-prudential regulation leak? Evidence from a UK policy experiment. Journal of Money, Credit and Banking, 46(1), pp. 181-214. Aldasoro, I., Caparusso, J. & Chen, Y., 2022. Global banks' local presence: A new lens. BIS Quarterly Review, pp. 31-43. Anginer, D., Demirguc-Kunt, A., Huizinga, H. & Ma, K., 2018. Corporate governance of banks and financial stability. Journal of Financial Economics, 130(2), pp. 327-346. Benincasa, E., Kabas, G. & Ongena, S., 2021. "There is No Planet B", but for Banks There are "Countries B to Z. CEPR Discussion Paper DP16665. BIS, 2021. Climate-Related Risk Drivers and Their Transmission Channels. Supervision, Basel Committee on Banking. Brucal, A., Javorcik, B. & Love, I., 2017. Pollution havens or halos? Evidence from acquisitions in Indonesia. Meetings Paper, Volume 306. Bruno, V. & Hauswald, R., 2014. The real effect of foreign banks. Review of Finance, Volume 18, pp. 1683-1716. Burck, J. et al., 2022. Climate change performance index: Background and methodolgy, Berlin: Germanwatch e.V., NewClimate Institue, Climate Action Network-International. Cai, X., Lu, Y., Wu, M. & Yu, L., 2016. Does environmental regulation drive away inbound foreign direct investment? Evidence from a quasi-natural experiment in China. Journal of Development Economics, Volume 123, pp. 73-85. Cetorelli, N. & Goldberg, L., 2012. Banking globalization and monetary transmission. The Journal of Finance, 67(5), pp. 1811-1843. Chung, S., 2014. Environmental regulation and foreign direct investment: Evidence from South Korea. Journal of Development Economics, Volume 108, pp. 222-236. Claessens, S. & Van Horen, N., 2015. The impact of the global financial crisis on banking globalization. IMF Economic Review, 63(4), pp. 868-918. De Haas, R. & Van Lelyveld, I., 2010. Internal capital markets and lending by multinational bank subsidiaries. Journal of Financial Intermediation, Volume 19, pp. 1-25. Duniz, N., Naqvi, A. & Monasterolo, I., 2021. Climate sentiments, transition risk, and financial stability in a stock-flow consistent model. Journal of Financial Stability, Volume 54, p. 100872. Giannetti, M. & Ongena, S., 2012. “Lending by example”: Direct and indirect effects of foreign banks in emerging markets. Journal of International Economics, 86(1), pp. 167-180. Hanna, R., 2010. US environmental regulation and FDI: Evidence from a panel of US-based multinational firms. American Economic Journal: Applied Economics, Volume 2, pp. 158-189. Houston, J. F. & Lin, C. M. Y., 2012. Regulatory arbitrage and international bank flows. The Journal of Finance, 67(5), pp. 1845-1895. 19 Kacperczyk, M. & Peydró, J., 2022. Carbon emissions and the bank-lending channel. https://ssrn.com/abstract=3915486 or http://dx.doi.org/10.2139/ssrn.3915486. Kalemli-Ozcan, S., Papaioannou, E. & Peydro, J.