1133_2446_abridged Liquidity Stress Testing PRIMA September 2006_presentation

8/3/2019 1133_2446_abridged Liquidity Stress Testing PRIMA September 2006_presentation

1/50

1Copyright 2006 Liquidity Risk Advisors

Practical Liquidity RiskStress Testing

Leonard Matz


2/50


Agenda

Why Stress Test Liquidity

General Considerations

Four Methods

Additional Considerations


3/50


Why Stress Test?

Stress testing is an important element for soundrisk management and contingency planning:

Stress scenarios highlight potential problemsThe untimely liquidation of assets can be costly

Good advance planning can, at least potentially,prevent insolvency


4/504Copyright 2006 Liquidity Risk Advisors

Why Stress Test?

The liquidity strategy should set out the generalapproach the bank will have to liquidity, includingvarious quantitative and qualitative targets. Thisstrategy should address the bank's goal of protectingfinancial strength and the ability to withstand stressfulevents in the marketplace.

Source: paragraph 7, Sound Practices for Managing Liquidity in Banking OrganizationsBasel Committee on Banking Supervision, Basel February 2000

BIS Guidelines Require Stress Testing



Why Stress Test?

Holding Liquidity is Not Free

No bank can hold enough liquidity to survive anything

close to a worst case liquidity crisis.The penalty for too little liquidity may be the failure of thebank but too much liquidity carries a penalty as well.

Optimal management of liquidityrequires a delicate balancebetween liquidity risk andincome.



Agenda


General ConsiderationsFour Methods



7/50


Definitions

Some risk professionals use the term scenario analysis to identifydeterministic tests while stress testing refers to probabilistic tests.Some risk professionals define stress testing as uni-varianttesting while scenario analysis is then defined as multi-varianttesting. Sometimes, sensitivity analysis is used to refer to uni-variant testing while stress testing refers to multivariate testing.

The BIS and defines scenarios to be the description of anintegrated future view. Sensitivity tests are uni-variant tests usedto establish the extent to which an outcome depends upon a singlevariable or single assumption. Stress tests are integrated, multi-

variant tests that show degrees of severity for scenarios.Integrated refers to the fact that assumptions for bothindependent and dependent variables reflect inter-relationshipsbetween variables.


8/50


Definitions - Source

A Survey of Stress Tests and CurrentPractice at Major Financial Institutions,report by a working group established by theCommittee on the Global Financial System,Bank for International Settlements, April2001, page 7.


9/50


What Liquidity Are We StressTesting?

Market Liquidity

Ability to sella Financial Instrumentin volume and time, without affectingthe market price

Funding Liquidity

Capacity to raise sufficientcashwhen neededto meet its paymentobligations

Standby LiquidityCapacity, in non-normal conditions,to raise sufficientcashto meet itspayment obligationsCapacity to maintain perceptions offunds providers


10/50


Sources of Contingent Risk

Unforeseen decreasesin funding available.

Unforeseen increasesin funding required.

Unexpected deposit withdrawals.

Requirements for increased collateral pledging.

Unexpected loan demand

Evaporation of unused funding capacity.

Erosion in the market value of unencumbered,marketable assets.

Unexpected credit draw downs and other commitmentdraws.

Source: Adapted from a slide created by Werner DHaese, Liquidity Contingency Plan: A Case Study at

Fortis, Fortis Central Risk Management, ALM Europe, London, October 2005.


11/50


The need for robust stress testing is

driven by three underlying truths:

Financial institutions can never avoid liquidity risk. Serving asliquidity intermediaries is one of the main ways that banks addvalue to the economy.

Financial institutions can do little to hedge liquidity risk. Somecontingent lending arrangements exist. But only a few remainin place in the event that the borrowing banks risk increases.

Most importantly, no financial institution can afford to holdenough liquidity to survive a severe or pro-longed fundingcrisis. Yet, since the least likely events can have devastatingconsequences, risk managers cant ignore them.


12/50


13/50


What is An

ExtremeEvent?


14/50


Agenda



Four Methods



15/50


Four Approaches for StressTesting

Sensitivity testing

VaR

Deterministic Scenarios

Probabilistic Scenarios


16/50


(1) Sensitivity Analysis: Uni-variant Stress Tests

Moving key variables one at a time:

simple and intuitive method

example:

the portfolio is ling the dollar vs. yen

we suppose the dollar could fall by 15% in one week; thisgives us a worst loss of $600 million

problem is with multiple sources of risk:

if the portfolio also contains positions in Japanese and USequities, we would have to predict movements in thesemarkets as well

we cannot assume the worst loss will occur at the sametime in all markets

Source: Professor Philippe Jorion, U.C.


17/50


(2) Historical VaR

Everyones favoriterisk metric.


18/50


Volatility of Savings Deposits

The good news: The bank has not experienced a severe loss of deposits.

