the summer of 1998.6
In industrialized countries, stock market volatility increases gradually from
1990 to 1999, with the exception of Japan, where it decreases, and France, where
it shows no trend. In most countries, volatility peaks in 1990-92, then decreases
until 1997, when it peaks again, reaching historical heights in 1998.
2.1.3 Covariance between stock market returns increases during cri-
sis periods.
Covariance of weekly returns is presented in Figure 2a-2d. Figure 2a confirms
that Asian countries are relatively unaffected by the Mexican crisis. Although
covariance is never nil during this crisis (as is in most tranquil periods), its level
is often lower than the peaks recorded before and after the crisis. Instead, the
impact of the Asian and the Russian/Brazilian crises on cross-country comove-
ments of stock returns is much higher. Covariance between weekly returns of
Indonesia, Korea, Malaysia, the Philippines and Thailand is record-high during
the Asian crisis, diminishes somewhat shortly after, and reaches new peaks in
1998-99. It comes back to pre-1997 levels only by the end of 1999. Covariances
between each of these five countries with Hong Kong, Japan, Singapore and the
U.S. follow a very similar pattern (Figures 2b and 2c).
In Latin America, covariances between returns of Argentina, Brazil and Mex-
ico sharply increase sharply during the three episodes of crisis in the second half
of the 1990s (Figures 2d). Comovements of returns in Latin American coun-
tries with the United States are not significantly different from tranquil periods
during the Mexican crisis, but are quite strong during the Asian and the Rus-
sian/Brazilian crises. Finally, covariances of Latin American countries and the
United States with Russia (for which data is available only from January 1996)
recorded sizable increments during the Asian and the Russian/Brazilian crises.
2.1.4 Correlation between stock market returns is not necessarily
larger during crisis periods than during tranquil periods.
Figures 3a to 3d show correlation coefficients of weekly returns for the stock
markets in the sample. For Asian countries, a first notable piece of evidence is
that, during the Asian crisis, correlation remains below or at the same level of
the peaks recorded between 1995 and 1997. That is to say, correlation is not
significantly higher during crisis periods than during tranquil periods.
A second notable piece of evidence is that correlation is on an increasing path,
both after the beginning of the Asian crisis, and after the Hong Kong crash in
October 1997. However, one cannot identify an analogous pattern during other
episodes of crisis. For instance, correlation across Asian stock markets during
the Russian/Brazilian crisis, either remains stable or decreases. By the same
token, there is no single correlation pattern during the ERM and Mexican crises.
In Latin America, correlation between the stock markets of Argentina, Brazil
and Mexico increases during the Mexican, Asian and Russian/Brazilian crises;
during the same crisis episodes, correlation of Latin American countries with
the United States increases as well. The magnitude of correlation between the
0Volatility of sovereign spreads followed a similar pattern during the period. It strongly
increased in 1997 and in 1998, then gradually decreased in 1999 (see Corsetti et al., 2000).