Policy Research Working Paper 10050 How Binding Is Supervisory Guidance? Evidence from the European Calendar Provisioning Franco Fiordelisi Gabriele Lattanzio Davide S. Mare Development Economics Development Research Group May 2022 Policy Research Working Paper 10050 Abstract This paper investigates whether banks respond differently guidance than after its enactment as a binding regulation. to supervisory guidance than to specific regulatory action. This finding is consistent with the notion that the sub- Using a sample of subsidiaries of European banks operating sequent formalization of the supervisory initiative within in developing countries, the study exploits the sequenc- a regulatory framework achieved limited results because ing in the supervisory and regulatory implementation of a it eliminated the flexibility the regulatory authority had reform on provisioning for credit losses for identification, concerning the stringency with which European calendar generally referred to as European calendar provisioning. provisioning was enforced. Finally, the study offers evidence While the reform achieved the intended goal of reducing of a mechanism through which policies in advanced econ- European banks’ nonperforming loan ratios, its effects were omies affect banking outcomes in developing countries to greater during the initial implementation of the supervisory which their local financial authorities should be alert. 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 dmare@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 How Binding Is Supervisory Guidance? Evidence from the European Calendar Provisioning Franco Fiordelisi, Gabriele Lattanzio, Davide S. Mare* JEL Classifications: G21; G28; G32. Keywords: Bank capital adequacy; Bank regulation; Cross-border financial institutions; Financial stability; non-performing loans ________________________________________________ * Franco Fiordelisi is at the University of Essex (UK), Gabriele Lattanzio is at the Nazarbayev University (Kazakhstan), and Davide S. Mare is at the World Bank (US) and University of Edinburgh (UK). We are grateful to Robert Cull, Bektemir Ismailov, Manju Puri, and Jerome Taillard for their comments and suggestions. We also thank Giulia Scardozzi for her excellent research assistance. Davide Mare acknowledges funding from the Knowledge for Change Program at the World Bank. This paper’s findings, interpretations, and conclusions are entirely those of the authors and do not necessarily represent the views of the World Bank, their Executive Directors, or the countries they represent. All errors and omissions are ours. E-mails: Franco Fiordelisi (franco.fiordelisi@essex.ac.uk); Gabriele Lattanzio (gabriele.lattanzio@nu.edu.kz); and Davide Salvatore Mare (dmare@worldbank.org). “The introduction of calendar provisioning represents one of the most disruptive and intrusive interventions among those made in recent years by European supervision, which certainly cannot be accused of timidity”. Italian Association of Financial Industry Risk Managers, Position paper n. 23, p. 7 1 Introduction Since the release of the first capital framework in 1988, the work of the Basel Committee on Banking Supervision (BCBS) has been at the center of the process of international harmonization of regulatory practices in banking. 1 This framework, known as Basel I, and its subsequent developments (Basel Committee on Banking Supervision, 2006: also known as Basel II; Basel Committee on Banking Supervision, 2010 and 2011: also known as Basel III) have indeed established general principles and standards that each member country has independently adopted and enforced over the last three decades. The resulting regulatory environment is thus the product of a global cooperative effort exercised by regulatory agencies, supervisory authorities, banks, and financial companies, all involved in an almost continuous process of consultation and coordination effort aiming at perfecting the resulting framework. Yet, 10 years after the initial enforcement of the second Basel Accord in Europe, the European Central Bank (ECB) first and the European Commission (EC) later set aside this international convention by introducing a unilateral regulatory change aiming at tackling the pressing issue of increasing levels of non-performing loans (NPLs) held by European banks. This initiative, generally referred to as the European calendar provisioning, is peculiar for three main reasons. First, it leaves aside the existing arrangements characterized by a global harmonized effort toward banking regulation. Second, its underpinning idea subverts the principles established by the Basel capital agreements themselves. Indeed, rather than relying on banks’ internal models and capital adequacy assessments, these new rules impose minimum loss coverage requirements to be achieved through write-downs or deductions from regulatory capital depending mechanically on the time elapsed since the default of the considered loan. Third, the complex 1 The Basel Committee on Banking Supervision (BCBS) is a committee of banking supervisory authorities that was established by the central bank governors of the Group of Ten countries in 1974. Over time, BCBS expanded its membership, and, as of year 2021, it is composed by 45 members from 28 jurisdictions, consisting of central banks and authorities with responsibility of banking regulation and supervision. 2 process of adoption and enforcement of the European calendar provisioning represents a unique setting to analyze how effective is supervisory guidance compared to similar regulatory interventions. This new regulation was indeed originally introduced in the form of supervisory guidance by the ECB in 2017. Later in 2018, the ECB provided banks and market participants with a clear statement of supervisory expectations about the provision of non-performing loans in an “Addendum” of the supervisory guidance issued in 2017. These expectations add to the Supervisory Review and Evaluation Process (SREP), carried out by ECB under the so-called Pillar 2. 2 As such, the guidelines released by the ECB were non-binding from a legal perspective, de facto representing a pure supervisory action enforceable exclusively under the principle of “comply or explain” (Keay, 2014). Subsequently, in 2019, the European Parliament and the European Council claimed their power to set regulations and published a minimum coverage framework for non-performing exposures (NPEs). This regulatory initiative (known as “backstop”) formalizes the system of deduction from banks’ common equity Tier 1 (CET1) triggered when the minimum coverage levels set out by the ECB calendar provisioning supervisory guideline are violated. Being a legislative act, this regulatory action is included in the Pillar 1 framework and, as such, is legally binding. 3 That is, following the approval of this regulation banks and supervisors have no residual flexibility concerning the degree to which the calendar provisioning should be enforced. These staggered supervisory and regulatory actions sparked a heated international debate among regulators and practitioners concerning two major issues. First, they effectively brought regulation back in time by foregoing model-based analyses of the risk associated with loan exposures to relying on a basic measure of time elapsed. Second, these interventions called into question the degree to which supervisory guidance issued under the comply or explain framework is perceived to be binding by commercial banks, as compared to the legally binding nature of regulatory actions. Building on this debate, we contribute to these important discussions and to the broader literature assessing the differential impact of supervisory and regulatory initiatives by investigating 2 Pillar 2 refers to the supervisory review process which broadly entails the revision of the assessment of banks of their exposures to risks such as credit risk, operational risk, and market risk. 3 Pillar 1 refers to the regulations concenting the minimum capital banks should hold to absorb eventual unexpected lossess for credit, market and operational risks. 