* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download International diversification of investment portfolios
Private money investing wikipedia , lookup
Environmental, social and corporate governance wikipedia , lookup
Socially responsible investing wikipedia , lookup
Short (finance) wikipedia , lookup
History of investment banking in the United States wikipedia , lookup
Investment banking wikipedia , lookup
Financial crisis wikipedia , lookup
Exchange rate wikipedia , lookup
Investment fund wikipedia , lookup
Currency intervention wikipedia , lookup
Stock exchange wikipedia , lookup
Bond University ePublications@bond School of Business Discussion Papers Bond Business School 12-1-1991 International diversification of investment portfolios: U.S. & Japanese perspectives Cheol S. Eun Bruce G. Resnick Follow this and additional works at: http://epublications.bond.edu.au/discussion_papers Part of the Finance Commons Recommended Citation Eun, Cheol S. and Resnick, Bruce G., "International diversification of investment portfolios: U.S. & Japanese perspectives" (1991). School of Business Discussion Papers. Paper 67. http://epublications.bond.edu.au/discussion_papers/67 This Discussion Paper is brought to you by the Bond Business School at ePublications@bond. It has been accepted for inclusion in School of Business Discussion Papers by an authorized administrator of ePublications@bond. For more information, please contact Bond University's Repository Coordinator. BOND UNIVERSITY School of Business DISCUSSION PAPERS "International Diversification of Investment PorHolios: US & Japanese Perspectives" Cheol S Eun & Bruce G Resnick DISCUSSION PAPER NO 18 December 1991 University Drive, Gold Coast,QLD,4229 SCHOOL OF BUSINESS DISCUSSION PAPERS Bond University was established by Act of Parliament in 1987 as an independant, prIvate University. The fIrst student intake occurred in May 1989. The School of Business offers degrees in the undergraduate (B Com and Diploma) and the graduate (MCom, MBA and PhD) levels. The School teaches and sponsors research in accounting, economics, econometrics, finance, marketing, management, organisational behaviour and related disciplines in hospitality and real estate fields. The Discussion Paper series is intended to foster research and comments are invited. The views expressed in the papers are the opinion of the authors and do not necessarily reflect the views of the School or the University. Lists of available Discussion Papers and copies of the papers (which are free of charge) may be obtained from: The Senior School Administrator School of Business Bond University University Drive GOLD COAST QLD 4229 Telephone: (075) 952244 Fax: (075) 951160 Dean: B o Professor Ashley W. Goldsworthy AO OBE N D u N I v E R s I T y INTERNATIONAL DIVRESIFICATION OF INVESTMENT PORTFOLIOS: U.S. AND JAPANESE PERSPECTlVES* by Cheol S. Eun ** and Bruce G. Resnick*** July 1991 *The authors are grateful to Salomon Brothers, Inc. for providing the bond data in this study. **The Wharton School, University of Pennsylvania, Philidephia, PA 19104-6367, U.S.A. College of Business and Management, University of Maryland, College Park, MD 20742, U.S.A.' ! ***School of Business, Bond University, Gold Coast, QLD 4229, Australia and School of Business, Indiana University, Bloomington, IN 47405, U.S.A, : INTERNATIONAL DIVERSIFICATION OF INVESTMENT PORTFOLIOS: U.S. AND JAPANESE PERSPECTIVES Abstract In this paper, we analyze the gains from international diversification of investment portfolios from the Japanese as well as the U.S. perspectives. The major findings of the paper include: First, the 'potential' gains from international, as opposed to purely domestic, diversification are much greater for U.S. investors than for Japanese investors. For U.S. investors, the gains accrue not so much in terms of lower risk as in terms of higher return, and the opposite holds for Japanese investors. Second, using various 'ex ante' international investment strategies designed to control parameter uncertainty, U.S. investors can realize substantial gains from international diversification in out-of-sample periods. Japanese investors, however, can gain little. Third, hedging exchange risk generally allows the U.S., but not Japanese, investors to benefit more from international diversification. For U.S. investors, the international bond diversification with exchange risk hedging offers a superior risk-return trade-off than the international stock diversification, with or without hedging. INTERNATIONAL DIVERSIFICATION OF INVESTMENT PORTFOLIOS: U.S. AND JAPANESE PERSPECTIVES I. Introduction Reflecting the trend toward a greater integration of world capital markets, international diversification of investment portfolios has recently received widespread attention at both the academic and practitioner levels. Originally, Grubel (1968) extended the concept of modern portfolio analysis, pioneered by Markowitz (1952) and Tobin (1958), to global markets. He argued that international portfolio diversification is the source of an entirely different world welfare gain, distinguishable from both the gains from trade and the productivity gains from international factor movements. This insight provided the stimulus for a series of studies, such as Levy and Sarnat (1970), Solnik (1974), and Lessard (1976), which collectively established a convincing case for international portfolio diversification. More recent studies, including Eun and Resnick (1988) and Jorion (1985), have shown i) that hedging foreign exchange risk can potentially increase the gains from international diversification, and ii) that it is important to control parameter uncertainty in order to capture the potential gains from international diversification. In other words, investors can substantially benefit from international diversification when they properly control foreign exchange and parameter uncertainties. When neither of these uncertainties are controlled, however, investors may not be able to realize enough of the potential benefits to justify international investment. Instead, they should invest domestically. It is pointed out that the previous literature was mostly focused on