Stress Testing at Major

Financial Institutions:
Survey Results and Practice

Yuko Kawai
Bank of Japan

Paper presented at the Expert Forum on Advanced Techniques on Stress Testing: Applications for Supervisors
Hosted by the International Monetary Fund
Washington, DC May 2-3, 2006

The views expressed in this paper are those of the author(s) only, and the presence of them, or of links to them, on
the IMF website does not imply that the IMF, its Executive Board, or its management endorses or shares the views
expressed in the paper.
Stress testing at major financial institutions:
survey results and practice
based on the report issued by a WG established by the CGFS (January 2005)

May 2006

Bank of Japan
Yuko Kawai

1
This presentation package is intended to summarize the results of the survey and
discussions conducted by the Working Group established by the Committee on the
Global Financial System (CGFS), BIS. The Working Group issued the report titled Stress
testing at major financial institutions: survey results and practice in January 2005, which
is available for download from BIS website (www.bis.org/publ/cgfs24.pdf).
Please note that the report was published in January 2005, based on the results of the
survey conducted in 2004, hence the contents of this presentation does not reflect
market/practice updates between the publication date and now. Also, the presentation
includes the speakers personal opinions, which may or may not be supported by WG
above mentioned or by Bank of Japan.

2
 AGENDA
1. Outline of the survey conducted by CGFS WG
2. The practice of stress testing
Definition, Types and Roles
Coverage, Formulation
3. Details
Scenario examples
Market risk, Credit risk, other risks
Funding liquidity, Market liquidity
New products
Integration
4. Conclusions

3
1. Outline of the survey conducted by
CGFS WG

4
 Stress Test Survey
WG of CGFS Collection of private
Year 2000 : 1st Survey sector stress tests
Report published in April 2001 conducted by each
Year 2004 : 2nd Survey (chaired by Bank of Japan) institution, not the
Report published in Jan 2005 macro top-down stress
test exercise
Local survey conducted in 2002 [Japan]
Participants and methods
62 financial institutions from 16 countries (mainly banks)
Collected real stress tests as of May 31 2004
Conducted follow-up interviews

Two years ago

Oil prices were just starting to rise
Markets expected US rate hikes
AND
Implementation of Basel II was at the preliminary stage at many banks
5
 Developments from 2000 to 2004
Developments observed between the last survey (year 2000)
and this survey (year 2004)
Stress tests are used more broadly
Wider range of applications
Wider range of coverage

Stress tests vary by institution

Issues to be solved
Stress testing on credit risk
Integration of risk management

6
 Overview of the survey results
Total number of tests collected: 963
Not necessarily covering
the full range of risks
By TYPE expected to realize in the
Historical
scenario, 212 markets with certain
probabilities, rather, the
Sensitivity,
risks related to the direct
458
Influenced by then- exposures
prevailing market
environments: we
may see more Hypothetical
CREDIT today scenario, 293

By RISK By Risk Region

North
Multiple America
Other
Property 7% 5% Interest 19%
3% Rates Global
38% 39% Europe
Credit
18% 9%

Japan
Commodities Survey 5%
4% FX Equities participants were Other Emerging Non-Japan
12% 13%
mostly BANKs 10% 5% Asia
13%
7
2. The practice of stress testing

8
 Definition
Definition
Risk management tool used to evaluate the potential impact on a firm (or a
division) of a specific event and/or movement in a set of financial variables

Stress tests may

cover
ORDINARY
events as well
A complement to
risks covered by
statistical models such STRESS TESTs
as VaR
VaRcapturing risks as
a plane
Stress test: capturing
risks as points unexpected risks

risks covered by
VaR

9
 Types
Stress tests generally fall into two categories
Scenario tests : Assume the background (i.e., source of risk) of stress
Sensitivity tests : Only risk factors specified

Historical scenarios
Scenario tests Hypothetical scenarios There are no
Stress tests standard
definitions/categori
Sensitivity tests Historical sensitivity
zations; hybrids
Simple sensitivity are common

Scenario vs. Sensitivity

Scenario Sensitivity
Source Risk factors No source No route Risk factors

price price price

change change change
price price price
change change Event change
Event
price price price
change change change
price price price
change change change
10
 Roles
Stress tests work as a communication tool and/or risk management tool
Formulation/running frequency depends on the purpose
The roles are not mutually exclusive; one stress test may have multiple
roles

