Bank capital (requirements) and credit supply

Journal of Corporate Finance 60 (2020) 101518
Contents lists available at ScienceDirect
Journal of Corporate Finance

journal homepage: www.elsevier.com/locate/jcorpfin
Bank capital (requirements) and credit supply: Evidence from pillar

T
2 decisions☆
Olivier De Jonghea, , Hans Dewachterb, Steven Ongenac
⁎
a
National Bank of Belgium and Tilburg University, Department of Finance, The Netherlands
b
National Bank of Belgium and KULeuven, Belgium
c
University of Zurich, SFI, CEPR, Switzerland
ARTICLE INFO ABSTRACT
Keywords: We analyze how time-varying bank-specific capital requirements affect bank lending to the non-
Capital requirements financial corporate sector as well as banks' balance sheet adjustments. To do so, we relate Pillar 2
Credit supply capital requirements to a comprehensive corporate credit register coupled with bank and firm
Credit register balance sheet data. Our analysis consists of three components. First, we investigate how capital
Bank regulation
requirements affect the supply of bank credit to the corporate sector, both on the intensive and
extensive margin, as well as for different types of credit. Subsequently, we document how bank
JEL classification:
and firm characteristics as well as the monetary policy stance impact the relationship between
G01
G21 bank capital requirements and the supply of credit. Finally, we examine how time-varying bank-
G28 specific capital requirements affect banks' balance sheet composition.
L5
1. Introduction
Following the Great Financial Crisis, there has been a robust consensus among policymakers that financial regulation needs a
macro-prudential dimension on top of an improved micro-prudential and resolution framework. A prime tool, alongside bor-
rower-based measures, of such micro- and macro-prudential supervision and policy is the use of time-varying bank capital re-
The authors would like to thank two anonymous referees, Margherita Bottero (discussant), Ferre De Graeve, Hans Degryse, Maite De Sola Perea,
☆
Mathias Efing (discussant), Stijn Ferrari, Giorgio Gobbi (discussant), Bjorn Imbierowicz, David Martinez Miera (discussant), Janet Mitchell, Klaas
Mulier, Gert Peersman, Alexandre Reginster, Glenn Schepens, Ibolya Schindele (discussant), Bent Vale (discussant), Rudi Vander Vennet, Matjaž
Volk (discussant), Raf Wouters. The paper has also benefited from discussions at various conferences: University of Groningen, Banque de France,
2016 Finest Winter Workshop, MPC Taskforce on Banking analysis for monetary policy (Vienna, 2017), Swiss Winter conference on Financial
Intermediation 2017 (Lenzerheide), DNB-EBC-CEPR conference on “Avoiding and Resolving Banking Crises” (2017), 1st Bristol Workshop on
Banking and Financial Intermediation (2017), IBEFA-ASSA 2018 (Philadelphia). This paper was initially prepared for the 2016 colloquium of the
National Bank of Belgium entitled “The transmission mechanism of new and traditional instruments of monetary and macroprudential instruments”.
The authors gratefully acknowledge the financial support from the National Bank of Belgium. The views expressed are solely those of the authors
and do not necessarily represent the views of the National Bank of Belgium. Ongena acknowledges financial support from ERC ADG 2016-GA
740272 lending.
⁎
Corresponding author.
E-mail addresses: olivier.dejonghe@nbb.be (O. De Jonghe), Hans.Dewachter@nbb.be (H. Dewachter), steven.ongena@bf.uzh.ch (S. Ongena).
https://doi.org/10.1016/j.jcorpfin.2019.101518
Received 24 October 2018; Received in revised form 26 June 2019; Accepted 22 September 2019
Available online 02 November 2019
0929-1199/ © 2019 Elsevier B.V. All rights reserved.
O. De Jonghe, et al. Journal of Corporate Finance 60 (2020) 101518
quirements.1 Higher own funds requirements not only foster bank stability but can also soften credit-led booms, either because
banks internalize more of the potential social costs of credit defaults (through a reduction in moral hazard by having more “skin
in the game”) or charge a higher loan rate due to the higher cost of bank capital (Morrison and White (2005), Adrian and Shin
(2010), Shleifer and Vishny (2010), Adrian and Boyarchenko (2012), Jeanne and Korinek (2013), Malherbe (2015)). Indeed, the
tax benefits of debt finance and asymmetric information about banks' conditions and prospects imply that raising external equity
finance may be more costly for banks than debt finance (Tirole (2006), Freixas and Rochet (2008), Hanson et al. (2011), Aiyar
et al. (2014b), Gornall and Strebulaev (2018)).
Despite the many research efforts now underway by academics to help develop macro-prudential policies (e.g., Galati and
Moessner (2013)), no empirical study so far has comprehensively estimated the joint impact of both time-varying, individual bank
required and actual capital ratios on the supply of bank credit to firms, and heterogeneity across bank and firm characteristics or due
to monetary conditions.
This paper aims to fill this void by analyzing a series of policy experiments with changes in individual bank capital requirements
in Belgium.2 From the adoption of the Twin Peaks supervisory model in April 2011 until the start of the Single Supervisory Me-
chanism in November 2014, the micro-prudential supervision of banks active in Belgium was the responsibility of the National Bank
of Belgium (NBB). One of its micro-prudential instruments to maintain or achieve financial stability was the bank-specific capital
requirement (Pillar 2 capital requirement under Basel II) communicated directly to the bank but otherwise kept confidential. The
resultant series of individual bank capital requirements coupled with comprehensive bank-, firm-, and loan-level data provide an
almost ideal setting for identification and comprehensive assessment of their impact.
In the first part of our analysis, using loan-level data, we find that the joint increase in the required and actual capital in particular
leads to a contraction in the supply of credit to firms, on all margins of credit granting. The fact that higher capital requirements lead
to lower credit supply seems to indicate that equity capital costs for banks are not necessarily negligible, challenging the views of
Admati et al. (2013) and Admati and Hellwig (2013). Therefore, in the second part of our analysis, mobilizing all aforementioned
datasets, we inspect whether or not the impact of required capital on credit supply varies with bank and firm characteristics. In
particular, we shed light on which frictions or costs affect raising actual equity as well as which firms will be affected more by banks'
goal to reduce risk-weighted assets in response to tougher capital requirements. We find that smaller, riskier, or less profitable banks
especially tend to reduce credit the most and that the resultant credit contraction mostly affects large, risky and low-borrowing-cost
firms. Banks' business model characteristics seem to play less of a role in the transmission of capital requirements in credit supply. In
the final part of our analysis, using bank-level data, we find that increases in required bank capital ratios shrink banks' balance sheets,
mortgage and term lending, securities holdings, and deposit collecting while spurring interbank lending and borrowing.
This study differs from existing bank capital literature in several ways. First, we differ from many earlier studies that focus on the
impact of shocks to actual bank capital and not on changes in regulatory requirements. Bernanke and Lown (1991), Berger and Udell
(1994), Berrospide and Edge (2010), among many others, rely on a panel, vector autoregression or matching analysis of bank-level
data, while Hubbard et al. (2002), Ashcraft (2006), Mora and Logan (2012) and Berger and Bouwman (2013), among others, focus
solely on bank-level credit growth or cost. Likewise, Peek and Rosengren (2000) and Puri et al. (2011), among others, exploit the
negative shocks to the profitability of multinational banks that occur abroad and that may affect actual bank capital. Moreover, most
of these studies use bank-level data.
Second, in terms of our individual bank policy experiments, we are closest to Aiyar et al. (2014a), Aiyar et al. (2014b), Bridges
et al. (2014), Mésonnier and Monks (2015), Aiyar et al. (2016), Labonne and Lamé (2018) and Imbierowicz et al. (2018), who use
similar policy experiments in the UK, EU and Denmark. However, these authors also mainly focus on the impact of changes in
required capital ratios on bank balance sheets and aggregate bank lending. In contrast to these papers, we can also study the bank-
firm-level granting of credit using a credit register.
In this respect, we follow Jiménez et al. (2017) who use a credit register to study the impact on the supply of credit of the singular
introduction and subsequent modifications of one macro-prudential policy, i.e., the dynamic provisioning in Spain, which affected all
banks concurrently. Fraisse et al. (2019) and Juelsrud and Wold (2019) similarly focus on a one-off change in capital requirements
that affected all banks, in respectively France and Norway, during the most recent crisis, while Auer and Ongena (2016) study the
compositional changes in banks' supply of credit using variation in their holdings of residential mortgages on which extra capital
requirements were uniformly imposed by the countercyclical capital buffer introduced in Switzerland in 2012. Gropp et al. (2018)
analyze how the outcome of the capital exercise conducted by the European Banking Authority in 2011 affect banks' balance sheets,
syndicate lending and firm outcomes. Finally, Célérier et al. (2016) study the effect of tax reforms abroad (in particular in Italy and
Belgium) and find that the resulting decrease in the cost of equity leads banks to raise their equity ratio, and to concurrently expand
their balance sheets by increasing the amount of credit supplied in Germany.
The main differences with the latter literature originate from studying many changes in individual bank capital ratios over time
1
Under the new international regulatory framework for banks - Basel III - regulators agreed to vary minimum capital requirements over the cycle,
by instituting procyclical bank capital requirements. In this terminology, procyclical bank capital requirements generate countercyclical capital
buffers that deal with the procyclicality of the financial system. Boosting equity in booms provides additional (countercyclical) buffers in downturns
that help mitigate credit crunches.
2
We customarily designate these changes in individual bank required capital ratios as “experiments”, though micro-policy shocks like these are
never (intentionally) randomized and banks dealing with different types of borrowers may be differentially affected. Therefore, both these con-
fidential shocks and comprehensive micro data are necessary for identification.
2
and using multiple micro-datasets. Our contribution is thus that we can comprehensively study policy experiments that change
individual required and actual bank capital ratios. We study their impact on bank balance sheets and credit granting, and study
various sources of heterogeneity in this relationship due to monetary, firm and bank conditions.
The rest of the paper proceeds as follows. Section 2 discusses the process that leads to Pillar 2 capital requirements. In Section 3,
we present the methodology and document the relationship between capital (requirements) and the supply of credit to firms. Section
4 consists of three subsections documenting the channels through which the identified relationship works. In particular, we examine
the heterogeneous impact of required capital on credit supply due to differences in banks' cost of equity, banks' business models and
firms' impact on risk-weighted assets. In Section 5, we present the outcome of many extensions and robustness checks. In particular,
we shed further light on the causality of the established relationships as well as analyzing potential anticipation effects of the policies
and asymmetric responses. In Section 6, we analyze how capital requirements affect broad balance sheet categories, allowing us to
assess what happens to other markets beyond lending to corporations. Section 7 concludes.
2. Pillar 2 capital requirements
We focus on an interesting period between the adoption of the Twin Peaks supervisory model in April 2011 and the start of the
Single Supervisory Mechanism in November 2014. During that period, the micro-prudential supervision of banks active in Belgium
was the responsibility of the National Bank of Belgium (NBB). Hence, during that time, the micro- and macro-prudential supervision
of the Belgian financial sector was integrated within one single institution, with the aim of maintaining and improving both the
micro- and macroeconomic resilience. Prior to April 2011, supervision was one of the responsibilities of the Banking, Finance and
Insurance Commission. Post-2014, with the introduction of the Single Supervisory Mechanism, the supervision of significant banks
has been centralized at the European Central Bank, whereas that of less significant institutions is still the responsibility of the NBB.
One of the micro-prudential instruments in the NBB's toolbox to maintain or achieve financial stability are bank-specific capital
requirements (Pillar 2 capital requirements under Basel II). The Pillar 2 regulatory capital requirements are the outcome of a (usually)
yearly Supervisory Review and Evaluation Procedure (henceforth, SREP or the evaluation process) for individual banks operating in
Belgium. The evaluation process is a continuous procedure that results in a Tier 1 common equity capital requirement that is privately
communicated to the bank by the end of a given year and becomes effective and binding as of January 1st of the following year. The
last capital requirements that are still set by the NBB (rather than the Single Supervisory Mechanism) are communicated in December
Fig. 1. Diagram of the supervisory review and evaluation process.

