Transcript Cross-Country Causes and Consequences of the Crisis Andrew K. Rose UC Berkeley, NBER and CEPR (based on work with Mark Spiegel) Causes and Consequences of.

Cross-Country Causes and
Consequences of the Crisis
Andrew K. Rose
UC Berkeley, NBER and CEPR
(based on work with Mark Spiegel)
Causes and Consequences of the Crisis
1
Financial Crisis has Renewed Calls for
Early Warning Systems
• “we agree … that the FSB [Financial Stability Board] should
collaborate with the IMF to provide early warning of
macroeconomic and financial risks and the actions needed to
address them”
– Final Communiqué G-20 Summit April 2, 2009
• “Any early warning system to detect impending dangers to the
world economy must find a way of bringing together the
scatter of international and national macrofinancial expertise.
We at the Fund have already begun intensifying our early
warning capabilities and will be strengthening our
collaboration with others involved in this area.”
– Dominique Strauss-Kahn
Causes and Consequences of the Crisis
2
Many calls …
• “An early warning system must be established to identify
upstream increases in risks...”
– Heads of State or Government of European Union, November 7 2008.
• “The Group recommends that the IMF … is put in charge of
developing and operating a financial stability early warning
system … to deliver clear messages to policy makers and to
recommend pre-emptive policy responses ...”
– De Larosiere Report, February 25 2009.
Causes and Consequences of the Crisis
3
But are Early Warning Models Likely to
have been able to predict this crisis?
• Models typically do better at predicting crosscountry incidence of crisis than timing
• We therefore examine ability of early warning
systems (EWS) to predict relative performance
across countries in the current crisis
• Also interested in examining which causes of
the financial crisis perform well as predictors
of relative future performance
Causes and Consequences of the Crisis
4
Success in Cross-Section is
Necessary but far from Sufficient!
In case of “Great Recession” we know lots of
inadmissible information (for EWS):
1.
2.
3.
4.
There was a crisis (EWS: many false signals)
The timing of the crisis (ditto)
The countries involved
Crisis involved banks, security markets, (at least some)
housing markets, and (at least some) exchange rates
– IMF: each of these is different
• Mix probably not comparable/predictable from
history!
Causes and Consequences of the Crisis
5
Non-Structural Methodology
• MIMIC (multiple-indicator, multiple cause)
model explicitly incorporate difficulties in
observing relative crisis severity
– Treats crisis as a latent variable
– Observed with error as function of observable
manifestations
– Also do OLS
• Sample is cross-section of 107 countries
Causes and Consequences of the Crisis
6
Crisis Performance is Latent Variable
Mapping Observable Causes to
Observable Manifestations
• 4 Manifestations: equity market collapse,
exchange rate devaluation, economic growth,
and change in creditworthiness
• Compare these to a large set of potential
causal variables
• Also obtain estimates of latent variable,
relative performance during financial crisis
Causes and Consequences of the Crisis
7
Results
• Plausible estimates of relative severity of crisis
– Ex: Iceland and Estonia identified as exceptionally bad
• Less success linking crisis to causal variables
– Very few variables consistently enter at statistically
significant levels
– Other equally-plausible variables, such as real estate
appreciation, fail to enter
• Conclusion: we can model the severity of the
crisis, but not its causes
Causes and Consequences of the Crisis
8
Results Bode Poorly for EWS
• Successful EWS must predict both incidence
and timing of crisis
• Our analysis fails even to predict incidence of
2008-09 crisis, should be (relatively) easy
• Potential reasons for poor performance
– Causes differ across countries
– Crisis might stem from a common “global shock”
• Both problematic for EWS!
Causes and Consequences of the Crisis
9
Potential Causes of Crisis
Causes and Consequences of the Crisis
10
Income included as Control
• Not obvious
– Crisis hit developed and EMEs alike
– Still, EMEs more risky and less potential for
counter-cyclical policies due to fiscal constraints
– Now widespread agreement of “progressive”
nature of Great Recession
Causes and Consequences of the Crisis
11
Often Include Size
• Size common in literature as crisis correlates
– Reinhart and Reinhart (2009), Calvo and Loo-Kung
(2009)
– Commonly believed small countries more exposed
(Ex.: Iceland)
– Small countries also more open
Causes and Consequences of the Crisis
12
Financial Policies
• Crisis widely perceived as demonstrating
regulatory flaws [e.g. Bernanke (2009)]
– Improper incentives within institutions
• Basel capital requirements
– Contributed to lending pro-cyclicality
– Codified role for rating agencies [Coval, et al (2009)]
– Encouraged opaqueness through securitization and
moving assets off balance sheets [Demirguc-Kunt and
Serven (2009)]
Causes and Consequences of the Crisis
13
Ex-Ante Measures of Financial Policy
Included as Potential “Causes”
• Economic Freedom of the World data base
– Bank Ownership: share of bank deposits in private banks
– Foreign Bank Competition: foreign bank license denial rates
– Interest Rate Controls/ Negative Real Interest Rate: credit
market controls
– Credit Market Regulation: quality of credit market regulation
• Barth, Caprio and Levine (2005) data
–
–
–
–
–
–
Overall Capital Stringency
Ability to Take Prompt Corrective Action
Capital Regulatory Index
Official Supervisory Power
Restructuring Power
Power to Declare Insolvency
Causes and Consequences of the Crisis
14
Financial Conditions
• Degree of maturity mismatch, due to shortterm debt obligations [e.g. Cecchetti (2008)]
• Risky lending practices
– Feldstein (2009): Appraised value of US mortgage
contracts grew from 70-80 % to 90-100 %
– Community Reinvestment Act encouraged lending
to broader set of borrowers [White (2008)]
• Exploding leverage of firms and households
Causes and Consequences of the Crisis
15
Measures of Financial Conditions
• Conditions products of policies (endogenous)
– Interested in prediction
• Measures included in specification:
–
–
–
–
–
–
–
Private Sector Domestic Credit as share of GDP
Domestic Bank Credit as share of GDP
Share of domestic credit consumed by Private Sector
Bank Liquid Reserves as a share of assets
Share of Non-Performing Loans
Bank Capital as a share of assets
Bank Claims as a share of deposits
Causes and Consequences of the Crisis
16
Asset Price Appreciation
• Widely cited as a source of fragility, in US and
elsewhere [e.g. Feldstein (2009), Teslik (2009)]
• Buiter (2009): Investment diverted to real estate from
“productive uses”
• Doms, et al (2007): US mortgage delinquencies highest
in areas that experienced highest amount of
appreciation
• Mian and Sufi (2008): U.S. areas with high latent
housing demand had highest decreases in denials
– Sub-prime lending fed appreciation
• Real estate appreciation mirrored in other financial
markets, particularly equity markets
Causes and Consequences of the Crisis
17
Measures of Real Estate and Equity
Price Appreciation
• Real estate
– Percentage Change in Real Estate Prices
• Based on data from the BIS and augmented by an Asiaspecific study by Glindro, et al (2008).
• Equity market appreciation
– Market Capitalization as a share of GDP
– Value of Stocks Traded relative to GDP
– Stock Market Growth.
Causes and Consequences of the Crisis
18
International Imbalances
• Many countries built up deficit international
financial positions over boom
• Arguments over source of “global imbalances”
– Asian and other country efforts to build up foreign
exchange reserves [e.g. Buiter (2007)]
– Lax policies in west [Buiter (2009)]
• Fratzscher (2009): Countries with high CA
deficits had larger exchange rate depreciations
– High exposure to U.S. also led to depreciation
Causes and Consequences of the Crisis
19
Measures of Imbalances and
International Reserve Adequacy
• Measures of the external balance position
–
–
–
–
–
Net External Position as % of GDP
Current Account as % of GDP
External Debt as % of GNI
Gross Financing in International Capital Markets as % of GDP
Real Effective Exchange Rate
• Measures of the adequacy of foreign reserve holdings
–
–
–
–
–
Total Reserves as % of external debt
Short-Term Debt as % of Reserves
Total Reserves over value of a Month of Imports
M2 as % of Total Reserves minus Gold
M2 as % of Central Bank Foreign Assets
Causes and Consequences of the Crisis
20
Macroeconomic Policies
• Easy monetary policy
– Taylor-rule indicated Fed Funds rate below levels
consistent with 2% inflation target 2003-2006
– “Greenspan put” exacerbated asset appreciation
[deLong (2009)]
• Lax fiscal policy
– Countries pursued unsustainable deficits
– Poor fiscal positions hindered counter-cyclical
policies during crisis [Buiter (2009)]
Causes and Consequences of the Crisis
21
Macro Policy Measures included
• Monetary policy
–
–
–
–
–
–
Currency Union dummy
Aggregate GDP of Monetary Zone
EU, but not EMU dummy
Inflation Targeter
M2 as % of GDP
M3 as % of GDP
• Fiscal policy
–
–
–
–
Government Budget Surplus/Deficit as % of GDP
Central Government Debt as % of GDP
Total Debt as % of GDP
Debt Service Burden as % of GDP.
• Macro conditions
– CPI inflation
– GDP growth
Causes and Consequences of the Crisis
22
Institutional Features
• Inferior institutional features associated with greater macro volatility [e.g.