Communication tool

B. Capture the impact on a portfolio

A. Analyze the of exceptional but plausible large G. Inputs to
risks under the loss events VaR
concern of
management C. Allocate/verify the limits/capital H. Measure
risks outside
D. Evaluate business risk
of VaR
E. Understand the risk profile
F. Measure the sensitivity of
portfolio

Risk management tool 11

 Coverage
Various parts of a companys financials are covered

Balance Sheet P/L statement

(assets) liabilities revenue
High liquidity (5) P/L planning

(1) Market risk

expense
(3) Funding

capital profit

(2) Loans, Real Estates (4) Capital

(loss, allocation)
Low liquidity
(6) Integrated scenario

(7) Operational risk (8) Business risk

12
BCP
 Formulation
Event-driven approach
What will happen if ?

Event Risk factors

(Step 1) Identify a risk source which (Step 2) By how much do risk

causes changes in financial markets parameters change is such an event
occurs?
Portfolio-driven approach
When do we lose money? more
common

Event Risk factors PORTFOLIO

(Step 2) What events might bring (Step 1) What are the risk parameter
about these changes? changes which result in a portfolio loss?

13
A. Analyze the risks under managements concern

Purpose
Know what risks are under the attention of managements, and measure
their impacts
Examples
Conduct the test to measure the influence on a banks portfolio of a
hypothetical prolonged Iraqi War/terrors
Formulation
By the request of managements
Event-driven approach
Trade-off envisaged between understandability and reality
Good way to make
numbers more
Asks for the specific comprehensive!
Board members Unit scenario running
managers Risk management unit
Return results
What will happen Ah, this risk must be
to the firm (unit) if considered! 14
B. Capture the impact on a portfolio of exceptional
but plausible large loss events
Purpose
Measure the impact of exceptional but plausible large loss events
Trigger to the contingency plan
Examples
Calculate the todays loss to be caused by Black Monday
Formulation
Mainly Portfolio-driven approach
Sensitivity analysis and Historical scenarios are more applicable
May focus on short-term impacts
May set risk parameters outside of VaR assumptions

What if Interest Rate

What if Black Monday
rises beyond
happens today?
VaR estimation?
15
B. Capture the impact on a portfolio of exceptional
but plausible large loss events
Multi-factor sensitivity analysis
Size of risk factors
The worst data during the observation period
The standard deviation for the observation period times multiplier (e.g.,
a number larger than the VaR assumption, data during the stressed
period)
Unit change e.g. +1-5pts
Combination of risk factors
Add-up of the worst data
Based on certain risk scenarios
Correlation assumed
Holding period
From the peak to the bottom, unit periods
Observation period
Conservativeness vs. plausibility
16
 C. Allocate/verify the limits/capital
Purpose
Direct/indirect inputs to the allocation of economic capital and/or risk limit
settings
Verify/diagnose the adequacy of allocation/limits otherwise established
Examples
Confirm that the loss under stress will not exceed the allocated capital

Economic Allocating Market risk Interest

capital economic rate
capital
Credit risk Equity
Results of a
certain stress
Operational FX test: Verify the
VaR, stress- risk
adequacy of
test results Other risks capital allocation
Others
and/or P/L
plan

17
 C. Allocate/verify the limits/capital
Formulation
Event-driven approach
Scenarios normally assume risk durations as long as 1 year
Include low-liquidity risks (e.g. loan portfolio, real estates related)
(If used as direct inputs)
Subjectivity (with reasonable plausibility) is a key
Economists, traders maybe involved from the early stage
Alternatively, worst case historical scenarios may be applied
Macro models maybe employed
Probability maybe assigned to stress scenarios (EVT-Extreme Value
Theory)
Feedback effects must be considered

Business unit Risk management

we want more Ah, but this is the
capital/limit ! risk amount to be
envisioned 18
 D. Evaluate business risks
Purpose
Inputs to the determination of the annual P/L plan and/or verification of
such plan
Examples
Main underlying scenario is the gradual recovery of economy, while
stressed risk scenario is the sudden acceleration
Formulation
Same as C (previous slide)
Include risk factors affecting P&L (not only asset value)

How much can we earn ?