-L., 2010. What lies beneath the euro's effect on financial integration? Currency risk, legal harmonization, or trade?. Journal of International Economics, 81(1), pp. 75-88. Karolyi, A. & Toboada, A., 2015. Regulatory arbitrage and cross-border bank acquisitions. The Journal of Finance, 70(6), pp. 2395-2450. Krueger, P., Sautner, Z. & Starks, L., 2020. The importance of climate risk for institutional investors. The Review of Financial Studies, 33(3), pp. 1067-1111. Mian, A., 2006. Distance constraints: The limits of foreign lending in poor economies. The Journal of Finance, 61(3), pp. 1465-1505. Miguel, F., Pedraza, A. & Ruiz-Ortega, C., 2022. Climate-change capital requirements: Bank lending and real effects. Working Paper. Morrison, A. & White, L., 2009. Leveling playing fields in international financial regulation. The Journal of Finance, 64(3), pp. 1099-1142. Ongena, S., Popov, A. & Udell, G., 2013. ‘‘When the cat’s away the mice will play’’: Does regulation at home affect bank risk-taking abroad. Journal of Financial Economics, Volume 108, pp. 727-750. Poelhekke, S., 2015. Do global banks facilitate foreign direct investment?. European Economic Review, Volume 76, pp. 25-46. Quian, J. & Strahan, P., 2007. How laws and institutions shape financial contracts: The case of bank loans. The Journal of Finance, 2(6), pp. 2803-2834. Reghezza, A. et al., 2022. Do banks fuel climate change?. Journal of Financial Stability, Volume 101049. Seltzer, L., Starks, L. & Zhu, Q., 2020. Climate regulatory risks and corporate bonds. Nanyang Business School Research Paper, pp. 20-05. 20 Figure 1. Climate policy measure (CPM) by regions Panel A. Developed Countries Panel B. Developing Countries Notes: The figure reports the yearly average values of the CPM across the three regions of the 35 developed countries (Panel A) and the six regions of the 23 developing countries (Panel B) for which data is collected. 21 Figure 2. Characteristics of global vs. domestic banks Notes: The figure reports the distribution of banks in our sample along four observable characteristics (i.e., environmental score, size, equity to asset ratio, and yearly credit growth) sorted by banks’ geographical focus. 22 Figure 3. Marginal effects: 1-standard deviation increase in the Climate Policy Measure Notes: The panels plot the point estimates and confidence intervals of the difference in credit and employment growth between: (i) global banks with E-scores above and below the median, (ii) global banks with E-scores above the median vs. all domestic banks, and (iii) global banks with E-scores below the median vs. all domestic banks. Estimates in Panels A and B control for fixed effects at the country and year level. Estimates in Panels C and D control for fixed effects at the country*year level. 23 Table 1. Summary Statistics Mean p50 p25 p75 SD # Obs. Panel A. Country-Year Data Developingc 0.39 0.00 0.00 1.00 0.49 694 CPMc,t 9.99 10.18 7.17 12.86 4.14 694 Number of banksc,t 8.45 5.00 3.00 11.00 9.69 694 Number of global banksc,t 6.25 4.00 2.00 8.00 5.62 694 Population growthc,t 0.67 0.63 0.08 1.27 0.90 694 GDP per capitac,t 9.84 9.92 9.15 10.71 1.02 694 Exchange ratec,t 0.99 1.00 0.93 1.05 0.10 692 Unemployment ratec,t 7.47 6.47 4.82 8.96 4.23 694 GHG per capitac,t 7.75 6.74 4.60 9.54 4.39 694 Panel B. Bank-Year Data Credit Growthb,c,t 0.09 0.06 -0.01 0.16 0.24 5,449 Employment Growthb,c,t 0.02 0.00 -0.04 0.05 0.29 3,012 CPMc,t 9.82 9.98 6.92 12.86 4.47 5,863 ENVb,t 0.81 1.11 -0.36 2.08 1.57 5,863 ENVb,t (Global banks) 1.26 1.61 0.46 2.27 1.34 4,339 GOVb,t 0.57 0.71 -0.19 1.46 1.22 4,852 Equity ratiob,c,t 0.14 0.10 0.07 0.14 0.13 5,744 Assetsb,c,t 9.05 8.99 7.48 10.63 2.35 5,838 Depositsb,c,t 8.34 8.50 6.71 10.25 2.77 5,516 NPL Growthb,c,t 0.00 