The flow chart depicts the various steps and ingredients in the Supervisory Review and Evaluation Process (SREP). On the left-hand side, the various
inputs to SREP are listed. NBB is shorthand for National Bank of Belgium and ICAAP stands for internal capital adequacy assessment process. The
inputs are used to perform an analysis and risk assessment, which are quantified by means of a scorecarding system. The output of the SREP are
prudential measures which could consist of three components: an action plan, a specific capital decision or other measures. The International
Monetary Fund writes in detail, in its 2013 Financial Sector Assessment Plan, on compliance of the SREP with the Basel Core Principles for effective
banking supervision. In particular, it mentions that “the NBB's approach to Pillar 2 is well developed using a scorecard as the primary tool for risk
analysis, taking into account qualitative and quantitative measures. At least on an annual basis, the NBB determines the minimum capital adequacy
requirements for all banks on a forward looking basis. The SREP and Internal Capital Adequacy Assessment Procedure analysis are important inputs
into the process as are, if available, outputs from banks' economic capital models. Stress testing is also taken into account as to ascertain whether the
bank is able to maintain capital buffers under stress conditions.” (see IMF Country Report No. 13/133).
Source: Prudential regulation and supervision, NBB Report 2015, p. 239.
3
2014, thus implying that our sample period ends in 2015. Fig. 1 provides a graphical presentation of the evaluation process that
involves multiple inputs to assess the risk profile of an institution and to determine appropriate supervisory actions.
Inputs in the evaluation process are, among other things, internal bank performance reports, reports from external and internal
auditors, the credit institution's Internal Capital Adequacy Assessment Process, as well as information obtained via contacts with the
institutions and other supervisors (via so-called supervisory colleges). Quantitative and qualitative assessment of risks as well as their
management by institutions are an essential component of the SREP. The structured analysis of risks under the SREP is provided by the
score-carding system, covering both a quantitative assessment of the level of risks and qualitative aspects of the quality of management
control. Together with the review of the Internal Capital Adequacy Assessment Process and stress tests, the output of the score-carding
process feeds into the overall risk assessment of the institution and forms the basis on which supervisory actions will be determined and
planned. Through this process, the final capital decision, which must be approved by the NBB Board, takes into consideration a full range of
risks. As micro-prudential supervisor, the NBB was, through the evaluation process, responsible for setting the Pillar 2 capital requirements
for the banks under its supervision. Up until 2014, this involved all banks with a Belgian banking license, both domestic banks as well as
subsidiaries of foreign banks operating in Belgium. On the basis of the evaluation process, the NBB sets capital requirements for all relevant
legal entities, including both capital requirements at group level as well as at the level of individual banks.
From the micro-prudential supervisor, we obtain information on both the required capital ratio as well as the actual capital ratio
at individual bank level. More specifically, we focus in this paper on the following two variables. First,
Previous quarter actual capital ratio captures the ratio of Tier 1 common equity to risk-weighted assets. It captures the amount of
regulatory capital in relation to risk-weighted assets at the beginning of the quarter over which loan growth is measured. Second,
Previous quarter required capital ratio is the required capital ratio (Pillar 2 capital requirement). The banks are informed of the
required capital ratio at the end of the last quarter of the preceding year and it is binding for the entire upcoming year. Hence, for all
quarters in a given year Y , the variable Previous quarter required capital ratio coincides with the announced required ratio in the last
quarter of year Y − 1. The difference between these two measures is the capital buffer.3
Crucial for the analysis is that the supervisor has discretion in setting the Pillar 2 capital requirements and that the decision is
exogenous with respect to the risks covered by Pillar 1 as well as past or future lending growth. Moreover, in the analysis, because of
the sets of fixed effects, we only exploit the within-bank variation as well as the within-time-period variation to avoid concerns that
time-invariant bank-specific characteristics or economic conditions jointly affect lending decisions as well as the capital require-
ments. Moreover, in the robustness section, we will conduct further analyses documenting that all established relationships can be
interpreted in a causal sense.
Summary statistics on these measures are provided in Table 1. Over the sample period, the average required capital ratio is 11.0%
of risk-weighted assets (RWA), with a standard deviation of 1.9%. The actual risk-weighted capital ratio has a mean of 14.9% of and a
standard deviation of 3.8%. Hence, both the mean and the dispersion in actual capital ratios are higher than those of the required
capital ratios. The average capital buffer, the difference between the actual and required capital ratio, equals 3.9% of RWA, indicating
that the average bank has a relatively large cushion.4 The range of the variation in (required) capital across banks corresponds with
the statistics reported by Aiyar et al. (2014b) who have information on the UK for the period 1998–2007. Using a longer time span
and a larger sample of banks, they find that the minimum required capital ratio was 8%, its standard deviation was 2.2%, and its
maximum was 23% of risk-weighted assets.
In panel B of Table 1, we report summary statistics on some bank characteristics for those banks for which we have information
about the regulatory capital ratio. In each of these characteristics, there is sufficient variation across banks and over time. The
average bank generates a low but stable quarterly return on equity, has a loan to total assets ratio of 53.4%, relies strongly on retail
deposits (the average share of demand and savings deposits is 80%) and 61% of its income stems from interest-generating activities.
During the sample period, the Belgian economy and banking sector experienced low, but stable growth.
The bank capital requirements are set for domestic banks as well as for subsidiaries of foreign banks. The NBB has no supervisory
authority over branches of foreign banks and hence these are not included in our sample. In Table A1 of the appendix, we provide
detailed information on the number of borrowers, the total amount of credit granted and aggregate total bank assets for the group of
banks for which we do and do not have information on regulatory capital. We also report the total number of banks in each group. For
three banks, the information on required capital is not available for the first year of our analysis (2013). While the number of banks in
each group is similar (in 2014 and 2015), note that the sample of banks with regulatory data contains the important banks. In each
quarter of 2014 and 2015, the SREP-sample banks cover at least 95% of all firm-bank relationships, total volume of credit granted and
total banking assets in Belgium. Note also that the share of corporate lending in total assets is lower in the group of foreign bank
branches (no Pillar 2 data available) compared to the group of Belgian banks or foreign subsidiaries (for which Pillar 2 requirements
are available).
3
Banks do not have much room to tweak or manipulate these capital requirements. While regulatory arbitrage cannot necessarily be avoided, if
observed (for instance by ‘playing’ with diversification or aggregation procedures), the supervisor can address this in its Pillar 2 add-ons.
4
The average capital buffer is substantially high. Banks are careful and keep sufficient voluntary (management) buffers to avoid a breach. Also,
supervisors do not necessarily wait for a breach to intervene and voice concerns on low management buffers ahead of actual problems. However, in
case of a breach of the requirements several actions can be taken. First, a breach implies a thorough discussion with the supervisor on plans to
restore, in the shortest time possible, the capital base to satisfy the capital requirement. Second, in the short run, this may imply several actions
including a prohibition to distribute dividends and interventions in remunerations of the leading management. Finally, in the event of prolonged or
structural breach, the supervisor can assign external managers to restructure the bank.
4
3. Regulatory capital and corporate credit supply
3.1. Measuring credit conditions
To identify the causal impact of capital requirements on corporate credit supply, we make use of information from the Corporate
Credit Register held by the National Bank of Belgium. The Corporate Credit Register contains information on credit granted by credit
institutions to legal entities (i.e., enterprises). Each month, a credit institution needs to provide information to the Credit Register on
all debtors. We extract all available credit data at the bank-firm-quarter level but only retain non-financial corporations (NFCs). This
implies that we filter out (i) borrowers from the financial or insurance sector (in line with most other corporate finance studies) and
(ii) borrowers active in public administration, education or household activities (as these are not NFCs, nor do they file balance
sheets). We use quarterly information rather than monthly as this matches the frequency of the bank balance sheet data and in-
formation on banks' capital ratios.
Using this database, we construct several variables that vary at the bank-firm-quarter level and provide an indication of credit
conditions at both the intensive and extensive margin. Summary statistics of these measures are reported in panel A of Table 2. First
of all, we compute quarterly authorized credit growth at the bank-firm level ( Credit growth). This is the quarterly change in the
natural logarithm of authorized credit at the bank-firm level. More importantly, we define credit as the total amount of authorized (or
granted) loans. The average growth in authorized credit granted is slightly negative at −2.9%, mainly due to amortizations. There is
substantial cross-sectional variation, with a standard deviation of 23.7%, indicating that some firms witness substantial drops in their
credit exposures (due to outright cuts or lower likelihood of renewals or roll-overs), whereas other bank-firm exposures grow sub-
stantially. However, less than 10% of the bank-firm credit exposures strictly increase over the course of a quarter.
Second, we create a dummy Large drop in credit which equals one if the firm's credit growth is in the lowest quartile of the credit
growth distribution of all the bank-firm observations in the sample. The quarterly growth rate of authorized credit at the 25th
percentile is −6.15%, whereas the average in this lowest quartile is −24% (neither statistics are reported in Table 2). This variable
proxies for authorized credit volumes that have been reduced substantially, or matured without having been rolled over. The two
aforementioned measures provide information on the intensive margin of bank-firm relationships. That is, they provide an indication
of banks' willingness to extend credit to incumbent borrowers. Our third measure is a dummy New relationships that equals one if a
firm has credit granted from a bank at time t, but was not borrowing from that bank at the end of the previous year (when the new
required capital ratio is communicated). in doing so, we will not only test the impact of capital requirements on actual loan growth,
but also the impact on banks' propensity to generate new bank-firm credit relationships (and thus the extensive margin of credit
provision). On average over all quarters, 2.2% of bank-firm relationships are newly established relationships.
The three aforementioned measures are based on authorized or granted amounts of credit. For credit lines, this implies that we
look at the total amount of credit that is available, and not at the portion that is taken up by the borrower. In this way, we make sure
that any changes in credit are not driven by a sudden draw-down of a credit line by a borrower (which is an indication of credit
demand rather than credit supply). However, as a fourth indicator, we also compute the Utilization rate of credit, which is the ratio
of the utilized amount at the bank-firm level over the authorized amount at the bank-firm level, measured at time t. The average
utilization rate is 66.7%, but there is huge variation across bank-firm pairs. Indeed, the standard deviation of the utilization rate is
39.1%.
Table 1
Summary statistics: Bank-level data.
Mean Standard deviation 5th Percentile Median 95th Percentile
Panel A: actual and required capital ratio
Required capital ratio 0.110 0.019 Not reported due to

Previous year required capital ratio 0.110 0.024 Data confidentiality reasons
Actual capital ratio 0.149 0.038
Previous year actual capital ratio 0.151 0.043
Panel B: bank characteristics
ln(Total assets) 9.867 1.672 7.077 9.835 12.247

Loans to total assets 0.534 0.174 0.035 0.544 0.766
Loans to deposits 0.945 0.639 0.553 0.784 2.556
Off-balance-sheet items to total assets 6.370 5.494 1.438 3.144 16.127
Share of demand and savings deposits 0.800 0.109 0.594 0.817 0.946
(Quarterly) return on equity 0.102 0.127 −0.075 0.102 0.285
(Quarterly) provisions to loans 0.001 0.001 −0.000 0.000 0.003
(Quarterly) interest income share 0.610 0.278 0.142 0.712 0.893
This table contains summary statistics on (regulatory) bank capital (panel A) and bank-specific characteristics (panel B). Data come either from the
SREP (bank capital) or the regulatory filings (balance sheet and income statement) and are on the bank-quarter level. Stock data are measured at the
end of the quarter. Flow data are changes accumulated over the quarter. The total number of observations is 132, but is unbalanced over 12 quarters
(2013Q1 to 2015Q4) and concerns 14 banks. Variables are winsorized at the 2% level.
5
Table 2
Summary statistics: firm-bank level data.
Variable Observations Mean Std.dev 5th percentile Median 95th percentile
Panel A: credit growth
Authorized credit amount 1,022,297 644,563 7,749,937 2500 75,723 1,569,095

Number of relationships 1,022,297 2.219 0.468 2.000 2.000 3.000
Credit growth 1,022,297 −0.029 0.237 −0.325 −0.012 0.182
Large drop in credit 1,022,297 0.233 0.423 0.000 0.000 1.000
New bank-firm relationship 1,067,376 0.022 0.148 0.000 0.000 0.000
Utilization rate 1,067,376 0.667 0.391 0.000 0.877 1.000
Term credit growth 577,073 −0.067 0.320 −0.549 −0.045 0.314
Collateralization rate 1,022,297 0.616 0.637 0.000 0.463 2.000
Panel B: bank characteristics (estimation sample, SREP banks)
Required capital ratio 1,022,297 0.109 0.015 Not reported due to

Previous year required capital ratio 1,022,297 0.107 0.015 Data confidentiality reasons
Actual capital ratio 1,022,297 0.155 0.029
Previous year actual capital ratio 1,022,297 0.161 0.033
(Lagged) log total assets 1,022,297 11.643 0.883 9.293 11.958 12.268
(Lagged) quarterly growth of common equity 1,022,297 0.005 0.060 −0.108 0.020 0.095
(Lagged) quarterly growth of deposits 1,022,297 0.011 0.063 −0.030 0.005 0.052
(Lagged) quarterly growth of total assets 1,022,297 −0.007 0.062 −0.083 0.002 0.081
(Lagged) return on equity 1,022,297 0.095 0.093 −0.082 0.092 0.233
Default probability 1,022,297 0.045 0.142 0.001 0.009 0.170
This table contains summary statistics on corporate credit (panel A) and bank characteristics (panel B). The unit of observation is a (firm, bank,
quarter) triplet. In panel A, we provide information on various aspects of corporate credit such as the authorized amount, the number of re-
lationships, credit provision at the intensive margin (credit growth and large drop in credit) as well as the extensive margin (new bank-firm
relationship), utilization rates of granted credit, term credit growth as well as the degree of collateralization. In panel B, we report summary
statistics on the independent variables included in subsequent regressions. Observations at the bank-firm-quarter level are winsorized at the 1%
level.
Fifth, we also focus on term loans only, for which there is no discrepancy between authorized and utilized amounts. Term loans
are either fixed-term credit, non-residential mortgage loans or non-mortgage instalment loans. The average (and median)
Term credit growth (−6.7%) is much lower compared to the average credit growth of all credit (−2.9%), precisely because of the
amortization schemes of the former. Finally, our sixth measure of credit conditions is the Collateralization rate, which is the ratio of
pledged collateral by a firm to a bank (for all its credit) over the total amount of authorized credit at the bank-firm level. On average,
61.6% of the granted credit amount is covered by pledged collateral, although the median is much lower (46.3%) as many firms
pledge more collateral than the borrowed amount (or the outstanding amount).
3.2. Methodology and hypotheses
The empirical specification used to document the relationship between (required) regulatory capital and the six credit variables is
the following:
Credit condition (Quarterly)b, f , t = 1 Required capital ratiob, t 1

+ 2 Previous year required capital ratiob, t 5
+ 3 Actual capital ratiob, t 1
+ 4 Previous year actual capital ratiob, t 5
+ Bank controlsb, t 1 + b + f , t + b, f , t (1)
The trio (b, f, t) stands for bank, firm and time, where the unit of time is a quarter. νb is a bank fixed effect that controls for
observed and unobserved time-invariant heterogeneity across banks such as e.g., ownership structure or managerial quality. νf, t is a
firm × time fixed effect that enables us to analyze the change in credit availability to the same firm at the same time by banks with
different regulatory capital ratios. In doing so, we control for observed and unobserved time-varying firm heterogeneity in loan
demand, quality and risk; an empirical technique pioneered by Khwaja and Mian (2008). Finally, εb, f, t is a zero-mean random noise
component. The standard errors are clustered at bank level.
It is crucial for the analysis that the supervisor has discretion in setting the Pillar 2 capital requirements and that the decision is
exogenous with respect to the risks covered by Pillar 1 as well as past or future lending growth. Moreover, in the analysis, because of
the sets of fixed effects, we only exploit the within-bank variation as well as the within-time-period variation to avoid concerns that
time-invariant bank-specific characteristics or economic conditions jointly affect lending decisions as well as the capital require-
ments. Moreover, in the robustness section, we will conduct further analyses documenting that all established relationships can be
6
interpreted in a causal sense.

There are four independent variables of interest based on the regulatory capital data. They are: Required capital ratiob, t−1,
Previous year required capital ratiob, t−5, Actual capital ratiob, t−1 and Previous year actual capital ratiob, t−5. Besides the one-
quarter lagged (required) capital ratio, we also include the previous year required capital ratio, Previous year required capital ratiob,
t−5, as the impact of capital requirements may take time to have effect. Previous year actual capital ratiob, t−5 is the one-year lagged
actual capital ratio, which is included in the specification for reasons of analogy with required capital.
First of all, the main coefficients of interest are β1 and β3. Our first two hypotheses test whether required and/or actual capital
ratios affect credit conditions. They are: H1a: β1=0 and H1b: β3=0. Second, besides individual significance of coefficients, we are
also interested in three different joint hypotheses. First, the effect of an increase in required capital, ceteris paribus (and hence
holding the actual capital ratio constant), is reflected in coefficient β1. Such a situation, in which a bank's actual capital ratio does not
react to an increase in requirements leads to a reduction in the capital buffer. If banks were to simultaneously also change their actual
capital ratio, to restore the buffer, then the effect on credit supply is given by β1+β3. We will thus also test whether the impact of
changes in regulatory capital requirements, when banks hold their buffer constant, is significantly different from zero (H2:
β1+β3=0). Note that the boundaries of the interval [β1, β1 + β3] correspond to no adjustment and complete adjustment of the actual
capital ratio in response to a regulatory change. In all cases where the buffer is only partially restored, the estimated economic effect
will be in between this interval.
Third, we also test whether or not the sum of the coefficients on the required and the previous year required capital ratio is
significantly different from zero (that is H3a: β1+β2=0). Likewise, we test whether or not the sum of the coefficients on the actual
and the previous year actual capital ratio is significantly different from zero. That is, we test hypothesis H3b: β3+β4=0, to analyze
whether the joint impact of current and lagged required capital is significantly different from zero. The latter two tests serve the
purpose of testing whether there is a long-lasting impact of capital requirements or whether there is a reversal (e.g., because of initial
overshooting).
Summary of the baseline hypotheses
H1a β1=0 Does the required capital ratio affect credit conditions?
H1b β3=0 Does the actual capital ratio affect credit conditions?
H2 β1+β3=0 Does the required capital ratio affect credit conditions, when banks hold the capital buffer constant?
H3a β1+β2=0 Is there is a long-lasting impact of required capital on credit conditions?
H3b β3+β4=0 Is there is a long-lasting impact of actual capital on credit conditions?
The vector Bank controlsb, t−1 consists of the following variables: the natural logarithm of total assets, quarterly growth in
common equity, quarterly growth in deposits, quarterly growth in total assets and quarterly return on equity. All bank variables have
been lagged one quarter such that they are in principle predetermined with respect to next quarter's credit growth at the bank-firm
level. In this set of controls, we also include one that varies at the bank-firm level, which is a bank-firm specific measure of credit risk
(as in Aiyar et al. (2016)). In the credit register, banks provide information on their assessment of the probability that the firm will
default during the ensuing year. This assessment is at the borrower level, rather than the loan (category) level, and is updated every
month. We include this measure of ex-ante borrower quality. Importantly, note that it varies at the bank-firm-time level.
Four issues on the specification are worth emphasizing. First, Eq. (1) is the most flexible specification which can be rewritten in
terms of a model that (i) includes the buffer and the required capital ratio, (ii) the buffer and the actual capital ratio, or (iii) changes
in the buffer or changes in the actual and required capital ratio. Note that a model with only the buffer would imply that β1=−β3,
which is rejected by the data.
Second, prior to running the main regression, we run auxiliary regressions to analyze the scope for collinearity due to correlation
between the actual and required capital ratio and/or correlation between the current and lagged values. Variance inflation factors
(VIFs) are low and range between 1.43 for the required capital ratio and 1.75 for the actual capital ratio, with values in between for
the previous year actual and required capital ratio. These VIFs are sufficiently low to give no concern about possible consequences of
collinearity.
Third, the choice of the lag structure is not determined by a statistical test but is driven by the nature of the required capital data
and the frequency of the analysis (quarterly). Actual capital and all other variables vary at the quarterly level, but the required capital
ratio is fixed for a whole year. We therefore only include the lags t-1 and t-5 in the specification, which means including the current
required capital ratio and the previous value of the required capital ratio (i.e., the one that applied in the previous calendar year).
Jointly including all intermediate lags would lead to spurious results due to highly correlated variables (because of the persistence)
and, accordingly, high VIFs.5 Given the staleness of the intermediate lags vis-à-vis the current or one-year lagged values and the
higher VIFs, we prefer the parsimony of Eq. (1).
Fourth, two of the dependent variables are binary indicators (‘Large drop’ and ‘New bank-firm relationships’). In the absence of
any fixed effects, a binary dependent variable would imply using a logit or probit model. However, in the presence of many fixed
5
The required capital ratio at the intermediate lags is either identical to the current required capital ratio or one-year lagged required capital ratio.
For example, for observations in the third quarter, the values of required capital at t-2 will be the same as on t and t-1. The value on t-3 (referring to
the fourth quarter of the previous year) will be the same as the one-year lagged required capital ratio.
7
effects (firm*time FE and bank FE), such non-linear models suffer from an incidental parameter problem. Therefore, we follow
common practice (see, e.g., Khwaja and Mian (2008)) and use a linear probability model that allows for (multiple sets of) high-
dimensional fixed effects.
3.3. Baseline results: loan level
The results of estimating Eq. (1) for the six different dependent variables that capture various dimensions of credit provision are
reported in Table 3.
Four general findings stand out. First, ceteris paribus, focusing on the estimated β1 coefficients in the different columns, we see
that raising the required capital ratio reduces the supply of credit on the intensive and extensive margin. We thus reject hypothesis
H1a: β1=0. Holding constant the level of actual capital (and hence a shrinking buffer), a one standard deviation increase in required
capital (1.5 pp) leads to a 0.19 pp. decrease in the quarterly credit growth rate. Put differently, the same firm borrowing from two
banks that differ only in the level of required capital will see slightly lower credit growth from the bank with the higher required
capital ratio. The economic significance of this effect is nevertheless moderate. In line with this finding, banks with higher capital
requirements are less likely to start new bank-firm relationships, and again the effect is statistically significant, but economically
moderate (column 3). The tightening of the credit supply by banks with a higher capital requirement implies a higher utilization rate
Table 3
Capital (requirements) and credit supply: baseline results.
Variables (1) (2) (3) (4) (5) (6)
Credit growth Large drop in New bank-firm Utilization rate Credit growth - term Collateral rate
credit relationships loans
Required capital ratio −0.129* 0.333 −0.808*** 0.285* −0.310** 0.526

