Content uploaded by Elisabetta Gualandri
Author content
All content in this area was uploaded by Elisabetta Gualandri on Feb 18, 2015
Content may be subject to copyright.
CEFIN – Centro Studi di Banca e Finanza
Dipartimento di Economia Marco Biagi – Università di Modena e Reggio Emilia
Viale Jacopo Berengario 51, 41121 MODENA (Italy)
tel. 39-059.2056711 (Centralino) fax 39-059 205 6927
ISSN 2282-8168
CEFIN Working Papers
No 50
MONITORING SYSTEMIC RISK
A SURVEY OF THE AVAILABLE MACROPRUDENTIAL
TOOLKIT
by Elisabetta Gualandri and Mario Noera
November 2014
1
MONITORING SYSTEMIC RISK:
A SURVEY OF THE AVAILABLE MACROPRUDENTIAL TOOLKIT
by
Elisabetta Gualandri
Center for Research in Banking and Finance
University of Modena and Reggio Emilia, Italy, Department of Economics “Marco Biagi”
Email: elisabetta.gualandri@unimore.it
Mario Noera
Bocconi University, Milan
Email: mario.noera@unibocconi.it
Abstract
Understanding the nature of systemic risk and identifying the channels of diffusion of the shocks are the
necessary prerequisite to anticipate and manage successfully the insurgence of financial crises. In order to
prevent financial distress and manage instability, the macroprudential regulator needs to track and measure
systemic risks ex-ante.
The aim of the paper is twofold: on one side, it reviews the theoretical frameworks which allow to assess the
different dimensions of systemic risk and, on the other, it classifies accordingly and analyzes the methodologies
available to assess in advance the occurrence of systemic distress.
The paper classifies the different definitions of systemic risk and discusses their significance during the 2007-08
crisis. It presents the tools available to extract real time information on market perception of risk from market
prices of securities and derivatives (i.e. CDS and equity options). The analysis is extended to the methods
focused on the measurement of the financial fragility due to the networks linkages within the financial system.
On the basis of the available empirical research, the paper also reviews the capacity of the different methods to
spot in advance the insurgence of the crisis prior to 2007-08 and draws some preliminary conclusions on the
completeness and consistency of the toolkit available to policy makers.
Keywords:systemicrisk,financialcrisis,prudentialregulation,financialinstitutions.
Jel classification: G01, G18, G 21, G28.
2
Introduction
The 2007-08 financial collapse has catalyzed the attention of scholars and policy-makers on the very nature of
financial crises, stimulating a new, fresh wave of research on the topic, aimed at defining and measuring systemic
risks. Understanding the nature of systemic risk and identifying the channels by which shocks spread are the
necessary prerequisite for anticipating and successfully managing the onset of financial crises. In order to
prevent financial distress and manage instability, the macroprudential regulator needs to track and measure
systemic risks ex-ante. Has the massive wave of research prompted by the crisis produced new tools able to
anticipate financial distress? How do these innovative tools work?
Conceptually, systemic risk has multiple dimensions. Systemic risk shows a time-varying pattern (which follows
the build-up of financial imbalances over time) and a cross-sectional structure (which determines the degree of
fragility of the system and governs its resilience to shocks). Financial shocks are endogenously fueled (stemming
from the co-dependent behaviors and chain-reactions of the financial institutions themselves). The time-varying
and cross-sectional dimensions of risk are compounded during the run up to a crisis. The disruptive impact of
the crisis transcends the resilience of each of the institutions involved. Individual soundness does not add up to
aggregate stability. The non-linear properties of financial networks represent a major challenge to the
microprudential approach to bank regulation.
The aim of the paper is twofold: on the one hand, it reviews the theoretical frameworks which allow the
assessment of the different dimensions of systemic risk, while on the other it classifies the methodologies
available for the advance assessment of the potential for systemic distress accordingly and moves on to analyze
them.
The paper is divided into four sections. In the first section, the paper classifies the different definitions of
systemic risk and discusses their significance during the 2007-08 crisis. In the second section, it presents the
tools available for the extraction of real-time information on market perception of risk from market prices of
securities and derivatives (i.e. CDS and equity options). In the third section, the analysis extends to the methods
focused on the measurement of financial fragility arising from the linkages between networks within the
financial system. In the fourth part, the paper reviews, on the basis of the empirical research carried forward
mostly by the IMF, the capacity of the different methods to spot in advance the insurgence of the crisis prior to
2007-08. The fifth briefly suggest how to organize operationally the informational content of different tools.
Some concluding remarks are put forward in the final section.
1. What is systemic risk?
1.1 Definitions of systemic risk
The notion of systemic risk usually refers to the probability of a collapse of the financial system prompted by
unidirectional and simultaneous downside co-movements of asset prices and/or by a generalized draught of
liquidity. Systemic risk is the risk of a banking crisis when the defaults of one or more banks appear to be chain-
connected.
Understanding the genoma of systemic risk and identifying the channels through which it spreads are the
conceptual and empirical prerequisites needed to anticipate the occurrence of financial and banking crises.
However, even the concept of systemic risk is not uniquely defined [de Bandt et al. 2009]. Sometimes it is
referred to as an exogenous and unexpected macro-shock affecting many banks at once (as in the case either of
a deep recession which feeds back into bad loans for most banks or of a fall in asset prices triggering a
generalized process of deleveraging); on other occasions, the notion of systemic risk relates to the chain reaction
prompted by the default of one debtor which translates into the default of its creditors and then, with further
cascade effects, into the default of the creditors of the latter. In this case, it is neither the original source of shock
3
nor its size that matters, but the nature of the endogenous self-fulfilling process of diffusion which makes the
crisis implode.
By nature, banking is highly exposed to both risks: banks are vulnerable to exogenous shocks because their
activity involves maturity mismatch between assets and liabilities. At the same time, banks are directly linked
through the network of interbank deposits. Furthermore, banks operate mostly on the same segments of the
financial market, often share the same business model and adopt the same risk management procedures: all
these features make them exposed to the same risk factors and make them prone to adopt similar behavior in
case of crisis.
In addition, there may also be indirect channels of distress transmission (i.e. channels not implying direct
connection between the subjects involved). For example, due to information asymmetries, even solvent banks
may be affected by the uncertainty generated by a bank default. The more similar their risk profile to that of the
defaulted bank, the higher the probability attached by market participants to the event that they may be heading
the same way (irrespective to their actual solvency) . It follows that fund withdrawals and liquidity shortages
could affect, at the same time, not just insolvent banks, but also banks which are perfectly sound, pushing them,
too, towards undeserved distress. [Aharony, Swary 1996, Revell1975]. Of course, faster the rate of contagion,
the higher the vulnerability of each single bank involved (i.e. lower capital ratio, higher leverage, higher maturity
mismatch etc.).
In order to define the perimeter of macroprudential control and identify adequate monitoring tools and policy
instruments, it is useful to dig into two complementary features of systemic risks: (i) the endogenous nature of
financial fragility; (ii) the structural complexity of financial systems.
i. Exogenous shocks vs. endogenous cycles. Before the 2007-08 financial crisis, most of the empirical literature on
financial distress focused on modeling exogenous shocks and their quantitative impacts over time.
Stress tests, and econometric simulations (such as vector auto-regression impulse-response analyses)
belong to this tradition, which is based on the assumption that the structure of the financial system is
given and does not modify over time. However there is also an alternative approach which focuses on
the internal dynamic of the financial system itself as a major engine of financial fragility and instability1.
This approach (often neglected because of the dominant paradigm of market efficiency) postulates that
the genesis of financial imbalances is rooted in the financial behavior prevailing during periods of
economic expansion. Those imbalances compound over time, increasing the fragility of the system, up
to the point where they turn out to be unsustainable. When the breaking point is reached the financial
implosion is sharp and huge. The extension of the crisis has no apparent relationship to the size of the
first shock triggering it (which is sometimes even undetectable), but depends rather on the size and
diffusion of the financial imbalances accumulated in the past. In this dimension, systemic risk is
correlated to the pro-cyclicality of agents’ behavior, it is dynamic in nature and it can be detected only
through observation over long time spans. The focus on the endogenous cycle of financial fragility
implies that in order to anticipate systemic financial distress, authorities must control the accumulation
of financial imbalances over time, by monitoring key indicators such as excessive credit expansion,
excessive leverage or asset bubble inflation. This dimension is labelled as the time-varying dimension of
systemic risk2.