0.00 0.00 0.00 0.02 4,132 Provisions Growthb,c,t 0.00 0.00 0.00 0.00 0.02 4,267 Deposits Growthb,c,t 0.11 0.07 0.00 0.17 0.28 5,229 GLOBALb,c,t 0.74 1.00 0.00 1.00 0.44 5,863 Number of countriesb,c,t 7.00 4.00 1.00 11.00 6.92 5,863 Languageb,c 0.49 0.00 0.00 1.00 0.50 5725 Distanceb,c 5.13 7.00 0.00 8.83 4.03 5725 DEVELOPING (home country)b,c,t 0.28 0.00 0.00 1.00 0.45 5,863 Notes: The table displays the summary statistics of the data set at the country-year level (Panel A) and at the bank-year level (Panel B). The sample is restricted to countries for which CPM data is collected banks with environmental scores. See Table A1 in the Appendix for variable definitions. 24 Table 2. Credit Growth Response to Movements in Climate Policy Continuous ENVb,t-1 Discrete ENVb,t-1 (1) (2) (3) (4) (5) (6) CPMc,t-1 0.253* 0.178 0.356*** [0.131] [0.176] [0.110] ENVb,t-1 0.001 0.013 0.025 0.039 [0.009] [0.012] [0.023] [0.036] ENVb,t-1 * CPMc,t-1 -0.087 -0.154 -0.315 -0.426 [0.084] [0.099] [0.191] [0.290] GLOBALb,c,t -0.021 0 -0.009 0.011 0.006 0.044 [0.017] [0.028] [0.022] [0.031] [0.017] [0.038] ENVb,t-1 * GLOBALb,c,t -0.023* -0.031** -0.065* -0.095** [0.013] [0.013] [0.034] [0.038] CPMc,t-1 * GLOBALb,c,t -0.002 -0.302 -0.14 -0.42 -0.340* -0.777** [0.140] [0.251] [0.169] [0.245] [0.173] [0.299] ENVb,t-1 * GLOBALb,c,t * CPMc,t-1 0.249* 0.310** 0.763** 1.030*** [0.129] [0.126] [0.337] [0.325] Constant 0.8 0.228 0.789 0.234 0.759 0.23 [0.522] [0.200] [0.512] [0.196] [0.519] [0.197] Observations 4,806 4,749 4,806 4,749 4,806 4,749 R-squared 0.271 0.402 0.272 0.403 0.272 0.404 Country FE Yes Yes Yes Year FE Yes Yes Yes Bank FE Yes Yes Yes Yes Yes Yes Country-Year FE Yes Yes Yes Notes: The table reports OLS estimates of regressions at the bank-year level summarized in equation 1 for the sample of countries for which CPM data is collected, and the sample of banks with environmental scores. The dependent variable corresponds to the yearly credit growth of a bank in a country. The variable ENVb,t-1 is the logistic transformation of the environmental score of bank b in year t-1. Columns 1, 4 and 5 restrict the sample to domestic banks. Columns 2, 6 and 7 restrict the sample to global banks. Columns 3, 8 and 9 pool domestic and global banks together. Controls at the country-year level include the change in exchange rate, GHG emissions per capita, lagged log GDP per capita, lagged log population growth, and lagged log unemployment rate. Other bank controls include the assets, deposits and equity ratios of banks, all in logs and lagged one year. Standard errors are reported in brackets and are doubled clustered at the bank and year levels. *, **, *** denote significance at the 10, 5 and 1 percent levels. Detailed variable definitions are provided in Table A1 in the Appendix. 25 Table 3. Bank Employment Growth Response to Movements in Climate Policy Continuous ENVb,t-1 Discrete ENVb,t-1 (1) (2) (3) (4) (5) (6) CPMc,t-1 0.338 0.3 0.296 [0.280] [0.332] [0.246] ENVb,t-1 0.007 0.003 0.011 -0.005 [0.016] [0.016] [0.043] [0.039] ENVb,t-1 * CPMc,t-1 0.085 0.072 -0.041 0.194 [0.245] [0.132] [0.424] [0.350] GLOBALb,c,t 0.016 0.026 0.074** 0.079 0.064* 0.084** [0.030] [0.025] [0.031] [0.046] [0.032] [0.033] ENVb,t-1 * GLOBALb,c,t -0.036* -0.029 -0.087** -0.098** [0.019] [0.022] [0.032] [0.042] CPMc,t-1 * GLOBALb,c,t -0.05 -0.076 -0.642 -0.659 -0.503 -0.555 [0.366] [0.305] [0.393] [0.402] [0.310] [0.344] ENVb,t-1 * GLOBALb,c,t * CPMc,t-1 0.368 0.360** 0.937 0.965** [0.301] [0.157] [0.530] [0.373] Constant -0.002 0.129 -0.185 0.147 -0.21 0.112 [1.052] [0.354] [1.091] [0.350] [1.099] [0.347] Observations 2,728 2,607 2,728 2,607 2,728 2,607 R-squared 0.161 0.374 0.164 0.376 0.163 0.376 Country FE Yes Yes Yes Year FE Yes Yes Yes Bank FE Yes Yes Yes Yes Yes Yes Country-Year FE Yes Yes Yes Notes: The table reports OLS estimates of regressions at the bank-year level summarized in equation 1 for the sample of countries for which CPM data is collected, and the sample of banks with environmental scores. The dependent variable corresponds to the yearly employment growth of a bank in a country. The variable ENVb,t-1 is the logistic transformation of the environmental score of bank b in year t-1. Columns 1, 4 and 5 restrict the sample to domestic banks. Columns 2, 6 and 7 restrict the sample to global banks. Columns 3, 8 and 9 pool domestic and global banks together. Controls at the country-year level include the change in exchange rate, GHG emissions per capita, lagged log GDP per capita, lagged log population growth, and lagged log unemployment rate. Other bank controls include the assets, deposits and equity ratios of banks, all in logs and lagged one year. Standard errors are reported in brackets and are doubled clustered at the bank and year levels. *, **, *** denote significance at the 10, 5 and 1 percent levels. Detailed variable definitions are provided in Table A1 in the Appendix. 26 Table 4. Growth of Global banks across subsidiaries (1) (2) (3) (4) (5) (6) Credit Growth Employment Growth CPMc,t-1 0.154 -0.268 -0.313 0.412 -0.184 -0.26 [0.211] [0.272] [0.280] [0.285] [0.257] [0.308] ENVb,t-1 * CPMc,t-1 0.270* 0.291* 0.343* 0.381* [0.138] [0.141] [0.166] [0.185] GDP per capitac,t-1 -0.056 -0.053 -0.059 0.066 0.069 0.065 [0.068] [0.066] [0.066] [0.064] [0.064] [0.065] Exchange ratec,t-1 -0.262** -0.261** -0.260** -0.041 -0.05 -0.055 [0.109] [0.108] [0.109] [0.149] [0.153] [0.159] Assetsb,c,t-1 0.017** 0.017** 0.017** 0.016 0.016 0.016 [0.007] [0.007] [0.007] [0.025] [0.025] [0.025] Equity ratiob,c,t-1 0.412*** 0.415*** 0.414*** 0.118 0.13 0.13 [0.084] [0.084] [0.096] [0.075] [0.085] [0.095] Population growthc,t-1 0.03 0.031 0.031 0.038 0.04 0.043 [0.025] [0.024] [0.024] [0.034] [0.035] [0.037] Unemployment ratec,t-1 -0.003 -0.002 -0.002 -0.003 -0.001 -0.001 [0.005] [0.005] [0.005] [0.006] [0.006] [0.006] Depositsb,c,t-1 -0.012 -0.012 -0.013* -0.021 -0.022 -0.023 [0.007] [0.007] [0.007] [0.027] [0.027] [0.028] GHG per capitac,t-1 -0.011 -0.011 -0.012 0.003 0.003 0.003 [0.012] [0.012] [0.012] [0.008] [0.008] [0.008] Languageb,c -0.012 0.015 [0.020] [0.052] Distanceb,c -0.003 0 [0.003] [0.008] Constant 0.855 0.823 0.909 -0.666 -0.691 -0.64 [0.668] [0.656] [0.668] [0.765] [0.764] [0.749] Observations 2,692 2,692 2,660 1,357 1,357 1,331 R-squared 0.307 0.309 0.307 0.324 0.325 0.326 Bank-Year FE Yes Yes Yes Yes Yes Yes Ctry of Origin-Year FE Yes Yes Yes Yes Yes Yes Host Country FE Yes Yes Yes Yes Yes Yes Notes: The table reports OLS estimates of regressions at the bank-year level summarized in equation 2 for the