L10-B- Macro-Stress Test - Systemic RIsk

Report
DEVELOPING MACRO-STRESS
TESTS: HOW TO IDENTIFY
SYSTEMIC RISK
SESSION 10
1
MINDAUGAS LEIKA
MACROPRUDENTIAL
POLICY FRAMEWORK
I. Macroprudential policy definition, targets, policy
transmission channels and relationships with other policies
(Monday)
II. Institutional structure (Tuesday)
III. Policy tools (Tuesday)
2
IV. Risk identification and quantification: stress testing (This
lecture)
AGENDA
How to identify systemic risk: role of stress testing
3
Advances in systemic risk research and monitoring
WHERE IS STRESS TESTING? (I)
Financial
stability
analysis
• Risk identification and scenario design
• Financial stability review
• Regional Financial Stability review
• Stress testing
• Early warning indicators
Risk
mitigation
policies
• Focus on more risky
banks and financial
institutions
• Dialog with financial
institutions
• Macroprudential
instruments
4
Data
Availability
• National accounts data
• Property prices data
• Data on cross border exposures
• Supervisory data: loan write offs, defaults
• Financial soundness indicators
• Meetings with financial institutions, surveys etc.
WHERE IS STRESS TESTING? (II)
Macro Stress testing
Systemic Risk
quantification
Macroprudential
policy tools
5
Systemic Risk
identification
DESCRIPTION VS.
PRESCRIPTION
6
Source: Christensen C., Carlile P. (2009)
STRESS TESTING CHALLENGES
-Data availability
-Data complexity
-Data relevancy
-Model uncertainty
7
-Scenario uncertainty
WHAT IS SYSTEMIC RISK?
Imbalances
Correlated exposures
Excessive exposures
Spillovers to the real economy
Disruption of financial transactions, flow of funds
Capital flight
Bubbles, contagion
Bank closures
8
Various studies offer different measures of risk. However, what is relevant
in one country might not be relevant in another one, and this is especially
important in Africa: various advanced tools might be too premature to be employed.
Many studies provide a good description of what happened in the past,
but can they predict the future?
IDENTIFICATION: METRICS
A. Lo at all (2012) provides an extensive list of various
systemic risk measures and data requirements:
-Macroeconomic: asset price bubbles, credit, investment
cycles; macroprudential regulation
-Granular and Network measures: default intensity, network
analysis and systemic financial linkages; scenario analysis;
risk and shock transmission; mark-to-market accounting and
liquidity pricing;
-Forward looking risk measures: Contingent claim analysis,
multivariate density estimation, housing sector simulations,
principal component analysis etc.
-Stress test measures: macro stress tests
-Cross-Sectional Measures: CoVaR, expected shortfall etc.
9
-Illiquidity and insolvency: risk topography, leverage cycle,
crowded trades, hedge fund trades etc.
IDENTIFICATION: DATA (1)
Asset price boom/bust cycles:
-National statistics and/or international data sources: IMF
International Financial Statistics
-Macroeconomic indicators: national statistics office, national
accounts data (e.g. financial accounts)
Bank funding risk and shock transmission:
-BIS locational banking statistics;
-Bank surveys (on financing terms); deposit statistics (interest
rates);
-Refinancing schedule (timing of bonds to be redeemed, loans
repaid etc.);
10
-Data on covenants (e.g. repayment obligation link with credit
rating).
Identification: data (2)
Corporate and consumer indebtedness:
- Banks’ proprietary data; credit register (both positive and
negative ones);
- National accounts statistics: corporate and household leverage
ratios;
Contingent claim analysis:
- Moody’s KMV, Bloomberg, MarkIt; CDS spreads;
CoVaR, currency trades, insurance premiums, etc.
11
- Market data, various service providers: Bollomberg, Reuters
Datastream, Moody’s, S&Ps etc. However, in many cases this is
not very relevant for African countries (dominated by plain vanilla
banks), as many instruments are not available in local markets.
