Transcript THE IMPACT OF EMU ON REAL EXCHANGE RATE VOLATILITY …

EARLY WARNING SYSTEMS
FOR BANKING CRISES
Course on Financial Instability at the Estonian Central Bank,
9-11 December 2009 – Lecture 7
E Philip Davis
NIESR and Brunel University
West London
[email protected]
www.ephilipdavis.com
groups.yahoo.com/group/financial_stability
Introduction
• 3 types of models for early warning, logit,
signal extraction and binary recursive tree
• We apply the models first to prediction of
crises in Asia
• And then outline a new logit approach
which predicts banking crises in OECD
countries
Early warning systems
• Multivariate logit model uses macroeconomic,
institutional and financial variables X as inputs to
calculate probability of a banking crisis Y as the
output via logistic function estimator. Suitable for
answering question “what is the likelihood of a
banking crisis occurring in the next t years?”
Pr obYit  1  F X it  
e  'Xit
1  e  'Xit
• Non-parametric signal extraction approach
tracks individual time series X prior to and
during crisis episodes to answer question “is
there a signal S of future crisis or not?” If an
input variable’s aberrant behaviour can be
quantitatively defined whenever that variable
moves from tranquil to abnormal activity, a crisis
is forewarned.
• { S ij = 1 } = { │ Xij │ > │ X*ij │ } or
• { S ij = 0 } = { │ Xij │ < │ X*ij │ }
• Binary Recursive Tree (BRT) can be used to
answer question “which non-linear variable
interactions make an economy more vulnerable to
crisis than others?” Argued that liquidity, credit
and market risks are all potentially non-linear.
Estimator identifies single most important
discriminator between crisis and non-crisis
episodes across the entire sample, thereby creating
two nodes. Nodes are further split into sub-nodes
based on the behaviour of splitter variables’ nonlinear interactions with previous splitter variables.
This generates nodal crisis probabilities and the
associated splitter threshold values.
Entire Sample: 72
crises
Figure 4: Schematic Diagram
of Binary Recursive Tree
(BRT)
PARENT NODE
X1≤ V1*
X1>V1*
Splitter Variable: X1
Child Node 1:
52 crises
Child Node 2:
20 crises
Splitter Variable: X2
X2≤ V2*
Terminal Node 3:
48 crises
Splitter Variable: X3
X2> V2*
Terminal Node 3:
4 crises
X3≤ V3*
Terminal Node 4:
17 crises
X3≥ V3*
Terminal Node 5:
3 crises
Advantages and disadvantages
• Logistic models are ideally suited to predicting a binary
outcome (1 = banking crisis, 0 = no banking crisis) using
multiple explanatory variables selected on the basis of their
theoretical or observed associations with banking crises.
• Logistic approach is also parametric, generating confidence
intervals attached to coefficient values and their
significance, but logit coefficients are not intuitive to
interpret and they do not reflect the threshold effects that
may be simultaneously exerted by other variables.
• Signal extraction non parametric and can use high
frequency data
• Logit approach is the most appropriate for use as a global
EWS, while signal extraction methods are more appropriate
for a country-specific EWS (Davis and Karim 2008).
• BRT is able to discover non-linear variable interactions,
making it especially applicable to large banking crises
datasets where many cross-sections are necessary to
generate enough banking crisis observations and
numerous factors determine the occurrence of systemic
failure.
• In BRT no specific statistical distribution needs be
imposed on the explanatory variables. Also not
necessary to assume all variables follow identical
distributions or that each variable adopts the same
distribution across cross-sections.
• Although logistic regression does not require variables
to follow any specific distribution, Davis and Karim
(2008) showed that standardising variables displaying
heterogeneity across countries improved the predictive
performance of logit models.
• Logistic regressions are also sensitive to outlier effects,
yet it is precisely the non-linear threshold effects exerted
by some variables that could generate anomalous values
in the data.
• In low risk, stable regimes, variables may conform to a
particular distribution which subsequently jumps to a
regime of financial instability. Non-parametric BRTs
should handle such data patterns better than logistic
regressions.
• BRT is extremely intuitive to interpret. The model
output is represented as a tree which is successively split
at the threshold values of variables that are deemed as
important contributors to banking crises.