Acemoglu, et al (2003)]
• Measures of institutional quality
–
–
–
–
–
–
–
–
–
–
–
–
–
EFW index of Credit/Labor/Business Regulation
Polity index
Constraints on the Executive (Polity data set)
Overall Economic Freedom
Common Law dummy
Control of Corruption index
Regulatory Quality index
Rule of Law index
Political Rights index
Civil Liberties index
Government Size
Security of Property Rights (EFW index)
Sound Money Access (EFW index)
Causes and Consequences of the Crisis
23
Geography
• Geographic features may have played a role in
relative performances
– Ex.: Iceland proximity to UK and Neth. enhanced
expansion of its financial sector
•
•
•
•
•
•
Geography controls
Log of Latitude
East Asian dummy
Central/Eastern Europe, Central Asian dummy
Commodity Exporter dummy
English Speaking dummy
Causes and Consequences of the Crisis
24
Empirics
Causes and Consequences of the Crisis
25
Methodology
• International cross section
– All information country-specific initially
– Then add linkages between countries (contagion)
– National characteristics may influence
vulnerability to foreign shocks (medical analogy)
Causes and Consequences of the Crisis
26
Empirical Strategy
• As crisis hit all types of countries, we include
both developed and EMEs
– All countries >$10,000 per capita GDP
– All countries >$4,000 per capita GDP, plus
population > 1 million
– Maximum of 107 countries in sample
• Use only publicly-available series available for
reasonable span of countries, time
– Necessary to be part of EWS
Causes and Consequences of the Crisis
27
Differences in Crisis Severity
(Manifestations/Consequences)
• Crisis intensity only observed with error [e.g. Berg, et al
(2004)]
– Abundant measurement error likely
• Initially model severity as a latent variable linked to
four observable indicators for 2008
1. GDP growth (%)
• Insensitive to using 2009 growth (collected in 3/09 or 9/09)
2. Change in National Equity Markets (stocks, %)
3. Change in multilateral SDR exchange rate (%)
4. Change in Institutional Investor credit rating
• March 2008 - March 2009
• Also use Euromoney index to check robustness
Causes and Consequences of the Crisis
28
2008 Consequences and Manifestations of the Crisis
Bottom
quartile
Upper
quartile
% Changes,
2008:
Iceland
Ukraine
Estonia
Argentina
Latvia
Ireland
Korea
New Zealand
UK
Hungary
Luxembourg
Denmark
Singapore
Swaziland
Finland
Japan
France
Netherlands
Thailand
Poland
Real GDP
II Rating
-4.7
2.1
-2.8
6
-4.6
-2.8
2.6
-.9
.7
.4
.6
-.9
1.2
2.7
1.4
-.5
.7
2
3
4.8
-32.5
-12.1
-9.4
-13.6
-8.3
-7.8
-7.3
-5.4
-5.5
-7.6
-2.6
-2.6
-3.8
-2.6
-2.6
-5.7
-2.6
-2.5
-3.5
-1.5
Causes and Consequences of the Crisis
Stock
Market
-90.0
-74.3
-63.0
-49.8
-55.1
-66.1
-40.7
-37.4
-31.5
-53.2
-59.5
-48.6
-48.9
3.9
-53.4
-42.1
-42.7
-52.3
-47.6
-51.1
Price of
SDR
90.0
48.6
1.7
6.9
-.3
3.1
30.9
30.4
33.9
6.1
3.1
1.5
-2.7
33.2
3.1
-22.4
3.1
3.1
1.3
18.6
29
Preliminary Factor Analysis
• Extract a common component from four
observables
• Estimate a single factor using factor analysis
• Default factor estimated using principal
factors with regression scoring
• Three alternatives for sensitivity analysis
– Euromoney credit rating substituted
– Drop exchange rate
– Maximum likelihood estimation
Causes and Consequences of the Crisis
30
First Principal Factor
(bottom quartile)
Iceland
Ukraine
Estonia
Argentina
Latvia
Ireland
Korea
New Zealand
UK
Hungary
Default
EuroMoney
not II
-5.5
-1.9
-1.1
-1.1
-1.0
-1.0
-.9
-.8
-.7
-.7
-3.3
-.7
-2.
.4
-1.5
-1.2
-.1
-1.0
-.8
-1.2
Drop
Exchange
Rate
-4.2
-1.4
-1.4
-.9
-1.3
-1.2
-.4
-.5
-.3
-.8
Causes and Consequences of the Crisis
MLE
Estimate
-6.7
-2.0
-1.4
-2.3
-1.1
-1.0
-.9
-.5
-.5
-1.0
31
Results
• Variables strongly positively correlated with each other and
obtain similar rankings
• List of countries hit hard by crisis plausible
– Iceland
– Baltics and Eastern Europe (Estonia, Latvia, Lithuania, Hungary,
Ukraine)
– Others: Korea, Ireland, UK
• Some surprises
– Japan characterized as unaffected
• Exchange rate performance subsequent to carry trade unwinding
– Provides initial caution about heterogeneity
• Difficult to quantify crises mechanically – well-known but forgotten
Causes and Consequences of the Crisis
32
Incidence and causes: First pass
• Link cross-country estimates of incidence to
potential causes
• To avoid endogenity as much as possible,
restrict causal data to 2006 or earlier
• Begin by comparing principal factor against
size and income
– Repeat with alternative measures
Causes and Consequences of the Crisis
33
Crisis Intensity Weakly Linked to Size
2
2
Qatar
Slovakia
Papua Ne
Peru
Barbados
Libya
Oman
El
Salva
Panama
Kyrgyz
R Saudi
Trinidad
Bahamas
Costa
Ri
Venezuel
Morocco
Iran Brazil
Lebanon
Tunisia
Bahrain
Colombia
Kuwait
Ar Egypt
Mauritiu
Israel
Sri
Lank
Slovenia
Macedoni
Indonesi
Switzerl
Cyprus
Malaysia
Malta
Botswana
Taiwan
Chile
Czech
Re
Jamaica
Hong
Kon
United
AEcuador
Germany,
Poland
Netherla
Thailand
France
Japan
Swazilan
Finland
Singapor
Denmark
Luxembou
Spain
Namibia
Croatia
Belgium
S
MexicoUnited
Greece
Canada
Portugal
Austria
Romania
Norway
Sweden
Italy
Bulgaria
South
Turkey
Af Russia
Australi
LithuaniHungary
Kazakhst
UK
New
Zeal
Korea
Ireland
Latvia
Argentin
Estonia
0
China
0
Ukraine
-2
-2
-4
Iceland
-6
-4
11
16
21
11
Default
16
21
Euromoney, not II
2
0
0
-5
-2
-4
-10
11
16
Drop Exchange Rate
21
11
16
21
Maximum Likelihood, not PF
Factors (y) against log Population (x)
Principal factors (y) against log Population (x)
Causes and Consequences of the Crisis
34
Stronger Negative Correlation with Income
2
2
Papua Ne
Kyrgyz R
0
Qatar
Slovakia
Peru
Barbados
Libya
ElColombia
Salva
Panama
Trinidad
Costa
Venezuel
Bahamas
Morocco
Egypt
IranRi Oman
Lebanon
China
Tunisia
Bahrain
Saudi
Kuwait
Ar Switzerl
Sri Lank
Mauritiu
Israel
Brazil
Slovenia
Indonesi
Macedoni
Malaysia
Cyprus
Ecuador
Botswana
Malta
Taiwan
Chile
Czech
Re
Jamaica
Hong
Kon
Germany,
United
A
Poland
Netherla
France
Japan
Swazilan
Finland
Singapor
Denmark
Spain
NamibiaThailand
Croatia
Belgium
United
SLuxembou
Mexico
Greece
Canada
Portugal
Romania
Austria
Norway
Italy
Sweden
Bulgaria
South
Turkey
AfRussia
Australi
Lithuani
Kazakhst
Hungary
UK
New Zeal Ireland
Korea
Latvia
Argentin Estonia
0
Ukraine
-2
-2
-4
Iceland
-6
-4
7
9
11
7
Default
9
11
Euromoney, not II
2
0
0
-5
-2
-4
-10
7
9
11
Drop Exchange Rate
7
9
11
Maximum Likelihood, not PF
Factors (y) against log real GDP per capita (x)
Principal factors (y) against log real GDP per capita (x)
Causes and Consequences of the Crisis
35
Regressions of Principal Factor on Size and Income
Crisis Measure:
Log (2006 Population)
Log (2006 Real GDP
per capita)
OECD Dummy
High-Income, nonOECD Dummy
Developing East Asia,
Pacific Dummy
Developing Eastern
Europe, Central Asia
Dummy
Developing Latin
American, Caribbean
Dummy
Developing South Asia
Dummy
Developing SubSaharan Africa
Dummy
Stock Market Growth,
2003-6
Default
-.01
(.08)
-.28**
(.08)
EuroM
.01
(.05)
-.44**
(.08)
No ER
-.02
(.06)
-.37**
(.07)
MLE
-.00
(.10)
-.26**
(.10)
Default
Default
.03
(.08)
-.38**
(.10)
-1.08**
(.24)
-.22*
(.09)
-.13
(.16)
-.93**
(.19)
-.23
(.14)
-.17**
(.03)
-.58**
(.15)
Causes and Consequences of the Crisis
-.23
(.12)
36
Principal Factor Regression Results
• Income robustly enters negatively
– Intuitive: rich countries hit badly
• Size insignificant
• Eastern Europe countries worst hit
• Also obtain significant negative coefficients for
Sub-Saharan and developing South-Asia
Causes and Consequences of the Crisis
37
MIMIC model
• MIMIC model consists of two sets of equations:
(1)
y    
i, j
j i
i
i   k xi,k   i
(2)
where yi , j is crisis indicator, xi ,k is an observation for potential crisis
cause;  i is latent variable representing severity of the crisis (or lack
thereof in our case), and i and  i are well-behaved disturbances
• Equation (1) links observable manifestations of the
crisis to latent variable
• Equation (2) links latent variable to causes of crisis
Causes and Consequences of the Crisis
38
Characteristics of MIMIC Model
• Substitute (2) into (1), eliminate latent variable
• MIMIC model is then a system of J equations with
right hand sides restricted to be proportional
• With normalization, system is identified
– We normalize on equity returns
• Desirable feature of MIMIC model is ability to
systematically address measurement error
• Estimate with STATA using GLLAMM model [RebeHesketh, et al (2004)]
Causes and Consequences of the Crisis
39
MIMIC Model Estimates with Only Size
and Income as Causes
Default
Log(2006
Population)
Log(2006 Real
GDP per capita)
-.98
(.95)
-7.79**
(2.44)
Euromoney instead of
Institutional Investor
-1.05
(.98)
-7.80**
(2.46)
Causes and Consequences of the Crisis
Drop SDR
Exchange Rate
-2.08
(1.12)
-10.1**
(2.66)
40
MIMIC Results: How Easily Can One
Model the Causes of the Crisis?
