Supranational rules, national discretion:

Increasing versus inflating regulatory bank capital?

This version: October 29, 2020

Abstract

How do higher capital requirements introduced at the supranational level affect regulatory capi- tal of banks across countries? Using the 2011 EBA capital exercise as a quasi-natural experiment, we find that treated banks increase their regulatory capital without increasing book equity and a corresponding contraction of credit default swap spreads. The incidence of regulatory capital in- flation is more pronounced in countries with powerful national supervisors and with tight credit supply, suggesting that national authorities do appear to apply discretion to forbear their banks to meet supranational regulatory requirements without an increase in risk absorbing capacity.

1 The regulatory framework introduced after the 2008/2009 financial crisis required banks to increase their regulatory capital ratios to foster financial stability. Capital ratios as a policy tool, however, are frequently criticized for being too complex, too opaque, and for their calculation being subject to too much discretion by supervisors (Haldane, 2012, 2013). While much of this debate has centered on the calculation of risk-weighted assets in the denominator of the capital ratio

(Acharya, Engle, and Pierret, 2014; Behn, Haselmann, and Vig, 2016; Plosser and Santos, 2018), the calculation of the numerator, regulatory capital, received much less attention so far.

This study attempts to close a part of this gap in the literature by examining how banks react to a sudden supranational demand for higher capital requirements. During our examination we uncover how banks rely on financial reporting tools to inflate their regulatory capital rather than increase their levels of book equity. These tools it turns out are discretionarily provided to them by their own national regulatory authorities.

The rules governing the definition of regulatory capital are complex and include a plethora of deductions and instruments that vary across countries. Discretion in these rules provide banks

(Koch, Schneider, Schneider, and Schroeck, 2017) and national authorities (Angeloni, 2014) with considerable leeway to inflate banks’ levels of regulatory capital without a commensurate increase in their levels of book equity.1 Furthermore, for the purpose of increasing regulatory capital ratios, boosting regulatory capital can achieve more bang for the buck” than a reduction in risk-weighted assets (e.g., (de Weert, 2012)).2

Indeed, in order to calculate regulatory capital, banks must deduct several items from their book equity that are not considered to increase a bank’s capability to absorb losses (BCBS, 2015).

If banks can discretionarily adjust these deductions, then such adjustments will allow banks to increase their levels of regulatory capital without issuing (costly) equity or retaining additional earnings. National authorities may assist banks’ efforts to inflate their regulatory capital by en- acting tax policies that boost regulatory capital or by making hybrid capital instruments eligible for regulatory capital.3 So while in theory supranational rules should be binding, in practice na-

1Angeloni (2014) points out that “. . . inconsistent application of national discretions, especially when leading to cross-border discrepancies in the level and the quality of capital, increases the potential reliance of banks on external support in certain constituencies relative to others”. 2For a bank with e.g. a 9 percent regulatory capital ratio, a marginal one dollar increase in regulatory capital increases the ratio as much as a marginal ten-dollar reduction in risk-weighted assets. 3Several European countries (Italy, Spain, Portugal and Greece) implemented changes in their tax regime to allow banks to convert part of their deferred tax assets into a transferable tax credit to boost capital ratios. In 2012, both

2 tional discretion may effectively undermine them.4 Causally identifying the use of discretion in the calculation of regulatory capital is, however, empirically most challenging. The amount of a bank’s regulatory capital depends on the underlying structure of its balance sheet, which makes it difficult to disentangle changes in regulatory capital to inflate capital ratios from changes due to other endogenous variation in a bank’s balance sheet. Thus, an ideal experiment would be to find an exogenous shock to banks’ needs to raise their regulatory capital ratios and then compare their changes in capital deductions to a control group of banks without an increased need to raise capital ratios.

The 2011 capital exercise, conducted by the European Banking Authority (EBA)5, provides such a quasi-natural experiment. The capital exercise raised the minimum required core tier 1 (CT1) capital ratio from 5% to 9% for a subset of European banks while leaving requirements unchanged for other European banks. The EBA, however, failed to specify enforcement actions related to their recommendations on how banks had to be recapitalized. This resulted in considerable discretion for national supervisors regarding which measures to take to reach the higher capital requirements and room to exercise forbearance. This empirical setting naturally lends itself to a difference-in- differences research design. For our baseline analysis, we compare changes in the ratio of regulatory capital relative to book equity between banks subject to the 2011 capital exercise (CE banks) and banks not subject to this exercise (non-CE banks). If banks increase their regulatory capital without a commensurate increase in their book equity, then this implies a reduction in capital deductions which, in turn, increases the ratio of regulatory capital relative to book equity.

Our main findings are as follows. First, we find that CE banks (i.e., our treatment group) increased their regulatory capital relative to their book equity via a reduction in capital deductions by 10 percentage points relative to comparable non-CE banks (i.e., our control group).6 The magnitude of this effect is economically significant. The average CE bank increased its regulatory

Italy and Portugal enacted decrees allowing the issuance of new hybrid instruments eligible as regulatory capital (See Table VII for more details). In addition, the national implementation of the Capital Requirements Directive gave national authorities national options and discretion (Maddaloni and Scopelliti, 2019). 4Regulatory forbearance could result in insufficiently stabilized banks, which in their turn continue lending to existing zombie borrowers (Acharya, Eisert, Eufinger, and Hirsch, 2019). 5The European Banking Authority (EBA) is an independent EU Authority which works to ensure an effective and consistent level of prudential regulation and supervision across the European banking sector. 6We adopt the following terminology: capital exercise (CE) banks are banks selected into the 2011 EBA capital exercise and therefore subject to the increase in capital requirements (treatment group); non-capital exercise (non- CE) banks are other European banks not selected into the EBA capital exercise and therefore not subject to the increase in capital requirements (control group).

3 capital from 2010 to 2012 by 17.0 percent, but its book equity by only 5.8 percent. Thus, for the average CE bank more than half of the increase in regulatory capital came from a reduction in capital deductions, and not from an increase in book equity.7 We interpret this finding as evidence that discretion was exercised in the calculation of regulatory capital at CE banks to inflate their capital ratios.

We further investigate how the capitalization of CE banks around the capital exercise would have changed in the absence of discretionary capital deductions. To that end, we calculate a

“shadow regulatory capital ratio” (i.e., book equity/risk-weighted assets) and investigate how this ratio changed compared to banks’ officially reported capital ratios (i.e., regulatory capital/ risk- weighted assets). We find that CE banks would not have achieved a material improvement in their reported capitalization relative to control group banks if they did not inflate regulatory capital.

Finally, the empirical evidence of capital inflation is stronger for CE banks with ex-ante lower levels of regulatory capital. This is consistent with weakly capitalized banks having a stronger incentive to engage in discretion to pass the EBA capital exercise.

Since the main objective of the capital exercise was to bolster confidence in the banking system by ensuring that banks are sufficiently capitalized to absorb unexpected losses, an increase in regulatory capital ratios shouldin principlereflect an increase in banks safety and soundness. We therefore study the effect of the capital exercise on market-based and accounting-based measures of bank risk. We find that changes in the CDS spreads of CE banks do not significantly differ from the CDS spreads of control group banks, neither around the announcement date of the capital exercise nor between the start and end date of the capital exercise. Moreover, we find a reduction in the z-score of weakly capitalized CE banks between 2010 and 2012 relative to the control group suggesting that the increase in regulatory capital ratios of CE banks did not reflect an improvement in the safety and soundness of the stress-tested banks.

The empirical results thus far suggest that exercising discretion in the calculation of regulatory capital is undesirable from a prudential point of view and it was fully understood by market participants. While the EBA capital exercise was a uniform supranational regulatory intervention,

7One mechanism to reduce capital deductions (and, hence, “inflate” regulatory capital) is by impairing goodwill. In 2011, the Italian government enacted Law Decree no. 98/2011, allowing tax-deductible goodwill impairments which can result in a net increase in regulatory capital after an impairment (See Online Appendix 2 for more details). This measure alone resulted in a 40 basis points increase in the regulatory capital of Banca Monte dei Paschi di Siena.

4 national supervisors were ultimately in charge of approving the measures by which banks intended to increase their capital ratios EBA (2011a). Thus, supervisors should be wary of banks inflating regulatory capital ratios. However, national supervisors might choose to be lenient on their banks for a variety of reasons. They might also be prone to regulatory capture and have a tendency to be too soft on their national champion banks (Goodhart, 2012; Schoenmaker, 2012; Haselmann,

Singla, and Vig, 2018; Bruno and Carletti, 2019), they might want to minimize disruptions to the real economy caused by bank failures (Walther and White, 2016), or their actions are constrained by political considerations (Brown and Din, 2005; Behn et al., 2016). Moreover, during the capital exercise, there was considerable heterogeneity in supervisory approaches across countries in Europe

(Barth, Caprio, and Levine, 2013; Nouy, 2017).

We document substantial heterogeneity across countries in the extent to which banks engaged in regulatory capital inflation suggesting that national supervisors play an important role. In par- ticular, we find that in countries in which national supervisors wield more discretionary power to exercise leniency (measured using the so-called Official Supervisory Power Index that was in- troduced by Barth, Caprio, and Levine (2013)), banks are more prone to inflate their reported capital ratios via reductions in capital deductions. We also find that in countries with a current ac- count deficitreducing the fiscal capacity to bail-out banksand in countries expecting tighter credit standards, banks are more likely to inflate their bank capital. The findings are consistent with the idea that local economic circumstances incentivize national authorities to exert discretion and forbearance., with an eye on short-term economic considerations.

Our paper contributes to three distinct parts of the literature. First, our paper contributes to the literature on centralized versus decentralized bank supervision (DellAriccia and Marquez,

2006; Calzolari, Colliard, and L´or´anth, 2019; Carletti, DellAriccia, and Marquez, 2020). While sev- eral studies show that local (national) supervisors may be more lenient as compared to centralized regulators in day-to-day oversight of the banks (Agarwal, Lucca, Seru, and Trebbi, 2014; Hasel- mann, Kick, Singla, and Vig, 2019), our study more sharply focuses on the distinctly heterogeneous response of national authorities to a one-time and uniform supranational recapitalization exercise.

Second, our paper also relates to the literature on supervisory discretion, forbearance and leniency (Martynova, Perotti, and Suarez, 2019; Walther and White, 2016). Previous studies show that regulators and supervisors exercise forbearance due to fiscal constraints (Acharya, Borchert,

5 Jager, and Steffen, 2020), electoral cycles (Brown and Din, 2005; Bian, Haselmann, Kick, and Vig,

2017), and too-many-to-fail effects (Brown and Din, 2011), with potentially adverse effects for the real economy (Gropp, Ongena, Rocholl, and Saadi, 2018). More specifically, some studies argue that national banking supervisors are especially prone to exercising leniency when it comes to preserving their so-called “national champions” (Goodhart, 2012; Schoenmaker, 2012; Haselmann,

Singla, and Vig, 2018; Bruno and Carletti, 2019). We contribute to this literature by documenting that national supervisors show discretionary leniency with regards to their own banks’ pursuit of regulatory capital arbitrage to pass a supranational recapitalization exercise.

