The Motivation Behind International Diversification for Portfolio Risk Reduction

It is often argued that the main benefit of international diversification is the reduction in portfolio risk rather than the increase in returns. The idea behind this concept lies in the fact that if investors randomly select stocks from a large market such as the NYSE the risk, measured by the standard deviation of portfolio returns, can substantially decrease as investors add foreign assets to their domestic portfolio. After a strong period of stock market gains last year, global equity fund managers shift their attention to developed markets which have been plunged during the financial crisis seeking for diversification benefits in their portfolios. Due to the time-varying nature of international diversification, a significant number of research papers tries to evaluate whether there has been an improvement in the risk-return trade off for investment portfolios in a number of ways such as hedging foreign returns with forwards and futures. Taking these into consideration, it would be worth reviewing if there are any benefits of diversification from the interaction of risk-adjusted returns with correlations(Elton and Gruber,1995) extending to conditional correlations in order to provide a dynamic structure between the markets using bootstrapping approaches(Efron,1979 and Hacker,2006)

LITERATURE REVIEW

International diversification benefits

The existence of the benefits of international portfolio diversification as provided by portfolio theory is documented by Elton and Gruber (1995) who constructed portfolios based on Markowitz(1958) and depicted the relationship between overall expected portfolio return and the riskiness of this return by producing efficient portfolio sets which have the minimum variance(risk) for a given expected return. Based on the weekly price data of Morgan Stanley Capital International for the period 1970-2000,they calculated risk-adjusted returns based on the Sharpe ratio and the minimum returns required for diversification to be beneficial. The tangency portfolio that was drawn according to the market equilibrium model by Sharpe(1964) had a large amount of proportions in Japan accompanied by low correlations of Japanese market with the US market. Also, despite the fact that Japan yields a higher risk-adjusted return from the point of view of an American investor than the UK does, it still pays to diversify in UK markets as long as the correlation is low. This is also confirmed by the Sharpe ratio of the UK market which is above the minimum required. This diversified portfolio consisting of the three markets yields a higher risk-adjusted return than the portfolio investing only on the US market(no diversification).However, there is one study (Hanna et al, 1999) that suggests that there are no benefits for a US investor who invests in the equity markets of Canada, US, France, Germany, Italy and Japan during the period 1988-1997.The findings were reported based again on risk-adjusted returns and correlations between the US and these markets which were not low enough to reveal any gains from diversifying. In other words, they found that the risk-adjusted returns of a portfolio consisting solely of the S&P 500 were considerably higher than those of the other G7 countries during the same period. Depending each time on whose point of view is taken, the results can be different and foreign returns can become lower than domestic returns.

Emerging markets and developed markets

The principle of diversification suggests that as long as the returns on the foreign assets do not have a perfect positive correlation with the domestic assets, investors can reduce but not eliminate the overall risk of their portfolio by altering the proportions held in the assets. Specifically, foreign asset returns have much lower correlations with domestic assets than the correlations among each of the domestic assets. In this paper, there is examined a number of variables that may explain the benefits from international diversification, if any. Specifically, benefits have been documented due to the low level of correlation among national equity markets (Grubel,1968 and Levy,1970) but there is a disagreement on the variables that cause this low correlation meaning that these low correlations may not be a true indication of potential gains. One of the papers that raised this issue, followed by several studies sharing the same view was that of Lessard(1974) who focused on common characteristics among returns within countries rather than across countries in order to determine possible gains. His key results were that national factors are more important than industry factors in defining securities sharing common characteristics and contribute more to risk portfolio reduction. In general, country and industry shocks have different impact on the returns of a portfolio because expected future cash flows are widely affected by economic activity, trading activity and liberalization of financial markets. As a result, the degree of economic integration of investment activities plays a significant role in explaining country and industry returns.

