Available via license: CC BY 4.0
Content may be subject to copyright.
MACROPRUDENTIAL SUPERVISION AND AGENTS'
INFORMATION: WHAT STRESS TESTS REALLY TELL
THE MARKETS
FAUSTO PACICCO
*
, LUIGI VENA
†
and ANDREA VENEGONI
‡
School of Economics and Management
LIUC - Universit
aCarlo Cattaneo
Corso Matteotti 22, 21053 Castellanza(VA), Italy
*
fpacicco@liuc.it
†
lvena@liuc.it
‡
avenegoni@liuc.it
Received 24 September 2020
Revised 12 February 2021
Accepted 18 November 2021
Published 7 January 2022
Central bank's macroprudential supervisory activities have to ful¯ll three distinct tasks: (i)
assessing the banking system's vulnerability to exogenous adverse turbulence, (ii) evaluating
the risk of systemic crisis originating from idiosyncratic shocks, and (iii) measuring ¯nancial
market's sensitivity to policy stimuli. Given that macroprudential stress tests are the center-
piece of this policy approach, it is important to establish whether they are up to the task. We
study how the 2011–2018 European Banking Authority stress tests a®ected market risk per-
ception and show that they provided agents with valuable information on the policy stances and
the vulnerabilities of the banking system, carrying out the above tasks successfully, especially
the second and third tasks.
Keywords: Disclosure policy; macroprudential supervision; risk assessment; market reaction;
stress tests.
JEL Classi¯cation: C32, G12, G14, G28.
1. Introduction
The impact of the 2008 ¯nancial crisis depleted the toolkit of nearly all economic
policy authorities. Finding themselves in extreme di±culties, policy makers were
forced to devise new instruments and a renewed approach in order to calm the
turbulence and revive the economic systems. The common trait to all the new
†
Corresponding author.
This is an Open Access article published by World Scienti¯c Publishing Company. It is distributed under
the terms of the Creative Commons Attribution 4.0 (CC BY) License which permits use, distribution and
reproduction in any medium, provided the original work is properly cited.
Journal of Financial Management, Markets and Institutions
(2021) 2150009 (32 pages)
#
.
cThe Author(s)
DOI: 10.1142/S2282717X21500092
2150009-1
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
strategies designed after this shock is the central role played by communication. The
banking system's supervisory activity was no exception to this. Competent author-
ities worldwide were forced to adopt a new paradigm, a fundamental pillar of which is
macroprudential practices, in particular macroprudential stress tests. These consist
of periodic assessments of the banks' capital adequacy under simulated adverse ex-
ternal shocks. The main aim of this novel approach is to enhance the transparency of
the banking industry by disclosing information which fosters the system's soundness
and resilience. Its goal is to enrich market agents' information, based on three
dimensions: (i) a \stamina" dimension, which assesses the system's vulnerability to
external factors or market failures, (ii) a \contagion" dimension, which measures the
interconnections of banking institutions and the risk of systemic crisis originating
from idiosyncratic shocks, and (iii) a \structural" dimension, which describes how
receptive markets are to policy stimuli.
a
The aim of this paper, therefore, is to assess whether the supervisory actions
carried out by European authorities performing macroprudential stress tests e®ec-
tively yielded the expected results under all three targeted dimensions.
Since the very ¯rst introduction of these new instruments in the supervisory
authorities' toolkit, the debate over their e®ectiveness and the impact of their in-
formation content has been heated between both academics and practitioners. As
reported by Lazzari et al. (2017), the literature on the topic focuses principally on
two issues: the ¯rst, mainly theoretical, concerns the proper design and execution of
the assessments (Borio et al. 2014,Acharya et al. 2017) whereas the second, pre-
dominantly empirical, tests their e®ectiveness in achieving the stated goals of both
fostering transparency and accountability in the banking sector and supporting the
market in sorting good from troubled banks.
Our research concerns the second issue. To evaluate the impact of stress test
exercises, existing works mainly employ value relevance frameworks. In general,
putting aside the peculiarities of each analyses, the modus operandi is as follows: each
time abnormal returns in stock prices or CDS premia are detected, by means of an
event-study approach, the studied exercises are deemed to have produced fresh news
for the market (cf. Petrella & Resti 2013,Morgan et al. 2014,Alves et al. 2015,
Candelon & Sy 2015,Sahin & De Haan 2016,Ahnert et al. 2020, among others). This
approach, although rigorous and sound, provides a narrow perspective on stress
tests' disclosure e®ects, both time and scope-wise; in addition, it has as of yet proved
unable to provide conclusive evidence on the di®erent relevant dimensions related to
the impact of the stress testing procedure. Time-wise, the main limitation of these
studies lies in their short time-horizon as the methodology applied allows only the
immediate e®ects of stress tests on ¯nancial markets to be appreciated, leaving their
impacts in the medium-long run barely explored. Given that macroprudential
practices aim to stabilize the ¯nancial markets, at least in the medium run, it is
a
This taxonomy resembles that which appears on the European Central Bank (ECB) website (https://
www.ecb.europa.eu/ecb/tasks/stability/html/index.en.html) which allows the main goals of macro-
prudential supervision to be quickly and simply represented.
F. Pacicco, L. Vena & A. Venegoni
2150009-2
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
necessary to develop an empirical framework that allows the focus of the analysis to
be extended. Scope-wise, existing works are focused almost exclusively on the anal-
ysis of the \stamina" dimension, assessing stress tests ability to disclose idiosyncratic
information and distinguish sound from weak banks. Finally, the evidence gathered
by the existing empirical analyses provide mixed indications, preventing a full as-
sessment of their e®ective outcome on ¯nancial markets dynamics. Hence, there is a
need to extend the existing literature with a more comprehensive analysis, both in
terms of time and scope, in the meantime trying to contribute to the areas of the
debate where no conclusive evidence has been obtained. Indeed, the existing litera-
ture (i) fails to provide clear-cut results on stress tests' ability to disclose valuable
information to distinguish vulnerable from sound banks (the \stamina" dimension)
b
;
(ii) fails to ascertain if these exercises are e®ective in conveying information on the
contagion threat and thus fostering the overall stability of the system more than that
of the single bank; and (iii) fails to provide unequivocal evidence of their impact on
the \structural" dimension, i.e. their ability to provide agents with a comprehensive
overview of the regulatory approach in the banking industry.
Some recent works attempt to overcome these limitations by employing di®erent
frameworks but still limiting their perspective to speci¯c features without providing
a comprehensive view of their e®ectiveness (cf. Lazzari et al. 2017,Philippon et al.
2017). Our work aims to ¯ll these gaps and provide a global evaluation of macro-
prudential stress tests impact on the information set of ¯nancial agents, assessing
how the markets have interpreted the information conveyed by the supervisory
procedures held in the European Union (EU) to date. We therefore seek to answer
three speci¯c research questions related to each of the single dimensions targeted by
macroprudential policy conduct, as follows:
(1) Concerning the \stamina"dimension: Do macroprudential stress tests provide
information that help agents to discriminate between \sound" and \weak"
institutions?
(2) Concerning the \contagion"dimension: Does the information disclosed from
macroprudential stress tests help market participants assess the risk that a bank-
speci¯c shock would lead to a systemic crisis?
(3) Concerning the \structural"dimension: Are macroprudential exercises a source
from which agents can infer the general attitude (\structural" dimension) of the
supervisory authority(ies)?
We address these queries following an innovative approach that investigates
the information contribution of macroprudential stress tests in terms of their
impact on market agents' risk perception. That is to say, we study the behavior of a
market-based measure of risk in the period around the disclosure date of the stress
b
Some works provide evidence corroborating the idea that some exercises have achieved this task
(cf. Georgescu et al. 2017,Ahnert et al. 2020), whereas others demonstrate that market reaction was not
related to the banks' performance in the exercise (see Morgan et al. 2014,Candelon & Sy 2015,Carboni
et al. 2017).
Macroprudential Supervision and Agents' Information
2150009-3
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
test exercises carried out at the EU-wide level to date. Thus, we employ a state-space
representation of the Fama–French three-factor model, as in Pacicco et al. (2020),
augmented with a panel regression analysis that allows us to infer the factors that
drive the variations of market risk perception in the weeks around the disclosure
dates of the stress tests' results. Indeed, this approach provides insights on how these
procedures are able to steer agents' behavior and whether they succeed in doing so in
the intended directions, assessing their e®ectiveness in all three dimensions of the
new macroprudential supervisory approach.
Regarding the \stamina" dimension, we observe that agents adjust their risk
perception notwithstanding idiosyncratic information. Indeed, we observe no rela-
tionship between risk perception adjustments and the banks vulnerability resulting
from the regulatory assessment. This con¯rms the evidence of Lazzari et al. (2017),
Morgan et al. (2014) that the markets are able to distinguish between sound and
weak institutions before stress tests results are disclosed, which add little to their
knowledge; thereby reinforcing the claim that macroprudential stress tests results are
predictable as they are based on publicly available consolidated banks' balance sheet
data (Cerutti & Schmieder 2014,Glasserman & Tangirala 2016,Carboni et al. 2017,
Philippon et al. 2017,Ahnert et al. 2020).
Regarding the \contagion" dimension, our ¯ndings indicate that macro-
prudential stress tests are able to provide valuable information on the contagion risk
in the banking system. Indeed, the results show that stress tests have a risk curbing
e®ect for both adequately and inadequately capitalized banks as their market sco-
move along a decreasing path after the disclosure of stress tests results.
Finally, stress test exercises can be useful announcement tools which are able to
convey information on the perspective supervisory stance (\structural" dimension).
The fact that the betas co-move, regardless of the magnitude of the curtailment
su®ered, suggests that agents perceive from stress tests the overall attitude of su-
pervisory authorities towards the banking system and not speci¯c asset class-related
weights. This evidence is additionally corroborated by the analysis of the market risk
perception evolution of the portfolios of examined and non-examined institutions.