1Minsky [1982]; Kindleberger, Aliber [2005]
2Kyotaki, Moore [1997]; Borio, Lowe [2002a; 2002b]; Borio, Drehmann [2009a; 2009b]; Brunnermeier [2001]; Borio
[2013]
4
ii. Structure of the financial system. The common feature of any systemic crisis is the velocity of diffusion,
which depends on the nature and strength of direct and indirect linkages among agents. As we have
already seen, the working of any banking system requires a wide network of direct financial connections,
through both the payment system (clearings) and interbank deposits. Being exposed to the same risk
factors, banks are also vulnerable through indirect channels (such as runs on deposits and/or assets
sales). Given their cross-country/cross-currency operations, banks are also the vehicle of international
diffusion of shocks. Microprudential tools (such as capital ratios and caps on leverage) may moderate
banks’ vulnerability, since the speed of contagion is higher when the banks involved along the
transmission chain are weaker. However, stronger defensive lines at the micro level may prove an
insufficient antidote against systemic crises: the overall structure of inter-linkages within the system
could overwhelm individual balance sheet equilibria, generating an explosive pattern of feed-backs. This
aspect of systemic risk is not adequately captured by the dynamic over time of financial aggregates (time-
varying dimension), but also requires a specific assessment of the structure of the banking/financial
network and the measurement of its internal interconnections at each point of time. This is called the
cross-section dimension of systemic risk [Allen, Babus 2008; Gai et al. 2011].
1.2 Systemic risk indicators and measurement metrics
For the combating of financial distress to be viable, systemic risk must be traceable and measurable. Both the
accountability of macroprudential authorities and their ability to prevent financial distress depend on the proper
measurement of systemic risk. In the first case (accountability) authorities can rely on ex-post indicators, which
signal the build-up of imbalances able to trigger a financial crisis. Whenever such imbalances reach a critical limit
(identified by looking back at past experiences of financial distress), action is within the domain of
macroprudential supervisors. However, ex-post evidence of unsustainable imbalances, albeit necessary, is not a
sufficient condition for triggering macroprudential action. Waiting for financial distress to show up (either at
micro or at macro level) could substantially weaken the effectiveness of prudential policy. In order to prevent
the occurrence of distress, macroprudential authorities need also to assess its probability in advance. In other
words, prevention requires ex-ante or forward-looking indicators of distress, able to measure both the potential
vulnerability of the system and the proximity of financial disruption. As a matter of fact, the financial system
could function and grow for very long periods even in the presence of major imbalances. As stated by Financial
Instability Hypotesis [Minsky 1982], it is actually during the good phases of the cycle that financial imbalances
build-up, because economic agents do not perceive the dangers of moral hazard and high leverage ratios are
generally considered a positive fuel for growth and profitability. Systemic risk is exposed to a paradox: it tends to
accumulate when liquidity is abundant, volatility is low and risk premiums are thin. In a nutshell, systemic risk
behaves like an asymtomatic pathology that works undetected, weakening the immune defences while the
patient is apparently sane, but exposes the body to major threats when the pathology manifests itself.
In order to act in time and monitor the effectiveness of policies, macroprudential authorities need diagnostic
tools that must be not only reliable but also available within a useful time span. This severely limits the universe
of data which can be processed within the necessary time. In addition, since systemic risk is a latent factor,
authorities must also rely on counterfactual approaches (such as, for example, stress tests and network
simulations), which are by nature conditional on discretionary model representations and calibration choices.
As mentioned in the previous section, systemic risk has multiple dimensions. Systemic risk has a time-varying
pattern (which follows the build-up of financial imbalances over time) and a cross-sectional structure (which
determines the degree of fragility of the system and governs its resilience to shocks at any given point of time).
It means that financial shocks are not only the effect of exogenous shocks, but are also fueled by endogenous
factors (stemming from the co-dependent behaviors and chain-reactions of the financial institutions
themselves). Time-varying and cross-sectional dimensions of risk compound during the run up of a crisis and the
5
disruptive impact of the crisis transcends the resilience of each of the single institutions involved. Individual
soundness does not sum up to aggregate stability. The non-linear properties of financial networks represent a
major challenge to the microprudential approach to bank regulation.
Each of these dimensions of systemic risk appears to be measurable using different tools [Noera 2013]. Since
the 2007-08 financial crisis a new wave of research has been trying to refine and test methods able to offer, on
the one hand a forward-looking approach to the measurement of systemic risk (mainly by extracting market
expectations from the pricing of securities and derivatives) and on the other hand able to highlight the
assessment of systemic vulnerability (by looking at the properties of the financial networks and the strength of
the inter-linkages among financial institutions). Table 1 offers a bird-eye view of the main indicators of systemic
risk now available to macroprudential authorities. They are classified according to: (a) the nature of the data
from which they are drawn (i.e. macro-statistics, accounting data; market data); (b) the methods of processing
(ratios; statistical-econometric estimates; model simulations); (c) their focus (either on individual institutions or
system-wide). However, the most important distinction among them is between time-varying and cross-section
indicators.
The distinction according to the nature of the data is sometimes straightforward: on one side, there are the usual
financial statistics referring to the financial system as a whole (mainly credit and/or debt aggregates) and/or
accounting data (either referred to single institutions or consolidated). Most of these indicators are simple ratios
or rates of growth of the quantities observed (financial soundness indicators, FSI; credit growth; debt-to-
income, DTI etc.). On the other hand, there are methodologically more complex indicators based on market
prices of both securities (bonds) and derivatives (equity options, CDS). Some of them are used to assess the
implied probability of default of single institutions (distance-to-default or DD; implicit probability of default or i-
PoD; higher moments analysis of the univariate probability of default). Other indicators apply to multiple
institutions and take into account the structural interconnections within the system (Co-Risk indicators). Some of
the latter can be observed also in their dynamics over time (time varying multivariate distress dependence) [Bisis
et al. 2012].
Table 1 – A Taxonomy of Systemic Risk Indicators
6
In general, the basic indicators of financial stress, both macro and micro3, are built on data available only at low
frequency (monthly, quarterly and even annually) and are backward-looking (ex-post accounting measures). By
nature, their predictive power is poor. The time patterns of these indicators is slow moving as they tend to
signal the progressive accumulation of disequilibria that, once calibrated on past stress episodes, may help the
macroprudential supervisors to identify critical thresholds beyond which the probability of a crisis is assumed to be
increasing4.
At the macro level, useful indicators of financial stress are the positive deviations both of the credit-to-output
ratios and asset prices (mainly equity and real estate) with respect to their respective medium term trends [Borio,
Lowe 2002a, 2002b; Borio, Drehmann 2009a]. The size of the deviation of asset prices from the trend signals
the progressive inflation of bubbles, the increasing likelihood of a burst and the severity of the subsequent
adjustment. The size of the deviation of the credit/GDP ratio from its trend captures the increasing
vulnerability of the financial system, as at higher levels of aggregate credit dependence the financial system
becomes less able to absorb losses.
The predictive power of these indicators improves significantly if they are analyzed jointly. By estimating a
probit model where probability of distress is a function of multiple risk factors and their
interactions(Credit/GDP gap and growth; equity and house prices growth; banking sector leverage etc.) and
calibrating a time-varying threshold for each risk indicator, Lund-Jensen [2012] shows that the joint assessment
of the probability of systemic distress improves significantly, reducing the costs associated with false signaling.
The methodology makes the observation of time-varying/slow-moving indicators more efficient and reliable,
allowing the monitoring of the build-up of potential systemic instability several years in advance.
Even though credit-to-output ratio and asset price imbalances are sometimes able to signal the risk of a financial
crisis in advance, they are incapable of identifying the exact point in time when the disruption actually occurs.
This means that additional information is needed on the prevailing mood of market participants. Indicators
based on market prices try to fill this informational gap. The rest of the paper is dedicated to this family of
forward-looking indicators (the so called near-coincident indicators).
2. Market assessment of default risk
2.1 Single institution i-PoD
Market sentiment indicators can be extracted directly from market prices (equity, bonds and credit default
swaps) in the form of implied probabilities associated by the market to the event of distress (either related to
individual institutions or to the system as a whole). The intuition behind this approach is that market pricing of
3 Among the most common macroindicators are: credit/GDP ratio, debt/GDP ratio (both public and private); equity
prices and real estate prices (all the variables may be observed both in terms of absolute level and in terms of rate of
growth). The most common balance-sheet microindicators measure: (a) capital adequacy (capital/assets ratio; tier1
capital/assets ratio; tier1 capital over RWA; tier1+tier2 capital over RWA); (b) asset quality (non-performing loans;
provisions); (c) leverage ratio (debt over capital; share of short term debt); (d) liquidity (loan/deposit ratio; loan/asset
ratio); (e) profitability (RoA; RoE); equity valuation (PE ratio; EPS; P/B ratio). IMF [2006; 2011a].