Finally, our results are also crucial for the design and implementation of supranational stress testing and bank supervision. While (Abbassi, Iyer, Peydr´o,and Soto, 2020) show that banks decrease the share of riskier securities and loans before the ECB’s 2014 asset quality review and reload them afterwards, our paper shows that the key regulatory intervention, i.e., to demand banks that failed the stress test to increase their regulatory capital, may result in inflated regulatory capital with no substantial improvement in financial stability. One of the main objectives of the

EU Single Supervisory Mechanism (SSM) is to harmonize supervisory practices by reducing options and national discretions. Although significant progress has made, national authorities still have substantial discretion in policy areas not covered by SSM, such as their national tax and bankruptcy codes. Our results show that agency problems between supranational and national authorities are an important force affecting the effectiveness of all supranational ruling.

The remainder of the paper proceeds as follows. Section I discusses the institutional background of the EBA capital exercise and the rules governing the calculation of regulatory capital. Section

II presents our empirical strategy and the data. Section III reports the results on the effect of the

EBA capital exercise bank capital structure and bank stability. Robustness checks are presented in Section IV. Section V concludes.

I. Institutional Background

A. The 2011 EBA Capital Exercise

The 2011 capital exercise, conducted by the European Banking Authority (EBA), was an- nounced on October 26, 2011, and required 61 European banks to reach and maintain a 9% core

6 tier 1 (CT1) capital ratio by the end of June, 2012. This constituted an economically significant increase compared to the previously required 5%. The selection of the 61 banks was based on total assets. In particular, for each country, the EBA included “banks in descending order of their market shares by total assets,” such that the exercise covered ”at least 50 percent of the national banking sectors in each EU Member State in terms of total consolidated assets as of end 2010”

(EBA, 2011b). This approach was similar to the 2011 EU-wide stress test whereby the selection procedure was based on total assets as of year-end 2010 such that the sample was not influenced by bank-specific events in the months leading to the exercise. Regardless, the capital exercise was not a stress test; it was a one-off exercise intended to improve the capital positions of European banks and for which the timing and magnitude was unexpected. The capital exercise occurred only a few months after the EU-wide stress test in June 2011 and was described as a “quick-fire regulatory health check” (Halstrick and Framke, 2011). The Financial Times reported that the 9% requirement was “well beyond the current expectations of banks and analysts” (Atkins, Jenkins, and Spiegel, 2011). Ex ante, the EBA recommended that “banks should first use private sources of funding to strengthen their capital position to meet the required target, including retained earnings, reduced bonus payments, [and] new issuances of common equity” (EBA, 2011a). Undercapitalized banks were required to submit their recapitalization plans to their respective national supervisory authorities outlining how they plan to reach the set target of 9%. The EBA, however, failed to specify enforcement actions related to their recommendations on how banks had to be recapitalized.

This resulted in considerable discretion for national supervisors regarding which measures to take to reach the higher capital requirements and room to forbear banks.

B. Regulatory Capital

Bank’s risk weighted assets (for regulatory purposes) differ from its book assets (for financial reporting purposes). Similarly does regulatory capital differ from book equity. This difference originates from the difference in objectives between financial reporting regulation and prudential regulation. While the aim of financial reporting is to provide information about the economic performance and condition of businesses, the objective of prudential regulation is to promote the safety and soundness of banks and the banking system. As such, the primary role of bank capital is to absorb losses arising from financial and economic stress. Core Tier 1 capital is the highest

7 quality of regulatory capital as it absorbs losses immediately when they occur. It consists of capital instruments of the highest quality (i.e., ordinary shares) plus reserves (retained earnings) less several

“regulatory deductions”. Regulatory deductions are items that, from a prudential point of view, are less effective in absorbing losses and, therefore, need to be deducted from (high quality) capital.

Examples include unrealized gains and losses on availableforsale instruments, intangible assets, deferred tax assets, and holdings in other financial institutions. The composition of regulatory capital is thus different from book equity and is governed by a complex set of rules.

C. Economic mechanisms explaining regulatory capital inflation

To increase its loss absorptive capacity, banks can adjust their business model or capital struc- ture in line with the capital rules as defined in the Basel framework. Some adjustments, however, result in higher levels of regulatory capital without a corresponding increase in equity. Whether an increase in capital ratios via such channels (i.e., inflation) is associated with stronger bank stability is an empirical question. If these adjustments enhance the capacity of banks to absorb losses, we would expect a positive association with bank stability.

Alternatively, capital inflation could be the result of regulatory arbitrage where banks merely exploit the complexity embedded in the rules governing regulatory capital. Banks have strong incentives to game capital rules by adopting strategies of regulatory capital arbitrage (Gerding,

2016) as it allows them to increase regulatory capital without the need of raising (costly) equity in the short run. This type of capital inflation increases capital ratios but does not improve bank stability as measured by bank credit default spreads.

Forbearance by national authorities forms a third explanation for capital inflation. National authorities confronted with banks violating the minimum capital requirements could decide to for- bear banks and postpone costly capital injections (Martynova, Perotti, and Suarez, 2019). The

Basel framework provides substantial discretion for national authorities to forbear domestic banks

(Maddaloni and Scopelliti, 2019). National authorities could also enact favorable tax regulation to boost bank capital without effectively improving the capital quality (Journal, 2014). Since the

EBA demanded national authorities to approve the bank recapitalization plans for the 2011 capital exercise, national supervisory incentives might have come into play resulting in them (potentially) turning a blind eye on arbitrage activities of banks to satisfy the newly imposed capital require-

8 ments. Though disentangling regulatory arbitrage versus regulatory forbearance at the individual country level is difficult, forbearance takes place at the national level which is why we predict substantial cross-country variation in bank capital inflation.

II. Empirical Strategy and Data

A. Empirical Strategy

Studying whether the rules governing the calculation of regulatory capital are exploited to inflate capital ratios is challenging in the absence of a true control group that is unaffected by these motives. We address this empirical challenge by exploiting the 2011 EBA capital exercise as an exogenous shock to banks’ capital requirements and thus banks’ needs to increase their regulatory capital ratios. The setup of the capital exercise, whereby the EBA required a subset of European banks to reach a 9% CET1 capital ratio by the end of June 2012, while leaving requirements unchanged for other European banks, naturally lends itself to a difference-in-differences research design (see Gropp, Mosk, Ongena, and Wix (2019)). As mentioned above, the EBA selected banks according to an explicit selection rule allowing a 50% coverage of each EU Member States’ banking sector. This country-specific selection threshold yields a considerable overlap in size and other variables between banks selected and not selected into the capital exercise, which allows us to control for bank-specific characteristics.

We propose a simple, intuitive measure capturing discretion in the calculation of regulatory capital, i.e. the change in the ratio of a bank’s regulatory capital relative to its book equity. If banks increase their regulatory capital without a commensurate increase in their book equity, then this implies a reduction in regulatory deductions and, therefore, an increase in the ratio of regulatory capital relative to book equity. We estimate the following difference-in-differences specification to test whether the EBA capital exercise affected the regulatory capital structure of treated banks:

k X k Yi = α + β × CEBi + θ Xi,c + γc + i,c (1) k where for our main analysis Yi is the change in regulatory capital to total equity ratio from 2010 to

2012. The variable CEBi takes on the value of 1 for banks selected into the capital exercise, and 0

9 otherwise. We control for the following bank characteristics as of 2010: log total assets, CET1 ratio, customer loans as a share of total assets, net interest income as a share of total operating revenue, depository funding as a share of total assets, and net income over total assets. Additionally, we include country fixed effects γc. We report standard errors clustered at the country level. We examine whether such an increase is driven by bank specific characteristics, such a proxies for bank asset complexity, or country specific characteristics, such bank supervisory measures and macro- economic measures, to test whether regulatory arbitrage or regulatory forbearance is more likely to explain the observed behavior. Finally, we examine the loss absorptive capacity hypothesis by studying the effect of the increase in capital requirements on bank risk. We use three measures of bank risk as dependent variables, the CDS Returns on the EBA capital exercise announcement date, the change in CDS spreads between October 2011 and June 2012, and the change in the bank

Z-Score between 2010 and 2012.

B. Data

In this section, we discuss the data sources we used in our analyses. We use annual bank balance sheet data from the SNL Financial Company database. Our initial sample contains all 61 CE banks and 494 non-CE European commercial and savings banks from the SNL Financial universe. We follow the sample construction procedure in (Gropp, Mosk, Ongena, and Wix, 2019) and exclude all subsidiaries of CE banks, non-CE banks, and foreign banks, all banks which were acquired during the sample period, all banks which received capital injections during the pre-treatment period, and all banks with negative levels of equity. This sample construction procedure finally leaves us with a sample of 48 CE banks (our treatment group) and 143 non-CE banks (our control group). The pre- and post-treatment periods in our analysis are 2010 and 2012, respectively, the years immediately before and after the capital exercise. All outcome variables are winsorized at the 5 percent level to reduce noise from extreme outliers. We obtain prices of five-year of CDS contracts on senior and junior bonds of European banks. CDS pricing series are only available for 51 banks (34 CE banks and 17 non-CE banks).

Table 1 shows summary statistics for all CE and non-CE banks in our sample for the pre- treatment year 2010. We use Welchs t-test to test for differences in means between the groups.

Due to the capital exercise being carried out on the largest banks in each country, the mean capital

10 exercise bank is more than 17 times larger than the mean non-capital exercise bank.8 The absolute euro-amount of Core Tier 1 capital is naturally also larger for CE bank. A crucial assumption for our analysis is that despite of these differences, the regulatory capital of CE and non-CE banks follows a similar pattern over time in absence of the capital exercise.

[Table I about here]

[Figure 1 about here]

III. Results

A. Univariate Results

Figure 1 graphically illustrates the evolution of book equity and regulatory capital over time for

CE banks and non-CE banks, normalized to the value of 1 in 2010. In the years prior to the capital exercise, both book equity and regulatory capital evolved similarly for both groups. However, while book equity and regulatory capital continued to evolve identically for non-CE banks, CE banks increased their regulatory capital visibly more than their book equity from 2010 to 2012.

This indicates a considerable reduction in CE banks regulatory deductions in response to the capital exercise. In Table II, we examine more closely the components of regulatory capital change between

2010 and 2012 and make two critical observations. First, we find that CE banks and non-CE banks increased their book equity over the sample period but CE banks have done this in a slower rate compared to non-CE banks. And second, while our control group demonstrates similar growth rates for both their regulatory deductions and their book equity, CE banks have substantially decreased their regulatory deductions over the sample period, boosting their regulatory capital. Based on

Table II, we conclude that while regulatory capital of both treatment and control group has grown at similar rates, the underlying cause of this increase is dissimilar for both groups. Non-CE banks have increased regulatory capital primarily through equity while CE banks have achieved this by decreasing their regulatory deductions.