Our panel regression analysis shows that after results disclosure, the market sof
non-participating banks bene¯t from a contraction higher than that experienced by
those of participating institutions. This can be read as indicating that market agents
derive from these exercises valuable information on the direction of the supervisory
action that decreases their risk perception and that, in the case of non-examined
banks, is given for \free," without incurring the cost of disclosing any idiosyncratic
information.
2. The Information Contribution of Stress Tests
Since their very ¯rst application, the credibility as well as the e®ectiveness of mac-
roprudential stress tests in conveying fresh and valuable information to the markets
has been called into question (Ong & Pazarbasioglu 2014). Many empirical works
F. Pacicco, L. Vena & A. Venegoni
2150009-4
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
address these issues, but no conclusive evidence has been found (cf. Petrella &
Resti 2013,Morgan et al. 2014,Candelon & Sy 2015,Sahin & De Haan 2016,Lazzari
et al. 2017,Carboni et al. 2017,Flannery et al. 2017,Miani et al. 2018,Ahnert
et al. 2020, among others). This is because, on the one hand, the literature is still in
evolution and empirical settings are undergoing a re¯nement process, whereas, on the
other hand, the exercises carried out until now by the di®erent supervisory au-
thorities worldwide display considerable heterogeneity under many perspectives
(from design to execution, from communication to results disclosure). Regarding the
macroprudential exercises performed at the EU-wide level, a number of studies have
assessed their impact on the markets, providing mixed evidence. There is a sub-
stantial agreement that the 2011 stress test has been e®ective in conveying fresh news
to the market and that such news was \bad", at least regarding the results disclosure
(cf. Petrella & Resti 2013,Candelon & Sy 2015). Regarding the 2014 Comprehensive
Assessment (CA), which combines a stress test with an Asset Quality Review (AQR)
procedure, the evidence is mixed, even if the argument that it conveys valuable
information to the market appears to prevail (cf. Sahin & De Haan 2016,Carboni
et al. 2017,Lazzari et al. 2017,Jabbour & Sridharan 2020). Furthermore, in this case
the market appeared to have learnt \bad" news from the supervisory practice, as
around the results disclosure date the majority of banks, notwithstanding their
outcome in the exercise, displayed negative abnormal returns (Lazzari et al. 2017).
Studying abnormal stock and CDS variations for the 2016 European stress test,
Georgescu et al. (2017) provide evidence that the publication of stress test results
enhanced price discrimination as the impact on bank CDS spreads and equity prices
tended to be stronger for the weaker performing banks. Summarizing the above
evidence, there is a substantial consensus on the fact that macroprudential stress
tests are valuable information providers.
However, either because of the con°icting evidence or due to the limitations of
the empirical settings chosen, literature contributions do not allow a thorough as-
sessment, with a reasonable degree of accuracy, of the impact of the additional
transparency brought by macroprudential stress tests. More precisely, regarding the
tests impact on the banking system soundness (the \stamina" dimension), no un-
equivocal empirical evidence has been found as opposite conclusions have been
reached in empirical works analyzing the same stress test exercise (cf. Morgan
et al. 2014,Candelon & Sy 2015,Georgescu et al. 2017,Carboni et al. 2017).
Regarding the contagion and the \structural" dimensions, the empirical setups so
far adopted have not allowed an assessment of either of the two as the event study
setting is focused on short-term market repercussions and the analyses so far con-
ducted have nearly exclusively focused on bank-by-bank reactions rather than on
the aggregate e®ect of stress test exercises. Regarding the \structural" dimension,
Bischof & Daske (2013) use a di®erent empirical setting, taking as reference the
2011 EU wide stress tests, to show that a mandatory one time disclosure induces an
increase in voluntary disclosures about sovereign risk in the following periods, thus
Macroprudential Supervision and Agents' Information
2150009-5
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
enhancing the overall stability of the system. Nevertheless, this study focuses on
only one single exercise and on a speci¯c asset class, being too speci¯c to allow its
¯ndings to be generalized.
Hence, our work aims to ¯ll this gap by designing an empirical framework that is
able to simultaneously address all three dimensions of the macroprudential strategy
and answer the related research questions.
3. Hypotheses Testing
As stated in the introduction, the aim of this paper is to thoroughly assess the impact
of the information disclosure provided by the macroprudential stress testing proce-
dures carried out in Europe from 2011–2018, extending the existing literature in both
time and scope dimensions. Time-wise, our empirical framework allows us to
appreciate how stress tests information disclosure a®ects market agents' risk per-
ception on a time-horizon wider than simply the few days surrounding the infor-
mation release. Regarding the scope of the analysis, our empirical approach makes it
possible to evaluate the impact of the transparency shock generated by macro-
prudential stress tests under three di®erent perspectives, namely, the \stamina,"
\contagion," and \structural" dimensions.
To achieve these aims, the empirical strategy employed includes three stages:
.Stage 1. We segment our sample, dividing the banks in groups according to the
features that are relevant to test each hypothesis;
.Stage 2. For each exercise considered, we graphically compare the trends of the
market betas estimated for each portfolio;
.Stage 3. To integrate the graphical investigation, we perform a regression analysis
on individual banks data in which we analyze the coe±cients estimated for the
dummy variables that represent the portfolios segmentation. Regression models
are run on the whole sample of exercises considered (2011, 2014, 2016, 2018 stress
tests) to preserve the robustness of the framework and of the coe±cients estimated
and to avoid degrees of freedom issues.
Since the very ¯rst application of macroprudential exercises, empirical investigations
have focused on the assessment of the impact of such procedures information dis-
closure and, most of all, targeted their ability to integrate the information set of
market agents and enhance their capacity to discriminate between sound institutions
(i.e. those adequately capitalized even subsequent to a particularly dire adverse
shock) and weak ones. However, whereas some contributions support the idea that
macroprudential procedures achieve this goal (cf. Georgescu et al. 2017,Ahnert et al.
2020), others conclude that market reaction was unrelated to the performance of
banking institutions in the exercise (see Morgan et al. 2014,Candelon & Sy 2015,
Carboni et al. 2017). The absence of an unequivocal pattern leaves room for
F. Pacicco, L. Vena & A. Venegoni
2150009-6
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
additional investigation, thereby leading us to the development of our ¯rst hy-
pothesis as follows:
Hypothesis 1. Stress tests are e®ective in ful¯lling the \stamina" dimension of
macroprudential supervision, conveying information useful for distinguishing sound
from vulnerable banking institutions.
We test this hypothesis by grouping participating banks in quartiles by their CET1
capital ratio, resulting in the application of the adverse scenario in each exercise;
subsequently, we compare the evolution of the market measure of risk () of the four
portfolios. Analogies (di®erences) in the pattern of each quartiles' market betas may
signal that market agents do not (do) learn anything new about the relative
soundness of each banking institution and so they do not (do) draw valuable idio-
syncratic information from the stress tests. Following the same rationale, we perform
a regression analysis plugging three dummy variables which identify the quartiles'
portfolios from the ¯rst (best performing banks) to the third (the fourth is excluded
and represents the benchmark). Any signi¯cance in the coe±cients related to the
dummies would signal a divergent pattern between the market betas of a given
quartile portfolio compared to those of fourth quartile portfolio; therefore signaling
that macroprudential exercises e®ectively convey information that enhance the
agents' ability to assess single banks' resilience to possible future adverse shocks.
While being highly debated, the \stamina" dimension is not the only relevant
feature targeted by the new macroprudential approach adopted by supervisory au-
thorities. Indeed, another important aspect highlighted by the 2008 ¯nancial crisis is
the contagion e®ect, i.e. the transmission of an idiosyncratic shock to the overall
banking system. Even before the crisis showed how disruptive the impact of ¯nancial
contagion can be, the role of public information in enhancing the overall ¯nancial
stability was highly debated (cf. Goldstein & Sapra 2014,Alvarez & Barlevy 2015,
Goldstein & Leitner 2018). Nonetheless, whereas a number of theoretical works have
addressed this extremely relevant aspect of ¯nancial markets' regulation and su-
pervision, there is scant empirical evidence that enables us to e®ectively assess how
regulatory disclosure a®ects the overall ¯nancial system stability. We aim to provide
contributions in this direction by testing the following hypothesis:
Hypothesis 2. Macroprudential stress tests successfully address the \contagion"
dimension by providing information on the degree to which an idiosyncratic shock is
bound to spread throughout the banking system.
To analyze if the \contagion" dimension is adequately assessed by the stress testing
procedures, we compare the market s pattern of the banks divided in portfolios
according to whether or not they met the 5% CET1 capital threshold under the
adverse scenario hypothesis. We take as reference this capital threshold, even though
not all the exercises considered have explicitly set a pass/fail condition. Nonetheless,
given that is has been proved that market agents have gone through a learning
Macroprudential Supervision and Agents' Information
2150009-7
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
process in which they have periodically learnt how to deal with macroprudential
stress tests and how to interpret their outcome (cf. Glasserman & Tangirala 2016,
Luu & Vo 2020), it is likely that this threshold, even when not clearly determined,
resembles a \line of scrimmage" distinguishing \sound" from \weak" banks. Indeed,
the relative movement of the two groups of banks selected using this criterion gives
information on the perceived degree of contagion present in the banking system. If
the market s of the two portfolios show a similar increasing pattern, it would mean
that market agents believe that the weak institutions that have not met the 5%
CET1 capital ratio under the adverse scenario are bound to constitute a threat for
the overall system. Contrarily, a shared downward trend would mean that agents are
reassured by the identi¯cation of the weak banks and believe that the system and
authorities are su±ciently strong to address these weaknesses. Finally, divergent
patterns indicate that markets do not infer any systemic indication. Accordingly, in
the regression analysis, a statistically signi¯cant coe±cient associated with the
dummy variable \pass" would signal a divergent pattern in the s of adequately and
non-adequately capitalized banks and suggest that stress tests do not contribute to
determining contagion risk in the banking system, whereas a statistically non-sig-
ni¯cant coe±cient combined with a negative and signi¯cant constant term would
indicate that these exercises have been successful in curbing the overall systemic risk.