(0.067) (0.264) (0.245) (0.152) (0.135) (0.440)
Previous year required capital ratio −0.251*** 0.395* −0.858*** 0.090 −0.283** −0.828
(0.050) (0.184) (0.198) (0.126) (0.101) (0.625)
Actual capital ratio −0.136** 0.434** 0.065 0.200** −0.258** −0.807*
(0.053) (0.167) (0.095) (0.072) (0.115) (0.426)
Previous year actual capital ratio −0.009 −0.034 −0.079 0.019 0.048 0.101
(0.048) (0.074) (0.100) (0.077) (0.127) (0.223)
(Lagged) log total assets 0.014 −0.038* −0.363*** −0.002 0.011 0.123***
(0.016) (0.018) (0.085) (0.011) (0.022) (0.037)
(Lagged) quarterly growth in −0.043** 0.047* 0.005 0.049*** −0.095*** −0.015
common equity (0.015) (0.025) (0.013) (0.008) (0.028) (0.034)
(Lagged) quarterly growth in 0.009 0.015 −0.057** −0.013 −0.012 −0.069*
deposits (0.006) (0.0110) (0.023) (0.015) (0.016) (0.034)
(Lagged) quarterly growth in total −0.019 0.054*** 0.101*** 0.079*** 0.043* −0.089***
assets (0.012) (0.016) (0.028) (0.017) (0.024) (0.019)
(Lagged) quarterly return on equity 0.025 −0.038 −0.038 −0.009 0.026 −0.006
(0.018) (0.033) (0.033) (0.009) (0.023) (0.028)
Default probability −0.040*** 0.032* −0.015*** 0.117*** −0.034*** 0.291***
(0.006) (0.016) (0.004) (0.016) (0.009) (0.051)
Observations 1,022,297 1,022,297 1,067,376 1,067,376 577,073 1,022,297

R-squared 0.47 0.50 0.51 0.58 0.48 0.54
Firm*time fixed effects Yes Yes Yes Yes Yes Yes
Bank fixed effects Yes Yes Yes Yes Yes Yes
Cluster Bank Bank Bank Bank Bank Bank
β1 + β3 −0.27 0.77 −0.74 0.48 −0.57 −0.28

p-value of test (β1 + β3=0) 0.03 0.09 0.01 0.04 0.02 0.74
p-value of test (β1 + β2=0) 0.00 0.11 0.00 0.11 0.01 0.76
p-value of test (β3 + β4=0) 0.16 0.12 0.93 0.12 0.32 0.29
This table contains estimation results from a regression relating various dimensions of credit growth to (regulatory) capital ratios. More specifically,
we run the following regression for six different dependent variables:
Credit conditions (Quarterly) b, f, t = β1 ∗ Required capital ratiob, t−1 + β2 ∗ Previous year required capital ratio b, t−5 + β3 ∗ Actual capital ratio b,
t−1 + β4 * Previous year actual capital ratiob, t−5 + γ ∗ Bank controlsb, t−1 + νb + νf, t + ϵb, f, t.
Besides the variables of interest (the labels of which are reported in bold), the model includes lagged control variables, which are: the natural
logarithm of total assets, quarterly growth in common equity, quarterly growth in deposits, quarterly growth in total assets and quarterly return on
equity. In all regressions, we also control for a bank-specific assessment of the firm's default probability. These variables have been lagged one
quarter such that they are in predetermined with respect to next quarter's credit growth at the bank-firm level. In addition, we also include a bank
fixed effect (νb). νf, t is a firm × time fixed effect that captures time-varying firm demand shifters. Standard errors are clustered at bank level. At the
bottom of the table, we also provide information on the p-values of various tests on the coefficients of interest.
***, ** and * denote p < 0.01, p < 0.05 and p < 0.1 respectively.
8
by the affected firms. The ratio of utilized (or drawn) credit to authorized credit increases. Note also that the coefficient on required
capital is 2.5 times larger when only focusing on term loans (fifth column), implying an economically larger effect for term loans
compared to all credit exposures. A one standard deviation increase in required capital leads to an effect of 0.5 pp. on term credit
growth. There is no significant effect of capital requirements on the collateralization rate.
Second, an increase in actual capital holdings, holding constant the required capital ratio (and thus an increase in the buffer) also
reduces the supply of credit. We thus also reject hypothesis H1b: β3=0. Whenever significant, β3 has a similar sign to β1. The
economic effects are again moderate and of comparable magnitude to an increase in the required capital ratio.
Third, combining the two aforementioned situations, i.e., an increase in both the required and actual capital ratios (e.g., a
situation where a bank is preserving its capital buffer in response to an increase in the capital requirement), leads to a stronger
contraction of credit supply. We report the sum of these two coefficients (β1+β3) in the first row of the lower panel and the p-value of
testing hypothesis two (H2: β1+β3=0) in the second row of the lower panel. Except in the last column, we can reject the hypothesis
that a joint and equal increase in required and actual capital ratios (thus holding constant the regulatory buffer) does not affect credit
supply. A one standard deviation increase in required capital (1.5 pp) combined with a 1.5 pp. increase in the actual capital ratio
leads to (i) a 0.40 pp. decrease in the quarterly credit growth rate (−0.80 pp. for term credit growth), (ii) a 1.14 pp. increase in the
probability of experiencing a large drop and (iii) a 1.11 pp. decrease in the probability of starting a new bank-firm relationship.
Moreover, the utilization rate of credit goes up with 0.72 pp. in such a case.
Fourth, the effect of changes in the required capital ratio is long-lasting. That is, the required capital ratio set in the previous year
also has a significant impact on the growth of credit (β2 is significant in four specifications), whereas the previous year actual capital
ratio does not have a significant impact on credit supply (β4 is never significant). In the next to last row of the table, the p-value of a
significance test of the sum of the effects of the current and previous year required capital ratio often hints at statistical significant
effects of the joint impact. H3a: β1+β2=0 is rejected in half of the cases, whereas H3b: β3+β4=0 is never rejected.
In sum, we do find statistically significant effects of bank capital requirements on the provision of credit both at the intensive and
extensive margins. However, the effects are moderate in economic magnitude and the methodology only enables an assessment of the
differential credit supply effects of capital requirements. Therefore, in the next subsection, we test whether higher capital require-
ments are effectively leading to higher credit constraints at the firm level. Put differently, we will test whether firms can set off the
reduction in credit from banks with higher capital requirements with credit from banks with lower capital requirements or with credit
from branches of foreign banks that are not supervised by the National Bank of Belgium.
3.4. Aggregate results: firm level
The results in Table 3 show that firms will receive less credit from banks with higher capital requirements. These loan-level results
imply a drop in credit supply, however, the effects could be mitigated if firms can obtain credit from less affected banks or from
foreign branches. To address the scope for mitigation, we conduct some additional analyses in line with the approach followed by
Khwaja and Mian (2008) and Jiménez et al. (2017) and assess an aggregate or macro effect of capital requirements using firm-level
estimations.
Therefore, we collapse the data to the firm-time level. The dependent variable is now the log change in credit received by a firm
from all banks for which we know the SREP capital requirement. The credit obtained after the shock can be from both “current” and
“non-current” banks (that did not lend to the firm prior to the shock). The bank-specific variables (and thus also the capital re-
quirements) are weighted averages, weighing each bank value by its loan volume to the firm over total bank loans taken by this firm.
As this is a regression at the firm level, we can no longer include a firm × time fixed effect. To control for firm demand, we include the
estimate firm × time fixed effect from Eq. (1). We also include a weighted average of the estimated bank fixed effect, weighing each
bank fixed effect by its loan volume to the firm over total bank loans taken out by this firm. The goal of this first additional regression
at the firm level is to analyze whether firms can offset the reduced credit supply from banks with a higher capital requirement with
credit from banks with a lower capital requirement. The results reported in the first column of Table 4 show that this is not the case.
Indeed, the coefficient on required capital is still negative and significant.
In the Belgian credit market, there are also branches of foreign banks active which are not subject to supervision by the NBB and
are thus not subject to Pillar 2 requirements by the NBB. For these branches, we do not know whether or not their home supervisor
imposes Pillar 2 requirements. If these branches' parent companies were to face lenient capital requirements, then they could sub-
stitute the loss of credit faced by firms because of more tightly regulated Belgian banks. In order to test that, we now use, as a
dependent variable, the log change in credit received by a firm from all banks (including foreign branches) active in Belgium. In the
second column of Table 4, we still find a negative and significant effect of required capital. Borrowers, who are receiving less credit
from domestic banks with higher capital requirements, cannot set off this by borrowing either from domestic banks with a lower
requirement or by borrowing from foreign branches.
In the third column, we use as dependent variable the growth rate of credit provided to a firm by foreign branches (not supervised
by the NBB). Note that this is a substantially smaller sample. Only a small fraction of firms (3.6%) borrow from both a regulated bank
and a foreign branch. Firms seem to receive more credit from outside branches if their domestic banks face higher capital re-
quirements (β1 is positive), but the coefficient is insignificant. Yet, while the credit loss cannot be offset, we do find a compositional
effect: the share of credit provided by NBB-supervised banks (in total credit received by the firm) is lower for firms borrowing from
banks with higher capital requirements (last column).
In sum, we find that tighter capital requirements lead to aggregate negative effects on credit supply. Firms cannot substitute the
reduction in credit by borrowing more from banks with lower capital requirements or from foreign branches operating in Belgium.
9
Table 4
Capital (requirements) and credit supply: aggregate effects.
Variables Firm-level credit growth Share of credit
SREP banks All banks Non-SREP banks SREP banks/All Banks
Required capital ratio −0.102*** −0.094*** 0.187 −0.016*

(0.024) (0.024) (0.229) (0.008)
Previous year required capital ratio −0.169*** −0.170*** 0.058 0.008
(0.020) (0.020) (0.206) (0.007)
Actual capital ratio −0.092*** −0.093*** −0.004 0.015***
(0.011) (0.011) (0.102) (0.003)
Previous year actual capital ratio −0.023** −0.020* 0.050 0.002
(0.010) (0.011) (0.100) (0.003)
(Lagged) log total assets 0.010*** 0.010*** 0.007* 0.000

(0.001) (0.001) (0.004) (0.000)
(Lagged) quarterly growth in common equity −0.045*** −0.041*** 0.049 −0.002**
(0.005) (0.005) (0.042) (0.001)
(Lagged) quarterly growth in deposits 0.006* 0.004 −0.062 0.003***
(0.003) (0.003) (0.038) (0.001)
(Lagged) quarterly growth in assets −0.011** −0.006 0.093** −0.002
(0.005) (0.005) (0.046) (0.001)
(Lagged) quarterly return on equity 0.013*** 0.012*** 0.001 −0.001
(0.003) (0.003) (0.027) (0.001)
Default probability −0.039*** −0.0389*** −0.013 0.001**
(0.001) (0.001) (0.012) (0.001)
Observations 475,770 475,770 17,296 471,853

R-squared 0.56 0.55 0.00 0.91
Cluster ILST ILST ILST ILST
Firm demand Estimated firm*time FE from column 1 of Table 3
Bank FE Weighted average estimated Bank FE from column 1 of Table 3
This table contains firm-level estimation results. In particular, we regress three different firm credit growth measures on (regulatory) capital ratios.
In column 1, the dependent variable is the growth in aggregate firm credit obtained by all banks in the SREP sample. In column 2, the dependent
variable is the growth in firm-level credit obtained from all banks (hence including credit from non-SREP banks). In column 3, we examine firm-
level credit growth obtained from non-SREP banks. Finally, in column 4, we use the share of credit by SREP banks in total firm credit as a dependent
variable. Using data at the firm level, we analyze whether the drop in credit because of stricter capital requirements can be offset by banks with
lower requirements or by banks not supervised by the National Bank of Belgium. More specifically, we run the following regression for these four
different dependent variables:
Credit conditions (Quarterly)f, t = β1 ∗ Required capital ratiob, t−1 + β2 ∗ Previous year required capital ratiob, t−5 + β3 ∗ Actual capital ratiob,
t−1 + β4 * Previous year actual capital ratiob, t−5 + γ ∗ Bank controlsb, t−1 + δ*Firm controlsf, t−1 + ϵf, t.
Besides the variables of interest (which are reported in bold), the equation includes control variables, which are weighted averages using the credit
shares at the firm-bank level as weights. They are: the natural logarithm of total assets, quarterly growth in common equity, quarterly growth in
deposits, quarterly growth in total assets, quarterly return on equity and firms' probability of default. All bank variables have been lagged by one
quarter such that they are in predetermined with respect to next quarter's credit growth. Firm controls are imputed fixed effects obtained from
column 1 of Table 1. In particular, firm demand is now controlled for by including the estimated fixed effect from column 1 of Table 3. Weighted
average estimated bank fixed effects (from column 1 of Table 3), which are firm-specific because of the weights, are also included. Standard errors
are clustered at the industry-location-size-time (ILST) level.
This lack of substitution by foreign branches could be either due to the small role they play in the credit provision to Belgian firms
(see Table A1 of the online appendix), or because these foreign branches might themselves also face capital requirements of a similar
magnitude (which we cannot observe as they would be set by a foreign regulator).
4. Regulatory capital and credit supply: documenting channels
So far, we have focused on the average effect of (required) capital on bank credit supply. However, this relationship need not be
homogeneous, but may vary with bank or firm characteristics. First of all, in Subsection 4.1, we test the hypothesis that frictions or
costs associated with raising equity affect the extent to which capital requirements impact credit supply. Subsequently, we test the
hypothesis that bank business models impact the way capital requirements restrict credit supply (Subsection 4.2). Finally, in
Subsection 4.3, we provide some insight into which firms will be affected more by banks' goal to reduce risk-weighted assets in
response to higher capital requirements. We thus test the hypothesis that banks faced with higher capital requirements treat all
borrowers equally or not.
10
4.1. Factors affecting banks' cost of capital
The Modigliani and Miller paradigm stipulates that a firm's capital structure is irrelevant for its operating decisions. In the
banking context, this would imply, for instance, that the rate a bank charges on its loans, as well as the volume of loans should be
independent of its funding structure, including its leverage and capital ratio. However, the real world may deviate in various ways
from the theoretical Modigliani-Miller set-up (see Kashyap et al. 2010 for an overview). Consequently, in the short run, a phasing of
higher capital requirements might prompt banks to meet them by contracting credit supply (lower volume and higher lending rates),
rather than issuing equity or increasing retained earnings though reductions in dividends. Unfortunately, we cannot directly test
deviations from the Modigliani-Miller world. However, we will test them indirectly by analyzing whether the established negative
relationship between regulatory capital ratios and credit supply to corporations is less pronounced for banks whose equity financing
is relatively cheaper, or that have less need to adjust equity.
First of all, we assess whether the relationship established in Section 3 varies with monetary policy. The monetary policy stance
directly affects the cost of and access to funding by banks as well as the market conditions for raising equity. Note that the period
under analysis is an exceptional period in terms of monetary policy conditions and interventions, affecting both credit supply as well
as banks' characteristics. The European Central Bank has been trying to stimulate bank lending with a series of unconventional
monetary policy measures. While we have been controlling for monetary policy (and other macro-economic conditions) by means of
time-fixed effects,(un)conventional monetary policy may also interact with (micro)prudential policy in influencing banks' lending
behavior.6 To that end, we introduce an interaction term between the required capital ratio and the quarterly growth in the balance
sheet of the European Central Bank (ECB). The growth of the ECB's balance sheet is used as a proxy for the monetary policy stance, as
both conventional and unconventional monetary policy may lead to change in the volume of assets held by the ECB.
Subsequently, we test the conjecture whether the impact of changes in required capital on the supply of credit to corporations
varies with bank-specific characteristics affecting banks' cost of equity. We assume that the negative relationship between the re-
quired capital ratio and credit supply should be more pronounced for small and risky banks, given that, in general, the cost of capital
is lower for larger and safer banks (Kashyap et al. (2010), Baker and Wurgler (2015), and Gandhi and Lustig (2015)). In the tests, we
include the natural logarithm of total assets (bank size) and the ratio of loan loss provisions to total assets (risk), to test the afore-
mentioned conjectures. Furthermore, the cost of capital is lower for more profitable banks. Profitable banks have a larger franchise
value, which reduces their incentives for excessive risk-taking. Moreover, they have more scope for internal capital generation by
retaining earnings and, hence, can manage capital passively (as opposed to active management via raising equity externally, see e.g.,
De Jonghe and Öztekin (2015)). Hence, we expect the impact of required capital on credit supply to be weaker for banks with a
higher return-on-equity (proxy for bank profitability). Finally, banks with higher past equity growth have built up some financial
slack and have been able to time the market/regulator. This may not necessarily affect the cost of raising equity, but reduces the need
or urgency to do so. We thus expect that the negative relationship is less pronounced for these banks.
We test the aforementioned hypotheses using the following regression framework:
MP BC
Credit growthb, f , t = ( 1 + 1 Monetary polt 1 + 1 Bank charb, t 1) Required capital ratiob, t 1
+ 2 Previous year required capital ratiob, t 5 + 3 Actual capital ratiob, t 1