4 Time varying indicators are based on “signal extraction” techniques, which look at the deviation of the observed variable (for
example: credit growth) with respect to its long term trend: when the deviation widens beyond a pre-definite threshold, the
indicator assumes the signal 1 (while during normal times it is set to 0). Theoretically, when the indicator is 1, a crisis
should follow. However, when tested in sample (i.e. on historical data), the indicator may give false signals. The errors are
classified in two categories: type 1 error is when the indicators fails to signal a crisis which actually occurs; type 2 error is when
it signals a crisis that fails to materialize. Therefore, calibration is needed in order to optimize the trade-off between the
two types of errors and get reliable signals that financial distress is actually building up [Borio, Drehmann, 2009a].
7
either CDS or equity options embeds the expectations about the probability that market values of the firm’s
assets fall short of liabilities (default threshold) (Figure 1).
Fig.1 - Equity volatility and implicit probability of default
The underlying assumption of this approach is that the market constantly monitors the soundness of banks and
the capabilities of the management and that this assessment is directly reflected in equity prices and CDS
spreads. Under the hypothesis that the market is efficient (i.e. that prices fully incorporate all the available
information), the information deduced from market prices is inherently forward-looking, as it reflects expectations.
In other words, equity option premiums and CDS spreads are the raw materials from which the implicit
probability of default (or i-PoD) expected by the market for each listed financial institution5 can be extracted.
Since i-PoDs try to capture the risk assessed by market participants at any point of time, they are very sensitive to
the likelihood of default (either actual or just perceived). Unlike (time varying) slow moving indicators, this category
of signals reacts with a very short lead-time (few months or weeks) as the critical point of disruption approaches:
that is why they are also known as near-coincident [Arsov et al. 2013].
3. The cross-section dimension of systemic risk
3.1 Multivariate distress dependence
When applied to a single institution, the i-PoD indicators do not account for systemic risk arising from direct
and indirect inter-linkages among financial bodies. One interesting research development of this approach is the
extension of similar techniques to estimation of the cross-probability of default of each institution conditional to
the probability of default of any other [Segoviano, Goodhart 2009]. Looking at the financial system as a
portfolio of banks, this approach estimates the multivariate probability distribution of distress of the whole
system and extracts a set of indicators of the joint probabilities of default of any pair of banks or groups of
banks from the multivariate distribution, implicitly taking into account the structure of their cross-correlations.
5Following the Merton [1974] contingent claims approach, implicit volatility can be used to estimate the expected future value
of the assets (A) of the bank (which is assumed to follow a stochastic path) and of its equity capital (E), which can be
treated as a call option on assets, with strike price equal to the maturing debt (L). Given the value of collateral plus
guarantees (H), the default risk for the creditors of the bank is equal to total liabilities net of collateral and/or guarantees.
Given the two equations (budget constraint A=D+E and risky debt D=L-H), the system can be solved for two unknowns
(assets valuation and implicit volatility of assets) which allow derivation of the implicit probability of default of the bank (i-
PoD). Tarashev, Zhou [2006; 2008]; Capuano [2008].
8
There are several indicators elaborated using this approach which have different focus: (a) the J-PoD (or joint
probability of distress) measures the probability that all the banks in the sample could default (this estimate is
equivalent to the tail systemic risk of default)6. Further systemic risk indicators can be calculated from the J-PoD
: (b) the Banking Stability Index (or BSI), which estimates the number of distressed banks associated to the case in
which at least one of the others is distressed (the larger the number of banks exposed to contagion, the less
stable the system); (c) the PAO or Probability-that-at least-one-bank-becomes-distressed as a consequence of the default
of a specified bank in the sample (i.e. Lehman Brothers; AIG etc), which can be also considered a measure of
the systemic relevance of each single institution7; (d) the Distress Dependence Matrix (or DDM), i.e. the double-
entry matrix of the cross-probabilities of distress of each bank, conditional on the probability of the distress of
each of the others (Table 2). The DDM shows the probability of distress of each bank indicated in any row,
conditional on the probability of default of each bank listed in any column. Even though there is no causal
direction in any bilateral linkage, the DDM maps the interconnectedness among institutions and accounts for the
non-linearities which characterize the contagion effect during episodes of financial stress.
Table 2 – Distress Dependence Matrix
Source: Segoviano, Goodhart [2009].
3.2 Co-Risk Measures
A similar strategy for monitoring systemic risk consists of the direct tracking of the linkage between the risk
exposure of several institutions. These co-risk indicators try to measure the variations of overall risk, conditional
on the event that one institution could default.
Inputs can be either the single Value-at-Risk (VaR) [Adrian, Brunnermeier 2009], CDS spreads (or bond risk
premia) or measures of entropy8. Since these indicators are also based on market prices, they implicitly account
for the market assessment of both direct (i.e. interbank) and indirect linkages (such as homogeneity of business
models, similar asset structure and risk management methods) and could also capture endogenous risk-feeding
factors.
The Co-VaR approach proposed by Adrian and Brunnermeier [2009] measures the Value-at-Risk (VaR) of any
financial institution conditional on the probability that other institutions fall in distress. The marginal
contribution to systemic risk of each financial institution is given by the difference between its own Co-VaR and
the total VaR of the whole system. Correlations between single Co-VaRs and the total VaR identify the extent
of contagion effects within the system, even though correlation analysis is unable to identify the causal structure
6 This technique is based on the estimation of a multivariate density function of the banking system (BSMD). The BMSD
function is estimated through the CIMDO-copula technique which captures both linear and non-linear correlations using
single PoDs as inputs (derived either from equity option or CDS spreads). Segoviano [2006].
7 An extension of the Segoviano and Goodhart [2009] methodology shows that the systemic importance of financial
institutions (SIFI) does not correlate to their size, but just to the probability that each one of them could influence the
stability of the others. Zhou [2010].
8 Chan Lau [2009]; Tarashev, Zhou [2006;2008]; Capuano [2008]; Segoviano, Goodhart [2009].
9
of systemic risk and is technically inadequate to capture the non-linearities which distort the significance of
correlation coefficients during financial crises9. In order to fully account for such non-linearities, other studies
have adopted either alternative approaches of risk measurement10 or different techniques of parameter
estimation11.
3.3 Network model simulations
After the 2007-08 crisis, the theory of complex systems (or networks) has been rediscovered and applied to
financial markets in order to obtain a better understanding of the role played by interconnectedness in financial
markets [Allen and Babus, 2008].
Network topology is a tool widely applied in several fields of research (physics, biology, ecology and
engineering), but it has been generally neglected in economic disciplines. Traditional economic and financial
theories are not endowed for the understanding of complex ecosystems. Even if individuals were rational and
markets were efficient, the aggregation of individual behaviors still does not sum-up linearly to the collective
behavior of the system as a whole. Since the structure of internal feed-backs is neither linear nor homogeneous,
complex systems are unstable. The more complex the system, the higher its potential fragility. Mainstream
economic theory, based on the paradigm of the representative agent, does not even see the issue [Haldane 2009].
A network is a set of agents (called vertices or nodes) linked by multiple connections (edges) of which the statistical
properties can be analyzed and appropriate measurement criteria (such as the length of the connecting paths or
the distribution degree) (Figure 2) can be defined. When dealing with very large networks (thousands of
vertices), it is useful to build simulation models to understand the internal dynamics of the network (i.e. how the
network assumes a particular shape and how the vertices interact). Based on the network structure and given
behavioral rules of the vertices, the model allows the researcher to observe the aggregate behavior of the system
[Newman 2003].
Applied to the banking system, network analysis allows empirical simulation of the final impacts of domino effects
(or chain reactions) beyond the point where the initial shock originates. Empirically, this methodology requires
the reconstruction of a double entry matrix of data, which collects the entire set of bilateral exposures of each
bank with respect to each of the others. Given the data matrix, it is possible to simulate a shock initially hitting
one or more institutions and to track the subsequent chain effects (direct and/or indirect) for a number of
successive rounds [Chan Lau, 2010]. Several studies have adopted a similar research strategy, exploring the role
played by payment systems, interbank markets or asset markets as channels for the spread of systemic shocks
[Allen, Babus 2008].
An alternative strategy in network analysis has been adopted by a group of economists at the Bank of England.
