Measures of systemic risk and

financial fragility in Korea

Jong Han Lee

The Bank of Korea

Dimitrios P. Tsomocos
Sad Business School
University of Oxford

Working paper series | 2013-14

This working paper is authored or co-authored by Sad Business School faculty. The paper is circulated for
discussion purposes only, contents should be considered preliminary and are not to be quoted or reproduced
without the authors permission.

September 2012

Jaemin Ryu
The Bank of Korea

Measures of Systemic Risk and

Financial Fragility in Korea
Jong Han Lee*, Jaemin Ryu** and Dimitrios P. Tsomocos***
This paper provides a quantitative metric for financial stability of Korean
banking system based on the Tsomocos (2003a, b) model, for which we use
market data as proxies for probabilities of default and equity valuation of the
banking sector. We estimate the effect of the probability of default and the
equity valuation of the banking sector on real output in order to measure the
financial fragility using a vector error correction model (VECM). In addition,
we estimate the contributions of individual banks to systemic risk using CoVaR
and MES (Marginal Expected Shortfall).

CoVaR is estimated based on the

methodology of Adrian and Brunnermeier (2010), and MES is estimated based on

Shapley value methodology which has been introduced by Tarashev, Borio and
Tsatsaronis. (2010).
Key Words: Financial Stability, Systemic Risk, JPoD, CoVaR, MES, Shapley value
JEL Specification: E30, E44, G01, G10, G18, G20, G28
* Economist, Marcroprudential Analysis Department, The Bank of Korea, 110, Namdaemunro
3-Ga, Jung-Gu, Seoul, 100-794, Republic of Korea. Email: ljh@bok.or.kr
** Economist, Marcroprudential Analysis Department, The Bank of Korea, 110, Namdaemunro
3-Ga, Jung-Gu, Seoul, 100-794, Republic of Korea. Email: rjm728@bok.or.kr
*** University Reader (Professor), Sid Business School and St. Edmund Hall, University of Oxford,
Park End Street, Oxford, OX1 1HP, United Kingdom, Email: dimitrios.tsomocos@sbs.ox.ac.uk
We are grateful to Gong Pil Choi, Charles Goodhart, Tae Soo Kang, Dae Sik Kim, Hoon Kim, Li
Lin, Juan Francisco Martinez, Byung Hee Seong, Jong Suk Won and the participants of the
Systemic Risk and Financial Stability seminar at the Bank of Korea in January 2012 for their
helpful comments. We especially thank Seung Hwan Lee and Miguel Segoviano for providing us
with excellent technical assistance and computer codes. However, all remaining errors are ours.
The views expressed herein are those of the author and do not necessarily reflect the official views
of the Bank of Korea.
This work is compiled with the financial support of the Bank of Korea. Dimitrios Tsomocos
gratefully acknowledge the support and the hospitality of the Bank of Korea.

Contents
. Introduction ................................................................................................ 1
. Alternative definitions of financial fragility ............................................. 5
. Composite Financial Stability Index ......................................................... 8
1. Introduction ............................................................................................................ 8
2. Model ..................................................................................................................... 8
3. Measuring financial fragility ................................................................................ 14

. CoVaR ........................................................................................................ 24
1. Introduction .......................................................................................................... 24
2. Definition of CoVaR ............................................................................................ 25
3. Estimation procedure and data ............................................................................. 27
4. Results .................................................................................................................. 30

. Marginal Expected Shortfall ................................................................... 35

1. Introduction .......................................................................................................... 35
2. Estimation procedure and data ............................................................................. 36
3. Results .................................................................................................................. 41

. Conclusion ................................................................................................. 43
References ....................................................................................................... 44
Appendices .................................................................................................... 49

. Introduction
As stated in the April 2009 issue of the IMFs Global Financial Stability Report,
For those responsible for safeguarding financial stability, monitoring measures of
systemic stress is now critical. This crisis has highlighted the dangers of focusing
supervisory practices and risk management simply on ensuring that individual
institutions are adequately capitalized and capable of surviving reasonable stress
events. The current crisis has demonstrated that a systemic approach is now urgently
needed. The issue now facing authorities is not whether to attempt to identify
systemic risks, but how best to do so in an interconnected global financial system with
incomplete information.
The problem is clearly how best to do so.

Despite central banks and

international policy making institutions producing a considerable number of studies

and publications on financial stability issues, there is still no obvious framework for
summarising developments in financial stability in a simple or single quantitative
manner.1 This is, to say the least, a considerable disadvantage.

As the same ECB

report put it,2 financial stability assessment as currently practised by central banks
and international organisations probably compares with the way monetary policy
assessment was practised by central banks three or four decades ago before there
was a widely accepted, rigorous framework.
But how can you have a rigorous framework if one cannot even define and
measure financial stability?3

Even five, or six, decades ago there was a comparative

plenitude of data relating to monetary policy, e.g. monetary aggregates, interest rates,
inflation.
1
2
3

By contrast, there is still no generally accepted way of providing

This was stated bluntly in the December 2005 ECB Financial Stability Report.
Ibid. See also Fell and Schinasi (2005).
See Borio and Drehmann (2009) for an analysis of the role of the definition and measurement of
financial fragility in developing an operational framework to secure financial stability and ihk (2007)
for a similar attempt.

-1-

quantitative comparisons of financial stability across countries, or over time in a single

country.4
Economics is a quantitative social science.

Without an ability to make

numerical comparisons, it becomes hard to undertake rigorous analysis. This is one

of the reasons why the growing multitude of central banks reports on financial
stability is mostly descriptive and backward looking in character, whereas reports on
monetary policy are more analytical and forward looking. While central banks have
two main responsibilities, to provide price stability and also systemic financial
stability, achieving the first is technically far easier than the second objective. As
Table 1 shows, when the pursuit of price stability is contrasted with that of systemic
financial stability, the former is generally far easier to undertake.
Table 1: Contrasts between Price and Financial Stability
Price Stability

Financial Stability

a) Measurement and Definition Yes, subject to technical queries

Hardly, except by its absence

b) Instrument for control

Yes, subject to lags

Limited, and difficult to adjust

c) Accountable

Yes

Hardly

d) Forecasting Structure

Central tendency of distribution

Tails of distribution

e) Forecasting Procedure

Standard Forecasts

Simulations or Stress Tests

f)

Simple

Difficult

Administrative Procedure

Even to define financial stability is not an easy task and it seems much easier to
characterise it as the absence of financial instability. Indeed a major bank crisis is all
too obvious to those involved.

So much of the analytical work on financial stability

has, at least until recently, been based on the identification of bank crises, and has then
analysed common antecedent factors, in the attempt to examine what factors may

For recent reviews of attempts to do so, see IMF, 2009, op. cit., Chapters 2 and 3, and Gadanecz and
Jayaram (2009).

-2-

cause instability. A number of examples of this approach are Berg (1999), ihk and
Schaeck (2007), Dermirguc-Kunt and Detragiache (1998), Disyatat (2001), Kaminsky
and Reinhart (1996, 1999), Logan (2000), Valls and Weistroffer (2008) and Vila
(2000). Illing and Liu (2003) provide a critique and a literature review of related
papers, especially those aiming to discover common early warning indicators of
assessed crisis events.
While this approach of studying the event of banking crisis has achieved
considerable success, it has several limitations. The dividing line between a crisis
event, and non-crisis, is bound to be fuzzy; the dating both of its onset and, even more
so, of its ending will be uncertain while the intensity of crises varies.

Any reduced

form relationship between a banking crisis and prior conditions is likely to be subject
to the Lucas critique, in that the identification of prior regularities will change the
behaviour of actors in the system, such as regulators, depositors, etc.

Finally, most of

the time financial systems are not in crisis mode, so focussing only on crisis events is
tantamount to throwing most of the observations away.
One of our main objectives has been to find a metric for measuring financial
stability. Two papers that have followed this same route are Hanschel and Monnin
(2005) and Illing and Liu (2003). In both cases the variables in their financial stress
index are not based on a structural model, and their exercises are limited to single
countries (Switzerland and Canada respectively). ihk (2007) puts together various
indices of financial sector distress and proposes a measure of financial instability that
includes probabilities of failure in individual financial institutions, loss given default,
and correlation of defaults across institutions.5 Nelson and Perli (2005), and San Jose
and Georgiou (2009) record selected indicators of financial stability, but without any
attempt to combine them into a single metric. Perhaps the most similar work to our
own is Christiano, Motto and Rostagno (2009), who embed default into a DSGE
model to obtain a quantitative estimate of the effect of changing risk on real output.
Estimation of the risk overarching the entire system closely relates to the
5

Further references can be found in the two review papers already mentioned, i.e., IMF Global Financial
Stability Report (April 2009) and Gadanecz and Jayaram (2009).

-3-

implementation of macroprudential policy. Macroprudential policy, differing from

microprudential policy, focuses on the stability of the whole financial system rather
than that of individual financial institutions.

Even though the objective of

macroprudential policies is established based from the perspective of the entire

system, the financial regulatory instrument and policy intervention is implemented to
individual financial institutions.

For example, a capital surcharge imposed on a bank

depends on its systemic importance so that the systemic risk caused by too-big-tofail banks can be addressed.

Accordingly, accurate measurement of the

contributions to systemic risk by individual banks is a key component of

macroprudential policy.
Many studies have been conducted on the methodologies of measuring the
contributions of individual banks to systemic risk. Tarashev, Borio, and Tsatsaronis
(2009) proposed an approach to estimate the systemic importance of individual
financial institutions using the Shapley Value methodology.

Acharya, Pedersen,

Philippon and Richardson (2010) emphasize that the financial system has become
vulnerable to macroeconomic shocks due to the lack of systemic risk management in
contrast to the adequate handling of risks embedded in individual financial
institutions. In order to address this problem, they estimate the individual financial
institutions' exposure to the systemic risk through Systemic Expected Shortfall (SES).
In a similar manner, Brownless and Engle (2010) develop an estimation methodology
of Marginal Expected Shortfall (MES).

Adrian and Brunnermeier (2009) develop

CoVaR that measures the Value at Risk (VaR) of the financial system conditional on
an institution being in distress. Using this measure, they estimate the contributions
of individual financial institutions to systemic risk (CoVaR).
There is no perfect methodology that precisely measures the contributions of
individual financial institutions to systemic risk. Relying on a single approach runs a
risk of errors, and therefore, various approaches need to be considered
contemporaneously when implementing macroprudential policy.

This paper

measures the contributions of Korean banks to systemic risk based on two approaches
proposed by Tarasheve et al. (2010) and Adrian et al. (2010), and assesses the

-4-

usefulness of the systemic risk contribution measurement.

The two different

approaches were employed because the use of MES, a top-down measure, and CoVaR,
a bottom-up measure, allows the systemic risk to be measured from different angles
(see, Drehmann and Tarashev (2011)).
In view of the revised Bank of Korea Act that was passed in August 31, 2011 in
the plenary session of the National Assembly (see, Appendix A), we embarked on
calculating measures of systemic risk and financial fragility for the Korean banking
system.

Our objective is to offer quantitative metrics of financial stability and

contributions of individual banks to systemic risk to be used for the conduct of

macroprudential policy. We consider the presence of these measures necessary for
the accountability of policy makers regarding the success of regulatory policy.
The remainder of this paper is organized as follows. In Section 2, we present
alternative definitions of financial fragility.

In Section 3, we develop a general

equilibrium model where default, money and bank intermediation play an essential
role, and calculate a single metric, the Composite Financial Stability Index for the
Korean banking system using the two factors; namely the joint probability of default
(JPoD) and the bank equity index of the banking sector.

