Transcript Slide 1

Stress Testing and Liquidity Risk
Management
Antonio García Pascual
Federico Galizia
Monetary and Capital Markets Department
International Monetary Fund
Quant Congress USA
New York, July 8-9, 2008
The views expressed in this presentation are those of the authors and do not necessarily
represent the views of the IMF or IMF policy
1
Motivation: market induced credit risk …
SIVs & subprime
Forex mortgages
•
•
•
SIVs may have
managed duration
risk ...
but overlooked
market-induced
credit risk (i.e. higher
defaults because of
higher interest rates)
•
Popular in Italy pre1992 (ECU) and in
CEEC currently (CHF)
Devaluation makes
debt service more
expensive while also
hurting the economy
(and jobs)
… can result in loss of market liquidity for
instruments and funding liquidity for institutions
2
Outline
•
•
•
•
Design of adequate stress tests and
scenario analysis
Modeling linkages: N-th to default
CDS basket
Linkages between credit, market and
liquidity risks
Strengthening liquidity management
3
Macro-scenario analysis can help risk management
Definition of
macro
scenario
Stressed
PDs
Macroeconometric
model
Measuring the
impact:
Measuring the
impact:
2nd round
effects,
feedbacks
Exposures
Stressed
&
LGDs Correlations
Algorithm /
Monte Carlo
Logit,
dynamic panel ...
Probability
Portfolio Loss Distribution for “Bank X”
under stressed scenario
99.9% VaR
EL = PDxLGD
Unexp. Loss
Losses
4
Calibrating a severe macro scenario (“1 in 20 years”)
Country X. Macro Scenario: "Sudden Stop"
Real GDP growth
Baseline
Stress Scenario
Average inflation
Baseline
Stress Scenario
Interest rates (policy rate)
Baseline
Stress Scenario
Credit growth
Baseline
Stress Scenario
Real Estate Prices (growth)
Baseline
Stress Scenario
2008
2009
2010
4.5
4.0
4.0
-0.5
4.0
1.0
7.5
7.0
6.0
5.0
5.0
4.0
5.0
6.0
4.8
8.0
4.5
6.0
25.0
10.0
20.0
5.0
15.0
4.0
-10.0
--15.0
--5.0
Forecast horizon: typically 2-3 years, in line with:
• The forecasting horizon of the macro model
• Tracing the lagged impact on the portfolio
• The medium-term strategic planning of the institution
5
Measuring the impact of a macro scenario
From supervisory agencies: A few examples of models linking
macro scenarios to bank-specific risk factors
Bundesbank
•
Dynamic panel for loan loss provisions:
LLP = f [credit growth, GDP growth, change in interest rates]
Banco de España
•
•
Corporates, dynamic panel for default rates:
DR = f [GDP growth, credit growth, interest rates]
Mortgages, dynamic panel for default rates:
DR = f[unemployment , interest rates]
LGD = f[housing prices and changes in interest rates]
Banque de France & Comission Bancaire
•
•
Logit model for default rates: PD = f [GDP, long and short interest rates]
Regression model for Net interest income:
NIC = f [interest rate, private demand, average PD]
Banca d’Italia
•
•
•
Preliminary PCA analysis on macro variables to reduce dimensionality
Dynamic panel for default rates:
DR = f [output gap, inflation , interest rate)
Regression for operating profits:
OP = f [economic activity, equity return, change in interest rates]
6
Outline
•
•
•
•
Design of adequate stress tests and
scenario analysis
Modeling linkages: N-th to default
CDS basket
Linkages between credit, market and
liquidity risks
Strengthening liquidity management
7
From macro to market to credit risk
An alternative to direct econometric modeling of PDs is a MtM
model à la Merton
Macro Scenario
(systemic risk)
Sectors Asset values
Idiosyncratic
risks
Portfolio structure:
•
Industry
•
Geography
•
Size (concentrations)
Loss Distribution
“Downturn
LGD”
Stressed
EDFs
(99% VaR)
Monte Carlo PLD
99 percentile
Expected
Loss
Unexp. Loss
Losses
8
From macro to market to credit risk: NTD CDS basket
(Avesani, Garcia Pascual, Li, 2006)
•
Portfolio of N different companies each associated with a given
CDS spread and a recovery rate.
•
Correlation of defaults is driven by m common factors and
creditworthiness for each company i (i=1,…,N) depends on its
asset value xi :
xi  ai1M1  ai 2 M 2  ...  aim M m  Zi 1  ai21  ai22  ...  aim2

The conditional default
probability of company i :

The risk-neutral PD that
company i defaults before
time t (forward default
hazard rate λi) :

Under a copula model, the
xi are mapped to ti (time
of default) :
 x  (a M  ...  a M ) 
i1
1
im
m