international diversification of stock portfolios. Despite the fact that the international bond market is at least as large as the international stock market in tenos of market capitalization value and is perhaps more integrated than the latter, international diversification of bond portfolios has received much less attention. This seems to mirror the fact that, in general, more empirical work has been done applying modem portfolio theory to the equities market than the fixed-income market. Moreover, the empirical studies of international bond diversification that have been done are still preliminary and not in complete agreement with one anotheLFor example. Levy and Lerman (1988) show in an ex post study that a U.S. investor who diversified across world bond markets could have earned more than twice the mean rate of return on a U.S. bond portfolio, at the same risk level. Similarly, Jorion (1987) showed that over the ten year period ending May 1987, a world valueweighted index of government bonds produced superior risk-return performance in comparison to a U.S. government bond index. Additionally, he showed that a hedged equal-weighted index would have produced about the same mean return as the U.S. government bond index, but with less than half the volatility. In contrast, Burik and Eunis (1990) claim that the risk and return characteristics on non-dollar bonds raise questions regarding their role in diversified portfolios of U.S. investors. They claim that U.S. investors receive no reliable compensation for bearing the currency risk inherent in foreign bonds. Moreover, while hedging can reduce the volatility due to exchange rate changes, the additional costs reduce expected return materially. They conclude that foreign bonds are a diversification opportunity many U.S. investors can afford to pass up. In addition, most of the previous literature examined the issue from the viewpoint of U. S. investors. Relatively little is known about international portfolio diversification from the perspective of non-U.S. investors. As a result, it is not clear at present whether or not and to what extent the general findings of the literature are applicable to non-U.S. investors. In this study, we analyze the gains from international diversification from the Japanese as well as the U.S. perspective, and make comparisons between the results obtained. Our analysis 2 encompasses both the international stock and bond markets. As is well known, Japan emerged as the world's largest creditor nation during the 1980s, heavily investing in financial securities of other nations. Specifically, the main objectives of the paper are to: i) evaluate the 'potential' gains from international diversification from the Japanese and U.S. perspectives; ii) analyze the effect of exchange rate uncertainty on international bond and stock portfolios, and iii) evaluate the out-ofsample period performance of a1temative 'ex ante' investment strategies designed to control for parameter uncertainty, both with and without exchange risk hedging. The organization of the paper is as follows. In Section II, we conduct an ex post analysis of the gains from international diversification of bond and stock portfolios from U.S. and Japanese perspectives, without considering the problem of parameter uncertainty. In Section III, we investigate the effect of exchange rate uncertainty. In Section IV, we evaluate the performance of various ex ante diversification strategies. Section V offers a summary and concluding remarks. ll. The Gains from International Diversification: An Ex Post Analysis In this section, the potential gains from international diversification are determined by solving for the optimal international (tangency) portfolios and then comparing their risk-return characteristics to those of domestic portfolios. Seven major markets are considered: Canada (CA), France (FR), Germany (GE), Japan (JA), Switzerland (SW), the United Kingdom (UK), and the United States (US).l In solving the optimal international portfolios, monthly return data for national bond and stock market indices from the period of 1978.1 through 1989.12 are used. The stock market return data are from Morgan Stanley Capital International Perspective and the bond market data are returns on Salomon Brothers World Government Bond Indices which are comprised of intermediate term bonds. From the historical return data, we first computed the mean, standard deviation and 3 correlation matrix in terms of both the U.S. dollar and the Japanese yen. The results are presented in Table I. A few points are noteworthy. First, stocks have higher mean returns than bonds in both the dollar and the yen; in fact, the highest return bond market, i.e., Japan, has a lower return than the lowest return stock market, Switzerland. Of course, risk tends to be higher for the stock markets than for bonds markets. It is also noted that in both numeraire currencies, Japan (Switzerland) registered the highest (lowest) return among international bond markets as well as among stock markets. Second, the 'inter-correlatiollS' among bond and stock markets tend to be somewhat lower than the 'intra-correlations' among stock markets or among bond markets. Specifically, in terms of the dollar (yen), the average value of the inter-correlations among bond and stock markets is .31 (.25) while the average value of intra-correlations is .51 (.37) for the bond markets and .44 (.41) for the stock markets. It is noted, however, that the inter-correlation is quite high between the bond and stock markets of the same country, with the exception of the numeraire currency country, which most likely reflects the same currency factor. Third, the correlation matrix confirms the well known fact that correlations tend to be high among the continental European countries, i.e., France, Germany and Switzerland, and also between the two North American countries, Canada and the U.S. Additionally, Table I presents the mean return and standard deviation of each national bond and stock index in the respective local currency numeraire. Comparison of the local currency returns with the corresponding dollar returllS implies that over the sample period the dollar