Next, we estimate the

contributions of individual banks to systemic risk of Korea using two approaches,

CoVaR and MES (Marginal Expected Shortfall). We report the results in Section 4
and 5, respectively.

Finally, our concluding remarks are presented in Section 6.

Appendices A and B contain the summary of the revised Bank of Korea Act and the
VECM estimation results, respectively.

. Alternative definitions of financial fragility

Numerous other authors have tackled financial stability, and at one level the term
seems to be familiar both from a theoretical as well as a practical viewpoint.
However, few attempts have been made to define and formally characterise it in an
analytically rigorous way.

Academics and policy-makers have offered various

-5-

definitions.6

For example, Crockett (1997) argues that financial stability (refers) to

the stability of the key institutions and markets that go to make up the financial
systemstability requires (i) that the key institutions in the financial system are stable,
in that there is a high degree of confidence that they continue to meet their contractual
obligations without interruption or outside assistance; and (ii) that the key markets are
stable, in that participants can confidently transact in them at prices that reflect
fundamental forces and that do not vary substantially over short periods when there
have been no changes in fundamentals. Mishkin (1994) offers a more informationbased definition.

Financial instability occurs when shocks to the financial system

interfere with information flows so that the financial system can no longer do its job of
channelling funds to those with productive investment opportunities.
Other authors have emphasised deviations from the optimal investment/savings
plan as a necessary ingredient of a definition of financial instability.

For example,

Haldane et al. (2004) propose the following definition; ...financial instability could be
defined as any deviation from the optimal savinginvestment plan of the economy that
is due to imperfections in the financial sector.

However, as Issing (2003) has

pointed out, ...the efficient allocation of savings to investment, though without doubt
a highly desirable feature of an economy, should not be part of a definition of financial
stability.

For example, no-one would say that savings were allocated efficiently to

investment opportunities in the Soviet Union between 1917 and 1991, but the Soviet
Union did not suffer from financial instability, except right at the end of its existence.
Issing (2003) and Foot (2003) have suggested that financial stability is related to
financial market bubbles, or more generally, volatility in financial market proxies.
Indeed, bubbles impair financial markets efficiency, however, in and of themselves,
they do not constitute a defining characteristic of financial fragility, and more
generally financial instability.

One can classify Minskys Financial Instability

Hypothesis in this family of definitions7 since he claims that the inherent financial
6

A survey of these definitions and extensive discussion can be found in Bank for International
Settlements (1998).
See Minsky (1985).

-6-

instability of financial markets is based on the overoptimistic behaviour of economic

agents.

Finally, Schwartz (1986) suggests that a financial crisis is fuelled by fears

that means of payment will be unobtainable at any price and, in a fractional reserve
banking system, leads to a scramble for high powered money..... In a futile attempt to
restore reserves, the banks may call in loans, refuse to roll over existing loans, or
resort to selling assets.

Allen and Wood (2006) offer a related definition.

The definition of financial fragility that we are proposing, for the reasons already
set out, is8
A combination of probability of default (PD) variously measured together
with bank profitability. This has the added advantage that he can be applied,
mutatis mutandis, at both the individual and aggregate levels.
Thus, financial instability is characterised by both high PDs and low profits.
Moreover, the authorities (government and/or the central bank) can influence the level
of debt above which (and the profit below which), a financial environment becomes
fragile.

Also note that this definition can be related to the welfare of the economy

and its distributional consequences.

This in turn conforms with deviations from the

optimal savings and investment plan in equilibrium, as has been suggested by Haldane
et al. (2004). The standard techniques and theorems of equilibrium theory can be
readily applied.

Equilibrium analysis is also amenable to comparative statics, for

example, by varying capital requirement rules one can affect default probabilities and
the welfare effects of a crisis.
This definition is sufficiently flexible to encompass most of the recent episodes
of financial instability.

The Mexican crisis of the early 1990s is a classic example of

such a crisis. The late 1990s east Asian crisis is characterised by a banking crisis and
economic recession as well as extensive default. Finally, the Russian crisis, the
Texas Banking crisis, and the U.S. Stock Market crash of 1987 conform to the
characterisation of a financially unstable regime generated by extensive default and
declines in bank profitability.

See Goodhart et al. (2004) and Tsomocos (2003a,b) for a formal definition.

-7-

In summary, the information-based definition of Mishkin, and of Issing and Foot,

and the institutionally oriented one offered by Crockett and Schwarz encompass
crucial aspects of financial instability; however, they do not capture the main reason
that policy-makers focus on instability, namely its welfare and distributional effects.
In other words, their definitions highlight the inefficiency that a financially unstable
regime generates but they are too general to be applicable for policy analysis.
Widespread default and a pronounced decrease in banks profitability eventually impair
markets so that eventually trade collapses altogether.

. Composite Financial Stability Index

1. Introduction
The main objective of this section has been to define metric for measuring
financial stability for Korean banking system.

Our work here is based on two

associated streams of research. The first involves the development of a general

equilibrium model in which default, money and bank intermediation play an essential
role in section 3.2. The second relates to the estimation of joint probabilities of
default (JPoD) and equity values for the banking sector; this is discussed, along with
other data-based issues in section 3.3.1. In section 3.3.2, we report the results of a
VECM over the period 2003 Q1 until 2011 Q4 in which the main variables of interest
are the two factors defining our metric of financial stability, namely banking sector
joint probabilities of default (JPoD) and bank equity values.

In section 3.3.3, we

show the metric produced by changes of the JPoD and the equity index.

2. Model
The model developed in Goodhart, Sunirand and Tsomocos (2006a) and

-8-

Tsomocos (2003a, b) incorporates heterogeneous banks and capital requirements in a

general equilibrium model with incomplete markets, money and default.
over two periods and all uncertainty is resolved in the second period.
place in both periods in the goods and equity markets.

It extends
Trade takes

In the first period agents also

borrow from, or deposit money with banks, mainly to achieve a preferred time path for
consumption. Banks also trade amongst themselves, to smooth out their individual
portfolio positions. The central bank intervenes in the interbank market to change
the money supply and thereby set the interest rate.

Capital adequacy requirements

(CARs) on banks are set by a regulator and penalties on violations of CARs, and on
the default of any borrower, are in force in both periods.

In order to achieve formal

completeness for the model, banks are liquidated at the end of the second period and
their profits and assets distributed to shareholders. Figure 1 below makes the time
line of the model explicit.
Figure 1: The time structure of the model

-9-

In the first period trades by all agents take place against a background of
uncertainty about the economic conditions (the state of nature) that will prevail in the
second period. Agents are, however, assumed to have rational expectations, and to
know the likelihood (the probability distribution) of good or bad states occurring when
they make their choices in period one.

In period two the actual economic

conjuncture (the state of nature) is revealed and all uncertainty is resolved.

The model incorporates a number of distinct, i.e. heterogeneous, commercial
banks, each characterised by a unique risk/return preference and different initial
capital. Since each bank is, and is perceived as being, different, it follows that there
is not a single market for either bank loans or bank deposits.

In addition, we

introduce limited access to consumer credit markets, with each household assigned (by
history and custom) to borrow from a predetermined bank. This feature allows for
different interest rates across the commercial banking sector. In sum, multiple credit
and deposit markets lead to different loan rates among various banks and to
endogenous credit spreads between loan and deposit rates.
Individual non-bank agents are also assumed to differ in their risk attitudes and
hence in their preferences for default. We model the incentive for avoiding default
by penalising agents and banks proportionately to the size of default. Banks that
violate their capital adequacy constraint are also penalised in proportion to the
shortfall of capital.

Both banks and households are allowed to default on their

financial obligations, but not on commodity deliveries.

Since default is possible in

equilibrium, financial fragility in this model emerges naturally as an equilibrium

phenomenon.

Liquidity and default premia co-determine interest rates, and both

regulatory and monetary policies have non-neutral effects. The model also indicates
how monetary policy may affect financial fragility, thus highlighting the trade-off
between financial stability and economic efficiency.9
As noted above, one of the purposes of the exercise is to design a framework
which could be used in practice for policy purposes, in any country, to assess and to

See also Goodhart et al (2006a).

- 10 -

estimate whether the banking system would be robust in the face of adverse shocks.
So the theoretical modelling has been followed up first with a numerical simulation
(Goodhart, Sunirand and Tsomocos, 2004), and subsequently by calibrating the model
to values for the UK banking system (Goodhart, Sunirand and Tsomocos, 2004 and
2006b).10
The calibration/simulation procedure is set out in Goodhart, Sunirand and
Tsomocos (2005 and 2006b).11

Having chosen the initial values of the relevant

variables, by calibration or exogenously, it is possible to identify an initial equilibrium

value for all the endogenous variables, including interest rates, bank profitability,
monetary aggregates and repayment rates, the latter being directly inversely related to
the probability of default. The next stage is to shock the initial equilibrium, for
example to simulate a recession and then see how the banking system responds. An
example is given in Table 2, when the central bank cuts the stock of base money by
10%.12
Table 2: % change in key variables given a 10% negative shock to M at t=1
Interest rates

10

11

12

rdb

2,8

ib iib

k ib

k iib

ib

0,005 0,003

0,35

0,38

-0,02

-0,04

0,03

0,08

0,5

0,6

0,005

-0,04

-0,02 0,03 0,005 0,005

0,47

0,5

0,003

-0,01

Bank

Bank

2,4 3,3 2,4 -0,12 0,3

Bank

2,4

3,3

0,06

0,1

eib

eiib

iib

GDPi

GDPii

-0,18

-0,17

Independent applications of this exercise have been done for Colombia (Saade, et al, 2006), Brasil
(Cajueiro and Tabak, 2009), Bulgaria (Tsenova, 2010), Belgium (de Walque, Pierrard and Rouabah,
2008), and Norway (Akram, Brdsen and Lindquist, 2007) .
In each period t, excluding the Lagrange multipliers, the optimality conditions imply a system of 56
equations in 143 unknown variables, 87 of which are exogenous variables/parameters in the model.
This implies that there are 87 variables whose values have to be chosen in order to obtain a numerical
solution to the model. Thus, they represent the degrees of freedom in the system and can either be set
appropriately or calibrated against the real data. In particular, the values of these variables have been
chosen such that they capture realistic features of the UK banking sector in 1997. It is important to
note that these variables, which are exogenous when solving the system of equations, do not
necessarily have to be those which are exogenous in the model.
This is taken from Aspachs et al (2006).

- 11 -

Legend:
rb = lending rate offered by bank b B = {, ,
borrowers in the interbank market,
rbd = deposit rate offered by bank b,
= interbank rate,
bs = profits of bank b in state of the world s = {i, ii},
ebs = capital held by bank b in state s,
kbs = ratio of capital to risk-weighted assets of bank b in state s,
bs = repayment rate of bank b to all its creditors in state s,
GDPs = GDP in state s.

As can be seen from the table 2, the interbank rate increases. Naturally, the net
lender bank switches from credit extension to its consumers towards investing into
the interbank market, while increasing its deposit demand. This portfolio adjustment
leads to higher deposit rates and lending rates and widening spreads due to lower
liquidity.

Net borrowers, i.e. banks , , lower interbank borrowing demand by

switching to the deposit market, while reducing loan supply to their customers. Thus,
deposit and lending rates of these two banks increase and credit spreads increase as
well. In sum, lower liquidity and the diversification effect increase deposit rates
whereas it reduces consumer credit extension due to higher funding costs.
The decrease in liquidity lowers future GDP and thus banks rationally anticipate
higher households default and, therefore, reduce credit extension further. This effect
is also reflected by the widening spreads of all banks.