Economy will be..
Markets are Will our profit target be achieved?
expected to be
Central bank will What is the potential downside?
19
 E. Understand the risk profile
Purpose
(1) Identify the concentration risk for the whole company (across
divisions/products)
(2) Identify the risk diversification effect (one divisions risk maybe offset with
the others)
Examples
(1) Calculate the effect of default of one big company (such as a US
automaker) over credit/equity/counterparty risk portfolios
(2) Measure the effect of the long-term interest rate rise coupled with stock
price rise (e.g. summer 2003, Japan)
Formulation
Event-driven approach

portfolio effects
COMPANY level

Division 1 Division 2 Division 3 simple aggregation

20
 G. Inputs to VaR
Purpose
Inputs to VaR risk measurement
Examples
Determine the probability distribution used to calculate VaR, based on the
sensitivity test result for each asset class and its correlation to other
parameters under the stressed situation
Formulation
VaR itself requires the periodical back-test
Historical scenario/sensitivity (in order to claim the subjectivity)

21
 H. Measure risks outside of VaR
Purpose
Measure the risk which by definition is not captured by VaR
Examples
Cessation of currency peg (no historical data)
Risk measurement of brand-new products (no historical data)
Non-linear profile risks
Formulation
Portfolio-driven approach
Cover the risks where statistical approaches are of no use

No Data VaR is not applicable /

may not fully capture the Stress Testing !
risk
Jump Risk
22
3. Specific Issues

23
 Specific Issues
Scenarios
List of risks covered
Major scenarios
Market Risk
Interest rates, Equities, FX
Real Estate Risk
Credit Risk
Integration of risk management
Others
Funding liquidity
Market liquidity
New products

24
 Scenarios
Major historical scenarios
Selection depends on the institution
As of May 2004

Year Event Year Event

1973 Oil shock 1998 Financial systemic crisis (Japan)
1987 Black Monday 2000 IT bubble burst
1990 Gulf War 2000 US credit deterioration
1992 EMS crisis 2000 Argentina crisis
1994 Interest rate rise (global) 2001 US Terrors
1994 Mexican Peso crisis 2002 US accounting scandals
1995 European bond price drop 2002 ABS spread widening
1997 Asian crisis 2003 Iraqi invasion
1998 Russian default 2003 Interest rate rise (global)
1998 LTCM failure 2003 GSE spread widening (US)
1998 USD/JPY nosedive 2003 SARS outbreak

Major scenarios : blue-shadowed

25
 Scenario examples

Historical [Hypothetical] scenarios

Source Scenario Source Event
Economy/Fiscal policy Exogenous events
Global Economy [Recovery][Recession] Real Esate price price drop
Japanese Economy [Recovery][Recession] Oil Oil shock
US Economy [Recovery][Recession] Other commodities [Price up/down]
European Economy [Recession] Catastrophy Earthquake, outbreak of SARS
Argentina crisis, Russian crisis, [cabinet resignation][assasination of
Politics VIP]
Emerging Economy Mexican Peso crisis, [Moratrium of
specific country] Financial system, corporate sector
Asian Economy Asian crisis Banks specific [own downgrade]
Inflation [Inflationary fear] accounting scandal, [downgrade of
Corporate sector loan borrowers]
Monetary Policy
Financial system LTCM, Japan(1997-9)
[rate hike][lift of quantitative easing
BOJ monetary policy policy] Financial Market
USD drop, EMS crisis, [de-peg of
FRB monetary policy Policy change (1993) Rate hike(1994) Foreign Exchange currency]
Other monetary policy [rate hike][rate cut] Equity IT bubble crash, Black Monday
Geopolitical risk/Terrors Interest Rate Global interest rate rise (1994, 2003)
Terror 9-11, Madrid train terror, [local terror] GSE/ABS spread widening, [credit
Other products derivatives turmoil][hedge fund crisis]
Geopolitical (Middle East) Iraqi ivasion, Gulf War
Geopolitical (Asia) [North Korea]
26
 Market Risk (1) Interest Rates
Background to scenarios
Economic recovery, Inflation fear, Monetary policy change, Selling
pressure in the market.. etc.
Gap between the market recognition and the reality
Risk factors
Interest rate for each tenor/currency, shape of yield curve, volatility
Basis (cash-future), currency basis, TED spread, NII