The bad news: The historical observations tell us NOTHING about a future stressenvironment.

-10,000

-8,000

-6,000

-4,000

-2,000

0

2,000

4,000

6,000

8,000

10,000

Red lines indicate 2 SD

L Di t ib ti


19/50


Loss Distribution

Se lec t edc o n f i d e n c e

l eve l

The ta i l o f the d is t r ibu t ionrepresents the quant i ty of

e x t r e m e l o s s .

M e a n

U n e x p e ct e d l o s s

FR

E

QUE

NC

Y

SEVER ITY


20/50


VaR Stress Tests

Probability

Severityof loss

Stress testing typically appliesstatistical tools to provide moreinformation about the tail.

VaRExtreme Value TheoryOther tools

14


21/50


VaR Stress Tests and Liquidity Risk

Probability

Severityof loss

We cant apply historical VaR toliquidity risk stress testing!

Very few banks have had theunfortunate experience of a near

death experience.

Those that have dont have any recentexperience.

The tail in any one banks data simplydoesnt include the sorts of stress

experiences that keep liquidity riskmanagers awake.

15


22/50


Whats Normal / Extreme ?

Market and Funding Liquidity Risk: concentrated inNormal Scenarios

What is normal?

Quantitative tools are applicable.

Liquidity Risk: is concentrated in Extreme Scenarios

What is extreme?

There is no Canonical Answer

Distinctions are derived from everyday life; however as noconsensus can be reached:

They are meaningless in a quantitative context

Adapted from material developed by Dr. Robert E Fiedler.


23/50


VaR Conclusions

Historical observation does not necessarily reflect what mighthappen (future events)

Modelling a (fat tail) distribution does not solve the problem

either:Outlying point or fat tail?

Risk is not linear in extreme events

The Question is not: What Risk will we get if we push out the

quantiles? The answer to that question is only a matter ofscaling and is therefore meaningless!

Instead, the question is: Is there a structural change that thebank should model?

Adapted from material developed by Dr. Robert E Fiedler


24/50


BIS Recommends Scenario Analysis

A bank should analyzeliquidity utilizing avariety of what if

scenarios.

BIS: Sound Practices For Managing Liquidity in BankingOrganizations, February 2000.

(3) Deterministic Stress Testing

B i M t D i i f


25/50


Basic Management Decisions forDeterministic Stress Tests

What risk scenarios are stressed? Market liquidity risk,systemic risk, bank specific risk?

What risk exposures are stressed? Deposit losses,

borrowings, off balance sheet in-flows, off balance sheetout-flows?

How much stress?

What time horizons?

What are the impacts/results?

What will the bank do about the indicated exposures?


26/50


Step 1

Define the scenarios and stress levels youwish to test.

Important Point: It is not a good idea to use

changes in the banks rating to define thestress levels. The ratings changes are laggingindicators of trouble and almost always followthe market. Identify and use precursors forratings downgrades instead.


27/50


Systemic Crises A Wide Variety

19871990199119921994199519971998

19992000

20012002

U.S. stock market crashcollapse of U.S. high yield (junk) bond marketoil price surgeERM (European Exchange Rate Mechanism) crisisU.S. bond market crashMexican CrisisAsian crisisRussian default, Ruble collapse. LTCM

gold pricesTMT (telecommunications, media & technology )sector collapseSeptember 11 payments system disruptionArgentine crisis

St f F di C i i


28/50


Stages of a Funding Crisis


29/50


Step 2

For each scenario, at each stress level,identify the contractual cash in-flows and out-

flows that can be expected to occur in eachfuture time period for your forecast. The netcontractual flows for each loan and depositare clearly denoted in the line itemdescriptions.


30/50


Step 3

For each scenario, at each stress level, estimate the cashin-flows and out-flows resulting from customer behaviourThese may be from the exercise of contractual optionssuch as loan prepayment options or options to withdraw

funds from indeterminate maturity deposit accounts. Usethe assumption estimation guidelines discussed in the priorsection. Customer behavior changes also include non-contractual actions such as new loans. Make sure thateach customer behavior driven change in your forecast isappropriate and consistent for the scenario and the stresslevel.


31/50


Step 4

Add the non-discretionary treasury cashin-flows and out-flows. These includecontractual maturities of investments

and borrowings. Note that if you havecallable securities, the call exerciseneeds to be forecast in a time bucketconsistent with the interest rate

assumption for that scenario at thatstress level.


32/50


Step 5

Sum the customer driven cash flows and thenon-discretionary treasury flows. This total isthe forecasted liquidity requirement for each

time period in each scenario at each stresslevel.


33/50


Important note: The contractual cash flows areshown separately from the assumption based cashflows. This permits you to easily identify net cash

flows that are particularly assumption driven.