3 the following two research questions. Was this non-model-based regulatory initiative effective in reducing non-performing loan ratios? To what degree do banks perceive binding de facto supervisory guidance as compared to the de iure nature of regulatory actions? To address these questions within a causal framework, our empirical strategy relies on a dynamic Difference-in-Differences (DiD) approach allowing us to compare changes in the riskiness, performance, and loan policies of subsidiaries of European banks operating in developing countries 4 with that of matched domestic banks. Because European banks consolidate worldwide credit exposures under their domestic regulatory framework, their subsidiaries operating in developing countries are directly affected by the adoption of calendar provisioning rules. Conversely, matched banks operating in developing countries are not exposed to the effects of these supervisory and regulatory actions. By exploiting this setting, we can thus eliminate any endogeneity concerns related to the passage of these supervisory and regulatory initiatives. European regulators are indeed at best unlikely to take any actions targeting directly and primarily European subsidiaries operating abroad. Furthermore, our approach allows us to ensure the comparability between the two groups of banks and restore the randomization conditions by adopting several matching approaches and resampling procedures. We document that European banks reacted to the release of the ECB calendar provisioning supervisory guideline by treating it as a binding requirement and decreasing their non-performing loan ratios. The recognition of these losses in banks’ income statements as charge-offs induced an immediate reduction in their regulatory capital (Tier 1 capital) and impaired loan reserves, weakening their capitalization profile. Yet, we document that these persistently higher regulatory costs do not cause a reduction in loan origination. Rather, subsidiaries of European banks appear to shift these increased costs to their customers by charging higher interest rates on new loans. We document that banks target primarily countries featuring weaker definitions of capital requirements and non-performing loans to implement this form of regulatory cost-shifting. This result is consistent with multinational banks exploiting the regulatory leniency of foreign countries to off- load – at least part of – the regulatory costs imposed by domestic regulations and apply lower bank lending standards (Ongena et al., 2013). 4 We define as developing countries those jurisdictions classified by the World Bank as non-high income based on the Gross National Income per capita in current US dollars. 4 All in all, these findings suggest that (1) the studied non-model based supervisory and regulatory interventions achieved the intended goal – that is reducing European banks’ NPL ratios – and (2) supervisory guidance issued under the “comply or explain” framework is perceived by banks as binding while providing supervisors with some flexibility concerning the degree to which such actions should be enforced over time and in the cross-section. Our paper contributes to several strands of the literature. First, we complement previous work on the supervision and regulation of banks. The difference between regulatory and supervisory actions is often blurred, with the vast majority of existing studies focusing on the direct impact of regulatory intervention targeting capital requirements (Bridges et al., 2014) and merger and acquisitions assessments (e.g., Morgan et al., 2004). Studies focusing on banking supervisory monitoring are limited and mostly analyze the information content of supervisory ratings (Hirtle and Lopez, 1999; Berger et al., 2000), and examinations (Berger and Davies, 1998). Recent work investigates specifically whether tougher supervisory standards cause lower loan growth or stricter origination standards, with most studies finding this to be the case (e.g., Bassett et al., 2015; Kiser et al., 2016; Bassett and Marsh, 2017). Other studies analyze the outcome of enforcement actions on bank risk (e.g., Delis and Staikouras, 2011; Pugachev, 2019). More recently, Hirtle et al. (2020) explore the effects of bank supervision on the riskiness, profitability, and growth of U.S. banks, documenting that supervisory activity has a distinct role in mitigating banking sector risk. We contribute to this literature by documenting for the first time whether and how non-legally binding supervisory guidance affects banks’ policies, as compared to the case of the passage of a specular regulatory intervention. Second, our paper provides novel evidence to worldwide policy makers to assess whether non-risk-based regulations are effective in tackling the emerging NPLs phenomenon. The COVID- 19 pandemic has renewed interest in the resolution and management of non-performing loans in the banking industry. Payment and enforcement moratoria have supported borrowers by allowing a temporary halt to their bank repayment obligations (Feyen et al., 2021). Yet, once support is removed, it is not clear which borrowers will be permanently affected and how debtors will adjust to the structural changes in the economy. Rising borrower distress is widely expected to translate into increases in non-performing loans in the banking sector, calling for immediate policy action (World Bank, 2022). These policies, especially in advanced economies, may hurt the stability and 5 functioning of the banking system in developing countries. Our paper provides evidence for up to 98 countries on such a mechanism. The rest of the paper is structured as follows. Section 2 provides background information on the introduction of calendar provisioning and formulates the hypotheses. Section 3 describes the data and variables employed in the estimation. Section 4 illustrates the empirical model. Section 5 discusses the results and section 6 concludes. 2 Institutional setting and hypotheses development The problem of non-performing loans has become particularly severe during the last decade in Europe (Figure 1), pressuring the European Central Bank to consistently set the reduction of NPLs among the supervisory priorities of the Single Supervisory Mechanism (SSM). 5 This growing concern and attention towards NPLs resulted in the development and adoption of calendar provisioning, which ultimately represents the most important supervisory effort exercised to tackle this issue. Figure 1: The trend of non-performing loans in Europe This figure plots the weighted mean level of the NPL ratio (non-performing loans to total gross loans) in percentage points between 2009 and 2020 for the following 17 European countries: Belgium, Bulgaria, Cyprus, Denmark, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Portugal, Spain, Sweden. The weights are computed for each country in a year using the information on GDP at the purchaser’s prices (constant 2015 prices, expressed in U.S. dollars). Data source: IMF Financial Soundness Indicators and World Bank’s World Development Indicators. 8 7.5 7.6 7.3 7 6.6 6.4 6 5.7 5.7 5.1 5.0 5 4.1 4 3.6 3.4 3 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 5 The euro area, and particularly countries in Southern Europe (i.e., Greece, Spain and Italy) was first hit by the 2007- 2009 Global Financial Crisis and then by the Sovereign Financial Crisis (Lane, 2012), bringing NPLs to 11.8% of total loans in 2016, corresponding to an outstanding amount of around €1 trillion. 6 The development of the European calendar provisioning occurred in various stages (Figure 2). In the first phase, the ECB, acting as the main supervisory entity within the single supervisory mechanism in the euro area, issued supervisory guidance for the treatment of non-performing loans. This guidance was launched in consultation in September 2016 and published in March 2017. The document provided banks with realistic and ambitious strategies in a holistic approach, without declaring quantitative targets, to address the NPLs problem. A few months later, the European Council proposed an action plan to tackle non-performing loans in Europe where it considers, within the framework of the ongoing review of the Capital Requirement Regulation/Capital Requirement Directive IV (CRR/CRD IV), to set prudential backstops in the form of compulsory prudential deductions from own funds of NPLs under the Pillar 1 framework addressing potential under-provisioning which would apply to newly originated loans. On October 4, 2017, the ECB reinforced its NPLs guidance for banks by publishing for consultation an addendum that sets out supervisory expectations for minimum levels of prudential provisioning for new NPLs. The final text of the “Addendum” was released in March 2018. The ECB addendum provides “significant banks” (the ones under the ECB direct supervision) with minimum loss coverage requirements, depending only on the time elapsed since the default, to be achieved through write-downs or deductions from regulatory capital. The mechanism, originally meant to apply only to loans defaulting after April 2018, was extended in July 2018 – through a communication of the Single Supervisory Mechanism (SSM) – to preexisting NPLs, which would also be subject to full coverage by 2026. 6 The release of the addendum, only a few months after the European Council proposal, is due to the ECB’s aim to urgently tackle the NPLs problem in a moment of favorable economic conditions in the euro area, 7 while the time implied for a regulatory decision (from the European Commission and Parliament) were significantly longer. That is, banks were asked to respond to the release of the ECB supervisory guidance in 2017 and 2018 while 6 Individual banks are subject to different timeframes, to be defined in their annual SREP. 7 The point is well clarified by Nouy (2017, page 2) “Obviously if the ECOFIN decides to go for a Pillar 1 measure, that’s to say legislative measures, once it is applicable and once it is addressing all the portfolios, we will adapt our own measures. So regarding our measures, these relate to the new NPEs, which is indeed a difference with the ECOFIN, which is targeting new loans. But we think that now is the moment to address NPEs, in particular because we are enjoying good economic conditions in the euro area. If we wait until there is a Pillar 1 text first, and then it is covering only new loans, we have to keep in mind that the full rolling over of the existing loan book can take a decade. So we have also to address, in my view, the future NPLs until we have measures that will be implemented and fully rolled over the whole portfolio. So there is a sequence and we are ready to adapt our own guidance once there is something else. But still, there will stay room for Pillar 2.” 