IMFFSB EARLY WARNING EXERCISE

1
IMF-FSB EARLY WARNING EXERCISE
A Sampling of Analytical Tools Crisis Risk
Models Asset Price Models Fiscal Risk
Tools Financial Market Tools Banking System
Contagion Scenario
2
Purposes of EWE

• Identify systemic vulnerabilities sufficiently in
  advance that corrective policies can be
  implementedflag raising
• Warn of imminent risks that suggest tail events
  about to materialize
• Prioritize policy recommendations and formulation
  of contingency plansbased on probability and
  impact

3
EWE SetupInputs and Outputs
close coordination/ exchanges
IMF
FSB
Quantitative tools (including VEE/VEA)
Vulnerability Assessments
Early Warning List
Qualitative/ heuristic approaches (including
consultations)
IMFC EWE presentation
4
What can EW systems realistically achieve?

• EWS modelspoor record at predicting timing of
  crises because precise trigger differs across
  crises and is unpredictable.
• More successful in identifying underlying
  vulnerabilities, i.e. predisposition to
  crisisasset booms, currency or maturity
  mismatches on sectoral balance sheets, high
  leverageas well as transmission
  channels/spillovers
• Link multiple vulnerabilities connect the dots
  to strengthen persuasiveness of flag and see
  where crisis heads next
• In benign times, vulnerabilities still relevant

5
The Role of Analytical Tools

• Formal tools complement qualitative/ heuristic
  approaches by
• Identifying systematically vulnerabilities
• Has the qualitative approach overlooked a
  potential source of vulnerability?
• Exploring cross-sector, cross-country linkages
• If a shock/risk is realized, how can it spread?
• Quantifying repercussions of realized risks
• How large is the potential shock? What will be
  its impact?
• Disciplining and informing judgment
• If formal models are raising flags, cannot
  ignore conversely, if not raising flags, risk
  may not be as severe (requires further
  exploration)

6
What is a crisis for purpose of EWE?

• Crisis (early warning) work traditionally focused
  on EME crises (sudden stop, BOP, external sector)
• This crisis demands a broader view, but...issues
• Is an event a crisis or just an unpleasant
  time?
• Is it a financial crisis? growth crisis?
• Probably want to consider range of alternative
  definitions
• For EWE, issue must be systemic (large country,
  systemic group of countries, major asset or
  commodity market, systemic financial
  institutions, major sovereign) or propagation of
  crisis across economies/regions

7
Types of Tools

• Ad hocdeveloped to explore particular risks in
  specific EWE rounds
• Creditless recoveries
• Vicious cycles of declining potential growth and
  weakening fiscal positions (including model-based
  scenario analysis)
• Standardwill likely be used in multiple EWE
  rounds
• General economic and financial setting
• Country risk models supporting sectoral models
• Cross-border linkages

8
Toolkit

• Global environment
• Risks to global growth, commodity prices (WEO fan
  charts)
• Global financial risks financial stability and
  volatility heat maps (GFSR)
• Other global variables S-I balances direction
  and composition of capital flows
  multilaterally-consistent exchange rate
  assessments (CGER)
• Country risks
• Three empirical crisis risk models external
  crises for EMEs financial crisis and growth
  crises for advanced countries
• Other country-based toolse.g. house price
  misalignment and impact sovereign risks default
  probabilities for financial institutions, crisis
  duration
• Spillover risks
• Common distress across financial institutions,
  corporates, and sovereigns
• Contagion potential through cross-border bank
  lending channels (upstream, downstream
  vulnerabilities)scenario analyses

9
Overall Assessments

• EWE not a mechanical exercisebalance between
  formal tools and heuristic approaches
• Overall assessments draw on
• Formal models (flags raised by crisis models and
  sectoral modules)
• Judgment informed by surveillance and broad
  consultationcritical when models produce
  counter-intuitive or mixed results
• Aggregation Informs both country and sectoral
  risks.