Important note: None of the cash flows depictedreflects any discretionary asset sales or borrowings.The black line is showing the risk quantity before anymanagement driven corrective actions areincorporated. Thus the forecasted risk quantity is inno way masked or hidden.

Format Suggestions


34/50


Step 6

Calculate the net cash flow coverage ratio foreach time bucket in each scenario at each

stress level. This is the amount by which theforecasted cash in-flows exceed (or fall shortof) the forecasted out-flows.


35/50


Important point: Forecasts for ordinary course of businessscenarios and for scenarios at minimal stress levelsshould always show a cushion. A positive margin for erroris required to offset potential model risk.

Important point: The forecasted cash flow cushion foreach time period, in each scenario at each stress levelcan be compared to either required liquidity risk minimumsor recommended guidance minimums.

Important point: The number of cash flow cushion ratios

can quickly get out of hand. For example, if you use 12time periods for bank specific scenarios done at each ofthree stress levels, you will have 36 cash flow cushionratios just from this single scenario type. The solution is toreport these in summary form.

Report and Analysis Suggestions


36/50


Step 7

Determine The Quantity of Standby Liquidity Available

The quantity of liquidity needed, the black line, must becompared to a quantity of liquidity available. (The

quantity of liquidity needed is sometimes called thecounter-balancing capacity.)

Important point: Unless we can compare somequantity of liquidity need to some quantity of liquidity

available, it is impossible to answer the central liquidityrisk management question: Is the current level of riskacceptable?

In order to forecast the quantity of


37/50


q ystandby liquidity available, we need to

do four things:

First, we need to determine what sources areavailable in each scenario at each stress level.

Second, we need to determine in what order we plan

to access each source. This can be described as acash flow waterfall.

Third, we need to forecast how long it will take toconvert each source to cash.

Fourth, but in no way least important, we need todetermine how much cash we can obtain. Haircutsfor marketable securities can very significantly withchanges in prevailing interest rates and changes incredit spreads.

Wh t i M k t bl ?


38/50


What is Marketable?

A 2005 survey of large U.S. andCanadian banks found that only 4 out of

17 reported that they varied theprojected quantity of liquid assets to fitthe scenario.


39/50


Step 8

Compare the Quantity of Forecasted Need tothe Quantity of Forecasted Funds Available.


40/50


Standby Liquidity Metrics

Net cash flow quantities by currency.

Days of sufficient cash flow.

St 9


41/50


Step 9

Make the Management Connection

No bank managers intend to preside over the failure of theirbank. When we say that the bank has a large enough standbyliquidity reserve to survive for five months, we definitely do notmean to imply that at the end of five months management locksto doors and gives the keys to the bank liquidators.

It is essential to view the liquidity reserve survival period as awindow of opportunity. It tells the banks managers how longthey would have to come up with additional cash in the event ofa scenario like the one projected. It also tells them how muchadditional cash they have to obtain.


42/50


(4) St h ti St T ti ?


43/50


(4) Stochastic Stress Testing?

Use Monte Carlo for cash flow forecasts.

At first glance, Monte Carlo stress testing for

liquidity risk appears to suffer from the samelimitations as historical VaR.

How do we estimate the parameters?

What about greatest hits , multi-period,Monte Carlo?

Multi Period Stochastic Stress


44/50


Multi-Period Stochastic StressTests

Initial state

Mean reversionVolatility

Can we study sets of similar,

historical liquidity events to

estimate Monte Carlo parameters?

A d


45/50


Agenda



Four Methods


Liquidity Stress Testing


46/50



Must focus on the interactions betweencredit risk, rate risk, market risk and liquidityrisk.

Must recognize that liquidity problems do notarise in a vacuum.

Must recognize that lack of sufficient liquidity

almost never TRIGGERS a liquidity crisis.



47/50



Must be relevant to the bank, its balancesheet and its business environment.

Must be relevant to the economicenvironment,

Must be severe enough to reveal tail riskexposure (i.e. improbable risk) without being

completely implausible.Must be detailed enough to highlightvulnerabilities.

Applying Stress Testing


48/50


Applying Stress Testing

Stress testing is not an end unto itself.Results have to be used in the riskmanagement process.

Results should be carefully consideredwhen setting risk limits.

Contingency plans for actions such as assetliquidations should be based on stress testresults.

Limitations of Stress Testing


49/50


Limitations of Stress Testing

Stress testing is only a tool for improving riskmanagement not a substitute for risk management.

Stress tests lack benchmarks or standards for

evaluating results.Stress tests do not reveal probabilities. We focuson plausible improbabilities without knowing howplausible or improbable.

Stress scenarios are deterministic and limited innumber.


50/50


QuestionsQuestions ??

[email protected]

1133_2446_abridged Liquidity Stress Testing PRIMA September 2006_presentation

Documents