Causes and Consequences of the Crisis
41
Methodology
• First, add each potential additional causal
variable one at a time
• Size and income included throughout
• Also include four columns of sensitivity analysis
– Euromoney index (instead of Institutional Investor)
– Drop exchange rate depreciation
– Alternative MIMIC model estimator
• (replace adaptive quadrature with Gauss-Hermite
quadrature)
– Substitute income and regional dummies for size and
income
Causes and Consequences of the Crisis
42
Add Causes to MIMIC Model, One by One
Extra Cause
Financial Policies
Overall Capital
Stringency, 2003
Capital Regulatory
Index, 2003
Official Supervisory
Power, 2003
Ability to Take Prompt
Corrective Action, 2003
Restructuring
Power, 2003
Declaring Insolvency
Power, 2003
Credit Market
Regulation, 2006
Private Bank
Ownership, 2006
Foreign Bank
Competition, 2006
Interest Rate Controls/negative real
interest rate, 2006
Default
Euromoney,
not II
Drop
Exchange
Rate
1.87
1.26
.82
(1.40)
(1.31)
(1.35)
1.19
.78
.84
(1.25)
(1.42)
(1.19)
.62
-.0006
.13
(.61)
(.0010)
(.61)
.70
.58
.16
(.91)
(.81)
(.87)
1.11
.98
.68
(2.41)
(2.40)
(2.26)
-1.65
-1.70
-1.84
(3.06)
(3.05)
(2.95)
.35
.65
1.38
(2.44)
(2.47)
(2.59)
.04
.11
.30
(.94)
(.95)
(1.12)
.81
.87
1.77
(1.63)
(1.64)
(1.69)
.72
.83
.48
Causes
and
Consequences
of
the
Crisis
(2.82)
(2.86)
(3.08)
Different
Estimator
Region/Income
Dummies
1.04
(1.22)
.51
(1.26)
.65
(.51)
.66
(.78)
1.85
(1.94)
-.34
(3.71)
.45
(2.31)
.04
(.95)
1.05
(1.49)
-.55
(2.52)
1.00
(1.25)
-.55
(1.11)
-.0004
(.0007)
1.57**
(.54)
2.01
(2.22)
-.25
(2.80)
3.74
(2.16)
1.38
(.84)
1.13
(1.52)
1.22
(2.48)
43
Continued
Extra Cause
Financial Conditions
Domestic Credit
Private Sector, %GDP
2006
Domestic Bank Credit,
%GDP 2006
Private Sector
Credit Access, 2006
Bank Non-Performing
Loans, % Loans 2006
Bank Liquid Reserves,
%Assets 2006
Bank Capital,
%Assets 2006
Bank Claims,
%Deposits 2006
Default
Euromoney,
not II
Drop
Exchange Rate
Different
Estimator
Region/Income
Dummies
-.06
(.05)
-.05
(.04)
-.05
(.04)
-.05
(.03)
-.091*
(.045)
-.06
(.04)
-.34
(1.66)
-1.00
(.53)
.03
(.11)
.21
(.69)
-9.1*
(4.3)
-.06*
(.03)
-.15
(1.68)
-1.04
(.53)
.05
(.06)
.24
(.62)
-6.39*
(2.92)
-.02
(.04)
.25
(1.80)
-1.00
(.53)
.01
(.10)
.61
(.89)
-8.53*
(3.94)
-.056*
(.025)
-.28
(2.63)
-1.10*
(.42)
.05
(.06)
.21
(.60)
-6.2*
(2.7)
-.09*
(.04)
.68
(1.51)
n/a
Causes and Consequences of the Crisis
-.06
(.08)
-.19
(.76)
.01
(.01)
44
Continued
Extra Cause
Default
Euromoney,
not II
Drop
Exchange
Rate
Different
Estimator
Region/Income
Dummies
Asset Price Appreciation
% Chg Real Estate Prices,
2003-6
% Chg Market Cap,
%GDP 2003-6
Stock Market Growth,
2006
Market Cap,
%GDP 2006
Stocks Traded,
%GDP2006
-2.96
(5.37)
-10.20**
(1.99)
-.06
(.10)
.01
(.03)
.02
(.03)
-3.28
(5.41)
-10.5**
(1.90)
-.11
(.07)
-.00
(.03)
.02
(.03)
-11.4
(5.8)
-10.5**
(2.1)
-.08
(.13)
.01
(.04)
.02
(.03)
-2.96
(5.37)
-10.6**
(1.84)
-.11
(.06)
-.00
(.03)
.02
(.03)
-3.42
(5.34)
-7.2**
(2.0)
-.03
(.08)
-.02
(.03)
-.02
(.03)
Causes and Consequences of the Crisis
45
Continued
Extra Cause
International Imbalances
Net External Position,
%GDP 2004
Current Account,
%GDP 2006
Debt Service,
% Exports 2006
External Debt,
%GNI 2006
Gross Financing via
international
capital markets, % GDP 2006
Real Effective Exchange
Rate 2006 (2000=100)
Total Reserves,
%external debt 2006
Short-Term Debt,
%Reserves 2006
Total Reserves,
import months 2006
M2,
%(total reserves–gold) 2006
M2, %(Central Bank
foreign assets) 2006
Default
Euromoney,
not II
Drop Exchange
Rate
Different
Estimator
Region/Income
Dummies
4.23
(2.58)
.56**
(.17)
-.17
(.42)
-.01
(.18)
2.32**
(.31)
5.07*
(2.11)
.57**
(.18)
-.17
(.41)
-.01
(.17)
.0000
(.0002)
3.36
(2.66)
.41**
(.17)
-.36
(.38)
n/a
5.07*
(2.12)
.54**
(.16)
-.01
(.16)
.24**
(.06)
2.20**
(.31)
2.1
(3.6)
-.08
(.16)
n/a
-.22
(.13)
-.00
(.01)
.36**
(.08)
.36
(.43)
0
-.22
(.13)
-.028**
(.007)
.00010*
(.00002)
.40
(.33)
-.000001
(.00003)
0
-.19
(.10)
-.002
(.004)
.13**
(.03)
.40
(.35)
-.00001
(.00006)
0
n/a
-.84
(.63)
-.23
(.15)
.01
(.01)
-.10
(.10)
.14
(.38)
.26
(.17)
1.7e-7
.09
Causes and Consequences of the Crisis
(1.6e-6)
(.05)
n/a
n/a
-.013*
(.005)
.36**
(.04)
-.15
(.36)
.02
(.14)
.-02
(.05)
46
Continued
Extra Cause
Macroeconomic
Policies
Currency Union
member, 2006
GDP of Monetary
Zone, 2006
EU but non-EMU
Member, 2006
Inflation Targeter,
2006
M2,
%GDP 2006
M3,
%GDP 2006
Gov’t Budget
Surplus/Deficit, % GDP 2006
Central Gov’t Debt,
%GDP 2006
Debt,
%GNP 2006
Debt Service,
% GDP 2006
CPI Inflation,
2006
GDP Growth,
2006
Default
Euromoney
, not II
Drop Exchange
Rate
9.1
11.9*
5.15
(5.1)
(4.4)
(4.97)
-2.9e-13
-3.0e-13
0
(2.7e-13)
(2.0e-13)
-10.8
-11.4*
-14.2**
(5.8)
(5.1)
(5.1)
.02
.02
.57
(.02)
(.03)
(4.9)
-9.8e-7
n/a
-.00
(6.3e-6)
(.05)
-8.3e-7
-1.1e-6
-.01
(4.7e-6)
(6.1e-6)
(.05)
.22
.23
.12
(.52)
(.52)
(.52)
-.01
-.01
-.03
(.07)
(.08)
(.08)
.09
.08
-.35**
(.20)
(.20)
(.13)
-1.81*
-1.76*
-.97
(.71)
(.70)
(.56)
.32
.36
.11
(.72)
(.72)
(.69)
-.15
-.10
-1.25
Causes
and
Consequences
of
the
Crisis
(.74)
(.74)
(.81)
Different
Estimator
Region/Income
Dummies
12.2**
(4.23)
2.9e-13
(2.7e-13)
-10.6*
(5.4)
.02
(.02)
-.00002
(.00002)
-.00001
(.00004)
.65*
(.31)
-.02
(.07)
.22**
(.06)
-2.61**
(.29)
-.18
(.65)
-.72
(.61)
-.01
(.01)
-2.2e-13
(1.6e-13)
-10.6*
(5.3)
-5.9
(8.6)
-.04
(.06)
n/a
-.34
(.49)
-.00
(.09)
n/a
n/a
.46
(.70)
-.29
(.81) 47
Continued
Extra Cause
Institutions
Credit/Labor/Business
Regulation, EFW 2006
Polity,
2006
Constraints on
Executive, 2006
Overall Economic
Freedom, 2006
Common Law
Country
Control of
Corruption
Regulatory
Quality
Rule of
Law
Political
Rights, 2006
Civil
Liberties, 2006
Government Size,
2006
Legal Security of Property
Rights, 2006
Sound Money
Access, 2006
Default
Euromoney
, not II
Drop Exchange
Rate
2.14
2.39
3.41
(2.73)
(2.75)
(2.82)
-.25
-.36
-.35
(.33)
(.24)
(.26)
-1.58
-1.57
1.58
(1.12)
(1.12)
(1.12)
2.64
3.34
3.90
(1.53)
(3.61)
(3.77)
2.98
5.57
8.49
(4.79)
(10.31)
(4.72)
.56
.44
6.20
(3.34)
(4.40)
(3.47)
.88
1.09
.21
(3.53)
(4.98)
(3.7)
-.51
-.68
2,77
(3.80)
(3.80)
(3.98)
1.22
1.24
.92
(1.12)
(1.12)
(1.15)
1.36
1.40
.88
(1.35)
(1.35)
(1.38)
2.64
2.85
3.26*
(1.53)
(1.53)
(1.59)
.03
.21
1.33
(2.24)
(2.26)
(2.33)
-.24 Causes and
-.20
-.78
Consequences of the Crisis
(2.02)
(2.05)
(2.18)
Different
Estimator
Region/Income
Dummies
1.04
(2.33)
-.34
(.23)
-1.49
(.89)
1.75
(2.92)
3.21
(8.17)
-.35
(4.48)
1.27
(4.72)
-.11
(4.30)
1.58
(.93)
1.47
(1.08)
2.69
(1.64)
-.04
(1.57)
.56
(1.71)
3.52