8See (Gropp, Mosk, Ongena, and Wix, 2019) for a detailed discussion on the difference between CE and non-CE banks and various sampling and matching strategies to reduce these differences.

11 B. Regulatory Capital Inflation

While the univariate descriptive statistics and the graphical evidence are suggestive that CE banks exercised discretion in the calculation of regulatory capital, these changes could conceivably be driven by other bank-specific factors. Table III therefore presents the estimation results of the difference-in-differences regression from Equation (1) in Section II. The first column provides the unconditional treatment effect of the capital exercise and shows that CE banks increased their regulatory capital relative to their book equity by 9 percentage points compared to non-CE banks.

The second column additionally controls for 2010 pre-treatment levels of log total assets, CT1 ratios, deposits over total assets, loans over total assets, net interest income over total operating revenue, and net income over total assets. In this specification, capital exercise banks increased their regulatory capital to book equity ratio by 7 percentage points compared to non-capital exercise banks, alleviating the concern that our results are driven by either bank size, profitability, banks’ business models, or funding strategies. The Core Tier 1 ratio coefficient (i.e., -0.004 with p- value¡0.05) suggests that lower capitalized banks are more pronounced to inflate their regulatory capital. The third column adds country dummy variables and compares CE banks and non-CE banks within countries, addressing concerns that the treatment effect disappears once controlling for country-specific factors. In this specification, CE banks increased the regulatory capital to book equity ratio by 10 percentage points. In the last column, we interact the CEB dummy variable with the 2010 Core Tier 1 ratio of the banks. The estimate (i.e., -0.032 with p-value¡0.01) shows that lower capitalized CE banks are more inclined to inflate their regulatory capital. CE banks with a one standard deviation lower capital ratio in 2010 increase their core tier 1 ratio with an additional 9 percentage points than the average CE bank. This result is consistent with (Lubberink, 2014), who shows that lower capitalized US banks are more likely to use regulatory adjustments to boost their regulatory capital ratios. Our result is consistent with both the regulatory arbitrage and forbearance hypothesis, because lower capitalized bank have more incentive to exploit in regulatory arbitrage strategies (e.g., Kisin and Manela (2016)) and national authorities are more likely to forbear low- capitalized banks (Brown and Din, 2011; Acharya, Borchert, Jager, and Steffen, 2020). It is not easy to disentangle these two mechanisms, because even if banks attempt to exploit loopholes in capital regulation, a form of forbearance by national authorities is to allow banks to exploit these loopholes.

12 This explanation might be relevant to our set up because national supervisors had to approve the recapitalization plans of the CE banks (EBA, 2011a). Finally, in column (5) we examine while larger banks are more likely to inflate their regulatory capital. Larger banks often have more intangible assets from subsidiaries, and a larger trading book, resulting in more regulatory deductions from their common equity to determine their regulatory capital. Since larger banks in general are more complex, they might have more arbitrage opportunities they could exploit. The results in the fifth column, however, show that capital inflation of the treated banks is not related with their size.

Variation in bank size and variation in bank size within the treatment group seem not to affect regulatory capital inflation, providing evidence that the difference in size between the treatment and control group banks does not play a significant role in the bank behavior we study.

[Table III about here]

C. Regulatory Capital Inflation and Financial Stability

As the main objective of the EBA capital exercise was to “restore confidence in the EU banking sector” (EBA, 2011a). In its final report, the EBA (2012) mentions that “the vast majority of the banks involved in the EBA capital exercise show a CT1, as of end of June, above the 9%

[. . . ]” and that “direct capital measures rise to [. . . ] 72% of the recapitalization amount”. Direct capital measures include items such as retained earnings and equity as well as liability management exercises and “other mitigating measures”. The latter category includes, inter alia, regulatory deductions which are subject to potential inflation. To test whether the increase in capital ratios is primarily driven by quality capital rather than by capital inflation, we contrast the effect of the capital exercise (i.e., CEB dummy) on two different capital ratios: (1) Core Tier 1 ratio (i.e., Core

Tier 1 Capital/RWA); and (2) a shadow bank capital ratio (i.e., Book Equity/RWA). Both ratios have the same denominator (RWA) but differ in terms of the capital definition used in the numerator

(Core Tier 1 Capital vs. Book Equity). Table IV, Column 1 and 2, reports the regression results of Equation (1) but with Core Tier 1/RWA and Book Equity/RWA as the dependent variable, respectively. Based on Column 1 and, we find evidence that CE banks did increase their reported capital ratios, in particular lower capitalized banks (column 2). However, if we focus on Column 3 and 4, we find the results do not hold when using the shadow capital ratio (i.e., Book Equity/RWA).

13 The CEB coefficient is statistically insignificant and considerably smaller in magnitude than for the reported CT1 ratios. Although the capitalization of the CE banks improved on paper post capital exercise, the results of this improvement is largely driven by the inflation of regulatory capital instead of an increase in equity.

[Table IV about here]

Since the prudential goal of demanding higher (quality) capital is to improve a bank’s ability to absorb unexpected losses, any increase in capital ratios should be associated with bank stability.

If our results are driven by managerial discretion (i.e., regulatory arbitrage) and/or preferential regulatory treatment (i.e., forbearance), then the market perception of the bank’s riskiness should remain unaffected. To investigate this, we look whether the capital exercise and the associated improvement in capital positions is associated with a change in the market perception of bank debt.

To that end, we first compare the three-day CDS return around the announcement of the capital exercise on October 26, 2011 of CE and non-CE banks. Panel A of Table V shows in the first three columns that the announcement of the capital exercise did not reduce the CDS spread of treatment banks. The interaction between the treatment dummy and the 2010 Core Tier 1 capital ratio (i.e.,

CEB × Core Tier 1 2010) tests for differences between high versus low capitalized banks. Since the capital exercise was aimed at improving the capital positions of weakly capitalized banks, we should expect their CDS spread to shrink after the announcement consistent with an improvement in bank stability. Column (4) of the table shows that this interaction coefficient is insignificant.

Subsequently, we investigate the effect of the capital exercise on the change in CDS spreads between the start (October, 2011) and the end (June, 2012) of the capital exercise. It is important to note that over the period of the capital exercise other events, such as the announcement of the LTRO

(Acharya, et al., 2019), might have affected bank CDS spreads. To alleviate some of these concerns, we compare the change in CDS spreads with our control group adding country fixed effects and the necessary control variables. The results show no significant effect on both the change in senior and junior CDS spreads. Finally, Figure 2 plots for each quarter the estimated difference in CDS spreads between CE and non-CE bank in a panel including the CDS spreads of CE and non-CE bank over the period 2011Q1 to 2012Q4. This figure illustrates that the CDS spreads of the CE bank are on a parallel trend in the quarter before the capital exercise and do not significantly drop

14 in comparison with the non-CE banks during the capital exercise (2011Q4-2012Q2). Summarized, we fail to document any evidence that the capital exercise was associated with a change in market perception of the riskiness of bank debt. A drawback of the use of CDS spreads though is that these panel series are only available for a subset of the banks in our sample (45 treatment versus 11 controls). Consequently, in Panel B we additionally estimate the same specifications but with the bank’s z-score as our dependent variable. The calculation of the z-score is based on SNL data which is available for all treatment and control group banks. It compares capital buffers (capitalization and returns) with bank riskiness (volatility of returns).9 The popularity of the z-score stems from the fact that it has a clear (negative) relationship to the probability of a financial institution’s insolvency, that is, the probability that the value of its assets becomes lower than the value of its debt. A higher z-score therefore implies a lower probability of insolvency.

[Table V about here]

[Figure 2 about here]

The results in Panel B, Column 1, show that the z-score of CE bank decreases between 2010 and

2012 relative to non-CE banks. However, if we include control variables and country fixed effects the treatment effect is not statistically significant. Finally, we interact our treatment dummy with the pre-treatment Core Tier 1 ratio and show in column (4) that the z-score of low capitalized CE banks decreases, suggesting a higher probability of insolvency (i.e., coefficient CEB x CT1 Ratio

2010 = 0.82 with p-value¡0.01). Although the z-score measure is based on financial reporting data instead of a market measure (e.g., CDS spreads) the results confirm that the EBA capital exercise did not improve the stability of treated banks. Our results are also in line with the findings of

Bostandzic, Irresberger, Juelsrud, and Wei (2020) who show that the EBA capital exercise did not result in an improvement of several market-based measures of bank risk (i.e., systematic risk, stock return volatility, market leverage, and Value-at-Risk) of treated banks. In sum, our results are inconsistent with our loss absorption capacity hypothesis which predicts that an increase in regulatory capital is associated with an improvement in the loss absorptive capacity of bank capital

9The z-score is defined as (ROA + Equity/ Total assets) / σ(ROA), ROA is the return as percent of assets, and σ is standard deviation of return on assets as a proxy for return volatility. Papers that used the z-score for analysis bank stability include Beck, Demirg¨u¸c-Kunt, and Levine (2006) and Laeven and Levine (2009).

15 and, thus, an improvement in the market perception of the riskiness of bank debt. Since we do not observe a reduction in CDS spreads of treated banks, it is more likely that the other two economic mechanisms (i.e., regulatory arbitrage and forbearance by national supervisors) explain the substantial increase in regulatory capital.

D. Regulatory Arbitrage and Regulatory Forbearance

Regulatory arbitrage and forbearance are difficult to disentangle at the country level because national authorities might knowingly allow banks to exploit arbitrage opportunities. The advantage of our empirical set up is that we could test whether the incidence of capital inflation differs across countries. One test we employ to examine whether capital inflation is driven by national authorities is to interact the CEB dummy with bank headquarter country dummies. Figure 3 plots the estimated coefficients and confidence intervals, showing substantial variation in the CEB treatment effect across countries. Specifically, the figure shows that Italian and Portuguese CE banks were more likely to inflate their regulatory capital than banks in other European countries.

Beside these two countries, which we will discuss in detail below, we also observe that German,

French, Norwegian and Slovenian banks are more likely to inflate their regulatory capital, although the magnitude of the coefficient is smaller. We report the estimated coefficients of the interaction between the CEB dummy with both the Italy and Portugal bank headquarter country dummy in Table VI. Column (1) shows that the magnitude of the CEB average treatment drops once we include these two country-interactions and becomes insignificant once we include our control variables. Since we established in the previous table that low capitalized banks drive our main results, we include the interaction term (CEB × CT1 Ratio 2010) in the third column. The regulatory capital to equity ratio of the average CE bank increased over the sample period relative to non-CE banks and this effect is more pronounced for lower capitalized banks but does not affect the magnitude of the CEB dummy-country interaction coefficients. Finally, we include the interaction term (CEB × Log Total Assets 2010) in the fourth column and show that the CEB dummy-country interaction coefficients are not driven by large CE banks.

[Figure 3 about here]

[Table VI about here]

16 The results in Figure 3 and Table VI strongly indicate that country specific factors explain a part of the regulatory capital inflation results. To understand the economic mechanism behind the observed cross-country heterogeneity, we follow two approaches. Firstly, we collect anecdotal evidence of country-specific policies that have been implemented to support banks around the time of the capital exercise. These results are illustrated in Table 7 and online in Appendix 2. Secondly, we interact the treatment bank indicator with characteristics of the supervisory institutions, local macroeconomic and political environment.