Liberalization and diversification benefits in emerging markets

New capital markets emerged when developing markets rebounded in the early 1990s followed by capital flows(fixed income and equity),foreign investments and a number of macroeconomic measures leading to a financial liberalization process which theoretically enables investors to embrace diversification benefits through emerging market integration with the global capital market. In reality, however, integration can increase the correlation between emerging market and world market returns leading to a reduction of possible gains from diversification toward emerging markets. Also, the extent to which the liberalization process can bring up diversification gains depends on the dynamics of capital flows in emerging markets. Specifically, many studies showed that there are significant benefits for investments in emerging markets because they yield returns that are less correlated with returns in the developed world(Divecha et al,1992).Thus, contrary to popular belief, investing in emerging markets can lead to lower risk since investment barriers had diminished and most importantly industry specific factors dominate country factors because stock returns in the emerging markets are more homogeneous than in developed. This view is also shared by Cavaglia et al(1994,1995) who casts doubts on earlier studies of Lessard(1974) by choosing to diversify across all industry sectors rather than across countries based on broad market indices. This optimal portfolio yields a lower risk than the broad market index portfolio since the correlation between industry returns in different countries is lower than the correlation between the returns on the broad market indices. In general, results suggest that industry factors are more important in countries whose industries depend on multinational productions and common production technologies because they have a high degree of industrial integration. This in turn suggests that industry shocks are important in explaining international portfolio returns. On the other hand, country factors seem to be important in countries whose economic activity is not affected by world economic activity. Although these studies make reasonable inferences based on the variables and determinants of country and industry returns, they are mainly driven by stock market behaviour which changes over time thus producing different results. A relevant study that was pioneering was that of Errunza et al(1999) testing the hypothesis that multinational corporations provide international benefits .While Agmon and Lessard(1977) showed that US investors can recognize benefits in holding multinational stocks and reduce the overall risk of their portfolios, Solnik(1978) concluded that multinational stocks are a 'poor substitute' of foreign traded securities. Errunza et al(2002) further showed that benefits can be obtained by investing in closed-end funds in emerging markets because these funds are often traded at a premium when their underlying assets are invested in closed or restricted markets. This means that the returns from holding the funds instead of their assets are different providing potential gains to investors. The same view is also shared by Bekaert and Urias(1996) who used closed-end funds and mean-variance spanning tests to depict potential international diversification benefits. They tested whether a set of asset returns when added to a four-index benchmark leads to a leftward shift in the mean-standard deviation frontier(reduction of risk).This research differentiates from previous studies in taking into consideration investment costs that previous studies ignored, thus making more strong the existence of diversification benefits in emerging markets.

International CAPM and weaknesses

The findings of Lessard based on 16 developed countries have documented the importance of local monetary policies and regulations as the key determinants of stock returns. The existence of these factors led to a multi-factor process generating returns named the international capital asset pricing model(CAPM).Using the international CAPM to estimate the expected gains from international diversification to a US investor according to Lessard poses difficulties in the choice of the world factor and if the true world market factor is used, a structure based on a single index will be misleading since national factors can not be fully diversified thus affecting overall expected returns. As a result, Agmon(1972) and Solnic(1973) integrated different national markets into a single multinational capital market. Solnik(1974) used an equally weighted portfolio of a wide set of possible stocks to determine how effective is international diversification in reducing risk by considering 20 randomly selected securities. He found that US investors could obtain almost full capacity of benefits from diversification by including up to 20 securities in their domestic portfolio achieving the minimum level of market(non-diversified risk) in each country chosen. The traditional CAPM postulates a linear relationship between the expected return on any asset and the covariance between that asset and the return on a world-wide portfolio. Also, it assumes that the market risk premium is the only relevant factor in international markets that are integrated and is restricted to be positive. That is because otherwise no risk-averse investor would willingly hold the stock when he could earn more by investing in the safe risk-free asset. However, during periods of high interest rates there may be a negative risk premium for which the model is unable to predict expected excess returns. Along with this, due to the increasing integration and globalization of the markets there is evidence that correlations between any two markets change over time (Engle and Wooldridge,1988).Their pioneering research involved the concept of time-varying covariance with the market as a more realistic measure for the expected excess return of assets. For this reason, they employed the parsimonious(simple) GARCH model which enabled them to test the restrictions of the model simultaneously on a large number of assets where asset returns depend on multiple risk factors. Dumas and Solnic(1995) and De Santis(1999) encompassed currency risk in addition to market risk. Specifically, Solnic's study was pioneering in considering perfectly hedging foreign currency returns by using a FX-forward contract but this holds only if you know for sure how much you will receive in one year's time in foreign currency. So, the drawback of this study is the assumption of perfect hedging as the actual return from holding foreign stocks is uncertain and thus the hedge will not be perfect. It is noteworthy that until 1973 the exchange rate was fixed (Bretton Woods), so the reduction of the non-diversifiable risk was not different when taking into account exchange rate fluctuations. However, nowadays the hedged international diversification strategy would not result in the same reduction of portfolio risk if exchange rate changes are considered. For this reason, Glen and Jorion,(1993) showed that currencies appear to play an important role in global portfolios using mean-variance tests and adjusting for different exchange rates.