27
 Market Risk (2) Equity
Background to scenarios
Economy, Terror (geopolitical risk), Oil surge, Political turmoil, Accounting
scandal etc.
Risk factors
Stock price (individual, index, by industry, beta), Volatility,
Correlation/spread (for long-short strategy)

Historical scenario examples (equity price drop)

Year Event
1987 Black Monday
1990/91 Gulf War
1997 Asian crisis
1998 Russian crisis/LTCM default
2000 IT bubble burst
2001 9-11 terror
2002 US accounting scandal
28
 Market Risk (3) FX
Background to scenarios
Political turmoil, De-peg, Fiscal deficit, Financial market crisis (especially in
the emerging market) etc.
There are equity drop/credit spread widening scenarios with FX as a
source of shock
Risk factors
FX rate (rise/fall, de-peg), volatility

Historical scenario examples (FX rate change)

Year Event
1992/93 EMS crisis
1997 Asian crisis
1998 Russian crisis/LTCM default
1998 USD nosedive against JPY
2000 IT bubble burst
2001 9-11 terror
2002 US accounting scandal 29
 Credit Risk
Background to scenarios
Macro factors such as Economy, Unemployment Rate
May use macro-model
Real estate price drop, terror, financial market crisis, accounting
scandals etc.
Risk factors
<for non-market exposures> Loan outstanding, Loan classification
(regulatory), Internal borrower ratings, Loss given default, Collateral value,
Probability of default
<for market exposures> Credit spreads, swap spreads, index value

more
products

Should have seen progresses

over the past two years market data
accumulation
regulations
30
 Issues (as of 2004/5) in Credit Risk management
Integration of risk management
Integrated framework to cover full range of credit exposures
Steps of credit risk management
Capture >> hedge >> active trade
Specific issues
Integration of risk scenarios

LOANS CORPORATE BONDS/CDS/STRUCTURED

Accrual accounting FINANCE PRODUCTS
Market price hardly Mark-to-market accounting
available Market price (partially) available
(PD/LGD risk factors) Bond/derivatives booking
Loan booking system system
Banking book Banking/Trading book

How should we add them up?

31
 Credit Riskcapture & measure risk
Various types of exposures to be covered:
Common quantity measurement required to add-up risk amount
Common risk factor to evaluate the whole portfolio
Market price may not be available for a substantial portion of the whole exposure

Credit exposure examples

Risk factors Market price available?
Loan (borrower) PD, LGD, collateral value, Depends on the country
(credit spread) and name

Bond (issuer) spread Yes

CDS (reference entity)

Financial contracts Spread, collateral value (almost) No

(counterparty)

32
 Credit Riskintegration
Risk management evolution <example>
Capture and add-up risks, reduce the excess
Common quantity measurement such as Loan Equivalent amount
System integration required to add them up
Internal ratings assigned (may be mapped to external ratings)
Reduce the over-exposure (sell loans, buy CDS production,
securitize)
(Quasi) mark-to-market and hedge economic capital
Mark all credit exposures (managerial accounting vs. reporting
accounting), with some estimations of credit spreads to be marked
hedge economic capital
Dynamic management
Actively traded credit markets
Buy and Sell the risks, to maximize the enterprise-wide risk-return

33
 Credit Riskrisk management issues
Accounting
Accrual vs. Mark-to-market
Loans (accrual) hedged by CDS (mark-to-market) produces accounting gaps

Market liquidity
Market prices are limited to traded names in CDS/Bond markets
The loan secondary market is limited in most countries, and SME names are
not traded even in the developed markets
Mark-to-market necessity vs. unavailability of input parameters

Data availability
Default rates/LGDs
Portfolio correlation
New risk factors
Structured finance correlations (CDO, basket/pooled/Index products)
Basis risk among bond spreads, loan spreads, CDS spreads
No historical data >> stress testing!