For our first approach, we identify forbearance by national authorities if it involves a public intervention (e.g., recapitalizations, guarantees, troubled asset relief programs or liquidity sup- port measures) since they require an official approval from the European Commission (Acharya,

Borchert, Jager, and Steffen, 2020). Other observable policies that we identify as a form of forbear- ance are changes in the national bank or tax code which require parliamentary approval. The first group of anecdotal forbearance measures are ad-hoc recapitalizations, such as the issuance of Core

Tier 1 eligible hybrid securities, underwritten by the state. These instruments increase regulatory capital but are not part of book equity. Both Italy (Banca Monte dei Paschi di Siena) and Portugal

(Banco BPI, Banco Comercial Portugues, Caixa Geral de Depositos) have underwritten hybrid securities to increase the regulatory capital of CE banks and often explicitly cite the EBA capital exercise as the motivation to provide state aid. For example, the state aid application of Banco

Comercial Portuguˆesnotes “on 8 December 2011, the EBA published a Recommendation related to banks’ recapitalisation needs. In the light of the foregoing, the BCP Group needed to raise substantial additional capital by 30 June 2012”.10 Besides direct capital support measures, several countries have implemented specific regulation aiming to increase the regulatory capital of banks during the time of the EBA capital exercise representing our second group of anecdotal forbearance measures. This includes the use of the tax code to support banks. We illustrate an example of how the tax treatment of goodwill impairments affects regulatory capital. Intangible assets and goodwill are part of regulatory deductions which means that they are deducted from CT1 capital.

However, these assets are typically large in magnitude relative to a bank’s regulatory capital (24% for the CE banks in our sample as of 2010). Banks could reduce the book value of intangible asset by impairing these assets. From an accounting perspective, this results in a decrease of the book

10State aid case SA.34724. Link: https://ec.europa.eu/competition/elojade/isef/case details.cfm?proc code=3 SA 34724

17 value of goodwill (equal to the amount of the impairment charge) but also reduces net income (by that same amount). The impact on CT1 capital is initially capital neutral because, on the one hand, the impairment charge reduces CT1-deductions but, on the other hand, it also reduces prots which is a component of CT1 capital. One way national authorities could support banks is to allow intangible asset impairments to be tax-deductible which would result in a net increase in regulatory capital after impairment. 11 In 2011, the Italian government enacted Law Decree no. 98/2011,12 allowing banks to boost their regulatory capital by introducing such tax-deductible impairments.

For example, this particular Decree resulted in a 40 basis points increase in the regulatory capital of Banca Monte dei Paschi di Siena’s.13 Another way for national authorities to support their banks was to allow Deferred Tax Assets (DTA’s) to be converted into government guaranteed tax credits. These DTA’s did not have to be deducted from Core Tier 1 capital as such boosting the banks regulatory capital. Online Appendix 2 discusses these forbearance measure in more detail and provides other examples of how national authorities can use tax instruments to increase the regulatory capital of their banking sector. While these anecdotes are suggestive of forbearance, they do not capture all policies from national authorities that might be driving our results. Actions of supervisory authorities are usually not observable, and one specific form of forbearance of the national supervisor, doing nothing, or turning a blind eye on specific bank behavior (Martynova,

Perotti, and Suarez, 2019) is inherently difficult to document. Table VII provides evidence of coun- tries that implemented (observable) policies which helped banks to comply with the supranational requirements of the EBA, consistent with Figure 3 and the results in Table VI.

[Table VII about here]

Besides anecdotal evidence, we also interact the treatment bank indicator (i.e., CEB) various country-level variables. National authorities might have different economic reasons to forbear their banks: they might also be prone to regulatory capture and have a tendency to be too soft on their national champion banks (Goodhart, 2012; Schoenmaker, 2012; Haselmann, Singla, and Vig, 2018;

11Assume a bank impairs goodwill for an amount of 1,000 EUR which is tax-deductible. The tax rate is 20%. This will result in a reduction of 1,000 EUR in capital deductions (thus increasing CT1 capital) whereas net income will decrease with 800 EUR (1,000*(1-0,2)) rendering an net positive effect on CT1 capital of 200 EUR due to a permanent difference between book and tax income. 12https://www.normattiva.it/uri-res/N2Ls?urn:nir:stato:decreto.legge:2011-07-06;98!vig= 13https://www.gruppomps.it/en/media-and-news/press-releases/banca-monte-dei-paschi-siena-tier-1-up-to-8-8- per-cent.html

18 Bruno and Carletti, 2019), they might want to minimize disruptions to the real economy caused by bank failures (Walther and White, 2016), or their actions are affected by political considera- tions (Brown and Din, 2005; Bian, Haselmann, Kick, and Vig, 2017). In Table VIII we examine these explanations by interacting our CEB dummy with several country characteristics. First, we examine whether national champions, defined as the largest bank in a country, drive our results.

Column (1) a negative, but insignificant relationship between national champions and our capital inflation measure, suggesting that collusion between national authorities and the largest national bank does not drive or results. Subsequently we examine whether differences in supervisory insti- tutions explains the cross-country variation in capital inflation. We employ Official Supervisory

Power Index (Barth, Caprio, and Levine, 2013) which measures the degree to which the country’s bank supervisory agency has the authority to take specific actions. Column (2) shows that coun- tries with more powerful supervisors are more likely to engage in capital inflation (i.e., coefficient of CEB × Official Supervisory Index = 0.062 with p-value¡0.05). We measure fiscal capacity, in line with Acharya, Borchert, Jager, and Steffen (2020), using the current account surplus/deficit as percentage of nominal GDP. If a country is borrowing from abroad on a net basis and has a current account deficit, it is subject to market discipline and possible reversal of capital flows when foreign investors become unwilling to roll over their funds, reducing tax income and thus increas- ing fiscal constraints. In line with this prediction, column (3) shows a negative, but insignificant correlation between the current account balance and our capital inflation measure. DellAriccia and

Marquez (2006) argue that supra-national bank supervision is more likely to be beneficial for more homogenous countries because the costs associated with a loss of flexibility is lower for countries that are similar. We use the share of banks expecting a credit standard tightening from the Bank

Lending Survey to proxy for homogenous countries and interact it with our CEB dummy to test whether capital inflation is more pronounced in countries that are similar to each other in terms of macro-economic environments. The interaction term in Column (4) shows that regulatory cap- ital inflation is indeed stronger in countries expecting a credit crunch (i.e., coefficient of CEB ×

Credit Standard Tightening = 0.004 with p-value¡0.01). This result is consistent with the idea that national authorities might choose to forbear banks in bad times, to avoid signaling additional bad news which could create a bank run (Walther and White, 2019). While this result does not show whether forbearance helps to avert a credit crunch, Acharya, Borchert, Jager, and Steffen (2020)

19 provide evidence that banks which are insufficiently stabilized reduce their loan supply less than better stabilized banks, but their credit is more likely to flow to low quality firms. Finally, we test whether capital inflation is less pronounced in countries that have a national election in less than 12 months after the EBA capital exercise. The insignificant effect of CEB x Before Election in Column (5) shows that the time of elections is not related with capital inflation. While prior studies usually document an association between political interests on supervisory actions (Brown and Din, 2005; Bian, Haselmann, Kick, and Vig, 2017), we fail to find any evidence of this on our dependent variable. We argue that the decision of the capital exercise is taken at the European level rather than at the individual country level rendering a disconnect with local political interest.

In the final two columns we estimate the CEB dummy country characteristics interaction jointly.

The results stress that our results are not drive by national champions (a negative significant co- efficient), while difference in supervisory institutions and macro-economic difference across EBA countries are likely drivers of regulatory capital inflation.

[Table VIII about here]

IV. Robustness Checks

A. Placebo Treatment Periods

If CE banks would systematically differ from non-CE banks with respect to characteristics relevant for capital deductions, we would expect to see differential changes in the ratio of regulatory capital to book equity between CE banks and non-capital exercise banks also in other periods. To examine this possibility, we conduct two placebo tests with alternative treatment periods: (1) before; and (2) after the 2011 EBA capital exercise. Table IX shows the results of this placebo regression exercise. For convenience of comparison, the first column of Table IX replicates our baseline result in the second column of Table and shows the treatment effect of the capital exercise from 2010 to 2012. The second and third column then compare the changes in regulatory deductions between CE banks and non-CE banks for the placebo periods from 2008 to 2010 and from 2012 to 2014, respectively. While there is a strong effect for the period around the capital exercise, CE banks and non-CE banks exhibit no differential change in the ratio of regulatory capital to book

20 equity during the two placebo periods. This alleviates concerns that CE banks and non-CE banks differ systematically with respect to characteristics relevant for regulatory deductions.

[Table IX about here]

B. Matching Results

For our main analysis, we rely on conventional (OLS) regression methods. However, if the co- variate distributions differ substantially by treatment status, then conventional regression methods can be sensitive to minor changes in the specification because of their heavy reliance on extrapola- tion. One approach to address this problem is the use of matching estimators which have favorable robustness properties with respect to a variety of data configurations (Imbens, 2014). Thus, as a second robustness check, we estimate the treatment effect of the capital exercise on banks’ engage- ment in regulatory capital arbitrage using the bias-corrected Abadie and Imbens (2011) matching estimator. Specifically, we adopt four different matching strategies based on Gropp, Mosk, Ongena, and Wix (2019). The full sample matching strategy matches four non-CE banks to each CE bank based on the six matching covariates using the full sample of 48 CE banks and 144 non-CE banks.

The overlap sample matching strategy matches one non-CE bank to each CE bank based on asset size only in the sample of banks which are larger than the smallest CE bank and smaller than the largest non-CE bank. The within-country matching strategy the two smallest CE banks and the two largest non-CE banks around the selection threshold within each country around. Finally, the within-region matching strategy matches CE banks to non-CE banks around the selection thresh- old within the same region (GIIPS countries and non-GIIPS countries). Online Appendix 3, which replicates Table 5 of Gropp, Mosk, Ongena, and Wix (2019) shows that the different matching strategies reduce the differences between the treatment and control group banks. The third row in Table X provides the results of the matching exercises. Our results are robust and similar in magnitude to our regression results when using the full sample, within-country, and within-region matching strategies. While the treatment effect is not significant when employing the overlap sam- ple matching strategy, the coefficient is very similar in magnitude. The associated p-value is 0.11 and the results are therefore borderline statistically significant at the 10% level.

[Table X about here]

21 C. Robustness of the results after excluding individual countries and CE banks

One concern is that the results in the paper are driven by one specific country or bank. Figure

3 shows that the inflation of regulatory capital of capital exercise banks is more pronounced in Italy and Portugal. Table VI, however, shows that the treatment effect is still positive and significant, even after including CEB × Italy and CEB × Portugal interaction term in the specification. In

Online Appendix 4 we replicate the results of Table III column (3) and (4) for different subsample by excluding specific countries and CE bank. This exercise provides additional evidence that the result of the paper are not driven by one specific country or bank.