• Signal extraction is also easier to interpret than logit, but
is vulnerable to ignoring multivariate patterns at core of
instability
Illustrative results – logit (Asia)
Variable
Coefficient
z-Statistic
Coefficient
z-Statistic
DCRED(-1)
GDPPC(-1)
FISCY(-1)
INFL(-1)
RIR(-1)
DEPREC(-1)
DCREDY(-1)
DTT(-1)
DGDP(-1)
M2RES(-1)
-0.033902
-0.000246
0.010451
-0.037791
0.114829
0.053493
0.022844
0.007492
-0.261366
-0.000549
-2.091298
-3.451172
0.153806
-1.212934
2.528462
2.724725
2.971898
0.322193
-3.853235
-2.232728
-0.032416
-0.000235
-2.046609
-3.535303
0.113567
0.044323
0.021231
2.612414
2.712526
2.959820
-0.276748
-0.000536
-4.192324
-2.190088
Expectation-Prediction Evaluation for Binary
Specification
Equation: IND_STAND
Date: 12/03/09 Time: 19:46
Success cutoff: C = 0.25
Estimated Equation
Dep=0
Dep=1
Total
P(Dep=1)<=C
P(Dep=1)>C
Total
Correct
% Correct
% Incorrect
Total Gain*
Percent Gain**
80
34
114
80
70.18
29.82
70.18
70.18
7
41
48
41
85.42
14.58
-14.58
NA
87
75
162
121
74.69
25.31
45.06
64.04
Signal extraction - Asia
2
1.8
1.6
NTSR
1.4
1.2
1
0.8
0.6
0.4
0.2
0
0.5
1
2
3.5
4
4.5
6
8
10
Percentile Threshold
GDP grow th
Change Terms of Trade
Depreciation
Real Interest Rate
Inflation
Fiscal Surplus/ GDP
M2/Reserves
GDP per Capita
Real GDP growth Fiscal surplus/GDP Depreciation
% crises correct
10
8
6
% no crises correct
99
98
98
% total correct
65
64
62
20
BRT - Asia
Node 1
Class Cases %
0
120 71.4
1
48 28.6
FISCY > -1.14
FISCY <= -1.14
Terminal
Node 4
Class Cases %
0
79 89.8
1
9 10.2
Node 2
Class Cases %
0
41 51.3
1
39 48.8
DGDP <= 4.75
Terminal
Node 1
Class Cases %
0
7 20.6
1
27 79.4
DGDP >
4.75
Node 3
Class Cases %
0
34 73.9
1
12 26.1
DCREDY <= 60.49
DCREDY > 60.49
Terminal
Node 2
Class Cases %
0
18 100.0
1
0
0.0
Terminal
Node 3
Class Cases %
0
16 57.1
1
12 42.9
Asia
% crises correct
46
% no crises correct 90
% total correct
84
Leading indicator selection
Asia
Logit
Real GDP Growth
Real Interest Rate
Inflation
Fiscal Surplus/ GDP
M2/ Foreign Exchange
Reserves
Real Domestic Credit Growth
Real GDP per capita
Domestic credit/GDP
Depreciation
Terms of Trade
Current account/GDP
External short term debt/GDP







Signal
Extraction
Tree







A new model for the OECD
• Existing work on early warning systems (EWS)
for banking crises generally omits bank capital,
bank liquidity and property prices, despite their
relevance to the probability of crisis in the mind of
bankers, policymakers and the public. One reason
for this neglect is that most work on EWS to date
has been for heterogeneous global samples
dominated by emerging market crises. For such
countries, time series data on bank capital
adequacy and property prices are typically absent,
while other variables affecting crises may also
differ in OECD countries.
• We argue results are misspecified
• Triggers of crisis depend on the type of economy and
banking system. In OECD countries with high levels of
banking intermediation and developed financial markets,
shocks to terms of trade are less important crisis triggers
than, say, property price bubbles.
• Also developed economy banking systems are more
likely to be regulated in terms of capital adequacy and
liquidity ratios
• Accordingly, we estimate logit models of crisis for
OECD countries only and find strong effects of capital
adequacy, liquidity ratios and property prices, such as to
exclude most traditional variables. Our results imply that
higher unweighted capital adequacy as well as liquidity
ratios has a marked effect on the probability of a banking
crisis, implying long run benefits to offset some of the
costs that such regulations may impose (e.g. widening of
bank spreads).
Methodology and data
• Multivariate logit with dependent variable being crisis
probability
• Problems of crisis dummies
– Definition of banking crises
– Start and end dates ambiguous
– Focus on switch date in core results
• Data partitioned to 1980-2006 and 2007 to leave subprime
crisis for out-of-sample
• Variables for bank regulation:
– Unweighted capital adequacy ratio - ratio of capital and reserves
for all banks to the end of year total assets
– Liquidity - ratio of the sum of cash and balances with central
banks and securities for all banks over the end of year total assets
Table of crises in sample
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
BG
CN
DK
FN
FR
GE
IT
JP
NL
NW
SP
SD
UK
US
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
1
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
Box 1: List of Variables (with variable key)
Variables used in
previous studies:
Demirguc-Kunt and
Detragiache (2005);
Davis and Karim (2008).