7 discounting the possibility of a future similar regulatory intervention, whose timing and content were however unknown and unforecastable at the time. Figure 2: The Timeline of Non-Performing Exposures supervisory and regulatory reforms This figure summarizes the formal announcements and publications of the main regulatory and supervisory reforms made in Europe to tackle the problem of non-performing exposures. The second phase of the adoption of the European calendar provisioning is represented by the regulatory change introduced by the European Commission in April 2019, when the European Regulation 2019/630 amended Regulation 575/2013 (the CRR) by introducing a mandatory calendar provisioning system (also known as “backstop”). This regulatory action dictates Pillar 1 requirements for all loans granted after the measure came into force. With the inclusion in the CRR, the ECB calendar provisioning rules thus changed their nature from supervisory expectations to binding regulatory requirements. Importantly, a few months after this regulatory change, the ECB revised in August 2018 its supervisory expectations for prudential provisioning for new NPLs to account for the sudden introduction of this new Pillar 1 requirement. The main changes concerned the scope of the ECB’s supervisory expectations for new non-performing exposures 8 (NPEs) 8 and the alignment of the relevant prudential provisioning time frames (the progressive path to full implementation and the split of secured exposures, as well as the treatment of NPEs guaranteed or insured by an official export credit agency) with the Pillar 1 treatment of NPLs set out in the new EU regulation. All other aspects, including specific circumstances, which may make prudential provisioning expectations inappropriate for a specific portfolio/exposure, remain as described in the original Addendum. As one can infer from this complex adoption process, calendar provisioning represents not only a regulatory and supervisory change, but also a turning point in banking regulation. Its underpinning idea is disruptive to the logic of the BCBS accords. Rather than relying on banks’ internal risk-based models (as was the case under the capital frameworks following the Basel I accord), ECB calendar provisioning rules provide banks with an old-fashioned regulation (not consulted and discussed in advance with financial intermediaries) that is based on minimum loss coverage requirements based on simple rules (non-linked to banks internal risk models, risk management sophistication, and the efficiency of judicial systems) depending only on the time elapsed since the default, to be achieved through write-downs or deductions from regulatory capital. In this sense, the first crucial hypothesis to be tested is whether these shocks reached the desired outcomes. H1: The supervisory guidance and regulatory change known as European calendar provisioning were effective in reducing NPL ratios. The timeline and process followed for introducing calendar provisioning within the European supervisory and regulatory frameworks constitute a unique setting to test the marginal effect of banking supervision and regulation. 9 In particular, we expand on recent studies documenting the importance of addressing separately bank supervision from bank regulation (Hirtle, 2020) and assess for the first time if and how binding supervisory guidances issued under the “comply or 8 NPEs is a broader category of non-performing assets which comprehends non-perofrminig loans. The scope of the ECB’s supervisory expectations is limited to NPEs arising from loans originated before 26 April 2019 (not subject to Pillar 1 NPE treatment). NPEs arising from loans originated from 26 April 2019 onwards are subject to Pillar 1 treatment, with the ECB paying close attention to the risks arising from them. 9 See, i.e., The World Bank (2020) and references therein for a review of the relevant literature. 9 explain” framework are. That is, we exploit our unique setting to test the following research hypothesis: H2: The marginal effect of the European Commission’s institutionalization of the ECB supervisory guidance within a regulatory framework is economically and statistically irrelevant. We test H1 and H2 in conjunction to find whether the supervisory guidance issued by the ECB was effective in reducing NPL ratios and whether the regulatory intervention adds to this supervisory intervention. In particular, failing to reject H1 and H2 would imply that supervisory guidance issued under the “comply or explain” framework is perceived as binding as it would spur a decline in NPL ratios, while its subsequent transformation into a regulatory action would not provide any statistically meaningful contribution. Next, we focus on the mechanism through which European banks eventually respond to this policy by reducing the size of their NPL portfolios. We identify various economic channels that may have driven the NPLs drop. First, banks may decrease NPLs by using their impairment loan reserves (H3). This would be the most immediate reaction to the ECB guidelines: banks use the equity reserves made in the past for credit losses to comply with the ECB’s expectations. In such a way, banks do not reduce their profits, but weaken their capitalization profile, thus limiting their ability to cope with sudden, large credit risk shocks. Second, banks can go further to decline their NPLs than using reserves by aggressively writing off all those NPLs that were potentially in conflict with the new calendar provisioning guidelines (H4). By recognizing losses in their income statements, banks would indeed be able to rapidly shrink their NPL ratios, ultimately reaching compliance with the new regulatory requirements. To sum up, the following research hypotheses enable us to identify the channel through which banks shrink their NPL ratios: H3: The adoption of the European calendar provisioning induces banks to increase their charge-off. H4: The adoption of the European calendar provisioning induces banks to decrease their impairment loan reserves. By reducing the incidence of non-performing loans on bank balance sheets, calendar provisioning may free up resources for fresh lending. Lower levels of NPLs may influence permanently the provision of credit by relaxing regulatory restrictions and decreasing funding 10 costs. Contrary, the increased regulatory costs imposed on European banks through more prudential and mechanical management of their NPL portfolio might have affected their ability and willingness to originate loans abroad. Thus, we investigate whether: H5: Both the supervisory guidance and the regulatory changes affect loan origination. Since calendar provisioning represents a costly burden for European banks, we analyze whether European banks transfer to customers the higher regulatory costs. We hypothesize that treated banks might be responding to the shock by charging higher rates on newly originated loans. H6: Both the supervisory guidance and the regulatory changes generated an increase in lending rates. A subset of studies exploits variation in global standards to identify the effect of regulation on a variety of banking outcomes focusing on the role of subsidiaries of multinational banks (Claessens et al., 2015; Berrospide et al., 2016). The presumption is that globalized banks while operating in multiple markets, adjust the allocation of financial resources according to regulatory circumstances (Chiuri et al., 2002). For example, Aiyar et al. (2014) show that changes in capital requirements imposed on UK-resident banks trigger a reduction in cross-border lending. Similarly, a recent stream of studies analyses how domestic macro-prudential regulation and monetary policy transmit to foreign countries via a change in the lending behavior of foreign subsidiaries (Morais et al., 2019). Thus, differences in domestic and host-country regulations correlate with lending standards (Ongena et al., 2013) and influence the association in risk-taking behavior of subsidiaries and their parent companies (Anginer et al., 2017). This leads us to the following hypothesis: H7: Banks engage in regulatory cost-shifting behaviors more aggressively in countries featuring more lenient banking regulatory environments. 