sample of global banks operating in countries for which CPM data is collected. In columns 1 to 3, the dependent variable corresponds to the yearly credit growth of a global bank in a country. In columns 4 to 6, the dependent variable corresponds to the yearly growth of employment of a global bank in a country. The variable ENVb,t-1 is the logistic transformation of the environmental score of bank b in year t-1. Standard errors are reported in brackets and are doubled clustered at the bank and year levels. *, **, *** denote significance at the 10, 5 and 1 percent levels. Detailed variable definitions are provided in Table A1 in the Appendix. 27 Table 5. Correlates of Environmental and Governance Scores (1) (2) (3) (4) (5) (6) ENV scores GOV scores Assetsb,c,t-1 0.081* 0.068 0.066 0.067** 0.081** 0.081** [0.041] [0.043] [0.044] [0.028] [0.032] [0.033] Equity ratiob,c,t-1 1.685** 1.789** 1.702* 1.881** 2.220*** 2.319*** [0.727] [0.786] [0.787] [0.633] [0.685] [0.719] GOVb,t-1 0.392*** 0.408*** 0.410*** [0.059] [0.060] [0.060] GLOBALb,c,t 1.855*** 1.637*** 1.631*** 0.581*** 0.726*** 0.743*** [0.208] [0.219] [0.220] [0.167] [0.179] [0.181] Number of countriesb,c,t-1 0.032** 0.030* 0.030* 0.026 0.021 0.024 [0.013] [0.014] [0.015] [0.015] [0.017] [0.017] DEVELOPING (home country)b,c,t 0.542** 0.532 0.554 -0.016 0.371 0.355 [0.239] [0.413] [0.420] [0.173] [0.353] [0.351] GLOBALb,c,t * DEVELOPING (home country)b,c,t -1.558*** -1.280*** -1.298*** -0.577** -0.725** -0.740** [0.285] [0.297] [0.296] [0.220] [0.247] [0.249] CPMc,t-1 (home country) 0 -0.016 -0.004 0.01 [0.015] [0.017] [0.011] [0.020] GHG per capitac,t-1 (home country) -0.061*** -0.059*** 0.027 0.027 [0.018] [0.018] [0.016] [0.016] GDP per capitac,t-1 (home country) 0.263 0.206 0.083 0.014 [0.166] [0.180] [0.169] [0.188] CPMc,t-1 (host countries) 2.15 -1.783 [1.682] [1.965] GDP per capitac,t-1 (host countries) 0.066 0.062 [0.108] [0.100] Constant -1.892*** -3.705* -3.827* -0.916** -2.326 -2.239 [0.482] [1.763] [1.781] [0.311] [1.762] [1.753] Observations 1,966 1,769 1,769 2,217 1,993 1,993 R-squared 0.456 0.468 0.469 0.112 0.14 0.142 Year FE Yes Yes Yes Yes Yes Yes Notes: The table reports OLS estimates of regressions at the bank-year level for the sample of banks with environmental and governance scores operating in countries for which CPM data is collected. The dependent variable corresponds to a bank's yearly logistic transformation of the environmental (columns 1-3) and governance (columns 4- 8) score. For global banks operating in multiple countries, host country variables are calculated as the weighted average of such variables, where the weights correspond to the share of total assets of the bank in each host country at a given year. Standard errors are reported in brackets and are doubled clustered at the bank and year levels. *, **, *** denote significance at the 10, 5 and 1 percent levels. Detailed variable definitions are provided in Table A1 in the Appendix. 