that, in a number of cases, the null hypothesis of interdependence would be
erroneously accepted when adopting those ‘adjusted’ or ‘corrected’ correlation
tests proposed by the literature, arbitrarily setting λ = 0.
The paper is organized as follows. Section 2 presents a few stylized facts
about stock market returns in the nineties, comparing their behavior during
crisis and tranquil periods. A notable point here is that financial crises are
characterized by an increase in the variance and covariance of returns across
markets, but not necessarily by an increase in correlation. Section 3 introduces
a factor model and derives a general empirical test. Section 4 discusses the
existing literature in light of our model. Section 5 carries on a test of financial
contagion from the the Hong Kong stock market crisis in October 1997, as a
representative case study.
2 Stylized facts
We start our analysis by presenting a set of stylized facts regarding the trans-
mission of shocks across stock markets.3 We single out four ‘stylized facts’
characterizing periods of international financial turmoil in our sample. The
first two are well understood and extensively discussed by the literature. These
are the concentration of sharp downward adjustments in stock prices and the
sharp increase in average volatility of daily returns. The other two are often
and somewhat surprisingly confused in both formal and informal discussions
of contagion: crises are systematically associated with a sharp increase in the
cross market covariance of assets’ returns; yet the direction of the change in
cross market correlation of asset returns is not homogeneous across countries
and crisis episodes — in several cases correlation actually drops during a crisis
relative to tranquil periods. This is more than a technical point, as it raises the
issue of assessing the relative importance of country-specific factors, as opposed
to global factors, underlying the increase in market volatility during periods of
turmoil.
Our data set includes 18 countries: the G7 countries, Argentina, Brazil,
Mexico, Russia, Hong Kong, Indonesia, South Korea, Malaysia, Philippines,
Singapore and Thailand. We use daily and weekly data, from January 1990
to March 2000; the source is Thomson Financial Datastream. For each stock
market in our sample, we examine levels and volatility of returns, calculated in
local currency, as well as covariance and correlation patterns with other mar-
kets. We allow for four periods of crisis in international financial markets: from
September 1992 to August 1993 (hereafter ERM crisis), from October 1994 to
June 1995 (hereafter Mexican crisis), from July 1997 to January 1998 (hereafter
Asian crisis), and from May 1998 to March 1999 (hereafter Russian/Brazilian
crisis ). The emphasis of the study is however on stock markets of emerging
economies during the second half of the 1990s.
3In a companion paper, we present empirical evidence for nominal exchange rates against
the U.S. dollar, overnight interest rates, and sovereign spreads of U.S. dollar denominated
bonds with corresponding U.S. assets (see Corsetti et al., 2000).