Benefits in emerging markets

Specifically, the approach used to calculate the optimal portfolio weights is similar to that used in the international CAPM but includes the forward premium which changes over time. Their findings suggest that conditional hedging strategies which take into account the time variation of hedge ratios yield higher returns without additional risk whereas with unconditional strategies there is little evidence of benefits from adding currencies in equity portfolios except for portfolios including bonds where the static currency hedging improves the risk-return trade-off. However their tests suggest benefits when investing in some individual Latin American or Asian countries but not when investing optimally in the combination of emerging markets which again confirms the weaknesses of the CAPM. This fact is attributable to the loss of power in mean-variance tests when more emerging markets are included(De Roon et al(2001).Papers examined by Glen, Jorion, De Roon et al make the assumption that short-selling is not allowed but they show that benefits disappear after imposing such restrictions,a matter that will be discussed later in the paper using the Bayesian inference. Further to the studies of Solnic(1995),De Santis and Gerard(1998) who acknowledged the pricing of currency risk premium, Claessens et al(1995),Carrieri and Majerbi(2005) found evidence of long-term risk premium considering an emerging-market-based portfolio that consists of equities, the returns of which are measured using the Morgan Stanley's Capital International World Index. The analysis includes seven Latin American countries. The factors that they consider is the home currency in which returns are measured and due to the fact that optimal weights change as forecasts of expected returns, variances and covariances change, they measure these variables in-sample data and out of sample data similar to the bootstrapping method used by Eun and Resnick(1988) .For a fully hedged portfolio, the risk-premium earned with respect to the local risk-free rate is equal to the risk premium with respect to the foreign risk-free rate. Thus the variance of a fully hedged global portfolio is just the variance of its dollar returns under the assumption that foreign currency rate will be paid with certainty(Claessens et al,1995).After analysing the long-term hedging decision ,they concluded that for open economies with flexible exchange rates currency hedging increases volatility followed by an increase in expected returns because hard currencies act as a 'natural hedge' against portfolio losses(Claessens et al,(1995),Carrieri and Majerbi(2005).Also, this holds due to the fact that there has been an upward trend in currency betas which correspond to the contribution of the local currency to the volatility of the global portfolio based in a developed market. Higher betas mean that the local currency is riskier for an international investor as long as the volatility of the exchange rate remains constant. When considering exchange rate volatility, the overall risk of the global portfolio increases as countries allow currencies to fluctuate and not be fixed as analysed in Brazil, Colombia and Mexico in 1999(Fischer,2001) when currency betas have tripled compared to 2000-2004.Again the conclusion reached is that hard currencies act as natural hedges against negative returns in global equity(Claessens et al(1995),Carrieri and Majerbi(2005).Further to these studies, Eun and Resnick(1988) showed that for an unhedged portfolio, the gains are better than a solely domestic investment from the point of view of an American investor irrespective of whether the international portfolio comprises of equal weights or the optimal weights given by mean-variance analysis. Specifically, they showed that the overall portfolio risk depends on the covariances among stock market returns and covariances among exchange rate changes. It is noteworthy that they employed two methods of exchange risk reduction namely the multiccurency diversification and hedging through forward contracts. If correlations among exchange rates are negative then fluctuating exchange rates will decrease portfolio risk. Their results suggested an overall risk of 4.8% as opposed to 2% in the absence of exchange rate volatility(page 202).This analysis shows that since exchange rate is nondiversifiable to a large extent there may be gains from using forward contracts to hedge risk. Indeed, they showed that hedging using forward contracts almost always produces better results than not hedging whether investors use the optimal weights or the equal weights. As a result, it appears that there are benefits by investing internationally even for an American investor whose market makes up half of the world stock market capitalization. Nevertheless, investors seem to prefer holding domestic portfolios for fear of government regulations on exchange rates despite the gains that they can achieve by investing internationally.

Unstable correlations of international stock markets

As discussed above, due to the increasing integration and globalization of the markets there is evidence that correlations between any two markets change over time and increase dependant on which is the market trend and not on the market volatility (Longin and Solnick,2002).They suggested that conditional correlation seems to increase in bear markets and not in bull markets which further enabled them to construct the optimum portfolio. In order to conclude reliably that the correlation is changing over time by simpling looking at different values of one or more variables is to specify each time the distribution of the conditional correlation which has not been done so far by previous studies examined. Since, financial time-series are not normally distributed, the parameter estimates from asymptotic tests are biased and the results are not reliable. Therefore, total reliance on correlations due to large volatility of return variables can be misleading. Theory shows that distribution of extreme returns can only converge to a distribution the shape of which is not well defined. Hence, in contrast to previous studies,(Efron, 1979 and Hacker,2006) use a bootstrapping approach with leveraged adjustments and new evidence is provided. They examined an international portfolio consisting of the US,UK and Japan markets as Elton and Gruber(1995) did but instead of using asymptotic tests, they applied causality tests. In detail, they took into account the dynamic structure between markets by using the bootstrap correlation coefficients which were measured by causality tests proposed by Granger(1969).The basic concept which differentiates this study from previous ones is that the degree of causality can indicate the size of the international diversification benefits that can be achieved. Granger tested whether movements in one variable follow movements in another variable and found that there is no evidence of causality between the three markets. This finding was reported in bootstrap correlation coefficients which were the same as the standard ones that Elton and Gruber(1995) estimated. Thus, there is additional support that international diversification can still provide gains to investors. This study can also be expanded to accommodate more markets and can also be done from the point of view of investors from other markets than the US one.