10
Crisis Risk Models
General economic and financial setting
Crisis Risk Models Advanced and Emerging Market
Countries
Crisis duration/ recovery models
Fiscal sustainability/ financing
Asset price misalignment
Financing gaps
How likely? How large? How long?
High frequency financial data
Cross border linkages

11
Crisis Risk ModelsObjective

• Identify sources of crisis vulnerability with
  sufficient specificity and warning that
  corrective policies and contingent plans can be
  put in place
• Provide a summary of country-level
  vulnerabilities
• Help identify sectors that might be source of
  vulnerabilityconsonance with sectoral
  models/tools
• Discipline drill down process
• Provide crisis probabilities for contagion tools

12
Crisis Risk ModelsApproach

• Define events (crises) of interest (financial,
  growth, external)
• Identify medium (boom phase) and short-term
  (bust phase) indicators (external, macro,
  fiscal, asset prices, financial, corporate)
• Determine thresholds such that vulnerability is
  appreciably greater if the indicator value
  exceeds the thresholdtrading off type I and type
  II errors
• Example (if threshold is 5-year avg. house price
  growth gt 10)
• Frequency(crisis avg house price growth lt 10)
  1.4
• Frequency(crisis avg house price growth gt 10)
  8.4
• Higher vulnerability ratio 6 times more likely
  to have crisis
• Look for consonance with other sectoral tools
• Aggregate indicators (weighting by signal/noise
  ratio) to obtain countrys crisis vulnerability

13
Crisis Risk ModelsOutput
14
Crisis Risk ModelsPerformance
Vulnerability to a Financial Crisis
2003-2007 (based on data available through 2006)
15
Asset Price Vulnerabilities
General economic and financial setting
Crisis Risk Models Advanced and Emerging Market
Countries
Asset price misalignment
Crisis duration/ recovery models
Fiscal sustainability/ financing
Financing gaps
How likely? How large? How long?
High frequency financial data
Cross border linkages

16
Real Estate Market Vulnerabilities

• Price misalignments
• Price-to-rent ratio (price-dividend)
• Price-to-income ratio (affordability)
• Regression model (WEO income, interest rates,
  credit, equity prices and demographic variables)
• Commercial real estate (prime rent, vacancy
  rates)
• Rule-of-thumb misalignment when deviation of
  more than 1 std. dev. from historical mean or
  model
• Economic impact of price correction
• Structural and VAR models of impact of house
  prices on private consumption, residential
  construction, GDP
• Real estate vulnerability index (price
  misalignment, commercial real estate rent and
  vacancy changes, structural characteristics of
  mortgage market, household balance sheet
  position, importance of real estate-related
  activity in the economy)

17
Equity Price

• Equity price misalignments indicate potential
  vulnerability to price correction trigger for
  further analysis
• Price misalignments are defined as greater than
  10 percent deviations of actual stock market
  index values from theoretical (fundamentals-based)
  values
• Theoretical equity price indices are constructed
  using the dividend-discount model and the
  arbitrage-pricing model (APM)
• ...Misalignments are also gauged by looking at
  standard equity valuation multiples (e.g.,
  price/earnings ratios)
• Sample MSCI local currency indices for 15
  countries
• Sample period December 1987-June 2009

18
Fiscal Risk Tools
General economic and financial setting
Crisis Risk Models Advanced and Emerging Market
Countries
Asset price misalignment
Crisis duration/ recovery models
Fiscal sustainability/ financing
Financing gaps
How likely? How large? How long?
High frequency financial data
Cross border linkages

19
Fiscal RisksObjectives and Approach

• Assess risks of sharp increase in fiscal deficits
  and public debt (including associated with
  financial sector), recognizing that market
  reaction could be discontinuous and abrupt
• Two types of tools utilized
• Market indicators to examine the extent to which
  markets reflect concerns about rising deficits
  and debt
• More fundamental indicators providing an
  assessment of fiscal adjustment required to
  maintain public debt sustainability

20
Fiscal RisksMarket Indicators

• Sovereign bond yields
• Widely-used indicator of pressure on sovereign
  credit market measure the cost of borrowing
• But conflates other factors and risks (risk
  appetite, business cycle, inflationary
  expectations/exchange rate risk, size and
  liquidity of bond markets)
• Sovereign CDS spreads
• Correspond more closely to sovereign default
  risk/credit eventthough in some cases, covers
  very small proportion of outstanding debt
• Relative asset swap (RAS) spreads
• RASi Ri RSWi
• Ri yield of 10-year bond issued by country i
• RSWi 10-year fixed rate on interest rate swaps
  in currency of country i
• (RAS expected to be negative in normal times)