Variables introduced in
this study.
1. Real GDP Growth (%) (YG)
2. Real Interest Rate (%) (RIR)
3. Inflation (%) (INFL)
4. Fiscal Surplus/ GDP (%) (BB)
5. M2/ Foreign Exchange Reserves (%) (M2RES)
6. Real Domestic Credit Growth (%) (DCG)
7. Liquidity ratio (%) (LIQ)
8. Unweighted capital adequacy ratio (%) (LEV)
9. Real Property Price Growth (%) (RHPG)
Pr ob Yit  1  F X it  
n
T
e  'Xit
1 e
 'Xit
Loge L   Yit loge F  ' X it   1  Yit  loge 1  F  ' X it 
i 1 t 1
Table 2: The General To Specific Approach
LIQ(-1)
LEV(-1)
RHPG(-3)
-0.118
(-3.55)
-0.333
(-2.85)
0.113
(2.8)
-0.124 -0.137
(-3.55) (-3.64)
-0.239 -0.315
(-1.90) (-2.24)
0.113
0.104
(2.87) (2.67)
-0.099 -0.10
(-1.82) (-1.97)
0.084
(1.37)
-0.135
(-3.55)
-0.247
(-1.64)
0.100
(2.59)
-0.10
(-1.86)
0.085
(1.40)
-0.00
(-1.0)
-0.135
(-3.45)
-0.271
(-1.67)
0.104
(2.67)
-0.10
(-1.99)
0.165
(1.41)
-0.00
(-1.0)
-0.13
(-0.8)
DCG(-1)
-
RIR(-1)
-
M2RES(-1)
-
-
-
INFL(-1)
-
-
-
-
YG(-1)
-
-
-
-
-
BB(-1)
-
-
-
-
-
-0.144
(-3.39)
-0.280
(-1.72)
0.108
(2.76)
-0.13
(-1.98)
0.173
(1.46)
-0.00
(-1.1)
-0.14
(-0.8)
0.116
(0.65)
-
-0.147
(-3.25)
-0.273
(-1.62)
0.110
(2.67)
-0.13
(-1.98)
0.166
(1.30)
-0.00
(-1.1)
-0.13
(-0.7)
0.125
(0.66)
-0.013
(-0.1)
Note: estimation period 1980-2006; t-statistics in parentheses; LIQ-liquidity ratio, LEV- unweighted capital
adequacy ratio, YG-real GDP growth, RPHG-real house price inflation, BB-budget balance to GDP ratio,
DCG-domestic credit growth, M2RES-M2 to reserves ratio, RIR-real interest rates, DEP-depreciation, INFLinflation.
Table 3: Comparing the Effects of Sample Period on Estimation Results
Estimation period
1980-2006 1980-2007
LIQ
LEV
PHG
-0.118
(-3.55)
-0.333
(-2.85)
0.113
(2.8)
-0.13
(-4.1)
-0.261
(-2.51)
0.106
(2.79)
 p(crisis) 
log  1 - p(crisis)  = - 0.333 LEV(-1) – 0.118 LIQ(-1) + 0.113 RHPG(-3)


(-2.85)
(-3.55)
(2.8)
Marginal effect of 1% rise in
variable on crisis probability
BG
CN
DK
FN
FR
GE
IT
JP
NL
NW
SD
SP
UK
US
LIQ
-0.17
-0.22
-0.05
-0.23
-0.78
-0.23
-0.17
-0.38
-0.56
-0.33
-0.12
-0.08
-1.19
-0.08
LEV
-0.49
-0.61
-0.14
-0.65
-2.17
-0.65
-0.46
-1.05
-1.57
-0.91
-0.34
-0.24
-3.32
-0.22
RHPG
0.17
0.21
0.05
0.22
0.74
0.22
0.16
0.36
0.53
0.31
0.12
0.08
1.13
0.07
BG
BG 80
BG 90
C 00
N
C 83
N
C 93
N
D 03
K
D 86
K
D 96
K
FN 06
FN 8 9
FR 9 9
FR 8 2
FR 9 2
G 02
E
G 85
E
G 95
E
-0
IT 5
-8
IT 8
JP 98
JP 81
JP 91
N 01
L
N 84
L
N 94
L
N - 04
W
N - 87
W
SD 97
SD 80
SD 90
SP 00
SP 83
SP 93
U 03
K
U 86
K
U 96
K
U 06
S
U 89
S
-9
9
Crisis probabilities
1.00
0.80
0.60
0.40
0.20
0.00
Probability
Crisis
In sample prediction
Total
Calls
Aftermath
Crises
of the
Crises
False
Calls
Timing of False Calls relative to Crisis Onset
BG
0
0
0
0
CN
6
1
1
4
DK
0
0
0
0
FN
10
1
1
8
FR
14
1
0
13
GE
4
0
0
4
IT
7
0
2
5
2nd and 3rd years
JP
15
1
6
8
Next 7 years, with a break on the 4th year