In 2008, Nier et al. [2008] built a laboratory model where banks are linked through interbank deposits and where
9Risk co-movements are not linear and non-linearity becomes more pronounced during periods of financial distress:
systemic risk increases more than proportionally with respect to the traditional risk measures based on (log)normal
distributions.
10 Application of the Extreme Value Theory (EVT) to multivariate distributions allows estimation of interdependencies
among tail risks (joint tail dependence) capturing the probability of extreme shocks. However, the ETV approach misses a
significant portion of data information, which makes it inapplicable when the time series available are too short. Poon et
al. [2004]; Rocco [2011].
11 An alternative way to track non-linearities during financial crises is quantile regression analysis. While traditional regressions
capture the average relationship among the variables, the quantile regression is estimated using only tail observations (i.e. the
95th quantile of data distribution), which represent just extreme states of financial stress. Koenker, Hallock [2001]; Chan
Lau [2009].
10
the behavior of the network is analyzed by setting alternative values for key-parameters (such as capital ratios,
size of reciprocal exposures, degree of interconnectedness, concentration, etc.). The exercise generated findings
common assumptions that the vulnerability of the system to shocks increases with the size of credit/debt
exposures and that the banks most resilient to contagion are the best capitalized ones; on the other hand, the
analysis also showed that immunization to shocks is not linearly proportional to banks’ capital ratios:
surprisingly, there is a level of aggregate capitalization below which minimum capital ratios are not sufficient to
stabilize the system. In addition, the study has discovered that the effect of connectivity among banks does not
behave monotonically: a small variation in connectivity may substantially increase the probability of contagion;
however, if connectivity grows beyond a certain threshold, it significantly improves the system’s capability to
dilute shocks. In other words, dispersed networks are more stable than concentrated networks.
Figure 2 – Network Topology
Source: Newman [2003].
Using the same approach, further studies at the Bank of England [Gai, Kapadia 2010; Arinaminpathy et al. 2012]
have shown that the probability of contagion does not fully capture the potential exposure to systemic risk: even
when such probability is low, minor shocks may have a very large negative impact due to the internal structure
of the network, where the degree of connectivity and the nature of the edges can compound the feed-back
effects within the system. For the same reason, shocks that appear similar in nature and magnitude may have
impacts which turn out to be very different, due to the relative importance of the institutions (vertices) first hit by
the shock12. If the institutions impacted first are either those with the largest exposures or those with the
highest degree of connection, the final effect tends to be stronger.
These results help to focus macroprudential policies on the structural characteristics of the system too, and
highlight the importance of concentration both in the size of single institutions (the well-known too-big-to-fail
issue) and in their connectedness (the newly discovered too-connected-to-fail issue).
4. Could the 2007-08 financial crisis have been overseen in advance?
As we have seen, the systemic risk shows up in multiple dimensions: exposure to the same risk factors,
procyclicality and network interlinkages. After the crisis most of the research efforts have been aimed at
identifying and measuring both the time varying and the cross-section faces of systemic risk. Most of the tools
now available have been applied ex-post to the 2007-08 financial crisis, in order to check their explanatory
power. Even though is goes beyond the scope of his paper to test empirically the alternative approaches, it is
12 It confirms that bank size is not a sufficient indicator of systemic importance. The contribution of banks to systemic
fragility increases more than proportionally with respect to their size as a function of the connectivity and concentration of
the system.
11
interesting to review also the evidences that could have been available to the authorities in the eve of the crisis, if
–at the time-they had been endowed with the proper toolkit13.
4.1 Balance-sheet indicators
According to IMF, the indicators based on balance-sheet data (which are backward looking/low frequency
indicators), would have not supplied signals useful to prevent the crisis. Indicators of capital adequacy were not
reliable at identifying ex-ante the institutions requiring intervention by the MAP. Though the working
assumption is that low capitalized institutions are the most vulnerable, the degree of capitalization of the banks
that actually required public intervention during the crisis was not ex-ante significantly lower than the
capitalization of the banks that did not required any intervention14.
The poor signalling power of capital/asset ratios could be attributed, to some extent, to the difficulties of
measuring the actual riskiness of financial assets15, to the imperfections of mark-to-market accounting16, to the
elusive nature of the shadow banking system17 and to regulatory arbitrage18.
For the same reasons, retrospective analysis shows that also common indicators of asset quality (as the
percentage of non-performing loans or NPL) were not good predictors of the crisis: also in this case the
incidence of NPL of banks that suffered severe financial stress was lower than the one of banks which required
public financial support during the crisis19. To the same token, also liquidity indexes did not show any significant
predictive power, since they were not able to capture the increasing importance of banks’ liability management
techniques in the interbank and repo markets20. .
More informative to detect in advance potential balance-sheet vulnerabilities would have been instead other
indicators as the degree of leverage, the return on assets (RoA) and most of the common stock market multiples
(as the P/E; P/BV and EPS ratios). In the eve of the crisis all these indicators showed up significantly higher in
the banks that subsequently required support21. These indicators are analytically consistent, since high levels of
RoA and leverage can be interpreted as indirect measures of the bank’s propensity to risk, aimed at maximizing
the shareholder value (return on equity), which in turn influences stock valuations.
A good warning signal would have been offered also by the percentage of banks’ short-term indebtedness
(interbank and repos):this evidence suggests that the degree of leverage, which represents the dominant
incubator of financial stress, must be coupled with indicators of maturity mismatch22.
13Further insight into the empirical results reviewed here are in IMF[2009; 2011b] to which this paragraph mostly refers.
14This evidence applies to US banks but does not extends to European and Asian banks, where capital/asset ratios were,
before the crisis, higher for those institutions that did not required any intervention IMF [2009].
15 Le Lesle, Avramova [2012].
16 There are significant differences between US accounting standards (GAAP) and Europeans’ (IAS-IFRS).
17. Feder, .Mitchell [2009].
18 Cannata, Castellina, Guidi [2012].
19 IMF [2009].
20 Morris, Shin [2008].
21 IMF [2009].
22 BCBS [2010]; A. Haldane [2012]; Liikanen Report [2012].
12
Balance-sheet indicators at micro level have a natural counterpart, at macro level, in the financial cycle indicators
(i.e. credit growth; credit/output or debt/output ratios; asset prices etc.)23. Applied to the US pre-crisis situation,
some of these indicators were releasing signals of imbalances which were growing unsustainable since early
2000s (Figure 3)24.
Fig. 3 – Financial cycle macroindicators
Source: Borio-Drehmann (2009)
The same indicators, applied to other countries, failed because they are unable to detect the stress originated by
external imbalances (i.e. foreign debt of the banking system) which were the main channels of international
contagion (Germany, Netherland, Switzerland; Canada). In this case, the indicators should have taken into
account the weighted exposure of each bank towards each other country. Such an exercise requires a very
granular database of banks’ cross-exposures25.
Since the 2007-08 crisis originated in the US and then spread around the world, the analysis of the interlinkages
across different banking systems is crucial to assess the potential chain-effects of the shocks originated outside
any single country. The IMF has extended the analysis of network effects, looking at the transmission of
financial contagion across countries26. Given the matrix of cross- exposures (where, on the asset side of each
bank’s balance-sheet, appears any single exposure towards any other bank and, on the liability side, any single
23 Borio, Lowe [2002a; 2002b]; Borio, Drehmann [2009a]; CGFS [2012].
24 Borio and Drehmann [2009a] adopt three indicators simultaneously (credit growth to the private sector; stock market
prices and real estate prices) and define an early signal of potential distress when at least two out of three variables go
beyond their threshold (see footnote for a definition). When calibrated over the period 1980-2003 and tested out-of-sample
over the period of financial crisis (2004-2008), both the credit growth indicator and the real estate indicator have been
able to jointly predict the crisis starting several years in advance (2000-2001). Significantly less efficient appeared the stock
market indicator, probably influenced (during the calibrating period) by he burst of the dot.com bubble. Given the relevant
weight of the stock market, also the composite index of stock and real estate prices do not perform well, while in previous
work of the same authors the stock market index (calibrated before 2000) approximated the threshold (even though failing
to release unambiguous signal of financial instability). Borio-Lowe (2002b).
25Borio, Drehmann [2009]
26 Chan-Lau [2010]; IMF [2009].
13
debt position), the IMF has studied the impact across the system of the default of either one or several banks,
simulating the chain effect through the transmission channel of the network of reciprocal exposures.