Constraints in international diversification

In all studies discussed so far, the results were reported without taking into consideration constraints such as short-selling and it was mentioned in the previous section that investors tend to hold a substantially larger proportion in domestic stocks because they fear of government regulations and find difficulties in taking short positions in many non-US equity markets. One advantage of the analysis undertaken by Elton and Gruber(1995) was that the measurement of international benefits was independent of expected returns since risk-averse investors can not forecast expected returns and seek to minimize the riskiness of their portfolio. That is why they measured the standard deviation of the global minimum-variance portfolio. Another advantage of this method is that the estimated weights may be more stable over different sample periods than returns. The surprising result is that they found diversification benefits after imposing short-sale constraints in emerging equity markets. However, the integration of world equity markets reduces but still there are benefits from diversifying in emerging markets after short-sell constraints are imposed. The results produced for portfolio efficiency subject to such constraints are examined through the Bayesian approach which unlike the asymptotic tests provides reasonable results. The sampling period examined is from 1976-1999.One measure of the diversification benefits employed follows the work of Kandel et al (1995) and Wang (1998) on portfolio efficiency. In general, it has long been argued that constraints on short sales could be a possible reason that the market portfolio is inefficient. However, Wang (1998) surprisingly found that when investors take large short positions(above 50%) can form a portfolio that dominates the market portfolio(e.g NYSE) by over 20% in annualized returns without incurring higher risk whereas when short positions are constrained to be less than 50%,the annualized returns fall. A general intuition is that investors will take a large positive position in an asset when they have favourable beliefs about the volatility of the returns and vice versa. As a result, the expected return of an efficient portfolio is sensitive to investors beliefs and constraints on portfolio weights prevent extreme long and short positions, thus reducing the variation in the efficient frontier. Thus, the expected return on the US equity index portfolio is either smaller or equal to the expected return on the internationally efficient portfolio(Kandel et al,1995 and Wang,1998).The difference between the expected returns are used to measure the size of benefits. Li et al (2003) also use a Bayesian approach and argue that international benefits from investing in emerging markets remain substantial even in the presence of short-sale constrains. However, De Roon et al (2001) argue the opposite with one exception when investing in some individual Latin American countries. Also, asset management industry supports that there are still substantial benefits but there has not been any formal econometric inference. In order to get a better understanding of the effect of short-sell constraints there has been reported a table with the means and standard deviations of the weights in the international portfolio that has the same variance as the US equity index. For the G7 countries the benefits before and after imposing short-sell constraints remain the same and for this reason short-sell constraints on emerging markets should not have a big impact whereas there is notable difference in developed countries such as Canada for which the optimum weights are zero when short-sell constraints are imposed. The standard errors of the portfolio weights of the developed countries indicate that there are massive errors when estimating portfolio weights meaning that the standard errors of the weights make the confidence intervals large(huge sampling variability).The standard deviation interval of the optimal weights on US is below 60% supporting the idea that when emerging markets and short-sell constraints are considered, institutional investors can have the came benefits by taking short positions on developed countries as well. As a result, small errors in those parameters can lead to huge mistakes on how investors should invest their wealth. The only substantial short position in emerging markets is on Singapore at around 11%.These results are also consistent with the results reported by Britten-Jones (1999). It is also notable that when constraints are imposed only on emerging markets, individual markets such as Brazil provide benefits similar to those when portfolio weights are unconstrained whereas when weights are constrained to be nonnegative Argentina and Chile do not offer any benefits when added to developed countries. This evidence determines that diversification benefits come from individual emerging markets and not from the combination of them. The concept of this analysis is that there are international diversification benefits after imposing short-sell constraints but the prediction of future benefits coming from emerging markets is not evident under the static analysis and for this reason the analysis is more reliable when dynamic diversified portfolios are considered.

Conclusion and recent data

Most prior studies identify the existence of international diversification benefits by showing the tangency portfolio and simultaneously by examining the correlation or integration between markets. The critical issue addressed is the extent of diversifications benefits as measured by causality tests between markets based on bootstrapping method rather than on asymptotic tests. Thus, dynamic structure between the markets provides significant contribution to the existing literature suggesting that investors can reap benefits from international diversification but it remains difficult for global stock picking especially the last two years that stock markets in the world have gone through a harsh depression. Since the correlation among international equity markets is higher than used to be in the past implies that investors should not hold a portfolio invested in different countries if their domestic portfolios yield higher risk-adjusted returns. However, the environment is now 'normalising' which means that investors should seek for global companies and especially in developed countries with long-term growth prospects. Also, investors can hedge currency risk if the prospect of the sterling strengthening against the dollar continues during this year.

Cite this page