21
Fiscal RisksSustainability Measures

• A number of different ways to define adjustment
  needed
• Ideally, estimate debt sustainability threshold
  empirically but difficult to do so for advanced
  countries for a variety of reasons
• Practical approaches
• Adjustment to primary balance needed to return to
  a given debt ratio e.g. the 2007 ratio, in ten
  years
• Adjustment to primary balance needed to satisfy
  the inter-temporal budget constraint (captures
  longer-term fiscal pressures)

22
Fiscal RisksSustainability Measures

• Comparison of ranking of required primary
  balance adjustment

30
25
Debt reduction to 2007
level
Intertemporal budget constraint
20
15
Country ranking
10
5
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
23
FiscalShocks and Rollover Risks

• Beyond baseline solvency, consider possible
  shocks
• Lower growth
• Contingent liabilities (including from financial
  sector see CCA)
• Interactionlow growth environment, more likely
  that guarantees will be called
• Beyond solvency, sustainability also requires
  liquidity (to finance deficits and roll over
  debt)
• Risk that costs of new borrowing will increase
• Risk that countries can borrow only at shorter
  maturities
• Risk of quantity constraint

24
Financial Sector Risk Tools
General economic and financial setting
Crisis Risk Models Advanced and Emerging Market
Countries
Crisis duration/ recovery models
Fiscal sustainability/ financing
Asset price misalignment
Financing gaps
How likely? How large? How long?
High frequency financial data
Cross border linkages

25
Financial Sector Risk ToolsObjectives and
Approaches

• Regime-Switching Models
• Identify periods when global market conditions
  move into higher or lower volatility states
• Noisy market data
• Individual market data often capture both
  idiosyncratic risk as well as market risks.
• Contingent Claims Analysis
• Early warning indicators of bank and corporate
  distress
• Risk transfer to sovereign from financial sector
  and
• Systemic CCA used to determine systemic
  tail-risk in the banking system and government
  contingent liabilities.

26
Financial Sector Risk ToolsRegime Switching
Models

• Regime-Switching Models
• Market conditions are often triggers that
  reveal latent fundamental-based vulnerabilities
• Can be a helpful tool for policymakers to
  evaluate when market conditions are such that
• Even relatively small shocks can lead to systemic
  events.
• Financial institutions and markets can become
  distressed as a result of unstable market
  conditions.
• When conditions have improved sufficiently to
  begin to withdraw government support.

27
Regime Switching ModelsE.g., VIX

• After the Lehman episode, the VIX moved to
  historic heights.
• The model also picks up most of the historically
  identified periods of market stress.

28
Regime Switching Model of VIX Improved Market
Uncertainty (but not all the way
down)(Probability of being in low-, medium-, and
high-volatility state)

• The VIX model decisively enters the
  medium-volatility state in April 2009...but has
  not returned to the pre-crisis state.

29
Contingent Claims Approach
Assets Equity PV of Debt Payments
Expected Loss due to default If equity is traded
in the market, CCA uses forward-looking equity
information plus balance sheet data If there is
not traded equity, direct estimation of assets
and asset volatility can be used instead
Asset Value
Exp. asset
Probability Distribution of Asset Value
value path
Distance to Distress standard deviations asset
value is from debt distress barrier
V
0
Distress Barrier (promised debt payments)
Probability of Default
T
Time
Explicit or implicit government guarantees mean
the government is taking over part of the banks
Expected Loss due to Default
30
Contingent Claims Approach Widely Applied Already
Median 1-year EDF for banking sectors in two
regions

Examples of Expected Default Frequencies (EDFs)
calculated daily for 35,000 financial and
corporate firms, and sectors, in 55 countries
using a CCA-type model (Source Moodys
KMV)
1-year EDF for two example banks
Median 1-year EDF for two EM corporate sectors
31
Higher Contingent Liabilities Can Increase
Sovereign Risk

• Country Example contingent liabilities are
  correlated and lead the sovereign default
  probability (inferred from sovereign CDS spreads)

Total contingent liabilities (LHS, USbn)
Est. sovereign default prob. 1-year (RHS, )
32
Contagion and Cross Border Linkages
General economic and financial setting
Crisis Risk Models Advanced and Emerging Market
Countries
Crisis duration/ recovery models
Fiscal sustainability/ financing
Asset price misalignment
Financing gaps
How likely? How large? How long?
High frequency financial data
Cross border linkages