Corresponding to the initial shock, several rounds of potential defaults have been identified and measured
(banks can be affected in the second round by banks affected in the first, and in turn, they become vehicles of
higher order contagion effects). The domino effect is strengthened by the assumption of a liquidity constraint
(i.e. interbank draught) and forced fire sales of assets. Applying this approach to the structure of cross-country
interconnections among banking systems at March 2008 (i.e. at the time of the distress of Bear Stearns and six
months before Lehman Brothers’ default), the IMF exercise confirms, on one side, the centrality of US and UK
banking systems as critical epicenters of financial shocks and, on the other, allows an ordering of countries
according to their relative financial vulnerability to different types of contagion. The IMF analysis shows, for
example, that Belgium. Netherland, Sweden and Switzerland appeared the most vulnerable countries to credit
shocks (shocks transmitted through direct and indirect credit exposures); adding a liquidity shock (the drought
of 45% of both interbank and repo markets and a 50% fall in asset prices) France appeared instead the most
fragile country outside US and UK [IMF 2009].
The monitoring of time-varying indicators of systemic risk have been tested by the IMF[2011b] on a broader base
of countries27 applying the methodologies explained in Lund-Jensen [2012]28. The tests show an encouraging
evidence that slow moving indicators bring precious operative information to policy makers about systemic risk:
(a) credit/GDP growth and credit/GDP gap have positive significant effect on systemic risk up to three years in
advance (for example 1 std deviation increase in credit/GDP gap increases the systemic risk for the median
country by 3.3%, 2.6% and 1.3% in the 3 following years respectively); (b) banking sector leverage appears to
have a significant effect on systemic risk two years in advance (for example: 1 std deviation increase in leverage
increases systemic risk within one year by 2.5% for the median country) 29. Asset price inflation and real
exchange rate appreciation are also found to have positive and significant effect on systemic risk up to three
years in advance. As far as the efficiency of signal extraction is concerned, joint observation of multiple factors
could release a more precise assessment of the danger embedded in the building up of financial imbalances: for
example a crisis signal could be extracted for the median country when credit/GDP growth is above 4.9%,
however the reliability of the signal can be controlled observing the joint impact on systemic risk also of banking
leverage (indicating financial vulnerability) and of the rate of growth in equity prices (detecting potential
bubbles)(Figure 4 ).
Fig. 4 – Systemic Risk and Crises Signals
Source: Lund-Jensen (2012)
27The model was estimated on an annual panel of 68 advanced and emerging countries over the period 1970-2010.
28See par.1.2 supra.
29The results reported here are based on the logit specification of the model of Lund-Jensen [2012].
14
4.3 Market based i-PoD
Estimating the implied probability of default (i-PoD) embedded either in option premia or CDS spreads30 allows
to detect in real time the sentiment prevailing in the market. It allows also to track the evolution of market
sentiment over time, through the analysis of the shape of the probability distribution (i.e. its higher order
moments: skewness, kurtosis and gamma31).
During financial distress, the implied probability distribution associated to equity prices deviates from
“normality”, reflecting non linearities linked to the systemic dimension of risk. It has been shown that, during
periods of distress, even very small changes in the perception of default risk generate disproportionate volatility
on equity prices. It follows that if the distribution is skewed towards the left (i.e. a negative skeweness), also the
distribution of firm’s asset values estimated by the market must be similarly skewed towards lower values. It
implies, in turn, that the perceived probability of the firm’s asset values falling short of its liabilities is higher (a
similar attitude could be read in CDS spreads32).
The analysis of higher moments of i-PoD referred to institutions distressed during the financial crisis33(Figure 5)
shows that in August 2007, shortly after that the major central banks had jointly injected massive liquidity into
the system34, the distributions overshooted with respect their normal shape, getting taller (higher positive
kurtosis), bending toward the right (higher positive skewness) and with a fatter right tail (higher gamma
positive): it was reflecting the jump reassessment of the probability of default following the government
intervention. Afterward (between September 2007 and the collapse of Lehman Brothers in September 2008)
skewness and kurtosis started to decrease towards zero (distribution was shifting towards the left and flattening),
while the fatness of tail risk (gamma) was initially deflating (the bulk of default risk probability was no longer
extreme but central) and then getting thicker on the left in the eve of the Lehman default. After the Lehman
collapse a new jump reassessment toward a normal shape occurred, following the renewed willingness of the
government to save the day (AIG bailout, TARP etc).
30 Under the assumptions of market efficiency and risk neutrality.
31 Skewness is equal to 0 when there is perfect symmetry; it is >0 when the distribution in skewed to the right and, vice
versa, it is <0 when the distribution is skewed to the left. When kurtosis is >0, the distribution is taller than he “normal”
(leptokurtic); when kurtosis is <0 the distribution is flatter than the “normal” (platkurtic). Gamma measures the fat tail risk: the
higher its value, the fatter the univariate tail risk.
32Capuano [2008].
33 IMF estimates are based on implied volatility derived from at-the-money put options, whose expiration was the closest
to the day of the distress.. The institutions involved in the i-PoD monitoring are (distress date in parenthesis): Bear Stearns
(14/3/2008), Lehman Brothers (15/9/2008), Merrill Lynch (15/9/2008), Wachovia (29/9/2008) and Citigroup
(24/11/2008). Ex-post the warning signals appear very informative for Bear Stearns, Merrill Lynch and Wachovia, while
appear weaker for Lehman Brothers and Citigroup, IMF [2009].
34 On August 9-10 2007, reacting to the crisis of three BNP mutual funds highly involved in ABS/CDO investments and
the freezing of interbank markets, Federal reserve, BCE and Bank of Japan jointly injected 330 billion US dollar of extra
liquidity in order to soften the liquidity crisis, signaling the determination of monetary authorities worldwide to stop the
crisis.
15
Fig.5 i-PoD Higher Moments Analysis
during 2007-08 crisis
Source: IMF ( 2009)
4.3 Market based J-PoD, BSI and PAO
The estimation of a multivariate probability of default allows to take into account of systemic risk in both its
cross-section and time-varying dimensions.
Applied to the 2007-08 period, J-PoD and BSI show that the default risk of each single institution is substantially
higher when the linkages with other institutions are taken into account than when the single institution is
analyzed in isolation. What is even more interesting is that, since J-PoD, BSI and PAO are extracted from values
observable on a daily basis (option premiums and CDS spreads), they can also be traced in their (high frequency)
time-varying dimension (Figure 6).
During 2007-08 cross-dependence among US financial institutions strengthened and the joint probability of
distress (J-PoD) increased more rapidly than the probability associated to the same single-out institution (i-PoD),
showing peaks in connection with the two major episodes of distress (Bearn Stearns and Lehman Brothers) and
starting to react 5-6 weeks in advance of the actual default.
The evidence confirms that, during crises, not only the probability of default of each bank is amplified
significantly, but also that the financial equilibrium of each institution becomes highly dependent on the health
of the others. It is also worth noting that the Banking Stability Index (BSI) –that measures the number of banks
defaulting if one defaults- starts increasing steadily at the beginning of 2007, adjusting temporarily only after
major public interventions. On the other hand, the Probability-of-at-least-one-default (PAO), which identify the
relative systemic importance of each institutions to produce cascade effects on the others applied to two of the
major hubs of the crisis (Lehman Brothers and AIG) shows an upward monotonic trend since the beginning of
2007, reaching the probability of 1 in September 2008 (certainty of a cascade effect approaching Lehman
collapse).
16
Fig. 6 Multivariate probability indicators of systemic risk during 2007-08
Source: Segoviano-Goodhart (2009); IMF ( 2009)
4.4 Co-Risk based on CDS spreads
Research at the International Monetary Fund35 has also applied quantile regression techniques to the linkages
across the CDS spreads of financial institutions in order to obtain estimates of the co-movements of their
default risk in the period July 1 2003-September 12 2008. The parameter estimates have been used to simulate
out-of-sample the CDS spread of each institutions conditional to the probability of default of each other
institution. The regression technique adopted (quantile regression) implies that only data falling in the extremes
tails of their distribution are taken into account: therefore, differently from OLS, this approach enables to
capture the non-linearities which characterize situations of acute financial stress.
The output of the exercise are double-entry matrices of a sort of augmented i-PoDs, that show the percentage
increase of the probability of default (i-PoD) of each single bank that stems from its linkages to any other. From
the same matrix it is also possible to build a vulnerability index of each bank, as a weighted average of the
sensitivity of its PoD conditional to the probability of default of all the others36.