appreciated against the Canadian dollar, French franc and British pound, and depreciated against the West German mark, Japanese yen and Swiss franc. A similar comparison of the local currency returns against the yen returns indicates that the yen appreciated versus all the others. Using the parameter values provided in Table I as input data, we solved the optimal international (tangency) portfolios for U.S. and Japanese investors. To simplify the analysis, we 4 assume that the monthly risk-free rate is zero. It is also assumed that investors can take short positions and use the short sale proceeds. Table II presents the compositions of optimal international portfolios. The optimal portfolios for U.S. investors ( or dollar-based investors) are examined first. In the optimal international bond portfolio, the U.S. receives the highest weight, 62.54%, followed by France (44.88%), Japan (28.38%), ,and the U.K. (8.95%); both Canada and Germany receive a weight of about 2 % while Switzerland receives a large negative weight, 48.96%. In the optimal stock portfolio, again the 'U.S. market receives the largest weight, 47.92%, followed by Japan (39.45%), U.K. (7.70%), Germany (5.50%) and France (5.09%); Switzerland receives a minuscule weight, .06%, and Canada receives a negative weight, -5.79%. In the optimal mixed portfolio of bonds and stocks, the largest positive weights are given to the U.S. bonds (51.45%) and Japanese stocks (36.5%), followed by the French and German bonds and U.S. stocks. When the Japanese yen is used as the numeraire, on the other hand, the Japanese market receives the largest weight in the optimal bond portfolio as well as the optimal stock portfolio. This reflects the high returns and low risks, in terms of yen, of the Japanese securities? Apart from the Japanese markets, the French and U.S. markets receive positive weights in the optimal international portfolios of the Japanese investors. It is also noted that the German securities tend to receive negative weights in the Japanese optimal portfolios, while the opposite is true in the U.S. optimal portfolios. This, of course, reflects the effect of numeraire currency. Table II also provides the risk-return characteristics of the optimal in~ernational portfolios, together with those of the domestic portfolios. As can be expected, Table II shows that each international portfolio has a higher Sharpe ratio (SHP), I.e., reward-to-variability, than the corresponding domestic strategy. This, of course, implies that investors, both U.S. and Japanese, can potentially benefit from international diversification; the potential gains, however, are much greater for U.S. investors than for Japanese investors. 3 5 Table I Summary Statistics of the Monthly Returns to Bonds and Stocks: 1978.1-89.12' (In U. S. Dollar and Japanese Yen) Stocks Bonds CA FR GE JA SW UK US CA FR GE JA .36 Canada 047 045 .83 -.06 .02 Switzer. .36 .68 U.K. 045 .37 France Germany (%) SO (%) ME (%) SO (%) .39 .25 .25 .14 .35 .32 .35 .88 3.91 .58 4.99 .90 3.29 .89 .68 .81 .52 .30 .08 .58 .51 044 .58 .29 .03 .83 4.06 045 3.23 .91 1.68 .64 .89 .56 .35 .12 .50 .58 .38 .62 .28 .00 .79 4.64 Al 3.81 .55 1.61 .66 .51 .27 .05 .39 .34 .73 .36 .25 -.02 1.07 4.97 .61 1.74 .61 1.74 .54 .30 .10 043 ,50 Al .65 .28 -.04 .55 4.60 .16 3.50 .28 1.07 .33 .28 .31 .31 .34 .33 .62 .10 .94 5.26 .57 4.78 .97 2.98 .32 .86 3.20 .56 4.60 .86 3.20 .25 -.17 .21 040 .50 .23 .49 .38 .23 .27 .19 .17 042 .31 -.09 046 -.08 .24 -.04 .32 -.10 .56 .25 .43 -.18 .35 -.14 .54 -.15 .15 .21 .12 Switzer. Al .56 .58 -.10 .61 .25 .39 046 .52 .74 .04 U.K. .44 .29 .26 -.09 .23 .63 .36 .67 044 .44 .18 U.S. .64 .31 .22 -.23 .15 .29 .67 .80 Al .40 .13 Japan ME (%) .76 .26 Germany SO 040 .59 France ME (%) .34 .43 Canada US .27 .86 Stocks UK Local 040 .11 .83 043 .40 U.S. SW Yen U.S. Dollar Bonds Japan Dollar i _. .02 .08 .37 -.11 .33 .39 .21 .18 .26 .14 .32 .20 040 .27 .22 .37 .60 .71 1.39 6.14 1.10 7.16 1.39 5.54 .61 043 .58 Al 2.06 7.73 1.66 7.13 2.13 6.56 .37 .76 048 044 .30 1.50 6.62 1.13 6.36 1.28 SAl .34 .35 .20 2.13 6044 1.66 4041 1.66 4041 045 .38 1.33 5.90 .96 5.66 1.08 4.47 .53 1.69 6040 1.35 6.55 1.76 5.41 1.34 4.60 1.07 6.13 1.34 4.60 045 049 .62 'The upper-right (lower-left) triangle provides the correlation matrix in dollar (yen) terms. ME and SO, respectively, denote the mean return and standard deviation of returns. Table III Decomposition of the Variance of Bond/Stock Returns in U.S. Dollars' (Monthly Data: 1978.1-89.12) Components of Var(R;s) Var(ej) Var(R;s) Var(R;) Canada 15.29 10.82 1.72 2.67 .08 France 16.48 2.82 12.74 .60 .32 Germany 21.53 2.59 13.84 4.91 .19 Japan 24.70 3.03 15.13 6.09 .45 Switzer. 21.16 1.14 17.64 2.34 .04 U.K. 27.67 8.88 12.39 6.08 .32 U.S. 10.24 10.24 .00 .00 .00 Canada 37.70 30.58 1.72 5.37 .03 France 59.75 43.03 12.74 3.75 .23 Germany 43.82 29.27 13.84 .00 .71. Japan 41.47 19.45 15.13 5.83 1.06 Switzer. 34.81 20.07 17.64 -3.76 .86 U.K. 40.96 29.27 12.39 -1.52 .82 U.S. 21.16 21.16 .00 l'he variances are computed using monthly percentage returns. .00 .00 2Cov(Ri,ei) AVar Bonds Stocks Examination of the Table II indicates that for Japanese investors, the gains from international diversification accrue in terms of lower risk, not in terms of higher return. For U.S. investors, on the other hand, the gains accrue not so much in terms of a lower risk as in terms of a higher return. Consider, for example, the case of stock investment. By holding the optimal international portfolio instead of the U.S. stock market index, U.S. investors can substantially increase the mean return from 1.34% to 1.72% and, at the same time, moderately reduce the risk from 4.60% to 4.19%. Japanese investors, on the other hand, can reduce the risk from 4.41 % to 3.65% at the cost of reducing the mean return from 1.66% to 1.55% by holding the optimal international portfolio, rather than the Japanese stock market index. m. The Effect of Exchange Rate Uncertainty Suppose that U.S. investors invest in the ith foreign market. Then the dollar rate of return, R;$, is given by = 1>. .l'i + e. + 1>·e· .I.'i I (I) l' where Ri is the local currency rate of return on the ith market and ei is the rate of appreciation of the local currency against the dollar. 