+ 4 Previous year actual capital ratiob, t 5 + Bank controlsb, t 1
+ b + f , t + b, f , t (2)
The specification is identical to Eq. (1), except for the interaction coefficients β1MP and β1BC, which stand for the interactive effect
on credit growth of the required capital ratio and, respectively, monetary policy or a bank characteristic on credit growth. The results
of these regressions are reported in Table 5. The monetary policy indicator and the bank characteristics have been standardized to
facilitate comparison of their economic magnitudes.
In the first column, we find that monetary policy affects the relationship between capital requirements and credit supply. The
interaction term is negative and significant, implying that an expansion of the European Central Bank's balance sheet will have a
weaker impact on credit supply for banks with higher capital requirements. The results in the first column thus indicate that there
might be a trade-off between (micro-)prudential capital requirements and monetary policy. An alternative way of interpreting the
negative interaction coefficient is that tightening capital requirements during expansionary monetary policy periods is more detri-
mental for credit supply, compared to a similar increase during monetary tightening. A one standard deviation increase in the growth
rate of the ECB's balance sheet raises the point estimate of the impact of required capital on credit supply from −0.201 (= 1) to
MP
−0.362 (= 1+ 1 ). Likewise, a one standard deviation decrease in the growth rate of the ECB's balance sheet results in an almost
MP
zero impact of required capital on credit supply ( 1− 1 =−0.04, to be precise). While this result is in contrast to our initial
hypothesis, it may be rationalized by the zero lower bound and the pressure that these unconventional policies put on banks' profit
margins.
In subsequent columns, we add an interaction term for the required capital ratio and the log of total assets (size), loan loss
provisioning ratio (risk measure), return on equity (scope for earnings retention), growth in common equity (market timing) and a
dummy that is one for banks in the lowest quartile of the capital buffer distribution (capital buffer = actual-required). This coincides
6
Theoretical models on the interaction of capital requirements and monetary policy have been developed by e.g., Angeloni and Faia (2013),
Angelini et al. (2014) and Du and Miles (2014), while the relationship has been empirically tested by Aiyar et al. (2016)
11
Table 5
Required capital and credit supply: banks' cost of capital.
Variables Credit growth Credit growth Credit growth Credit growth Credit growth Credit growth
Required capital ratio −0.201** −0.184* −0.267** −0.202* −0.211* −0.251**

(0.092) (0.100) (0.095) (0.101) (0.100) (0.099)
Required capital ratio × monetary policy −0.161** −0.163** −0.145 −0.156** −0.191** −0.214**
(0.075) (0.074) (0.083) (0.067) (0.078) (0.078)
Required capital ratio × Bank characteristic 0.058* −0.045* 0.111* 0.123*** 0.234**
(0.029) (0.024) (0.062) (0.038) (0.104)
Previous year required capital ratio −0.284*** −0.287*** −0.286*** −0.297*** −0.314*** −0.472***
(0.055) (0.054) (0.051) (0.060) (0.063) (0.096)
Actual capital ratio −0.219** −0.217** −0.211** −0.209** −0.254** −0.295**
(0.094) (0.095) (0.095) (0.089) (0.106) (0.101)
Previous year actual capital ratio −0.031 −0.029 −0.052 −0.015 −0.028 −0.075
(0.054) (0.054) (0.051) (0.050) (0.047) (0.070)
(Lagged) Bank characteristic Log total assets Loan loss provisions Return on Quarterly growth Small buffer
to total loans equity in common equity
Observations 1,022,297 1,022,297 1,022,297 1,022,297 1,022,297 1,022,297

R-squared 0.47 0.47 0.47 0.47 0.47 0.47
Bank controls Yes Yes Yes Yes Yes Yes
This table contains estimation results from a regression relating quarterly growth in authorized credit to (regulatory) capital ratios. The required
capital ratio also enters in interaction with a bank characteristic related to a bank's cost of capital. We add these interaction terms one-by-one in
subsequent columns. More specifically, we run the following regression:
Credit growth (Quarterly)b, f, t = (β1 + β1MP∗ Monetary policyt + β1BC ∗ Bank characteristicb, t−1) ∗ Required capital ratiob, t−1 + β2 ∗ Previous year
required capital ratiob, t−5 + β3 ∗Actual capital ratiob, t−1 + β4 ∗Previous year actual capital Ratiob, t−5 + γ ∗Bank controlsb, t−1 + νb + νf, t + ϵb, f, t.
Besides the variables of interest (of which the coefficients are reported), the equation includes control variables (similar to the baseline regression)
that have been lagged by one quarter. νf, t is a firm × time fixed effect that captures time-varying firm demand shifters. νb is a bank fixed effect. We
indicate which bank characteristic enters as interaction term in the row following the interaction term. They are, respectively, a proxy for bank size,
bank credit risk, bank profits, equity growth and a dummy if a bank's regulatory capital buffer is small. Standard errors are clustered at bank level.
with a buffer of less than 3%. Each interaction term enters as a one-quarter lag. The point estimates of the interaction terms provide
support for each of the aforementioned hypotheses. First of all, firms are more shielded from a reduction in lending due to higher
capital requirements at larger banks.7 We also find that more risky banks reduce lending to a larger extent, in response to an increase
in required capital.8 This provides support for the hypothesis that the cost of equity (which is higher for riskier banks) is the
constraining factor leading to negative effects of capital on lending.
Second, more profitable banks constrain credit supply less in response to higher capital requirements. More profitable banks can
use passive capital management (earnings retention) to restore their capital buffers and can protect their borrowers more. Third, we
also find that the negative impact is reduced for banks that expanded their capital buffer in the previous period. Indeed, the in-
teraction term on equity growth is positive and significant. Finally, in the last column, we find that banks with a smaller buffer reduce
credit supply less for a given increase in required capital, compared to banks with a larger buffer. A possible explanation for this
finding might be that banks with a low buffer that face a higher requirement may act via increasing equity. By relying on equity
issuance rather than asset-shrinking, the bank would be able to more swiftly restore the buffer (or even increase it). In contrast, when
the buffer is large, banks may opt for a more gradual restoration of its buffer (when requirements increase) by lowering credit growth.
What do these point estimates imply in economic terms? From an economic point of view, the obtained coefficients imply that a
BC
1 pp. increase in required capital reduces credit supply by 0.24 pp. for small banks (one standard deviation below the mean, 1− 1 ),
BC
and only 0.125 pp. for large banks (one standard deviation above the mean, 1+ 1 ). The same firm's credit supply is reduced by
almost twice as much at small banks facing a higher capital requirement than at large banks facing exactly the same capital re-
quirement increase (in the same quarter). The impact of heterogeneity in credit risk across banks is of similar economic magnitude.
The effects are larger, in economic magnitude, when looking at high and low profits (losses) or fast and slow equity growth (as the
standardized point estimates are larger). The implied effect of a 1 pp. increase in required capital for low-profit banks (one standard
deviation below the mean) is −0.313 pp., whereas it is 0.091 pp. for highly profitable banks (mean plus one standard deviation).
7
Aiyar et al. (2016) also find a positive interaction effect, though it is statistically insignificant.
8
Our measure of credit risk, i.e., the loan loss provisioning ratio, is an indicator of the quality of the existing loan portfolio. Our finding is in line
with the theoretical prediction of Bahaj et al. (2016) that lending is less sensitive to a change in capital requirements when legacy assets are healthy.
12
Concerning equity growth, we find similar economic effects of fast equity growth, as for highly profitable banks. Turning to the last
column, in economic terms, the impact of higher capital requirements on credit supply is almost non-existent for banks with a buffer
belonging to the lowest quartile of the distribution.
4.2. Factors affecting banks' business model
All banks in the sample are commercial banks. Yet, commercial banks may have different business models. The specific hypothesis
being tested in this subsection is that banks with features associated with a wholesale commercial bank business model transmit
higher capital requirements less into credit supply than retail banks. The latter have less flexible sources of funding and less di-
versification alternatives. Therefore, when confronted with higher capital requirements, they are more constrained in their actions
and will squeeze borrowers more. In our analysis, we do not classify banks according to a business model typology,9 but rather
investigate the role of variables associated with retail versus wholesale commercial bank types. We therefore include interaction
effects between required capital and four characteristics featured in banks' business models. These four characteristics are: bank asset
growth, the reliance on stable deposit funding, the importance of wholesale and interbank funding, and the importance of non-
interest income sources. In addition to being the driving forces behind bank business model clusters, these factors have been shown to
be defining characteristics of bank performance (Fahlenbrach et al. 2012). The advantage of our approach over a classification by
bank types is that we can pinpoint which specific aspect could create heterogeneity (and to what extent) in the relationship between
required capital and corporate credit supply.
We again use Eq. (2) to estimate these relationships and we report the results in Table 6. In contrast to variables related to banks'
cost of financing, we hardly find any support for the hypothesis that business model characteristics matter for the transmission of
capital requirements to credit supply. Heterogeneity among banks in asset growth does not affect how capital requirements affect
credit supply. Furthermore, the composition of banks' funding also does not create heterogeneities in banks' responses to higher
capital requirements. The interaction effect is not significant, neither for banks with more stable funding (higher share of demand and
savings deposits) nor for banks with a strong reliance on interbank funding. The only significant interaction term in Table 6 is
between required capital and the share of interest income in total income, which is negative and significant. Banks that are more
dependent on typical intermediation income reduce credit supply more in response to tighter capital requirements. Banks with a
lower share of interest income in total income seem to shield the corporate lending market and potentially curb their exposures in
markets generating non-interest income. In economic terms, a bank with a high (low) reliance on interest income (one standard
deviation above (below) the mean) reduces credit supply by 0.25 pp. (0.083 pp) in response to a 1 pp. increase in capital require-
ments.
Comparing the results in Table 5 and Table 6, we find that characteristics related to banks' cost of financing are a more likely
source of heterogeneity in the impact on credit supply, than characteristics related to banks' business models. Hence, the channel
through which bank capital requirements affect lending is through banks' financing constraints.
4.3. Borrower heterogeneity
Thus far, we have shown that higher capital requirements lead to lower credit supply and that banks for which the cost of capital
is higher will constrain credit supply to a larger extent. To meet the higher required capital ratio, banks reduce lending to adjust their
actual risk-weighted capital ratio. As not all firms are equal in terms of impact on risk-weighted assets, we should expect banks to
treat borrowers differentially. In this subsection, we thus test the hypothesis that banks discriminate between borrowers in con-
straining credit, in response to increased capital requirements.
Cutting credit more to larger firms (ln of total assets) and riskier firms (measured by Altman Z,10 financial leverage or col-
lateralization rate) will enable banks to adjust their volume of risk-weighted assets more swiftly. Furthermore, if banks have to cut
credit, they might be less likely to cut it for firms that pay higher interest rates (cost of borrowing is measured as total financial
expenses over total loans), as this protects bank profits and leaves scope for earnings retention.
Next, we also include an interaction term with firm age. The predicted sign for this variable is ambiguous because of two opposite
forces. Older firms are, on the one hand, on average less risky (survivorship bias), but, on the other hand, they are larger and also
have larger loans (because firms grow over the life cycle). The pairwise correlations between firm age, on the one hand, and firm size,
Altman Z and financial leverage, on the other hand, are respectively, 0.36, 0.09 and − 0.18. Finally, we measure the relationship
length between a bank and a firm (in months) and expect the effect to be smaller for firms with longer relationships with their banks,
as these firms are less risky (stable access to financing) and pay higher interest rates (locked into their relationship). Characteristics
on Belgian corporations are obtained via filings of their balance sheets and income statements to the NBB. The average firm, in our
sample, is nineteen years old (standard deviation of 12), with 3,187,731 euro in assets (standard deviation is 8,909,007 euro). The
9
Note that a thorough business model analysis only became part of the SREP Guidelines as of 2014 (EBA/GL/2014/13) and was thus not yet in
place during our sample period. Moreover, even now, there is still no consensus on which methodology to use to identify bank business models
(Cernov and Urbano 2018).
10
We employ the Altman Z for privately held corporations, which is a linear combination of five financial ratios. More precisely, Altman
Z = 0.717 x Working capital to total assets +0.847 x Retained earnings over total assets +3.107 x EBIT over total assets +0.42 x Book value of
equity to total liabilities +0.998 x Operating revenues over total assets.
13
Table 6
Required capital and credit supply: banks' business models.
Variables Credit growth Credit growth Credit growth Credit growth
Required capital ratio −0.207* −0.078 −0.170* −0.167*

(0.105) (0.123) (0.093) (0.086)
Required capital ratio * Monetary policy −0.166* −0.252** −0.190** −0.197**
(0.082) (0.089) (0.083) (0.074)
Required capital ratio * Bank characteristic 0.008 −0.002 −0.011 −0.084**
(0.053) (0.031) (0.068) (0.035)
Previous year required capital ratio −0.287*** −0.364*** −0.252*** −0.274***

(0.058) (0.104) (0.062) (0.043)
Actual capital ratio −0.221** −0.269** −0.202** −0.215**
(0.096) (0.114) (0.093) (0.088)
Previous year actual capital ratio −0.032 −0.067 −0.035 −0.039
(0.052) (0.074) (0.064) (0.047)
(Lagged) bank characteristic Quarterly growth in assets Retail deposit share Interbank funding share Interest income share
Observations 1,022,297 1,022,297 1,022,297 1,022,297

R-squared 0.47 0.47 0.47 0.47
Firm*time fixed effects Yes Yes Yes Yes
Bank fixed effects Yes Yes Yes Yes
Bank controls Yes Yes Yes Yes
Cluster Bank Bank Bank Bank
capital ratio also enters in interaction with a bank characteristic related to a bank's business model. We add these interaction terms one by one in
subsequent columns. More specifically, we run the following regression:
Credit growth (Quarterly)b,f,t = (β1 + β1MP∗ Monetary policyt + β1BC ∗ Bank characteristicb, t−1) ∗ Required capital ratiob, t−1 + β2 ∗ Previous year
required capital ratiob, t−5 + β3 ∗ Actual capital ratiob, t−1 + β4 ∗ Previous year actual capital Ratiob, t−5 + γ ∗ Bank controlsb, t−1 + νb + νf, t + ϵb, f,
t.
that have been lagged one quarter. νf, t is a firm × time fixed effect that captures time-varying firm demand shifters. νb is a bank fixed effect. We
indicate which bank characteristic enters as interaction term in the row following the interaction term. They are, respectively, bank asset growth,
the share of retail deposits funding in deposits, the share of interbank funding and the share of interest income in total income. Standard errors are
clustered at bank level.
average financial leverage ratio in the sample is 26% (standard deviation is 22%). The average Altman Z score is 0.83, but has a
standard deviation of 1.27.
We estimate these interaction effects using the following specification, which is akin to Eq. (2):
MP FC
Credit growthb, f , t = ( 1 + 1 Monetary polt 1 + 1 Firm charf , t 1) Required capital ratiob, t 1
+ 2 Previous year required capital ratiob, t 4 + 3 Actual capital ratiob, t 1