3 Data and variables We gather information from different sources for the period from 2015 to 2019. Bank-level financial statements are obtained through Fitch Connect. Data on bank regulation and supervision is garnered through the Bank Regulation and Supervision Survey (Anginer et al., 2019). Since the goal of our study is to assess the effect of changes in bank regulation that may affect bank behavior in countries where European foreign subsidiaries operate, we restrict our sample to banks that 11 operate in developing countries where different standards and regulations apply. This allows us to use as a counterfactual the observed performance of banks in a country not subject to the regulatory change. We are also able to investigate whether specific features of a country’s regulatory environment interact with the reform implemented at the euro area level. We apply standard data cleaning procedures to control for the influence of outliers or inconsistent information. In detail, we drop information on banks for which loans over total assets are less than 10 percent and deposits over total assets are less than 5 percent. This filter is applied to ensure that the core business of banks in the sample is credit intermediation. Our final sample consists of 7,084 bank-year observations distributed across up to 91 countries (Table 1, panel A). Around 5 percent of the observations are from subsidiaries of EU banks. The number of banks by country and year is reported in Table A1 in Appendix A. As we aim to uncover new evidence on how subsidiaries of European banks react to the introduction of calendar provisioning, we employ different dependent variables in the estimations. The first group of variables aims at understanding the effect of the regulatory change on non- performing loans and whether this change is related to changes in reserves for impaired loans or net charge offs. 10 We compute three ratios: non-performing loans to total gross loans (NPL ratio); impaired loans reserves to total loans; and net charge-offs to gross loans. The second group of variables measures credit intermediation both in terms of volumes and prices. We consider three variables: the natural logarithm of gross loans; the growth in gross loans; and interest income to total assets. Overall, our analyses provide a comprehensive picture of both adjustments in credit risk exposure and spillovers to the real economy. Our main variable of interest is a dichotomous indicator taking the value of 1 if a bank is a subsidiary of a European bank (EU bank) operating in one of the countries in the sample, 0 otherwise. We observe up to 96 EU subsidiaries over the sample period operating across 42 countries. The total number of observations of EU bank subsidiaries is 405. 11 The economies with 10 Decreases in non-performing loans ratios may be achieved not only through write-offs, but also through other means. For example, increases in loan growth may dilute the relevance of non-performing exposures. Securitization and non- performing loans sales may be alternative mechanisms to reduce non-performing loan ratios. 11 The overall number of observations of bank subsidiaries in the sample is 576. 12 the largest number of observations of EU subsidiaries are China (41), the Russian Federation (26), Brazil (22), and Bosnia and Herzegovina (20). As control variables, we include a parsimonious set of bank-specific controls employed in previous studies investigating the determinants of non-performing loans (see, among many others, Berger and DeYoung, 1997; Louzis et al., 2012; Ghosh, 2015). The direction of causation is nevertheless not clear. Bank size may be related to excessive risk-taking as very large banks exploit their status of too-big-to-fail (Kaufman, 2014). Yet larger banks have in principle more diversified bank loans which lower credit risk (Louzis et al., 2012). We account for the role of size in explaining non-performing loans by including the logarithm of total assets. Profitability may be also associated with non-performing loans as poor performance could be related to a lack of skills in the screening, selection, and monitoring of the borrowers. Yet a too-high return on investment may also indicate moral hazard and greater exposure to credit risk. To control for bank performance, we compute bank returns on assets. Finally, we account for the role of bank capitalization by controlling for the level of bank capitalization measured as equity over total assets. Bank capital may work both as an incentive device to reduce the riskiness of the bank loan book (Furlong and Keeley, 1989) and increase risk-taking at higher levels of capital (Calem and Rob, 1999). As documented in Table 1, Panel B, subsidiaries of European banks have lower non- performing loans over the studied period. The difference in the mean values between European subsidiaries and other banks is around 1.2 percentage points, 12 which is substantial as this value is approximately 16 percent of the sample mean of non-EU subsidiaries. European banks operating in emerging markets are on average larger than their domestic counterparts and show higher overall profitability (i.e., operating ROA). However, their interest income to total assets is lower because of their relatively smaller loan portfolios. 12 This difference is statistically significant at the 1 percent level according to a two-sample t-test imposing unequal variance. 13 Table 1: Summary Statistics Table 1, Panel A reports the time-series distribution of the available observations. Panel B reports summary statistics for the 8,209 banks for which data is available over the period 2015-2019. Source: own elaborations using data from Fitch. Panel A: Time-Series Distribution Number of Number Number of Treated Year Countries of Banks Subsidiaries (EU) 2015 90 1,860 89 2016 88 1,867 85 2017 90 1,579 82 2018 87 1,486 78 2019 82 1,417 71 Panel B: Bank-level data Non-EU subsidiary EU subsidiary Variable N Mean Std. Dev. N Mean Std. Dev. NPL Ratio 7,804 7.116 9.089 405 5.965 7.110 Net Charge-Offs to Gross Loans 4,814 1.299 2.291 296 1.383 2.056 Reserve for Impaired Loans to Gross Loans 7,759 6.229 6.951 398 4.965 4.633 Log of Gross loans 7,804 6.389 2.219 405 7.347 1.612 Gross Loans growth 7,800 6.927 24.581 405 2.530 22.577 Total Regulatory Capital Ratio 5,886 21.022 15.950 312 19.491 6.448 Interest Income to Total Assets 7,585 6.764 4.236 397 4.555 3.560 Log of Total Assets 7,804 6.933 2.103 405 8.113 1.493 Operating ROA 7,804 1.590 2.467 405 2.074 2.082 Equity to Total Assets 7,804 14.705 10.318 405 13.450 6.076 Loans to Total Assets 7,804 57.105 17.414 405 49.073 19.174 Deposits to Total Assets 7,804 63.573 19.820 405 61.587 19.024 4 Identification Our empirical setting is based on comparing financial and economic outcomes between European subsidiaries and domestic banks operating in developing countries. In particular, we employ a difference-in-differences approach to isolate the causal effect of calendar provisioning on banks’ policies. As previously discussed, the supervisory and regulatory interventions were introduced in two subsequent stages. To capture these staggering effects, we include dichotomous variables for each year starting from 2016 to closely follow each phase's implementation in the introduction of calendar provisioning and control for any specific time trend affecting the studied dependent 14 variable. The inclusion of a dummy variable for 2016 enables us to verify if there are statistically significant differences in trends between banks in the treated and control groups before the first introduction of the calendar provisions in 2017. The lack of statistical significance for the 2016 dummy interacted with the European Subsidiary dummy (SUB EU) would indeed support the hypothesis that the two groups follow statistically indistinguishable trends over the two years preceding the regulatory change. The remaining year dummy variables enable us to capture the average treatment effects after the first issuance of the ECB guidelines (2017), the ECB’s addendum (2018), and the EU introduction of the prudential backstop regulation within the Pillar 1 framework (2019). These considerations result in the following specification: , = 1 , + 2 , × 2016 + 3 , × 2017 + 4 , × 2018 + (1) 5 , × 2019 + ,−1 + + + , Where corresponds to one of the dependent variables employed in the analysis: the NPL ratio; impaired loans reserves to total loans; net charge-offs to gross loans; the natural logarithm of gross loans; the growth in gross loans; and interest income to total assets. Sub is a dummy variable for European bank subsidiaries. is a vector of bank controls, namely the natural logarithm of total assets, Return on Assets (ROA), and equity over total assets. We also control for time- invariant bank-specific fixed effects ( ) and for bank-invariant time-specific fixed effects ( ). Standard errors are clustered at the individual bank level. To provide broader support to the validity of our findings we employ three alternative samples. First, since EU bank subsidiaries tend to be larger than other banks operating in a country (see Table 1, panel B), we restrict the sample to large banks only (i.e., above the median of the distribution of total assets in a country in a year). Second, as mentioned above, EU subsidiary operates in only 42 countries out of the 91 countries included in our sample. For this reason, we restrict the sample to countries and years where at least one EU subsidiary bank exists. Finally, we perform propensity score matching (PSM) to compare EU subsidiaries with a matched sample of banks based on characteristics observed before the introduction of the calendar provisioning. 