NL
18
0
0
18
NW
14
1
2
11
next 2 years
SD
6
1
1
4
next year
SP
2
0
0
2
UK
20
2
0
18
US
0
0
0
0
total
116
8
13
95
next year
next year
Out of sample predictions
BG
CN
DK
FN
FR
GE
IT
JP
NL
NW
SD
SP
UK
US
2007
X
X
X
X
X
X
X
-
2008
X
X
X
X
X
X
-
definition1 definition2
X
X
-
X
-
X
X
X
X
X
-
X
X
-
Country elimination tests
Final
panel
LIQ(-1)
LEV(-1)
PHG(-3)
-0.118
(-3.55)
-0.333
(-2.85)
0.113
(2.8)
US and
Norway
UK not
US not Japan not
Japan not
not
included included included
included included
-0.143
(-2.99)
-0.3
(-1.78)
0.152
(3.44)
-0.125
(-3.55)
-0.339
(-2.79)
0.119
(2.82)
-0.111
(-3.28)
-0.344
(-2.94)
0.111
(2.74)
-0.119
(-3.29)
-0.349
(-2.86)
0.118
(2.76)
-0.124
(-3.59)
-0.282
(-2.38)
0.089
(2.04)
Finland
not
included
Sweden
not
included
-0.121
(-3.5)
-0.293
(-2.43)
0.083
(1.84)
-0.115
(-3.41)
-0.343
(-2.87)
0.107
(2.58)
Alternative crisis dates
Final
version
LIQ(-1)
LEV(-1)
PHG(-3)
-0.118
(-3.55)
-0.333
(-2.85)
0.113
(2.8)
Japanese
US crisis
crisis at
at 1984
1992
-0.119
(-3.56)
-0.332
(-2.85)
0.113
(2.8)
-0.12
(-3.58)
-0.317
(-2.73)
0.104
(2.56)
Aftermath elimination and
subprime runup
Final
version
LIQ(-1)
LEV(-1)
PHG(-3)
-0.118
(-3.55)
-0.333
(-2.85)
0.113
(2.8)
Aftermath
of the
Crisis
-0.111
(-3.48)
-0.329
(-2.91)
0.111
(2.74)
Final
version
LIQ(-1)
LEV(-1)
RHPG(-3)
-0.118
(-3.55)
-0.333
(-2.85)
0.113
(2.8)
LIQ(-1)b
-
LEV(-1)b
-
RHPG(-3)b
-
1980-2007
estimation
with break
-0.128
(-3.4)
-0.241
(-1.94)
0.106
(2.85)
-0.029
(-0.34)
-0.045
(-0.19)
0.006
(0.05)
Further lags and systemic crises
LIQ (-2)
LEV (-2)
PHG (-3)
-0.104
(-3.27)
-0.385
(-3.22)
0.119
(3.00)
LIQ (-1)
LEV (-1)
PHG (-3)
-0.121
(-2.49)
-0.768
(-3.59)
0.235
(3.71)
Conclusions
• 3 approaches complementary
• Traditional approaches fruitful for EMEs such as Asia but
not for OECD countries
• Found relevance of bank capital, liquidity and property
prices absent from traditional EWS, exclude traditional
variables
• Can predict crises out of sample and specification is robust
• Warrants policy focus on bank regulation – of capital,
liquidity but also of terms of mortgages loans
• Also supports measures to reduce procyclicality, adjusting
capital or provisions countercyclically – and use of simple
leverage ratio as well as risk weighted capital adequacy
References
• Davis, E P and D Karim (2008a), "Comparing early
warning systems for banking crises", Journal of Financial
Stability, 4, 89-120
• Davis E P and Karim D (2008b), "Could early warnings
systems have helped to predict the subprime crisis?",
National Institute Economic Review, 206, 25-37 and
Brunel University Economics and Finance Working
Paper No 08-27
• Barrell R, Davis E P, Karim D and Liadze I (2009),
"Bank Regulation, Property Prices And Early Warning
Systems For Banking Crises In OECD Countries",
NIESR Discussion Paper No. 330