+ 4 Previous year actual capital ratiob, t 5 + Bank controlsb, t 1
+ b + f , t + b, f , t (3)
The results of these regressions are reported in Table 7. The monetary policy indicator and the firm characteristics have been
standardized to facilitate comparison of their economic magnitudes. To begin with, we find that the impact of an increase in required
capital on credit supply is more negative for larger firms (measured in total assets). The economic magnitude of the size effect is large.
Credit growth will be substantially lower for larger firms that borrow from banks with higher capital requirements
FC
( 1 + 1 =−0.159–0.243 = −0.402), whereas small firms (one standard deviation below the mean) will have no (statistically)
significant effect on their credit growth (the hypothesis β1−β1FC=0 cannot be rejected for small firms).
With respect to the risk indicators, higher values of the Altman Z score or pledged collateral11 indicate less risk, whereas firms
with more financial leverage are riskier. The estimated coefficients thus indicate that riskier firms are penalized more in terms of
lower credit supply by banks with higher capital requirements. In the absence of actual (firm-specific) risk weights, this is the best
indication that banks shift their credit supply in order to reduce risk-weighted assets. The economic magnitude of the risk effects is
sizeable, though smaller than the size effect (for each of the three risk characteristics). Furthermore, the interaction term with a firm's
implied interest rate is positive and significant. The implied interest rate is computed as the ratio of firms' financial costs over the sum
of long- and short-term loans. The negative impact of higher capital requirements on credit supply is weaker for high versus low
implied interest rate firms. Firms with a one standard deviation higher (lower) cost of borrowing will have a 0.25 pp. (0.13 pp)
11
Pledged collateral is a dummy that is one if the ratio is in the highest quartile, and zero otherwise.
14
Table 7
Required capital and credit supply: heterogeneity due to firm characteristics.
Variables Credit Credit Credit Credit Credit Credit Credit
growth growth growth growth growth growth growth
Required capital ratio −0.159 −0.186* −0.187* −0.238** −0.193** −0.178 −0.327***
(0.105) (0.101) (0.101) (0.092) (0.089) (0.103) (0.093)
Required capital ratio × Monetary policy −0.178** −0.173** −0.174** −0.136* −0.158** −0.173** −0.119
(0.072) (0.073) (0.074) (0.072) (0.070) (0.073) (0.074)
Required capital × Firm characteristic −0.243*** 0.069** −0.103*** 0.320** 0.059*** −0.152*** 0.714***
(0.040) (0.029) (0.017) (0.137) (0.011) (0.025) (0.106)
Previous year required capital ratio −0.261*** −0.269*** −0.269*** −0.309*** −0.289*** −0.267*** −0.270***
(0.071) (0.065) (0.065) (0.072) (0.069) (0.068) (0.057)
Actual capital ratio −0.213** −0.216** −0.216** −0.230** −0.200** −0.215** −0.233**
(0.098) (0.097) (0.097) (0.103) (0.088) (0.098) (0.093)
Previous year actual capital ratio −0.027 −0.023 −0.023 −0.028 −0.019 −0.025 −0.038
(0.065) (0.060) (0.060) (0.055) (0.056) (0.063) (0.056)
(Lagged) firm characteristic Firm size Altman Z Financial Collateralization Cost of Firm age Relationship
leverage rate borrowing length
Observations 969,700 969,626 969,700 1,022,297 874,109 969,700 1,022,293

R-squared 0.47 0.47 0.47 0.47 0.46 0.47 0.47
Firm*time fixed effects Yes Yes Yes Yes Yes Yes Yes
Bank fixed effects Yes Yes Yes Yes Yes Yes Yes
Bank controls Yes Yes Yes Yes Yes Yes Yes
Cluster Bank Bank Bank Bank Bank Bank Bank
capital ratio also enters in interaction with various firm characteristics. We add these interaction terms one by one in subsequent columns. More
specifically, we run the following regression:
Credit growth (Quarterly)b, f, t = (β1 + β1MP∗ Monetary policyt + β1FC∗ Firm characteristicb, t−1) ∗ Required capital ratiob, t−1 + β2 ∗ Previous year
required capital ratiob, t−5 + β3 ∗Actual capital ratiob, t−1 + β4 ∗Previous year actual capital Ratiob, t−5 + γ ∗Bank controlsb, t−1 + νb + νf, t + ϵb, f, t.
that have been lagged by one quarter. νf, t is a firm × time fixed effect that captures time-varying firm demand shifters. νb is a bank fixed effect. We
indicate which firm characteristic enters as interaction term in the row following the interaction term. They are, respectively, a proxy for firm size,
Altman Z, firm financial leverage, the collateralization rate, firm cost of borrowing, firm age and the length of the bank-firm relationship. Standard
errors are clustered at bank level.
reduction in credit supply in response to a 1 pp. increase in capital requirements.

We find a negative interaction effect on firm age. Note that the predicted sign was unclear as older firms are, on average, both less
risky but larger. The size effect seems to dominate the risk effect as an increase in required capital leads to more constrained credit
supply for older firms, relative to younger firms. In economic terms, the impact is smaller than that found in the first column,
precisely because of the offsetting risk effect. While cutting lending more to larger firms enables swifter adjustment of the risk-
weighted capital ratio, there might also be another reason why banks restrict credit supply more to larger and older firms, in response
to higher capital requirements. Banks can simultaneously facilitate access to alternative sources of financing to larger and older firms,
via debt markets or syndicated loans. Unfortunately, data availability does not permit us to investigate this conjecture in further
detail.
Finally, the result obtained in the last column is inconsistent with the hypothesis that banks protect firms with longer relation-
ships more than relatively new borrowers. The dummy variable called Relationship Length, which equals one for borrowers in the
lowest quartile of the distribution (relationship length of 2.5 years or less) and zero otherwise, is positive and significant. In fact, the
results indicate that banks even increase credit supply (in response to higher capital requirements) to their fresh borrowers at the
expense of incumbent borrowers. While this finding may be counterintuitive, it could be due to other factors correlated with the
length of the bank-firm relationship. Borrowers with a short relationship may be younger and smaller firms, which are growing faster
and hence also need more credit.
In additional regressions, reported in Appendix A2, we analyze whether the effect is different for single-individual firms. In the
absence of access to a household credit register, this would be the closest equivalent to analyze the credit supply effects on
households. An important remark here is that even these single-individual firms operate under limited liability, as unlimited liability
firms do not have to file any annual accounts with the Central Balance Sheet Office. We find that the impact of an increase in capital
requirements reduces credit supply less for single-individual corporations (relative to all other firms). The impact is economically
sizeable and statistically significant (in line with the size effect) when included as the only interaction. However, when including an
interaction term between size (ln(total assets)) and required capital in this specification, we no longer find any statistically significant
effect for the single-individual NFCs, suggesting that it is the size, rather than the mere fact of being the only individual responsible
for the credit, that matters.
15
5. Extensions and robustness
In the previous two sections, we have shown that capital requirements affect credit supply and mentioned the channels through
which this happens. In this section, we conduct some further analysis on the baseline result, as reported in Table 3. In particular, in
Subsection 5.1, we conduct a robustness check that should further mitigate concerns about endogeneity of the set required capital
levels. Subsequently, we analyze whether banks anticipate the requirement and already alter their credit policy before the an-
nouncement (Subsection 5.2). In the same subsection, we also investigate non-linearities in the response. We will examine whether
the response differs when requirements are relaxed versus tightened. Finally, in the last subsection, we explore the robustness of our
results in two alternative samples (Subsection 5.3).
5.1. Exogenous capital requirements
Pillar 2 capital requirements are add-ons to the Pillar 1 requirements that cover for various risks (especially credit risk). As such,
they ought to be orthogonal to these risks and could really be considered as exogenous with respect to lending decisions. To further
alleviate concerns that changes in bank balance sheet characteristics and various risks are reflected in banks' individual Pillar 2
capital requirements, we follow a test procedure akin to De Marco and Wieladek (2016). In particular, we test whether bank balance
sheet variables that supervisors had access to at the time of the regulatory decision can statistically predict the regulatory capital
ratio. In a first-stage regression (of which the results are reported in Table A3 of the online appendix), we find little to no role for
balance sheet characteristics in explaining the set capital requirements. We include a large set of contemporaneous and/or lagged
bank balance sheet characteristics, as well as borrower risk (at a bank's portfolio level), and find that an F-test indicates that they are
jointly insignificant (bottom of the table). This indicates that the Pillar 2 capital requirements are exogenous to balance sheet
characteristics.
De Marco and Wieladek (2016) label the residuals from such a first stage the “non-balance-sheet-based” capital requirement (as
the residual is the capital requirement orthogonalized with respect to balance sheet information). In a second step, we then use these
“non-balance-sheet-based” capital requirement to verify if our main findings are altered. More specifically, we substitute in baseline
Eq. (1) the Required capital ratio with the residuals of column 4 of Table A3, which we label Residual required capital ratio. The
results are reported in Table 8. Compared to the main results, we find that, if anything, the results become (slightly) stronger, rather
than weaker. Compared to Table 3, the coefficients in Table 8 are slightly larger in absolute value, indicating that the economic
impact might even be underestimated without the orthogonalization.
The tests conducted in this subsection thus provide further evidence that we estimate and then quantify the causal impact of
capital requirements on various aspects of credit conditions.
5.2. Anticipation and asymmetry
The results from the previous subsection suggest that our main findings can be interpreted in a causal sense as Pillar 2 re-
quirements are exogenously set by the supervisor. Yet, banks may anticipate changes in the requirements and may already start
responding immediately, such that we may underestimate the full response. In fact, in Section 2, we explain that the SREP conducted
by the NBB consists of several steps. At some point in the process, typically around the beginning of the last quarter of the year, banks
may learn about possible future capital requirement decisions. At that stage of the SREP, they may find out through explicit or
implicit communication between management teams and supervisors/regulators. Therefore, we conduct a test to examine whether
there is a significant pre-announcement effect in the fourth quarter. That is, we investigate whether credit supply in the fourth quarter
of year Y is not only driven by the capital requirement set and communicated at the beginning of year Y, but also by the capital
requirement that will be officially announced at the beginning of year Y + 1 (but may have already been anticipated by the bank
management team).
To that end, we add the following term to the baseline Eq. (1): β5 ∗ Required capital ratiob, t+1 ∗ I(Q4) and estimate the wider
specification. I(Q4) is a dummy that equals one in the fourth quarter only. This variable captures the response in the fourth quarter to
the capital requirement that will become effective in the next quarter. The results of this additional analysis are reported in the first
column of Table 9. We find that coefficient 5 is negative, statistically significant at the 1% level and larger (in absolute value) than
the point estimate 1 (effect of Required capital ratio). The latter seems to indicate that the leaking of the future capital decision in the
fourth quarter and the immediate response to that news is larger than the effect of the required capital ratio actually in place. To shed
further light on these differential magnitudes, we run an additional test, where we break down the effect of the Requir-
ed capital ratiob, t−1 into an effect in (i) the first quarter (immediate response to the new requirement), (ii) an effect in quarters two
and three (no news, nor leak) and (iii) an effect in the fourth quarter where this could be a response to either the actual requirement
announced in the first quarter of the ongoing year or a response to the new requirement which will be in place as of the first quarter of
the next year. We obtain a number of interesting findings from the regression results (reported in the second column).
In the first quarter, there is a significant and negative effect of the capital requirement on credit supply (point estimate of
−0.239). In the two middle quarters of the year, the effect is statistically not distinguishable from zero. Interestingly, in the fourth
quarter, credit supply is affected both by the capital decision made in the first quarter of the same year, as well as by the capital
decision that will be officially announced in the first quarter of the following year (but may have already been anticipated or leaked).
The point estimate of the former (−0.534) is nearly twice as large as the point estimate of the latter (−0.283). Hence, in terms of
16
Table 8
Capital (requirements) and credit supply: residual of required capital.
Variables (1) (2) (3) (4) (5) (6)
Credit growth Large drop New bank-firm Utilization rate Credit growth - Term Collateral rate
in credit relationships Loans
Residual required capital ratio −0.252* 0.366 −0.872*** 0.261* −0.578** 1.137**
(0.124) (0.305) (0.210) (0.133) (0.265) (0.477)
Previous year required capital ratio −0.275*** 0.403* −0.875*** 0.087 −0.334** −0.706
(0.065) (0.202) (0.185) (0.132) (0.133) (0.577)
Actual capital ratio −0.162** 0.448** 0.034 0.202** −0.313** −0.683
(0.064) (0.186) (0.086) (0.074) (0.143) (0.408)
Previous year actual capital ratio −0.023 −0.041 −0.061 0.005 0.017 0.176
(0.059) (0.071) (0.100) (0.073) (0.153) (0.210)
(Lagged) log total assets 0.012 −0.038** −0.361*** −0.003 0.008 0.134***
(0.014) (0.017) (0.085) (0.010) (0.020) (0.030)
(Lagged) quarterly growth in common equity −0.042** 0.045 0.010 0.047*** −0.091*** −0.023
(0.015) (0.025) (0.014) (0.009) (0.026) (0.035)
(Lagged) quarterly growth in deposits 0.007 0.014 −0.054** −0.016 −0.016 −0.059*
(0.006) (0.011) (0.021) (0.015) (0.014) (0.033)
(Lagged) quarterly growth in total assets −0.019 0.057*** 0.094*** 0.082*** 0.042 −0.090***
(0.013) (0.018) (0.027) (0.018) (0.027) (0.016)
(Lagged) quarterly return on equity 0.027 −0.038 −0.036 −0.009 0.031 −0.016
(0.019) (0.034) (0.033) (0.009) (0.025) (0.026)
Default probability −0.040*** 0.032* −0.015*** 0.117*** −0.034*** 0.291***
(0.006) (0.016) (0.004) (0.016) (0.009) (0.051)
Observations 1,022,297 1,022,297 1,067,376 1,067,376 577,073 1,022,297

R-squared 0.47 0.50 0.51 0.58 0.48 0.54
β1 + β3 −0.41 0.81 −0.84 0.46 −0.89 0.45

p-value of test (β1 + β3=0) 0.04 0.11 0 0.03 0.04 0.57
p-value of test (β1 + β2=0) 0.01 0.14 0.00 0.12 0.02 0.63
p-value of test (β3 + β4=0) 0.14 0.13 0.86 0.12 0.27 0.41
This table contains estimation results from a regression relating various dimensions of credit growth to (regulatory) capital ratios. Compared to the
baseline regression, we use the residual required capital ratio, which is the capital requirement orthogonalized with respect to current and lagged
balance sheet information. The results of the first stage regression to obtain the residuals are reported in the appendix (Table A3). In the second
stage, we run the following regression for six different dependent variables:
Credit conditions (Quarterly)b, f, t = β1 ∗ Residual required capital ratiob, t−1 + β2 ∗ Previous year required capital ratiob, t−5 + β3 ∗ Actual capital
ratiob, t−1 + β4 ∗ Previous year actual capital ratiob, t−5 + γ ∗ Bank controlsb, t−1 + νb + νf, t + ϵb, f, t.
Besides the variables of interest, the equation includes control variables, which are: the natural logarithm of total assets, quarterly growth in
common equity, quarterly growth in deposits, quarterly growth in total assets and quarterly return on equity. In all regresions, we also control for a
bank-specific assessment of the firm's default probability. All bank variables have been lagged by one quarter such that they are predetermined with
respect to next quarter's credit growth at the bank-firm level. In addition, we also control for a firm × time fixed effect (νf, t) that captures time-
varying firm demand shifters and a bank fixed effect (νb). Standard errors are clustered at bank level. At the bottom of the table, we also provide
information on the p-values of various tests on the coefficients of interest.
effective impact on credit supply, the requirement effectively in place in the fourth quarter has a twice as large effect on credit supply
as the requirement that will be implemented in the following quarter. In sum, these additional tests show that the actual effect of a
change in the capital requirement may be even larger and that the effect of the capital requirement on credit supply predominantly
materializes in the first and fourth quarter. This is indicative of both a swift response in the first quarter, combined with substantial
end-of-year changes.
In the last column of Table 9, we report the results of a test assessing whether the impact of actual and required capital on credit
supply is different when requirements increase strictly, rather than decrease. To that end, we add two additional variables to the
baseline specification and estimate:
17
Table 9
Capital (requirements) and credit supply: anticipation effects in Q4, timing and asymmetry.
Variables Credit growth Credit growth Credit growth
Required capital ratio −0.233** −0.590*

(0.085) (0.321)
Previous year required capital ratio −0.387*** −0.353*** −0.040
(0.086) (0.083) (0.257)
Actual capital ratio −0.164** −0.143** −0.045
(0.067) (0.066) (0.140)
Previous year actual capital ratio −0.081* −0.055 −0.076
(0.045) (0.045) (0.100)
Next quarter required capital ratio*I(quarter 4) −0.571*** −0.283***

(0.166) (0.086)
Required capital ratio*I(quarter 1) −0.239**
(0.095)
Required capital ratio*I(quarter 2 ∣ quarter 3) −0.016
(0.049)
Required capital ratio*I(quarter 4) −0.534***
(0.134)
Required capital ratio*I(Reqt−1 > Reqt−5) 0.310**
(0.111)
Actual capital ratio*I(Reqt−1 > Reqt−5) −0.182**
(0.066)
Observations 875,421 875,421 1,022,297

R-squared 0.47 0.47 0.47
Firm*time fixed effects Yes Yes Yes
Bank fixed effects Yes Yes Yes
Bank controls Yes Yes Yes
Cluster Bank Bank Bank
In this table, we examine some extensions of the baseline specification. In the first column, we examine whether there are anticipation effects of the
new capital requirement which already affect credit supply in the fourth quarter of the year before they become binding. In the second column, we
include these anticipation effects in a specification that further examines whether the impact of required capital on credit supply differs across the
various quarters of the year. Finally, in the third column, we explore the role of non-linearities by interacting actual and required capital with a
dummy that is one if the requirement has increased compared to the previous year. Besides the variables of interest (of which the coefficients are
reported), the equation includes control variables, which are: the natural logarithm of total assets, quarterly growth in common equity, quarterly
growth in deposits, quarterly growth in total assets and quarterly return on equity as well as firms' probability of default. All bank variables have
been lagged one quarter such that they are predetermined with respect to next quarter's credit growth at the bank-firm level. In addition, we add a
firm × time fixed effect that captures time-varying firm demand shifters and a bank fixed effect. Standard errors are clustered at bank level.
Credit condition (Quarterly)b, f , t = 1 Required capital ratiob, t 1