15 5 Results We begin our analysis by examining whether the introduction of the European calendar provisioning caused the subsidiaries of European banks to reduce their NPL ratios, as compared to that of their untreated peers (H1). As documented in Table 2, we find this to be the case. European banks reacted immediately to the release of the initial guidance published by the ECB in March 2017 by shrinking their NPL ratios. In particular, the 2017-year dummy interacted with the European Subsidiary dummy (SUB EU) included in our specifications captures an immediate relative decline in European banks’ NPL ratios ranging between 22% (Column 1) and 44% (Column 5), confirming that banks reacted immediately to the first release of regulatory guidance. As one would have expected, the estimated effect further increases for the calendar year 2018, when the ECB formalized its supervisory expectations. In the European banks’ eyes, the ECB’s Addendum provided new criteria that generated an additional and greater decline in the NPL ratio. In particular, our estimates suggest that subsidiaries of European banks operating in developing countries decreased their NPL ratios in 2018 by a further 10%-15%. Overall, the supervisory guidance issued by the ECB was effective in reducing NPL ratios. In this sense, these two non-binding initiatives appear to have acted as material shocks for European banks. These results are robust across specifications. In detail, consistent estimates are identified using either (1) the whole sample of available observations (Table 1, Column (1) – without controls – and Column (2) – with controls), (2) the sample including large commercial banks, exclusively (Table 1, Column 3), (3) the sample including only those countries in which at least one European bank’s subsidiary is operating (Table 1, Column 4), and (5) the propensity scored matched sample (Table 1, Column 5). Notably, in all models the 2016-year dummy interaction with the European Subsidiary dummy (SUB EU) is not statistically significant, supporting the argument that the two groups are statistically indistinguishable before the first supervisory change in 2017. Furthermore, the estimated economic effect increases as the matching requirements tighten, indicating that eventual endogeneity concerns are likely biasing the resulting coefficients towards 0. Next, we focus on the European Commission’s regulatory initiative institutionalizing calendar provisioning within the existent regulatory framework. Our estimates highlight an additional drop in the NPL ratio in 2019 when the European Commission introduced the calendar provision within the Pillar 1 regulation. The coefficient estimates for the 2019-year dummy interacted with the EU 16 subsidiary indicator are indeed negative and statistically significant, confirming that lasting material effect induced by the adoption of the European calendar provisioning, providing further support for the validity of H1. That is, we document that the calendar provision policy achieved the intended direct goal of reducing European banks' exposure to NPLs, despite its non-model- based nature. Table 2: Calendar provisioning and non-performing loans ratio of bank subsidiaries This table reports the results of the estimates of equation (1). The dependent variable is the NPL ratio. In Column (3) we restrict the sample to banks above the median total assets in a year in a country. In Column (4) we consider instead countries wherein each year at least one foreign-owned subsidiary operates. In Column (5) we employ propensity score matching to select a control group with similar characteristics to the treatment group pre-introduction of the calendar provisioning. Robust standard errors clustered at the bank level appear in parentheses. Subsidiary EU x 2018 indicates the p-value for an F-test for the equality between the two coefficients. ***p < 0.01, **p < 0.05, *p < 0.1. (1) (2) (3) (4) (5) Subsidiary EU 1.081** 0.696 0.326 0.981 1.977 (0.441) (0.635) (0.740) (0.767) (4.744) Subsidiary EU x 2016 -0.200 -0.360 -0.305 -0.322 -0.699 (0.359) (0.356) (0.363) (0.361) (0.799) Subsidiary EU x 2017 -1.319* -1.501** -1.511** -1.660** -2.625*** (0.709) (0.711) (0.716) (0.727) (0.829) Subsidiary EU x 2018 -1.660** -1.872*** -1.861*** -2.239*** -2.780*** (0.685) (0.665) (0.680) (0.683) (0.841) Subsidiary EU x 2019 -2.306*** -2.472*** -2.251*** -2.932*** -3.136*** (0.756) (0.753) (0.773) (0.778) (0.865) Natural logarithm of -0.895** -0.702 -1.351** -0.673 total assets (0.433) (0.528) (0.639) (0.415) Return on assets -0.429*** -0.572*** -0.560*** -0.472*** (0.100) (0.157) (0.120) (0.055) Equity over total assets 0.010 0.072 0.015 0.058** (0.039) (0.072) (0.039) (0.024) Constant 7.011*** 13.947*** 12.486*** 17.383*** 11.350*** (0.009) (3.253) (4.561) (4.906) (3.208) Subsidiary EU x 2018 = 0.0675 0.1065 0.3254 0.3736 0.6840 Subsidiary EU x 2019 ? Calendar Year FE Yes Yes Yes Yes Yes Bank FE Yes Yes Yes Yes Yes Only Countries Only Large PSM Sample Full Full with a Banks Sample Subsidiary Observations 7,644 7,644 4,540 5,981 4,370 R-squared 0.774 0.779 0.790 0.789 0.792 17 Was the marginal effect of the European Commission’s effort to institutionalize the ECB supervisory guidance within a regulatory framework economically and statistically material? Or did banks treat the initial non-binding supervisory guidance as if it was a regulatory intervention? To answer this question, we run a set of F-tests to assess whether the coefficient for the 2018 dummy interacted with the European Subsidiary dummy (SUB EU) is statistically indistinguishable from the coefficient for the 2019 dummy interacted with the European Subsidiary dummy (SUB EU). Finding these two coefficients are not statistically different would indeed imply that the regulatory action through which the ECB supervisory guidance introduced the European calendar provisioning had no material effects. We find this to be the case. As reported the Table 2, these two coefficients are only significant in Column 1, with the p-value increasing as the quality of the matched samples increases. That is, we cannot reject H2, thus leading us to the conclusion that banks react to the release of supervisory guidances issues under the “comply or explain” framework as if it is a regulatory action while providing banks and supervisors with some residual flexibility concerning the degree to which these initiatives should be enforced over time and in the cross-section. To assess the robustness of our conclusions, we recognize that the identified reduction in NPL ratios should necessarily pair with a reduction in banks’ equity reserves for credit losses. In theory, this is indeed the most immediate and almost mechanical reaction to the ECB guidelines: banks use the equity reserves made in the past for credit losses to comply with ECB’s expectations. To this aim, we focus on the ratio of Impaired Loan Reserves to Total Loans around the adoption of the ECB calendar provision discipline to evaluate if this change in European regulation affected banks’ ability to absorb credit risk shocks. As documented in Table 3, our estimates suggest that the impaired loan reserves relative to total loans dropped significantly in 2017, 2018, and 2019. Once again, we show that banks reacted immediately to the disclosure of the ECB guidelines. Impaired Loan Reserves to Total assets further decline following the release of the Addendum in 2018 and the European Union act in 2019, but the bulk of the effect can be attributed to the 2017 supervisory initiative, confirming banks’ perception of supervisory guidelines as virtually binding requirements. Indeed, the F-tests testing for the equality of the coefficient for the 2018 dummy interacted with the European Subsidiary dummy (SUB EU) and that for the 2019 dummy interacted with the European Subsidiary dummy (SUB EU) produce consistently high p-values, ultimately 18 confirming that European banks are responding to supervisory guidance as if they were regulatory shocks. These findings also corroborate our third research hypothesis (H3). The European calendar provisioning induced an immediate deterioration of banks’ reserves for impaired loans, which declined by about 15% (Column 1) to 30% (Column 5). Again, the average treatment effects increase as the matching conditions tighten, indicating that eventual biases are pushing our estimates towards the zero bound, rather than in our favor. All in all, this finding – which appears to be robust to the use of the 5 proposed specifications – provides support for the materiality of calendar provisioning, further suggesting that while the passage of this new supervisory and regulatory discipline contributed to deflating European banks’ NPL ratios, it also caused a persistent weakening of their capitalization profile, thus limiting their ability to copying with sudden, large credit risk shock. Finally, these results confirm the primary role played by the initial supervisory guidance issued by the ECB, indicating that banks are indeed treating these non-binding initiatives as if they were legally binding regulatory actions. Next, we focus on the mechanism through which European banks immediately responded to the passage of this policy by reducing the level of their NPL ratios. In particular, we evaluate whether banks wrote off aggressively all those NPLs that were potentially in conflict with the new calendar provisioning guidelines. By recognizing losses in their income statements, banks would indeed be able to rapidly shrink their NPLs, ultimately reaching compliance with the new regulatory requirements. As documented in Table 4, we identify support for this possibility. Consistent with the estimates reported in Table 2 and Table 3, we identify abnormally high levels of write-offs recognized by the subsidiaries of European banks. 