28 Table 6. Horse race Continuous ENV and GOV scores Discrete ENV and GOV scores (1) (2) (3) (4) Credit Growth Employment Growth Credit Growth Employment Growth ENVb,t-1 0.024* -0.003 0.054 0.006 [0.012] [0.015] [0.032] [0.037] ENVb,t-1 * CPMc,t-1 -0.179 0.01 -0.412 -0.085 [0.127] [0.158] [0.314] [0.367] GLOBALb,c,t 0.031 0.083 0.079 0.090* [0.038] [0.050] [0.045] [0.045] ENVb,t-1 * GLOBALb,c,t -0.03 -0.023 -0.079* -0.1 [0.017] [0.031] [0.040] [0.059] CPMc,t-1 * GLOBALb,c,t -0.648* -0.676 -1.033** -0.609 [0.307] [0.511] [0.372] [0.530] ENVb,t-1 * GLOBALb,c,t * CPMc,t-1 0.296* 0.340* 0.812* 1.112** [0.153] [0.183] [0.417] [0.439] GOVb,t-1 0.005 0.019 0.008 -0.002 [0.011] [0.031] [0.024] [0.043] GOVb,t-1 * CPMc,t-1 -0.119 -0.144 -0.065 0.147 [0.139] [0.258] [0.371] [0.477] GOVb,t-1 * GLOBALb,c,t -0.012 -0.046 -0.04 -0.031 [0.010] [0.027] [0.030] [0.034] GOVb,t-1 * GLOBALb,c,t * CPMc,t-1 0.131 0.174 0.282 -0.057 [0.136] [0.141] [0.389] [0.223] Constant 0.035 0.151 0.034 0.2 [0.251] [0.464] [0.246] [0.458] Observations 3,936 2,396 3,936 2,396 R-squared 0.4 0.387 0.4 0.386 Bank FE Yes Yes Yes Yes Country-Year FE Yes Yes Yes Yes Notes: The table reports OLS estimates of regressions at the bank-year level for the sample of countries for which CPM data is collected, and the sample of banks with environmental scores. In columns 1 and 3, the dependent variable corresponds to the yearly credit growth of a bank in a country. In columns 2 and 4, the dependent variable corresponds to the yearly employment growth of a bank in a country. In columns 1 and 2, the variables ENVb,t-1 and GOVb,t-1 correspond to the logistic transformations of the environmental and governance scores of bank b in year t- 1. In columns 3 and 4, the variables ENVb,t-1 and GOVb,t-1 correspond to indicator variables that equal one if the domestic (global) bank b has a score above the median of domestic (global) banks in year t-1 and zero otherwise. Controls at the bank-country-year level include the assets, deposits and equity ratios of banks, all in logs and lagged one year. Standard errors are reported in brackets and are doubled clustered at the bank and year levels. *, **, *** denote significance at the 10, 5 and 1 percent levels. Detailed variable definitions are provided in Table A1 in the Appendix. 29 Appendix Table A1. Variable Definition Panel A. Country-Year Data Developingc Indicator variable equal to 1 if country c is a developing country and zero otherwise. CPMc,t Climate policy component of the Climate Policy Component of country c in year t, scaled by 100, with values ranging from 0 to 0.2. Number of banksc,t Number of banks of country c in year t. Number of global banksc,t Number of global banks of country c in year t. Population growthc,t Yearly population growth of country c in year t. GDP per capitac,t GDP per capita (in USD logs) of country c in year t. Exchange ratec,t Yearly percentage change of the USD exchange rate of country c in year t. Unemployment ratec,t Unemployment rate of country c in year t. GHG per capitac,t GHG emissions over population of country c in year t. Panel B. Bank-Year Data Credit Growthb,c,t Percentage change of the gross total loans of bank b in country c in year t. Employment Growthb,c,t Percentage change of the number of staff of bank b in country c in year t. CPMc,t Climate policy component of the Climate Policy Component of country c in year t, scaled by 100, with values ranging from 0 to 0.2. ENVb,t Environmental scores of bank b in year t, measured as the log transformation of the score. GOVb,t Governance scores of bank b in year t, measured