V. Conclusion

This paper studies the effect of higher capital requirements on the composition of regulatory capital. Exploiting the sudden increase in capital requirements during the 2011 EBA capital exer- cise, we find evidence that treated banks increase their regulatory capital, without a commensurate increase in book equity, thereby increasing their levels of regulatory capital without issuing eq- uity or retaining earnings. Although this increase in regulatory capital boosted their capital, we do not find that the capital exercise improved the market perception of the credit risk of bank debt of treated banks. This inflation of regulatory capital is more pronounced in countries with weaker supervisory standards and tightening credit standards, suggesting that national authorities forbear their banks to meet supranational regulatory requirements. The aim of the ECB’s Single

Supervisory Mechanism (SSM), introduced in 2014, was to ensure a level playing field and the equal treatment of all supervised institutions. Our results show that in our sample period, before the introduction of the SSM, even though the capital exercise banks were treated with the same intervention, discretionary powers on behalf of national supervisors resulted in a heterogeneous response of the treated banks.

22 REFERENCES

Abadie, Alberto, and Guido W. Imbens, 2011, Bias-corrected matching estimators for average

treatment effects, Journal of Business & Economic Statistics 29, 1–11.

Abbassi, Puriya, Rajkamal Iyer, Jos´e-LuisPeydr´o,and Paul E. Soto, 2020, Stressed banks? Evi-

dence from the largest-ever supervisory review, Deutsche Bundesbank Discussion Paper 26/2020.

Acharya, Viral V., Lea Borchert, Maximilian Jager, and Sascha Steffen, 2020, Kicking the can

down the road: Government interventions in the European banking sector, Review of Financial

Studies forthcoming.

Acharya, Viral V., Tim Eisert, Christian Eufinger, and Christian Hirsch, 2019, Whatever it takes:

The real effects of unconventional monetary policy, Review of Financial Studies 32, 3366–3411.

Acharya, Viral V., Robert Engle, and Diane Pierret, 2014, Testing macroprudential stress tests:

The risk of regulatory risk weights, Journal of Monetary Economics 65, 36–53.

Agarwal, Sumit, David Lucca, Amit Seru, and Francesco Trebbi, 2014, Inconsistent regulators:

Evidence from banking, Quarterly Journal of Economics 129, 889–938.

Angeloni, Ignazio, 2014, The SSM sails past the starting line: Seeking high-quality supervision and

level playing field, Speech London, 10 November 2014.

Atkins, Ralph, Patrick Jenkins, and Peter Spiegel, 2011, Europe’s banks face 9% capital rule,

Financial Times October 11, 2011.

Barth, James R., Gerard Caprio, and Ross Levine, 2013, Bank regulation and supervision in 180

countries from 1999 to 2011, NBER Working Paper 18733.

BCBS, 2015, The interplay of accounting and regulation and its impact on bank behaviour: Liter-

ature review, Working Paper 28.

Beck, Thorsten, Asli Demirg¨u¸c-Kunt, and Ross Levine, 2006, Bank supervision and corruption in

lending, Journal of Monetary Economics 53, 2131–2163.

23 Behn, Markus, Rainer Haselmann, and Vikrant Vig, 2016, The limits of model-based regulation,

ECB Working Paper Series 1928.

Bian, Bo, Rainer Haselmann, Thomas Kick, and Vikrant Vig, 2017, The political economy of bank

bailouts, Mimeo .

Bostandzic, Denefa, Felix Irresberger, Ragnar Juelsrud, and Gregor Wei, 2020, Do capital require-

ments make banks safer? Evidence from a quasi-natural experiment, Leeds University Business

School Working Paper 18-02.

Brown, Craig O., and I. Serdar Din, 2005, The politics of bank failures: Evidence from emerging

markets, Quarterly Journal of Economics 120, 1413–1444.

Brown, Craig O., and I. Serdar Din, 2011, Too many to fail? Evidence of regulatory forbearance

when the banking sector is weak, Review of Financial Studies 24, 1378–1405.

Bruno, Brunella, and Elena Carletti, 2019, The next SSM term: Supervisory challenges ahead, In-

Depth Analysis requested by the European Parliament’s Committee on Economic and Monetary

Affairs .

Calzolari, Giacomo, Jean-Edouard Colliard, and Gyongyi L´or´anth, 2019, Multinational banks and

supranational supervision, Review of Financial Studies 32, 2997–3035.

Carletti, Elena, Giovanni DellAriccia, and Robert Marquez, 2020, Supervisory incentives in a bank-

ing union, Management Science forthcoming. de Weert, Frans, 2012, Bank and insurance capital management, West Sussex (UK): John Wiley

& Sons Ltd .

DellAriccia, Giovanni, and Robert Marquez, 2006, Competition among regulators and credit market

integration, Journal of Financial Economics 79, 401–430.

EBA, 2011a, Eba recommendation on the creation and supervisory oversight of temporary capital

buffers to restore market confidence (EBA/REC/2011/1), December 8, 2011 .

EBA, 2011b, EU-wide stress test: Methodological note, March 18, 2011 .

24 EBA, 2012, Final report on the implementation of capital plans following the EBAs 2011 recom-

mendation on the creation of temporary capital buffers to restore market confidence, october 3,

2012 October 2012.

Gerding, Erik F., 2016, The dialectics of bank capital: Regulation and regulatory capital arbitrage,

Washburn Law Journal 55.

Goodhart, Charles, 2012, Funding arrangements and burden sharing in banking resolution, In:

Banking Union for Europe: Risks and Challenges Thorsten Beck (ed.).

Gropp, Reint, Thomas Mosk, Steven Ongena, and Carlo Wix, 2019, Bank response to higher

capital requirements: Evidence from a quasi-natural experiment, Review of Financial Studies

32, 266–299.

Gropp, Reint, Steven Ongena, J¨orgRocholl, and Vahid Saadi, 2018, The cleansing effect of banking

crises, Mimeo .

Haldane, Andrew G., 2012, The dog and the frisbee, Speech given at the Federal Reserve Bank of

Kansas City’s 36th Economic Policy “The Changing Policy Landscape” Jackson Hole, Wyoming.

Haldane, Andrew G., 2013, Constraining discretion in bank regulation, Speech given at the Federal

Reserve Bank of Atlanta Conference on “Maintaining financial stability: holding a tiger by the

tail(s)”, Federal Reserve Bank of Atlanta, Atlanta, 9 April 2013. .

Halstrick, Philipp, and Andreas Framke, 2011, Europe hits banks with tougher capital test, Reuters

October 11, 2011.

Haselmann, Rainer, Thomas Kick, Shikhar Singla, and Vikrant Vig, 2019, Capital regulation,

market-making, and liquidity, Mimeo .

Haselmann, Rainer, Shikhar Singla, and Vikrant Vig, 2018, Political economy of financial regula-

tion, Mimeo .

Imbens, Guido W., 2014, Matching methods in practice: Three examples, NBER Working Paper

Series 19959.

25 Journal, Wall Street, 2014, ECB should stop southern european games with bank capital October

13, 2014.

Kisin, Roni, and Asaf Manela, 2016, The shadow cost of bank capital requirements, Review of

Financial Studies 29, 1780–1820.

Koch, Stefan, Roland Schneider, Sebastian Schneider, and Gerhard Schroeck, 2017, Bringing Basel

IV into focus: How banks can mitigate 120 billion euro in capital requirements and avoid an

ROE haircut., McKinsey & Company Risk Insights November 2017.

Laeven, Luc, and Ross Levine, 2009, Bank governance, regulation, and risk taking, Journal of

Financial Economics 93, 259–275.

Lubberink, Martien, 2014, A primer on regulatory bank capital adjustments, Mimeo .

Maddaloni, Angela, and Alessandro D. Scopelliti, 2019, Rules and discretion(s) in prudential regu-

lation and supervision: evidence from EU banks in the run-up to the crisis, ECB Working Paper

Series 2284.

Martynova, Natalya, Enrico C. Perotti, and Javier Suarez, 2019, Bank capital forbearance, CEPR

Discussion Paper 13617.

Nouy, Dani`ele,2017, Gaming the rules or ruling the game? How to deal with regulatory arbitrage,

Speech at the 33rd SUERF Colloquium, Helsinki, 15 September 2017 .

Plosser, Matthew C., and Jo˜ao A. C. Santos, 2018, Banks’ incentives and inconsistent risk models,

Review of Financial Studies 31, 2080–2112.

Schoenmaker, Dirk, 2012, Banking supervision and resolution: The European dimension, DSF

Policy Paper Series 19.

Walther, Ansgar, and Lucy White, 2016, Rules versus discretion in bank resolution, Mimeo .

26 Figure 1. The Ratio of Regulatory Capital to Book Equity Over Time. This figure shows the evolution of the mean of the ratio of regulatory core tier 1 capital to book total equity over time for 48 capital exercise banks (CEB) (solid blue line) and 143 non-capital exercise banks (Non-CEB) (dashed red line), normalized to the value of 1 in 2010. The dotted lines indicate the 95 percent confidence intervals. The two dashed vertical lines mark 2010 and 2012, the years immediately before and after the capital exercise.

27 Figure 2. CDS Spreads of CE and Non-CE Banks Over Time. This figure shows the estimated coefficients and confidence intervals of the specification: Senior CDS spread = β1CEB × Y ear − Quarter + CY × YearQuarter FE + Bank FE + ε. Standard errors are clustered at the bank level.

28 Figure 3. Coefficients Plot Capital Exercise Bank x Country Interactions This figure plots the estimated coefficients and confidence intervals of the specification P k k ∆(CET1 Capital /Book equity)i,2012−2010 = α + β × CEBi × CYc + k θ Xi = εi, where the dependent variable is the change in the CET1 capital / book equity ratio from 2010 to 2012. The variable CEBi takes on the value of 1 for banks selected into the capital exercise, and 0 otherwise, and CYc are bank headquarter country dummies. In addition, we control for the following bank characteristics as of 2010: log total assets, CET1 ratio, customer loans as a share of total assets, net interest income as a share of total operating revenue, depository funding as a share of total assets, and net income over total assets. Standard errors are clustered at the country level. Online Appendix 4 shows that our results are not driven by one specific country or bank.