For example, the evidence shows that, measured before the Lehman collapse, the average probability of default
for any other bank correlated to the case of a Lehman default was 22% higher at the beginning of the subprime
crisis (July 2007) and was almost twice as much (41% higher) one year later (August 2008), i.e. immediately
before the Lehman event. Therefore the MAP authorities could have spotted one year in advance the
destabilizing potential of the Lehman default (as of any other major institution).
35Chan Lau [2009]; IMF [2009]
36 The CDS spreads analyzed covered the following institutions: in USA: AIG, Bank of America, Bear Stearns, Citigroup,
Goldman Sachs, JP Morgan, Lehman Brothers, Merrill Lynch, Morgan Stanley, Wachovia, and Wells Fargo; in Europe:
Fortis, Paribas, SoGen, Deutsche Bank , Commerzbank, BBVA, B. Santander, Credit Suisse, UBS, Barclays, HSBC; in
Japan: Mitsubishi, Mizuho and Sumitomo.
17
5. Organizing the toolkit
The availability of operative tools aimed at quantifying systemic risk proliferated epidemically in the last few
years, giving MAP a much improved oversight on potential systemic events and a deeper capability to anticipate
distress. This paper has just selected the most significant indicators, but more are at hand [Bisis et al. 2012].
However only by putting the pieces in the correct position the puzzle delivers a coherent picture.
Some effort was already devoted to organize the information in an order that could turn out operationally useful
[FSB-IMF 2010; IMF 2011a; Blancher et al. 2013]. Properly managing systemic risk is a complex task that must
be based on solid knowledge and clear-cut decision processes. On this ground further work has still to be done
to link all the pieces of information available in an organic tableau de bord .
Despite such effort is beyond the scope of the paper, a preliminary ordering criteria of the tools reviewed has
been implicitly traced and it is worth making it explicit here (Figure 7). Phasing the crisis process (pre-crisis;
near-crisis; post-crisis), most of the indicators reviewed in the paper find their proper informational position: the
long term, slow-moving build-up of financial imbalances (that beyond some threshold turn out unsustainable);
the near-coincident dominance of a negative self-fulfilling market sentiment (which announces the actual
occurrence of the crisis); the post-crisis transmission of domino effects to the rest of the system (depending on
the network structure). As we have seen, for each phase, a different set of tools may be available, with different
technical features and different time horizons.
Fig. 7 Informational content of systemic risk indicators
6. Concluding remarks
In 2007-08, the international financial system reacted to a minor shock which originated in the US subprime
mortgage market with a self-fulfilling loss of confidence among banks, mainly due to the ubiquity of toxic assets
and the similar structure of balance sheets. The increase in credit and counterparty risks perceived by agents led
to illiquidity of securitized assets, drought of interbank deposits and forced deleveraging for most banks (which
because inadequately capitalized). Up to the Lehman default, government support was able to anticipate any
chain reaction, absorbing the losses and isolating the distressed institutions; however when the too-big/too-
connected hub of the network was hit (Lehman), the system reacted like a wounded ecosystem, irradiating
destabilizing waves through a tangled web of systemic inter-linkages (repos, interbank exposures, holdings of
18
illiquid assets, credit derivatives etc.), shifting abruptly towards the total collapse of the financial system37.
During the 2007-08 crisis, both dimensions of systemic risk (time-varying and cross-section) came into play,
reinforcing each other. The long period of credit expansion had been inflating the real estate bubble,
encouraging excessive leverage and maturity mismatch across the economy, while low interest rates and
abundant liquidity had been encouraging risk tolerance and moral hazard. Financial innovation and deregulation
had increased the complexity and interconnectedness of financial institutions, compounding financial fragility
Onado [2009]
Macroprudential supervision has been identified as the appropriate answer to the 2007-08 financial crisis [FSB-
IMF-BIS 2009] and during the last few years it has been assuming a clearer operative profile. A significant effort
has been dedicated to defining the final objectives specific to macroprudential policies, starting from the
identification of the market failures (or externalities) that could trigger systemic financial distress[De Nicolò et
al. 2012; ESRB 2013]. It has been recognized that some externalities arise endogenously from the behavior of
financial institutions themselves, amplifying the cross-correlations among the risk exposure of individual firms
(high leverage, similar business models; same risk management procedures); other externalities depend on
market and liquidity risks due to fire sales of assets, which could simultaneously damage the balance sheets of
multiple banks; a further source of externalities is the complex network of financial interconnections which link
institutions to each other. Appropriate policy instruments have been associated with each macroprudential
objective [ ESRB, 2013; CGFS, 2012; IMF, 2011a, 2011b, 2013a, 2013b; Gualandri, Noera 2014a].
However the monitoring of systemic risk is the necessary pre-requisite for timely and effective implementation
of macroprudential policies. Macroprudential policies face a continuously changing financial environment, where
several risk factors could combine unexpectedly. Since macroprudential action is mostly pre-emptive, it requires
the ex-ante evaluation and measurement of systemic risks [Goodhart, Perotti 2013].
Under the pressure of the 2007-08 experience, a wide and diversified range of diagnostic tools have been
developed and there seem to be grounds for concluding that now, should the conditions of a new financial crisis
occur, they could be spotted in advance. Some of the available indicators rely on the backward observation of
the build-up of imbalances over time (time-varying dimension) and would allow tracking of the probability of
distress several years before its actual occurrence. Other tools focus on the extraction of risk perceptions from
market prices, delivering forward-looking probabilities of distress, which may signal that a disruption is as close
as few months or weeks ahead. More complex applications allow measurement (or simulation) of the non-
linearities embedded in the system (cross-section dimension), helping supervisors to identify those institutions that
are systemically important and deserve special attention.
A look at the set of tools available leads to the conclusion that decisive progress has been achieved in technical
knowledge and in the capability for preventing systemic shocks. Maybe some further effort is needed to collect
the plurality of indicators into a single, and operatively manageable tableau de bord. Each available indicator has its
own properties and limits and their joint use would require them to be organized according to a consistent
syntax [Arsov et al, 2013; Blancher et al. 2013; Lund-Jensen 2012].
However, for macroprudential policies to be effective, the major challenge left appears no longer to be
theoretical or technical. By itself, the efficient monitoring of systemic risk is unable to deliver suggestions either
in the dominion of policies (the issues of when and how macroprudential action must be put in place) or in that of
institutions (the issue of who is in charge of taking decisions)[Gualandri, Noera 2014b].
37Brunnermeier [2008]; Gorton [2010]; Borio, Lowe [2002a]; Cifuentes et al. [2005]; Adrian, Shin [2010]; Brunnermeier,
Pedersen [2008].
19
Firstly, even though the activation of macroprudential policies requires some degree of freedom, it is doubtful
that they could be totally discretionary. The need to act pre-emptively requires the macroprudential supervisor to
take restrictive decisions when the cycle is still in its positive phase (or it is widely perceived to be so): it follows
that macroprudential supervisors are exposed to considerable pressure (from government and market agents) to
dilute, to delay or even to give up intervention. Such interference, in a fully discretionary decision process, could
turn out to amplify the conservative attitude and the risk aversion of decision makers, lengthening the process
and weakening its timeliness. Therefore the requirements of the institutional, organizational and functional
independence of the macroprudential supervisor must be pinpointed in predefined and transparent policy rules,
which trigger non-discretionary action. Secondly, the working of macroprudential regulation and policies may
overlap (due to the selection of instruments) with the areas of competence of other authorities (microprudential
and monetary policies, in particular), giving rise to the institutional issue either of centralizing competences or of
coordinating different agencies [Agur, Sharma 2012].
The review of both policy and institutional aspects goes far beyond the scope of this paper, but it is worth
keeping in mind that the ability to spot systemic risk is just the first link in a very long chain in the
macroprudential supervision process.
20
References
Adrian, T. and Brunnermeier, M. [2009], Co-Var, in “Federal Reserve Bank of New York Staff Reports”, August
2009, n.348.
Adrian, T. and Shin, H. S. [2010], Liquidity and Leverage, in “Journal of Financial Intermediation”, vol.19(3).
Agur, I. and Sharma, S. [2013], Rules, Discretion, and Macro-Prudential Policy, IMF Working Paper, 13/65, March.
Aharony, J. and Swary, I. [1996], Additional Evidence of the Information Based Contagion Effects of Bank Failures, in
“Journal of Banking and Finance”, January pp. 57-69.
Allen, F. and Babus, A. [2008], Networks in Finance, Wharton Financial Institution Center, Working Paper n.08-
07, August.