RECOMMENDATIONS
Avoid “Sophistication trap”, i.e. the more sophisticated
analysis is, the more credibility it has. This is wrong
approach, especially in countries where simple financial
systems dominate. Avoid sophisticated models, if data is
not reliable and even not available.
Concentrate on data availability and timing issues: simple
ratios can tell much more, than distance to default or first
principal component.
12
Try to build simple models and for systemic risk, create
network of individual exposures, observe common
patterns.
EXAMPLE: FROM INDIVIDUAL
BANK RUN TOWARDS SYSTEMIC
CRISIS
13
INDIVIDUAL BANK RUN (1)
As soon as even a small fraction of deposits is channeled
towards illiquid assets (long-term loans, investments etc.) every
bank won’t be able to meet its obligations towards creditors
(depositors), hence there is always a risk of liquidity crisis.
14
 Imagine, that we have a three period game: deposit D is placed
at t=0 and can be withdrawn at t=1 (impatient depositors) and t=2
(patient depositors). Bank invest deposits into long term assets
that yield rate of return R. R>1 at time period t=2, however R<1 at
time period t=1. Hence, if bank needs to liquidate investment
before it matures, it makes a loss (R<1) (e.g. fire-sale of assets). If
bank knows the proportion of depositors  who are impatient and
withdraw at time t=1, it invests D share of deposits into liquid
assets and (1- )D share of deposits into less liquid (long-term)
assets.
INDIVIDUAL BANK RUN (2)
This system is called a Fractional Reserve Banking
system. In practice, banks need to make their own liquidity
forecasts to obtain , or stick to minimum reserve and
liquidity requirements imposed by regulators.
This system works well, if  is stable, however is  is
higher than anticipated, the bank will have to liquidate
higher proportion of assets under depressed prices and
suffer loss. If this happens, patient depositors realize, that
they won’t be able to get back their deposits at time t=2 in
full amount. Bank run in this case is a rational strategy;
depositors’ expectations are self-fulfilling.
15
Of course, there is a case when R is much higher than 1,
so higher  might not necessarily lead to bank liquidation.
INDIVIDUAL BANK RUN (3)
16
These assumptions were incorporated in the classical DiamondDybvig framework.
If depositors type is observed, it is possible to prohibit early
withdrawal (contract design or by law); otherwise banks need to
rely on historical observations. However a lack of confidence in a
given bank would lead to a coordination failure among depositors.
Hence, the bank needs to obtain liquidity in the market and/or
from a central bank or liquidate investments.
If interbank market is functional and liquidity needs are bank
specific rather than system, bank might be able to obtain liquidity
by borrowing. It is also possible, that the bank simply securitizes
its long-term assets (the case of Northern Rock).
Due to inter-linkages among banks, other financial institutions,
payment and settlement systems etc. individual bank run might
lead to the general loss of confidence in the banking system and
as a result – systemic crisis.
Alternative solutions: suspension of convertibility (also “bank
holidays”), deposit insurance, debt-equity swap.
SYSTEMIC CRISIS (1)
Due to inter-linkages among banks, other financial
institutions, payment and settlement systems etc.
individual bank run might lead to the general loss of
confidence in the banking system and as a result –
systemic crisis.
Systemic crisis can be triggered via system of crossholdings among banks.
17
Market topology is important: a system where each
bank borrows from one or several banks is more fragile
than a more diversified interbank system.