13 In particular, this effect is concentrated in the year 2017, while it dissipates afterward – a finding that is again robust to all the five proposed specifications. This important result supports our fourth research hypothesis (H4), further confirming that banks complied with the initial supervisory guidelines by exploiting the most direct tool at their disposal to comply with it – that is, recognizing losses in their income statements. Furthermore, this result may suggest that once the ECB explicitly stated its expectations in 2018 and the European Union introduced the new regulation as a Pillar 1 13 This result is marginally insignificant in Column 5, possibly due to a reduction in power caused by the smaller sample size characterizing the matched sample. 19 requirement, banks decreased NPLs using other more expensive tools, such as loan sales or securitization. 14 Table 3: Calendar provisioning and impaired loans reserves of bank subsidiaries This table reports the results of the estimates of equation (1). The dependent variable is the ratio between Impaired Loans Reserves and Total Loans. In Column (3) we restrict the sample to banks above the median total assets in a year in a country. In Column (4) we consider instead countries wherein each year at least one foreign-owned subsidiary operates. In Column (5) we employ propensity score matching to select a control group with similar characteristics to the treatment group pre-introduction of the calendar provisioning. Robust standard errors clustered at the bank level appear in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. (1) (2) (3) (4) (5) Subsidiary EU 1.682*** 1.273*** 1.188*** 1.404*** 1.406*** (0.526) (0.341) (0.432) (0.432) (0.416) Subsidiary (EU) x 2016 -0.130 -0.297 -0.114 -0.434 -0.536 (0.394) (0.389) (0.402) (0.399) (0.507) Subsidiary (EU) x 2017 -0.905** -1.077*** -0.984*** -1.207*** -1.515*** (0.362) (0.372) (0.370) (0.413) (0.481) Subsidiary (EU) x 2018 -1.361*** -1.570*** -1.493*** -1.823*** -1.691*** (0.428) (0.423) (0.433) (0.455) (0.508) Subsidiary (EU) x 2019 -1.626*** -1.785*** -1.662*** -2.046*** -1.872*** (0.505) (0.503) (0.530) (0.540) (0.595) Logarithm of total assets -0.697* -0.402 -1.492** -1.350** (0.411) (0.539) (0.616) (0.635) Return on assets -0.453*** -0.439*** -0.533*** -0.586*** (0.099) (0.153) (0.133) (0.125) Equity over total assets 0.026 0.067 0.016 0.051 (0.035) (0.084) (0.044) (0.053) Constant 6.178*** 11.512*** 9.013* 17.825*** 16.578*** (0.023) (2.976) (4.716) (4.674) (4.899) Subsidiary EU x 2018 = 0.3565 0.4593 0.5685 0.4594 0.6026 Subsidiary EU x 2019 ? Calendar Year FE Yes Yes Yes Yes Yes Bank FE Yes Yes Yes Yes Yes Only Large Only Countries Sample Full Full PSM Sample Banks with a Subsidiary Observations 7,644 7,644 4,540 5,981 4,370 R-squared 0.824 0.830 0.805 0.822 0.821 14 We are unable to directly test for this possibility as it is difficult to gather information systematically on these off- balance sheet transactions. 20 Table 4: Calendar provisioning and write-offs of bank subsidiaries This table reports the results of the estimates of equation (1). The dependent variable is the ratio between Net Charge- Off and Gross Loans. In Column (3) we restrict the sample to banks above the median total assets in a year in a country. In Column (4) we consider instead countries wherein each year at least one foreign-owned subsidiary operates. In Column (5) we employ propensity score matching to select a control group with similar characteristics to the treatment group pre-introduction of the calendar provisioning. Subsidiary EU x 2018 indicates the p-value for an F-test for the equality between the two coefficients.Robust standard errors clustered at the bank level appear in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. (1) (2) (3) (4) (5) Subsidiary EU -0.165 0.022 -0.034 0.036 0.168 (0.203) (0.178) (0.211) (0.185) (0.254) Subsidiary (EU) x 2016 0.416* 0.227 0.172 0.246 0.016 (0.217) (0.208) (0.211) (0.210) (0.172) Subsidiary (EU) x 2017 0.854** 0.620* 0.654* 0.574* 0.235 (0.382) (0.332) (0.382) (0.333) (0.379) Subsidiary (EU) x 2018 0.309 0.117 0.116 0.085 0.145 (0.369) (0.314) (0.378) (0.318) (0.444) Subsidiary (EU) x 2019 0.223 0.118 0.150 0.107 -0.085 (0.310) (0.294) (0.307) (0.299) (0.324) Logarithm of total assets -0.146 -0.115 -0.230 -0.267 (0.160) (0.249) (0.203) (0.345) Return on assets -0.182*** -0.302*** -0.159*** -0.272*** (0.038) (0.061) (0.043) (0.053) Equity over total assets 0.013 0.006 0.004 0.025 (0.012) (0.027) (0.014) (0.025) Constant 1.344*** 2.520** 2.699 3.275* 3.594 (0.005) (1.261) (2.171) (1.671) (2.752) Calendar Year FE Yes Yes Yes Yes Yes Bank FE Yes Yes Yes Yes Yes Only Large Only Countries PSM Sample Full Full Banks with a Subsidiary Sample Observations 4,967 4,967 3,248 3,995 2,827 R-squared 0.616 0.607 0.603 0.639 0.659 The simultaneous deterioration of European banks’ risk profile and NPLs portfolio size, paired with the temporary increase in write-offs, leads us to a second, crucial consideration: did European banks significantly change their loan origination policies abroad to adjust their credit risk profile to the new regulatory environment? Indeed, the increased regulatory costs imposed on European banks through more prudential and mechanical management of their NPL portfolio might have affected their ability and willingness to originate loans abroad. To test this hypothesis, 21 we study the time-series dynamics of gross loans (Table 5) around the studied supervisory and regulatory change. 15 We identified no evidence that supports the hypothesis that these higher regulatory costs caused a reduction in loan origination by subsidiaries of European banks. Table 5: Calendar provisioning and gross loans This table reports the results of the estimates of equation (1). The dependent variable is the Natural logarithm of gross loans. In Column (3) we restrict the sample to banks above the median total assets in a year in a country. In Column (4) we consider instead countries wherein each year at least one foreign-owned subsidiary operates. In Column (5) we employ propensity score matching to select a control group with similar characteristics to the treatment group pre- introduction of the calendar provisioning. Robust standard errors clustered at the bank level appear in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. (1) (2) (3) (4) (5) Subsidiary EU 0.028 -0.035 -0.051** -0.006 0.001 (0.039) (0.025) (0.025) (0.026) (0.024) Subsidiary (EU) x 2016 -0.140 -0.044 -0.045 -0.047 -0.086 (0.026) (0.024) (0.024) (0.024) (0.069) Subsidiary (EU) x 2017 -0.102*** -0.008 -0.007 -0.044 -0.022 (0.039) (0.032) (0.033) (0.033) (0.039) Subsidiary (EU) x 2018 -0.095* 0.000 0.003 -0.021 -0.021 (0.051) (0.039) (0.040) (0.039) (0.041) Subsidiary (EU) x 2019 -0.106* -0.013 -0.004 -0.039 -0.027 (0.057) (0.049) (0.050) (0.049) (0.041) Natural logarithm of total assets 0.655*** 0.637*** 0.587*** 0.5548*** (0.047) (0.063) (0.031) (0.037) Return on assets 0.005 -0.000 0.009 0.013* (0.006) (0.005) (0.007) (0.007) Equity over total assets -0.001 0.000 -0.002 -0.001 (0.002) (0.004) (0.003) (0.003) Constant 6.521*** 1.888*** 2.364*** 2.466*** 2.641*** (0.001) (0.350) (0.526) (0.239) (0.293) Calendar Year FE Yes Yes Yes Yes Yes Bank FE Yes Yes Yes Yes Yes Only Countries PSM Sample Full Full Only Large Banks with a Subsidiary Sample Observations 7,644 7,644 4,540 5,981 4,370 R-squared 0.987 0.991 0.991 0.993 0.9937 15 As robustness test, we report in Table B1 in Appendix the results of the regressions with changes in gross loans. Results remain qualitatively the same. 22 How do European banks cope with the increase in these regulatory costs then? We hypothesize that treated banks might be responding to the shock by offsetting the increased compliance costs onto their customers by charging higher rates on newly originated loans. Building on the previously discussed results, such behavior should materialize in an increase in their interest income. As documented in Table 6, we find this to be the case. Subsidiaries of European banks appear to systematically shift these increased regulatory costs to their customers by charging higher interest rate spreads on new loans. Table 6: Calendar Provisioning and Interest Income This table reports the results of the