Arinaminpathy, N., Kapadia, S. and May, R. [2012], Size and Complexity in Model Financial Systems, Bank of
England, Working Paper n.465, October.
Arsov, I., Canetti, E., Kodres, L. and Mitra, S., [2013], Near-Coincident Indicators of Systemic Stress, IMF Working
Paper 13/115, May.
Basel Committee on Banking Supervision (BCBS) [2010], Calibrating Regulatory Minimum Capital Requirements and
Capital Buffers: A Top Down Approach.
Bisis, D., Flood, M., Lo A.W. and Valvanis. S. [2012], A Survey of Systemic Risk Analytics, US Department of the
Treasury, Office of Financial Research, January.
Blancher, N., Mitra, S., Morsy H., Otani, A., Severo, T. and Verderrama L. [2013], Systemic Risk Monitoring
(SysMo) Toolkit. A User Guide, IMF Working Paper 13/168, July.
Borio, C. [2013], Macroprudential policy and the financial cycle: Some stylized facts and policy suggestions, IMF Rethinking
Macro Policy II Conference, First Steps and Early Lessons, April, 16-17.
Borio, C. and Drehmann, M. [2009a], Assessing the Risk of Banking Crises-Revisited, BIS Quarterly Review, March.
Borio, C. and Drehmann, M. [2009b] Towards an Operational Framework for Financial Stability: Fuzzy Measurement and
Its Consequences, BIS Working Papers n. 284, June.
Borio C. and Lowe, P. [2002a], Asset Prices Financial and Monetary Stability: Exploring the Nexus, BIS Working
Papers, n. 114, July.
Borio C. and Lowe, P. [2002b], Assessing the Risk of Banking Crises, BIS Quarterly Review, December.
Brunnermeier, M. [2001], Asset Pricing Under Asymmetric Information: Bubbles, Crashes, Technical Analysis and Herding,
Oxford University Press.
Brunnermeier, M. [2008], Deciphering the 2007-2008 Liquidity and Credit Crunch, NBER Working Paper, December.
Brunnermeier, M. and Pedersen, L. H. [2008], Market Liquidity and Funding Liquidity, in “Review of Financial
Studies”, November.
21
Cannata, F., Castellina, S. and Guidi, G. [2012], Inside the Labyrinth of RWA: How Not to Get Lost, Banca d’Italia,
Occasional Papers n. 132.
Capuano, C. [2008], The Option i-PoD. The probability of Default implied in Option Prices Based on Entropy, IMF
Working Papers n. 194, August.
Chan Lau, J. [2009], Default Risk Codependence in a Global Financial System: Was the Bear Stearns Bail-out Justified?, in
G.Gregoriou (ed.), The Banking Crisis Handbook, CRC Press, Boca Raton USA.
Chan Lau, J. [2010], Balance-Sheet Network Analysis of Too-Connected-to-Fail Risk in Global and Domestic Banking
Systems, IMF Working Papers n. 107, April.
Cifuentes, R., Ferrucci, G. and Shin, H. [2005], Liquidity Risk and Contagion, in “Journal of European Economic
Association”, 3, pp.556-66, April-May.
Committee on the Global Financial System (CGFS), BIS, [2012], Operationalising the Selection and Application of
MacroprudentialInstruments, Report submitted by a Working Group established by the Committee on the Global
Financial System The Group was chaired by José Manuel González-Páramo, then European Central Bank,
December, CGFS Papers No 38.
de Bandt, O., Hartmann, P. and Peydro, J. [2009], Systemic Risk in Banking: An Update, in Berger, Moliyneux,
Wilson (eds.), Oxford Handbook in Banking, Oxford University Press.
De Nicolò, G., Favara, G. and Ratnovsky, L. [2012], Externalities and Macroprudential Policy, IMF Discussion Note
SDN 12/05, June.
European Systemic Risk Board (ESRB) [2013], Recommendation of April 4, 2013, on Intermediate objectives and
instruments of macroprudential policy (ESRB/20/2013/1), (2013/C 170/01), June 20.
Feder, I. and Mitchell, J., [2009], The Future of Securitization. How to Align Incentives, BIS Quarterly Review,
September.
Financial Stability Board, International Monetary Fund, Bank of International Settlements (FRB-IMF-BIS)
[2009], Guidance to Assess the Systemic Importance of Financial Institutions, Markets and Instruments: Initial considerations,
Report to G-20 Finance Ministers and Central Bank Governors, October.
Financial Stability Board, International Monetary Fund (FSB-IMF) [2010], Early Warning Exercise. Design and
Methodological Toolkit, September.
Gai, P. and Kapadia, S. [2010], Contagion in Financial Networks, Bank of England Working Paper n.383, March.
Gai, P., Haldane, A. and Kapadia, S. [2011], Complexity, Concentration and Contagion, in “Journal of Monetary
Economics”, 58, pp.453-70.
Goodhart C. and E. Perotti [2013], Preventive Macroprudential Policy, in “Journal ADEIMF”.
Gorton, G. [2010], Slapped by the Invisible Hand. The Panic of 2007, Oxford University Press.
Gualandri, E. and Noera, M. [2014a], Rischi sistemici e regolamentazione macroprudenziale, in Bisoni C., Gualandri E.,
Landi A., Lusignani G.(editors), Lo stato della finanza, Il Mulino, Bologna.
Gualandri, E. and Noera, M. [2014b], Towards a Macroprudential Policy in EU: Main Issues, CEFIN WP n. 49.
22
Haldane, A. [2009], Rethinking the Financial Network, Bank of England, Speech delivered at the Financial Student
Association, Amsterdam, April.
Haldane, A. [2012], The Dog and the Frisby, Bank of England, paper given at the Federal Reserve Bank of Kansas
City’s 36th economic policy symposium in Jackson Hole, Wyoming, August.
International Monetary Fund (IMF) [2006], Financial Soundness Indicators: Compilation Guide. Washington.
International Monetary Fund (IMF) [2009], Global Financial Stability Report, ch. 2, Assessing the Systemic Implications
of Financial Linkages; ch.3 Detecting Systemic Risk, April.
International Monetary Fund (IMF) [2011a], Macroprudential Policy: an Organizing Framework, Washington.
International Monetary Fund (IMF) [2011b] Global Financial Stability Report, ch. 3, Towards Operationalizing
Macroprudential Policies: When to Act?, Washington, September.
International Monetary Fund (IMF) [2013a], The Interaction of Monetary and Macroprudential Policies, Washington,
January.
International Monetary Fund (IMF) [2013b], Key Aspects of Macroprudential Policy, Washington, June.
Kindleberger, C. and Aliber, R. [2005], Manias, Panics and Crashes, 5th ed., Wiley&Sons.
Koenker, R. and Hallock, [K. 2001] Quantile Regression, in “Journal of Economic Perspectives”, 15, n. 4.
Kyotaki, N. and Moore, J. [1997], Credit Cycles, in “Journal of Political Economy”, 105, n.2, April, pp. 211-248.
Le Lesle, V. and Avramova, S. [2012], Revisiting Risk Weighted Assets, IMF working papers 12/90.
Liikanen Report [2012], High Level Group on Reforming the Structure of the Banking Sector, European Commission,
October.
Lund-Jensen, K. [2012], Monitoring Systemic Risk Based on Dynamic Thresholds, IMF Working Papers 12/159, June.
Merton, R.C. [1974], On the Pricing of Corporate Debt: The Risk Structure of Interest Rates, Journal of Finance, 29(2)
pp.449-70.
Minsky, H. P. [1982], Can It Happen Again? Essays on Instability and Finance, M.E. Sharpe, Armonk, N.Y.; trad.it.
Potrebbe ripetersi?Instabilità e finanza dopo la crisi del’29, Torino, Giulio Einaudi, 1984.
Morris, S. and Shin, H. [2008], Financial Regulation in a System Context, Brooking Papers of Economic Activity,
Fall.
Newman, M. E. J. [2003], The Structure and Function of Complex Networks, SIAM Review, 45, pp. 167-256 , March
Nier, E., Yang, J., Yorulmazer, J. and Alentorn, A. [2008], Network Models and Financial Stability, Bank of England
Working Paper n.346, April.
Noera, M. [2013] Il rischio sistemico, mimeo. [2013].
Onado, M. [2009] I nodi al pettine, Laterza, Bari.
Poon, S., Rockinger, M. and Tawn, J. [2004], Extreme Value Dependence in Financial Markets: Diagnostics, Models and
Financial Implications, in “Review of Financial Studies”, 17, n. 2.pp. 581-610.