SYSTEMIC CRISIS (2)
Bank A
Bank A
Bank B
Bank C
Incomplete structure: more
risky to contagion effects
Bank D
Bank B
Bank C
Complete structure: less
risky to contagion effects
18
Bank E
ORIGINS OF LIQUIDITY SHOCKS
Credit losses
Operational losses
Change in investors expectations
Rating downgrade
Liquidity ratio below
threshold
Negative cash flow
19
Market events
Cross-border contagion
Loss in confidence
Speculations (rumors)
Interbank contagion
VICIOUS CIRCLE
1. Run on a bank
4. Assets value, market
prices
2. Liquidity crisis
20
3. Interbank interest
rates, Real interest
rate 
Source: Bech M. et all (2009)
21
STRESS IN THE FEDWIRE
ADVANCES IN SYSTEMIC RISK
RESEARCH AND MONITORING
22
FROM INDIVIDUAL RISK TO
SYSTEMIC RISK
Bank
C
Systemic risk
Most of the studies
before the GFC focused
on individual or firm
level risks: VaR,
portfolio concentration,
diversification, credit
ratings etc. There were
studies on joint
probability of default,
but not on bank, but
firm level
SR studies flourished during this crisis: CoVaR,
measures of interconnectedness, spillover effects, risk transmission,
deleveraging etc.
23
Bank
A
Bank
B
STUDIES (1)
IMF Cross-Border contagion model. Utilizes BIS statistics and
uses Input-Output type of matrices of cross-border
exposures.
http://www.imf.org/external/np/pp/eng/2010/090110.pdf
Adrian, Brunnermeier (2009) CoVaR. Co means: conditional,
contagion, comovement. It is VaR of the whole financial
sector conditional on institution i being in distress.
http://www.princeton.edu/~markus/research/papers/CoVaR.pd
f
24
Information cascades and Big data analytics. Abreu,
Brunnermeier (2003).
http://www.princeton.edu/~markus/research/papers/bubbles_
crashes.pdf
STUDIES (2)
Goodhart, Segoviano (2009). Banking Stability Measures.
Measures distress dependence among banks.
http://www.imf.org/external/pubs/ft/wp/2009/wp0904.pdf
25
Abbe, Khandani, Lo (2011) Privacy-Preserving Methods for
Sharing Financial Risk Exposures. Data collection: how to
obtain aggregate number without disclosing individual
exposure? http://bigdata.csail.mit.edu/node/23
EXAMPLE: REAL ESTATE
LINKED LOANS (1)
800
1,600
40000
Aggregate
700
1,400
1,200
JPMorgan
30000
Wells Fargo
500
1,000
Aggregate Value
A
Indiividual Bank Value
600
Bank of America
JP Morgan to Ba
Wells Fargo to B
35000
Bank of America
25000
400
800
300
600
200
400
10000
100
200
5000
20000
15000
Jun 92
Jun 91
Jun 90
Jun 89
Jun 88
Jun 87
Jun 86
Jun 09
Jun 10
Jun 07
Jun 08
Jun 04
Jun 05
Jun 06
Jun 03
Jun 01
Jun 02
Jun 99
Jun 00
Jun 96
Jun 97
Jun 98
Jun 95
Jun 93
Jun 94
Jun 90
Jun 91
Jun 92
Jun 88
Jun 89
Jun 87
Jun 86
0
Source: Abbe, Khandani, Lo (2011)
26
(a)
EXAMPLE: REAL ESTATE
LINKED LOANS (2)
40000
Bank of America to JP Morgan
JP Morgan to Bank of America
Wells Fargo to Bank of America
35000
Bank of America to Wells Fargo
JP Morgan to Wells Fargo
Wells Fargo to JP Morgan
25000
20000
15000
10000
5000
Jun 09
Jun 08
Jun 07
Jun 06
Jun 05
Jun 04
Jun 03
Jun 02
Jun 01
Jun 00
Jun 10
27
Source: Abbe, Khandani, Lo (2011)
Jun 99
Jun 98
Jun 97
Jun 96
Jun 95
Jun 94
Jun 93
Jun 92
Jun 91
Jun 90
Jun 89
Jun 88
Jun 87
0
Jun 86
Aggregate Value
A
30000
Source: Billo, Getmansky, Lo, Pelizzon (2012). Measuring Systemic Risk
in the Finance and Insurance Sectors
28
MEASURES OF INTERCONNECTEDNESS: HOW
THE WORLD BECAME MORE AND MORE
INTERCONNECTED

similar documents