Higher default coupled with

lower loan supply reduces the borrowing banks profits (the lending bank benefits
from the higher return on interbank market investments). However, since banks are
subject to a capital requirement and therefore need to accumulate capital, they respond
by adopting riskier strategies to counteract the negative effect on profits of the credit
contraction.
The point to note here is not so much the details of the table but that the crucial
aspects of the impact of shocks on the banking system are contained in two variables,
bank profitability and bank repayment rate, which in turn is equivalent to its
probability of default.13
13

In normal economic cycles bank repayment rates will be

See also Tsomocos and Zicchino (2005).

- 12 -

positively correlated with profitability. When economic conditions improve, bank

profitability rises and probabilities of default fall. However there can be shifts in
preferences, or other conditions, in which the previous normal relationship can be
reversed: a decrease in banks probabilities of default coincides with a decline in
banks profits. One such example involves an increase in the penalties imposed on
banks should they default. The outcome is shown in Table 3 below.14
Table 3: % change in key variables given a 2% increase
in default penalties imposed on banks on both states of the world
Interest rates

rdb

rb

ib iib

eib

eiib

k ib

k iib

ib

iib

Bank

-9,2

-3,3

-14.8 -27,6 -1,4

-0,9

-0,38

0,2

0,24

0,14

Bank

-18

-10

-3,2 -13,2 -34,4 -2,8

-8,9

-1,8

-8

0,24

0,58

Bank

-18

-10

-16.9 -23.3 -3.5

-3.5

-2,6

-2,6

0,44

0,29

GDPi GDPii

-0,9

-1,5

Legend:
rb = lending rate offered by bank b B = {, , }, where is a net lender and and are net
borrowers in the interbank market,
rbd = deposit rate offered by bank b,
= interbank rate,
bs = profits of bank b in state of the world s = {i, ii},
ebs = capital held by bank b in state s,
kbs = ratio of capital to risk-weighted assets of bank b in state s,
bs = repayment rate of bank b to all its creditors in state s,
GDPs = GDP in state s.

This table reports the consequences of the regulator increasing the penalties on
banks who default on their debt (to depositors and other banks). We assume two
percent increase in both states of the world. Since defaulting is now more costly,
banks increase their repayment rates (the percentage changes of bi and bii are positive
for all banks, as shown in Table 3). Banks' more prudent investment choices induce
a decline in profits (and, therefore, in capital and capital to risk-weighted asset ratios),

14

For more details see Aspachs et al (2006).

- 13 -

aggregate credit and output.15

The point of all this is that the effects of shocks on the stability of the overall
banking system cannot be reasonably represented by a one factor model, either
probabilities of default or profitability, since the relationship between the two factors
may change with the nature of the shocks hitting the system.16 With a two factor
model, quantification requires giving weights to each factor to arrive at a single
number, or metric, and the choice of weights will be conditioned by the results of the
particular empirical model employed.

3. Measuring financial fragility

The composite financial stability index, our metric of financial fragility as a
weighted two factor model, is derived from our prior theoretical modeling.17 Our
hypothesis, based on simulations and calibrations of the general equilibrium model,
developed in Goodhart et al. (2005, 2006a, b) is that whenever banks' default rates
increase and banks' profitability decrease, i.e., when the economy is more financially
fragile, GDP (our proxy of welfare) falls.18 Our aim here is to test whether these two
measures of banking sector's distress do have the predicted impact on output, and
hence provide empirical support for the theoretical model.

15

16

17
18

The effect on credit of an increase in penalties (a regulatory shock) is negative. Because banks and
increase their repayment rates to all creditors considerably, bank is willing to invest more in the
interbank market. As a result, the interbank rate decreases. Since and are able to borrow more
and at a lower cost from the interbank market, their demand for deposits decreases and so do their deposit
rates. The overall level of aggregate credit to households decreases as a result of the negative
households' wealth effect of lower bank equity values.
If the normally expected inverse correlation between profitability and PD had been very high (close to
-1.00) in practice, there could have been a case for concentrating only on PD. However, it is far too
low to allow us to ignore the separate effect of bank profitability.
See Tsomocos (2003a, b).
In the general version of the model, an increase in default and a decrease in profitability are, typically,
associated with a reduction in agents' welfare (see, Goodhart, Sunirand and Tsomocos (2004)).

- 14 -

3.1 Data
We sought variables that would give a good measure of default probabilities and
banking profitability. We use the IMF methodology (Segoviano and Goodhart, 2009)
to estimate a time series for banks joint probabilities of default, which will be referred
to as JPoD. We take the percentage change in equity values of the banking sector as
our index of the markets perception of the change in the present value of returns to
bank.
The data set to calculate the composite financial stability index includes the CDS
spreads, bond spreads, and equity values of Hana Bank (Hana), Industrial Bank of Korea
(IBK), Kookmin Bank (KB), Korea Exchange Bank (KEB), Shinhan Bank (Shinhan),
and Woori Bank (Woori) (6 banks in total) over the period from 2003 Q1 until 2012
Q119. Bond spread means the difference of the yields between the bank debenture (3
years) and Treasury bond (3 years). Data of Hana, KB, Shinhan and Woori beyond a
specific point (Hana: '05.12.12, KB: '08.10.10, Shinhan: '03.1.1, Woori: '03.1.1) are
stock prices of their holding companies. Data frequency is 5 week day format.
We calculate the Joint Probability of Distress or Default (JPoD) defined as a
measure of the probability of all the banks in the system (portfolio) becoming
distressed, which represents the tail risk of the system, using PDs of individual banks.
We use CDS spreads and bond spreads to estimate the PDs of individual banks.
general, CDS spreads are used to calculate the PDs of banks.

In

However, in the case

of Korea, CDS spreads seem to have several shortcomings to represent the PDs of
banks.

CDS spreads of individual banks depend on macroeconomic fundamentals

(e.g., GDP, growth rate, fiscal balance), and foreign sector variables (e.g., current
account, foreign exchange reserves) (see, Seo and Lee (2010)). We argue that this
feature is more conspicuous in Korea when considering that CDS spreads of most
Korean banks have very similar values. As an alternative, EDF (expected default
frequency) provided by Moodys can be considered to measure the PDs of banks.
19

The data for 2012 Q1 are the average of those for 2012 January and February.

- 15 -

However, it is also often pointed out that the estimation process of EDF is not
transparent (Kim, Choi and Hyung, 2011).

In other words, one cannot know

explicitly the model and the parameters used for estimations of EDF (Kim, Moon and
Aydin, 2011). Thus, we develop a new data set to calculate the PDs of banks by
combining both bond spreads and CDS spreads.

We assume that bond spreads

mainly reflect business characteristics of the banks, whereas CDS spreads primarily
reflect credit risk. Even though this assumption is not entirely satisfactory, we argue
it is compatible with the realities of the Korean financial system. Consequently, we
find that the JPoD estimation based on the new data set more reasonably describes
Koreas financial system more accurately. Finally, we rescale bond spreads so that
their mean and standard deviation to be the same as those of the corresponding CDS.
We use the CIMDO approach (Segoviano and Goodhart, 2009) to estimate a time
series for banks JPoDs (joint probabilities of default).

This measure takes into

account the impact of individual banks distress on the rest of the banking sector.
Banks distress dependence is based on the fact that banks are usually linked either
directly through the interbank deposit market and exposures in syndicated loans, or
indirectly through lending to common sectors and proprietary trades. This distress
dependence among banks is also a key feature of the model described before. Banks
distress dependence varies across the economic cycle and tends to rise in times of
distress since the fortunes of banks decline concurrently through either contagion of
idiosyncratic shocks or through negative systemic shocks.

Therefore, in such periods,

the banking systems joint probability of distress, i.e., the probability that all the banks
in the system experience large losses simultaneously, may experience larger - and
highly non-linear - increase than those experienced by the probabilities of distress of
individual banks.20
Figure 2 shows the JPoDs of Korean banking sector. We think that the JPoDs
based on new data justifies better our intuition about the systemic risk of Korean
banking sector rather than other JPoDs based on CDS spreads or bond spreads solely.

20

More details on the estimation of JPoD (joint probabilities of default) are available in the Box 1.

- 16 -

Figure 2: Joint probability of distress (JPoD) of Korean banking sector

0.050

0.020

JPoD (CDS spreads, LHS)

JPoD(New, LHS)

0.040

0.016

JPoD (Bond spreads, RHS)

0.030

0.012

0.020

0.008

0.010

0.004

0.000

0.000
03

04

05

06

07

08

09

10

11

12

We take the percentage change in equity values of the banking sector as our
index of the markets perception of the change in the present value of returns to banks.
As expected, figure 3 shows that the equity growth index largely deteriorated during
the crisis period in 2008.
Figure 3: Equity index growth of Korean banking sector
(QOQ, %)
40
30
20
10
0
-10
-20
-30
-40
03

04

05

06

07

- 17 -

08

09

10

11

12

Considering together figure 2 and 3, we observe that although the JPoD dropped
precipitously since the 4th quarter of 2008, the continuing deterioration of the bank
equity index suggests that financial fragility persists. Hence, we arguably need a
composite financial fragility index.

3.2 The empirical model

As already mentioned in 3.3, our aim is to investigate whether our two indicators
of banking sectors distress, namely, JPoD and equity, have the expected impact on
output. We thus measure the impact on output (GDP) of the two indicators. We
use the VAR (or VECM) methodology, which treats all the variables in the system as
endogenous, to derive the weights of two indicators by variance decomposition of the
VAR (or VECM).
As the variables such as GDP, equity, JPoD or other variables are not stationary and
they are cointegrated, we use the VECM approach. Our baseline model is a threevariable vector, {gdp, eq, jpod}, where gdp is the quarter on quarter growth rate of real
GDP, jpod is the measure of the banking sectors default risk, and eq is the quarter on
quarter growth rate of the bank equity index. Because the period of estimation is
relatively short in Korea, we also estimate the VECM using a monthly data. In this case,
we use the Industrial Production Index (IPI) as a proxy of GDP due to the difficulty of
obtaining monthly GDP data. The ordering of variables is gdp, eq, and jpod, which is
determined by the degree of linkage to external factors. We check the robustness of the
baseline models result by performing different sets of tests. For example, we run
additional regressions adding either only real TB03, which means the real interest rate (3
year Treasury rate expected inflation) or both real TB03 and CPI. In addition, we run
several regressions based on monthly data, such as {ipi, eq, jpod}, {ipi, real TB03, eq,
jpod}, and {ipi, cpi, real TB03, eq, jpod}. Our estimation results using these models are
shown in the figures of Appendix B. They all show that ceteris paribus, a positive shock
to the banks probability of default or a negative shock to the banks equity value has a
negative impact on output. These results are consistent with our predictions.

- 18 -

We use a two year average of the impact in the variance decomposition to estimate
the weights of two indicators. In our baseline model, the effect of JPoD on GDP is
always more important than that of equity. Based on a two year average value of
variance decomposition, the bank equity index explains 7.8% of the variation of GDP,
while the probability of default of the banking sector explains 11.9%. Hence, we assign
the weight of JPoD as 60% [11.9 / (7.8 + 11.9)], and that of equity as 40% [7.8 / (7.8 +
11.9)]. The results of the variance decompositions for the different specifications show
that JPoD explains 4.4% - 16.2%, and the bank equity index explains 4% - 7%.
Accordingly, the relative weights of JPoD and bank equity are respectively 53% - 70%,
and 30% - 47%. The results of our robustness check demonstrate that the predictions of
the baseline model are reasonable. Consequently, we assume the relative weights of
JPoD and equity to be around 60% and 40% respectively in the Korean banking system.