Finally, there is a third dimension to be considered which we have labeled,
according to the ECB de¯nition, \structural." Now that macroprudential supervi-
sion has become a standard practice and stress test exercises are routinely held,
market participants might use them as a \userguide" to infer the authorities' overall
supervisory stance and, in turn, policy makers might use it as a \forward guidance"
tool to steer markets' behavior. This is most likely the feature less investigated in the
literature. Nevertheless, some empirical works provide insights. Lazzari et al. (2017)
show that markets, beyond simply getting fresh news on the soundness of single
institutions, also used the 2014 comprehensive assessment to infer the policy stance
that the newly established European supervisory framework (the SSM) would have
adopted. In contrast, Gambetta et al. (2019) demonstrate that capital curtailments
resulting from stress tests exercises are predictable as they are determined by precise
banking features, and market agents have the means to detect which these are. To
understand if agents use stress tests as an informative source on the supervisory
attitude, we test the following hypothesis:
Hypothesis 3. Macroprudential exercises e®ectively represent tools from which the
market infers the policy stance of supervisory authorities (the \structural" dimen-
sion).
To test this hypothesis, we estimate and analyze the pattern of the market sof
portfolios representing the quartiles of the adverse scenario curtailment distribution.
The relative evolution of the market of these portfolios is a valid indicator of how
the supervisory authorities convey their policy approach to the market. Di®erences in
F. Pacicco, L. Vena & A. Venegoni
2150009-8
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
the risk perception evolution of these portfolios (indicated by statistically signi¯cant
coe±cients associated with the dummy variable) would signal that the market learns
from the exercises the relative risk weights attributed by the regulators to the dif-
ferent balance sheet exposures, whereas similar patterns (non-statistically signi¯cant
coe±cients of the dummy variables) would deny this signaling power, in turn
meaning that stress tests provide market participants with information about the
overall perspective supervisory policy stance rather than just providing clues on the
risk weights attributed to single asset classes. To further corroborate our empirical
assessment and take into account the fact that the European banking system is not
just represented by the scrutinized banks but also by certain institutions that are
excluded from these supervisory procedures for various reasons, we setup an addi-
tional empirical exercise which estimates the evolution of the market s for the
portfolios of participating and non-participating listed banks for each macro-
prudential stress test. This allows us to gather a more reliable and more detailed
picture of the contribution of these procedures to the \structural dimension" of
macroprudential supervision. Indeed, if agents generalize the information gathered
on the examined banks to the ones which have not undergone the tests, the sof
participating and non-participating banks will move accordingly, otherwise they will
show di®erent patterns. As for the regression analysis, a non-signi¯cant coe±cient
associated with the dummy variable which identi¯es the banks that have partici-
pated in the four exercises would signal that a \structural" e®ect is at work. Even
better, a negative signi¯cant coe±cient would indicate that the non-participating
banks exploit a \free riding" advantage as they do not su®er from any idiosyncratic
disclosure \cost" from sharing with participating institutions the bene¯ts derived
from the information about the perspective direction of the supervisory stance.
4. Study Design
Empirical contributions on the information content of macroprudential stress tests
follow a fairly standardized approach, classifying each exercise as a valid information
provider if signi¯cant abnormal returns in stock prices or CDS premia are detected
when the results are disclosed (Morgan et al. 2014).
Although this approach has provided valuable insights on the information con-
tribution of stress tests procedures, it leaves many relevant questions unanswered. In
particular, it does not allow an assessment of how these exercises contribute to each
of the dimensions of the supervisory strategy. To ¯ll this gap, we formulate three
hypotheses, each of which tackles a speci¯c feature of the macroprudential super-
visory activity. To test these, the empirical framework is split into two steps: ¯rst we
employ a time varying Fama–French three-factor model to estimate market s
evolution and subsequently a panel regression framework to infer the factors that
a®ect their variation around the stress tests' results disclosure dates. Employing the
methodology proposed by Pacicco et al. (2020), this paper tests the research hy-
potheses by investigating the e®ectiveness of macroprudential procedures in terms of
Macroprudential Supervision and Agents' Information
2150009-9
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
their impact on banks' risk perception. That is to say, we study how each of the EU
macroprudential stress tests shapes a market-based measure of risk, utilizing a state-
space representation of the Fama–French three-factor model to do so. The practical
relevance of this methodological approach goes beyond the mere scope of assessing
how a stress test shapes betas as this framework helps to gain precious insights on the
nature of the new information, if any. More precisely, to the extent that market s
move with, and react to, ¯rm-speci¯c information (Patton & Verardo 2012,Kha-
tua 2015), the information disclosed to markets by stress tests communication is
bound to a®ect the dynamics of the market betas of the single banks and the port-
folios analyzed, re°ecting the impact of stress tests information disclosure on market
risk perception, if any.
To consolidate the empirical evidence obtained through the estimation of the
dynamic market s and the graphical analysis of their relative patterns, we perform a
panel regression analysis which allows us to study the determinants of the market s
patterns around each stress tests disclosure data. This further step allows us to better
identify the nature of the information conveyed by stress test exercises and to assess
if and how they achieve the supervisory strategy. As previously pointed out, we
assess the e®ectiveness of all the European stress tests so far conducted for which
bank-by-bank results have been disclosed. Thus, as no idiosyncratic results have been
published in the CEBS 2009 and 2010 exercises, we had to exclude the ¯rst two
European macroprudential stress tests ever conducted. We also had to exclude the
2015 CA and the 2018 ECB stress test as those procedures involved Greek institu-
tions only. We also do not consider the European Banking Authority (EBA)
transparency exercises (2013, 2015, 2017) as they represent a complementary tool of
the European macroprudential strategy apart from the stress testing procedures.
Accordingly, we focus on four macroprudential exercises, i.e. those conducted in
2011, 2014, 2016, and 2018 (the EU-wide one). Because of the full disclosure policy,
these four procedures constitute a one-of-a-kind in the panorama of EU macro-
prudential supervision. Indeed, in light of the absolute transparency which char-
acterizes them from the beginning of the procedure (when market participants have
been informed about the sample, the methodology and the input data) to the dis-
closure of results (when bank-by-bank outcomes have been disclosed) the four EU
exercises above represent, at least in theory, four information providers able to
provoke a transparency shock.
4.1. Sample selection
To evaluate the e®ectiveness of the European macroprudential procedures, we apply
our models to a sample composed by the 61 banks that have been subject to at least
one of the EU-wide stress test exercises for which bank-by-bank results have been
disclosed (Table 1).
Italy and Spain are the most represented countries, each contributing 10 ¯nancial
institutions to the sample, followed by Portugal (6 banks), Germany, Poland, and
F. Pacicco, L. Vena & A. Venegoni
2150009-10
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
Greece (5). The UK and France contribute 4 each and Ireland 3, whereas of the
remaining 12 countries, 6 contribute with 2 institutions and the other 6 are repre-
sented by only a single bank.
In 2011, 43 institutions were analyzed, and 35 (8) of these were found to be
adequately (inadequately) capitalized. Three years later, in 2014, at the eve of the
start of the SSM, the ECB (jointly with EBA) directly assessed the soundness of 51
banks. Of these, 13 institutions (25.5% of the sample) ended the exercise with a
capital shortfall, whereas the vast majority (38) were shown to be adequately resil-
ient even to the harshest scenario, ending the procedure with their capital above
threshold. In the last two exercises (2016 and 2018), the listed institutions involved
decreased (31). Of these, only one resulted with capital needs in 2016, a ¯gure that
dropped to zero in the subsequent one held two years later (in 2018).
To answer our research question and investigate stress tests ability to provide
information under all the three dimensions of the supervisory activity scope, we
partition our sample in di®erent portfolios.
For each of the procedures under scrutiny, participant banks are jointly inves-
tigated by means of the equally weighted portfolio composed by involved institu-
tions, the remaining ones being part of the portfolio on non-scrutinized banks.
Di®erences and similarities in the patterns of the betas of these two portfolios are one
of the dimensions used to investigate the ability of such exercises to a®ect the
\structural dimension".
Furthermore, as Table 1reports, in each procedure we also distinguish partici-
pant banks according to their performance in the exercise, thereby generating a
three-fold classi¯cation criterion. First, we classify them in two subgroups, including
those institutions who exceeded the 5% CET1 capital ratio in the adverse scenario
(portfolio pass) and those who fell short of it (portfolio fail) in each test, respectively
(second column of each panel). Even if a pass/fail threshold was not set in all the
exercises, it is reasonable to assume that the canonical 5% threshold still constituted
a landmark to which market agents referred. This is because it has been found that a
learning e®ect is at work so that market agents use the experience gained from each
stress test exercise performed during time to interpret and anticipate the outcome of
the subsequent procedures (Glasserman & Tangirala 2016,Luu & Vo 2020).
Following a reviewer's suggestion, we also check the sensitivity of results with dif-
ferent thresholds. That is, for those procedures in which a speci¯c cut-o® point
between su±ciently and insu±ciently capitalized banks was not set by the regulator,
we use di®erent, stricter, thresholds, such as 8%, 10%, and 15% of the CET1 ratio
under the adverse scenario. The results (available upon request) qualitatively con-
¯rm our ¯ndings. The same holds when models are estimated to exclude the pro-
cedures for which the regulator has not explicitly set a pass/fail threshold. The
second and third classi¯cations divide banks into four portfolios corresponding to the
quartiles of, respectively, the capital curtailments and the CET1 level resulting from
the application of the adverse scenario (last two columns of each panel). These
segmentations of our sample allow us to assess if stress tests are able to ful¯ll all three
Macroprudential Supervision and Agents' Information
2150009-11
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
Table 1. Involved institutions and portfolios by stress test.
