Are Crises Threatening the Benefits from International Portfolio Diversification?

Only available on StudyMode
  • Download(s) : 59
  • Published : October 21, 2012
Open Document
Text Preview
Are crises threatening the benefits from international portfolio diversification?

This essay examines whether rising international stock market correlations during market crises are weakening the efficacy of modern portfolio theory, which promises benefits from international portfolio diversification. The importance of the combining assets that are not perfectly correlated is discussed, as are reasons why we might expect markets to move more closely during crises. If it is clear that markets are moving more closely together during crises, investors are losing diversification benefits at the most critical time. However, because of assumptions, it is apparent that there is considerable debate over the validity of the correlation coefficient for use in explaining the co-movement of market returns. It may be that markets are moving closely together at all times. Also, over the long run, short term rises in correlations may have negligible effects for the investor. For these reasons, we cannot discredit the worth of international portfolio diversification.

The correlation coefficient is a key statistic for devising the optimal portfolio. In accordance to Modern Portfolio Theory, where risk is to be minimized for a given level of return, the correct combination of assets relies critically on the correlation between those assets. When defining the risk of a portfolio as its standard deviation, when assets that are less that perfectly correlated are combined, the standard deviation of the portfolio is actually less that the weighted average of each individual asset’s standard deviation, although the expected return from a portfolio is simply the weighted average of the expected return of each individual asset (Bodie, Kane, Marcus, Perrakis, &Ryan, 2008, p.208). This is why combining assets with correlations of less than one (proper diversification) is referred to as a ‘free lunch.’ It is also apparent that this principle also extends to international portfolio diversification, where market returns are less that perfectly correlated.

From examining the correlation data from Table 2, the financial crisis of the late 2000’s seems to be yet another example of international stock market correlations rising during bear markets. The selective correlations of the UK’s FTSE 100 Index, Japan’s Nikkei 225 Average, and Brazil’s Bovespa Index to the S&P 500 come from simple fifteen year data series’ on monthly index returns. All correlations were higher from 2008 to March 2009 (the substance of the S&P 500’s decline during the financial crisis) relative to correlations for the entire sample period, and even higher relative to correlations for the bull run of 2003-2008. This correlation behaviour may create misgivings for an investor who values his ‘free lunch,’ and raises the questions: do markets move more closely during crises and why ?

When using the correlation coefficient to measure the comovement of market returns, we are making assumptions about the underlying data (stock market returns.) An important assumption is that the underlying data follow a normal distribution. A study by Campbell, Forbes, Koedijk, and Kofman (2007) examined annualized mean return data for five stock indices from 1990-2005 and found that each data set failed to conform to normality, by displaying significant skewness and kurtosis. They also test conditional correlation estimators under normality and under the student-t distribution, and find that “earlier studies may have overestimated the excess in conditional correlation by assuming bivariate normality,” and that under the student-t distribution excess conditional correlation ceases for the left tail (bear markets) of returns (para.4). Furthermore, Forbes and Rigobon (2002) assert that the correlation coefficient is further biased as an estimator of market comovement because of the heteroskedasticity of market return data (volatility changes), while the correlation coefficient assumes...
tracking img