3.3 Results
What we have done so far is to test the hypothesis that our two variables of
banking sector fragility, JPoD and equity, have a significant effect on welfare as
proxied by GDP. Now we want to combine these two factors to obtain a single
quantitative metric, an index for financial fragility. We assume the weights of the
JPoD and equity for a financial stability index are 60% and 40% as shown by our
empirical analysis above. We construct our financial fragility index by combining
the two indicators applying the weights.

Before applying the weights, we rescale

equity so that its mean absolute value and standard deviation are the same as those of
JPoD. The metric is a weighted average of the JPoD and the banking sector equity
index. This metric represents GDP losses due to financial instability produced by
changes of the JPoD and the equity index:
Metrict = 60%*JPoDt 40%*[eqt + av(JPoDt) - av(eqt)]
where av(eqt) = average of the transformed equity series (eqt),
av(JPoDt) = average of the JPoD.

- 19 -

(1)

The financial stability index for Korea is reported in Figure 4. The increase of
the value of the index indicates the intensification of financial instability. It also
shows that financial stability of the Korean banking sector declined during the 2008
crisis, and it improved by the policy response of central banks.

In the second half of

2011, financial stability deteriorated within narrow limits due to the European
sovereign debt crisis.

However, the Longer Term Refinance Operation (LTRO) of

ECB in 2012 Q1 improved financial stability.

Figure 4: Financial Stability Index for Korea
0.025

0.025

0.020

0.02

0.015

0.015

0.010

0.01

0.005

0.005

0.000

-0.005

-0.005
03

04

05

06

07

08

- 20 -

09

10

11

12

Box 1: Joint probability of default (JPoD)

The JPoD is a measure of the tail (extreme) risk of the financial system.

This

statics provides us an assessment of the probability of all the banks in the systemic
defaulting. This framework treats the banking system as a portfolio of banks; it
recovers the joint statistical distribution that characterizes the implied asset values of
the portfolio of banks that represent the system, termed the banking system
multivariate density or BSMD, which characterizes both the individual and joint
asset value movements of the chosen banks.

The BSMD thus captures banks

linear (correlations) and non-linear distress dependence, and their changes

throughout the economic cycle, reflecting the fact that dependence increases in
periods of distress.

21

The entire approach takes three steps: (i) calculating the

probability of distress of individual banks (PoD); (ii) estimating the BSMD based on
the collected PoDs; and (iii) constructing various measures of the banking sector
stability by inspecting various conditional probabilities using the estimated BSMD.
Step 1: Estimation of Probabilities of Distress of Individual Banks
This methodology can be flexibly implemented, since the PoDs of individual
banks, which represent the input variables, can be estimated using alternative
approaches; i.e., (i) market-based PoDs, including PoDs estimated using the
structural approach (SA), CDS spreads and out-of-the-money option prices; and (ii)
historical-distress based PoDs, including PoDs estimated using Z-score type models
and historical frequencies of default. One or another type of indicator is usually

21

In contrast to correlation calculations, which only capture linear dependence, copula functions
characterize the whole dependence structure; i.e., linear and non-linear dependence, embedded in
multivariate densities (Nelsen, 1999). Thus, in order to characterize banks distress dependence we
employ a novel non-parametric copula approach; i.e., the CIMDO-copula (Segoviano, 2009),
described below.

- 21 -

accessible to analysts; thus the data set that is necessary to estimate the input PoDs is
available most of the time; hence, the proposed BSMDs can be constructed for most
countries. Being able to establish such a set of measures with a minimum of basic
components makes it feasible to undertake a wider range of comparative analysis,
both time series and cross-section.
The data that were necessary to estimate historical-based PoDs were not available
in Korea. Therefore, in order to estimate the JPoD employed in this paper, we
analysed the alternative market-based PoDs.

Athanasopoulou, Segoviano and

Tieman (2009) presented a detailed study of these alternative market-based

indicators and concluded that the best available indicator of distress for the banks
under analysis was CDS-PoDs.22

However, we emphasize that PoDs are an input to

our model; thus, any market-based PoDs (risk neutral) or historical-distress based
PoDs (actual probabilities) can be employed, if it is believed to be an adequate
measure.

Put differently, the estimation of the proposed BSMDs is not intrinsically

related to CDS-PoDs.
Step 2: The Banking System Multivariate Density with Time-Varying DistressDependence
Using the individual PoDs as exogenous input variables, by whatever method used
for the individual PoD estimation, the Consistent Information Multivariate Density
Optimization (CIMDO) methodology (Segoviano 2006) allows us to recover the
BSMD.23 The BSMD embeds banks distress dependence structure characterized
by the CIMDO-copula function (Segoviano, 2009) that captures linear and non-

22

23

They analyzed the theoretical properties of the PoDs estimated using the structural approach, CDS
spreads and out-of-the-money option prices, focusing on the theoretical sensitivity of the different
models to different underlying variables. They also analyzed the empirical models performances.
This is done through event-based analysis, which measures how well the models signal distress events
that can be found in the data.
The CIMDO methodology is a non-parametric framework based on the cross-entropy approach.

- 22 -

linear distress dependencies among the banks in the system, and allows for these to
change throughout the economic cycle.24
Step 3: The Joint Probability of Distress: A measure of Tail Risk in the System
The BSMD characterizes the probability of distress of the individual banks
included in the portfolio, and the structure of distress dependence amongst these
banks.

From this, we can obtain the Joint Probability of Distress (JPoD) that is a

measure of the probability of all the banks in the system (portfolio) becoming
distressed, i.e., the tail risk of the system.
The JPoD is defined as

| |

and it is estimated by integrating the density

function (BSMD) as follows:

, where, x, y, and

r denote logarithmic returns of asset in banks X, Y, and R. xid denotes a threshold

point of x above which the bank i is distressed.
Since the JPoD is estimated from the BSMD, it captures linear and non-linear
distress dependencies among the banks in the system, and allows these to change
throughout the economic cycle, reflecting the fact that distress dependence increases
in periods of distress.

So, systemic risks rise faster than individual banks risks.

Therefore, systemic risk may have larger and nonlinear increases than the
probabilities of distress (PoDs) of individual banks. Consequently, measures of
financial stability that are based on averages and/or indexes that assume fixed
correlation parameters through time could be misleading.

24

As compared to traditional methodologies to model parametric copula functions, the CIMDO-copula

avoids the difficulties of explicitly choosing the parametric form of the copula function to be used and
calibrating its parameters, since CIMDO-copula functions are inferred directly (implicitly) from the
joint movements of the individual banks PoDs. The CIMDO-copula appears more robust than
parametric copula functions under the PIT criterion (Diebold, Hahn and Tay, 1999).

- 23 -

. CoVaR
1. Introduction
CoVaR was developed to assess the systemic risk that reflects externalities and
ripple effects of the financial sector, as the problems of microprudential supervision,
especially through VaR (value at risk), had come to light after the 2008 financial
crisis.25

Unlike VaR, which assesses the risk of individual institutions, CoVaR refers

to the value of VaR assessed under the condition that a certain financial institution is
at risk.

Here, Co means conditional, co-movement, contagion, and contribution of

individual banks to systemic risk.

CoVaR attempts to assess the financial systemic

risk when one financial institution realizes low earnings.

Through CoVaR, the

impact of insolvency of a certain financial institution on systemic risk can be assessed

thus enabling the quantification of systemic importance of individual financial
institutions.

The use of CoVaR could also assess the financial institutions

vulnerability to systemic risk or interconnectedness among specific institutions,

although this paper focuses on estimating only the contributions of individual financial
institutions to systemic risk.

This implies that financial supervisory authorities can

use CoVaR as a useful macroprudential supervisory tool.

For example, the

Macroprudential Supervision Group (MPG) of the Basel Committee on Banking

Supervision (BCBS) has discussed ways to use CoVaR to complement the indicatorbased methodology of assessing systemic importance, in selecting G-SIBs. (However,
agreement that more review is needed was reached). By using CoVaR, we assess the
contributions of domestic banks to systemic risk, and review the usability in terms of
macroprudential policy.

25

Co-developed by Tobias Adrian and Markus K. Brunnermeier (2008, 2009, 2010)

- 24 -

2. Definition of CoVaR
CoVaR is the Value-at-Risk (VAR) of financial institutions conditional on other
institutions being under distress.

For a specific financial institution and sector (i),

CoVaR refers to VaR of the given financial institution and sector (j) under the
condition that they are under stress.

In other words, CoVaR is a conditional VaR,

VaR of institution j under the condition that

|

(
where

(2)

= return of financial institution i.

Equation (2) states that the probability that losses of financial institution j is
greater than CoVaR equal to q when the return of a financial institution i falls below a
threshold value.
CoVaR can estimate the contribution of institution i to the risk of institution j
through the difference between CoVaR of j when institution i is under stress
(
(

|
|

) and CoVaR of j when institution i is under average condition

),

(3)

In equation (3), if we assume that j is the whole financial system, the degree of
specific institution i contributing to systemic risk is as follows:

26

The initial work of Adrian and Brunnermeier on CoVaR (2008, 2009) defines
difference between CoVaR and VaR (In other words,

- 25 -

(4)26

as the
|

This paper estimates the impact of individual institution on systemic risk

|

), using equation (4).

Box 2: Value at Risk (VaR)

VaR (Value at Risk) calculates the maximum loss (minimum value) that could
arise at a given confidence level (q), and therefore the possibility that an asset yield
is lower than VaR is equal to q. The extreme loss possibility that return of financial
institution i,

, is smaller than

is q.

Figure 5: Value at Risk

Distribution of changes in value of

portfolio X
Prob

q%

VaR

), but in 2010 elaborate the definition.

- 26 -

Box 3: CoVaR under various conditions

By using the basic definition of CoVaR, CoVaRs under various conditions
can be estimated, and macroprudential policies can be implemented under different
scenarias :
(i) Contributions of individual institutions to systemic risk (contributional CoVaR)
|

(ii) Vulnerabilities of individual institutions from systemic risk (exposure CoVaR)

|

(iii) Impact of financial institution i on financial institution j (network CoVaR)

|

This paper estimates (i) Contributions of individual institutions to systemic risk.

3 Estimation procedure and data

3.1 Data
We calculate the rate of weekly change in the asset value (market-valued total
assets,

) of individual financial institutions and the entire financial system, using

the equity market capitalization and the leverage ratio of ten Korean banks (Jan. 2003
- Dec. 2011, weekly data): Busan, Daegu, Hana, IBK, Jeju, Jeonbuk, KB, KEB,
Shinhan, and Woori.

where


ME = equity market capitalization,
BE = equity book value,
BA= book value of total assets.

- 27 -

(5)

3.2 CoVaR
Quantile regression estimates expected value of financial system yield on
quantile q of the given institution i.
+

where

(6)

= expected return of the overall system on quantile q,

= return of financial institution i.

In equation (6), quantile analysis on quantile q shows that expected value

(

) is VaR of the whole system under the condition of return of institution i.

Therefore, the expected value of systemic return under the condition of

|

means
{

. It is conditional VaRq of entire system in the event of

}.

(7)

The level of contribution of individual institution to systemic risk

(

) is calculated by equation (8) below:

50% ).

- 28 -

50%

50%

(8)

Box 4: Quantile regression

In regression analysis, ordinary least squares (OLS) is useful for understanding
linear information on mean relationship among independent variables and dependent
variables. However, it does not capture nonlinear effects by providing imperfect
information on extreme events such as the financial crisis.