Scrutinized
Yes/No (1/0)
Pass/Fail (1/0)
Quartiles
by curtailment
Quartiles
by CET 1
Scrutinized
Yes/No (1/0)
Pass/Fail (1/0)
Quartiles by
curtailment
Quartiles
by CET 1
Scrutinized
Yes/No (1/0)
Pass/Fail (1/0)
Quartiles
by curtailment
Quartiles
by CET 1
Scrutinized
Yes/No (1/0)
Pass/Fail (1/0)
Quartiles
by curtailment
Quartiles
by CET 1
BanknStress test Country 2011 2014 2016 2018
Aareal bank DE 0 0 0 0
Abn amro bank NL 1 1 2 2 1 1 3 2 1 1 4 2 1 1 2 1
Alior bank PL 0 111 2 0 0
Allied Irish banks IE 1 0 4 4 1 0 4 4 1 1 4 4 1 1 4 2
Alpha bank GR 1 1 4 3 1 0 4 4 0 0
Banca carige IT 0 104 4 0 0
Banca MPS IT 1 0 3 4 1 0 4 4 1 0 4 4 0
BPER banca IT 0 113 3 0 0
BP Sondrio IT 0 103 4 0 0
BBVA ES 1 1 1 2 1 1 2 2 1 1 2 3 1 1 1 3
Banco bpi PT 0 114 1 0 0
Banco Bpm IT 1 1 2 4 1 0 3 4 1 1 3 2 1 1 2 4
Banco Com. Portugus PT 1 0 3 4 1 0 4 4 0 0
Banco de sabadell ES 1 1 3 4 1 1 2 2 1 1 2 3 1 1 4 4
Banco pastor ES 1 0 4 4 0
0 0
Banco popular espaol ES 1 1 3 4 1 1 3 3 1 1 4 4 0
Banco santander ES 1 1 1 2 1 1 2 3 1 1 3 3 1 1 1 3
Bank handlowy PL 0 111 1 0 0
Bank Och. srodowiska PL 0 112 2 0 0
Bank of Ireland IE 1 0 4 4 1 0 4 4 1 1 4 4 1 1 3 2
Bank of Valletta MT 1 1 2 1 1 1 2 2 0 0
Bank Pekao PL 0 0 0 111 1
Bankinter ES 1 1 2 3 1 1 1 1 0 0
Barclays UK 1 1 4 3 1 1 1 3 1 1 3 4 1 1 4 4
Bfa-Bankia ES 1 0 4 4 0 0 0
F. Pacicco, L. Vena & A. Venegoni
2150009-12
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
Table 1. (Continued )
Scrutinized
Yes/No (1/0)
Pass/Fail (1/0)
Quartiles
by curtailment
Quartiles
by CET 1
Scrutinized
Yes/No (1/0)
Pass/Fail (1/0)
Quartiles by
curtailment
Quartiles
by CET 1
Scrutinized
Yes/No (1/0)
Pass/Fail (1/0)
Quartiles
by curtailment
Quartiles
by CET 1
Scrutinized
Yes/No (1/0)
Pass/Fail (1/0)
Quartiles
by curtailment
Quartiles
by CET 1
BanknStress test Country 2011 2014 2016 2018
Bnp paribas FR 1 1 3 2 1 1 2 2 1 1 1 3 1 1 2 4
Caixabank ES 0 0 0 0
Caja Mediterraneo ES 0 0 0 0
Commerzbank DE 1 1 3 3 1 1 3 3 1 1 4 4 1 1 4 3
Credito emiliano IT 0 112 2 0 0
Danske bank DE 1 1 1 1 1 1 2 1 1 1 1 1 1 1 3 2
Deutsche bank DE 1 1 3 3 1 1 4 3 1 1 4 4 1 1 4 4
Dnb bank group NO 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1
Erste group bank AT 1 1 2 2 1 1 2 3 1 1 3 3 1 1 3 4
Eurobank ergasias GR 1 0 4 4 1 0 4 4 0
0
Credit Agricole FR 1 1 1 2 1 1 2 2 1 1 2 2 1 1 3 2
Getin noble bank PL 0 111 2 0 0
Hellenic bank GR 0 0 0 0
Hsbc holdings UK 1 1 3 2 1 1 1 2 1 1 2 3 1 1 4 3
Hypo real estate DE 0 0 0 0
Intesa Sanpaolo IT 1 1 2 3 1 1 3 2 1 1 2 2 1 1 2 2
Jyske bank DK 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 2
Kbc bank BE 1 1 2 1 1 1 4 3 1 1 3 2 1 1 2 2
Liberbank ES 0 103 4 0 0
Lloyds banking UK 1 1 3 2 1 1 3 3 1 1 2 2 1 1 4 4
Cyprus popular bank CY 1 0 4 4 0 0 0
Mediobanca IT 0 112 3 0 0
Nat. bank of Greece GR 1 1 4 2 1 0 4 4 0 0
Nordea bank FI 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
Op-pohjola group FI 1 1 2 1 1 1 3 1 1 1 3 1 1 1 3 1
Macroprudential Supervision and Agents' Information
2150009-13
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
Table 1. (Continued )
Scrutinized
Yes/No (1/0)
Pass/Fail (1/0)
Quartiles
by curtailment
Quartiles
by CET 1
Scrutinized
Yes/No (1/0)
Pass/Fail (1/0)
Quartiles by
curtailment
Quartiles
by CET 1
Scrutinized
Yes/No (1/0)
Pass/Fail (1/0)
Quartiles
by curtailment
Quartiles
by CET 1
Scrutinized
Yes/No (1/0)
Pass/Fail (1/0)
Quartiles
by curtailment
Quartiles
by CET 1
BanknStress test Country 2011 2014 2016 2018
Otp bank HU 1 1 1 1 1 1 3 1 1 1 3 2 1 1 1 2
Permanent tsb IE 0 104 4 0 0
Piraeus bank GR 1 1 4 3 1 0 4 4 0 0
PKO Bank Polski PL 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
Rai®eisen bank AT 1 1 2 2 0 0 112 3
Royal bank of Scotland UK 1 1 4 3 1 1 2 3 1 1 4 3 1 1 4 3
Skandinaviska E. banken SE 1 1 2 1 1 1 1 1 1 1 1 1 1 1 2 1
Societe Generale FR 1 1 3 3 1 1 2 2 1 1 2 3 1 1 2 3
Swedbank SE 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
Sydbank DK 1 1 1 1 1 1 1 1 0 0
Unicredit IT 1 1 3 3 1 1 2 3 1 1 2 4 1 1 3 3
UBI banca IT 1 1 2 3 1 1 3 2 1 1 2 2 1 1 2 4
Notes: The table lists all the banks pertaining to our sample. For each of them, the table shows also if it took part in stress test (scrutinized ¼1)orifitdid
not (scrutinized ¼0), if they resulted as su±ciently (pass/fail ¼1) or insu±ciently capitalized (pass/fail ¼0), as well as the partitioning between sub-
samples according to its performance in each exercise.
F. Pacicco, L. Vena & A. Venegoni
2150009-14
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
dimensions of macroprudential supervision, namely (i) the \stamina" dimension, (ii)
the \contagion" dimension, and the (iii) \structural" dimension.
Tables 2–5show the descriptive statistics of the portfolios so far mentioned for
each procedure under assessment. More precisely, each table shows, by stress-test,
the average market return of portfolios; their standard deviation, minimum, median,
and maximum values computed over the two-year period surrounding the release
date of results, composed by the 52 trading weeks before and after it.
In all cases, the portfolios of involved and non-involved institutions performed
worse than the overall market. Indeed, although these portfolios could not do better
than a negative performance of 0.15% (Involved banks in 2014), the average market
return is positive in all the sub-samples (minimum 0.01%, maximum 0.14%).
Considering the exercise that highlighted more than one bank as undercapitalized
(2011 and 2014), the portfolios of banks without capital shortfall outperformed those
of banks that ended the exercise with their capital below the established threshold. In
both exercises, the average market return of the former was almost a half percentage
point more than that of the latter.
Examining the portfolios by quartiles, only in 2011 is there a relationship between
the capital curtailment su®ered by banks, their capital level after the adverse sce-
nario, and the average market returns. Indeed, only in this case is there a linear
relationship between the robustness of institutions classi¯ed in each portfolio and
Table 2. Descriptive statistics of portfolios of banks scrutinized and non-scrutinized in the EU
2011 stress test.
Portfolio NMean St Dev. Min Median Max
Scrutinized banks 105 0.80% 4.26% 13.41% 0.78% 10.97%
Non-scrutinized banks 105 0.66% 3.73% 13.25% 0.65% 10.27%
Su±ciently capitalized banks 105 0.62% 4.40% 13.67% 0.85% 11.52%
Insu±ciently capitalized banks 105 1.76% 5.50% 15.55% 2.15% 9.08%
Quartiles by curtailment
Q1 105 0.23% 3.95% 13.49% 0.24% 10.37%
Q2 105 0.42% 4.32% 11.42% 0.53% 13.45%
Q3 105 0.93% 4.81% 15.37% 1.05% 12.96%
Q4 105 1.69% 6.27% 16.15% 1.98% 13.97%
Quartiles by CET 1 ratio
Q1 105 0.11% 3.30% 10.93% 0.11% 9.17%
Q2 105 0.55% 4.94% 14.16% 0.36% 13.44%
Q3 105 1.19% 5.73% 16.85% 1.36% 14.30%
Q4 105 1.57% 5.08% 13.91% 1.90% 8.75%
Fama–French factors
Market 105 0.10% 3.29% 11.32% 0.27% 9.28%
Small–Big Cap 105 0.05% 1.01% 3.58% 0.06% 2.44%
High–Low B/M 105 0.22% 1.13% 3.30% 0.18% 2.58%
Notes: The table shows the mean, standard deviation (St Dev.), minimum (Min), median and
maximum (Max) values of market return of portfolios described in Table 1during the two years
surrounding the EU 2011 stress test.