29 Table I Descriptive Statistics

Table 1 provides the summary statistics for the EBA Banks and Non-EBA Banks in our sample (48 capital exercise banks and 144 noncapital exercise banks). The paper tests for differences in means using Welchs t-test. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

CEB Non-CEB ∆

Log Total Assets 5.26 5.21 1.46 2.00 2.21 1.58 3.27∗∗∗ CT1 Capital Ratio (in %) 9.86 9.21 3.12 11.43 10.64 4.99 −1.57∗∗∗ Deposits / Total Assets (in %) 40.93 40.47 15.59 55.46 56.25 20.49 −14.53∗∗∗ Loans / Total Assets (in %) 56.73 60.03 15.65 66.53 70.59 17.72 −9.80∗∗∗ NII / Operating Revenue (in %) 60.42 57.94 14.86 67.65 68.97 22.59 −7.23∗∗∗ Net Income / Total Assets (in %) 0.39 0.40 0.43 0.41 0.29 0.54 −0.02∗∗∗

30 Table II Descriptive Statistics

Table 2 presents the components of CT1 capital (total book equity, including common share capital, and retained earnings; regulatory deductions, including intangible assets), and regulatory capital to total equity ratio for capital exercise banks (CEB) and non-capital exercise banks (Non-CEB) for the years 2010 and 2012, mean percentage change for these variables, as well as differences- in-differences between CEB and Non-CEB banks. The paper tests for differences in means using Welch’s t-test and for difference-in-differences using a regression framework. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

CEB Non-CEB DID

Variable 2010 2012 ∆ 2010 2012 ∆

Common Share Capital 9.88 10.72 8.5% 0.55 0.64 16.6% 0.59∗∗∗ Retained Earnings 12.19 12.07 -1.0% 0.89 0.99 11.2% 0.14

Book Equity 23.08 24.25 5.1% 1.39 1.59 14.4% 0.85∗∗∗

Regulatory Deductions 7.34 6.03 -17.9% 0.25 0.28 10.7% −1.31∗∗∗

CET1 Capital 15.74 18.22 15.8% 1.13 1.30 15.2% 2.16∗∗∗

CET1 Capital / Book Equity 0.78 0.85 7.3 pp. 0.88 0.88 -0.7 pp. 0.09∗∗∗

31 Table III Changes in the Ratio of Regulatory Capital to Book Equity

Table 3 presents the estimation results of the following specification:

k X k ∆(Regulatory Capital/Book Equity)i,2010−2012 = α + β × CEBi + θ Xi + γc + i k where the dependent variable is the change in the CET1 capital / book equity ratio from 2010 to 2012. The variable CEBi takes on the value of 1 for banks selected into the capital exercise, and 0 otherwise. In addition, we control for the following bank characteristics as of 2010: log total assets, CET1 ratio, customer loans as a share of total assets, net interest income as a share of total operating revenue, depository funding as a share of total assets, and net income over total assets. Standard errors are clustered at the country level. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Dependent variable ∆(Regulatory Capital/Book Equity)2010−2012

Capital Exercise Bank (CEB) 0.093∗∗ 0.074∗ 0.100∗∗ 0.394∗∗∗ 0.439∗∗ (0.034) (0.041) (0.045) (0.101) (0.156)

CEB × CT1 Ratio 2010 −0.032∗∗∗ −0.032∗∗∗ (0.008) (0.008)

CEB × Log Total Assets 2010 -0.008 (0.013)

Log Total Assets 2010 0.002 -0.003 0.001 0.002 (0.006) (0.008) (0.008) (0.007) CT1 Ratio 2010 −0.004∗∗ -0.001 -0.000 -0.000 (0.002) (0.001) (0.001) (0.001) (Total Deposits/TA) 2010 -0.000 -0.000 -0.000 -0.000 (0.000) (0.000) (0.000) (0.000) (Customer Loans/TA) 2010 0.000 0.001 0.001 0.000 (0.001) (0.001) (0.001) (0.001) (Net Int. Inc./Op.Rev) 2010 -0.001 -0.000 -0.000 -0.000 (0.000) (0.001) (0.001) (0.001) (Net Income/TA) 2010 0.021 0.022 0.028∗ 0.028∗ (0.014) (0.015) (0.014) (0.014)

Country FE Yes Yes Yes

N 191 191 191 191 191 R2 0.129 0.176 0.326 0.401 0.402

32 Table IV Changes in CT1 and TE/RWA Ratios

Table 4 presents the estimation results of the following specification:

k X k ∆Yi,2010−2012 = α + β × CEBi + θ Xi + γc + i k where the dependent variable ∆Yi,2010−2012 is the change in regulatory CT1 capital to risk-weighted assets ratio from 2010 to 2012 on the first column and the change in the total equity to risk-weighted assets ratio in the second column. The variable CEBi takes on the value of 1 for banks selected into the capital exercise, and 0 otherwise. In addition, we control for the following bank characteristics as of 2010: log total assets, CET1 ratio, customer loans as a share of total assets, net interest income as a share of total operating revenue, depository funding as a share of total assets, and net income over total assets. Standard errors are clustered at the country level. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

∆(CT1 Capital/RWA)2010−2012 ∆(TE/RWA)2010−2012

CEB 1.003 3.213∗∗ 0.300 -0.156 (0.781) (1.343) (0.742) (1.450)

CEB × CT1 Ratio 2010 −0.238∗ 0.049 (0.122) (0.163)

Bank-Level Controls Yes Yes Yes Yes Country FE Yes Yes Yes Yes

N 191 191 190 190 R2 0.306 0.322 0.207 0.208

33 Table V The EBA Capital Exercise and Bank Risk

Table 5 presents the estimation results of the following specification:

k X k ∆Yi,2010−2012 = α + β × CEBi + θ Xi + γc + i k where the dependent variable ∆Yi,2010−2012 is the three-day CE announcement CDS returns around October 26 in columns (1)-(4), 2011 and the senior and junior CDS spread change between October 2011 and June 2012 in column (5) and (6) respectively. The variable takes on the value of 1 for banks selected into the capital exercise, and 0 otherwise. The dependent variable of Panel B is the z-score change of the bank ([ROA + Equity/ Total assets] / σ(ROA)). In addition, we control for the following bank characteristics as of 2010: log total assets, CET1 ratio, customer loans as a share of total assets, net interest income as a share of total operating revenue, depository funding as a share of total assets, and net income over total assets. Standard errors are clustered at the country level. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Panel A. CDS Spreads.

CDS Returns: CE Announcement ∆CDS: Oct11-Jun12

∆CDSSEN ∆CDSSUB

CEB −8.444∗ 6.408 4.903 -19.414 -38.983 62.89 (3.920) (6.776) (6.898) (41.034) (130.329) (240.514)

CEB × CT1 Ratio 2010 2.913 15.959 -2.469 (4.878) (17.889) (24.249)

Controls YES YES YES YES YES Country FE YES YES YES YES

N 53 53 53 53 51 41 R2 0.034 0.276 0.443 0.457 0.678 0.679

Panel B. Z-Scores.

∆Z-Score2010−2012

CEB -2.04* -1.85 -1.25 -8.86*** (1.17) (1.12) (2.02) (2.63) CEB × CT1 Ratio 2010 0.82*** (0.22)

Controls YES YES YES Country FE YES YES

N 190 190 190 190 R2 0.02 0.05 0.20 0.20 34 Table VI Regulatory Capital Inflation Across Countries

Table 6 presents the estimation results of the following specification:

k X k ∆(Regulatory Capital/Book Equity)i,2010−2012 = α + β × CEBi × CYc + θ Xi + γc + i k where the dependent variable is the change in the CET1 capital / book equity ratio from 2010 to 2012. The variable CEBi takes on the value of 1 for banks selected into the capital exercise, and 0 otherwise. CYc are country dummies taking the value of one if bank i is headquartered in country c, and 0 otherwise. In addition, we control for the following bank characteristics as of 2010: log total assets, CET1 ratio, customer loans as a share of total assets, net interest income as a share of total operating revenue, depository funding as a share of total assets, and net income over total assets. Standard errors are clustered at the country level. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. Online Appendix 4 shows that our results are not driven by one specific country or bank.

Dependent variable ∆(Regulatory Capital/Book Equity)2010−2012

CEB 0.050∗∗ 0.036 0.218∗∗∗ 0.232∗∗∗ (0.021) (0.025) (0.065) (0.071) CEB x CT1 Ratio 2010 −0.018∗∗∗ −0.019∗∗∗ (0.006) (0.005) CEB x Log Total Assets 2010 -0.002 (0.006)

CEB x Italy 0.187∗∗∗ 0.199∗∗∗ 0.144∗∗∗ 0.145∗∗∗ (0.021) (0.019) (0.026) (0.026) CEB x Portugal 0.450∗∗∗ 0.441∗∗∗ 0.398∗∗∗ 0.396∗∗∗ (0.021) (0.015) (0.016) (0.017)

Controls Yes Yes Yes Country FE Yes Yes Yes Yes

N 191 191 191 191 R2 0.458 0.471 0.491 0.491

35 Table VII Regulatory Forbearance of National Authorities: Anecdotal Evidence

Table 7 present an overview of the anecdotal evidence on measures implemented by national au- thorities around the EBA capital exercise to increase the capitalization of domestic banks. The table provides for each measure the country, a short description, if applicable a reference to the national law or decree introducing the measure and the date. At the bottom, the table include two examples of regulatory forbearance after the EBA capital exercise. Online appendix 2 provides a detailed description about the working of the specific policies.

Country Description Reference Date

Panel A. Around the EBA capital exercise.

Germany Recapitalisation of German Nord LB Law 16/3567 Apr 14, 2011 Italy New instruments eligible as CT1 capital Decree 95 Jul 6, 2012 Italy Hybrid capital injection Banca MPS Jul 6, 2012 Italy Tax treatment of goodwill Decree 98/2011 Jul 17, 2011 Italy Conversion of DTAs to tax credits Decree 201 Dec 22,2011 Portugal Amendment recapitalization scheme Law 4/2012 Jan 11, 2012 Portugal Hybrid capital injection Banco BPI Jun 29, 2012 Portugal Hybrid capital injection Banco Comercial Jun 29, 2012 Portugal Hybrid capital injection CGD Jun 28, 2012 Slovenia Hybrid capital injection Nova Ljublj. Banka May 16, 2012 Slovenia Hybrid capital injection Nova Kreditna Banka Dec 5, 2012

Panel B. After the EBA capital exercise.