4 The variance of the dollar rate of return can be decomposed as follows: Var(R;$) = Var (R;) + Var (eD + 2Cov(R;,ei) + tSar, (2) where Ll.Var represents the contribution of the cross product term, R;ei' to the variance of the dollar rate of return. Of course, the same formula will apply to the variance of the yen rate of return, Var(R;I), when the variable ei is redefined as the rate of appreciation of the local currency against the yen. Table III presents the decomposition of the variance of dollar returns into different components during the sample period of 1978.1 - 89.12, while Table Iy presents the same decomposition for Japanese yen. It is clear from examination of the tables that exchange rate 6 uncertainty substantially increases the risk of foreign investment, whether bonds or stocks, regardless of the numeraire currency. However, it is also noted from the tables that the effect of exchange rate uncertainty on the risk of foreign investment is much more significant for bond investment than for stock investment. This is due to the fact that stocks have much higher volatility of local currency returns than bonds. In fact, the variance of exchange rate changes is always greater than that of local bond returns but less than that of local stock returns in both numeraire currencies, with the sole exception of the Canadian bond in dollar terms. Consider, for instance, the case of Swiss bonds. The Swiss bond market has a local currency variance of only 1.14 percent squared but the variance becomes 21.16 percent squared in terms of the U.S. dollar. This dramatic increase in variance reflects the variance of the exchange rate, 17.64 percent squared, as well as the covariance of the exchange rate change with the local bond market return, 2.34 percent squared. Examination of the tables shows that, more often than not, the covariances between the exchange rate changes and local market returns add to the risk of foreign investment. It is noted that the contribution of the cross product term to the risk of foreign investment, AVar, is rather insignificant. Considering that investors are likely to hold multi-currency portfolios, it is useful to extend the above analysis to a portfolio context. The variance of dollar portfolio returns can be written as follows: Var~$) = where Xi I:iI:jXiXjCov(R;,R) + I:iI:jXi:<.iCov(ei,ej) + 2I:iI:jXi:<.iCov(Ri,ej), represents the fraction of wealth invested in the ilb bond or stock market. The exchange rate uncertainty contributes to the overall risk of the portfolio through the second and the third terms of equation (3). To empirically examine the effect of fluctuating exchange rates on the portfolio risk, we construct equally-weighted portfolios ofthe seven markets. Table V provides the variances of portfolio returns and their decompositions. In the case of stock portfolios, exchange rate 7 (3) Table IV Decomposition of the Variance of Bond/Stock Returns in Japanese Yena (Monthly Data: 1978.1-89.12) Components of Var(R;r) Var(R;r) Var(R;) Canada 24.90 10.82 14.75 -1.01 .34 France 10.43 2.82 8.58 -1.08 .11 Germany 14.52 2.59 9.86 2.02 .05 3.03 3.03 .00 .00 .00 Switzer. 12.25 1.14 10.43 .69 -.01 U.K. 22.85 8.88 11.42 2.93 -.38 U.S. 21.16 10.24 14.59 -3.91 .24 Canada 51.27 30.58 14.75 5.96 -.02 France 50.84 43.03 8.58 -.38 -.39 Germany 40.45 29.27 9.86 .68 .64 Japan 19.45 19.45 .00 .00 .00 Switzer. 32.04 20.Q7 10.43 1.16 .38 U.K. 42.90 29.27 11.42 2.45 .24 U.S. 37.58 21.16 14.59 liThe variances are computed using monthly percentage returns. 2.11 -.28 Var(e;) 2Cov(R;,e;) AVar Bonds Japan Stocks Table V Decompostiion of the Variance of Returns of the Equally-Weighted Portfolios" (Monthy Data: 1978.1-89.12) Components of Portfolio Risk VarCRp) E E (l/NiCov(Rj,R} I J E E(l/NiCov(ej,e} I J 2 E E (llNiCov(Rj,e) I J U.S. Dollar Bonds 11.31 2.61 (23.2%) 6.03 (53.5%) 2.63 (23.3%) Stocks 20.52 14.13 (69.9%) 6.03 (29.9%) .05 (.2%) Bonds & Stocks 12.49 5.01 (40.5%) 6.03 (48.7%) 1.34 (10.8%) Japanese Yen Bonds 7.48 2.61 (35.1 %) 5.55 (74.6%) -.73 (-9.7%) Stocks 19.92 14.13 (70.6%) 5.55 (27.8%) .33 (1.6%) Bonds & Stocks 10.38 5.01 (48.3%) 5.55 (53.6% ) -.20 (-1.9%) 'The portfolio variances are computed using the monthly percentage returns. The relative contribution of the individual components to the total portfolio risk appear in parentheses. uncertainty accounts for about 30% of portfolio risk in terms of both the dollar and the yen. It is noted that the cross-covariances among local market returns and the exchange rate changes contribute relatively little to the risk of stock portfolios. In the case of bond portfolios, on the other hand, the exchange rate uncertainty accounts for about 77% of the variance of dollar portfolio returns and 65% of the variance of yen portfolio returns. It is noted that the crosscovariance terms contribute significantly to the risk of dollar bond portfolio, but reduce the risk of yen bond portfolio. In the case of mixed portfolios comprising both bonds and stocks, the exchange rate uncertainty accounts for about 59 % (52 %) of the variance of dollar (yen) portfolio returns. The preceding analysis indicates that exchange risk is not very much diversifiable in the context of multi-currency portfolios, and that hedging exchange risk can potentially increase the gains from international diversification, especially in the case of bond investment. This observation leads one to consider the use of foreign exchange forward contracts as a tool for exchange risk management. As an example, assume that the U.S. investor sells the expected foreign currency proceeds forward. In dollar terms, it amounts to exchanging the uncertain dollar return (1 + E(R;»(l + e) - 1 for the certain dollar return (1 + E(RD)(1 + fD - I, where E(Ri) is the expected rate of return on the i'h foreign stock market in terms of the foreign currency and 11 is the forward exchange premium. The unexpected foreign currency proceeds, however, will have to be converted into U.S. dollars at the uncertain future spot exchange rate. The dollar rate of return under the hedging strategy is thus given by R j H$ = [1 + E(Rj )](1 + f) + [R j - E(R)J (1 + e) - 1, (4a) (4b) Obviously, parameter uncertainty enters into implementing the hedging strategy described 8 by equation (4) through E(R;). To a degree, the success of the hedge depends on the accuracy of the estimate of E(RJ. Poor estimates of E(RJ will result in poor hedging results. If one assumes that E(Ru cannot be estimated with accuracy, then instead of hedging the expected foreign currency proceeds, an alternative is to hedge only the principal and convert any return in the spot market on the horizon date. This hedging strategy can be described by (Sa) (5b) In what follows, we refer to the hedging arrangement described by equation (4) as 'back-end' hedging and that described by equation (5) as 'front-end' hedging. The key advantage of front-end hedging is that the hedging decision can be made separately from the parameter estimation. 