Transcript On the Empirics of Sudden Stops

“Does Openness to Trade Make
Countries More Vulnerable to Sudden
Stops, or Less? Using Gravity to
Establish Causality”
Comments
Alejandro Izquierdo
Second Meeting of the
Latin Finance Network
December 4, 2004
In a Nutshell





Thorough empirical paper that builds upon
results by Calvo et al (2004)
Reassuring that it confirms relevance of
openness as a determinant of SS
Does a nice job in controlling for endogeneity
using a gravity model (Calvo et al use a
Rivers-Vuong approach to control for
endogeneity of a similar variable)
Goes over a painstaking set of robustness
checks to show the resilience of openness
across specifications
Openness seems to be an important
determinant both for SS and currency crises
Definition of Crisis








Looking for a credit crisis?
Originated in systemic capital market factors?
In markets that are financially integrated (or
broader definition)?
Currency crisis?
Costly crisis?
Timing will vary
Determinants will vary
Groups of countries affected will vary
Sudden Stops and Large
Depreciation
In % of total
Emerging
Markets
Developed
Economies
Depreciations associated with Sudden Stop
Of which: First Sudden Stop, then depreciation
First depreciation, then Sudden Stop
63
42
21
17
9
9
Depreciations not associated with Sudden Stop
37
83
Note: The total number of large devaluations is 19 in emerging markets and 23 in developed
economies. From Calvo et al (2004)
Measuring SS




SS is a large and unexpected event
Is mean and volatility that of the whole
time span, or that prevailing at time t-1?
“Costly” criterion: Ruling out positive
shocks.
Criticism: This may bias the set of
determinants (“disqualifies” short-term
policies:not much can be done when
output collapses)
Measuring SS


Calvo et al (2004) stress systemic factors:
change in SS definition to include swings in
regional spreads, besides large capital flow
reversals
Interpretation: The probability of a full-fledged
SS depends on both the probability of a systemic
shock (or incipient SS) and the conditional
probability of a full-fledged SS, given an
incipient SS:
P(SSF) = P(SSF/SSI) P(SSI)

The empirical exercise amounts to finding
determinants of this conditional probability
Balance Sheet Effects




Interaction between prices and dollarization
In principle, potential price effects are missing
But are they? CADt-1/GDPt-1 is a proxy
When CADt is driven down to zero (what a
country cannot avoid), given Y and S (fixed):
CADt = Zt
CADt-1 / Zt-1 = – Zt / Zt-1

Combining this with standard homothetic
preferences, and a constant supply of nontradables:
rert = ( / ) CADt-1 / Zt-1 = ( / ) (1-t-1)
t = (Yt – St) / Zt (openness a la Calvo et al)
Balance Sheet Effects




Why CAD/GDP and not CAD/Z?
Openness controlled via gravity, how about
CAD/GDP?
Interesting that both traditional measure of
openness and CAD/GDP (proxying for the
leveraged portion of the current account deficit)
come out significant
What additional factors is the openness measure
capturing besides potential price effects? Are
more open economies subject to a quicker
response from tradable sectors after RER
depreciation?
Dollarization





Measure 1: Lacks dollar deposits (FB/M)
Measure 2 (D*/D+D*): relative measure of
degree of deposit dollarization, but what about
size?
Is (D*+FB)/GDP more indicative of potential
contingent liabilities (a proxy for dollar loans
assuming bank currency matching)?
Interesting that dollarization measure comes up
significant in some specifications with SS, but
not with currency crises (as in Arteta (2003))
Are credit crises linked to balance sheet effects,
but not currency crises (developed countries)?
Linear Probability Estimates

Probit has interactions built up given non-linear
specification
But linear estimations should specify interactions
(CAD/GDP and dollarization)
Not controlling for
endogeneity of w
Controlling for
endogeneity of w
(A)
(B)
1.00
Probability of a sudden stop

1.00
Low DLD
Average DLD
High DLD
0.75
0.75
0.50
0.50
0.25
0.25
0.00
0.00
0.75
1.00
Omega
1.25
1.50
0.75
1.00
Omega
1.25
1.50
Domestic Policies - Controls





Domestic variables like debt, lag of reserves,
effectiveness of government do not come up
significant (similar to Calvo et al)
Criticism: This in part reflects that when GDP
collapses, there may not be much room left for
policies.
Measure replacing output fall criterion for
regional spreads criterion yields similar results
Domestic policies don’t matter? Measures such
as openness and liability dollarization may
represent summary statistics of past poor trade,
fiscal and monetary policies
Could control for differences between EMs and
developed countries (EM dummy), and external
variables such as TOT.
Output Loss Estimations



Openness measure is less robust
But this is consistent with the fact that
output outcomes depend on the crisis
resolution process and the associated
transfers that go with it.
Edwards (2004) for example, finds that
dollarization, interacted with a dummy for
current account crisis is not significant in
explaining output behavior
“Does Openness to Trade Make
Countries More Vulnerable to Sudden
Stops, or Less? Using Gravity to
Establish Causality”
Comments
Alejandro Izquierdo
Second Meeting of the
Latin Finance Network
December 4, 2004