Russian and the U.S. stock markets has gradually increased between 1994 and
2000.
As for industrial countries, correlation of the U.S. stock returns vis-à-vis
France, United Kingdom, Italy and Canada is rising from the low values recorded
in 1993-95, reaching a peak in 1999. Somewhat surprisingly, the correlation of
both the United States and the European countries with Germany decreases
from 1990 until end 1998. No clear trend is observable in the correlation between
the Japanese and the US stock prices.
2.2 A synthesis through a case-study
Figure 4 below presents a case-study that summarizes well the typical patterns
emphasized in our analysis above. The figure shows the pattern of volatility,
covariance and correlation for Hong Kong and the Philippines. In order to
disentangle the largest price movements, we also show an indicator of price
reversal, calculated as the ratio between the value of the stock market in period
t and its maximum value up to period t (xt∕max{xh}th_0) — called CMAXi.
While the Mexican crisis has a limited impact on most Asian countries, stock
prices in both Hong Kong and the Philippines record some decline and a rise
in volatility. Cross-market linkages between the two countries at first record
a decrease in covariance and correlation, due to the sharper movements in the
Hong Kong prices. Then, both covariance and correlation follow an inverted V.
During the Asian and the Russian/Brazilian crises, the drop in prices as
well as the increase in volatility and covariance are quite striking. In particular,
covariance between the two markets rises from nil to its decade-record high, with
a sharp step up around October 1997, when the Hong Kong stock market crisis
erupts. Covariance remains on high levels until the first quarter 1998, then
decreases somewhat, before rising again in correspondence with the Russian
turmoil. Correlation increases steadily during the Asian crisis although it does
not appear significantly larger than in 1996; it decreases somewhat between May
and September 1998 and is fairly stable thereafter.
In light of the stylized facts discussed above, what strikes market participants
as evidence of contagion is the magnitude of asset price movements occurring
more or less simultaneously in different regions of the world, as measured by
the dramatic increase in covariance and volatility. Correlation seldom rises
above the level recorded in tranquil periods, even during ‘extreme’ episodes of
international transmission of shocks.
3 A factor-model approach to the analysis of
contagion
3.1 The model
This section lays out a simple factor model to approach the issue of testing for
structural breaks in the international transmission mechanism. For the purpose
of comparison with the current literature, in what follows we focus on correla-
tion analysis, casting our argument in the framework of a single factor model.
A meaningful generalization of our argument to multi-factor models is best ac-
complished without using correlation-based tests — a task that is left to future