5 Because the third term in equations (4b) and (5b) will be small in magnitude as will be the fourth term in equation (4b), the following approximation results for both hedging strategies: (6) Equation (6) suggests that much of the effect of exchange rate changes on the risk of the foreign stock market investment can be offset by means of the forward exchange contract. Moreover, the empirical results presented in Tables IV and V generally suggest that H Var(R'$) < Var(R'$) and Cov(R,f,Rjf) < Cov(R'$,RjJ since fi is a constant. Additionally, since the forward exchange premium is known to be a nearly unbiased predictor of the future change of the exchange rate, I.e., fi '" E(e,), hedging offers the potential of reducing risk without adversely affecting return. Tables IV and V suggest that the analogous statements hold for the. yen-based investors as well. 6 9 IV. Ex-Ante Portfolio Strategies In this study the ex ante international portfolio strategies we examine, both with and without forward exchange hedging, are the strategies developed by Jobson and Korkie (1980, 1981), Jorion (1985, 1986) and extended by Eun and Resnick (1988). This literature has established that the expected return vector is the critical input for successfully implementing modern portfolio theory, Le., identifying the ex ante 'optima!' investment weights. Conventional estimation of the inverse of the variance-covariance matrix is satisfactory. In lhis study we use the unbiased estimator (see Jobson and Korkie), (T - N - 2) S-1 , (T - 1) (7) where S is the usual N x N sample variance-covariance matrix. Let us examine the unhedged strategies first, using the expected return equation A Ii = (1 - w) A X+ w 1. Yo, (8) where a bar under a variable symbol denotes a vector, Y is the N x 1 ex post (historical) sample mean-return vector of the N assets, 1. is a vector of ones, Yo denotes the mean return from the ex post minimum-variance portfolio, and ~ represents the estimated shrinkage factor for shrinking the elements of Y toward Yo. Equation (8) is a Bayes-Stein expression derived by Jorion for estimating the ex ante expected-return vector to use in solving the portfolio problem. It is, however, general enough to encompass other models. If ~ = 0, the resulting vector of estimated expected returns is the ex post classical sample means. Coupling these estimates with t- 1 results in identifying the weights of the ex post (or historical) tangency portfolio as the ex ante 'optimal' investment weights. This method implicitly assumes no estimation risk,in the classical sample ..estimates and is labeled the certainty-equivalence-tangency (CET) portfplio strategy. A second 10 strategy, which is due to the simulation results of Jobson and Korkie (1980, 1981) is to arbitrarily, set ~ = 1. Combined with t·1, this strategy identifies as the optimal ex ante investment weights those of the ex post minimum-variance portfolio (MVP). The MVP strategy implicitly assumes that there is no useful asset-specific information in Y because it is not required as input to solve the portfolio problem. A third strategy is the Bayes-Stein strategy developed by Jorion, which uniquely estimates the shrinkage factor according to the equation. ~ = -'-(N_+_2,-'-)('-T_-~1).:...., (N + 2)(T - I) + (X - Yol)' TS-'(T - N - 2) ex - --,Yol)' where T represents the length of the time series of the sample observations and S is as defined before. Using ~ in equation (8) and t·1, the Bayes-Stein (BST) optimal ex ante tangency portfolio can be determined. Equation (8) can potentially result in a uniform improvement on the ex post classical sample mean or Yo as estimates of the expected return because it relies on a more general model that includes them as special cases. Whether using the Bayes-Stein estimates in -the portfolio problem results in a more efficient ex ante optimal portfolio is an empirical question. A fourth unhedged strategy is to construct an equally weighted (EQW) portfolio. This approach can be viewed as a naive diversification strategy in the attempt to capture some of the potential gains from international diversification. Alternatively, it can be viewed as if the investor believes there is no useful information in the historical return data that can distinguish one asset versus another. If the investor selects an unhedged strategy, realized returns are defined by equation (I). To implement either the CET, MVP or the BST strategy requires obtaining a historical time series sample of dollar (or yen) returns, R i $ (R,t) (i= I, ... , N), to calculate the Y , Yo, S, and ~ needed for equations (7) and (8). If the investor selects a hedged (H) strategy, realized returns are defined by (4) if a back-end hedging (BH) strategy is implemented and (5) if a front-end hedging (PH) strategy is selected. Equation (6), which approximates equations (4) and (5), suggests that when a 11 hedged strategy is employed, the variability in R¥s (R¥i) will be primarily due to the variability in the local currency return, 1<; (i= 1, ... ,N). This, in turn, suggests that for either the ex ante backend or front-end hedging strategy, the expected-return vector be estimated (using $ as the example notation) as (9a) A E<Ef) = (1 - where Y, Yo, S, and A w) X W, + wI Yo + (9b) 1, and thus R, are calculated from a historical time series of local currency returns, R; (i = 1, ... ,N). The vector f is not estimated from historical