+ 3 Actual capital ratiob, t 1
+ 5 Required capital ratiob, t 1 I ( Reqt 1 > Reqt 5)
+ 6 Actual capital ratiob, t 1 I ( Reqt 1 > Reqt 5)
+ Bank controlsb, t 1 + b + f , t + b, f , t (4)
Including these two additional variables changes the interpretation of two other variables. First, coefficient β1 (β3) now indicates
the response of credit supply to the required (actual) capital ratio, when the required capital ratio decreases or remains constant.
Second, the sum of the coefficients β1 + β5 indicates the response of credit supply to the required capital ratio, when the required
capital ratio strictly increases. Third, the sum of coefficients β3 + β6 indicates the response of credit supply to the actual capital ratio,
when the required capital ratio strictly increases. We find that a reduction in the capital requirement increases lending ( 1 = 0.59).
An increase in the capital requirement reduces lending ( 1 + 5 = 0.59 + 0.31 = 0.28 ). However, in absolute value, the impact of a
reduction in capital requirements is thus larger than the impact of an equally large increase in the requirements. Moreover, we find
that the responsiveness of lending to actual capital changes is larger when banks see an increase in their capital requirement ( 3 = 0
and 6 < 0 ).
In sum, we do find some preliminary evidence of asymmetric adjustments. In general, the asymmetric response of lending to
capital has been investigated using actual capital rather than required capital. A notable exception is Imbierowicz et al. (2018), who
study balance sheet responses to both capital requirement increases and decreases using a sample of Danish banks over the period
2007–2014. To the best of our knowledge, there are no papers investigating asymmetric effects of capital requirements on lending
using micro data, which makes it difficult to benchmark our results. Nevertheless, the results on required capital are in line with the
18
arguments of De Jonghe and Öztekin (2015). They document that under-capitalized banks actively raise equity (active capital
management), whereas over-capitalized banks expand lending and reduce earnings retention (passive capital management). On the
one hand, when banks are over-capitalized (or face a reduction in the requirement), they will adjust by expanding their balance sheet
(holding equity constant). On the other hand, when they are under-capitalized (and especially when they face an increase in the
requirement) they will more likely adjust via active capital management (raising equity) rather than genuine downsizing. We leave a
more thorough analysis of asymmetric responses (and other non-linearities) for future research, ideally using a longer timespan and
wider range of banks (e.g., all banks supervised in the Single Supervisory Mechanism).
5.3. Opposite capital changes and single-bank borrowers
While we restrict the sample to multiple bank borrowers to better isolate supply from demand effects, it could still be that
borrowers match with specific banks. If firms borrow from multiple banks, but are similar in terms of (required) capital ratio, it may
be hard to identify statistical relationships. Therefore, we redo the analysis using a slightly different sample. More specifically, we
only keep firms in the sample if they borrow, in one and the same quarter, from at least one bank that experiences a negative change
in the required capital ratio, and at least one bank that experiences a strictly positive change in the required capital ratio. In panel A
of Table A4 in the online appendix, we report these results. The specification is similar to the baseline Eq. (1) but we thus impose an
additional constraint on the multiple borrower sample. This additional criterium implies that the sample size shrinks to about 70% of
the sample used in Table 3. Nevertheless, the results in both tables are very similar, except for the utilization rate regression. An
increase in required and/or actual capital ratios reduces credit supply. Moreover, capital requirements have a long-lasting effect.
In panel B of Table A4, we report results for a substantially larger sample, which has nearly three times as many observations as
the baseline results. The difference stems from the inclusion of single-bank borrowers. All regression specifications reported in
Table 3 include firm × time fixed effects, and are hence comparing credit supply to the same firm in the same quarter, by two banks
with different regulatory capital. However, this implies that one can only include firm-quarter observations if that firm borrows
simultaneously from at least two banks. The multiple-bank borrower sample consists of 1,022,324 bank-firm-quarter observations
involving 64,183 firms. Firms borrowing from more than one bank typically borrow from two banks, with a maximum of six banks.
However, in the Belgian context, the majority of firms borrow from a single banks. The full sample, including single-bank borrowers,
has 3,338,729 bank-firm-quarter observations covering 316,969 unique firms. Hence, only about a fifth of the firms borrow from two
or more banks in a given quarter, which explains why the average (median) firm in our sample has 1.373 (1) bank relationships.
While focusing on multiple-bank borrowers may have methodological merit, as it makes it possible to control for demand like Khwaja
and Mian (2008), it also implies a substantial reduction in the sample size, especially dropping smaller firms, potentially leading to
misguided conclusions for the entire universe of firms; i.e., single-bank and multiple-bank borrowers (see Degryse et al. (2019)).
Therefore, we also redo the analysis using the full sample. In order to mitigate concerns about confounding credit demand effects,
we now include a ‘group’ fixed effect to control for credit demand. The group is defined as the firm itself, in the case of a firm with
multiple bank relationships in a given quarter. The single-bank firms are, in each quarter, grouped on the basis of sector affiliation,
firm location, and size (hereafter referred to as ILS or industry-location-size). More specifically, these firms are grouped according to
the deciles of loan size in the Credit Register, the two-digit NACE code and the two-digit postal code (which broadly coincides with
the district level). A similar approach is used by Edgerton (2012), Morais et al. (2019), Degryse et al. (2019) and De Jonghe et al.
(2019). The results for the full sample estimation are reported in panel B of Table A4. In cases where the point estimate is significantly
different from zero, we find that the established relationships are quantitatively similar. That is, point estimates are by and large the
same in the full sample and the multiple-bank borrower sample. However, we do observe that some of the previously found re-
lationships are no longer statistically significant. This could be due to two reasons. On the one hand, it could be caused by only
imperfectly controlling for firm demand, if firms in an industry-location-sector group have differential demand shocks in a given
quarter. On the other hand, even if firms in such bins are homogeneous, the results could still differ if the characteristics of the firms
that borrow from a single bank are different from those of the multiple-bank firms. Two dimensions in which they may differ are firm
size and firm age. Degryse et al. (2019) show (in their Table 1) that single-bank firms are, on average, significantly smaller and
younger than multiple-bank firms. As we have shown that the impact of required capital on credit supply is firm size-dependent, it
may also explain the different effects in the multiple borrower sample versus the full sample.
6. Regulatory capital and balance sheet effects
Using detailed bank-firm level data on credit and applying micro-econometric techniques has many advantages, especially with
regard to identification and causality, but also with respect to contributions to the literature. One drawback, however, is that the
analysis is then focused on one lending market only, namely corporate lending, and also ignores other balance sheet changes banks
make in response to higher capital requirements. In this section, we empirically test how regulatory capital may affect the compo-
sition of the balance sheet. Balance sheet and income statement data come from (confidential) filings with the National Bank of
Belgium (i.e., Format A).
The empirical specification used to document the relationship between (required) regulatory capital and bank balance sheet
effects mimics that of Eq. (1), with the crucial difference that the level of observation is at the bank-quarter level:
19
Growth (Quarterly) of Xb, t = 1 (Residual) Required capital ratiob, t 1

+ 3 Actual rapital ratiob, t 1
+ Bank controlsb, t 1 + b + t + b, t (5)
The dependent variable, Growth (quarterly) of Xb, t, is the quarterly percentage change in a specific aggregate balance sheet item.
We relate quarterly growth rates in nine broad asset categories and six broad funding categories to lagged actual and required
regulatory capital ratios.12 The asset classes we consider are: cash interbank assets, mortgages, term loans (domestic), term loans
(foreign), other loans, securities, other assets, total assets (domestic) and total assets. Regarding banks' funding sources, we look at
quarterly growth in interbank liabilities, retail deposits, wholesale deposits, other debt, total liabilities and common equity. Summary
statistics on these variables are reported in Table 10. In panel A, we provide information on the growth rates, and on their share in
total assets in panel B. The four independent variables of interest are based on the regulatory capital data (and are defined as before).
The vector Bank Controlsb, t−1 consists of the following variables: bank size (natural logarithm of total assets), loans to total assets
ratio, loans to deposits ratio, off-balance sheet to total assets ratio, share of demand and savings deposits in total deposits, quarterly
return on equity, provisions to total loans, and the share of interest income in total income, which are all lagged by one quarter to
mitigate reverse causality concerns. νb is a bank fixed effect and νt is a time fixed effect. Among other things, the former is a crude
proxy for time-invariant heterogeneity in banks' borrower pools, which may create heterogeneous demand for that asset type;
whereas the latter accounts for general macro-economic factors.
Importantly, rather than estimating Eq. (5) separately for each of the fifteen growth rates, we estimate them as a system of
equations (that has a similar right-hand-side structure for each equation). In particular, using a seemingly unrelated equation esti-
mator makes it possible to take into account cross-equation (residual) correlation. Seemingly related regression estimation also allows
for more flexibility in the modeling of the variance/covariance matrix, such as clustering standard errors at bank level.
We report the estimation results in Table 11, in which each row corresponds to one regression in the system. The first column
mentions the dependent variable, whereas the other columns provide information on the coefficients of interest, the number of
observations and the R-squared. In the last column, we also report the p-value of a test of the hypothesis that β1+β3=0. We thus
again also test whether the impact of changes in regulatory capital requirements, when banks hold their buffer constant, is sig-
nificantly different from zero. The results on assets are reported in panel A of Table 11, while those on liabilities are in panel B.
When inspecting the results in panel A and focusing exclusively on the impact of the required capital ratio, we find that all but one
asset categories are reduced when capital requirements are higher, and significantly so for seven out of nine dependent variables. The
coefficients also allow for an easy comparison of the magnitude of the effects across asset classes. The three largest asset classes
(mortgages, domestic term loans, securities) represent 70% of the average bank's balance sheet (see panel B of Table 10). The effects
on the first two are relatively similar in magnitude and is slightly larger (in absolute value) for securities growth. The effects are much
more sizeable for foreign term loans and other loans, but these are much smaller asset categories. Consequently, the impact on
(domestic) total assets is in line with the estimated effects on the three most important asset categories.
Turning to bank funding, we find that an increase in capital requirements is associated with a reduction in both retail and
wholesale deposits. The effect on wholesale deposits is economically very large, but, for the average bank, this type of funding is
relatively unimportant (6.7%) vis-à-vis retail funding (66.7%). Higher capital requirements also lead to larger equity growth (but not
significantly so). Unfortunately, detailed information on dividend policies, earnings retention or active equity issuance are not
available, which limits the scope to analyze what drives equity growth and thus show which mechanisms banks use to build up their
capital levels. In sum, we find sizeable deleveraging in response to higher capital requirements.13
As regards the other independent variables, whenever significant, they have a similar sign to the required capital ratio. The results
discussed above pertain to a model where we use the residual required capital ratio (see Subsection 5.1). Results using the required
capital ratio are reported in the online appendix A5. Like the case of the microdata, the coefficients are slightly larger (in absolute
value) when using the residuals (i.e., the “non-balance-sheet-based” capital requirement).
To conclude this section, the aggregate balance sheet data analysis shows effects of (required) capital changes on bank activities.
While the results are intuitive and interesting, the set-up is not necessarily perfect. First, we examine broad asset class and hence
ignore the scope for heterogeneity within such an asset class for a given bank (a bank can substitute risky securities for safe securities)
or between banks (clientèle effects). Moreover, an analysis with aggregate data may suffer from other biases, such as imperfectly
controlling for unobserved firm demand or borrower quality, leading to biased estimates. The results in the last section thus mainly
12
As with the micro data, prior to running the main regression, we run auxiliary regressions to analyze the scope for collinearity due to correlation
between the actual and required capital ratio and/or correlation between the current and lagged values. Variance inflation factors (VIFs) are low and
range between 1.15 for the required capital ratio, and 1.47 for the previous year required capital ratio. These VIFs are sufficiently low not to have to
worry about possible consequences of collinearity.
13
Various editions of the Financial Stability Report issued by the National Bank of Belgium describe that there were no complementary measures
or deleveraging pressures/incentives during our sample period. Most, if not all, of the forced deleveraging following the global financial crisis of
2007–08 was finalized before the start of our sample period. The initial deleveraging (during 2008–2009) was mainly because of agreements with
the European Commission regarding state aid received by some of the major Belgian players. Compared to their European peers, Belgian banks
started the deleveraging process earlier and more extensively. The remaining minor deleveraging in 2012 and 2013 is mainly observed in the
derivatives portfolio (off-balance sheet), and thus does not affect our results.
20
Table 10
Summary statistics: aggregate balance sheet items.
Mean Standard deviation 5th Percentile Median 95th Percentile
Panel A: growth rates
Cash and interbank assets 0.016 0.284 −0.512 −0.002 0.492

Mortgages 0.027 0.066 −0.034 0.016 0.159
Term loans (domestic) 0.011 0.054 −0.066 0.010 0.064
Term loans (foreign) 0.022 0.218 −0.349 0.000 0.511
Other Loans 0.005 0.209 −0.288 0.001 0.503
Securities −0.004 0.059 −0.073 −0.007 0.112
Other assets −0.022 0.128 −0.304 −0.010 0.139
Total assets (domestic) 0.009 0.057 −0.059 0.006 0.049
Total assets 0.006 0.052 −0.068 0.005 0.074
Interbank liabilities −0.021 0.495 −0.794 −0.001 0.661
Retail deposits 0.022 0.102 −0.036 0.008 0.096
Wholesale deposits −0.000 0.153 −0.198 −0.009 0.262
Other debt −0.020 0.161 −0.300 −0.011 0.238
Total liabilities 0.006 0.055 −0.069 0.004 0.071
Common equity 0.006 0.064 −0.120 0.022 0.104
Panel B: share in total assets
Cash and interbank assets 0.121 0.125 0.014 0.101 0.533