(2.20)
-.39
(.39)
-1.75
(1.29)
2.78
(3.09)
.002
(.004)
-3.8
(4.8)
-.97
(3.4)
-3.40
(3.36)
1.36
(1.26)
1.08
(1.49)
1.68
(1.41)
-.15
(2.07)
.80
(2.00)
48
Continued
Extra Cause
Default
Geography
Log of
Latitude
East
Asian
Central/Eastern European
or Central Asian
Commodity
Exporter
English
Language
-4.8
(2.5)
3.9
(6.8)
-16.7**
(5.2)
1.5
(4.7)
6.23
(4.71)
Euromoney,
Drop
not II
Exchange Rate
-5.67*
(1.95)
.75
(7.9)
-16.9**
(5.4)
.73
(4.55)
6.33
(4.70)
-7.54**
(2.70)
-3.6
(6.65)
-22.0**
(4.73)
5.58
(4.37)
12.8**
(4.3)
Causes and Consequences of the Crisis
Different
Estimator
Region/Income
Dummies
-5.7**
(1.9)
-.004
(.003)
-16.2**
(5.3)
-1.9
(4.8)
8.46
(6.76)
-.56
(2.45)
n/a
n/a
1.67
(4.11)
4.37
(4.61)
49
Adding Causes to the MIMIC Model Simultaneously
Extra Cause
Financial Policies
Overall Capital
Stringency, 2003
Financial Conditions
Domestic Bank Credit,
%GDP 2006
Bank Claims/Deposits,
2006
Asset Price Appreciation
% Chg Market Cap,
%GDP 2003-6
International Imbalances
Current Account,
%GDP 2006
Short-Term Debt,
%/Reserves 2006
Macroeconomic Policies
Currency Union member,
2006
EU but non-EMU member,
2006
Institutions
Polity,
2006
Geography
Log of
Latitude
Central/Eastern European
or Central Asian
Observations
Default
Without
ST Debt
Drop Poor
(GDP
p/c<$5k)
Drop Small
(Pop<1m)
1.8
(1.9)
1.1
(1.2)
.1
(1.8)
1.8
(1.9)
.04
(.07)
9.7
(8.9)
.00
(.04)
1.7
(4.4)
.01
(.06)
10.8
(7.6)
.04
(.07)
10.8
(8.9)
-5.6*
(2.4)
-6.6**
(2.0)
-2.3
(2.3)
-5.6
(2.4)
.43
(.64)
-.08
(.10)
.01
(.24)
.04
(.57)
-.07
(.08)
.46
(.64)
-.09
(.10)
10.9
(9.3)
1.0
(10.4)
-2.8
(4.9)
-9.3
(6.4)
18.5
(9.2)
1.1
(9.7)
12.3
(9.2)
1.1
(10.4)
.19
(.56)
-.22
(.43)
.04
(.54)
.21
(.56)
3.6
1.0
2.8
(4.4)
(2.7)
(4.1)
-19.8
-15.4*
-37.0**
(10.8)
(6.9)
(10.1)
40
68
32
Causes and Consequences of the Crisis
4.4
(4.4)
-20.3
(10.8)
39
50
Results Disappointing and Weak
• Few potential causes are robustly significant
• In particular, real estate appreciation cause
insignificant
– Widely cited as a principal “cause of crisis”
• Same is true for almost all other causal
variables
• Not just an artifact of MIMIC framework
– Shows up in scatter plots of each category of
causal variables
Causes and Consequences of the Crisis
51
Financial Policy Insignificant
2008 Crisis Manifestations against Capital Regulatory Index
Depreciation against SDR
-50
-100
0
-50
50
0
100
Stock Market Change
2
4
6
8
2
6
8
GDP Growth Rate
-5
-30
5
-10
15
10
Country Credit Rating Change
4
2
4
6
8
2
4
6
8
Barth, Caprio and Levine 2003 Capital Regulatory Index
Causes and Consequences of the Crisis
52
Stock Market Run-up Does Best
2008 Crisis Manifestations against Stock Market Runup
Depreciation against SDR
-50
-100
0
-50
50
0
100
Stock Market Change
0
200
400
600
0
400
600
GDP Growth Rate
-5
-30
5
-10
15
10
Country Credit Rating Change
200
0
200
400
600
0
200
400
600
2003-06 Change in Market Capital (%GDP), WDI
Causes and Consequences of the Crisis
53
Domestic Credit Growth Insignificant
2008 Crisis Manifestations against Domestic Credit Growth
Depreciation against SDR
-50
-100
0
-50
50
0
100
Stock Market Change
-100
0
100
200
300
-100
Country Credit Rating Change
0
100
200
300
-5
-30
5
-10
15
10
GDP Growth Rate
-100
0
100
200
300
-100
0
100
200
300
2006 Domestic Bank Credit (%GDP), WDI
Causes and Consequences of the Crisis
54
Bank Claim/Deposits at 10% Confidence
2008 Crisis Manifestations against Bank Claim/Deposit Ratio
Depreciation against SDR
-50
-100
0
-50
50
0
100
Stock Market Change
0
1
2
3
4
0
Country Credit Rating Change
1
2
3
4
-5
-30
5
-10
15
10
GDP Growth Rate
0
1
2
3
4
0
1
2
3
4
2006 Bank Claims (%Deposits), IFS
Causes and Consequences of the Crisis
55
Real Estate Appreciation
Insignificant
2008 Crisis Manifestations against Real Estate Price Runup
Depreciation against SDR
-50
-100
0
-50
50
0
100
Stock Market Change
0
50
100
150
0
100
150
GDP Growth Rate
-5
-30
5
-10
15
10
Country Credit Rating Change
50
0
50
100
150
0
50
100
150
2003-06 Change in Real Real Estate Prices
Causes and Consequences of the Crisis
56
Bank Leverage Insignificant
2008 Crisis Manifestations against Bank Capital Adequacy
Depreciation against SDR
-50
-100
0
-50
50
0
100
Stock Market Change
5
10
15
20
25
5
Country Credit Rating Change
10
15
20
25
-5
-30
5
-10
15
10
GDP Growth Rate
5
10
15
20
25
5
10
15
20
25
2006 Bank Capital (%Assets), WDI
Causes and Consequences of the Crisis
57
Current Account Comes In, but not
with regional dummies included
2008 Crisis Manifestations against Current Account
Depreciation against SDR
-50
-100
0
-50
50
0
100
Stock Market Change
-40
-20
0
20
40
60
-40
Country Credit Rating Change
-20
0
20
40
60
-5
-30
5
-10
15
10
GDP Growth Rate
-40
-20
0
20
40
60
-40
-20
0
20
40
60
2006 Current Account (%GDP), WDI
Causes and Consequences of the Crisis
58
Budget Deficits Didn’t Matter
2008 Crisis Manifestations against Government Budget
Depreciation against SDR
-50
-100
0
-50
50
0
100
Stock Market Change
-10
0
10
20
-10
10
20
GDP Growth Rate
-5
-30
5
-10
15
10
Country Credit Rating Change
0
-10
0
10
20
-10
0
10
20
2006 Budget Surplus/Deficit (%GDP), WDI
Causes and Consequences of the Crisis
59
Also weak results relative to latent variable
Latent Crisis Variable against Potential Causes
2006 Causes unless noted
Log Population
Real Estate Price Runup
Stock Market Runup
2003-06 price change
2003-06 price change
10
15
20
7
8
9
10
2
0
-6 -4 -2
-6 -4 -2
-6
-6
-4
-4
-2
-2
0
0
0
2
2
2
Log Real GDP per capita
11
0
.5
1
1.5
Bank Capital
%GDP
%Assets
4
6
Bank Claims/Deposits
1
2
3
4
5
6
-2
0
0
100
200
300
400
-6
-4
-6 -4 -2
-6 -4 -2
-6 -4 -2
0
0
0
2
2
2
2
Domestic Bank Credit
2003
2
Overall Capital Stringency
0
5
10
15
20
25
0
1
2
3
4
Government Budget
Current Account
Real Eff. Exchange Rate
2003
%GDP
%GDP
REER=100 in 2000
2
4
6
8
2
2
0
0
-10
0
10
20
-6 -4 -2
-6 -4 -2
-6 -4 -2
-6 -4 -2
0
0
2
2
Capital Regulatory Index
-40
-20
0
20
Causes and Consequences of the Crisis
40
60
80
100
120
140
160
60
Results suggest that Measurable Pre-existing
Conditions had Little Common Impact
• Few of the potential causes robust across slightly
different specifications
• Some exceptions
– Equity market appreciation
– Current account deficits (weak)
– Weaker evidence for high credit growth and banking
sector leverage
• Still, these variables alone unlikely to generate
acceptable EWS model
• Moreover, unlikely to dominate following these
characteristics in a non-parametric way
Causes and Consequences of the Crisis
61
What Happens with Updated Data?