Spain Conversion of DTAs to tax credits Decree 14/2013 Nov 29, 2013 Portugal Conversion of DTAs to tax credits Law 61/2014 Aug 26, 2014

36 Table VIII Table 8: Regulatory Capital Inflation and Country Characteristics

Table 8 presents the estimation results of the following specification:

k X k ∆(Regulatory Capital/Book Equity)i,2010−2012 = α + β × CEBi × CY Measurec + θ Xi + γc + i k where the dependent variable is the change in the CET1 capital / book equity ratio from 2010 to 2012. The variable CEB i takes on the value of 1 for banks selected into the capital exercise, and 0 otherwise. CY Measurec are the following country specific measures: National champion takes the value of 1 if bank i is the largest bank in country c, and 0 otherwise; Official Supervisory Power measures the degree to which national supervisory agencies have the authority to take specific actions (Barth et al., 2013). CA Balance is the current account balance to GDP; Credit Standards measures the share of banks expecting to tighten their credit standards in the next quarter. Before Election takes the value of 1 if the capital exercise takes place 6 months before the next national congressional election, and 0 otherwise. In addition, we control for the following bank characteristics as of 2010: log total assets, CET1 ratio, customer loans to total assets, net interest income to total operating revenue, depository funding to total assets, and net income to total assets. Standard errors are clustered at the country level. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. 37

Dependent variable ∆(Regulatory Capital/Book Equity)2010−2012

Capital Exercise Banks (CEB) 0.111∗∗ 0.093∗∗ 0.104∗∗ 0.014 0.109∗∗ 0.051∗ 0.191∗∗ (0.050) (0.039) (0.040) (0.027) (0.051) (0.027) (0.078) CEB × CT1 Ratio 2010 0.015∗ (0.008) National champion -0.058 −0.073∗∗ −0.064∗∗ (0.037) (0.030) (0.028) CEB x Official Supervisory Power 0.062∗∗ 0.040∗∗∗ 0.030∗∗ (0.024) (0.009) (0.011) CEB x CA Balance -0.011 −0.006∗∗ -0.004 (0.007) (0.003) (0.003) CEB x Credit Standards 0.004∗∗∗ 0.003∗∗∗ 0.003∗∗ (0.001) (0.001) (0.001) CEB x Before election -0.057 -0.050 -0.043 (0.058) (0.040) (0.037)

Controls Yes Yes Yes Yes Yes Yes Yes Country FE Yes Yes Yes Yes Yes Yes Yes N 191 184 191 181 191 181 181 R2 0.339 0.376 0.368 0.428 0.330 0.470 0.481 Table IX Placebo Treatment Periods

Table 9 presents the estimation results of the change in the ratio of regulatory core tier 1 (CT1) capital to total equity around the 2011 EBA capital exercise and around two placebo treatment periods:

k X k ∆(Regulatory Capital/Book Equity)i = α + β × CEBi + θ Xi + i k where the dependent variable is the change in the ratio of CT1 capital to total equity from 2010 to 2012, 2008 to 2010, and 2012 to 2014, respectively. The variable CEB i takes on the value of 1 if the bank is an capital exercise bank, and 0 otherwise. In addition, we control for the following k bank characteristics Xi as of 2010: log total assets, CT1 ratio, deposits over total assets, loans over total assets, net interest income over total operating revenue, and net income over total assets. Robust standard errors are given in parentheses. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

∆2010−2012 ∆2008−2010 ∆2012−2014 Treatment Placebo Placebo

Capital Exercise Bank (CEB) 0.07∗∗ 0.00 0.00 (0.03) (0.02) (0.02)

Bank-Level Controls Yes Yes Yes Country FE Yes Yes Yes

N 191 178 157 R2 0.18 0.02 0.08

38 Table X Matching Results

Table 10 presents the estimation results of the change in the ratio of regulatory core tier 1 (CT1) capital to total equity around the 2011 EBA capital exercise using the matching methodology based on Gropp, Mosk, Ongena, and Wix (2019) described in Section IV. In each column, the first row contains the difference in the outcome variable for capital exercise banks (CEB) between the period before (2010) and the after (2012) the capital exercise; the second row contains the difference in the outcome variable for matched control group (control) banks over the same period. The paper tests for differences-in-means using Welch’s two-sample t-test. The third row contains the estimate for the average treatment effect on the treated (ATT) based on the bias-corrected Abadie and Imbens (2011) matching estimator. Column 1 presents the results for the full sample matching strategy, Column 2 the results for the overlap matching strategy, Column 3 the results for the within-country matching strategy, and Column 4 the results for the within-region matching strategy. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

∆(Regulatory Capital/Book Equity)2010−2012

Full Overlap Within Country Within Region

CEB: After - Before 0.09∗∗∗ 0.09∗∗∗ 0.09∗∗ 0.08∗∗ Control: After - Before 0.01 0.00 −0.01 −0.02 Bias-Corrected ATT 0.08∗∗ 0.09 0.10∗∗∗ 0.09∗∗ Number of Observations 48 48 25 26

39 Table XI Changes in the Ratio of Intangible Assets to Book Equity

Table 11 presents the estimation results of the following specification:

k X k ∆(Intangible Assets/Book Equity)i,2010−2012 = α + β × CEBi + θ Xi + γc + i k where the dependent variable is the change in the intangible assets / book equity ratio from 2010 to 2012. The variable CEBi takes on the value of 1 for banks selected into the capital exercise, and 0 otherwise. In addition, we control for the following bank characteristics as of 2010: log total assets, CET1 ratio, customer loans as a share of total assets, net interest income as a share of total operating revenue, depository funding as a share of total assets, and net income over total assets. Standard errors are clustered at the country level. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Dependent variable ∆(Regulatory Capital/Book Equity)2010−2012

Capital Exercise Bank (CEB) -0.05 0.05 -0.07 (0.04) (0.06) (0.17) CEB × CT1 Ratio 2010 0.01 (0.02)

CEB × Italy −1.41∗∗∗ −1.41∗∗∗ −1.37∗∗∗ (0.04) (0.05) (0.08) CEB × Portugal -0.02 −0.07∗ -0.04 (0.04) (0.04) (0.05)

Controls Yes Yes Country FE Yes Yes Yes

N 189 189 189 R2 0.74 0.76 0.76

40 Online appendix for:

“Supranational rules, national discretion:

Increasing versus inflating regulatory bank capital?”

1

Online Appendix 1: Book Equity, Capital Deductions, and Regulatory Capital This table illustrates the calculation of regulatory common equity tier 1 (CET1) capital via the application of capital deductions to book equity (Common Equity Tier 1 capital before regulatory deductions) according to the social Basel disclosure template (BCBS, 2011).

Item Common share capital plus related stock surplus + Retained earnings + Accumulated other comprehensive income (AOCI) + Directly issued capital subject to phase out from CET1 + Common share capital issued by subsidiaries and held by third parties = Common Equity Tier 1 capital before regulatory deductions (Book Equity) - Prudential valuation deductions - Goodwill (net of related tax liability) - Intangibles other than mortgage-servicing rights (net of related tax liability) - Deferred tax assets - Cash-flow hedge reserve - Shortfall of provisions to expected losses - Securitisation gain on sale - Gains and losses due to changes in own credit risk on fair valued liabilities - Defined-benefit pension fund net assets - Investments in own shares - Reciprocal cross-holdings in common equity - Investments in the capital of financial institutions [. . . ] (above 10% threshold) - Significant investments in financial institutions [. . . ] (above 10% threshold) - Mortgage servicing rights - Deferred tax assets arising from temporary differences - Amount exceeding the 15% threshold of which: Significant investments in the common stock of financials of which: mortgage servicing rights of which: deferred tax assets arising from temporary differences - National specific regulatory deductions - Deductions applied to CET1 due to insufficient AT1 and T2 = Common Equity Tier 1 capital (CET1) (Regulatory Capital)

2

Online Appendix 2: Anecdotic evidence of forbearance by national authorities

We hand collected from annual reports, pillar 3 reports and EC state aid cases anecdotic evidence of forbearance by national authorities. Below we discuss three different methods how national authorities could forbear their banks.

1. Recapitalizations

One group of forbearance measures are ad hoc recapitalizations, such as the issuance of core tier 1 eligible hybrid securities, underwritten by the state. Germany (NORD/LB), Italy (Banca Monte dei Paschi di Siena), Portugal (Banco BPI, Banco Comercial Portugues, Caixa Geral de Depositos) and Slovenia (NLB, Nova Kreditna Banka Maribor) provided these instruments and often explicitly referred to the EBA capital as an important reason to justify the state aid measure. For example, the state aid application of Banco Comercial Português notes “On 8 December 2011,the EBA published a Recommendation related to banks’ recapitalisation needs. … In the light of the foregoing, the BCP Group needed to raise substantial additional capital by 30 June 2012.”.1 Capital assistance measures need a legal basis and thus approval from the parliament. The legal basis of the assistance to Banco Comercial Português is Law No 4/2012 which has been enacted on January 11th, 2012.2 Besides direct capital support measures, several countries implemented around the EBA capital exercise regulation aiming to increase the regulatory capital of banks. We discuss several of these measures below.

2. Conversion of deferred tax assets into tax credits

Deferred tax assets are items on the bank’s balance sheet that may be used to reduce taxable income in the future. Capital Requirements Regulation No. 575/2013 requires banks to deduct deferred tax assets from core tier 1 capital, because their dependence on future income, which depresses bank capital. The Italian government, however, enacted decree-law no. 201 on 6 December 20113, allowing banks to convert their deferred tax assets into tax credits, which do not have to be deducted because they are guaranteed by the government. Similar regulatory changes have been implemented in the subsequent years by Spain, Portugal and Greece, resulting in a European Commission investigation whether these measures constitute illegal state aid.

1 State aid case SA.34724. Link: https://ec.europa.eu/competition/elojade/isef/case_details.cfm?proc_code=3_SA_34724 2 https://dre.pt/application/conteudo/477151 3 https://www.normattiva.it/uri-res/N2Ls?urn:nir:stato:decreto.legge:2011;98 3

3. Deductions of tax treatment of goodwill impairments

For prudential purposes, the book value of intangible assets and goodwill is deducted from CT1 capital because they may become worthless when banks go bankrupt. However, these assets are typically large in magnitude relative to a bank’s regulatory capital (24% for capital exercise banks in our sample as of 2010). Banks could reduce intangible asset deductions by impairing these assets. From an accounting perspective, this results in a decrease of the book value of goodwill (equal to the amount of the impairment charge) but also reduces net income (by that same amount). The impact on CT1 capital is initially capital neutral because, on the one hand, the impairment charge reduces CT1-deductions but, on the other hand, it also reduces profits which is a component of CT1 capital. One way national authorities could support banks is to allow tax-deductible intangible asset impairments, resulting in a net increase in regulatory capital after an impairment.4 In 2011, the Italian government enacted Law Decree no. 98/20115, allowing banks to boost their regulatory capital by impairing intangible assets. For example, the measure resulted in a 40 basis points increase in the regulatory capital of Banca Monte dei Paschi di Siena's.6

4 Assume a bank impairs goodwill for an amount of 1,000 EUR which is tax-deductible. The tax rate is 20%. This will result in a reduction of 1,000 EUR in capital deductions (thus increasing CT1 capital) whereas net income will decrease with 800 EUR (1,000*(1-0,2)) rendering an net positive effect on CT1 capital of 200 EUR due to a permanent difference between book and tax income. 5 https://www.normattiva.it/uri-res/N2Ls?urn:nir:stato:decreto.legge:2011-07-06;98!vig= 6 https://www.gruppomps.it/en/media-and-news/press-releases/banca-monte-dei-paschi-siena-tier-1-up-to-8-8- per-cent.html 4

Online Appendix 3: Pre-Treatment Characteristics of Banks This table provides pre-treatment summary statistics on Capital Exercise banks, Non-Capital Exercise banks and control group banks (mean comparison). The paper tests for differences in means using the Student’s t-test. (*, **, and *** indicate statistical significance at the 10%, 5%, and 1% level respectively). Panel A compares the mean values of the 48 Capital Exercise banks and 145 Non-Capital Exercise banks in the unmatched sample. Panel B compares the 48 Capital Exercise banks to the full sample of matched control group banks based on the Mahalanobis matching estimator. Panels C to E compare Capital Exercise banks to the sample of matched control group banks using the overlap, within country and within region matching strategy respectively. Table I lists the matching covariates for each matching strategy.