data but rather contains as elements the current market-determined forward exchange premiums. Examination of equations (4) and (5) show that R¥s (i;eUS) is dependent on how E(R;) is estimated. For both the BH and FH versions of the EQW(H) strategy, the E(R;) values are each estimated as the grand mean of the Y elements. This is also done for the MVP(H) strategy. For the CET(H) strategy, the E(R;) values are the respective Yj elements. In the BST(H) strategy, the E(R;) values are the respective elements from the vector .R in equation (9a).7 In the next section, we empirically test the unhedged and hedged ex ante international diversification strategies in the dollar and yen numeraires and compare their performance results with one another and with domestic investment in the United States (US) and, respectively, Japan (JA). V. Empirical Results A. The Data and Test Structure We use the Morgan Stanley Capital International Perspective Stock Index Monthly return data and the Salomon Brothers World Government Bond Index Monthly ,return data described in Section II for the seven countries previously mentioned as the primary data. The stock data are in 12 the U.S. dollar numeraire and the bond data are provided in both the U.S. and the local currencies. Using the U.S. and local currency bond returns, a corresponding time series of exchange rate changes versus the dollar and the yen are constructed for each country via solving equation (1) for The return data and exchange rate change data are used to test the performance of the ex ante investment strategies discussed in Section IV. In conducting the tests, it is assumed that the investor has a twelve-month investment holding period. To estimate the 'optimal' ex ante investment-weight vector, it is assumed that the investor has knowledge of the 60 monthly returns prior to the beginning of the holding period. For the hedging strategies, the twelve-month forward premiums are calculated as of the inception date of the holding period from spot and twelve-month forward exchange rates obtained from the Chicago Mercantile Exchange Statistical Yearbook volumes. The twelve-month forward premiums are then converted to monthly premiums to calculate the ex ante expected-return vector and the holding-period returns. We attempt to examine the performance results for each strategy for thirty-six out-Df-sample (overlapping) holding periods using the 144 months of data. The sample periods are structured as follows: For the first holding period covering months 61 through 72, the estimation period covers months 1 through 60. For the second holding period of months 63 through 74, the estimation period covers months 3 through 62. Each subsequent pair of estimation and holding periods is shifted forward in time by two months. This methodology was successful in constructing 36 outof-sample tests when the numeraire curren~y was the yen. For the dollar, however, only 29 out- of-sample periods could be constructed. For the other seven, a CET tangency portfolio on the positively sloped section of the ex post efficient frontier did not exist. 8 Consequently, since all strategies could not be compared, those seven out-Df-sample test periods were eliminated. B. Test Results Tabl VIII presents the performance results for bonds, stocks, ~d combined portfolios of 7 bonds and stocks from employing the various ex ante strategies in the twenty-nine out-Df-sample 13 holding periods using the dollar as the numeraire currency. For each strategy, the table shows the average portfolio mean return and standard deviation stated in percentage per month. The table also shows the average Sharpe (SHP) measure of portfolio performance. Table IX presents the corresponding results for the thirty-six out-of-sample holding periods when the yen serves as the numeraire currency. Before we discuss the performance results for the alternative ex ante international investment strategies, we briefly examine the portfolio weights that produce the performance results. Tables VI and VII present the average portfolio weight vectors for each strategy, both with and without hedging, for the U.S. and Japanese investors, respectively. A few points are noteworthy. First, hedging exchange risk induces both U.S. and Japanese investors to substantially reduce investment in their domestic market and increase investment in foreign markets, especially in the Swiss market. Second, the portfolio weight vectors under hedging for U.S. and Japanese investors look remarkably similar for all asset classes. The reason for this is that regardless of which numeraire currency is used, a particular hedging strategy will identify a similar ex ante optimal investment weight vector to the extent that the forward premiums or discounts are small relative to the local currency mean returns, as they typically will be. B.!. U.S. Results Panel A of Table VIII presents the performance results for the unhedged strategies. Examination of the results show that for all asset classes the three international diversification strategies cbntrolling estimation risk, I.e., MVP, EQW and BST, have higher average SHP values than the corresponding U.S. strategy, and also the CET strategy which does not attempt to control estimation risk. The only exception is the BST strategy for the mixed bond/stock diversification that failed to outperform the U.S. strategy of 50% bonds and 50% stocks. This result shows that as long as estimation risk is properly controlled, U.S. investors can actually realize gains from international diversification. Panel A of Table VIII also shows that for each of the international investment strategies, the 14 Table VI Average Portfolio Weigbts for Out-of-Sample Periods: U.S.' A. Unhedged Approach Tangency Portfolio Minimum-Variance Portfolio Bayes-Stein Market Bonds Stocks Bonds/Stocks Bonds Stocks Bonds/Stocks Bonds Canada .2769 -.2850 .3002/-.6523 -.0590 -.0393 -.2377/.0081 .0826 -.1241 -.1214/-.2598 France -.3187 .0049 -2.1407/-.1382 .3680 -.0537 .1841 -.0282 -.1856/-.0853 Germany .7337 -.0150 2.3896/-.5855 -.3390 .1016 -.2014/.1889 .0804 .0695 .0872/.1912 Japan .8034 .3363 .029511.1988 .0617 .1838 -.2021/.1997 .3946 .2349 -.2536/.5624 -1.3536 .1303 -4.750412.8086 .3052 .1360 .01731.1833 -.4439 .1348 -1.4935/1.1936 U.K. .0700 .1572 .2107/.7309 -.0013 .0652 -.02461.0734 -.0048 .0943 .1126/.2566 U.S. .7883 .6712 .7811/.8177 .6644 .6064 .4724/.4195 .7071 .6188 .50911.4866 Switzer. .1664/-.0632 Stocks Bonds/Slocks B. Hedged Approach Tangency Portfolio Bayes-Stein Minimum-Variance Portfolio Market Bonds Stocks Bonds/Stocks Bonds Stocks Bonds/Stocks Bonds Stocks Canada·· -.0521 -.2130 -.0354/-.0394 -.0757 .0621 -.0797/-.0182 ·.0601 -.0801 -.0539/-.0315 France .1591 .0945 .14041-.0100 .2192 -.0089 .02051.0068 .0921 -.0122 .06621-.0103 Germany .1720 -.0068 .1800/-.0305 .0070 .0540 .0809/~.0019 .1639 .0242 .1707/-.0169 Japan .2159 .3823 .2006/.0647 .1825 .3055 .1541/.0581 .2021 .3639 .1769/.0689 Switzer. .5277 .4805 .44881.0768 .6826 .3856 .7791/.0435 .6522 .5419 .57051.0850 U.K. -.0019 .1972 .01001.0146 -.0253 .0953 -.0293/-.0311 -.0277 .0658 -.00471-.0095 .0096 .1062 -.02241.0396 -.0226 .0965 -.0498/.0384 -.0208 .0654 -.0527/.0322 U.S. llEach. number represents the average of 29 out-of-sample values. Bonds/Stocks Table VII Average Portfolio Weights for Out-Qf-Sample Periods: Japan' A. Unhedged Approach Tangency Portfolio Minimum-Variance Portfolio Bayes-8tein Market Bonds Stocks Bonds/Stocks Bonds Stocks Bonds/Stocks Bonds Canada .0330 -.2555 .0741/-.1027 -.0396 -.0396 -.0550/-.0130 -.0061 -.1121 -.0103/-.0.427 France .2297 .0772 .1817/-.0275 .2239 .0196 .1709/-.0235 .2470 .0415 .1855/-.0284 Germany -.0717 -.0869 -.0113/-.0382 -.2684 .0078 -.21791.0597 -.1841 -.0189 -.1459/.0299 Japan .9241 .6287 .8116/.1374 .7541 .5071 .6271/.1260 .8345 .5472 .7219/.1182 Switzer. -.1882 .2617 -.3191/.1777 .2230 .2852 . 1298/.0549 .0276 .2814 -.0740/.1163 U.K. -.0120 .1324 -.0467/.0773 -.0103 .0598 -.0164/.0141 -.0141 .0835 -.0227/.0326 U.S. .0851 .2424 -.0436/.1293 .1173 .1600 .0083/.1348 .0952 .1775 -.0041/.1238 Stocks Bonds/Stocks B. Hedged Approach Tangency Portfolio Mimimum-Variance Portfolio Bayes-Stein Market Bonds Stocks Bonds/Stocks Bonds Stocks Bonds/Stocks Bonds Stocks C~ada'_ -.0239 -.2189 .0085/-.0710 -.0990 .0310 -.0798/-.0327 -.0385 -.0942 -.0135/-.0696 France .0217 .0796 .2621/.0039 .2528 -.0071 -.0865/.0086 -.1029 -.0281 -.1891/-.0200 Germany .2844 -.0180 .3028/-.0592 -.0142 .0579 .0975/-.0019 .2702 .0270 .2718/-.0264 Japan .3014 .3913 .2069/.1497 .1951 .3171 .1331/.1021 .2636 .3761 . 151l1.1639 Switzer. .4505 .5105 .6605/-.0249 .6599 .4066 .8174/.0923 .7057 .5866 .5576/.2188 U.K. .0645 .2164 -.0237/.0468 -.0273 .0859 -.0204/-.0334 -.0030 .0683 .0563/-.0037 U.S. -.0986 .0390 -.1322/.0200 .0326 .1087 -.03821.0419 -.0951 .0642 8Each number represents the average'ot 36 out-of-sample values. Bonds/Stocks -.1310/.0343 Table VIII Average Out-of-Sample Performance Results of the Ex Aote Investment Strategies: U.S." Bonds Stocks Bonds and Stocks A. Unhedged ME(%) 5D(%) SHP 1.50 6.48 .32 1.65 4.98 .37 6.74 21.10 .31 MVP ME(%) SD(%) SHP 1.20 2.61 ..40 1.86 4.38 .51 2.23 4.71 .41 EQW ME(%) SD(%) SHP 1.29 3/J7 .41 2.07 4.40 .55 1.68 3.08 .54 BST ME(%) SD(%) SHP 1.3.3 3.92 .37 1.80 4.46 .48 3.77 14.68 .35 US ME(%) SD(%) 5HP .91 2.67 .31 1.32 4.84 .33 1.11 3.14 .36 CET B. Back-End Hedged CET ME(%) SD(%) SHP .78 1.09 .74 1.77 4.78 .49 .88 1.27 .72 MVP ME(%) SD(%) SHP .73 .90 .87 1.72 4.39 .51 .86 1.12 .81 EQW ME(%) SD(%) SHP .83 1.37 .63 1.60 4.24 .53 1.21 2.37 .60 BST ME(%) SD(%) SHP .75 1.05 .75 1.79 4.58 .51 .90 1.21 .79 C. Front-End Hedged CET ME(%) SD(%) SHP .79 1.10 .74 1.78 4.77 .49 .89 1.27 .72 MVP ME(%) SD(%) SHP .73 .92 .87 1.73 4.38 .51 .86 1.13 .81 ME(%) SD(%) SHP .83 1.38 .63 1.61 4.23 .53 1.22 2.37 .60 EQW BST "In each cell, ME(%) .76 1.80 SD(%) 1.06 4.56 .74 SHP .51 the three numbers represent the average ot 29 out-ot-sample va!~es. .90 1.22 .79 Table IX Average Out-of-Sample Performance Results of the Ex Ante Investment Strategies: Japan' Bonds Stocks Bonds and Stocks .92 1.15 2.66 .46 .93 2.00 .51 A. Unbedged ME(%) SD(%) SHP 1.96 .53 1.34 4.57 .35 MVP ME(%) SD(%) SHP .65 1.47 .53 1.53 4.21 .47 EQW ME(%) SD(%) SHP .37 2.14 ME(%) SD(%) SHP .78 1.59 1.18 4.38 .37 1,48 4.21 .45 ME(%) SD(%) SHP .75 1.64.59 2.18 5.24 .44 1.46 2.97 .52 CET BST JA .17 .59 .78 2.87 .32 1.07 2.13 .54 B. Back-End Hedged CET ME(%) SD(%) SHP .59 1.28 .49 1.28 5.83 .26 .64 3.81 .21 MYP ME(%) SD(%) SHP .52 .89 .64 1.81 5.30 .40 .14 4.76 .04 EQW ME(%) SD(%) SHP .67 1.33 .52 1.89 5.40 .43 1.04 3.24 .38 BST ME(%) SD(%) SHP .04 4.85 .06 1.45 5.52 .30 4.47 .14 .44 C. Front-End Hedged CET ME(%) SD(%) SHP .59 1.28 .49 1.28 5.80 .26 .64 3.80 .21 MYP ME(%) SD(%) SHP .52 .89 .64 1.81 5.28 .40 .15 4.72 .04 EQW ME(%) SD(%) SHP .67 1.33 .52 1.88 5.37 .43 1.04 3.23 .38 BST ME(%) .04 1.45 SD(%) 4.83 5.50 SHP .06 .30 'In each celt. the three numbers represent the average ot 36 out-ot-sample values. .44 4.46 .14 stock portfolio has a higher average SHP ratio than the corresponding bond portfolio. In addition, the stock portfolio is found to have a higher average SHP ratio than the corresponding mixed bond/stock portfolio. Considering the relatively low correlations between stocks and bonds, this result is somewhat unexpected. This result implies that in the presence of estimation risk involving different classes of assets, expanding the opportunity set to include both stocks and bonds will not always allow the investor to realize a superior risk-return trade-off. A few points are of interest from Panel B, which presents the back-end hedging strategy results. First, the average SHP ratio of the international bond portfolio increases substantially with hedging. For example, the average SHP ratio of the MVP (EST) strategy increases from .40 (.37) without hedging to .87 (.75) with hedging. As a result, each of the hedged international bond investment strategies has an average SHP ratio which is more than twice as large as the SHP ratio of the U.S. bond strategy. In addition, the average SHP ratio of the mixed portfolio of bonds and stocks also increases substantially with hedging, outperforming the corresponding U.S. strategy. Second, examination of the average SHP values for the hedged stock strategies shows that each of the international strategies outperforms the U.S. strategy. However, with the exception of the