Quantile regression

analysis (Koenker, 2005) is appropriate for understanding loss distribution of

financial institutions under financial stress by examining non-linear effects, which
could vary depending on regime, to dependent variables on a given quantile:
|

[ | | + 1

where

| |]

<0
|

3.3 Time-varying CoVaR

We estimate the trend of changes in joint probability distribution of returns
between individual financial institutions and the entire system, by using a function of
state variables. The CoVaR explained above produces only one value during a
given sample period while the time-varying CoVaR produces results of time-series
during the period.
First, we estimate a regression analysis on conditional quantile of a state variable
(M) 27:
27

Under the assumption that the return of financial institution is the function of state variables M, we
estimate time-varying VaR using quantile regression. We include a set of state variables M that
are well known to capture time variation in conditional moments of asset return, and liquid and easily
tradable (Adrian and Brunnermeir, 2010).

- 29 -

+
|

(9)
|

(10)

= return of financial institution i,

= state variable vector.
Note that
state variable vector) consists of the VIX(KOSPI200 Volatility Index), Short
term liquidity spread (CD (91 days)- Government bonds (3-month)), Changes of government
bonds (3-month), Changes in yield curve(Difference of Government bonds (10-year) Government bonds (3-month)), Changes in credit spread(Difference of corporate bonds (BBB-,
3-year) - Government bonds (3-year)), and Stock market return (KOSPI volatility rate).

Next, we estimate time-varying CoVaR and VaR through the predicted value
produced by the quantile regression analysis:
+

(11)
|

Finally, the degree (

(12)

) of an individual institutions contribution

to the entire systemic risk is estimated:

|

50%

50% ).

(13)

4. Results
4.1 CoVaR
Using CoVaR and time varying CoVaR, we estimate a contribution of
individual bank on the financial stability of entire banking system. Table 4 below
provides the CoVaRs for individual banks of Korea.
correspond to the 5

th

st

percentile and 1

respectively.

- 30 -

It shows CoVaRs which

percentile of the return distribution

The individual banks that contribute the most to systemic risk are Bank 1, Bank 2
and Bank 3 using 5% CoVaR, and Bank 1, Bank 2 and Bank 5 using 1% CoVaR.
Table 4: CoVaRs of Korean banks
Quantile

Bank 1

Bank 2

Bank 3 Bank 4

Bank 5

Bank 6

Bank 7

Bank 8

5%

5.86

4.87

4.77

Ranking

1%

9.82

Ranking

Bank 9 Bank 10

3.97

3.92

3.92

3.73

3.60

2.58

0.91

10

9.51

9.13

7.48

9.30

8.64

6.40

7.49

5.12

2.61

10

As shown in Table 4, the correlation between CoVaR and VaR of individual

financial institutions is very high. Accordingly, the scatter plot in Figure 6 shows a
positive correlation between institutions' risk (VaR) and institutions' contribution to
systemic risk (CoVaR). It suggests that bank regulation relying on VaR may be
valid in the case of Korean banks.

On the other hand, Adrian and Brunnermeier

(2010) analysis showed a very low correlation, and they used this as evidence that the
CoVar can replace the VaR.

The high degree of correlation among financial

institutions implies that the applicability of CoVaR may be limited in Korea.

-2.0

-2.0

-4.0

5%

-3.0

-5.0

-4.0

-8.0

-6.0

-10.0

-6.0
-7.0
-8.0

CoVaR

0.0

1%

0.0
-1.0

CoVaR

Figure 6: CoVaRs vs. VaR in Korean banks

-7.0

-6.0
5% VaR

-5.0

-12.0
-16.0

-4.0

- 31 -

-14.0

-12.0 -10.0
1% VaR

-8.0

-6.0

4.2 Time-varying CoVaR

As a result of estimation of Time-varying CoVaR, we also estimate the
contributions of individual banks on the entire financial system. Table 5 below
describes the time-varying CoVaRs for individual banks in Korea.

It shows time-

varying CoVaRs which correspond to the 5th percentile and 1st percentile of the
return distribution respectively.

The estimated results resembles those of the

CoVaR. That is, the individual banks that contribute to the most to the systemic
risks are Bank 1, Bank 3 and Bank 2 using 5% CoVaR, and Bank 1, Bank 3 and
Bank 5 using 1% CoVaR.

However, the order below the ranking 4 is somewhat

different from that of the estimated results of CoVaR.

Table 5: Time Series Average1) - CoVaRs of Korean banks
Quantile

Bank 1

Bank 2

Bank 3

Bank 4

Bank 5

Bank 6

Bank 7

Bank 8

Bank 9 Bank 10

5%

5.42

4.30

4.88

3.12

4.17

3.44

2.97

3.12

2.00

0.53

Ranking

10

1%

8.71

5.65

6.47

4.10

5.84

5.04

5.15

3.59

3.42

1.19

Ranking

10

Note: 1) Average value of time-varying CoVaR at each period

When comparing the estimated results of the time-varying CoVaR with the
VaR of individual financial institutions, there is a positive correlation similar to the
CoVaR case. However, it is somewhat lower than the correlation between the
CoVaR and the VaR.

- 32 -

Figure 7: Time Series Average - CoVaRs vs. VaR in Korean banks

0.0

-2.0
-3.0

-4.0

5%

CoVaR

-1.0

-5.0
-6.0
-7.0
-8.0

-7.0

-6.0
5% VaR

-5.0

-4.0

Meanwhile, Figure 8 shows that the contributions of all banks to systemic risk
were high during the 2008 financial crisis.
Figure 8: Time-varying 5% CoVaRs of Korean banks
25

25

Bank 1

Bank 2

Bank 3

Bank 4

Bank 5

Bank 6

Bank 7

Bank 8

Bank 9

Bank 10

20

20

15

15

10

10

0
03

04

05

06

07

08

- 33 -

09

10

11

4.3 Remarks
CoVaR purports to estimate the contributions of individual banks to the entire
financial systems risk. However, we could not calculate the aggregate systemic risk
using this approach.

From figure 8, we can assume that entire systemic risk sharply

increased in the late 2008. Yet we cannot directly estimate the level of the systemic
risk from CoVaR approach. Despite such disadvantage, CoVaR could be useful tool
to find the risk of individual bank that takes into account its exposure to common
factor. Thus, CoVaR measures could be used as an ancillary tool.
We estimate the contributions of individual banks to systemic risk through
CoVaR using the quantile regression analysis. It should be noted that the ranking
of banks contribution to the systemic risks change depending on the quantile level.
Furthermore, some major Korean banks without information on stock prices were
excluded in the analysis. Meanwhile, the correlation between the CoVaR and the
VaR is somewhat high in this study, suggesting that the usefulness of the CoVaR may
be limited in the Korean financial system.

CoVaR, however, takes into consideration

of the interconnections of financial industry sectors that VaR cannot detect, and,
hence, is able to identify banks contribution to systemic risks.

In this sense, its

usefulness still seems to be valid in Korea.

For example, it can be used for checking

the systemic importance of domestic banks.

Finally, it can be used for measuring the

systemic risk changes of individual banks before or after a certain period, sectorial
systemic risk, etc.

- 34 -

. Marginal Expected Shortfall

1. Introduction
Macroprudential policy differs from microprudential policy in that it takes greater
interest in the stability of the entire financial system.

However, even if

macroprudential policy goals are determined from the systemic perspective, financial
supervisory tools and policy intervention are implemented on individual financial
institutions. For example, capital surcharge can be imposed on banks depending
upon their systemic importance to mitigate systemic risk caused by too-big-to-fail
banks.

Consequently, it is essential to identify an individual financial institutions

contribution to the entire systemic risk to effectively implement macro-prudential

policy. From this perspective, MES (Marginal Expected Shortfall), the marginal
contribution of an individual bank, can be a useful tool for financial supervision for
macroprudential policy implementation. Although, various MES estimation methods
have been developed, this paper intends to estimate MES of Korean banks based on
the methodology introduced by Tarashev, Borio and Tsatsaronis (2009) 28 with
priority placed upon a fair distribution of contribution to systemic risk. MES refers
to an individual banks loss in the tail of the aggregate sectors loss distribution. In
other words, it identifies the marginal contribution of an individual financial
institution/firm i to the systemic risk through expected losses it will be incurred during
times of significant slowdown in the entire financial system.
MES is similar to CoVaR proposed by Adrian and Brunnermeier (2010) in that
they both estimate the contribution of an individual financial institution to systemic
risk, but differs in following points.
First, CoVaR evaluates an individual banks contribution separately from the

28

Acharya et al. (2010) and Brownless (2010) have also estimated MES.

- 35 -

entire systemic risk measurement (bottom-up approach), but MES measures the entire
systemic risk first and then evaluates an individual banks contribution to it (top-down
approach). Put differently, if an individual financial institution is in a crisis situation,
CoVaR captures the contribution of any other individual financial institution by
estimating risk of the entire system. If the entire system faces crisis, however, MES
captures the contribution of an individual financial institution by estimating the extent
of losses of an individual financial institution. Second, CoVaR is calculated based
on VaR (Value at Risk), while MES is based on ES (Expected shortfall).29 Finally,
the sum of MES is identical to the value of the entire systemic risk. However, the
systemic risk cannot be estimated with the sum of CoVaR.

2. Estimation procedure and data

An individual banks contribution to systemic risk is calculated by estimating the
ES of the entire financial system first and then allocating it to each bank.
ES (Expected Shortfall) is an average loss that can occur if the loss exceeds VaR
(conditional expected loss in the event of losses exceeding VaR):

[ |

(14)

where is a confidence level and R is portfolio profits and losses

VaR is an index of measuring the maximum losses that can occur within a
confidence level, while ES takes into account the case where losses exceed VaR.
Thus, ES is a more conservative index.
In order to estimate MES, we first calculate expected shortfall (ES). Expected
shortfall, also known as expected tail loss, is the measure of systemic risk we use in all
numerical examples. It is defined as the expectation of default-related losses in the
system, conditional on a systemic event.
29

This event occurs when system-wide losses

Please, see Box 5 for more detail on expected shortfall (ES).

- 36 -

equal or exceed some percentile of their probability distribution.

We quantify

expected shortfall using Monte Carlo simulations30 that take as inputs the following
parameters for each institution i: si (the size of the liabilities of institution i), LGDi
(loss-given-default), PDi, i (the loading on the common (or systematic) factor). PDi
is calculated by using CDS spreads.

We use the correlation between profitability

(equity growth rate) of institution i and entire system as a proxy of common factor (i),
which is an individual financial institutions exposure to the systemic risk.
ES is estimated using the formula presented below using indicators representing
the three factors stated previously.
The system-wide losses following a banks failure can be estimated as follows:

(15)

where si = the weight of the size of liabilities of bank i relative to the entire system,
1,
LGDi = system losses following a failure of bank i and it is assumed to be 55% of
liabilities (si) of bank i,
Ii = 1, if bank i fails, and 0, otherwise.

An individual bank is assumed to fail, when its assets value falls below a specific
threshold. In the model, bank i fails, if the indicator representing its assets value (Vi)
is lower than the threshold value that is equal to the individual banks default
probability. That is,

+ 1

(16)

where M = risk factor affecting all financial institutions,

Zi = factor affecting Financial Institution i,
M, Zi follow the normal distribution
PDi = probability of default of Financial Institution i,
-1= reverse function of standard normal cumulative distribution function,
i= the loadings on the common (or systematic) factor.

30

Monte Carlo simulations are usually used in cases where there is a lack of previous direct time series of
variables to be measured, credibility is low due to insufficient information on previous time series or ample
noise, and it is impossible to measure time series of variables directly.

- 37 -

ES of financial system is calculated with the use of the previous formula and
Monte Carlo simulations.

Random numbers are created to satisfy statistical

characteristics of risk factors, M and Zi (normal distribution assumed).

Individual banks contribution to systemic risk is estimated by allocating the
entire systemic risk to each bank by calculating its Shapley value.