as the log transformation of the score. Equity ratiob,c,t Common equity over total assets of bank b in country c and year t. Assetsb,c,t Total assets (in million USD logs) of bank b in country c and year t. Depositsb,c,t Total deposits (in million USD logs) of bank b in country c and year t. NPL Growthb,c,t Yearly percentage change of the non-performing loans of bank b in country c and year t. Provisions Growthb,c,t Yearly percentage change of the provisioning rate of bank b in country c and year t. Deposits Growthb,c,t Yearly percentage change of the deposits of bank b in country c and year t. GLOBALb,c,t Indicator variable that equals one if bank b operating in country c at year t is classified as global (based on the matching of Fitch data with GeoRev data). Number of countriesb,c,t Number of countries of operation of global bank b in year t. Languageb,c Indicator variable that equals one if the country of origin of global bank b and the country of operations c is the same, zero otherwise. Distanceb,c Geographic distance (in log km) between the country of origin of global bank b and the country of operations c. DEVELOPING (home country)b,c,t Indicator variable that equals one if bank b in country c at year t is headquartered in a developing country. 30 Appendix Table A2. Countries with CCPI data Developed countries Developing countries Australia Algeria Austria Argentina Belgium Belarus Canada Brazil Croatia Bulgaria Cyprus Chile Czechia China Denmark Egypt, Arab Rep. Estonia India Finland Indonesia France Iran, Islamic Rep. Germany Kazakhstan Greece Malaysia Hungary Malta Iceland Mexico Ireland Morocco Italy Romania Japan Russian Federation Latvia Saudi Arabia Lithuania South Africa Luxembourg Thailand Netherlands Türkiye New Zealand Ukraine Norway Poland Portugal Singapore Slovak Republic Slovenia Korea, Rep. Spain Sweden Switzerland United Kingdom United States Notes: Only three countries do not have information for the entire 2007-2020 period. These countries are Chile, which was added to the CCPI in 2020, and Iceland and Singapore, whose data is available until 2017. 31 Appendix Table A3. Bank Lending Response to Movements in Climate Policy (Including all domestic banks in Fitch) Credit Growth Employment Growth (1) (2) (3) (4) (5) (6) CPMc,t-1 0.03 0.029 -0.042 -0.043 [0.053] [0.053] [0.024] [0.024] GLOBALb,c,t -0.055*** -0.011 [0.018] [0.026] CPMc,t-1 * GLOBALb,c,t 0.289 0.304 [0.165] [0.187] GLOBAL-LowENVb,c,t -0.041** -0.006 0.024 0.042 [0.016] [0.014] [0.025] [0.025] GLOBAL-HighENVb,c,t -0.075** -0.052** -0.038 -0.03 [0.030] [0.018] [0.038] [0.040] GLOBAL-LowENVb,c,t * CPMc,t-1 0.168 -0.122 -0.054 -0.189 [0.172] [0.134] [0.132] [0.178] GLOBAL-HighENVb,c,t * CPMc,t-1 0.419* 0.324* 0.661* 0.631* [0.198] [0.172] [0.356] [0.349] Constant 0.364 0.362 0.693*** 0.316 0.3 0.408*** [0.368] [0.368] [0.093] [0.183] [0.186] [0.076] Observations 174,973 174,973 174,971 134,377 134,377 134,341 R-squared 0.284 0.284 0.318 0.153 0.153 0.167 Country FE Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Bank FE Yes Yes Yes Yes Yes Yes Country-Year FE Yes Yes F-test (pval) 0.118 0.013 0.0805 0.0477 Notes: The table reports OLS estimates of regressions at the bank-year level for the sample of countries for which CPM data is collected. The