Mortgages 0.244 0.185 0.025 0.220 0.636
Term loans (domestic) 0.204 0.119 0.006 0.196 0.403
Term loans (foreign) 0.034 0.044 0.000 0.009 0.143
Other loans 0.060 0.049 0.004 0.045 0.184
Securities 0.253 0.094 0.088 0.260 0.397
Other assets 0.091 0.064 0.019 0.060 0.199
Total assets (domestic) 0.734 0.180 0.429 0.673 0.975
Total assets 1.000 0.000 1.000 1.000 1.000
Interbank liabilities 0.109 0.196 0.002 0.069 0.876
Retail deposits 0.667 0.198 0.013 0.661 0.876
Wholesale deposits 0.067 0.082 0.001 0.035 0.276
Other debt 0.096 0.087 0.019 0.048 0.285
Total liabilities 0.940 0.021 0.919 0.938 0.970
Common equity 0.060 0.021 0.030 0.062 0.081
This table contains summary statistics on broad asset and liability classes. The quarterly growth rate is reported in panel A. The share in total assets
is reported in panel B. The total number of observations is, in general, 124, and are unbalanced over 12 quarters (2013Q1 to 2015Q4) and concerns
14 banks. However, the number of observations can be less than 124, as not all banks have each type of asset or liability. Data are winsorized at the
2% level.
serve the purpose of showing what happens in credit markets (other than those that serve NFCs), for which we unfortunately do not
have such granular data. An interesting avenue for further research would explore, within a given country and with micro data from a
household credit register, a corporate credit register and a securities register, how banks strategically behave in different markets in
response to higher capital requirements.
7. Conclusion
In general, macro-prudential policy has the explicit goal to safeguard the resilience of the financial system. One instrument at its
disposal to mitigate the impact of credit or financial cycles is the countercyclical capital buffer. The build-up of capital buffers during
booms provides financial room in downturns that helps mitigate credit crunches. Moreover, in some cases, higher capital require-
ments can also help to slow down credit booms, if banks internalize more of the potential social costs of defaults. Micro-prudential
capital requirements do not have the objective of affecting credit supply, but rather aim at enhancing the soundness and stability of
individual financial institutions. Nevertheless, micro-prudential capital requirements may also affect bank activity and lending, if
raising capital internally or externally is costly. Whether this is the case is a widely debated issue, as it would imply that there are
costs or frictions associated with bank capital that lead banks to pass up on otherwise profitable loans.
Using two alternative approaches, we document that higher capital requirements correspond to lower credit supply to corporations as
well as balance sheet adjustments. We also show that the effects are, first, less pronounced for banks for which the cost of capital (internally
or externally) is lower and, second, more prevalent for firms that facilitate swifter adjustments to banks' risk-weighted assets ratio.
Moreover, the heterogeneous treatment of borrowers by banks that face higher capital requirements are rational and may even be welfare-
improving by reducing credit less to safer firms as well as younger firms. Finally, we also find that the impact on aggregate corporate
lending is, in economic terms, moderate, suggesting that the impact on real activity might be limited.
Overall, the unintended consequences of micro-prudential capital requirements on credit supply are present but may be small,
21
Table 11
Capital requirements and growth of balance sheet items.
Equation Dependent variable Residual required One year lagged Actual One year lagged Observations R-squared p-Value
capital ratio required capital capital ratio actual capital ratio (β1 + β3=0)
ratio
Panel A: growth in asset types
(1) Δln cash and interbank assets 1.609 7.608*** 1.956** −0.163 124 0.48 0.26
(2.724) (0.792) (0.874) (0.972)
(2) Δln mortgages −0.853* −0.865** −0.806*** −0.472*** 124 0.33 0.02
(0.471) (0.388) (0.282) (0.173)
(3) Δln term loans (domestic) −0.896* 0.184 −0.021 −0.072 124 0.29 0.21
(0.495) (0.643) (0.432) (0.284)
(4) Δln term loans (foreign) −3.949** −5.125** −2.109*** −1.542 112 0.42 0.00
(2.010) (2.149) (0.697) (1.009)
(5) Δln other loans −3.962** −1.252 −0.981 −0.097 124 0.28 0.03
(1.756) (2.985) (0.927) (1.245)
(6) Δln securities −1.117*** −1.078* −0.685** −0.316 124 0.41 0.00
(0.377) (0.558) (0.280) (0.215)
(7) Δln other assets −1.911 3.309 −0.113 −0.160 124 0.40 0.27
(1.683) (2.036) (0.462) (0.442)
(8) Δln total assets (domestic) −0.773** 0.194 −0.045 −0.155 124 0.29 0.18
(0.376) (0.643) (0.316) (0.309)
(9) Δln total assets −0.976** 0.712 −0.341 −0.273 124 0.39 0.03
(0.487) (0.654) (0.273) (0.219)
Panel B: growth in funding
(10) Δln interbank liabilities 6.342 6.765*** 0.733 0.443 122 0.37 0.26
(6.269) (2.509) (1.670) (2.264)
(11) Δln retail deposits −1.173* −3.380 −0.831** −1.511 124 0.37 0.03
(0.680) (2.094) (0.414) (1.192)
(12) Δln wholesale deposits −5.046*** 0.131 −0.837 0.063 124 0.30 0.01
(1.738) (1.610) (0.896) (0.366)
(13) Δln other debt −1.772 1.884 −0.923** −0.926 124 0.38 0.04
(1.321) (1.563) (0.410) (0.621)
(14) Δln total liabilities −1.070** 0.782 −0.349 −0.281 124 0.39 0.02
(0.507) (0.677) (0.283) (0.234)
(15) Δln common equity 0.396 −0.031 −0.260* −0.270 124 0.49 0.69
(0.313) (0.414) (0.137) (0.219)
This table contains estimation results from a regression relating growth rates of bank asset or funding classes to (regulatory) capital ratios. We use
the residual required capital ratio, which is the capital requirement orthogonalized with respect to current and lagged balance sheet information
(see Table A3 of the appendix). The table consists of two panels in which each row depicts the results of an equation. The dependent variables in
panel A are growth rates of asset types, whereas they are growth rate of funding in panel B. More specifically, we run the following regression:
Growth (Quarterly) of Xb, t = β1 ∗ Residual Required capital ratiob, t−1 + β2 ∗ Previous year required capital ratiob, t−5 + β3Actual capital ratiob,
t−1 + β4 ∗ Previous year actual capital ratiob, t−5 + γ ∗ Bank controlsb, t−1 + νb + νt + ϵb, t.
Besides the variables of interest (of which the point estimates and standard errors are reported), the equation includes control variables as well as
bank and time fixed effects, i.e.,(νb) and (νt) respectively. Moreover, the 15 regressions are estimated as a system of equations in order to allow for
cross-equation error correlation. In addition, standard errors are clustered at bank level. In the last three columns, we report the number of
observations in the regression, the regression R-squared as well as the p-value of the test of a constant buffer effect. That is, we test whether the
impact of a joint increase in required and actual capital (thus holding the buffer between both constant) has a significant impact on the dependent
variable. We thus test the hypothesis: β1 + β3=0.
under the right conditions. For example, the required adjustments to bank capital were small and gradual, making it possible for
banks to meet them fairly smoothly. In addition, banks were dealt with individually, and additional requirements did not necessarily
correlate in time, so that the capacity existed both internally in the bank and externally in the markets and among firms to absorb the
increases in individual bank capital requirements. The fact that, despite this benign environment, our paper still provides new well-
identified evidence of an economically relevant impact of changes in bank capital requirements on bank lending to corporations and
other bank balance sheet items, should come as a poignant reminder to academics and policymakers alike not to overlook the possible
market frictions that banks face when raising new capital.
Finally, an interesting avenue for research is exploring the impact of these Pillar 2 requirements on credit supply in a multi-
country set-up. Since the introduction of the Single Supervisory Mechanism in 2014, the European Central Bank is in charge of the
micro-prudential supervision of significant financial institutions of the euro area. It could be interesting to explore whether country-
specific features are an additional source of heterogeneity (alongside bank and firm characteristics) in the impact of capital re-
quirements on credit supply.
22
Appendix
Table A1
Banks covered in the SREP vis-à-vis the other banks: comparing number of borrowers, volume of loans and assets.
O. De Jonghe, et al.
Year Quarter SREP banks Non-SREP banks Share of SREP banks in
Number of Aggregate firm-bank Aggregate Number Number of Aggregate firm-bank Aggregate Number All firm-bank Aggregate firm-bank Aggregate Total
firm-bank re- credit exposure total assets of banks firm-bank re- credit exposure total assets of banks relationships credit exposure total assets number
lationships (million euro) (million euro) lationships (million euro) (million euro) (million euro) (million euro) of banks
2013 Q1 298,956 96,922 628,056 11 83,013 32,622 214,481 17 0.78 0.75 0.75 0.39
2013 Q2 302,272 97,887 629,150 11 83,543 30,661 208,372 17 0.78 0.76 0.75 0.39
2013 Q3 302,126 97,033 613,582 11 83,879 30,800 203,970 17 0.78 0.76 0.75 0.39
2013 Q4 305,951 96,932 578,778 11 83,851 30,769 193,217 17 0.78 0.76 0.75 0.39
2014 Q1 371,011 120,623 751,995 14 15,342 6523 41,098 14 0.96 0.95 0.95 0.50
2014 Q2 369,473 120,984 767,089 14 15,012 6354 39,450 14 0.96 0.95 0.95 0.50
2014 Q3 366,790 121,526 778,424 14 16,697 6490 40,058 14 0.96 0.95 0.95 0.50
2014 Q4 373,748 133,141 777,075 14 16,581 6442 39,118 14 0.96 0.95 0.95 0.50
2014 Q1 370,489 134,097 821,347 14 16,765 6660 40,374 14 0.96 0.95 0.95 0.50
2014 Q2 369,956 136,132 788,704 14 16,811 6658 40,946 14 0.96 0.95 0.95 0.50
2014 Q3 370,060 136,919 782,849 14 16,878 6353 40,209 14 0.96 0.96 0.95 0.50
2014 Q4 369,152 141,002 754,160 14 16,731 6414 39,886 13 0.96 0.96 0.95 0.52
This table provides information on the number of borrowers, the total amount of corporate credit (in million euro), aggregate volume of total assets (in million euro) as well as the number of banks for two
groups of banks. In the left panel, we report the information for the banks covered in the SREP (and hence the sample used in this paper). The middle panel provides information for the other banks. In the
23
rightmost panel, we report the share of the “SREP” sample in the total sample for the number of borrowers as well as the volume of corporate credit and assets. The information is provided for each quarter
in the sample used in the analysis, running from the first quarter of 2013 to the last of 2015. The jump in the first quarter of 2014 is due to the inclusion in the SREP group as of 2014 of one of the four
large banks in Belgium (as well as two other smaller banks).
Journal of Corporate Finance 60 (2020) 101518
Table A2
Required capital and credit supply: robustness check on self-employed.
Variables Credit growth Credit growth Credit growth
Required capital ratio −0.159 −0.241** −0.168

(0.105) (0.110) (0.114)
Previous year required capital ratio −0.261*** −0.267*** −0.261***
(0.071) (0.066) (0.071)
Actual capital ratio −0.213** −0.215** −0.212**
(0.098) (0.097) (0.098)
Previous year actual capital ratio −0.027 −0.024 −0.027
(0.065) (0.062) (0.065)
Required capital ratio × monetary policy −0.178** −0.176** −0.178**

(0.072) (0.073) (0.072)
Required capital ratio × ln(Total assets) −0.243*** −0.239***
(0.040) (0.041)
Required capital ratio × I(Self-employed) 0.175** 0.031
(0.074) (0.078)
Observations 969,700 969,700 969,700

R-squared 0.47 0.47 0.47
Firm*time fixed effects Yes Yes Yes
Bank fixed effects Yes Yes Yes
Cluster Bank Bank Bank
This table contains the results of a robustness check concerning self-employed individuals with a registered
corporate activity. We regress quarterly growth in authorized credit on (regulatory) capital ratios and add an
interaction of required capital with firm size (column 1), a dummy if it concerns a firm with no additional
employees, in that case I(self-employed) = 1 (column 2) and the two aforementioned interactions jointly
(column 3). Besides the variables of interest (of which the coefficients are reported), the equation includes
control variables that have been lagged one quarter as well as the interaction between required capital and
monetary policy. The equation also includes a firm × time fixed effect that captures time-varying firm demand
shifters as well as a bank fixed effect. Standard errors are clustered at bank level.
Table A3
Residual required capital ratio: first stage orthogonalization.
Variables Required capital ratio Required capital ratio Required capital ratio Required capital ratio
Loans to total assets −0.062* −0.024

(0.035) (0.049)
Equity to total assets −0.011 −0.481
(0.121) (0.486)
Deposits to total assets 0.063* 0.047
(0.033) (0.050)
Interbank liabilities to total assets 0.024 0.015
(0.026) (0.048)
Interest income share 0.002 0.001
(0.008) (0.008)
Quarterly return on equity −0.001 −0.004
(0.006) (0.008)
Quarterly growth of total assets −0.013 −0.049
(0.015) (0.041)
Quarterly growth of equity 0.009 0.043
(0.013) (0.034)
Quarterly growth of deposits 0.002 −0.004
(0.006) (0.008)
Default probability −0.006 −0.020 −0.012
(0.030) (0.032) (0.034)
Lagged loans to total assets −0.073* −0.058
(0.037) (0.053)
Lagged equity to total assets −0.002 0.500
(0.104) (0.459)
Lagged deposits to total assets 0.064* 0.026
(0.037) (0.050)
Lagged interbank liabilities to total assets 0.020 0.008
(0.027) (0.045)
(continued on next page)
24
Table A3 (continued)
Variables Required capital ratio Required capital ratio Required capital ratio Required capital ratio
Lagged interest income share 0.003 0.006

(0.007) (0.008)
Lagged quarterly return on equity 0.002 −0.001
(0.007) (0.008)
Lagged quarterly growth of total assets 0.008 0.014
(0.013) (0.015)
Lagged quarterly growth of equity 0.000 0.002
(0.013) (0.015)
Lagged quarterly growth of deposits −0.003 −0.006
(0.005) (0.007)
Observations 128 128 128 128
R-squared 0.89 0.89 0.89 0.90
Bank dummies Yes Yes Yes Yes
F-stat 0.677 0.482 0.573
This table contains estimation results from a regression of the required capital ratio on a large set of bank characteristics. These regressions serve the
purpose of documenting that the Pillar 2 requirements are by and large independent of observed balance sheet characteristics. In column 2 (3), we
include contemporaneous (lagged) values of the bank characteristics, whereas in column 4 we include both sets jointly. The first column is a
regression of the capital requirements on bank fixed effects of which the R-squared serves as a benchmark for the additional explanatory power of
the bank characteristics. In the last row of the table, we report the p-value of an F-test for the joint significance of the included regressors. The
residuals of these regressions can then be considered as “non-balance-sheet-based” capital requirements and used to test whether the main results
are robust to this first-stage orthogonalization. This additional test should provide further support for the causality of the documented relationships.
Table A4
Required capital and credit supply: opposite capital change.
Variables Credit growth Large drop in credit New bank-firm relation- Utilization rate Credit growth - term loans Collateral rate
ships
Panel A: multiple-bank borrowers
Required capital ratio −0.163* 0.465 −0.677** 0.117 −0.524*** −0.101

(0.080) (0.301) (0.268) (0.290) (0.107) (0.513)
Previous year required −0.282*** 0.617*** −0.708*** 0.237 −0.333*** −0.687
capital ratio (0.058) (0.200) (0.218) (0.193) (0.090) (0.740)
Actual capital ratio −0.179*** 0.537*** 0.097 0.239* −0.274** −1.565**
(0.049) (0.158) (0.117) (0.128) (0.091) (0.568)
Previous year actual capital 0.018 −0.096 −0.046 −0.007 0.164 −0.228
ratio (0.030) (0.078) (0.083) (0.099) (0.101) (0.467)
Observations 713,294 713,294 744,624 744,624 388,411 713,294

R-squared 0.46 0.49 0.49 0.58 0.47 0.53
Sample Multiples Multiples Multiples Multiples Multiples Multiples
Panel B: single- and multiple-bank borrowers
Required capital ratio −0.217** 0.431 −0.899** 0.026 −0.244 0.287

(0.094) (0.257) (0.366) (0.201) (0.150) (0.513)
Previous year required −0.292*** 0.633** −0.944** 0.328 −0.256** −1.474
capital ratio (0.074) (0.229) (0.316) (0.273) (0.105) (0.855)
Actual capital ratio −0.169* 0.376* 0.185 0.073 −0.149* −1.143*
(0.079) (0.192) (0.128) (0.155) (0.079) (0.548)
Previous year actual capital −0.071 −0.029 −0.132 −0.177 −0.048 −0.150
ratio (0.087) (0.159) (0.153) (0.102) (0.175) (0.280)
25
Variables Credit growth Large drop in credit New bank-firm relation- Utilization rate Credit growth - term loans Collateral rate
ships
Observations 2,985,800 2,985,800 3,065,124 3,065,124 2,211,122 2,985,800