estimates of equation (1). The dependent variable is the ratio between Interest Income and Total Assets. In Column (3) we restrict the sample to banks above the median total assets in a year in a country. In Column (4) we consider instead countries where in each year at least one foreign-owned subsidiary operates. In Column (5) we employ propensity score matching to select a control group with similar characteristics to the treatment group pre-introduction of the calendar provisioning. Robust standard errors clustered at the bank level appear in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. (1) (2) (3) (4) (5) VARIABLES Interest Income to Total Assets Subsidiary EU 0.192 0.010 0.289 0.057 -1.987*** (0.880) (0.912) (0.896) (1.374) (0.159) Subsidiary (EU) x 2016 -0.016 0.084 -0.016 0.097 0.089 (0.109) (0.116) (0.115) (0.121) (0.116) Subsidiary (EU) x 2017 0.125 0.294 0.161 0.308 0.307 (0.170) (0.184) (0.183) (0.190) (0.210) Subsidiary (EU) x 2018 0.300 0.517** 0.252 0.568** 0.641*** (0.191) (0.215) (0.212) (0.223) (0.228) Subsidiary (EU) x 2019 0.337* 0.568*** 0.261 0.580*** 0.647** (0.176) (0.199) (0.198) (0.208) (0.260) Natural logarithm of total assets 0.877*** 0.856*** 0.924*** 1.183*** (0.122) (0.131) (0.175) (0.237) Return on assets 0.034** -0.003 0.041** 0.026 (0.016) (0.021) (0.019) (0.023) Equity over total assets -0.041*** -0.023* -0.042*** -0.034*** (0.009) (0.012) (0.010) (0.012) Constant 6.559*** 1.158 -0.651 0.703 -1.148 (0.041) (0.865) (1.093) (1.273) (1.726) Calendar Year FE Yes Yes Yes Yes Yes Bank FE Yes Yes Yes Yes Yes Only Countries PSM Sample Full Full Only Large Banks with a Subsidiary Sample Observations 7,644 7,644 4,540 5,981 4,370 R-squared 0.910 0.915 0.920 0.916 0.919 23 Table 7: Calendar provisioning and bank regulation This table reports the results of the estimates of equation (1). The dependent variable is the ratio between Interest Income and Total Assets. Bank controls are size (the natural logarithm of total assets), profitability (return on assets), and capitalization (equity over total assets). Robust standard errors clustered at the bank level appear in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. (1) (2) (3) (4) w = Overall Restrictions on Banking w = Stringency in the definition of Activities NPLs Subsidiary EU -0.002 -0.002 0.015 0.021* (0.007) (0.008) (0.011) (0.012) Subsidiary (EU) x 2016 0.006 0.005 -0.009 -0.011 (0.006) (0.006) (0.008) (0.009) Subsidiary (EU) x 2017 0.011 0.012 -0.009 -0.006 (0.009) (0.010) (0.011) (0.011) Subsidiary (EU) x 2018 0.018** 0.021** -0.016* -0.017* (0.009) (0.010) (0.008) (0.009) Subsidiary (EU) x 2019 0.020** 0.026*** -0.001 -0.011 (0.009) (0.010) (0.012) (0.011) W -0.001*** -0.001*** 0.001* 0.001* (0.000) (0.000) (0.000) (0.000) Subsidiary (EU) x 2016 x -0.001 -0.001 0.002 0.002 w (0.001) (0.001) (0.001) (0.002) Subsidiary (EU) x 2017 x -0.001 -0.001 0.002 0.002 w (0.001) (0.001) (0.002) (0.002) Subsidiary (EU) x 2018 x -0.002** -0.002* 0.004** 0.004** w (0.001) (0.001) (0.002) (0.002) Subsidiary (EU) x 2019 x -0.002* -0.003** 0.004* 0.003* w (0.001) (0.001) (0.002) (0.002) Constant 0.084*** 0.038* 0.070*** 0.020 (0.003) (0.022) (0.002) (0.021) Additional Controls No Yes No Yes Calendar Year FE Yes Yes Yes Yes Bank FE Yes Yes Yes Yes Sample Full Full Full Full Observations 8,850 8,476 8,886 8,509 R-squared 0.922 0.909 0.922 0.908 24 How can European banks exercise this form of market power while keeping a constant level of new loan originations like their untreated peers? We hypothesize that subsidiaries of European banks might exploit the regulatory leniency characterizing certain emerging markets to offload their domestic compliance costs to foreign consumers. If this is the case, we should observe that the increased interest income documented in Table 6 is concentrated in banking systems featuring weaker regulatory environments. We find this to be the case. Using data from the World Bank’s Bank Regulation and Supervision Survey, we identify countries featuring stricter overall restrictions on banking activities and providing stricter regulatory decisions for NPLs, we document that this regulatory cost-shifting mechanism is significantly stronger in countries with a weaker regulatory environment (Table 7). This result is consistent with multinational banks exploiting the regulatory leniency of foreign countries to off-load – at least part of – the regulatory costs imposed by domestic regulations. 6 Conclusion The European calendar provisioning represents not only a major regulatory and supervisory change, but also a turning point in banking regulation. Its underpinning idea is disruptive to the logic of the regulatory framework established by the Basel Committee on Banking Supervision. Rather than relying on banks’ internal risk-based models, the European calendar provisioning imposes an old-fashioned regulation that hinges on minimum loss coverage requirements based on simple rules depending only on the time elapsed since the default. Furthermore, the process of adoption and enforcement of this regulation is unique: originally introduced in the form of supervisory guidance in 2017 by the ECB, it became stricter in 2018, and it turned into a new Pillar 1 regulation in 2019. The regulatory process followed for the introduction of this new regulation enables us to assess for the first time how binding supervisory guidances issued under the comply or explain framework are as compared to the introduction of a similar regulatory intervention. Our empirical strategy relies on a Difference-in-Differences (DID) approach allowing us to compare changes in the riskiness, performance, and loan policies of subsidiaries of European banks operating in developing countries with that of matched domestic banks. Because European banks consolidate worldwide credit exposures under their domestic regulatory framework, their subsidiaries 25 operating in developing countries are directly affected by the adoption of calendar provisioning rules. Conversely, matched banks operating in developing countries are not exposed to the effects of these supervisory and regulatory actions. Our empirical analyses identify two novel and unique contributions to the ongoing academic and regulatory debate on banking supervision and regulation. First, non-model-based supervisory and regulatory shocks can achieve their intended goal - reducing European banks’ NPL ratios in this case. Second, supervisory guidance issued under the “comply or explain” framework is perceived by banks as binding while providing supervisors with some flexibility concerning the degree to which such actions should be enforced over time and in the cross-section. Such flexibility is valuable in a rapidly evolving industry. Regulatory and supervisory agencies should exercise coordinated efforts to minimize redundancy and to avoid inducing over-regulation costs which may ultimately result in negative social consequences. Finally, changes in regulatory and supervisory practices in advanced economies transmit to developing countries and should be monitored carefully by domestic financial authorities to ensure effective regulation and supervision. 26 References Aiyar, S., Calomiris, C.W., Hooley, J., Korniyenko, Y., Wieladek, T., 2014. The international transmission of bank capital requirements: Evidence from the UK. J. financ. econ. 113, 368–382. https://doi.org/10.1016/j.jfineco.2014.05.003 Anginer, D., Bertay, A.C., Cull, R., Demirgüç-Kunt, A., Mare, D.S., 2019. Bank regulation and supervision ten years after the Global Financial Crisis. Policy Research Working Papers (No. WPS9044). The World Bank, Washington, DC, USA. Anginer, D., Cerutti, E., Martínez Pería, M.S., 2017. Foreign bank subsidiaries’ default risk during the global crisis: What factors help insulate affiliates from their parents? J. Financ. Intermediation 29, 19–31. https://doi.org/10.1016/j.jfi.2016.05.004 Basel Committee on Banking Supervision, 2011. Basel III: A global regulatory framework for more resilient banks and banking systems. Bank for International Settlements, Basel, Switzerland. Basel Committee on Banking Supervision, 2006. International Convergence of Capital Measurement and Capital Standards. A Revised Framework. Bank for International Settlements, Basel, Switzerland. Bassett, W.F., Lee, S.J., Spiller, T.P., 2015. Estimating changes in supervisory standards and their economic effects. J. Bank. Financ. 60, 21–43. https://doi.org/10.1016/j.jbankfin.2015.07.010 Bassett, W.F., Marsh, W.B., 2017. Assessing targeted macroprudential financial regulation: The case of the 2006 commercial real estate guidance for banks. J. Financ. Stab. 30, 209–228. https://doi.org/10.1016/j.jfs.2016.06.001 Berger, A.N., Davies, S.M., 1998. The information content of bank examinations. J. Financ. Serv. Res. 14, 117–144. https://doi.org/https://doi.org/10.1023/A:1008011312729 Berger, A.N., Davies, S.M., Flannery, M.J., 2000. Comparing Market and Supervisory Assessments of Bank Performance: Who Knows What When? J. Money, Credit Bank. 32, 641– 667. https://doi.org/10.2307/2601200 Berger, A.N., DeYoung, R., 1997. Problem loans and cost efficiency in commercial banks. J. Bank. Financ. 