Macroprudential Supervision and Agents' Information
2150009-15
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
their average market return, with market performances positively (negatively) re-
lated to the soundness (fragility) of banks.
4.2. Estimating the cycle-free betas
The source and nature of information conveyed through the EU macroprudential
procedures are investigated using the model developed by Pacicco et al. (2020). For
the sake of terseness and readability, we limit the description of the model to the
basics: (i) providing su±cient information to comprehend its functioning and in-
terpret the results and (ii) highlighting the di®erences between our setup and the
original one. We thus refer the interested reader to Pacicco et al. (2020) which gives a
technical, complete, and exhaustive description of the framework.
Starting from the model description, to ful¯ll our research aim, we must estimate
the market agents risk perception evolution around the disclosure of macro-
prudential stress tests results for the portfolios of banks in which we segment our
sample. To achieve this goal, it is necessary to set up a time-varying framework able
to measure the evolution over time of a valid proxy of market risk perception.
Therefore, the model we employ estimates a time-varying Fama–French three-factor
model (Fama & French 1993) which allows us to appreciate the market beta evo-
lution, employed as our measure of agents risk perception (Sahin et al. 2020).
Table 3. Descriptive statistics of portfolios of banks scrutinized and non-scrutinized in the EU
2014 stress test.
Portfolio NMean St Dev. Min Median Max
Scrutinized banks 105 0.15% 2.96% 7.14% 0.17% 6.58%
Non-scrutinized banks 105 0.28% 3.40% 11.03% 0.38% 8.24%
Su±ciently capitalized banks 105 0.06% 2.50% 6.22% 0.39% 5.52%
Insu±ciently capitalized banks 105 0.76% 5.95% 21.68% 0.10% 12.00%
Quartiles by curtailment
Q1 105 0.04% 2.17% 5.42% 0.03% 5.07%
Q2 105 0.08% 2.73% 7.26% 0.19% 6.50%
Q3 105 0.16% 3.02% 6.49% 0.24% 6.97%
Q4 105 0.71% 5.66% 20.34% 0.16% 11.35%
Quartiles by CET 1 ratio
Q1 105 0.14% 1.97% 4.69% 0.41% 4.13%
Q2 105 0.08% 2.78% 7.19% 0.24% 7.48%
Q3 105 0.06% 3.11% 8.12% 0.10% 7.88%
Q4 105 0.76% 5.95% 21.68% 0.10% 12.00%
Fama–French factors
Market 105 0.01% 1.81% 4.24% 0.21% 5.31%
Small–Big Cap 105 0.04% 0.94% 2.13% 0.00% 2.49%
High–Low B/M 105 0.16% 0.87% 2.13% 0.20% 3.16%
Notes: The table shows the mean, standard deviation (St Dev.), minimum (Min), median and
maximum (Max) values of market return of portfolios described in Table 1during the two years
surrounding the EU 2014 stress test.
F. Pacicco, L. Vena & A. Venegoni
2150009-16
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
In line with Pacicco et al. (2020), we plug into our framework a de-cycling pro-
cedure to purge the business cycle component from the market betas evolution,
obtaining an unbiased measure that gauges the evolution of agents' risk perception.
This is a fundamental feature of our analysis as it allows us to obtain a \pure"
estimation of the communication e®ects yielded by stress tests results disclosure
without the conditioning of the overall economic conditions. Brie°y, the steps that
our framework completes to produce time-varying cycle-free measures of market risk
perception are as follows:
(1) Dynamic estimation of the three Fama & French (1993) factors. The return of
each portfolio is disentangled into four components related to the three risk
factors (market risk factor market, size risk factor small–big, style risk factor
high–low [book to market]) with which are associated the related time-varying
coe±cients plus a constant term.
(2) Decomposition of the coe±cients. Each time-varying parameter is shaped to be
properly disentangled in three parts: an auto-recursive component, a cycle e®ect
and, consequently, a \cycle-free" update.
(3) Construction of the cycle-free risk perception measure. By summing the \cycle-
free" updates, for each point in time we calculate the time series of the proxy for
agents' risk perception, which is therefore sterilized from the e®ect of the economic
cycle.
Table 4. Descriptive statistics of portfolios of banks scrutinized and non-scrutinized in the EU
2016 stress test.
Portfolio NMean St Dev. Min Median Max
Scrutinized banks 105 0.20% 3.24% 7.23% 0.67% 7.66%
Non-scrutinized banks 105 0.50% 3.55% 9.14% 0.50% 7.99%
Su±ciently capitalized banks 105 0.12% 3.13% 6.89% 0.55% 7.93%
Insu±ciently capitalized banks 105 2.39% 10.12% 32.78% 0.00% 45.74%
Quartiles by curtailment
Q1 105 0.17% 2.54% 6.37% 0.17% 6.75%
Q2 105 0.11% 4.11% 9.82% 0.18% 10.98%
Q3 105 0.11% 2.64% 6.71% 0.08% 6.49%
Q4 105 1.02% 4.62% 11.48% 1.27% 12.84%
Quartiles by CET 1 ratio
Q1 105 0.15% 2.15% 5.55% 0.03% 5.57%
Q2 105 0.04% 3.45% 8.68% 0.06% 7.81%
Q3 105 0.13% 3.78% 8.50% 0.17% 9.80%
Q4 105 1.00% 4.79% 10.89% 1.49% 12.85%
Fama–French factors
Market 105 0.14% 1.89% 6.11% 0.13% 5.31%
Small–Big Cap 105 0.13% 0.98% 2.72% 0.11% 3.29%
High–Low B/M 105 0.03% 1.08% 2.12% 0.13% 3.16%
Notes: The table shows the mean, standard deviation (St Dev.), minimum (Min), median and
maximum (Max) values of market return of portfolios described in Table 1during the two years
surrounding the EU 2016 stress test.
Macroprudential Supervision and Agents' Information
2150009-17
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
In formula, for each ith equally weighted portfolio, we estimate the following state-
space Fama & French (1993) three-factor model:
Ri;t¼i;tFtþi;t;ð1aÞ
i;tþ1¼Tti;tþi;t;ð1bÞ
where Ri;tis the market return of the portfolio i.Ftand i;tare, respectively, the
three factors plus the constant term array and their relative coe±cients, with each of
them being shaped as a time-varying process thanks to Eq. (1b). This latter allows to
us properly ¯lter out the cyclical component of each beta (for better reference on the
business cycle in°uence on betas cf. Jagannathan & Wang 1996,Lewellen &
Nagel 2006,Adrian & Franzoni 2009). Speci¯cally, as shown by Eq. (2) which makes
explicit Eq. (1b), the period-by-period update (i;t) of the market-based risk measure
is not in°uenced by the business cycle as it is considered by including the cycle proxy
in the equation (gov yieldi;t).
i;tþ1¼i;tþi;tgov yieldi;tþi;t:ð2Þ
Starting from Eq. (2), once all parameters are estimated, we algebraically com-
pute the cycle-free update as described by the following equation:
i;t¼i;tþ1i;ti;tgov yieldi;t:ð3Þ
Table 5. Descriptive statistics of portfolios of banks scrutinized and non-scrutinized in the EU
2018 stress test.
Portfolio NMean St Dev. Min Median Max
Scrutinized banks 105 0.20% 2.45% 5.39% 0.34% 7.60%
Non-scrutinized banks 105 0.28% 2.28% 6.45% 0.02% 4.42%
Su±ciently capitalized banks 105 0.20% 2.45% 5.39% 0.34% 7.60%
Insu±ciently capitalized banks
Quartiles by curtailment
Q1 105 0.08% 2.21% 4.39% 0.15% 6.79%
Q2 105 0.18% 2.82% 6.52% 0.17% 8.32%
Q3 105 0.30% 2.54% 7.19% 0.32% 6.97%
Q4 105 0.26% 2.82% 5.78% 0.07% 8.22%
Quartiles by CET 1 ratio
Q1 105 0.07% 2.05% 4.37% 0.13% 6.74%
Q2 105 0.19% 2.21% 5.30% 0.30% 6.53%
Q3 105 0.31% 3.08% 6.73% 0.35% 8.54%
Q4 105 0.26% 3.16% 7.43% 0.21% 8.72%
Fama–French factors
Market 105 0.01% 1.88% 6.32% 0.14% 4.56%
Small–Big Cap 105 0.09% 0.72% 1.88% 0.04% 1.61%
High–Low B/M 105 0.09% 0.83% 2.44% 0.11% 3.84%
Notes: The table shows the mean, standard deviation (St Dev.), minimum (Min), median and
maximum (Max) values of market return of portfolios described in Table 1during the two
years surrounding the EU 2018 stress test.
F. Pacicco, L. Vena & A. Venegoni
2150009-18
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
Thus, exploiting the characteristic of the i;t, we retrieve a synthetic series of
\cycle-free" betas by means of the algebraic procedure described by the following
equation:
It is worth noting that, once \purged" from the cyclical component, the betas
level becomes no more informative, a fact that does not impair our analysis given
that we are only interested in studying its dynamic. Accordingly, betas below 1 are
not understood as \defensive" betas, just as negative betas do not signal counter-
cyclical institutions. As claimed by Pacicco et al. (2020), the fact that the betas' level
becomes no more informative is due to the auto-recursive structure of the coe±cients.
Indeed, because of this, in each period, the di®erence between the estimated (i;t)
and cycle-free betas (~
i;t) does not measure the periodical impact of the business
cycle. This aspect, although preventing us from testing whether betas signi¯cantly
di®er across groups, does not impair the analysis of betas' dynamic, i.e. the statistical
tests for di®erences in betas over time.
