Does the Euro affect the dynamic interactions of stock markets in Europe? Evidence from France, Germany and Italy

Abstract

The dynamic links between stock market indices are analyzed in a GARCH-M framework, using daily data from France, Germany, Italy and the USA. It is shown that indices in the periods before and after the introduction of the Euro as a single currency display a very distinct behaviour. Consistent with the literature, in the earlier period price changes are found to have an impact the next day on other markets. In the latter period this type of co-movement disappeared within Europe. Feedback trading has been shown to induce (negative) autocorrelation in national stock markets. In this paper an international version of the feedback trading model is used to illustrate that the lead–lag relationships across countries and the strength of these links depend on the currency regime.

Keywords:

• Euro
• feedback trading model
• GARCH model
• stock market indices

Notes

¹Earlier studies on the correlation of national stock markets, including Levy and Sarnat (1970), Grubel and Fadner (1971), Hillard (1979), found only little evidence of co-movement in data from the 1960s and 1970s.

²Note that predictability implied by this equation does not mean that the efficient market hypothesis is violated. As the strength of autocorrelation is linked to the volatility of the index, it may not be worthwhile to exploit a forecastable rise in the index.

³In the case of national holidays in one of the countries, this data point was eliminated from all series.

⁴Modeling stock price dynamics as a simple VAR-process may violate the assumptions of OLS estimation such as constant variance of the errors.

⁵In spite of the drawback in the interpretability of the USA coefficient, it is important to include the USA in the regression. If a relevant variable as the US markets are not included, the remaining variables in the regression may be biased.

⁶In order to further assess the relative importance of price shocks in different countries, we also decomposed the stock market price index forecast error variances from a VAR in first differences into parts that are attributable to shocks emanating from the stock indexes in France, Germany, Italy and the USA.

In the first time period, news from the German stock market had the strongest impact on all European markets during the first two days after a shock. Even larger than shocks from the US market, which has the strongest effect on Germany. On a longer horizon (from the third to the seventh day), The US shocks explain the largest fraction of non-domestic forecast error variance. While in France and Germany most volatility is attributable to domestic shocks at this forecast horizon, the external US shocks are more important that the domestic ones for Italy. This is consistent with the findings of Knif and Pynnonen (1999), who argue that this evidence reflects the fact that not all information born in one market is processed efficiently in another market during the same day.

In the second period, the results changed again consistent with the out earlier findings. The Impact of the German stock market has decreased substantially. At shorter horizons, each country's stock market is now mainly explained by news coming from the domestic market. On longer horizons the USA has become the most important external market in all cases. Results of the VAR exercise are available from the Authors upon request.

⁷Parameter values in the variance equation are highly significant for most countries and the variance in the mean equation is significant for France at the 10% level in the latter period. This is consistent with most modern models of asset price dynamics.

⁸The results are unaffected, if the week before and after the introduction of the Euro are excluded from the regression. They are therefore not due to irregularities during the changeover.