Should investors reduce the ‘home bias’ of their portfolios by diversifying into overseas markets? The standard argument in favour of international investment is that it lowers risk without sacrificing return. Stocks and bonds traded in the same national market tend to move up and down together since they are affected in similar ways by the national business cycle, economic growth and interest rate movements. By diversifying overseas, an investor can hope to reduce some of that national market risk.

But there will only be benefits from international diversification if asset returns differ widely across markets, that is, if markets move independently of each other. Their degree of independence is measured by the correlation coefficient: zero implies total independence while one implies that market movements are perfectly synchronized. Over the last four decades, the correlation coefficients between the US equity market and other major markets have been 0.52 with the UK, 0.44 with France, 0.39 with Germany, and 0.27 with Japan. The correlation between two typical stocks in the same country is much higher, suggesting that there is room for successful diversification.

However, recent studies show that correlation is not constant but varies considerably over time depending on the relative importance of domestic and worldwide forces in national markets. In some periods, when there are no global economic shocks, equity markets are primarily affected by country-specific factors. But at other times, all markets are affected by the same global forces, for example, the oil shock of 1974, the Gulf War of 1990 and, in the late 1990s, the turmoil in emerging markets.

These studies have also linked correlation to market volatility, the idea being that markets are more synchronized in times of high turbulence, perhaps because the dominance of global factors is associated with volatile markets. New research by François Longin and Bruno Solnik – ‘Extreme Correlation of International Equity Markets’, CEPR Discussion Paper No. 2538 (August 2000) – questions these findings. These researchers argue that the standard technique for comparing asset returns across markets can give the impression that there is higher correlation during volatile periods than during tranquil periods even though the true correlation is constant.

Longin and Solnik believe that the key to determining whether correlation really increases during volatile periods is to focus on what they call ‘extreme correlation’ – the correlation between returns in either the negative or positive tail of the distribution of asset returns. Using monthly data on the five largest stock markets from 1958–96, they find that correlation is not related to market volatility per se but rather to the market trend. Correlation increases in bear markets but not in bull markets.