Turning to the di®erences between our approach and that of Pacicco et al. (2020),
there are two such di®erences due to the need to adapt the framework to the research
question and the sample analyzed.
First, in addition to estimating the model on all securities, we rely on equally
weighted portfolios. Such a methodological choice allows us to properly discern the
source of information macroprudential procedures convey, if any, allowing us to
study in depth the communication e®ects under the three di®erent dimensions which
constitute the object of this research.
The second di®erence between our model and Pacicco et al. (2020) is that they
proxy the US business cycle using either ¯nancial (three-month T-bill, BAA–AAA
corporate spread) or real variables (imports, vehicle sales, unemployment claims).
With regards to the European cases, an adaptation becomes necessary for at least
two reasons. First, due to the asynchronous business cycles of European countries,
rather than a common EU proxy, country-speci¯c proxies must be used. Second, no
weekly time series for real variables are available. Accordingly, we use 10-year
government bond yields as, especially after the sovereign debt crises, they mostly
re°ect the heterogeneity of European countries business cycles patterns. However, as
our analysis is conducted at the portfolio level, govies' yields must be aggregated thus
to re°ect the construction of the portfolio itself. More precisely, to each portfolio we
pair a portfolio of government bonds to re°ect the cross-country portfolio composi-
tion. In other words, if the portfolio is composed of Spanish and French banks by 20%
each, Italian banks by 35%, and German banks by 25%, the portfolio proxying the
cycle is constructed by aggregating the government yields using the same weights:
20% each for Spain and France, 35% for Italy, and 25% for Germany. However, as a
robustness check, we also estimate our models by proxying the business cycle with
Macroprudential Supervision and Agents' Information
2150009-19
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
the German Bund's yields only. Results, available upon request, qualitatively con-
¯rm the ¯ndings presented here.
4.3. The contribution of stress testing to macroprudential supervisory
activity
The aim of this research is to assess if stress test exercises represent a valuable tool
that supervisory authorities can exploit to pursue their macroprudential goals. To do
so, our empirical framework is centered on the analysis of the dynamic evolution of
the market s of the institutions involved in such exercises. Once time-varying s
have been estimated, we analyze the features, if any, that a®ect their pattern around
the results disclosure date of each stress test. We thus resort to a set of panel
regression models where the market variation for each institution in the time
window that goes from 26 weeks prior to (pre-event period) the disclosure date (in
order to reasonably control for possible anticipation e®ects) to 26 weeks after it
(post-event period) is used as a dependent variable. More precisely, to assess the
ability of stress test procedures to address the \stamina" dimension, we regress the s
variation on three dummy variables that mimic the segmentation of scrutinized
banks in quartiles by their CET1 capital ratio due to the outcome of the adverse
scenario (see Eq. (5)). In other words, these three variables take the value of 1 if,
respectively, each of the banks considered in the sample belong to the ¯rst, second,
and third quartiles of the distribution per CET1 capital ratio after the application of
the adverse scenario simulation, leaving the betas' variation of the banks that belong
to the fourth quartile as the benchmark.
i;t¼0;iþ1Q1CET1
i;tþ2Q2CET1
i;tþ3Q3CET1
i;tþX0
tþi;t:ð5Þ
In this manner, signi¯cant coe±cients associated with the dummy variables
would signal a di®erence in the market s pattern of a quartile group compared to
the baseline, corroborating the hypothesis that stress tests e®ectively convey new
idiosyncratic information on single banks and so contribute to the \stamina" feature
of the supervisory activity.
Turning to the \contagion" dimension, i.e. the ability of the stress test exercises
to convey information regarding the soundness of the overall banking system and its
vulnerability to idiosyncratic shocks, we regress the changes in s on a dummy
variable equal to 1 for banks for which the CET1 ratio was higher than the 5% ratio
after the adverse scenario simulation and 0 vice versa, as outlined by the following
equation:
i;t¼0;iþ1ACi;tþX0
tþi;t:ð6Þ
The statistical signi¯cance of the coe±cient associated with this variable (ACi;t)
would thus signal that the market uses stress test results only to update their bank-
by-bank risk assessment. In contrast, should parameter 1fail to reach statistical
signi¯cance, this may indicate the ful¯lment of the \contagion" dimension, with
F. Pacicco, L. Vena & A. Venegoni
2150009-20
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
market participants using this piece of information to assess the overall soundness of
the ¯nancial system.
Finally, to detect whether stress tests tackle the \structural" dimension, we
perform the two-fold analysis described by Eqs. (7) and (8).
The former includes three dummy variables identifying the ¯rst three quartiles of
scrutinized banks by the capital curtailment due to the adverse scenario simulation.
i;t¼0;iþ1Q1curt
i;tþ2Q2curt
i;tþ3Q3curt
i;tþX0
tþi;t:ð7Þ
Again, the fourth quartile represents the benchmark. A statistically signi¯cant co-
e±cient associated with one or more of the three dummy variables considered here
would indicate that the market learnt from the exercises the relative risk weights
attributed by the regulators to the di®erent balance sheet exposures, whereas it
would otherwise indicate that these exercises provide agents with a signal of the
overall supervisory stance, thus addressing the dimension we target with this setup.
The latter includes the dummy variable PBi;twhich distinguishes between par-
ticipating (1) and non-participating (0) banks for each exercise.
i;t¼0;iþ1PBi;tþX0
tþi;t:ð8Þ
In this case, a positive signi¯cant coe±cient 1would indicate that the infor-
mation provided by stress tests are not extended to the overall banking system by
market agents, that, in turn, deem as riskier the non-scrutinized, and thus more
opaque, banks. In contrast, a negative signi¯cant coe±cient would signal that non-
participating banks bene¯t from a market risk perception reduction relative to the
examined institutions. This may be because agents have gathered information about
the general policy stance that also concerns non-examined banks, which, in turn, did
not have to \pay the price" of having potentially negative idiosyncratic news dis-
closed. Finally, a non-signi¯cant coe±cient would highlight an aligned reaction be-
tween participating and non-participating institutions which, again, ful¯lls the
\structural" dimension and suggests that stress tests are a useful tool to convey
information regarding the overall perspective supervisory attitude.
In addition to the variables that directly assess our hypotheses, we employ a
standard set of control variables (X0
t) that are bound to a®ect the market-based
measure of risk. Thus, in line with the literature (cf. Ben-Zion & Shalit 1975,Dia-
mond & Rajan 2000,Ellul & Yerramilli 2013), we consider the e®ects exerted by: size
(proxied by total assets, in log), solidity, and soundness (the ratio of non-performing
exposures to total exposures coverage ratio), capital structure, and leverage
(represented by the Tier 1 capital ratio and the ratio of total debt to total asset). We
plug these variables into our model, considering their last value before the stress test
announcement, given that these are the data that regulatory authorities consider in
their stress-testing procedure and so are the ones to which market agents should refer
when forming expectations of the exercises' results. Nevertheless, according to the
investors' rationality principle, they are bound to react to fresh news, being
Macroprudential Supervision and Agents' Information
2150009-21
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
in°uenced by possible pre-emptive actions that banks may take in the period be-
tween the announcement of the procedure and the publication of the results. For this
reason, we perform a robustness check using data taken from the last quarterly
report prior to the disclosure date of each considered stress test. The evidence holds
and the results are qualitatively very similar.
Last, all models include both individual and time ¯xed e®ects. The former allows
us to control for the impact of idiosyncratic features of each institution that remain
constant over time (such as nationality, exposures to some particular asset class and
business model), whereas the second allows us to control for the macroeconomic
context in which each procedure was carried out and their speci¯c targets, thereby
¯ltering out those time-variant e®ects relevant to all institutions under scrutiny (for
example, the 2011 test was run during the sovereign crisis, the 2014 test at the start
of the SSM, and the 2016 test to provide pillar 2 information to the supervisory
authorities).
5. Results and Discussion
The sample segmentation described in the previous section allows us to analyze all
three facets of macroprudential supervision: the (i) \stamina", (ii) \contagion," and
(iii) \structural" dimensions. Our analysis is not limited to the observation of the
betas pattern as we complement the insights o®ered by the graphical analysis with
statistical tests for di®erences in the value of market betas registered in the post-
event period and the pre-event period.
c
Starting from the ¯rst of the dimensions we consider, Fig. 1shows the pattern of
market risk perception for each quartile of the distribution of the CET1 capital ratios
resulting from the adverse scenario application in every exercise. For three out of four
stress tests (2011, 2016, and 2018), the evidence tells us that agents did not dis-
criminate between sounder and weaker banks as they adjusted their risk perception
in the same direction for all the quartiles. The only exception is the combined stress
test and AQR procedure held in 2014. For this exercise, we observe a behavior
opposite to that which we would have expected. The top three quartiles show a
statistically signi¯cant increase in the market risk perception in the post-event pe-
riod, whereas the one that groups the least capitalized institutions displays a marked
and statistically signi¯cant decrease, signaling that agents considered the informa-
tion disclosed to be risk quelling for those banks. This is a surprising piece of evidence
as we should have expected agents to revise upward their risk perception related to
the most fragile institutions if these were unexpected news. However, this seemingly
surprising pattern aligns with the theoretical predictions of Faria-e-Castro et al.
(2017), who demonstrate how a credible ¯scal backstop facility drives markets'
c
We choose this timespan as we deem it a good compromise to observe a durable e®ect and to avoid bias to
the results, including other economic and ¯nancial events (i.e. the launch of the Public Sector Purchase
Program by the ECB in February 2015). Nevertheless, for the sake of rigor, we perform several robustness
checks with longer and shorter windows, and the results remain qualitatively similar.
