1 Introduction
In the past few years, currency and financial crises originating in one country
or group of countries have often spread internationally. In periods of instabil-
ity, asset price movements and comovements across markets and across borders
have increased visibly relative to more tranquil periods. The size of these co-
movements during crises have led many economists to raise the question as of
whether ‘tranquil periods’ and ‘crises’ are to be interpreted as different regimes
in the international transmission of financial shocks; that is, as of whether there
are discontinuities in the international transmission mechanism.1
The headline of the theoretical and policy debate on this issue is usually ‘con-
tagion’.2 Contagion - as opposed to ‘interdependence’ - conveys the idea that
international transmission mechanism is discontinuous, as a result of financial
panics, herding, or switches of expectations across instantaneous equilibria. Al-
though there is considerable ambiguity on what contagion exactly is and how we
should measure it, several authors have proposed empirical tests in an attempt
to address the issue of contagion versus interdependence on empirical grounds
(see Forbes and Rigobon (1999a and 1999b), Boyer et al. (1999), among others).
The idea underlying these studies is to compare cross-market correlation
in tranquil and crisis periods and define contagion as structural breaks in the
parameters of the underlying data generating process. Suppose that a crisis
is caused by shocks to some global factors in the world economy. For a given
mechanism of international transmission, changes in the volatility of asset prices
in one market can be expected to be correlated with changes in volatility in other
markets. During a period of financial turmoil, some comovements across markets
are therefore an implication of interdependence. Contagion will instead occur
when the observed pattern of comovement in asset prices is too strong, relative
to what can be predicted when holding constant the mechanism of international
transmission.
Using a simple factor model, this paper presents a critical assessment of the
empirical literature on correlation analysis of contagion. Key to this literature
is the specification of a theoretical measure of interdependence, suitable to cap-
ture the international effects of an increase in the volatility of asset prices for a
given transmission mechanism. We show that many leading contributions de-
rive such measure by (implicitly) making a specific yet arbitrary identification
assumption about a key parameter. This is the ratio between the variance of
the country-specific shock and the variance of the global factor weighted by its
factor loading; we refer to it as the ‘variance ratio’, and denote it by λ. Tests
that are conditional on a low value of λ, tend to accept the null hypothesis
of interdependence, while tests that are conditional on a high value of λ tend
to reject the null of no contagion. We provide some estimates of λ suggesting
1The possibility of such discontinuities are a concern for both investors and policy mak-
ers. If correlation across assets is abnormally high during financial crises, diversification of
international portfolios may fail to deliver exactly when its benefits are needed the most. By
the same token, excessive comovements of asset prices may spread a country-specific shock to
other economies, even when these have better domestic fundamentals.
1A partial list of contributions to this debate include Baig and Goldfajn (1998), Bordo and
Mushid (2000), Buiter et al. (1998), Calvo (1999), Calvo and Mendoza (1999), Caramazza
et al. (2000), Claessens et al. (2000),Edwards (1998), Eichengreen et al. (1998), Jeanne and
Masson (1998), Kaminsky and Reinhart (2000), Kaminsky and Schmukler (1999), Kodres and
Pritsker (1999), and Schinasi and Smith (1999).