It allocates the

total impact to members depending on their respective degree of contribution (Shapley,

1953). The total risk is allocated to individual financial institutions in the same manner as
the total impact is shared according to the Shapley value calculation.31 We calculate the
), by evaluating the level of risk of each subsystem.32

Shapley value, (

i
||

{} )

(17)

where = entire financial system,

= all subgroups of the entire financial system () including bank i,
|| = the number of banks within subgroups,
C(ns) = the number of subgroups (including bank i) when the number of banks is ns
(=

),

= systemic risk of the subgroup,

(risk of subgroup which does not include any banks) = 033.

We estimate MES using the data of six Korean banks (2003 Q1 - 2011 Q4,
quarterly data): Hana, IBK, KB, KEB, Shinhan, and Woori. The data set to estimate
MES includes the CDS spreads, equity values and liabilities of six banks.

31
32

33

Please, see Box 6 for a detailed explanation of Shapley value and how to use the formula (17).
Allocation part using Shapley value methodology in the program for MES estimation is developed by
Lee Seung Hwan, Marcroprudential Analysis Department of Bank of Korea. We verified the
accuracy of the program. We would like to thank him for his help.
If any financial institution does not exist in financial system, the systemic risk is 0.

- 38 -

Box 5: Expected shortfall (ES)

Let X be the random variable showing portfolio profits and losses and the VaR
with quantile % be VaR(X). Expected Shortfall, ES(X), is defined as follows:

[ |

i.e., average for all VaR of the area between 0 and .

Figure 9: Comparison between VaR and ES
Prob

Distribution of changes in value of

porfolio X

ES
%

VaR

return

Source: Comparison of ES and VaR (Artzner, 1999)

In the case of VaR, if a loss from a negative shock is greater than a critical value,
it cannot be captured.

This implies that very risky trades can be hidden.

However, in the case of ES, all losses greater than a certain value are measured.
Such characteristics are important in that a lot of massive losses requiring
government bailouts are those exceeding a VaR critical value.

The sum of every

VaR can be greater than the entire portfolio VaR, but this problem does not occur in
the case of ES. This implies that ES is a more coherent risk measure index.

- 39 -

Box 6: Example: Estimation of Marginal Contribution to Systemic Risk

Under the assumption that Bank A, B and C are included in the system, marginal
contributions to systemic risk of individual banks can be calculated.
Table 6: Calculation of Shapley Value
Marginal contribution of individual bank
Bank A
Bank B
Bank C
4
4
4
5
5
6
6
7
7
4
5
6
4.5
5
5.5

Systemic Risk
(ES)

Subgroup
A
B
C
A, B
A, C
B, C
A, B, C
Shapley Value

4
4
4
9
10
11
15

After calculating all systemic risks (ES) of subgroups of all different scales,
marginal contributions of individual banks within each subgroup are calculated.
For instance, Bank As marginal contribution 5 to systemic risk (ES) of the subgroup
(A, B) is calculated by subtracting 4, Bs systemic risk, from 9, the subgroup.
Likewise, As marginal contribution 4 to the entire system (A, B, C) is calculated by
subtracting systemic risk 11 of the subgroup (B, C). The Shapley value of each
bank is calculated by averaging each banks marginal contribution to systemic risk of
subgroups of all different scales.
The Shapley value can be also calculated with use of a general formula as shown
below:

i
||

- 40 -

{ } ).

So, bank As Shapley value is:

{

1 1
[ {}
3 1

1 1
[ 4
3 1

1
1
0 + ( { }
{} ) + ( {
2
2
1
+ { }
{ } ]
1
1
1
1
0 + 9 4 + 10 4 + 15 11 ]
2
2
1

{ } )

4.5

3. Results
According to our results, individual banks contribution to systemic risk varies
somewhat depending on the time period. In general, however, Bank A shows the
largest contribution followed by Bank B and Bank C, respectively. The ES of the
entire system (sum of all MES) has risen since 2007 to peak in the second half of 2008,
and has been higher during the post-crisis period than during the pre-crisis period.
Figure 10: Estimation of Banks MES
MES of 5% Quantile Level
0.6

MES of 2% Quantile Level

0.6

Bank F
Bank E
Bank D
Bank C
Bank B
Bank A

0.5

0.6

0.6

Bank F
Bank E
Bank D
Bank C
Bank B
Bank A

0.5

0.5

0.4

0.4

0.3

0.3

0.2

0.2

0.2

0.2

0.1

0.1

0.1

0.1

0.0

0.0

0.4
0.3

0.0
03

04

05

06

07

08

09

10

11

0.4
0.3

0.0
03

- 41 -

0.5

04

05

06

07

08

09

10

11

Table 7: Systemic Importance of Korean banks

Bank A

Bank B

Bank C

Bank D

Bank E

Bank F

MES (5%)

0.044

0.038

0.035

0.022

0.020

0.014

MES (2%)

0.076

0.056

0.056

0.032

0.030

0.019

Note: Average MES figure for each time period. MES figures rise as falls (or confidence level rises)

While the absolute level of marginal contribution to systemic risk varies

depending upon the quantile level (or confidence level), the relative level and ranking
of each banks marginal contribution to systemic risk are generally stable during the
sample period.
Regardless of the quantile level (5% or 2%), Bank A shows the largest degree of
marginal contribution to systemic risk (systemic importance), followed by Bank B and
C. Evaluation of marginal contribution through the Shapley value methodology is
effective for fair judgment of each banks marginal contribution under the assumption
that the measurement of systemic risks is accurate. These characteristics can be
utilized when assessing the systemic importance of each bank in the course of
implementing macroprudential policy.

For example, in future discussions on D-SIBs

regulation system, the Shapley value may be useful to calculate the systemic
importance of individual banks.
accuracy of the index.

Efforts should be maintained to enhance the

This can be achieved by performing sensitivity analysis of

basic input variables.

- 42 -

. Conclusion
We began this paper by stating that there was no obvious framework for
measuring financial fragility, though much work is currently being undertaken in this
field. It has been our purpose here to demonstrate that such a framework can be
obtained. This is clearly a first shot at what has been a difficult problem. We hope
and expect others to refine and to improve our methodology, but we contend that it
can be done.

Our results here suggest that the two variables that we identified in the

composite financial stability index are highly significant in determining GDP, and,
indeed, the most important factors over longer horizons when they are considered
contemporaneously.
A metric for financial stability may contribute towards efficient crisis prevention
and management. In addition, policy makers will inevitably be faced with clearly
defined objectives and, therefore, be accountable for breaching them. We hasten to
add that the component of the composite financial stability index of financial fragility
is JPoD, and it is possible to predict its fluctuations (see, Goodhart, Hofmann and
Segoviano (2006)). The construction of metrics for financial stability may enable us
to adopt the appropriate regulatory policy and implement effective regulatory
measures. Ultimately, the aim is to build an evidence-based and analytically rigorous
counter-cyclical regulatory structure for prudential regulation to replace the present
pro-cyclical one.

- 43 -

References
Acharya, V., Pedersen, L., Philippon, T. and Richardson, M., (2010), Measuring
Systemic Risk, Working Paper.
Adrian, T. and Brunnermeier, M., (2008), CoVaR, Federal Reserve Bank of New
York Staff Report No.348.
, (2009), CoVaR, Working Paper, Princeton University.
, (2010), CoVaR, Federal Reserve Bank of New York Staff Report
No.348, revised.
Akram, G. Brdsen and K-G. Lindquist, (2007), Pursuing financial stability under an
inflation-targeting regime, Annals of Finance, 2007, vol. 3, issue 1, 131-153.
Allen, B.,Wood,G.E., (2006), Defining and achieving financial stability, J Financ
Stability 2. 152-172.
Artzner, P., Delbaen, F., Eber, J. and Heath, D., (1999), Coherent Measures of Risk,
Mathematical Finance Vol.9. 203-228.
Aspachs-Bracons, O., Goodhart, C.A.E., Segoviano, M., Tsomocos, D. and L.
Zicchino, (2006), Searching for a Metric for Financial Stability, LSE FMG
Special Paper167.
Aspachs-Bracons, O., Goodhart, C.A.E., Tsomocos, D. and L. Zicchino, (2007),
Towards a measure of financial fragility, Annals of Finance, 2007, vol. 3, issue
1. 37-34.
BCBS, (2011), A finalised proposal on the G-SIB identification methodology and
request for national input on the measurement of global systemic importance,
BS/11/24.
Berg, A., (1999), The Asia crises: Causes, policy responses and outcomes, IMF
Working Paper 99/138.
Borio, C. and M. Drehmann, (2009), Towards an operational framework for financial
stability: fuzzy measurement and its consequences, BIS working paper 284.

- 44 -

Brownless C. and Engle. R., (2010), Volatility, Correlation and Tails for Systemic
Risk Measurement, Working Paper.
Cajueiro, D. O. and

B. Tabak, (2009), Financial fragility in a general equilibrium

model, the Brasilian Case, Bank of Brazil, working paper.

Chaiwatana, T., Espinoza, R., Malik, S., and M. Segoviano, (2009), Probabilities of
Default and the Market Price of Risk in a Distressed Economy, IMF Working Paper.
Choi, P. and Min, I., (2010), Estimating the CoVaR for Korean Banking Industry,
KDI Journal of Economic Policy Vol.32(3) 71-99. (Korean).
Christiano, L., Motto, R. and M. Rostagno, (2009), Financial Factors in Economic
Fluctuations, preliminary draft, May 31, paper presented at DG Ecfin 6th
Annual Research Conference, Brussels, October 16, ECB Working Paper.
Crockett, A., (1997), Maintaining Financial Stability in a Global Economy, Federal
Reserve Bank of Kansas Citys Symposium, Jackson Hole, Wyoming.
ihk, M (2007), Systemic Loss: A Measure of Financial Stability, Czech Journal
of Economics and Finance, 57, 5-26.
ihk, M. and K. Schaeck, (2007), How Well Do Aggregate Bank Ratios Identify
Banking Problems?, IMF working paper No. 07/275.
Dermirguc-Kunt, A. and E. Detragiache, (1998), The determinants of banking crises
in developing and developed countries, IMF Staff papers 45(1), 81-109.
de Walque, G, Pierrard, O and A Rouabah, (2008),

Financial (in)stability,

supervision and liquidity injections, a dynamic general equilibrium approach,

National Bank of Belgium Working Paper No 148.
Diebold, F., Hahn, J., and A. Tay, (1999), Multivariate Density Forecast Evaluation
and Calibration in Financial Risk Management: High-Frequency Returns on
Foreign Exchange, The Review of Economics and Statistics, 81(4), 661-673.
Disyatat, P., (2001), Currency crises and the real economy: The role of banks, IMF
working paper No. 01/49.
European Central Bank, (2005), Financial Stability Review, (December).
Fell, J. and G. Schinasi, (2005), Assessing financial stability: exploring the boundaries
of analysis, National Institute Economic Review, No. 192, April, 101-117.