21, 849–870. https://doi.org/10.1016/S0378-4266(97)00003-4 27 Berrospide, J., Correa, R., Goldberg, L., Niepmann, F., 2016. International Banking and Cross-border Effects of Regulation: Lessons from the United States. NBER Working Paper (No. 22645). National Bureau of Economic Research, Cambridge, MA, USA. https://doi.org/10.3386/w22645 Bridges, J., Gregory, D., Nielsen, M., Pezzini, S., Radia, A., Spaltro, M., 2014. The impact of capital requirements on bank lending. Bank of England Working Paper (No. 486), Bank of England Working Paper. London, UK. https://doi.org/10.2139/ssrn.2388773 Calem, P., Rob, R., 1999. The Impact of Capital-Based Regulation on Bank Risk-Taking. J. Financ. Intermediation 8, 317–352. https://doi.org/10.1006/jfin.1999.0276 Chiuri, M.C., Ferri, G., Majnoni, G., 2002. The macroeconomic impact of bank capital requirements in emerging economies: Past evidence to assess the future. J. Bank. Financ. 26, 881– 904. https://doi.org/10.1016/S0378-4266(01)00267-9 Claessens, S., Van Horen, N., Berger, A., Buch, C., Cerutti, E., De Haas, R., Gobat, J., Laeven, L., Lambert, F., Saldias, M., 2015. The Impact of the Global Financial Crisis on Banking Globalization 63, 868–918. https://doi.org/10.1057/imfer.2015.38 Delis, M.D., Staikouras, P.K., 2011. Supervisory effectiveness and bank risk. Rev. Financ. 15, 511–543. https://doi.org/10.1093/rof/rfq035 Feyen, E., Alonso Gispert, T., Kliatskova, T., Mare, D.S., 2021. Financial sector policy response to COVID-19 in Emerging Markets and Developing Economies. J. Bank. Financ. 133. https://doi.org/10.1016/J.JBANKFIN.2021.106184 Furlong, F.T., Keeley, M.C., 1989. Capital regulation and bank risk-taking: A note. J. Bank. Financ. 13, 883–891. https://doi.org/10.1016/0378-4266(89)90008-3 Ghosh, A., 2015. Banking-industry specific and regional economic determinants of non- performing loans: Evidence from US states. J. Financ. Stab. 20, 93–104. https://doi.org/10.1016/j.jfs.2015.08.004 Hirtle, B., Kovner, A., Plosser, M., 2020. The Impact of Supervision on Bank Performance. J. Finance jofi.12964. https://doi.org/10.1111/jofi.12964 Hirtle, B.J., 2020. Banking supervision: The perspective from economics. Federal Reserve Bank of New York, New York, NY, USA. 28 Hirtle, B.J., Lopez, J.A., 1999. Supervisory information and the frequency of bank examinations. FRBNY Econ. Policy Rev. 5. Federal Reserve Bank of New York, New York, NY, USA. Kaufman, G.G., 2014. Too big to fail in banking: What does it mean? J. Financ. Stab. 13, 214–223. https://doi.org/10.1016/j.jfs.2014.02.004 Keay, A., 2014. Comply or explain in corporate governance codes: in need of greater regulatory oversight? Leg. Stud. 34, 279–304. https://doi.org/10.1111/lest.12014 Kiser, E.K., Prager, R.A., Scott, J.R., 2016. Supervisory ratings and bank lending to small businesses during the financial crisis and great recession. J. Financ. Serv. Res. 50, 163–186. https://doi.org/10.1007/s10693-015-0226-x Lane, P.R., 2012. The European sovereign debt crisis. J. Econ. Perspect. 26, 49–68. https://doi.org/10.1257/jep.26.3.49 Louzis, D.P., Vouldis, A.T., Metaxas, V.L., 2012. Macroeconomic and bank-specific determinants of non-performing loans in Greece: A comparative study of mortgage, business and consumer loan portfolios. J. Bank. Financ. 36, 1012–1027. https://doi.org/10.1016/j.jbankfin.2011.10.012 Morais, B., Peydro, J.-L., Roldan-Pena, J., Ruiz-Ortega, C., 2019. The international bank lending channel of monetary policy rates and QE: Credit supply, reach‐for‐yield, and real effects. J. Finance 74, 55–90. https://doi.org/10.1111/jofi.12735 Morgan, D.P., Rime, B., Strahan, P.E., 2004. Bank integration and State business cycles. Q. J. Econ. 119, 1555–1584. https://doi.org/10.1162/0033553042476161 Ongena, S., Popov, A., Udell, G.F., 2013. “When the cat’s away the mice will play”: Does regulation at home affect bank risk-taking abroad? J. financ. econ. 108, 727–750. https://doi.org/10.1016/j.jfineco.2013.01.006 Pugachev, L., 2019. The risk-shifting value of payout: Evidence from bank enforcement actions. J. Bank. Financ. 105595. https://doi.org/10.1016/j.jbankfin.2019.07.015 The World Bank, 2022. World Development Report 2022. Finance for an Equitable Recovery. The World Bank, Washington, DC, USA. 29 Appendix A: Descriptive statistics Table A1: Number of banks by country and year Table A1 reports information on the number of banks by country and year included in our sample. Economy 2015 2016 2017 2018 2019 Total Afghanistan 0 0 0 2 2 4 Albania 13 14 13 11 9 60 Algeria 4 6 5 3 3 21 Angola 12 9 11 13 10 55 Argentina 40 38 38 32 42 190 Armenia 15 12 13 13 12 65 Azerbaijan 21 14 16 14 13 78 Bangladesh 40 43 44 44 43 214 Belarus 19 21 20 21 20 101 Benin 0 0 2 2 2 6 Bhutan 2 3 3 3 2 13 Bolivia 7 6 8 9 9 39 Bosnia and Herzegovina 21 20 19 20 20 100 Botswana 7 5 6 4 5 27 Brazil 66 65 72 79 75 357 Bulgaria 17 16 16 16 14 79 Burkina Faso 3 3 2 0 2 10 Burundi 0 2 2 2 0 6 Cabo Verde 5 5 5 4 3 22 Cambodia 29 29 15 15 11 99 Cameroon 7 6 4 2 0 19 Chad 0 2 0 0 0 2 China 124 127 124 119 112 606 Colombia 38 35 38 39 39 189 Congo, Dem. Rep. 4 4 3 0 2 13 Costa Rica 18 18 18 16 16 86 Djibouti 0 0 0 2 0 2 Dominican Republic 15 37 40 36 33 161 Ecuador 18 18 19 18 18 91 Egypt, Arab Rep. 23 23 22 21 21 110 El Salvador 10 7 8 8 9 42 Eswatini 4 4 4 2 0 14 Ethiopia 5 9 11 12 11 48 Gabon 3 0 2 0 0 5 Gambia, The 2 0 2 2 0 6 Georgia 16 14 13 12 14 69 Ghana 20 18 15 15 15 83 Guatemala 17 20 20 21 20 98 Guinea 0 2 2 2 3 9 Guyana 6 6 6 6 6 30 Haiti 5 5 5 5 4 24 Honduras 17 15 16 22 22 92 India 69 74 80 73 54 350 30 Economy 2015 2016 2017 2018 2019 Total Indonesia 86 99 100 99 93 477 Jamaica 3 4 5 5 5 22 Jordan 11 11 10 10 10 52 Kazakhstan 20 19 19 19 19 96 Kenya 33 35 31 33 33 165 Kosovo 2 2 2 2 2 10 Kyrgyz Republic 3 0 0 0 0 3 Lao PDR 5 6 5 5 4 25 Lebanon 26 26 27 23 11 113 Lesotho 3 2 3 0 0 8 Liberia 2 2 3 3 3 13 Macedonia, FYR 8 7 7 8 7 37 Madagascar 3 3 2 0 0 8 Malawi 6 5 5 4 4 24 Malaysia 24 25 24 24 23 120 Mali 3 3 3 2 3 14 Mauritania 3 3 3 2 2 13 Mexico 51 56 51 48 50 256 Moldova 9 9 9 10 8 45 Mongolia 3 4 4 4 4 19 Montenegro 9 9 9 8 10 45 Morocco 7 7 7 7 8 36 Mozambique 8 6 7 8 7 36 Namibia 4 4 4 3 3 18 Nepal 42 41 21 33 35 172 Nicaragua 4 4 5 4 4 21 Nigeria 18 17 15 15 14 79 Pakistan 24 25 25 23 24 121 Papua New Guinea 0 2 0 0 0 2 Paraguay 14 14 14 14 12 68 Peru 15 16 15 14 15 75 Philippines 18 19 19 19 18 93 Russian Federation 381 363 97 59 68 968 Rwanda 5 7 7 8 8 35 Senegal 6 4 3 2 0 15 Serbia 17 18 21 20 21 97 Sierra Leone 4 0 0 0 0 4 South Africa 13 14 13 16 12 68 Sri Lanka 16 17 18 22 21 94 Syrian Arab Republic 10 10 10 11 12 53 Tajikistan 2 0 0 0 0 2 Tanzania 23 27 28 20 17 115 Thailand 19 22 20 21 21 103 Togo 2 3 3 2 0 10 Tunisia 10 11 12 12 10 55 Turkey 28 28 26 26 25 133 Uganda 17 16 16 9 7 65 Ukraine 23 18 21 22 23 107 31 Economy 2015 2016 2017 2018 2019 Total Uzbekistan 9 6 7 7 9 38 Venezuela, RB 22 21 22 5 3 73 Vietnam 21 21 21 19 19 101 West Bank and Gaza 2 0 0 2 2 6 Yemen, Rep. 0 0 2 0 0 2 Zambia 10 11 9 7 7 44 Zimbabwe 11 10 7 7 5 40 Total 1860 1867 1579 1486 1417 8209 32 Table A2: Number of countries by income level Table A2 reports information on the number of countries by income level and year included in our sample. Lower-middle Upper-middle Year Low income Total income income 2015 16 37 37 90 2016 16 36 36 88 2017 17 36 37 90 2018 14 37 36 87 2019 13 33 36 82 Total 76 179 182 437 33 Appendix B: Robustness checks Table B1: Calendar provisioning and growth in gross loans This table reports the results of the estimates of equation (1). The dependent variable is the percentage change in gross loans. In Column (3) we restrict the sample to banks above the median total assets in a year in a country. In Column (4) we consider instead countries where in each year at least one foreign-owned subsidiary operates. In Column (5) we employ propensity score matching to select a control group with similar characteristics to the treatment group pre- introduction of the calendar provisioning. Robust standard errors clustered at the bank level appear in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. (1) (2) (3) (4) (5) Subsidiary EU -26.297* -22.956 -22.764 -19.978 2.144 (14.091) (14.769) (14.591) (15.037) (2.158) Subsidiary (EU) x 2016 -3.845 -3.679 -3.741 -3.112 -2.615 (2.466) (2.445) (2.504) (2.488) (2.515) Subsidiary (EU) x 2017 7.495** 3.643 4.979 0.487 2.069 (3.498) (3.568) (3.603) (3.619) (4.390) Subsidiary (EU) x 2018 6.475** 2.550 3.796 1.128 0.049 (3.008) (3.111) (3.160) (3.205) (3.225) Subsidiary (EU) x 2019 5.053* 1.219 1.380 -1.854 -2.297 (2.864) (2.562) (2.652) (2.574) (2.923) Natural logarithm of total assets -26.123*** -26.683*** -34.459*** -35.040*** (2.896) (3.955) (2.155) (2.529) Return on assets 0.348 0.690* 0.566* 0.851** (0.284) (0.383) (0.339) (0.403) Equity over total assets 0.221 0.066 0.042 -0.117 (0.168) (0.249) (0.202) (0.255) Constant 8.082*** 189.676*** 222.619*** 259.652*** 265.033*** (0.706) (21.315) (32.357) (16.808) (19.965) Calendar Year FE Yes Yes Yes Yes Yes Bank FE Yes Yes Yes Yes Yes Only Only Countries Sample Full Full PSM Sample Large Banks with a Subsidiary Observations 7,644 7,644 4,540 5,981 4,370 R-squared 0.472 0.529 0.530 0.537 0.512 34