prob( xi  xi |  )  H i  i
2
2


1  a i1 ...  a im
t
i ( u ) du

0
Qi  ti  t   1  e

xi  F 1 Qi t  
;
prob  xi  xi |    Qi t |  
9
CDS spreads: 15 Large complex financial institutions
80
"Higher risk" profile (LEH, MER, GS, MS, JPM)
70
60
50
Up until Jun 2007:
LEH
"Medium risk" profile (CS, DB, BoA, Citi)
• “Higher” risk profile: (LEH, MER, GS, MS)
• "Lower
“Medium”
risk profile
: (JPM,SG,
CS,
DB,BAR,
BoA, BNP
Citi) )
risk" profile
(UBS, HSBC,
ABN,
• “Lower” risk profile: (UBS HSBC, SG, ABN, BAR, BNP)
MER
GS
MS
JPM
CS
DB
40
BoA
30
Citi
UBS
20
HSBC
SG
10
ABN
BAR
Apr-07
Jan-07
Oct-06
Jul-06
Apr-06
Jan-06
Oct-05
Jul-05
Apr-05
Jan-05
Oct-04
Jul-04
Apr-04
Jan-04
Oct-03
Jul-03
Apr-03
Jan-03
0
BNP
10
Correlation of equity returns
SG
BNP
DB
ABN
SG
1.00 0.87 0.70 0.80
BNP
1.00 0.69 0.73
DB
1.00 0.67
ABN
1.00
HSBC
BARC
UBS
CS
BoA
CITI
JPM
LEH
ML
GS
MS
HSBC BARC UBS
0.65
0.62
0.55
0.68
1.00
0.66
0.67
0.61
0.68
0.78
1.00
0.73
0.67
0.65
0.75
0.68
0.73
1.00
CS
BoA
CITI
JPM
LEH
ML
GS
MS
0.67
0.59
0.64
0.63
0.63
0.62
0.81
1.00
0.01
0.04
0.27
0.05
-0.06
0.04
-0.01
0.09
1.00
0.08
0.11
0.24
0.11
0.16
0.16
0.04
0.13
0.76
1.00
0.06
0.08
0.29
0.13
0.06
0.09
0.08
0.05
0.75
0.66
1.00
0.04
0.04
0.21
0.06
0.06
0.15
0.05
0.07
0.61
0.51
0.53
1.00
0.06
0.09
0.36
0.08
0.15
0.15
0.09
0.10
0.71
0.67
0.68
0.71
1.00
0.05
0.04
0.30
0.01
0.02
0.07
0.03
0.13
0.59
0.50
0.53
0.82
0.67
1.00
0.08
0.13
0.33
0.04
0.16
0.16
0.00
0.14
0.62
0.67
0.46
0.64
0.75
0.64
1.00
11
NTD probability computed under stress
Baseline PDs
a
PD0
PD1
PD2
PDs when all factors in recession
PD3
1.00
0.80
Left scale
0.60
0.40
Right scale
0.20
0.00
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 2 0
0.45
0.40
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0.00
1.00
0.45
0.40
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0.00
0.80
0.60
0.40
0.20
0.00
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 2 0
Avesani, Garcia Pascual and Li (2006)
… can provide a measure of systemic counterparty
risk
12
Outline
•
•
•
•
Design of adequate stress tests and
scenario analysis
Modeling linkages: N-th to default
CDS basket
Linkages between credit, market
and liquidity risks
Strengthening liquidity management
13
Stress testing and counterparty risk: CDS pricing
Value of Protection Leg (a function of PD*)
S
" risky" DV01 (a function of PnD*)
PD
PD*
PnD
•
•
C
PnD*
Upon an adverse scenario, counterparty
risk “punches a hole” in the protection set …
… implying that protection is now overpriced
and value of CDS lower (Hull & White 2004)
14
Standard CDS valuation doesn’t work under stress
Non-defaulted states
(credit migration)
Upon default
Standard
approach
Use “standard”
formulas with forward
discount rates and
default probabilities
Assume value of the
CDS to be equal to
the LGD on the
reference obligation
Pitfalls
It overlooks impact of
counterpart risk on
valuation
(CDS value falls)
It overlooks potential
for double-default &
no-recovery events
15
Is it market, credit or liquidity risk?
I still trust
you … in ten
years!
What happened
to my 2-year
benchmark?
16
Outline
•
•
•
•
Design of adequate stress tests and
scenario analysis
Modeling linkages: N-th to default CDS
basket
Linkages between credit, market and
liquidity risks
Strengthening liquidity management
17
Policy recommendations on funding liquidity risk
A private sector
perspective on
governance and
organizational structure;
analytical framework for
measuring, monitoring,
and controlling liquidity;
and stress testing and
contingency planning
-- Stress testing practices for
Risk Mgmt and capital planning
-- Central bank operational
frameworks should be flexible =>
frequency and maturity of
operations, instruments, and the
range of counterparties and
collateral, to deal with
extraordinary situations
Principles for both
financial institutions and
the supervisory process
18
What can we do about market liquidity risk?
Market participants should act promptly to ensure that the
settlement, legal and operational infrastructure underlying
OTC derivatives markets is sound. Cash settlement of
obligations stemming from a credit event … Automate trade
novations; trade data submissions and the timeliness of
resolutions of trade matching errors for OTC derivatives … A
longer-term plan for a reliable operational infrastructure supporting
OTC derivatives
Federal Reserve Bank of New
York (June 9, 2008) Market
participants and regulators
agreed on … developing a central
counterparty for credit default
swaps that, with a robust risk
management regime, can help
reduce systemic risk
Current gaps in price discovery
mechanisms need to be addressed. The
standardization of securitized instruments
would help in this regard … A centralized
over-the-counter (OTC) registry that would
19
collect and distribute transaction data ...
Summary
•
Design of adequate stress tests and scenario
analysis
 Macroeconomists, risk managers and quants team up
•
Modeling linkages: N-th to default CDS basket
 Macro to market to credit risk (systemic counterparty risk)
•
Linkages between credit, market and liquidity risks
 Counterparty risk appears in stress scenarios (and reality)
•
Strengthening liquidity management
 Funding liquidity versus market liquidity
20
Stress Testing and Liquidity Risk
Management
Antonio García Pascual
[email protected]
Federico Galizia
[email protected]
The views expressed in this presentation are those of the authors and do not necessarily
represent the views of the IMF or IMF policy
21