Title: Does Openness to Trade Make Countries More Vulnerable to Sudden Stops, or Less? Using Gravity to Establish Causality
1Does Openness to Trade Make Countries More
Vulnerable to Sudden Stops, or Less? Using
Gravity to Establish Causality
- Eduardo A. Cavallo
Jeffrey A. Frankel - CID and KSG Harvard University
NBER and KSG Harvard
University - Second Workshop of the Latin American Finance
Network - Cartagena, Colombia, December 3-4 2004
2Motivation
- Textbook Result Countries that trade a smaller
share of GDP are prone to larger swings in the
real exchange rate in the aftermath of external
shocks (expenditure-switching) and/or to larger
reductions in spending (expenditure-reduction). - One type of external shock is sudden stops in
capital flows large and unexpected fall in
capital inflows Calvo (1998). - An equal-sized shock triggers, ceteris paribus, a
larger real exchange rate depreciation and/or
more expenditure reduction in a country that
trades a smaller share of GDP than in an
otherwise identical country Calvo et al. (2003),
Cavallo (2004).
3Motivation (Cont.)
- Real exchange rate depreciations are costly in
terms of output-loss in countries with balance
sheets problems Balance Sheet Literature
Krugman (1999).... Sharp reductions in spending
are also painful. - In more open economies sudden stops are less
costly (i.e. less contractionary ex-post)
Sachs (1985), Guidotti et. al. (2003). - Butdoes this make sudden stops more or less
likely? - There is no reason, a priori, why something
(openness) that makes the consequences of sudden
stops better (less contractionary) should also
necessarily make them less frequent.
4The Questions
- (1) What is the effect of trade openness on the
vulnerability to sudden stops implemented by a
probit model measuring the probability of a
sudden reduction in the magnitude of net capital
inflows Calvo et. al. (2003). - (2) What is the effect of trade openness on the
vulnerability to currency crashes implemented by
a probit model measuring the probability of a
sudden devaluation Frankel and Rose (1996).
5The Problem
- Potential identification problem the endogeneity
of trade. - Possible channels of endogeneity of trade
- Via income richer countries tend to liberalize
Frankel and Romer (1999). - Via Washington Consensus forces.
- Experience with crises might itself cause
liberalization. - Feedbacks between trade and financial openness
Aizenman and Noy (2004).
6Proposed Solution
- Use of gravity estimates as instrumental
variables for trade quantities. - Gravity estimates are the predicted trade to
GDP ratios, where the prediction is based on
geographical characteristics of the countries. - Being based on geography, gravity estimates are
quite plausibly exogenous, yet they are highly
correlated to the true trade to GDP ratios.
7Empirical Investigation
- We test whether countries that trade more are
(all else equal) more or less prone to sudden
stops in capital flows or to currency crises. -
- SSi,t c f(Trade Openness)i,t ?1(Foreign
Debt/GDP)i,t-1 ?2(Liability Dollarization)i,t-1
? (CA/GDP) i,t-1 ? Z µi,t - Given that the dependent variable is binary (0,1)
and endogenous explanatory variables the method
of estimation is IV probit Newey (1987). - Dataset is a stacked cross-section (141
countries, 1970-2002).
8Main Variables
- Sudden Stops binary (0,1). A sudden stop occurs
during the year in which there is a noticeable
reduction in the current account deficit that is
accompanied by a disruptive (i.e., recessionary)
reduction in foreign capital inflows. Data
IMF-IFS. - Currency Crashes binary (0,1). Foreign Market
Pressure Index fall in reserves fall in
the value of the currency. The index measures the
fall in demand for the countrys currency. A
crisis episode is defined when there is an
increase in the index of at least 10 over the
preceding period with an exclusion window of 3
years. Data Frankel and Rose (1996) updated in
Frankel and Wei (2004).
9Main Variables (cont.)
- Trade Openness
- Instrumented variable Trade to GDP ratio (XM /
Y).
Data WB-WDI. - Instrument aggregate of bilateral gravity
estimates.
Data Andrew Rose website. - Liability Dollarization
- Foreign Liabilities / Money.
Data IMF-IFS (Line 26C/Line
34). - Foreign Currency Deposits / Total Deposits.
Data Arteta (2003).
10Sudden Stops and Currency Crashes are less
frequent in open economies
11...and the pattern is even more marked with the
instrument
12Results (Dependent variable Sudden Stops)
Probit IV Probit IV Linear IV-GLS RE (linear)
Openness t -0.53 (0.259) -2.451 (0.813) -0.066 (0.022) -0.066 (0.026)
Foreign Debt / GDP t-1 -0.080 (0.217) 0.196 (0.275) 0.0066 (0.0182) -0.006 (0.0155)
Liability Dollarization t-1 0.316 (0.195) 0.591 (0.256) 0.027 (0.0169) 0.027 (0.0149)
Current Account / GDP t-1 -4.068 (1.297) -7.386 (2.06) -0.317 (0.10) -0.317 (0.095)
Obs. 778 1062 1040 1040
Statistically significant at 10, 5, and
1
Additional Controls Constant term, Year FE,
Regional Dummies, International Reserves / Months
of Imports, Institutional Quality, GDP per
capita, Short Term Debt, FDI/GDP, Dummy for
Nominal Exchange Rate Rigidity.
13Results (Dependent variable Currency Crashes)
Probit IV Probit
Openness t -0.57 (0.269) -1.73 (0.918)
Foreign Debt / GDP t-1 0.23 (0.231) 0.59 (0.373)
Liability Dollarization t-1 0.027 (0.249) 0.18 (0.234)
Exchange Rate Rigidity Index t-1 0.13 (0.094) 0.22 (0.113)
Ln Reserves in Months of Imports t-1 -0.26 (0.082) -0.37 (0.099)
Obs. 557 841
Additional Controls Constant term, Year FE,
Regional Dummies, CA/GDP, Institutional Quality,
GDP per capita, Short Term Debt, FDI/GDP.
14Conclusions
- Countries that trade less are, ceteris paribus,
more prone to sudden stops and to currency
crashes. - Raising the Trade / GDP by 10 percentage points
(Argentina to Australia) reduces the probability
of a sudden stop by approximately 32. - At good times closed economies can borrow, but
at bad times investors seem to anticipitate
that closed economies will suffer more in the
aftermath of a shock and thereby are more likely
to attack them (self-fulfilling pessimism). - To be safer remain open to trade.