F. Pacicco, L. Vena & A. Venegoni
2150009-22
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
Notes: The ¯gures show the betas of the four portfolios grouping banks by quartiles according to their CET 1 ratio due to the outcome of the adverse scenario in the
EU stress tests (from left to right, 2011, 2014, 2016 and 2018). Banks in the ¯rst quartile are those with the highest CET 1 ratio, while banks in the fourth quartile
are those that appeared less capitalized. The bootstrapped t-test for di®erence between the average betas before and after the exercise is provided in square brackets.
Fig. 1. Betas of portfolios of banks by their CET 1 ratio under the adverse scenario of the EU-wide stress tests.
Macroprudential Supervision and Agents' Information
2150009-23
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
reaction to the disclosure of stress tests results. Indeed, in 2014, banks that did not
meet the minimum capital threshold were forced to recapitalize via an ad-hoc pro-
gram \through the provision of public funds (backstops)", if required (European
Central Bank 2014, p. 126). Thus, it should come as no surprise if the market revised
the risk perception in light of the backstop rather than the fresh information dis-
closed by the stress test procedure.
The panel regression model (cf. column 4 of Table 6) con¯rms the evidence
gathered from the graphical analysis as there are no statistically signi¯cant coe±-
cients associated with the quartile dummy variables, re°ecting the CET1 ratio level
after the application of the adverse scenario. Summarizing the evidence, it appears
that agents do not gather valuable information from these exercises to discriminate
banking institutions according to their capital resilience. This leads us to conclude
that stress tests add little if anything to the \stamina" dimension of the macro-
prudential activity as they do not disclose information that enhances agent's
knowledge regarding the soundness of banking institutions; instead, they most likely
Table 6. The determinants of betas variations.
Dependent variable: delta beta
Adequately capitalized banks 0.242
(0.5787)
Non-scrutinized banks 0.507***
(0.0094)
Q1 by curtailment 0.427*
(0.0610)
Q2 by curtailment 0.243
(0.2439)
Q3 by curtailment 0.084
(0.6929)
Q1 by CET1 0.046
(0.8362)
Q2 by CET1 0.188
(0.3027)
Q3 by CET1 0.198
(0.1580)
Constant 5.266* 2.448 5.024 4.009
(0.0898) (0.3847) (0.1123) (0.2145)
N131 191 131 131
R20.36 0.26 0.40 0.38
Adj R20.32 0.23 0.35 0.33
Control vars YES YES YES YES
Time ¯xed-e®ects YES YES YES YES
Individual ¯xed e®ects YES YES YES YES
Notes: The table shows output of the panel regression models used to disentangle the
changes in betas. Robust p-values are provided in parentheses, whereas *, ** and ***
denote statistical signi¯cance at the 10%, 5% and 1% con¯dence level, respectively.
F. Pacicco, L. Vena & A. Venegoni
2150009-24
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
disclose information they already have (cf. Cerutti & Schmieder 2014,Lazzari et al.
2017,Pacicco et al. 2020).
Regarding the \contagion" dimension, we compare the betas of institutions that
met or failed to meet the 5% CET1 capital threshold under adverse scenario con-
ditions in each procedure (Fig. 2).
d
This distinction does not allow us to gather
information from all the procedures that we consider as, in 2016 and 2018, respec-
tively, only one or even no banks showed a CET1 capital below the 5% threshold
under the adverse scenario hypotheses. To address the \contagion" dimension, we
observe if the market risk measure associated with the two portfolios moves in the
same direction around the results disclosure date or not. An upward co-movement
would indicate a high degree of contagion in the economy (the \threat" posed by
weak institutions to the system also highly a®ects the sounder banks), whereas a
shared downward trend or dissociated patterns can be interpreted as a decrease in
the contagion threat. This is because, on the aggregate level, a negative reaction
associated with banks that \passed" the exercise can be due to an overall increase in
the perceived weakness of the banking system derived from the performance of
weaker institutions. The limitation to the ¯rst two exercises (2011 and 2014) does not
impair our analysis but provides extremely valuable insights. Indeed, it is possible to
observe how the betas patterns around the results disclosure date for the two dif-
ferent procedures co-move in a very di®erent manner. The similarities in the betas
evolution which are evident in the 2011 case (both show a statistically signi¯cant
increase in the post-event period) are no longer present in 2014.
Notably, in 2014 there seems to be a \convergence" in the agents risk perception
of sounder and weaker banks, with the betas of the former increasing and those of the
latter decreasing. This provides clear evidence that the release of stress tests results
in quelling the contagion risk, most likely because undercapitalized banks were forced
to subscribe to recapitalization plans aimed at fostering their soundness, removing
their threat to the overall systemic stability.
The panel regression supports this claim as it is possible to observe a statistically
non-signi¯cant coe±cient associated with the \adequately capitalized" dummy (cf.
Table 6, ¯rst column), with the markets revising downward the betas of all banks as
the constant term signals. In summary, this evidence shows that stress tests' pro-
cedures are used by the market to infer the threat posed by the weakest institutions
to the overall soundness of the banking system, hence serving as an e®ective tool to
accomplish the \contagion" dimension of macroprudential supervision strategy.
Finally, to analyze the \structural" dimension, we observe the market risk
measure evolution of the portfolios of examined banks divided in quartiles according
to the distribution of the capital curtailment su®ered under the adverse scenario
d
Following a reviewer's suggestion, we test for the \contagion" hypothesis by substituting in Eq. (6) the
dummy variable AC with a continuous one that measures the post-exercise CET1 ratio. In this manner, we
are able to gauge not only the number of \weak" banks but also the degree of intensity of the under-
capitalization phenomenon. Indeed, the more undercapitalized the CET1 ratio <5%group, the larger the
contagion to the good banks. The results are remarkably stable.
Macroprudential Supervision and Agents' Information
2150009-25
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
Notes: Time varying cycle-free betas of portfolios of banks who resulted as su±ciently (a) or insu±ciently capitalized (b) in the EU stress tests (from left to right,
2011, 2014, 2016 and 2018). The bootstrapped t-test for di®erence between the average betas before and after the exercise is provided in square brackets.
Fig. 2. Portfolios of banks betas by their outcome in the EU stress tests: su±ciently versus insu±ciently capitalized.
F. Pacicco, L. Vena & A. Venegoni
2150009-26
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
conditions (Fig. 3). Quartiles are classi¯ed for increasing levels of curtailments, so the
¯rst one includes the banks that su®ered the lowest curtailment and the fourth
includes the institutions whose capital decreased the most. Indeed, this analysis
allows us to examine whether market agents use stress tests to gather information on
the regulatory attitude towards di®erent asset classes or otherwise take a more
systemic approach and extract a general signal about the prospective severity of the
supervisory conduct.
Observing the results, for the 2011, 2014, and 2018 exercises, it is possible to
recognize a substantial alignment in the betas pattern around the result's disclosure
date. That is, an overall increase is evident, with the notable exception of the fourth
quartile portfolio for the 2014 exercise. Regarding the 2016 stress test, it displays a
risk curbing e®ect on the top ¯rst and third quartiles while substantially has no
impact on the second and fourth. This highlights how, in general, more than gath-
ering insights on the risk weights attributed to speci¯c asset class exposures, agents
seem to infer from the stress test results the supervisory authority attitude toward
the banking system. This claim is further supported by the analysis of the reaction of
the risk measure of scrutinized and non-scrutinized banks portfolios in each exercise
(Fig. 4).
It is possible to trace similar patterns in the betas movements around the results
disclosure date for all the EBA stress tests considered as well as a relative risk
reduction of non-participating banks compared to participating ones for the 2016
and 2018 exercises. The panel regression analysis, the results of which show that the
coe±cient associated with the dummy \non scrutinized banks" is statistically sig-
ni¯cant and negative, highlights that macroprudential stress tests bring a risk re-
duction premium to non-participating banks. This can be attributed to the fact that
such institutions bene¯t from the disclosure of information relative to the supervisory
attitude without having to disclose any possibly negative idiosyncratic data.
Summarizing the evidence described above, the analysis performed with our time-
varying Fama–French three-factor model, dividing our sample into ad-hoc portfolios,
allows us to thoroughly assess the information contribution of this supervisory tool.
In particular, the conventional wisdom is challenged as, rather than using the in-
formation provided to distinguish weak from sound banks, markets use it to assess
the contagion threat over the system and to gather systemic information on the
severity of the policy stances.
This empirically demonstrates that macroprudential stress tests are a valuable
communication tool but in a di®erent manner to that commonly highlighted. Indeed,
among the three dimensions monitored by supervisory activity, the one for which
these exercises seem to be less informative is the \stamina" dimension, given that
they do not appear to contribute much to the market knowledge regarding banks
soundness and resilience to negative external shocks. What they do instead is en-
hance the other two dimensions by clarifying how strong the banking system as a
whole is (and, hence, how probable is the spreading, through contagion, of a systemic
crisis) and informing agents of the degree of strictness of the prospective policy
Macroprudential Supervision and Agents' Information
2150009-27
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
Notes: The ¯gures show the betas of the four portfolios grouping banks by quartiles according to their performance in the EU stress tests (from left to right, 2011,
2014, 2016 and 2018). Banks in the ¯rst quartile are those with the lowest curtailments, while banks in the fourth quartile are those mostly penalized by the exercise.
The bootstrapped t-test for di®erence between the average betas before and after the exercise is provided in square brackets.
Fig. 3. Betas of portfolios of banks by their curtailment in the EU 2011 stress test.
F. Pacicco, L. Vena & A. Venegoni
2150009-28
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
Notes: Time varying cycle-free betas of banks involved (a) and non-involved (b) in the EU stress tests (from left to right, 2011, 2014, 2016 and 2018). The
bootstrapped t-test for di®erence between the average betas before and after the exercise is provided in square brackets.
Fig. 4. Betas of banks scrutinized and non-scrutinized in the EU-wide stress tests.