sample includes all domestic banks in Fitch and global banks with environmental scores. The dependent variable corresponds to the yearly credit growth of a bank in a country. The dependent variables correspond to the yearly credit (columns 1-3) and employment (columns 4-6) growth of a bank in a country. The indicator variable GLOBAL-LowENV (GLOBAL-HighENV) equals one for global banks with environmental scores in year t-1 below (above) the median environmental score. Controls at the country-year level include the change in exchange rate, GHG emissions per capita, lagged log GDP per capita, lagged log population growth, and lagged log unemployment rate. Other bank controls include the assets, deposits and equity ratios of banks, all in logs and lagged one year. F-test (pval) report the p-values of an F-test comparing if GLOBAL-LowENVb,c,t * CPMc,t-1 = GLOBAL-LowENVb,c,t * CPMc,t-1. Standard errors are reported in brackets and are doubled clustered at the bank and year levels. *, **, *** denote significance at the 10, 5 and 1 percent levels. Detailed variable definitions are provided in Table A1 in the Appendix. 32 Appendix Table A4. Alternative Mechanisms Behind Movements in Climate Policy (1) (2) (3) (4) (5) (6) Provisions Growth NPL Growth Deposits Growth CPMc,t-1 -0.038* -0.032* 0.111 [0.018] [0.015] [0.191] ENVb,t-1 0 0 0 -0.001 0.015 0.025** [0.001] [0.001] [0.001] [0.001] [0.010] [0.011] ENVb,t-1 * CPMc,t-1 -0.001 -0.005 0.01 0.013 -0.098 -0.133 [0.008] [0.008] [0.009] [0.008] [0.115] [0.137] GLOBALb,c,t 0.001 0.002 0.001 -0.001 0.001 -0.007 [0.002] [0.002] [0.002] [0.002] [0.024] [0.028] ENVb,t-1 * GLOBALb,c,t -0.001 -0.001 -0.001 0.001 -0.005 -0.018 [0.001] [0.001] [0.001] [0.001] [0.014] [0.016] CPMc,t-1 * GLOBALb,c,t 0.015 0.007 0.004 0.005 0.115 0.26 [0.019] [0.019] [0.014] [0.015] [0.202] [0.270] ENVb,t-1 * GLOBALb,c,t * CPMc,t-1 0.01 0.012 0.003 -0.011 0.072 0.104 [0.012] [0.012] [0.011] [0.009] [0.171] [0.177] Constant -0.043 -0.023 -0.107 -0.034 2.674*** 0.418 [0.061] [0.019] [0.062] [0.019] [0.510] [0.302] Observations 3,828 3,765 3,723 3,617 4,827 4,771 R-squared 0.102 0.33 0.19 0.478 0.276 0.412 Country FE Yes Yes Yes Year FE Yes Yes Yes Bank FE Yes Yes Yes Yes Yes Yes Country-Year FE Yes Yes Yes Notes: The table reports OLS estimates of regressions at the bank-year level summarized in equation 1 for the sample of countries for which CPM data is collected, and the sample of banks with environmental scores. The dependent variables correspond to the yearly growth of loan loss provisioning (columns 1 and 2), of NPLs (columns 3 and 4) and of deposits (columns 3 and 6) of a bank in a country. The variable ENVb,t-1 is the logistic transformation of the environmental score of bank b in year t-1. Controls at the country-year level include the change in exchange rate, GHG emissions per capita, lagged log GDP per capita, lagged log population growth, and lagged log unemployment rate. Other bank controls include the assets, deposits and equity ratios of banks, all in logs and lagged one year. Standard errors are reported in brackets and are doubled clustered at the bank and year levels. *, **, *** denote significance at the 10, 5 and 1 percent levels. Detailed variable definitions are provided in Table A1 in the Appendix. 33