• 2008-09 and 2008 Data Highly Correlated
Four Crisis Manifestations
2008 against 2008-09 Changes
2008-09
50
100
-100
-50
0
2008-09
50
Country Risk Change
GDP Growth
100
5
-5
0
-20 -10
0
2008
10 15
0
10 20
-50
2008
-50
-100
0
2008
50
0
Stock Market Growth
-50
2008
100
SDR Depreciation
-20
-10
0
2008-09
10
20
-20
-10
Causes and Consequences of the Crisis
0
2008-09
10
20
62
Parameterization
• Four Consequences – 2008-09 changes in: a)
GDP; b) stocks; c) country credit rating; d) SDR
exchange rate
• Three Successful Crisis Causes from RS:
– 2006 Real GDP per capita as control
– 2003-06 stock market run-up
– Proportion 2006 trade with US
Causes and Consequences of the Crisis
63
2008-09 MIMIC Estimates
Log Real GDP per capita, 2006
Growth in Market Capitalization
2003-06
Trade with USA, % total 2006
trade
Current Account %GDP, 2006
-16.4**
(3.9)
-8.3**
(3.2)
.32
(.20)
Credit Market Regulation, 2006
Growth in Bank Credit %GDP,
2000-06
Real House Price growth, 200006
ST External Debt
% GDP, 2006
Observations
81
-12.9*
(5.4)
-8.9**
(3.45
.36
(.21)
.2
(.3)
-2.0
(3.4)
-.14
(.11)
75
-6.4
(5.2)
-10.7**
(3.2)
.61**
(.21)
.2
(.3)
2.0
(3.1)
-.23**
(.08)
-.01
(.05)
43
-8.6
(13.9)
-8.6
(5.3)
.25
(.41)
.1
(1.1)
-2.1
(5.9)
-.1
(.4)
-.6
(.9)
38
Results Similar
• Negative, usually significant effects on real
GDP per capita and stock market growth
• Little else
• But MIMIC methodology may be special
– Will pursue shortly … but first …
Causes and Consequences of the Crisis
65
What About Contagion/CrossCountry Influences?
Causes and Consequences of the Crisis
66
Link Crisis Epi-Center to Other
Countries
• Now go beyond national causes, allow
international linkages
– Not Simply Common Shocks
• National characteristics make country more/less
vulnerable to common shocks
• Have already included multilateral features
– Now incorporate bilateral linkages between crisis
origins and other countries
Causes and Consequences of the Crisis
67
Allow Two “Channels” of Linkage
• Real/Trade Channel
– Country x devalues, gains competitiveness,
induces devaluation in others
• Size of effect proportional to trade/GDP
• Gerlach-Smets (1995); Eichengreen et al (1996)
• Asset Channel
– Epi-center asset values decline, others lose wealth
• Size of effect proportional to financial exposure/GDP
• Forbes-Chinn (2004), Ehrmann - Fratzscher (2009)
Causes and Consequences of the Crisis
68
Some Potential Channels Omitted
• Trade Competitiveness Channel
– Country in crisis devalues, other countries lose
competitiveness [e.g. Baig and Goldfajn (1999)]
– Primary crisis impact lost export opportunities in
U.S., not disadvantage in 3rd countries
• Common Creditor Channel
– Crisis country defaults, weakens creditors’ ability
to lend to others [e.g. Kaminsky and Reinhart
(2000)]
– Crisis impacted entire global financial system
Causes and Consequences of the Crisis
69
Allow for Multiple Epi-Centers
• USA seems most likely origin of crisis
– But different possible epi-centers exist
• Allow for other potential centers:
– UK, Germany, Spain
– Japan, Korea in Asia
– Small Europeans (Iceland, Ireland, Ukraine,
Estonia, Latvia, and Lithuania) – collectively
Causes and Consequences of the Crisis
70
(14) Measures of Trade Linkages
• Exports to USA (as proportion of all exports)
– Also consider other epi-centers
– Also consider 2-way trade (not just exports)
Causes and Consequences of the Crisis
71
Curious Export Results
Linkage
(2006)
Exposure to
Default
Exports
USA
Exports
UK
Exports
Germany
Exports
Japan
Exports
Spain
Exports
Korea
Exports
Small Crises
.28**
(.10)
-.55
(.35)
-.42
(.25)
-.26
(.36)
.18
(.45)
.17
(.55)
-1.02**
(.39)
Drop Exchange
Rate
Consequence
.32**
(.10)
-.53
(.36)
-.50
(.26)
-.24
(.37)
.22
(.46)
.03
(.56)
-1.11**
(.39)
Euromoney,
not II
.29**
(.10)
-.66
(.35)
-.44
(.25)
-.26
(.36)
.18
(.45)
.18
(.55)
-1.04**
(.39)
Causes and Consequences of the Crisis
Condition on
Trade (%GDP),
not size
.31**
(.09)
-.51
(.29)
-.43
(.25)
-.27
(.34)
.16
(.42)
-.0003
(.0008)
-.91**
(.35)
Condition on
Exports (%GDP),
not size
.30**
(.09)
.0002
(.0007)
-.45
(.24)
-.29
(.35)
.14
(.42)
-.0002
(.0002)
-.89**
(.35)
72
Exposure to America Helped!
• No obvious explanation, but economically and
statistically significant effect
– Effect of Dollar Appreciation?
• Results broadly insensitive
• Export exposure to “Small Crisis” countries
hurt, as expected
• Results broadly similar when total 2-way trade
used instead of just exports
Causes and Consequences of the Crisis
73
Similar Results for Trade
Linkage
(2006)
Exposure to
Default
Trade
USA
Trade
UK
Trade
Germany
Trade
Japan
Trade
Spain
Trade
Korea
Trade
Small Crises
.38**
(.12)
-.68
(.50)
-.37
(.36)
-.36
(.44)
.21
(.49)
.05
(.74)
-1.29**
(.48)
Drop Exchange
Rate
Consequence
.43**
(.12)
-.58
(.52)
-.45
(.26)
-.36
(.46)
.21
(.51)
-.20
(.76)
-1.39**
(.50)
Euromoney,
not II
.38**
(.12)
-.69
(.51)
-.39
(.26)
-.35
(.44)
.20
(.50)
.06
(.74)
-1.30**
(.49)
Causes and Consequences of the Crisis
Condition on
Trade (%GDP),
not size
.39**
(.11)
-.67
(.53)
-.37
(.26)
-.40
(.44)
.20
(.42)
.05
(.65)
-1.11*
(.41)
Condition on
Exports (%GDP),
not size
.38**
(.11)
-.61
(.57)
-.38
(.25)
-.45
(.45)
.22
(.42)
-.0003
(.0004)
.0003
(.0003)
74
(34) Measures of Financial Linkage
• Share of overseas assets held in USA (as
proportion of all overseas wealth)
– Also consider other potential epi-centers
– Also consider debt, long-term debt
– Also consider BIS Consolidated banking data
– Even more (only for American exposure): US TIC
assets (as proportion of national GDP)
• Also consider American equity/long
debt/debt/treasuries/long treasuries
– Also consider $PPG debt denominated in yen/$
Causes and Consequences of the Crisis
75
Again, American Exposure Helps!
Linkage
(2006)
Exposure to
Default
CPIS Asset Share
USA
CPIS Asset Share
UK
CPIS Asset Share
Germany
CPIS Asset Share
Japan
CPIS Asset Share
Spain
CPIS Asset Share
Korea
CPIS Asset Share
Small Crises
.27*
(.10)
-.28
(.14)
-.30
(.38)
.26
(.88)
-.44
(.75)
.16
(1.97)
-.82
(.62)
Drop Exchange
Rate
Consequence
.36**
(.09)
-.15
(.16)
-.68
(.38)
1.02
(.95)
-.70
(.88)
-1.01
(2.47)
-1.24
(.73)
Euromoney,
not II
.29*
(.10)
6e-6
(.00003)
-.36
(.39)
.32
(.91)
-.51
(.76)
.19
(1.98)
-.87
(.63)
Causes and Consequences of the Crisis
Condition on
C/acc (%GDP),
not size
.19*
(.10)
-.28**
(.10)
-.25
(.33)
.0003
(.0007)
-.0001
(.0003)
-1.06
(1.36)
-.43
(.44)
Condition on NFA
(%GDP), not size
.23*
(.10)
.0002
(.0002)
-.0007
(.0006)
.0003
(.0007)
-.0002
(.0005)
.0008
(.002)
.0001
(.0005)
76
A Few Significant Results
Linkage
(2006)
Exposure to
Default
Drop Exchange
Rate
Consequence
Euromoney,
not II
Condition on
C/acc (%GDP),
not size
Condition on NFA
(%GDP), not size
CPIS Debt Share
USA
.19*
(.09)
.28**
(.08)
.21*
(.09)
.10
(.09)
.13
(.09)
CPIS Long Debt
Share
USA
-.64
(1.26)
.28**
(.09)
.20*
(.09)
.11
(.09)
.14
(.09)
US TIC
Assets/GDP
USA
.19
(1.39)
-.04
(1.40)
.17
(1.39)
.10
(1.81)
.42
(1.61)
% PPG Debt in $
USA
.21*
(.10)
.21*
(.10)
.22*
(.10)
.18**
(.06)
.15*
(.07)
Causes and Consequences of the Crisis
77
Some Sensitivity, Bigger Mystery
• Why does exposure to America (most likely
epi-center of 2008 crisis) seem to help?