- 45 -

Gadanecz, B., and Jayaram, K., (2009), Measures of Financial Stability A Review,
Irving Fisher Committees Bulletin on Central Bank Statistics, vol. 31, 365-380.
Goodhart, C.A.E., Sunirand, P. and D.P. Tsomocos, (2004), A model to analyse
financial fragility: applications, Journal of Financial Stability, Vol. 1(1) pp. 1~30.
, (2005), A risk assessment model for banks, Annals of Finance, Vol.
1, 197-224.
, (2006a), A model to analyse financial fragility, Economic Theory,
Vol. 27, 107-142.
, (2006b), A time series analysis of financial fragility in the UK
banking system, Annals of Finance, Vol. 2(1), 1-21.
Haldane, A., Hall, S., Saporta, V., Tanaka,M., (2004), Financial stability and
macroeconomic models, Bank of England Financial Stability Review, 80-88.
Hanschel, E. and P. Monnin, (2005), Measuring and forecasting stress in the banking
sector. Evidence from Switzerland, BIS working paper No. 22, 431-449.
Illing, M. and Y. Liu, (2003), An index of financial stress for Canada, Bank of
Canada working paper No. 14.
Issing, O., (2003), Monetary and financial stability: is there a trade-off? Conference
on Monetary Stability and the Business Cycle, BIS, Basel, 2829 March 2003.
IMF, (2009), Assessing the systemic implications of financial linkages, Global
Financial Stability Report.
Jorion P., (1997), Value at Risk: the New Benchmark for Controlling Market Risk,
McGraw-Hill, 3rd edn.
Kaminsky, G. and C. Reinhart, (1996), The twin crises: The cause of banking and
balance-of-payment problems, Board of Governors of the Federal Reserve
System International Finance Discussion Paper no. 544.
Kim, J. and Kim, B., (2010), Estimating Capital Surcharge for Systemically
Important Financial Institutions; Case of Korea, KIF Financial Research Report
2010-07 (Korean).
Kim, J. and Kim, Y., (2010), Analysis on contagion across financial sectors using CoVaR,
Journal of Social Science Research Vol. 23, 75-103 (Ewha University, Korean).

- 46 -

Kim, M., Choi, K. and Hyung, N. (2011), Spillover Effect of Volatility of Credit Risk
of Financial Institutions and Assessment of Systemic Risk, Collection of Papers
on Financial Stability, Bank of Korea (Korean).
Koenker R., (2005), Quantile Regression, Cambridge University Press.
Logan, A., (2000), The early 1990s small banks crisis: Leading indicators, Bank of
England Financial Stability Review 9, 130-145.
Minsky, H.P., (1992), "The financial instability hypothesis", Working Paper 74,
Jerome Levy Economics Institute, Annandale on Hudson, NY.
Mishkin, F.S., (1994), Global Financial Instability: framework, events, issues,
Journal of Economic Perspectives 13, 3-25.
Nelsen, R., (1999), An introduction to copulas, Lecture Notes in Statistics 139,
Springer.
Nelson, W.R. and R. Perli, (2005), Selected Indicators of Financial Stability, Irving
Fisher Committees Bulletin on Central Bank Statistics, vol. 23, 92-105.
Saade, A., Osorio, D. and D. Estrada, (2006), An equilibrium approach to financial
stability: the Colombian case, Annals of Finance, 3(1), 75-106.
San Jose, A., and A. Georgiou, (2009), Financial Soundness Indicators (FSIs):
Framework and Implications, Irving Fisher Committees Bulletin on Central
Bank Statistics, vol. 31, 277-282.
Schwartz, A., (1986), Real and pseudo financial crises : Financial Crises and the
World Banking System, Capie, F.H., Wood, G.E., (eds.) Macmillan, London.
Segoviano, M., (2006a), The Conditional Probability of Default Methodology,
Financial Markets Group, London School of Economics Discussion Paper 558.
, (2006b), The Consistent Information Multivariate Density Optimizing
Methodology, Financial Markets Group, London School of Economics
Discussion Paper 557.
, (2009), The CIMDO-Copula. Robust Estimation of Distress
Dependence under Data Restrictions, IMF Working Paper, forthcoming.
Segoviano, M., and W. Cui, (2009), A Methodology for Reconstructing CDS Spreads
Back in Time, IMF Working Paper, forthcoming.

- 47 -

Segoviano, M. and C.A.E. Goodhart, (2009), Banking Stability Measures, IMF

Working Paper, No. 09/4.
Seo, B. and Lee, Y. (2010), Explaining Credit Default Swap Premium of Large
Banks through the Global Financial Crisis, KIF Financial Research Report
2010-01 (Korean).
Tarashev, N., Borio, C. and Tsatsaronis, K., (2009), The Systemic Importance of
Financial Institutions, BIS Quarterly Review.
, (2010), Attributing Systemic Risk to Individual Institutions, BIS
Working Papers No.308.
Tsomocos, D.P., (2003a), Equilibrium analysis, banking and financial instability,
Journal of Mathematical Economics, 39 (5-6), 619-655.
, (2003b), Equilibrium analysis, banking, contagion and financial
fragility, Bank of England Working Paper, No. 175.
Tsomocos, D.P. and L. Zicchino, (2005), On modelling endogenous default,
Financial Markets Group Discussion Paper, No. 548.
Tsenova, T (2010), General equilibrium model with heterogeneous banking sector for
risk scenario assessments, Bulgarian National Bank, mimeo.
Valls, V. and C. Weistroffer, (2008), Monitoring Banking Sector Risks: An applied
approach, Deutsche Bank working paper series, 29.
Vila, A., (2000), Asset price crises and banking crises: Some empirical evidence,
BIS conference papers No. 8, (March) 232-252.

- 48 -

Appendix A: Summary of the revised Bank of Korea Act

The Bank of Korea (BOK) Act has been revised, with the passage on August 31,
2011, in the Plenary Session of the National Assembly.

As the importance of

macroprudential policies has grown globally based on lessons drawn from the 2008
global financial crisis this revision of the Bank of Korea Act is directed toward
strengthening the financial stability role of the central bank in line with this trend.

It

will consequently contribute to elevating the international profiles of both Korea and
its central bank, by enhancing Korean financial system stability while at the same
time heightening Korea's external credibility.
Major Revisions to the Bank of Korea Act are as below.
Bank of Korea (BOK) Act was amended to impose new mandates for Financial
Stability: the Bank of Korea shall conduct monetary and credit policies with
due consideration on Financial Stability. (Article 1)
Greater Accountability: A report on the status of the nation's macro- financial
stability to be compiled at least twice a year and submitted to the National
Assembly
Enhanced Access to B/S information of both Banks and Non-Banks: More
Institutions Required to Submit Materials Requested by the BOK.

Joint BOK-

Financial Supervisory Service (FSS) Examinations initiated with a BOKs

request when needed.
More

Responsibility

to

Respond

Systemic

Risks:

Improved

Reserve

Requirement System (e.g. Enhanced Liabilities Subject to Reserve Requirements)

With this revision of the Bank of Korea Act, the Bank of Korea will faithfully
carry out the new responsibilities assigned to it, notably its financial stability role as

- 49 -

the central bank.

While striving above all else for price stability, the Bank of Korea,

in accordance with the revision of the BOK Act and its intent, will pay close attention
to financial stability in carrying out its monetary and credit policies, and will
effectively discharge its newly assumed functions under the revised Act, including the
regular publication of a Macrofinancial Stability Report.

- 50 -

Appendix B: VECM estimation results

Figure 11: Impulse response of a VECM with GDP, Equity, JPoD (baseline model)
Response to Cholesky One S.D. Innovations
Response of GDP to GDP

Response of GDP to EQUITY

Response of GDP to JPOD

1.0

1.0

1.0

0.5

0.5

0.5

0.0

0.0

0.0

-0.5

-0.5

-0.5

10

Response of EQUITY to GDP

10

12

12

-4

-4

-4

-8

-8
2

Response of JPOD to GDP

10

.0010

.0005

.0005

.0005

.0000

.0000

.0000

-.0005

-.0005

-.0005

-.0010

-.0010

-.0010

-.0015

-.0015
3

10

10

10

10

Response of JPOD to JPOD

.0010

Response of JPOD to EQUITY

.0010

-8
1

10

Response of EQUITY to JPOD

12

Response of EQUITY to EQUITY

-.0015
1

10

Table 8: Variance Decomposition of GDP

Period

GDP

EQUITY

JPOD

100.0000

0.0000

0.0000

82.7691

8.4028

8.8282

81.9766

8.3010

9.7225

80.5439

8.1653

11.2908

80.3684

8.0740

11.5576

80.0019

7.7111

12.2870

79.6720

7.4023

12.9257

79.0050

7.1941

13.8008

- 51 -

Figure 12: Impulse response of a VECM with GDP, Real TB03, Equity, JPoD
Response to Cholesky One S.D. Innov ations
Response of GDP to GDP

Response of GDP to REAL_T B03

Response of GDP to EQUIT Y

Response of GDP to JPOD

1.5

1.5

1.5

1.5

1.0

1.0

1.0

1.0

0.5

0.5

0.5

0.5

0.0

0.0

0.0

0.0

-0.5

-0.5

-0.5

-0.5

10

Response of REAL_T B03 to GDP

10

Response of REAL_T B03 to REAL_T B03

10

Response of REAL_T B03 to EQUIT Y

.6

.6

.6

.6

.4

.4

.4

.4

.2

.2

.2

.2

.0

.0

.0

.0

-.2

-.2

-.2

-.2

10

Response of EQUIT Y to GDP

10

Response of EQUIT Y to REAL_T B03

10

Response of EQUIT Y to EQUIT Y

12

12

12

-4

-4

-4

-4

-8
1

10

-8
1

Response of JPOD to GDP

10

Response of JPOD to REAL_T B03

10

.0010

.0010

.0005

.0005

.0005

.0000

.0000

.0000

.0000

-.0005

-.0005

-.0005

-.0005

-.0010

-.0010

-.0010

-.0010

-.0015

-.0015

-.0015

10

10

10

10

10

10

-.0015
1

10

Table 9: Variance Decomposition of GDP

Period

GDP

REAL TB03

EQUITY

JPOD

100.0000

0.0000

0.0000

0.0000

93.1850

0.2321

0.1700

6.4129

86.6390

4.7797

2.3761

6.2052

80.8822

4.4534

6.9196

7.7448

79.0745

4.4225

8.7788

7.7242

78.9196

4.0404

9.0280

8.0120

78.8528

3.8755

9.0283

8.2434

77.9211

3.8138

9.6167

8.6484

- 52 -

Response of JPOD to JPOD

.0010

Response of JPOD to EQUIT Y

.0005

-8
1

.0010

Response of EQUIT Y to JPOD

12

-8

Response of REAL_T B03 to JPOD

Figure 13: Impulse response of a VECM with GDP, CPI, Real TB03, Equity, JPoD,
Response to Cholesky One S.D. Innovations
Response of G DP to G DP