Macroprudential Supervision and Agents' Information
2150009-29
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
stance. This, although a novelty in the ¯nancial system supervision framework, is
perfectly in line with the evolution of the economic policy development in other
adjoining areas. Indeed, communication that e®ectively conveys the intentions and
the prospective attitude of central banks has nowadays become a central pillar of
monetary policy, and many studies conclude that the struggle of ¯scal policy to
succeed in a®ecting the economic cycle is due to a lack of credible communication.
6. Concluding Remarks
Communication has gained a central role in economic policy activity since the 2008
¯nancial crisis. Authorities, seeing that their standard tools were becoming pro-
gressively ine®ective, have been forced to devise innovative solutions in which en-
hanced transparency and clearer forward guidance on their prospective conduct play
a fundamental role. In this respect, regarding supervisory practices, the shift from a
micro to a macroprudential approach was centered on a renewed attitude towards
communication and, in particular, implied the enhanced disclosure of the results of
the implemented procedures. The three cardinal points that constitute the goals of
the new approach are the following: (i) enhancing the ¯nancial systems resilience to
adverse shocks (the \stamina" dimension), (ii) reducing the system's opacity and
thus the internal propagation e®ect of a single institution's crisis (the \contagion"
dimension) and, ¯nally, (iii) successfully communicating the regulatory approach at
the systemic level in order to convey the intended messages to market agents (the
\structural" dimension). As stress tests procedure are one of the main tools with
which macroprudential supervision is carried out, it becomes relevant to assess their
ability to cover all three dimensions as this provides authorities with insights on their
capacity to pursue the general goals of the overall strategy.
To do so, we employ a time-varying Fama–French three-factor model able to
evaluate the pattern of a market measure of risk around the disclosure date of the
four EU-wide stress tests for which bank-by-bank results have been disclosed to date.
By segmenting the sample of listed banks in portfolios according to their par-
ticipation and performance in each test, we assess their contribution to each of the
identi¯ed dimensions of supervisory activity. In particular, we highlight that mac-
roprudential stress tests (i) do not make any tangible contribution to enhance the
\stamina" dimension, (ii) e®ectively contribute to quell the contagion risk, and (iii)
are a valuable tool through which authorities can convey their perspective line of
action.
In light of this, policymakers should be aware that after a decade of macro-
prudential stress tests implementation, market agents have learnt where the real
value of such exercises resides. Indeed, by performing these procedures, more so than
providing the market with fresh news on the single banks' soundness, supervisors
disclose information on their regulatory stance and the extent to which they perceive
the overall system to be exposed to adverse shocks. These are the features that agents
value the most and that they use to caliber their risk perception, both in crisis and
F. Pacicco, L. Vena & A. Venegoni
2150009-30
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
non-crisis periods. For this reason, such exercises are an extremely valuable tool to be
used by supervisory authorities to steer agents expectations and to keep possible
mounting fears under control.
Acknowledgments
We would like to thank Valter Lazzari, Vincenzo Capizzi, Massimiliano Serati,
Danilo Drago, Matteo Formenti, Luca Corazzini, Rodolfo Helg, Elisa Borghi,
Ra®aele Gallo and two anonymous reviewers for their more than helpful comments.
References
V. V. Acharya, L. H. Pedersen, T. Philippon & M. Richardson (2017) Measuring systemic risk,
The Review Financial Studies 30 (1), 2–47.
T. Adrian & F. Franzoni (2009) Learning about beta: Time-varying factor loadings, expected
returns, and the conditional CAPM, Journal of Empirical Finance 16 (4), 537–556.
L. Ahnert, P. Vogt, V. Vonho® & F. Weigert (2020) Regulatory stress testing and bank
performance, European Financial Management 26(2), 1449–1488.
F. Alvarez, & G. Barlevy (2015) Mandatory disclosure and ¯nancial contagion, Technical
report, National Bureau of Economic Research.
C. Alves, V. Mendes & P. Pereira da Silva (2015) Do stress tests matter? a study on the impact
of the disclosure of stress test results on European ¯nancial stocks and CDS markets,
Applied Economics 47 (12), 1213–1229.
U. Ben-Zion & S. S. Shalit (1975) Size, leverage, and dividend record as determinants of equity
risk, The Journal of Finance 30 (4), 1015–1026.
J. Bischof & H. Daske (2013) Mandatory disclosure, voluntary disclosure, and stock market
liquidity: Evidence from the EU bank stress tests, Journal of Accounting Research
51 (5), 997–1029.
C. Borio, M. Drehmann & K. Tsatsaronis (2014) Stress-testing macro stress testing: Does it
live up to expectations? Journal of Financial Stability 12,3–15.
B. Candelon & A. N. Sy (2015) How did markets react to stress tests? Technical report.
M. Carboni, F. Fiordelisi, O. Ricci & F. S. S. Lopes (2017) Surprised or not surprised? the
investors reaction to the comprehensive assessment preceding the launch of the banking
union, Journal of Banking & Finance 74, 122–132.
E. Cerutti & C. Schmieder (2014) Ring fencing and consolidated banks stress tests, Journal of
Financial Stability 11,1–12.
D. W. Diamond & R. G. Rajan (2000) A theory of bank capital, The Journal of Finance
55 (6), 2431–2465.
A. Ellul & V. Yerramilli (2013) Stronger risk controls, lower risk: Evidence from us bank
holding companies, The Journal of Finance 68 (5), 1757–1803.
European Central Bank (2014) Aggregate report on the comprehensive assessment.
E. F. Fama & K. R. French (1993) Common risk factors in the returns on stocks and bonds,
Journal of Financial Economics 33 (1), 3–56.
M. Faria-e-Castro, J. Martinez & T. Philippon (2017) Runs versus lemons: Information dis-
closure and ¯scal capacity, The Review of Economic Studies 84 (1), 1683–1707.
M. Flannery, B. Hirtle & A. Kovner (2017) Evaluating the information in the Federal Reserve
stress tests, Journal of Financial Intermediation 29,1–18.
N. Gambetta, M. A. García-Benau & A. Zorio-Grima (2019) Stress test impact and bank risk
pro¯le: Evidence from macro stress testing in Europe, International Review of Eco-
nomics & Finance 61, 347–354.
Macroprudential Supervision and Agents' Information
2150009-31
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
O. M. Georgescu, M. Gross, D. Kapp & C. Kok (2017) Do stress tests matter? Evidence from
the 2014 and 2016 stress tests. ECB Working Paper No. 2054, Available at SSRN:
https://ssrn.com/abstract=2968952 or http://dx.doi.org/10.2139/ssrn.2968952.
P. Glasserman & G. Tangirala (2016) Are the Federal Reserves stress test results predictable?
The Journal of Alternative Investments 18 (4), 82–97.
I. Goldstein & Y. Leitner (2018) Stress tests and information disclosure, Journal of Economic
Theory 177,34–69.
I. Goldstein & H. Sapra (2014) Should banks' stress test results be disclosed? an analysis of the
costs and bene¯ts, Foundations and Trendsrin Finance 8(1), 1–54.
R. S. Jabbour & N. Sridharan (2020) The ECB's in-comprehensive SSM-ent: The higher they
go, the harder they fall, The European Journal of Finance 26 (15), 1506–1528.
R. Jagannathan & Z. Wang (1996) The conditional CAPM and the cross-section of expected
returns, The Journal of Finance 51 (1), 3–53.
S. Khatua (2015) The impact of ¯rm-speci¯c news like quarterly earnings announcements on
stock beta, IUP Journal of Applied Finance 21 (4), 51.
V. Lazzari, L. Vena & A. Venegoni (2017) Stress tests and asset quality reviews of banks:
A policy announcement tool, Journal of Financial Stability 32,86–98.
J. Lewellen & S. Nagel (2006) The conditional CAPM does not explain asset-pricing anom-
alies, Journal of Financial Economics 82 (2), 289–314.
H. N. Luu & X. V. Vo (2020) Macroprudential policy and bank risk-taking behaviour:
Implications from regulatory stress test, International Review of Financial Analysis, 75,
101586.
S. Miani, J. Floreani & A. Paltrinieri (2018) Do capital adequacy and credit quality a®ect
systematic risk? investigation of a sample of european listed banks in light of EBA stress
tests, Quarterly Journal of Finance 8(04), 1840006.
D. P. Morgan, S. Peristiani & V. Savino (2014) The information value of the stress test,
Journal of Money, Credit and Banking 46 (7), 1479–1500.
L. L. Ong & C. Pazarbasioglu (2014) Credibility and crisis stress testing, International Journal
of Financial Studies 2(1), 15–81.
F. Pacicco, L. Vena & A. Venegoni (2020) Communication and ¯nancial supervision: How
does disclosure a®ect market stability? Journal of Empirical Finance 57,1–15.
A. J. Patton & M. Verardo (2012) Does beta move with news? ¯rm-speci¯c information °ows
and learning about pro¯tability, Review of Financial Studies 25 (9), 2789–2839.
G. Petrella & A. Resti (2013) Supervisors as information producers: Do stress tests reduce
bank opaqueness? Journal of Banking & Finance 37 (12), 5406–5420.
T. Philippon, P. Pessarossi & B. Camara (2017) Backtesting European stress tests, Technical
report, National Bureau of Economic Research.
C. Sahin & J. De Haan (2016) Market reactions to the ECBs comprehensive assessment,
Economics Letters 140,1–5.
C. Sahin, J. de Haan & E. Neretina (2020) Banking stress test e®ects on returns and risks,
Journal of Banking and Finance, 117, 105843.
F. Pacicco, L. Vena & A. Venegoni
2150009-32
J. Fin. Mngt. Mar. Inst. Downloaded from www.worldscientific.com
by 176.105.248.202 on 01/08/22. Re-use and distribution is strictly not permitted, except for Open Access articles.