– Special Advantage of “Exorbitant Privilege”?
– Interesting Future Research topic
• Still, results not very strong
– Most measures of contagion insignificant
– Also, we know epi-center and timing of 2008 crisis
– Doesn’t help EWS much
Causes and Consequences of the Crisis
78
What About Newer Data and
Literature?
Causes and Consequences of the Crisis
79
Emerging Literature
• Four Big Differences from Our Early Work
1. Measures of Crisis Intensity
2. Potential Causes (Covariates)
3. Estimator linking causes, intensity
4. Country Sample
Causes and Consequences of the Crisis
80
Key References
• Berkmen, Gelos, Rennhack, Walsh (2009)
“BGRW”
• Blanchard, Faruqee and Das (2010) “BFD”
• Claessens, Dell’Arriccia, Igan, Laeven (2010)
“CDIL”
• Frankel and Saravelos (2010) “FS”
• Giannone, Lenza and Reichlin (2010) “GLR”
• Lane and Milesi-Fettetti (2010) “LMF”
Causes and Consequences of the Crisis
81
Measures of Crisis Intensity
•
•
•
•
•
•
•
Default: Real 2008-09 GDP growth (from EIU)
Real GDP growth change, 2008-09 - 2005-07 (LMF)
Real GDP growth change, 2008-09 - 1990-07 (BFD)
Revision to WEO 2009 growth forecast (BGRW)
2009 output gap (from OECD)
2008-09 consumption growth
First Principal Factor from 4 RS variables
Causes and Consequences of the Crisis
82
Crisis Intensity Measures
Correlated
Different Crisis Measures
2008-09
Growth
20
2008-09
Growth '05-'07
Growth
0
-20
-40
20
2008-09
Growth '90-'07
Growth
0
-20
-40
10
Revised
WEO
Forecast
for 2009
Growth
0
-10
-20
0
Output
Gap
2009
-5
-10
100
Consumption
Growth
2008-09
50
0
2
Extracted
Principal
Factor
0
-2
-20
0
20 -40
-20
0
20
-40
-20
0
20
-20
-10
0
10
-10
-5
Causes and Consequences of the Crisis
0
0
50
100
83
Causes: Many
Unsuccessful Attempts
• Credit %GDP
• Debt %GDP
• Domestic Banking Sector Characteristics
• Fiscal Policy
• Trade Flows
• Capital Flows
Causes and Consequences of the Crisis
84
Causes: Some Successes
(Investigate All Below)
•
•
•
•
•
•
•
•
•
House Price Appreciation (CDIL)
Credit Growth (BGRW, CDIL, LMF)
Credit Market Regulation (GLR)
Current Account %GDP (CDIL, LMF)
Financial Leverage (BGRW)
Fixed Exchange Rate Regime (BFD, LMF)
Reserves (FS, Obstfeld et al for depreciation)
Short-Term External Debt (BGRW)
Trading Partner Growth (BFD, LMF)
Causes and Consequences of the Crisis
85
Different Country Samples
1.
2.
3.
4.
5.
6.
Full sample (107)
(51) World Bank High Income
(74) IMF non-Advanced
(89) non-Oil Exporters
(91) non-Small Financial Centers
(51) non-oil, non-FC High/Upper-Middle
Income
Causes and Consequences of the Crisis
86
Causes of Crisis Intensity
• Begin to link causes to manifestations using
simple bivariate OLS
– Causes from 2006 (relevant time horizon for
policy; reduce endogenous policy problem)
– Include intercept
– Control for Log 2006 real GDP per capita
– Different Measures of Crisis Intensity
– Different Country Samples
Causes and Consequences of the Crisis
87
Informal Graphical Evidence
2008-09 Growth Against Key Regressors
Credit Market Regulation
Bank Credit Growth
2006; Corr=-.47
% GDP, 2000-06; Corr=-.40
20
10
10
-20 -10
-20 -10
0
0
10
0
-20 -10
2008-09 Growth
20
20
Current Account
% 2006 GDP; Corr=.31
6
-40
-20
0
20
40
10
-50
50
100
150
2006, Corr=-.32
20
40
60
80
10
0
-20 -10
0
-20 -10
0
10
20
Log Real GDP per capita
2003-06; Corr=.17
20
Growth in Market Cap'n
20
Trade with US
10
0
0
% 2006 Trade; Corr=.05
-20 -10
2008-09 Growth
7
8
9
Higher = Tighter
0
2
4
6
Causes and Consequences of the Crisis
7
8
9
10
11
88
Effect of Exchange Rate Regime
(Dummy for 2006 Fix)
All
2008-09
Growth
2008-09 –
’05-’07 Gr
2008-09 ’90-’07 Gr
Rev. WEO
’09 Gr.
Output
Gap ’09
Cons’n Gr
2008-09
Principal
Factor
-.02
(.01)
.00
(.01)
-.00
(.01)
.01
(.01)
.00
(.01)
-.06
(.04)
-.002
(.002)
High
No
No
Income
Advanced
Oil
-.05**
-.01
.01
(.01)
(.02)
(.01)
-.03
.01
.03*
(.01)
(.02)
(.01)
-.02
.01
.02
(.02)
(.02)
(.01)
-.00
.01
.02*
(.01)
(.01)
(.01)
.00
n/a
.00
(.01)
(.01)
-.11
-.07
-.00
(.07)
(.07)
(.01)
-.005*
.001
.001
(.002)
(.003) of the Crisis
(.002)
Causes and Consequences
No Fin’l
Centers
-.01
(.01)
.01
(.01)
.01
(.01)
.01*
(.01)
.01
(.01)
-.06
(.04)
-.000
(.002)
No Poor,
Oil, or FCs
.02
(.02)
.03*
(.02)
.02
(.01)
.02*
(.01)
.01
(.01)
.01
(.01)
.003
(.002) 89
Current Account
(2006, % GDP)
All
2008-09
Growth
2008-09 –
’05-’07 Gr
2008-09 ’90-’07 Gr
Rev. WEO
’09 Gr.
Output
Gap ’09
Cons’n Gr
2008-09
Principal
Factor
.23**
(.05)
.20**
(.06)
.21**
(.06)
.02
(.03)
.07
(.04)
.18*
(.07)
.03**
(.01)
High
Income
.21**
(.06)
.17*
(.07)
.17*
(.07)
.02
(.03)
.07*
(.03)
.16
(.11)
.02*
(.01)
No
Advanced
.26**
(.06)
.26**
(.08)
.26**
(.08)
.05
(.03)
n/a
.20*
(.09)
.03**
(.01)
No
Oil
.25**
(.07)
.26**
(.09)
.23*
(.09)
.01
(.05)
.07
(.04)
.25**
(.07)
.03**
(.01)
Causes and Consequences of the Crisis
No Fin’l
Centers
.22**
(.06)
.21**
(.07)
.22**
(.07)
.02
(.03)
.04
(.07)
.19*
(.08)
.02**
(.01)
No Poor,
Oil, or FCs
.27*
(.12)
.33*
(.14)
.28
(.14)
.01
(.07)
.04
(.07)
.26**
(.09)
.02
(.02)
90
Trading Partner Growth (2005-07)
All
2008-09
Growth
2008-09 –
’05-’07 Gr
2008-09 ’90-’07 Gr
Rev. WEO
’09 Gr.
Output
Gap ’09
Cons’n Gr
2008-09
Principal
Factor
14.8
(65.4)
-46.9
(71.7)
-52.9
(68.2)
73.9
(26.9)
29.2
(43.1)
45.6
(85.2)
1.0
(8.6)
High
Income
211.7
(88.1)
156.3
(89.3)
141.2
(86.2)
-66.8
(42.9)
16.2
(41.5)
245.2
(149.8)
20.2
(6.5)
No
Advanced
-33.0
(87.1)
-83.6
(99.0)
-95.6
(92.1)
-58.7
(34.2)
n/a
-35.1
(93.5)
-14.4
(14.7)
No
Oil
-51.6
(64.4)
-106.9
(71.8)
-99.4
(65.3)
-110.2**
(27.4)
29.2
(43.1)
-63.5
(53.4)
-1.7
(9.5)
Causes and Consequences of the Crisis
No Fin’l
Centers
-12.6
(85.1)
-79.6
(92.8)
-89.0
(88.3)
-66.6*
(32.1)
11.1
(48.7)
76.3
(124.6)
-10.9
(12.4)
No Poor,
Oil, or FCs
-21.9
(169.0)
-89.4
(194.9)
-98.3
(179.2)
-85.8
(71.5)
11.1
(48.7)
-95.5
(117.9)
-12.5
(16.5)
91
Credit Mkt Regulation (2006, EFW)
All
2008-09
Growth
2008-09 –
’05-’07 Gr
2008-09 ’90-’07 Gr
Rev. WEO
’09 Gr.