Core Total Customer Total Tier 1 Deposits / Loans / Net Interest Income / Net Income / # Banks Assets Ratio Total Assets Total Assets Operating Revenue Total Assets Panel A: Unmatched Sample Capital Exercise Banks 48 454.31 9.86 40.93 56.73 60.42 0.39 Non-Capital Exercise Banks 144 24.43 11.41 55.54 66.62 67.69 0.41 Δ 429.87 *** -1.55 ** -14.61 *** -9.89 *** -7.27 ** -0.02

Panel B: Full Sample Capital Exercise Banks 48 454.31 9.86 40.93 56.73 60.42 0.39 Matched Control Group 48 107.14 10.30 47.89 64.80 64.62 0.41 Δ 347.17 *** -0.44 -6.95 *** -8.07 *** -4.19 *** -0.02

Panel C: Overlap Sample Capital Exercise Banks 36 161.32 9.98 41.97 59.78 61.95 0.40 Matched Control Group 36 156.10 10.95 53.80 57.06 71.89 0.38 Δ 5.22 -0.96 -11.83 ** 2.72 -9.94 0.02

Panel D: Within Country Capital Exercise Banks 25 320.88 9.96 43.51 59.08 58.80 0.40 Matched Control Group 25 80.92 10.80 43.21 61.72 71.22 0.42 Δ 239.96 *** -0.84 * 0.31 -2.64 -12.42 * -0.02

Panel E: Within Region Capital Exercise Banks 26 310.18 10.01 44.85 59.77 58.99 0.45 Matched Control Group 26 108.31 9.73 48.26 65.69 61.31 0.45 Δ 201.87 ** 0.28 -3.41 -5.92 *** -2.31 0.00

5

Online Appendix 4: Robustness of the results after excluding individual countries and banks Online Appendix 4 replicates the results of Table 3 column (3) and (4) and subsequently estimates equation (1) for subsamples excluding one specific country (Panel A) or CE bank (Panel B). Panel A: Robustness of the results after excluding individual countries CEB Controls CY FE Obs. R-sq CEB CEB x CT1 Ratio 2010 Controls CY FE Obs. R-sq All CY's 0.100** YES YES 191 0.326 0.394*** -0.032*** YES YES 191 0.401 Excluding: Austria 0.103** YES YES 184 0.335 0.401*** -0.032*** YES YES 184 0.413 Belgium 0.098** YES YES 189 0.326 0.393*** -0.032*** YES YES 189 0.401 Cyprus 0.100** YES YES 190 0.316 0.394*** -0.032*** YES YES 190 0.392 Denmark 0.106** YES YES 167 0.339 0.437*** -0.037*** YES YES 167 0.418 Finland 0.102** YES YES 188 0.330 0.394*** -0.032*** YES YES 188 0.403 France 0.098* YES YES 185 0.322 0.396*** -0.032*** YES YES 185 0.398 Germany 0.098* YES YES 151 0.361 0.416*** -0.034*** YES YES 151 0.436 Greece 0.100** YES YES 191 0.326 0.394*** -0.032*** YES YES 191 0.401 Hungary 0.100** YES YES 190 0.326 0.394*** -0.032*** YES YES 190 0.401 Ireland 0.100** YES YES 190 0.324 0.394*** -0.032*** YES YES 190 0.399 Italy 0.083 YES YES 162 0.338 0.325** -0.025** YES YES 162 0.388 Luxembourg 0.104** YES YES 189 0.328 0.395*** -0.032*** YES YES 189 0.401 Malta 0.101** YES YES 189 0.327 0.394*** -0.032*** YES YES 189 0.401 Netherlands 0.108** YES YES 186 0.336 0.415*** -0.033*** YES YES 186 0.414 Norway 0.108** YES YES 167 0.329 0.402*** -0.032*** YES YES 167 0.406 Poland 0.106** YES YES 188 0.336 0.394*** -0.031*** YES YES 188 0.408 Portugal 0.064* YES YES 185 0.264 0.321*** -0.027*** YES YES 185 0.344 Slovenia 0.101** YES YES 188 0.326 0.415*** -0.034*** YES YES 188 0.406 Spain 0.112** YES YES 181 0.337 0.408*** -0.032*** YES YES 181 0.412 Sweden 0.103** YES YES 184 0.339 0.395*** -0.032*** YES YES 184 0.403 United Kingdom 0.097* YES YES 176 0.327 0.387*** -0.031*** YES YES 176 0.401

6

Panel B: Robustness of the results after excluding individual CE banks Country CEB Controls CY FE Obs. R-sq CEB CEB x CT1 Ratio 2010 Controls CY FE Obs. R-sq All banks 0.100** YES YES 191 0.326 0.394*** -0.032*** YES YES 191 0.401 Excluding: BNP Paribas France 0.100** YES YES 190 0.327 0.395*** -0.032*** YES YES 190 0.401 Monte dei Paschi di Siena Italy 0.081* YES YES 190 0.325 0.319*** -0.025*** YES YES 190 0.380 Banco BPI Portugal 0.086** YES YES 190 0.283 0.393*** -0.033*** YES YES 190 0.380 BBVA Spain 0.101** YES YES 190 0.328 0.395*** -0.032*** YES YES 190 0.402 Banco Comercial Portugues Portugal 0.092** YES YES 190 0.282 0.344*** -0.027*** YES YES 190 0.344 Banco Popolare Italy 0.099** YES YES 190 0.320 0.416*** -0.034*** YES YES 190 0.400 Banco Popular Espanol Spain 0.107** YES YES 190 0.333 0.400*** -0.032*** YES YES 190 0.407 Banco Santander Spain 0.100** YES YES 190 0.326 0.397*** -0.032*** YES YES 190 0.402 Bank of Valletta Malta 0.101** YES YES 190 0.327 0.394*** -0.032*** YES YES 190 0.401 Banque et Caisse d'Epargne Luxembourg 0.104** YES YES 190 0.328 0.395*** -0.032*** YES YES 190 0.401 Barclays United Kingdom 0.100** YES YES 190 0.327 0.393*** -0.032*** YES YES 190 0.401 Caixa Geral de Depositos Portugal 0.100** YES YES 190 0.316 0.394*** -0.032*** YES YES 190 0.392 Credit Agricole Group France 0.100** YES YES 190 0.325 0.395*** -0.032*** YES YES 190 0.401 DNB Bank Norway 0.100** YES YES 190 0.327 0.396*** -0.032*** YES YES 190 0.402 Danske Bank Denmark 0.099** YES YES 190 0.326 0.394*** -0.032*** YES YES 190 0.401 DekaBank Germany 0.100** YES YES 190 0.324 0.397*** -0.032*** YES YES 190 0.400 Deutsche Bank Germany 0.100** YES YES 190 0.324 0.395*** -0.032*** YES YES 190 0.399 DZ bank Germany 0.099** YES YES 190 0.322 0.395*** -0.032*** YES YES 190 0.397 Erste Group Bank Austria 0.099** YES YES 190 0.326 0.394*** -0.032*** YES YES 190 0.400 Espirito Santo Portugal 0.104** YES YES 190 0.344 0.419*** -0.034*** YES YES 190 0.428 La Caixa Spain 0.099** YES YES 190 0.325 0.395*** -0.032*** YES YES 190 0.400 Groupe BPCE France 0.099** YES YES 190 0.324 0.394*** -0.032*** YES YES 190 0.399

7

Panel B: Robustness of the results after excluding individual CE banks (Continued) Country CEB Controls CY FE Obs. R-sq CEB CEB x CT1 Ratio 2010 Controls CY FE Obs. R-sq HSBC United Kingdom 0.100** YES YES 190 0.327 0.394*** -0.032*** YES YES 190 0.401 Intesa Sanpaolo Italy 0.098** YES YES 190 0.319 0.393*** -0.032*** YES YES 190 0.393 Jyske Bank Denmark 0.101** YES YES 190 0.327 0.399*** -0.032*** YES YES 190 0.402 Landesbank Berlin Germany 0.099** YES YES 190 0.324 0.398*** -0.032*** YES YES 190 0.399 Landesbank Hessen-Thuringen Germany 0.103** YES YES 190 0.332 0.413*** -0.033*** YES YES 190 0.413 Lloyds Banking Group United Kingdom 0.101** YES YES 190 0.328 0.394*** -0.032*** YES YES 190 0.402 NORD/LB Germany 0.094* YES YES 190 0.320 0.382*** -0.031*** YES YES 190 0.382 Nordea Bank Sweden 0.099** YES YES 190 0.326 0.397*** -0.032*** YES YES 190 0.402 Nova Kreditna banka Maribor Slovenia 0.099** YES YES 190 0.326 0.394*** -0.032*** YES YES 190 0.400 Nova Ljubljanska Banka Slovenia 0.100** YES YES 190 0.327 0.407*** -0.033*** YES YES 190 0.405 Nykredit Realkredit Denmark 0.101** YES YES 190 0.329 0.420*** -0.035*** YES YES 190 0.405 OP Financial Group Finland 0.103** YES YES 190 0.328 0.393*** -0.032*** YES YES 190 0.400 OTP Bank Nyrt. Hungary 0.100** YES YES 190 0.326 0.394*** -0.032*** YES YES 190 0.401 Permanent TSB Group Ireland 0.100** YES YES 190 0.324 0.394*** -0.032*** YES YES 190 0.399 Powszechna Bank Poland 0.106** YES YES 190 0.333 0.393*** -0.031*** YES YES 190 0.405 Rabobank Group Netherlands 0.100** YES YES 190 0.327 0.400*** -0.032*** YES YES 190 0.402 Raiffeisen Bank Austria 0.102** YES YES 190 0.328 0.400*** -0.032*** YES YES 190 0.404 SNS Bank Netherlands 0.104** YES YES 190 0.329 0.412*** -0.033*** YES YES 190 0.408 Skandinaviska Enskilda Banken Sweden 0.100** YES YES 190 0.328 0.391*** -0.031*** YES YES 190 0.401 Societe Generale France 0.099** YES YES 190 0.324 0.394*** -0.032*** YES YES 190 0.399 Svenska Handelsbanken Sweden 0.100** YES YES 190 0.326 0.403*** -0.033*** YES YES 190 0.403 Swedbank Sweden 0.101** YES YES 190 0.328 0.396*** -0.032*** YES YES 190 0.400 Sydbank Denmark 0.102** YES YES 190 0.328 0.397*** -0.032*** YES YES 190 0.401 UniCredit Italy 0.099** YES YES 190 0.322 0.395*** -0.032*** YES YES 190 0.397 UBI Italy 0.097** YES YES 190 0.318 0.397*** -0.032*** YES YES 190 0.393 WGZ bank Germany 0.101** YES YES 190 0.326 0.395*** -0.032*** YES YES 190 0.400

8