23
Revell, J. [1975], Solvency and Regulations of Banks, Bangor Occasional Papers in Economics, 5, Cardiff, University
of Wales Press.
Rocco, M. [2011], Extreme Value Theory for Finance: A Survey, Banca d'Italia, Questioni di Economia e Finanza,
n.99, July.
Segoviano, M. [2006], Consistent Information Multivariate Density Optimizing Methodology, London School of
Economics, Financial Markets Group Discussion Paper n.557, March.
Segoviano, M. and Goodhart, C. [2009], Banking Stability Measures, IMF Working Paper 09/04, January.
Tarashev, N. and Zhou, H. [2006], The Pricing of Portfolio Credit Risk, BIS Working Paper n.214.
Tarashev, N. and Zhou, H. [2008], Market Perceptions of Systemic Risk in the Banking Industry, BIS Quarterly Review
pp.6-8, March.
Zhou, C. [2010] Are Banks Too-Big-to-Fail? Measuring Systemic Importance of Financial Institutions, in “International
Journal of Central Banking”, December.
CEFIN – Centro Studi di Banca e Finanza
Dipartimento di Economia Marco Biagi – Università di Modena e Reggio Emilia
Viale Jacopo Berengario 51, 41121 MODENA (Italy)
tel. 39-059.2056711 (Centralino) fax 39-059 205 6927
CEFIN Working Papers ISSN (online) 2282-8168
49 Towards a macroprudential policy in the EU: Main issues, by Gualandri, E., Noera, M.
(November 2014)
48 Does homeownership partly explain low participation in supplementary pension
schemes?, by Santantonio, M., Torricelli, C., and Urzì Brancati M.C.,
(
September 2014
)
47 An average-based accounting approach to capital asset investments: The case of
project finance, by Magni, C.A.
(
September 2014
)
46 Should football coaches wear a suit? The impact of skill and management
structure on Serie A Clubs’ performance, by Torricelli, C., Urzì Brancati M.C., and
Mirtoleni, L. (July 2014)
45 Family ties: occupational responses to cope with a household income shock, by
Baldini, M.,Torricelli, C., Urzì Brancati M.C. (April 2014)
44 Volatility co-movements: a time scale decomposition analysis, by Cipollini, I., Lo
Cascio I., Muzzioli. S. (November 2013)
43 The effect of revenue and geographic diversification on bank performance, by
Brighi, P., Venturelli, V. (October 2013)
42 The sovereign debt crisis: the impact on the intermediation model of Italian banks,
by Cosma,S., Gualandri, E. (October 2013)
41 The financing of Italian firms and the credit crunch: findings and exit strategies, by
Gualandri, E., Venturelli, V. (October 2013)
40 Efficiency and unbiasedness of corn futures markets: New evidence across the
financial crisis, by Pederzoli, C., Torricelli, C. (October 2013)
39 La regolamentazione dello short selling: effetti sul mercato azionario italiano
(Short selling ban: effects on the Italian stock market), by Mattioli L., Ferretti R.
(August 2013)
38 A liquidity risk index as a regulatory tool for systematically important banks? An
empirical assessment across two financial crises,by Gianfelice G., Marotta G.,
Torricelli C. (July 2013)
37 Per un accesso sostenibile delle Pmi al credito (A sustainable access to credit for
SMEs), by Giuseppe Marotta (May 2013)
36 The unavoidable persistence of forum shopping in the Insolvency Regulation, by
Federico M. Mucciarelli (April 2013)
35 Rating Triggers, Market Risk and the Need for More Regulation, by Federico
CEFIN – Centro Studi di Banca e Finanza
Dipartimento di Economia Marco Biagi – Università di Modena e Reggio Emilia
Viale Jacopo Berengario 51, 41121 MODENA (Italy)
tel. 39-059.2056711 (Centralino) fax 39-059 205 6927
Parmeggiani (December 2012)
34 Collateral Requirements of SMEs: The Evidence from Less–Developed Countries,
by Elmas Yaldiz Hanedar, Eleonora Broccardo, Flavio Bazzana (November 2012)
33 Is it money or brains? The determinants of intra-family decision power, by
Graziella Bertocchi, Marianna Brunetti, Costanza Torricelli (June 2012)
32 Is financial fragility a matter of illiquidity? An appraisal for Italian households, by
Marianna Brunetti, Elena Giarda, Costanza Torricelli (June 2012)
31 Attitudes, personality factors and household debt decisions: A study of consumer
credit, by Stefano Cosma and Francesco Pattarin (February 2012)
30 Corridor implied volatility and the variance risk premium in the Italian market, by
Silvia Muzzioli (November 2011)
29 Internal Corporate Governance and the Financial Crisis: Lessons for Banks,
Regulators and Supervisors, by Elisabetta Gualandri, Aldo Stanziale, and Enzo
Mangone (November 2011)
28 Are defined contribution pension schemes socially sustainable? A conceptual
map from a macroprudential perspective, by Giuseppe Marotta (October 2011)
27 Basel 3, Pillar 2: the role of banks’ internal governance and control function, by
Elisabetta Gualandri (September 2011)
26 Underpricing, wealth loss for pre-existing shareholders and the cost of going
public: the role of private equity backing in Italian IPOs, by Riccardo Ferretti and
Antonio Meles (April 2011)
25 Modelling credit risk for innovative firms: the role of innovation measures, by
Pederzoli C., Thoma G., Torricelli C. (March 2011)
24 Market Reaction to Second-Hand News: Attention Grabbing or Information
Dissemination?, by Cervellati E.M., Ferretti R., Pattitoni P. (January 2011)
23 Towards a volatility index for the Italian stock market, by Muzzioli S. (September
2010)
22 A parsimonious default prediction model for Italian SMEs, by Pederzoli C.,
Torricelli C. (June 2010)
21 Average Internal Rate of Return and investment decisions: a new perspective, by
Magni C.A. (February 2010)
20 The skew pattern of implied volatility in the DAX index options market, by Muzzioli
S. (December 2009)
19 Accounting and economic measures: An integrated theory of capital budgeting, by
Magni C.A. (December 2009)
18 Exclusions of US-holders in cross-border takeover bids and the principle of
equality in tender offers, by Mucciarelli F. (May 2009).
17 Models for household portfolios and life-cycle allocations in the presence of labour
income and longevity risk, by Torricelli C. (March 2009)
16 Differential evolution of combinatorial search for constrained index tracking, by
Paterlini S, Krink T, Mittnik S. (March 2009)
15 Optimization heuristics for determining internal rating grading scales, by Paterlini
S, Lyraa M, Pahaa J, Winker P. (March 2009)
CEFIN – Centro Studi di Banca e Finanza
Dipartimento di Economia Marco Biagi – Università di Modena e Reggio Emilia
Viale Jacopo Berengario 51, 41121 MODENA (Italy)
tel. 39-059.2056711 (Centralino) fax 39-059 205 6927
14 The impact of bank concentration on financial distress: the case of the European
banking system, by Fiordelisi F, Cipollini A. (February 2009)
13 Financial crisis and new dimensions of liquidity risk: rethinking prudential
regulation and supervision, by Landi A, Gualandri E, Venturelli V. (January 2009)
12 Lending interest rate pass-through in the euro area: a data-driven tale, by Marotta
G. (October 2008)
11 Option based forecast of volatility: an empirical study in the Dax index options
market, Muzzioli S. (May 2008)
10 Lending interest rate pass-through in the euro area, by Marotta G. (March 2008)
9 Indebtedness, macroeconomic conditions and banks’ losses: evidence from Italy,
by Torricelli C, Castellani S, Pederzoli C. (January 2008)
8 Is public information really public? The role of newspapers, Ferretti R, Pattarin F.
(January 2008)
7 Differential evolution of multi-objective portfolio optimization, by Paterlini S, Krink
T. (January 2008)
6 Assessing and measuring the equity gap and the equity, by Gualandri E,
Venturelli V. (January 2008)
5 Model risk e tecniche per il controllo dei market parameter, Torricelli C, Bonollo M,
Morandi D, Pederzoli C. (October 2007)
4 The relations between implied and realised volatility, are call options more
informative than put options? Evidence from the Dax index options market, by
Muzzioli S. (October 2007)
3 The maximum LG-likelihood method: an application to extreme quantile
estimation in finance, by Ferrari D., Paterlini S. (June 2007)
2 Default risk: Poisson mixture and the business cycle, by Pederzoli C. (June 2007)
1 Population ageing, household portfolios and financial asset returns: a survey of
the literature, by Brunetti M. (May 2007)