Output
Gap ’09
Cons’n Gr
2008-09
Principal
Factor
-2.38**
(.68)
-2.54**
(.78)
-2.72**
(.74)
-1.02**
(.34)
-.87
(.49)
-1.63**
(.61)
-.21**
(.07)
High
Income
-1.15
(1.37)
-1.30
(1.56)
-1.17
(1.57)
-.21
(.72)
-.55
(.44)
-.77
(1.17)
-.08
(.14)
No
Advanced
-2.95**
(.82)
-3.10**
(.91)
-3.38**
(.88)
-1.33**
(.36)
n/a
-1.93*
(.74)
-.27**
(.08)
No
Oil
-2.76**
(.77)
-2.84**
(.91)
-3.07**
(.85)
-.97*
(.38)
-.87
(.49)
-1.46*
(.63)
-.22**
(.07)
Causes and Consequences of the Crisis
No Fin’l
Centers
-2.56**
(.71)
-2.75**
(.81)
-2.97**
(.75)
-1.13**
(.35)
-1.07*
(.50)
-1.60*
(.63)
-.21**
(.07)
No Poor,
Oil, or FCs
-2.22
(1.12)
-2.51
(1.27)
-2.82*
(1.24)
-.89
(.53)
-1.07*
(.50)
-1.46
(.74)
-.20
(.10)
92
ST External Debt (2006, % GDP)
All
2008-09
Growth
2008-09 –
’05-’07 Gr
2008-09 ’90-’07 Gr
Rev. WEO
’09 Gr.
Output
Gap ’09
Cons’n Gr
2008-09
Principal
Factor
-.27**
(.10)
-.32**
(.12)
-.28*
(.12)
-.12**
(.03)
n/a
-.12**
(.03)
-.04**
(.01)
High
Income
n/a
n/a
n/a
n/a
n/a
n/a
n/a
No
Advanced
-.27**
(.10)
-.32**
(.12)
-.28*
(.12)
-.12**
(.03)
n/a
No
Oil
-.30**
(.11)
-.35**
(.12)
-.30*
(.12)
-.11**
(.03)
n/a
No Fin’l
Centers
-.27**
(.10)
-.32**
(.12)
-.28*
(.12)
-.12**
(.03)
n/a
No Poor,
Oil, or FCs
-.25*
(.12)
-.21*
(.13)
-.24
(.13)
-.09*
(.04)
n/a
-.11
(.15)
-.04**
(.01)
-.10
(.15)
-.04**
(.01)
-.11
(.15)
-.04**
(.01)
-.04
(.16)
-.03
(.01)
Causes and Consequences of the Crisis
93
Housing Price Change (2000-06)
All
2008-09
Growth
2008-09 –
’05-’07 Gr
2008-09 ’90-’07 Gr
Rev. WEO
’09 Gr.
Output
Gap ’09
Cons’n Gr
2008-09
Principal
Factor
-.06**
(.02)
-.08**
(.02)
-.07**
(.02)
-.02**
(.01)
-.00
(.01)
-.05**
(.01)
-.007**
(.002)
High
Income
-.04*
(.02)
-.06*
(.02)
-.06*
(.02)
-.02
(.01)
-.01
(.01)
-.04**
(.01)
-.005**
(.002)
No
Advanced
-.07*
(.02)
-.09**
(.02)
-.08**
(.02)
-.02*
(.01)
n/a
-.04*
(.01)
-.007**
(.002)
No
Oil
-.06**
(.02)
-.08**
(.02)
-.07**
(.02)
-.02*
(.01)
-.00
(.01)
-.05**
(.01)
-.007
(.002)
Causes and Consequences of the Crisis
No Fin’l
Centers
-.06**
(.02)
-.08**
(.02)
-.07**
(.02)
-.02*
(.01)
.00
(.01)
-.04**
(.01)
-.006**
(.002)
No Poor,
Oil, or FCs
-.06*
(.02)
-.07*
(.03)
-.07*
(.03)
-.02
(.01)
.00
(.01)
-.04**
(.01)
-.005*
(.002)
94
Bank Credit Growth (2000-06, %GDP)
All
2008-09
Growth
2008-09 –
’05-’07 Gr
2008-09 ’90-’07 Gr
Rev. WEO
’09 Gr.
Output
Gap ’09
Cons’n Gr
2008-09
Principal
Factor
-.07**
(.02)
-.06**
(.02)
-.06*
(.02)
-.01
(.01)
-.01*
(.01)
-.11*
(.05)
-.011**
(.002)
High
Income
-.07**
(.03)
-.06*
(.02)
-.05
(.02)
-.02
(.01)
-.01*
(.01)
-.14
(.08)
-.011**
(.002)
No
Advanced
-.11**
(.04)
-.11*
(.05)
-.11*
(.04)
-.02
(.02)
n/a
-.15
(.10)
-.014**
(.004)
No
Oil
-.06*
(.03)
-.05
(.03)
-.05
(.03)
-.01
(.01)
-.01*
(.01)
-.06**
(.01)
-.010**
(.002)
Causes and Consequences of the Crisis
No Fin’l
Centers
-.09**
(.03)
-.08*
(.03)
-.08*
(.03)
-.01
(.02)
-.02
(.01)
-.13
(.08)
-.010**
(.003)
No Poor,
Oil, or FCs
-.07
(.04)
-.07
(.05)
-.08
(.05)
-.01
(.02)
-.02
(.01)
-.07
(.03)
-.007
(.005)
95
Bank Credit Level (2000-06, %GDP)
All
2008-09
Growth
2008-09 –
’05-’07 Gr
2008-09 ’90-’07 Gr
Rev. WEO
’09 Gr.
Output
Gap ’09
Cons’n Gr
2008-09
Principal
Factor
-.02
(.01)
.00
(.01)
-.00
(.01)
.01
(.01)
.00
(.01)
-.06
(.04)
-.002
(.002)
High
Income
-.05**
(.01)
-.03
(.01)
-.02
(.02)
-.00
(.01)
.00
(.01)
-.11
(.07)
-.005*
(.002)
No
Advanced
-.01
(.02)
.01
(.02)
.01
(.02)
.01
(.01)
n/a
-.07
(.07)
.001
(.003)
No
Oil
.01
(.01)
.03*
(.01)
.02
(.01)
.02*
(.01)
.00
(.01)
-.00
(.01)
.001
(.002)
Causes and Consequences of the Crisis
No Fin’l
Centers
-.01
(.01)
.01
(.01)
.01
(.01)
.01*
(.01)
.01
(.01)
-.06
(.04)
-.000
(.002)
No Poor,
Oil, or FCs
.02
(.02)
.03*
(.02)
.02
(.01)
.02*
(.01)
.01
(.01)
.01
(.01)
.003
(.002)
96
International Reserves (2006, % GDP)
All
2008-09
Growth
2008-09 –
’05-’07 Gr
2008-09 ’90-’07 Gr
Rev. WEO
’09 Gr.
Output
Gap ’09
Cons’n Gr
2008-09
Principal
Factor
.068*
(.028)
.029
(.026)
.025
(.031)
-.047*
(.019)
.082
(.052)
.073
(.049)
.007
(.005)
High
Income
.068
(.042)
-.015
(.022)
-.019
(.022)
-.065**
(.022)
.067
(.051)
.104
(.127)
.007**
(.002)
No
Advanced
.066
(.045)
.057
(.041)
.035
(.053)
-.034
(.027)
n/a
.066
(.059)
.006
(.010)
No
Oil
.075*
(.036)
.028
(.033)
.016
(.029)
-.054*
(.023)
.082
(.052)
.039*
(.019)
.0003
(.005)
Causes and Consequences of the Crisis
No Fin’l
Centers
.066
(.036)
.045
(.033)
.032
(.044)
-.042
(.024)
.082
(.052)
.115
(.083)
.007
(.008)
No Poor,
Oil, or FCs
.007
(.075)
.017
(.080)
-.028
(.071)
-.051
(.050)
.082
(.052)
.038
(.040)
-.003
(.010)
97
Results
• Not easy to link crisis causes to intensity
robustly
• Failures:
– Exchange Rate Regime
– Trading Partner Growth
– Credit/GDP ratio
– International Reserves
Causes and Consequences of the Crisis
98
Successes
•
•
•
•
Current Accounts (surplus helps)
Bank Credit Growth (less helps)
Credit Market Regulation (tighter helps)
Housing Price Growth (less helps – 45
observations!)