33
Contagion and Cross Border LinkagesObjectives
and Approaches

• Financial Market Co-dependence
• Allows for correlations to depend on extreme
  events, and to evolve over time
• Common Distress (CDS credit events)
• Joint Probability of Distress (JPoD) Likelihood
  of common distress of all Financial Institutions
• Bank Stability Index (BSI) Expected number of
  FIs in distress given that at least one becomes
  distressed
• Cross Border Bank Claims Analysis
• Provides vulnerability measures of creditor
  countries resulting from exposures to main
  borrowers, and of borrowers countries resulting
  from exposures to main creditors
• Assesses the propagation of financial shocks
  across borders through default and deleveraging
  (scenario analysis)

34
Financial Market Codependence

• Correlation coefficients
• Capture average co-movement between e.g., equity
  prices
• But may be misleading in times of distress/
  extreme events
• Codependence (multivariate, non-linear,
  time-varying dependence structureGeneralized
  Extreme Value Theory)
• allows co-movement to be non-constant and to
  depend on size of movement

35
Financial Market Codependence
Correlation coefficient 0.4
36
Common Distress

• Help assess financial stability from three
  complementary perspectives
• Tail risk in the system
• Joint Probability of Distress (JPoD) Likelihood
  of common distress of all FIs in the system
• Bank Stability Index (BSI) Expected number of
  FIs in distress given that at least one becomes
  distressed
• Distress between any two specific financial
  institutions (FIs)
• Cascade effects in the system triggered by
  distress of specific FI A measure of systemic
  importance
• Note Distress risk Larger coverage than
  default risk.
• For example, CDS embedded distress
  probabilities may include default and other
  types of credit events.

37
Common Distress Modeling Approach

• Step 1
• View the financial system as a portfolio of
  banks.

Step 4 Estimate Financial Stability Measures.
Based on conditional Probabilities of distress
PoD of Bank X
Step 3 Recover the Financial System
Multivariate Density It characterizes the implied
asset values of the portfolio of FIs and the
distress dependence amongst these FIs.
PoD of Bank Y
Step 2 Estimate individual FIs Probabilities of
Distress (PoDs)

• Can be constructed from very limited data, using
  market-based or micro-founded approaches
  flexible approach that easily incorporates
  non-bank FIs, e.g., insurance companies, pension
  funds, hedge funds.

38
Common DistressExample
Joint Probability of Distress (JPoD) Likelihood
of common distress of all the FIs in the system.
Banking Stability Index Expected number of FIs
in distress given that at least one became
distressed.
39
Common DistressExample of cascade effects
Probability that at least one bank becomes
distressed given that Lehman Brothers becomes
distressed
Lehman Brothers
40
Cross Border Bank Claims Analysis

• Models cross border propagation of financial
  shocks
• Shocks in one countrys banking, non-financial
  private or public sector propogate to the banking
  systems of other countries, causing defaults and
  deleveraging
• Provides vulnerability measures of creditor
  countries resulting from exposures to main
  borrowers, and of borrowing countries resulting
  from exposures to main creditors

41
Scenario Analysis

• Stylized bank balance sheet
• ASSETS CAPITAL Other LIABILITIES
• domestic
• foreign
• The scenario analysis focuses on asset/funding
  shocks
• - Losses on assets deplete capital (1st round)
• - A sound capitalization ratio is restored by
  deleveraging (assume no recapitalization) (2nd
  round)
• - Banks reduce their lending to other banks
  (funding shocks), causing fire sales, and further
  deleveraging (3rd round)
• - Potential bank failures causing additional
  losses on other banks
• Deleveraging achieved by
• (i) Loans not being rolled-over
• (ii) Sale of assets.
• Convergence when no further deleveraging occurs

42
IllustrationX percent loss on foreign assets
leads to proportional reduction of Z percent of
domestic and foreign assets to restore
CAPITAL / ASSETS 0.08
43
Simulated data
44
Conclusions

• Formal quantitative tools help
• Identify vulnerabilities that might be missed in
  the qualitative/heuristic approach
• Quantify the likelihood, impact, and contagion of
  crises
• Explore scenarios that might be suggested by the
  qualitative approach
• And thus help discipline and inform judgment