R-squared 0.18 0.26 0.16 0.37 0.21 0.30
Sample All All All All All All
ILS*time fixed effects Yes Yes Yes Yes Yes Yes
Compared with the table containing the baseline results (Table 3), we present results from using two alternative samples. In panel A, we only include
firm-quarter observations for firms that simultaneously borrow from a bank with a strict increase in required capital ratio as well as a bank with a
decrease in the required capital ratio. In panel B, we additionaly include single-bank borrowers. The latter firms are pooled into groups based on
industry, size and location and within each pool at least one firm is required to borrow from a bank with a strict increase in required capital ratio and
at least one firm must borrow from a bank with a decrease in the required capital ratio. For each of these samples, we re-run the baseline regressions
for six different dependent variables:
Credit conditions (Quarterly)b, f, t = β1 ∗ Required capital ratiob, t−1 + β2 ∗ Previous year required capital ratiob, t−5 + β3 ∗ Actual capital ratiob,
t−1 + β4 ∗ Previous year actual capital ratiob, t−5 + γ ∗ Bank controlsb, t−1 + νb + νf, t (or νILS, t) +ϵb, f, t.
Besides the variables of interest (of which the coefficients are reported), the equation includes control variables, which are: the natural logarithm of
total assets, quarterly growth in common equity, quarterly growth in deposits, quarterly growth in total assets and quarterly return on equity. In all
regresions, we also control for a bank-specific assessment of the firm's default probability. All bank variables have been lagged by one quarter such
that they are in principle predetermined with respect to the next quarter's credit growth at the bank-firm level. In panel A, νf, t is a firm × time fixed
effect that captures time-varying firm demand shifters. In panel B, which includes single bank borrowers, νILS, t is a ILS × time fixed effect that
captures time-varying firm demand shifters. ILS allows for the inclusion of single-bank borrowers, by creating groups (for each quarter separately)
based on firms their industry, size and location. νb is a bank fixed effect. Standard errors are clustered at bank level.
Table A5
Capital requirements and growth of balance sheet items: robustness.
Equation Dependent variable Required One year lagged re- Actual One year lagged Observations R- p-value
capital ratio quired capital ratio capital ratio actual capital ratio squared (β1 + β3=0)
Panel A: growth in asset types
(1) Δln cash and interbank assets 0.434 7.378*** 1.838** −0.196 124 0.48 0.47
(2.718) (0.886) (0.840) (0.947)
(2) Δln mortgages −0.555 −0.788** −0.771*** −0.459*** 124 0.33 0.06
(0.452) (0.394) (0.282) (0.159)
(3) Δln term loans (domestic) −0.592 0.264 0.016 −0.059 124 0.28 0.47
(0.465) (0.659) (0.444) (0.281)
(4) Δln term loans (foreign) −3.013 −4.850** −1.970*** −1.510 112 0.41 0.02
(1.998) (2.038) (0.765) (0.989)
(5) Δln other loans −3.743** −1.056 −0.909 −0.057 124 0.28 0.03
(1.518) (2.970) (0.920) (1.238)
(6) Δln securities −1.322*** −1.060** −0.687** −0.309 124 0.42 0.00
(0.408) (0.513) (0.282) (0.230)
(7) Δln other assets −1.946 3.384* −0.090 −0.142 124 0.40 0.23
(1.566) (2.053) (0.446) (0.432)
(8) Δln total assets (domestic) −0.915** 0.206 −0.046 −0.150 124 0.29 0.14
(0.384) (0.646) (0.318) (0.312)
(9) Δln total assets −0.920* 0.760 −0.324 −0.263 124 0.39 0.06
(0.526) (0.658) (0.280) (0.220)
26
Equation Dependent variable Required One year lagged re- Actual One year lagged Observations R- p-value
capital ratio quired capital ratio capital ratio actual capital ratio squared (β1 + β3=0)
Panel B: growth in funding
(10) Δln interbank liabillities 4.512 6.246*** 0.500 0.354 122 0.37 0.40
(6.125) (2.330) (1.662) (2.180)
(11) Δln retail deposits −1.224* −3.338 −0.819** −1.501 124 0.37 0.03
(0.677) (2.070) (0.410) (1.189)
(12) Δln wholesale deposits −4.215** 0.457 −0.702 0.122 124 0.29 0.02
(1.788) (1.599) (0.896) (0.325)
(13) Δln other debt −1.129 2.048 −0.847** −0.900 124 0.38 0.10
(1.239) (1.631) (0.418) (0.627)
(14) Δln total liabilities −1.004* 0.836 −0.330 −0.270 124 0.39 0.05
(0.555) (0.681) (0.289) (0.235)
(15) Δln common equity 0.398 −0.048 −0.266** −0.273 124 0.49 0.64
(0.267) (0.426) (0.135) (0.217)
This table contains estimation results from a regression relating growth rates of bank asset or funding classes to (regulatory) capital ratios. Unlike in
Table 11 in the paper, we now use the Required capital ratio, rather than the residuals of a first-stage regression. The table consists of two panels in
which each row depicts the results of an equation. The dependent variables in panel A are growth in asset types, whereas they are growth in funding
in panel B. More specifically, we run the following regression:
Growth (Quarterly) of Xb, t = β1 ∗ Required capital ratiob, t−1 + β2 ∗ Previous year required capital ratiob, t−5 + β3 ∗ Actual capital ratiob, t−1 + β4 ∗
Previous year actual capital ratiob, t−5 + γ Bank controlsb, t−1 + νb + νt + ϵb, t.
Besides the variables of interest (of which the point estimates and standard errors are reported), the equation includes control variables as well as
bank and time fixed effects, denoted νb and νt respectively. Moreover, the 15 regressions are estimated as a system of equations in order to allow for
cross-equation error correlation. Standard errors are clustered at bank level. In the last three columns, we report the number of observations in the
regression, the regression R-squared as well as the p-value of the test of a constant buffer effect. That is, we test whether the impact of a joint
increase in required and actual capital (thus holding the buffer between both constant) has a significant impact on the dependent variable. We thus
test the hypothesis: β1 + β3=0.
References
Admati, A.R., Hellwig, M.F., 2013. The bankers’ New Clothes: what’s Wrong with Banking and What to Do about it. Princeton University Press, Princeton NJ.
Admati, A.R., DeMarzo, P.M., Hellwig, M.F., Pfleiderer, P.C., 2013. Fallacies, irrelevant facts, and myths in the discussion of capital regulation: why bank equity is not
socially expensive. In: Stanford GSB Working Paper No. 2065.
Adrian, T., Boyarchenko, N., 2012. Intermediary Leverage Cycles and Financial Stability. vol. 567 Federal Reserve Bank of New York, Staff Reports.
Adrian, T., Shin, H.S., 2010. Liquidity and leverage. J. Financ. Intermed. 19, 418–437. https://doi.org/10.1016/j.jfi.2008.12.002.
Aiyar, S., Calomiris, C., Hooley, J., Korniyenko, Y., Wieladek, T., 2014a. The international transmission of bank capital requirements: evidence from the UK. J. Financ.
Econ. 113, 368–382. https://doi.org/10.1016/j.jfineco.2014.05.003.
Aiyar, S., Calomiris, C., Wieladek, T., 2014b. Does macro-prudential regulation leak? Evidence from a UK policy experiment. J. Money Credit Bank. 46, 181–214.
https://doi.org/10.1111/jmcb.12086.
Aiyar, S., Calomiris, C.W., Wieladek, T., 2016. How does credit supply respond to monetary policy and bank minimum capital requirements? Eur. Econ. Rev. 82,
142–165. https://doi.org/10.1016/j.euroecorev.2015.07.021.
Angelini, P., Neri, S., Panetta, F., 2014. The interaction between capital requirements and monetary policy. J. Money Credit Bank. 46, 1073–1112. https://doi.org/10.
1111/jmcb.12134.
Angeloni, I., Faia, E., 2013. Capital regulation and monetary policy with fragile banks. J. Monet. Econ. 60, 311–324. https://doi.org/10.1016/j.jmoneco.2013.01.003.
Ashcraft, A.B., 2006. New evidence on the lending channel. J. Money Credit Bank. 38, 751–775.
Auer, R., Ongena, S., 2016. The Countercyclical Capital Buffer and the Composition of Bank Lending. Mimeo.
Bahaj, S., Bridges, J., Malherbe, F., O’Neill, C., 2016. What determines how banks respond to changes in capital requirements? In: Bank of England Working Paper No.
593.
Baker, M., Wurgler, J., 2015. Do strict capital requirements raise the cost of capital? Bank regulation, capital structure, and the low-risk anomaly. Am. Econ. Rev. 105,
315–320. https://doi.org/10.1257/aer.p20151092.
Berger, A., Bouwman, C.H., 2013. How does capital affect bank performance during financial crises? J. Financ. Econ. 109, 146–176. https://doi.org/10.1016/j.jfineco.
2013.02.008.
Berger, A., Udell, G.F., 1994. Did risk-based capital allocate bank credit and cause a “credit crunch” in the United States? J. Money Credit Bank. 26, 585–628.
Bernanke, B., Lown, C., 1991. The credit crunch. Brook. Pap. Econ. Act. 22, 205–239.
Berrospide, J., Edge, R., 2010. The effects of bank capital on lending: what do we know, and what does it mean? Int. J. Cent. Bank. 6, 5–54.
Bridges, J., Gregory, D., Nielsen, M., Pezzini, S., Radia, A., Spaltro, M., 2014. The impact of capital requirements on bank lending. In: Bank of England Working Papers.
vol. 486.
Célérier, C., Kick, T., Ongena, S., 2016. Changes in the Cost of Bank Equity and the Supply of Bank Credit: When it Rains in Milano or Brussels, Does it Drizzle in
Frankfurt? Mimeo.
Cernov, M., Urbano, T., 2018. Identification of EU bank business models. In: EBA Staff Paper Series. vol. 2.
De Jonghe, O., Öztekin, Ö., 2015. Bank capital management: international evidence. J. Financ. Intermed. 24, 154–177. https://doi.org/10.1016/j.jfi.2014.11.005.
De Jonghe, O., Dewachter, H., Mulier, K., Ongena, S., Schepens, G., 2019. Some Borrowers are More Equal than Others: Bank Funding Shocks and Credit Reallocation.
Review of Finance (forthcoming).
De Marco, F., Wieladek, T., 2016. The real effects of capital requirements and monetary policy: evidence from the United Kingdom. In: CEPR Discussion Paper No.
DP11265.
Degryse, H., De Jonghe, O., Jakovljevic, S., Mulier, K., Schepens, G., 2019. Identifying credit supply shocks with bank-firm data: methods and applications. J. Financ.
27
Intermed. 40, 100813 (forthcoming).

Du, C., Miles, D., 2014. Interaction Between Monetary Policy and Regulatory Capital Requirements. Centre for Economic Policy Research DP10200.
Edgerton, J., 2012. Credit Supply and Business Investment during the Great Recession: Evidence from Public Records of Equipment Financing. Working Paper.
Fahlenbrach, R., Prilmeier, R., Stulz, R.M., 2012. This time is the same: using bank performance in 1998 to explain bank performance during the recent financial crisis.
J. Financ. 67, 2139–2185.
Fraisse, H., Lè, M., Thesmar, D., 2019. The real effects of bank capital requirements. Manag. Sci (forthcoming).
Freixas, X., Rochet, 2008. Microeconomics of Banking. MIT Press, Cambridge MA.
Galati, G., Moessner, R., 2013. Macroprudential policy: a literature review. J. Econ. Surv. 27, 846–878. https://doi.org/10.1111/j.1467-6419.2012.00729.x.
Gandhi, P., Lustig, H., 2015. Size anomalies in U.S. bank stock returns. J. Financ. 70, 733–768. https://doi.org/10.1111/jofi.12235.
Gornall, W., Strebulaev, I., 2018. Financing as a supply chain: the capital structure of banks and borrowers. J. Financ. Econ. 129, 510–530.
Gropp, R., Mosk, T., Ongena, S., Wix, C., 04 2018. Banks Response to Higher Capital Requirements: Evidence from a Quasi-Natural Experiment. The Review of
Financial Studies vol. 32, 266–299. doi:https://doi.org/10.1093/rfs/hhy052, arXiv:http://oup.prod.sis.lan/rfs/article-pdf/32/1/266/27185080/hhy052.pdf.
Hanson, S.G., Kashyap, A.K., Stein, J.C., 2011. A macroprudential approach to financial regulation. J. Econ. Perspect. 25, 3–28. https://doi.org/10.1257/jep.25.1.3.
Hubbard, G., Kuttner, K.N., Palia, D., 2002. Are there bank effects in borrowers’ costs of funds? evidence from a matched sample of borrowers and banks. J. Bus. 75,
559–581.
Imbierowicz, B., Kragh, J., Rangvid, J., 2018. Time-varying capital requirements and disclosure rules: effects on capitalization and lending decisions. J. Money Credit
Bank. 50, 573–602.
Jeanne, O., Korinek, A., 2013. Macroprudential Regulation Versus Mopping up After the Crash. National Bureau of Economic Research, WP, pp. 18675.
Jiménez, G., Ongena, S., Peydró, J.-L., Saurina, J., 2017. Macroprudential policy, countercyclical bank capital buffers, and credit supply: evidence from the spanish
dynamic provisioning experiments. J. Polit. Econ. 125, 2126–2177. https://doi.org/10.1086/694289.
Juelsrud, R., Wold, E.G., 2019. Risk-weighted capital requirements and portfolio rebalancing. J. Financ. Intermed (forthcoming).
Kashyap, A., Stein, J.C., Hanson, S.G., 2010. An Analysis of the Impact of ‘Substantially Heightened’ Capital Requirements on Large Financial Institutions. Mimeo.
Khwaja, A., Mian, A., 2008. Tracing the impact of bank liquidity shocks: evidence from an emerging market. Am. Econ. Rev. 98 (4), 1413–1442.
Labonne, C., Lamé, G., 2018. Credit Growth and Bank Capital Requirements Identification Using Supervisory Ratings. Unpublished manuscript.
Malherbe, F., 2015. Optimal Capital Requirements Over the Business and Financial Cycles. European Central Bank (WP1830).
Mésonnier, J.-S., Monks, A., 2015. Did the EBA capital exercise cause a credit crunch in the euro area? Int. J. Cent. Bank. 11, 75–117.
Mora, N., Logan, A., 2012. Shocks to bank capital: evidence from UK banks at home and away. Appl. Econ. 44, 1103–1119. https://doi.org/10.1080/00036846.2010.
537639. (arXiv:http://dx.doi.org/10.1080/00036846.2010.537639).
Morais, B., Peydro, J.L., Roldan-Pena, J., Ruiz Ortega, C., 2019. The International Bank Lending Channel of Monetary Policy Rates and QE: Credit Supply, Reach-for-
Yield, and Real Effects. J. Financ. 74, 55–90 (forthcoming).
Morrison, A.D., White, L., 2005. Crises and capital requirements in banking. Am. Econ. Rev. 95, 1548–1572.
Peek, J., Rosengren, E., 2000. Collateral damage: effects of the Japanese bank crisis on real activity in the United States. Am. Econ. Rev. 90, 30–45.
Puri, M., Rocholl, J., Steffen, S., 2011. Global retail lending in the aftermath of the US financial crisis: distinguishing between supply and demand effects. J. Financ.
Econ. 100, 556–578.
Shleifer, A., Vishny, R.W., 2010. Unstable banking. J. Financ. Econ. 97, 306–318. https://doi.org/10.1016/j.jfineco.2009.10.007.
Tirole, J., 2006. The Theory of Corporate Finance. Princeton University Press.
28

Bank capital (requirements) and credit supply

Uploaded by

Copyright:

Available Formats

Bank capital (requirements) and credit supply

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Bank capital (requirements) and credit supply

Uploaded by

Copyright:

Available Formats

Journal of Corporate Finance 60 (2020) 101518

Contents lists available at ScienceDirect

Journal of Corporate Finance

Bank capital (requirements) and credit supply: Evidence from pillar

ARTICLE INFO ABSTRACT

2. Pillar 2 capital requirements

Fig. 1. Diagram of the supervisory review and evaluation process.

3. Regulatory capital and corporate credit supply

3.1. Measuring credit conditions

Panel A: actual and required capital ratio

Required capital ratio 0.110 0.019 Not reported due to

Panel B: bank characteristics

ln(Total assets) 9.867 1.672 7.077 9.835 12.247

Panel A: credit growth

Authorized credit amount 1,022,297 644,563 7,749,937 2500 75,723 1,569,095

Panel B: bank characteristics (estimation sample, SREP banks)

Required capital ratio 1,022,297 0.109 0.015 Not reported due to

3.2. Methodology and hypotheses

Credit condition (Quarterly)b, f , t = 1 Required capital ratiob, t 1

interpreted in a causal sense.

Summary of the baseline hypotheses

3.3. Baseline results: loan level

Required capital ratio −0.129* 0.333 −0.808*** 0.285* −0.310** 0.526

Observations 1,022,297 1,022,297 1,067,376 1,067,376 577,073 1,022,297

β1 + β3 −0.27 0.77 −0.74 0.48 −0.57 −0.28

3.4. Aggregate results: firm level

SREP banks All banks Non-SREP banks SREP banks/All Banks

Required capital ratio −0.102*** −0.094*** 0.187 −0.016*

(Lagged) log total assets 0.010*** 0.010*** 0.007* 0.000

Observations 475,770 475,770 17,296 471,853

4. Regulatory capital and credit supply: documenting channels

4.1. Factors affecting banks' cost of capital

+ 2 Previous year required capital ratiob, t 5 + 3 Actual capital ratiob, t 1

Required capital ratio −0.201** −0.184* −0.267** −0.202* −0.211* −0.251**

Observations 1,022,297 1,022,297 1,022,297 1,022,297 1,022,297 1,022,297

4.2. Factors affecting banks' business model

4.3. Borrower heterogeneity

Required capital ratio −0.207* −0.078 −0.170* −0.167*

Previous year required capital ratio −0.287*** −0.364*** −0.252*** −0.274***

Observations 1,022,297 1,022,297 1,022,297 1,022,297

+ 2 Previous year required capital ratiob, t 4 + 3 Actual capital ratiob, t 1

Observations 969,700 969,626 969,700 1,022,297 874,109 969,700 1,022,293

reduction in credit supply in response to a 1 pp. increase in capital requirements.

5. Extensions and robustness

5.1. Exogenous capital requirements

5.2. Anticipation and asymmetry

Observations 1,022,297 1,022,297 1,067,376 1,067,376 577,073 1,022,297

β1 + β3 −0.41 0.81 −0.84 0.46 −0.89 0.45

Required capital ratio −0.233** −0.590*

Next quarter required capital ratio*I(quarter 4) −0.571*** −0.283***

Observations 875,421 875,421 1,022,297

Credit condition (Quarterly)b, f , t = 1 Required capital ratiob, t 1

5.3. Opposite capital changes and single-bank borrowers

6. Regulatory capital and balance sheet effects

Growth (Quarterly) of Xb, t = 1 (Residual) Required capital ratiob, t 1

Panel A: growth rates

Cash and interbank assets 0.016 0.284 −0.512 −0.002 0.492

Panel B: share in total assets

Cash and interbank assets 0.121 0.125 0.014 0.101 0.533

Panel A: growth in asset types

Required capital ratio −0.102* −0.094* 0.187 −0.016*

(Lagged) log total assets 0.010* 0.010* 0.007* 0.000

Previous year required capital ratio −0.287* −0.364* −0.252* −0.274*

Next quarter required capital ratioI(quarter 4) −0.571 −0.283*

Required capital ratio × monetary policy −0.178 −0.176 −0.178**

Required capital ratio −0.163* 0.465 −0.677 0.117 −0.524* −0.101

Required capital ratio −0.217 0.431 −0.899 0.026 −0.244 0.287