Response of G DP to CPI

Response of G DP to REAL_TB03

Response of G DP to EQ UIT Y

Response of G DP to JPO D

1 .5

1 .5

1 .5

1 .5

1 .5

1 .0

1 .0

1 .0

1 .0

1 .0

0 .5

0 .5

0 .5

0 .5

0 .5

0 .0

0 .0

0 .0

0 .0

-0 .5

-0 .5
2

10

-0 .5
2

Response of CPI to G DP

0 .0

-0 .5

10

Response of CPI to CPI

-0 .5

10

Response of CPI to REAL_TB03

10

Response of CPI to EQ UIT Y

.8

.8

.8

.8

.6

.6

.6

.6

.6

.4

.4

.4

.4

.4

.2

.2

.2

.2

.0

.0

.0

.0

.0

-.2

-.2

-.2

-.2

-.2

10

Response of REAL_TB03 to G DP

10

Response of REAL_TB03 to CPI

10

Response of REAL_TB03 to EQ UITY

.6

.6

.6

.6

.4

.4

.4

.4

.4

.2

.2

.2

.2

.2

.0

.0

.0

.0

-.2
2

10

-.2
2

Response of EQ UIT Y to G DP

Response of EQ UIT Y to CPI

10

Response of EQ UITY to EQ UITY

10

10

10

10

-5

-5

-5

-5

-5

-1 0
2

10

-1 0
2

Response of JPO D to G DP

-1 0

10

Response of JPO D to CPI

10

Response of JPO D to REAL_TB03

10

Response of JPO D to EQ UIT Y

.00 10

.00 10

.00 10

.00 05

.00 05

.00 05

.00 00

.00 00

.00 00

.00 00

.00 00

-.00 05

-.00 05

-.00 05

-.00 05

-.00 05

-.00 10

-.00 10

-.00 15
2

10

10

10

10

10

-.00 10
-.00 15
2

10

Table 10: Variance Decomposition of GDP

Period

GDP

CPI

REAL TB03

EQUITY

JPOD

100.0000

0.0000

0.0000

0.0000

0.0000

95.1457

0.0077

0.1023

0.3474

4.3969

88.8940

1.2381

4.0836

1.6384

4.1459

83.5214

2.6974

3.8596

4.7101

5.2115

81.4337

3.2268

4.0846

6.0859

5.1690

81.2176

3.3469

3.7408

6.3137

5.3810

81.3152

3.3348

3.6599

6.2001

5.4901

80.6196

3.5658

3.6319

6.4680

5.7148

- 53 -

.00 05

-.00 15
2

.00 10

-.00 10

-.00 15
2

10

Response of JPO D to JPO D

.00 05

-.00 15

-1 0
2

.00 10

-.00 10

Response of EQ UIT Y to JPO D

10

-1 0

-.2

10

Response of EQ UITY to REAL_TB03

10

.0

-.2

10

Response of REAL_TB03 to JPO D

.6

-.2

.2

10

Response of REAL_TB03 to REAL_TB03

Response of CPI to JPO D

.8

10

Figure 14: Impulse response of a VECM with IP, Equity, JPoD (Monthly data basis)
Response to Cholesky One S.D. Innovations
Response of IPI to IPI

Response of IPI to EQUITY

Response of IPI to JPOD

1.5

1.5

1.5

1.0

1.0

1.0

0.5

0.5

0.5

0.0

0.0

0.0

-0.5

-0.5

-0.5

-1.0

-1.0
5

10

15

20

25

-1.0
5

30

Response of EQUITY to IPI

10

15

20

25

30

Response of EQUITY to EQUITY

10

10

10

-2
10

15

20

25

Response of JPOD to IPI

10

15

20

25

30

Response of JPOD to EQUITY

.0012

.0012

.0008

.0008

.0008

.0004

.0004

.0004

.0000

.0000

.0000

-.0004

-.0004

-.0004

-.0008
5

10

15

20

25

30

10

30

15

20

25

30

25

30

-.0008
5

10

15

20

25

30

10

Table 11: Variance Decomposition of IPI

Period
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

25

Response of JPOD to JPOD

.0012

-.0008

20

-2
5

30

15

-2
5

10

Response of EQUITY to JPOD

IPI
100.0000
88.2632
75.9789
75.7860
76.1319
76.9122
76.5245
75.7580
75.6234
75.5293
75.4641
75.3662
75.1480
74.9940
74.8377
74.6967
74.5738
74.4197
74.2704
74.1247
73.9787
73.8410
73.6977
73.5549

EQUITY
0.0000
7.0884
6.7954
7.0368
6.9366
6.7074
7.1585
7.2646
7.2472
7.2896
7.3121
7.3808
7.4509
7.4933
7.5510
7.5943
7.6420
7.7005
7.7512
7.8032
7.8525
7.9016
7.9537
8.0037

- 54 -

JPOD
0.0000
4.6484
17.2257
17.1771
16.9315
16.3804
16.3170
16.9775
17.1294
17.1811
17.2239
17.2531
17.4011
17.5128
17.6113
17.7091
17.7842
17.8798
17.9785
18.0721
18.1688
18.2574
18.3486
18.4414

15

20

Figure 15: Impulse response of a VECM with IP, Real TB03, Equity, JPoD (Monthly data basis)
Response to Cholesky One S.D. Innov ations
Response of IPI to IPI

Response of IPI to REAL_T B03

Respons e of IPI to EQUIT Y

Response of IPI to JPOD

1.5

1.5

1.5

1.5

1.0

1.0

1.0

1.0

0.5

0.5

0.5

0.5

0.0

0.0

0.0

0.0

-0.5

-0.5

-0.5

-0.5

-1.0

-1.0
5

10

15

20

25

30

-1.0
5

Response of REAL_T B03 to IPI

10

15

20

25

30

-1.0
5

Response of REAL_T B03 to REAL_T B03

10

15

20

25

30

Res pons e of REAL_T B03 to EQUIT Y

.3

.3

.3

.3

.2

.2

.2

.2

.1

.1

.1

.1

.0

.0

.0

.0

-.1

-.1

-.1

-.1

-.2

-.2
5

10

15

20

25

30

-.2
5

Response of EQUIT Y to IPI

10

15

20

25

30

10

15

20

25

30

Response of EQUIT Y to EQUIT Y

-2
5

10

15

20

25

30

Response of JPOD to IPI

10

15

20

25

30

10

15

20

25

30

Response of J POD to EQUIT Y

.0012

.0012

.0012

.0008

.0008

.0008

.0008

.0004

.0004

.0004

.0004

.0000

.0000

.0000

.0000

-.0004

-.0004

-.0004

-.0004

-.0008
10

15

20

25

30

-.0008
5

10

15

20

25

30

10

15

10

15

IPI
100.0000
88.4442
74.1233
68.5104
68.7661
69.6652
69.2742
68.3371
68.2042
67.9965
67.9086
67.8722
67.7449
67.6049
67.4884
67.3746
67.2962
67.1869
67.0751
66.9719
66.8659
66.7633
66.6620
66.5588

REAL TB03
0.0000
0.8170
5.5543
12.5793
12.4864
12.1828
12.1174
12.6348
12.6688
12.6392
12.6759
12.6582
12.6668
12.6522
12.6304
12.6200
12.6015
12.5878
12.5790
12.5624
12.5474
12.5322
12.5177
12.5045

- 55 -

25

30

20

25

30

-.0008
5

10

15

20

25

30

10

15

Table 12: Variance Decomposition of IPI

Period
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

20

Res ponse of J POD to J POD

.0012

30

-2
5

Response of JPOD to REAL_T B03

-.0008

25

-2
5

20

Response of EQUIT Y to JPOD

-2

15

-.2
5

Respons e of EQUIT Y to REAL_T B03

10

Respons e of REAL_T B03 to JPOD

EQUITY
0.0000
5.6159
5.3765
5.0457
4.9929
4.8427
5.2296
5.1645
5.2142
5.3264
5.3411
5.4020
5.4563
5.5173
5.5978
5.6410
5.6956
5.7620
5.8193
5.8788
5.9357
5.9934
6.0530
6.1093

JPOD
0.0000
5.1229
14.9459
13.8646
13.7545
13.3093
13.3789
13.8637
13.9129
14.0380
14.0743
14.0675
14.1320
14.2257
14.2834
14.3644
14.4067
14.4634
14.5267
14.5869
14.6510
14.7111
14.7674
14.8273

20

25

30

Figure 16: Impulse response of a VECM with IP, CPI, Real TB03, Equity, JPoD (Monthly data basis)
Response to Cholesky One S.D. Innovations
Response of IPI to IPI

Response of IPI to CPI

Response of IPI to REAL_T B03

Response of IPI to EQ UIT Y

Response of IPI to JPO D

1 .5

1 .5

1 .5

1 .5

1 .5

1 .0

1 .0

1 .0

1 .0

1 .0

0 .5

0 .5

0 .5

0 .5

0 .5

0 .0

0 .0

0 .0

0 .0

0 .0

-0 .5

-0 .5

-0 .5

-0 .5

-0 .5

-1 .0

-1 .0
5

10

15

20

25

30

-1 .0
5

Response of CPI to IPI

10

15

20

25

-1 .0

30

Response of CPI to CPI

10

15

20

25

30

-1 .0
5

Response of CPI to REAL_TB03

10

15

20

25

30

Response of CPI to EQ UIT Y

.4

.4

.4

.4

.4

.3

.3

.3

.3

.3

.2

.2

.2

.2

.2

.1

.1

.1

.1

.0

.0

.0

.0

.0

-.1

-.1

-.1

-.1

-.1

-.2

-.2

-.2

-.2

-.2

10

15

20

25

30

Response of REAL_T B03 to IPI

10

15

20

25

30

Response of REAL_TB03 to CPI

10

15

20

25

30

10

15

20

25

30

Response of REAL_TB03 to EQ UITY

.3

.3

.3

.3

.2

.2

.2

.2

.2

.1

.1

.1

.1

.0

.0

.0

.0

.0

-.1

-.1

-.1

-.1

-.1

-.2

-.2

-.2

-.2

-.2

10

15

20

25

30

Response of EQ UIT Y to IPI

10

15

20

25

30

Response of EQ UIT Y to CPI

10

15

20

25

30

10

15

20

25

30

Response of EQ UITY to EQ UITY

-4

-4

-4

-4

-4

10

15

20

25

30

Response of JPO D to IPI

10

15

20

25

30

Response of JPO D to CPI

10

15

20

25

30

Response of JPO D to REAL_TB03

10

15

20

25

30

Response of JPO D to EQ UIT Y

.00 10

.00 10

.00 10

.00 10

.00 05

.00 05

.00 05

.00 05

.00 05

.00 00

.00 00

.00 00

.00 00

.00 00

-.00 05

-.00 05

-.00 05

-.00 05

-.00 05

-.00 10
5

10

15

20

25

30

-.00 10
5

10

15

20

25

30

-.00 10
5

10

15

20

25

30

10

15

20

25

30

10

15

20

25

30

10

15

20

25

30

Response of JPO D to JPO D

.00 10

-.00 10

30

Response of EQ UIT Y to JPO D

25

.1

Response of EQ UITY to REAL_TB03

20

Response of REAL_TB03 to JPO D

.3

15

.1

Response of REAL_TB03 to REAL_TB03

10

Response of CPI to JPO D

-.00 10
5

10

15

20

25

30

10

15

20

25

30

Table 13: Variance Decomposition of IPI

Period
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

IPI
100.0000
87.5458
74.7336
68.6723
68.8998
69.1909
68.0768
67.4245
67.3807
67.1628
66.9303
66.7878
66.6573
66.5707
66.4761
66.3599
66.2702
66.1551
66.0490
65.9600
65.8624
65.7642
65.6643
65.5649

CPI
0.0000
2.5528
2.1559
1.9788
1.9573
2.6920
3.2255
3.1914
3.2412
3.2816
3.3852
3.4985
3.5594
3.5672
3.5665
3.5671
3.5909
3.6207
3.6532
3.6696
3.6800
3.6935
3.7118
3.7323

REAL TB03
0.0000
0.3415
4.2264
11.9193
11.8522
11.4387
11.2922
11.7939
11.7857
11.7353
11.7753
11.7756
11.7551
11.7403
11.7237
11.7203
11.6982
11.6754
11.6570
11.6381
11.6228
11.6054
11.5863
11.5678

- 56 -

EQUITY
0.0000
5.1387
4.9947
4.6174
4.5769
4.4490
4.9464
4.9010
4.9202
5.0291
5.0339
5.0867
5.1436
5.2014
5.2952
5.3396
5.3898
5.4563
5.5109
5.5729
5.6326
5.6906
5.7512
5.8067

JPOD
0.0000
4.4212
13.8895
12.8122
12.7138
12.2294
12.4590
12.6891
12.6722
12.7913
12.8753
12.8515
12.8847
12.9204
12.9386
13.0131
13.0509
13.0925
13.1299
13.1595
13.2023
13.2463
13.2865
13.3283