34
 Risk management and Market development
As the hedging activities increase (based Credit
on risk analysis), the credit market Market
expands and more data become
available for risk analysis to further fortify Risk Management
the risk management mechanism analysis
technology

Risk analysis based on market data

Credit
Warning to the management Market

Early managerial decision making

Hedging transactions executed

Risk
Credit market liquidity increases analysis Management
technology
35
 Credit RiskScenario integration
Macro economic model
Example 1: Simulate macro factor (such as unemployment rate) change
over collateral value/corporate sector earnings
Combined with simulation over market risks (e.g., FX, interest rates),
an enterprise-wide testing conducted
Market factor model
Example 2: Simulate real estate price drop, with the 1st effect on collateral
value and 2nd effect on corporate credit deterioration
3rd effect over market risks maybe combined
Trigger could be other risks (e.g., interest rates)

36
 Other risks
Real Estates
Background to scenarios
Slowdown of economy
Real estate price drops are mostly given as exogenous risk factors
Risk exposures being tested
Real estate price (direct holding/collateral to loan), mortgage loan
value
Measuring direct impact of holdings/loan value
Generic default probability/recovery ratio of loans
Measuring indirect impact on overall loans
Influence to FX/equity
Measuring effects on market risks
Other risks: Commodities, Emerging economy events etc.

37
 Funding liquidity
Fluctuation of funding needs
Withdrawal of deposits (insurance contracts, funds) or difficulty in
market financing
(possible cause) own downgrade, change in economy
Increased drawdown of commitment lines
(possible cause) borrowers credit deterioration
Rise of funding costs
Own credit deterioration
alternative finance tools, back-up lines
External event-driven (e.g., terror, financial systemic crisis)
support from central banks assumed

May lead to trigger contingency plans

Succeeding set of tests to evaluate the liquidation of
collateral/own assets for financing

38
 Market liquidity
No magic solutions
Use empirical data
Extend the holding periods / widen the range of price change
Distinguished the low-liquidity assets and apply conservative
parameters

Feedback effect [amplification of a negative market trend owing to loss-

cutting transactions by the institutions themselves]
May assume certain price changes (tough to quantify)

Aftermath of Major dealer exit

Game-theory?
May assume certain price changes/liquidity draining

39
 New products
Spreads of structured finance products
ABSCDOMBS
Basis risks
CDS-bond spread
Correlation risks
Credit correlation (basket/pool)
Cross-product correlation (e.g., equity-credit)
Hedge Fund (index)

40
 Integration of Risk Management
Organization
Common reporting line
Sum-up of similar risks
System integration
Common measurement of exposure quantity
Common risk factor
Common event/scenario setting
Applicable to all products and divisions
Applicable to assets, liabilities, and P/L
Recognition/evaluation of concentration/diversification
Consider diversification effect in allocating capital
Integration to include operational risks?
Common scenarios maybe applied

41
4. Conclusions

42
 Conclusions of the Report
Stress testing is becoming more integrated into risk
management frameworks
Various roles, not limited to the tool to capture exceptional risks
Credit risk management is a key focus for most institutions
Practice varies by institution and/or by market, while the most advanced
institutions claim there still are issues
Development of credit derivatives market may support the mark-to-market
approach
Integration of risk management has various aspects, such
as a common platform and an enterprise-wide scenario

Dry-up of market liquidity is being noticed, while no

legitimate risk measurement is invented yet

43
 Some thoughts
Stress testing is evolving dynamically,
Stress testing started as an attempt to measure the risks which are not
appropriately captured by regular statistical methods, such as VaR
Now the applications of stress testing are expanding dramatically
With its flexibility as a key element;
Flexibility and affinity to the institutions risk profile are the hearts of stress
tests
Therefore, there is no ideal framework / single exponent
of stress testing
Each institution builds its own set of stress testing, depending on its
portfolio/risk profile, market environment, expected roles of stress testing.

44
 In reading survey results ..
The list of scenarios/parameters is not a complete set of the likely events
expected by market participants
Institutions choose risk parameters to affect their risk profile (e.g.
interest rate up vs. down, fall in equity prices vs. rise), and may not
pick up the changing leading to profits
Stress test may mean completely different type of tests, depending on
who uses the term
Macro tests to check Financial Stability (FSAP type) vs. each
institutions risk management (covered by this survey)
Market risk only vs. including non-market risk
VaR supplement vs. stand alone tests
Asset evaluation vs. P/L planning, capital allocation
Operational risk

Do not try to read what are likely future events expected by market
participants through stress test results
Be specific in asking practitioners for information 45