• Short Term External Debt (less helps – 52 obs)
Causes and Consequences of the Crisis
99
Including Causes Simultaneously
1. 2006 Current Account, % GDP
2. 2006 Credit Market Regulation
3. 2000-06 Growth in Bank Credit (% GDP)
• Extra Controls
1. log 2006 real GDP per capita
2. 2000-06 growth in stock market (% GDP)
3. Proportion 2006 trade with US
Causes and Consequences of the Crisis
100
Full Sample, Different Regressands
2008-09
Growth
2008-09
- ’05-07
2008-09
- ’90-07
Revised
WEO Fc
Output
Gap ’09
Co Gr
2008-09
Principal
Factor
CA %
Cred Mkt Bk Cred
US Tr
Mkt Cap Ln rl GDP
GDP ‘06 Reg ‘06 Gr ‘00-06 %Tr 2006 Gr ‘03-06 pc ’06
.22**
-2.01**
-.012
-.002
.85
-3.19**
(.08)
(.70)
(.025)
(.046)
(.99)
(.1.21)
.15
-2.30**
-.027
.021
.80
-1.52
(.09)
(.84)
(.024)
(.053)
(1.08)
(1.52)
.17*
-2.50**
-.021
.021
1.00
-1.94
(.08)
(.77)
(.025)
(.049)
(1.02)
(1.29)
-.03
-.91*
-.017
.049
-.00
.05
(.04)
(.38)
(.010)
(.019)
(.40)
(.55)
.04
-.77
-.005
-.021
.59
-1.01
(.08)
(.55)
(.014)
(.038)
(.82)
(2.06)
.18
-1.20*
-.025
-.037
-.11
-.44
(.11)
(.46)
(.015)
(.045)
(.43)
(.76)
.01
-.14*
-.006*
.001
-.07
-.38**
(.01)
(.06)
(.003)
(.005)
(.09)
(.10)
Causes and Consequences of the Crisis
R2
Obs
.41
75
.29
74
.36
74
.18
75
.23
26
.40
71
.52
70
101
Different Samples, 2008-09 Growth
Sample
Period:
All
Only
High Inc
No
Adv’d.
Drop
Oil
Drop
FCs
No Poor
Oil, FCs
CA %
Cred Mkt Bk Cred
US Tr
Mkt Cap Ln rl GDP
GDP ‘06 Reg ‘06 Gr ‘00-06 %Tr 2006 Gr ‘03-06 pc ’06
.22**
(.08)
.20
(.11)
.24
(.12)
.25*
(.10)
.22*
(.09)
.30
(.15)
-2.01**
(.70)
-.03
(1.26)
-2.44**
(.80)
-2.13**
(.77)
-2.09**
(.70)
-1.30
(.91)
-.012
(.025)
-.020
(.025)
-.065
(.068)
-.007
(.026)
-.031
(.035)
-.039
(.038)
-.002
(.046)
.014
(.061)
-.021
(.058)
.023
(.053)
-.001
(.051)
.020
(.077)
.85
(.99)
4.84
(2.42)
1.03
(1.07)
.52
(1.83)
.77
(.99)
1.49
(1.32)
Causes and Consequences of the Crisis
-3.19**
(.1.21)
-4.27
(4.10)
-3.97
(2.38)
-3.36*
(1.28)
-2.95*
(1.40)
-3.41
(2.42)
R2
Obs
.41
75
.41
36
.46
45
.44
68
.41
65
.37
41
102
Sobering Results
• Poor fit consistently
• Despite pre-filtering, many insignificant
regressors
– Stock market run-up (RS)
– Trade linkage with US (RS)
• Some Weak “Successes”
– Credit Growth significant at 5% for 1/7 measures
– Current Account, income for 2/7
Causes and Consequences of the Crisis
103
More …
• One Real Success (GLR)
– Credit Market Regulation: significant 6/7
• 2008-09 Real GDP growth
– Three pre-crisis vulnerabilities significant
– Use that crisis intensity measure to maximize odds
of successful EWS
• Even sticking with growth though
– Results sensitive to exact sample of countries
Causes and Consequences of the Crisis
104
Narrowing the Focus Further
Sample
Period:
All
All (no ADV
no FCs)
Only High
Income
Drop
Advanced
Drop Oil
Drop Oil
Drop Financial
Centers
Drop Poor,
Oil, FCs
Drop Poor,
Oil, FCs
CA % GDP
‘06
Cred Mkt
Reg ‘06
Ln rl GDP pc
’06
.21**
(.06)
.05
(.15)
.21**
(.07)
.26**
(.08)
.22**
(.07)
.08
(.16)
.22**
(.07)
.27*
(.12)
-.01
(.21)
-2.16**
(.67)
-2.38*
(.93)
-.51
(1.30)
-2.73**
(.79)
-2.39**
(.68)
-2.59*
(.97)
-2.41**
(.69)
-1.86
(.95)
-1.52
(1.47)
-2.98**
(.83)
-1.53
(2.31)
-4.93
(3.09)
-3.15
(1.71)
-2.92**
(.87)
-1.16
(2.35)
-2.85**
(.92)
-3.32
(1.80)
-1.73
(10.)
ST Ext Debt
%GDP, 2006
-.36*
(.13)
-.37**
(.14)
-.43*
(.19)
Causes and Consequences of the Crisis
R2
Obs
.38
84
.44
42
.30
40
.39
52
.42
76
.48
38
.38
72
.29
4
.53
18
105
Even Restricting to Positive Results
• Only short-term external debt consistently
statistically significant
– Current Account positive; real GDP per capita, and
credit regulation have consistently negative signs
• Fit still poor
• Weighting by population has little effect
• Non-linearities (thresholds?) delivers little
Causes and Consequences of the Crisis
106
Non-Parametric (LWR) Estimation
2008-09 Growth against Key Regressors
LWR Estimation on Whole Sample Residuals; bandwith=.5
0
-20 -10
-20 -10
0
10 20
Credit Market Regulation
10 20
Current Account %GDP
-20
0
20
40
-3
-1
0
1
2
0
-20 -10
-20 -10
0
10 20
ST External Debt %GDP
10 20
Log Real GDP per capita
-2
-2
-1
0
1
2
-20
-10
Causes and Consequences of the Crisis
0
10
20
30
107
Other Global Recessions: 1991-92
All
Only High
Income
Drop
Advanced
Drop Oil
Drop FCs
Drop Poor,
Oil, FCs
CA % GDP
‘90
.29**
(.08)
.30
(.16)
.24*
(.09)
.33**
(.12)
.29**
(.09)
.57**
(.19)
Cred Mkt
Reg ‘90
.59*
(.29)
.26
(.58)
.83*
(.33)
.62
(.32)
.59
(.30)
.85*
(.32)
Ln rl GDP pc
’90
-1.15*
(.46)
-2.53
(2.67)
-.43
(.68)
-1.26*
(.49)
-1.15*
(.48)
-3.69**
(.84)
Causes and Consequences of the Crisis
R2
Obs
.20
103
.16
31
.25
76
.21
92
.20
95
.42
42
108
Other Global Recessions: 2001-02
All
Only High
Income
Drop
Advanced
Drop Oil
Drop FCs
Drop Poor,
Oil, FCs
CA % GDP
‘00
-.03
(.04)
.05*
(.02)
-.05
(.06)
-.06
(.07)
-.05
(.05)
-.06
(.05)
Cred Mkt
Reg ‘00
-.36
(.26)
.55*
(.22)
-.49
(.33)
-.35
(.29)
-.41
(.28)
.71*
(.34)
Ln rl GDP pc
’00
-.42
(.31)
-3.77**
(.82)
-.33
(.57)
-.43
(.32)
-.45
(.33)
-.90*
(.36)
Causes and Consequences of the Crisis
R2
Obs
.09
112
.44
39
.07
80
.10
101
.10
102
.10
49
109
Results Quite Sensitive to Time
• Most Successful variables do not have stable
(let alone significant) coefficients
• Credit Market Regulation: “best” result for
Great Recession (negative, often significant)
– Positive, often significant, for earlier crises
Causes and Consequences of the Crisis
110
Summary and Conclusion
Causes and Consequences of the Crisis
111
Examine causes and consequences
of Great Recession
• MIMIC Methodology allows us to explicitly
confront fact that “crisis severity” observed with
error
• (But OLS delivers same results)
• Look at a broad set of potential causes commonly
cited in literature
• Success in such a cross-sectional specification
necessary, but not sufficient, condition for
reliable EWS
– Only trying to explain relative incidence, not timing
Causes and Consequences of the Crisis
112
Results Disappointing for Proponents
of Early Warning Systems
• We were able to model incidence of crisis well,
but not link it to observable causes
• Not an artifact of MIMIC
– Scatter plots, principal factor, OLS exercises yield
similar results
• Disappointing, given voluminous analysis of
causes of observed crisis
• Ex ante for future crises: much harder!
Causes and Consequences of the Crisis
113
Why are results so weak?
• Not Data limitations
– Initially collected in spring 2009; crisis was still developing
– Still, measures of incidence of crisis plausible, correlated
– Updated data deliver same results
• Focus mostly on national characteristics
– Inappropriate if crisis international due to contagion or
common shock – but few sensible signs of contagion when
included (American exposure helps!)
– Vulnerability to common shocks typically viewed as
dependent on national fundamentals
• Causes may differ across countries
– Japan: little financial exposure, but downturn in trade
Causes and Consequences of the Crisis
114
Some Causes seem to
“work” for others
• But Positive Findings sensitive to:
– Measures of Crisis Intensity
– Sample of Countries
– Time Period
Causes and Consequences of the Crisis
115
Message
• Natural to generalize from salient facts, but
often inappropriate
– Credit growth high pre-2008 in Australia, Canada,
South Africa – but small crises
– Housing Prices fell pre-2008 in Japan, Germany,
Portugal (no bubble!) – but big crises
Causes and Consequences of the Crisis
116
Reasons for Failure Bode Poorly for EWS
• Difficult to create EWS model that accounts for global
shocks or contagion
• Need successful model for both time-series and cross-section
• Don’t know where next global shock will originate
• Also difficult for EWS models to allow for parameter
heterogeneity across groups of countries
• Ehrmann, et al (2009) demonstrate that exposure to U.S. partly
explains (negatively) performance of equity portfolios in crisis
• But difficult to identify idiosyncratic vulnerabilities in advance
• Since crisis and response ongoing, this is “early warning”
about ability of EWS to predict performance in similar
crises
Causes and Consequences of the Crisis
117