CET(H) strategy, the average SHP values of the hedged strategies are just about the same size as the corresponding values for the unhedged stock strategies. In other words, hedging internationally diversified stock portfolios is not of much benefit to the U.S. dollar-based investors. This, of course, is in sharp contrast to the case of bond investment. This different hedging result appears to be attributable to the fact that exchange risk is much more significant a factor in bond investment than in stock investment, as can be seen from examination of Table V. Third, once exchange risk is hedged, for each strategy the international bond portfolio has a higher average SHP value than the corresponding stock portfolio. This is in contrast to the unhedged results in Panel A, where for each strategy the stock portfolio has a higher average SHP value than the bond portfolio. It is also found that for the CET(H), MVP(H) and EQW(H) strategies, the bond portfolio has a higher average SHP value than the corresponding mixed 15 portfolio of stocks and bonds. When the results for all three asset classes are considered, the MVP(H) strategy emerges as the best overall strategy among all hedging strategies. Panel C of Table vm presents the performance results of the front-end hedging strategies. Examination of the panel indicates that the results are very similar in all cases to those obtained for the back-end hedging strategies. This suggests that a greater qu,antity risk associated with front-end hedging may be largely diversifiable in the context of a multi-eurrency portfolio. Note that the front-end and back-end versions of a particular hedging strategy have identical ex ante investment weight vectors but that the realized returns are computed differently because the back-end approach calls for hedging the entire expected terminal value, whereas the front-end approach only hedges the principal investment against exchange risk. Since the results are similar, practical considerations suggest that the easier, front-end hedging be used, rather than the back-end hedging that is based on the estimation of the expected future foreign currency position. 9 B.Z. Japanese Results Table IX provides the performance resuits of the alternative international investment strategies for Japanese investors. Perhaps, the most striking finding of the table is that the yenbased investors could not go too far wrong by just investing in Japanese bonds. The average SHP value for investing solely in the Japanese bond market is larger than that for all other unhedged strategies, except the BST strategy which has the same average SHP value. Similarly, the Japanese bond strategy outperforms all the hedged strategies, except the MVP(H) strategy for international bond investment. It is noted from Panel A of Table IX that with the exception of EQW strategy, the bond portfolio has a higher average SHP value than either the stock portfolio or the mixed portfolio of bonds and stocks. This is in contrast to the U.S. case, where the stock portfolio outperformed both the bond and mixed portfolios. In addition, Panel A shows that it is difficult for Japanese investors to substantially benefit from holding internationally diversified stock or mixed portfolios. 16 The Japanese stock or mixed investment strategies performed nearly as well, or better, than the corresponding international strategies considered. Examination of Panel B and C, shows that the back-end and front-end hedging essentially produce the same results. Unlike the case for U.S. investors, however, hedging fails to allow the Japanese investors to achieve a superior risk-return trade-off. This result appears to be attributable to the fact that the hedging strategies call for reduced investment in the best performing Japanese market and increased investment in such weak markets as the Swiss. It is noted that among the unhedged strategies, the EQW strategy performed the worst for the same reason, Le., reduced investment in the Japanese market. VI. Summary and Conclusions As world capital markets have become more integrated, both individual and institutional investors have begun to diversify their portfolios internationally. In this paper, we have evaluated the 'potential' gains from international diversification from the perspectives of both U.S. and Japanese investors, considering bonds as well as stocks. In addition, we have examined the extent to which U.S. and Japanese investors can actually capture the gains from international diversification under the conditions of exchange rate and parameter uncertainties. The major findings of the paper are summarized as follows. First, the ex post analysis of the gains from international diversification indicates that, in the absence of parameter uncertainty, both U.S. and Japanese investors can potentially benefit from international diversification; the magnitude of the potential gains, however, is much higher for U.S. investors than for Japanese investors. For U.S. investors, the gains accrue not so much in terms of a lower risk as in terms of a higher returns. For the Japanese investors, on the other hand, the gains accrue mostly in terms of a lower risk. Second, in the presence of a parameter uncertainty, U.S. investors can capture gains from international diversification by using various ex ante international investment strategies designed to 17 control estimation risk. In contrast, Japanese investors could gain little from employing the ex ante strategies. This appears mainly due to the fact that the potential gains are relatively modest to begin with for the Japanese investors, reflecting the strong performance of the Japanese capital markets. Third, exchange rate uncertainty is an important component of foreign investment risk for both V.S. and Japanese investors, especially for bond investment. Hedging exchange risk is found to generally increase the benefits from international investment for V.S., but not Japanese, investors. It was also found that for V.S. investors, the international bond portfolio with hedging substantially outperformed the international stock portfolio, with or without hedging. This implies that V.S. investors should seriously consider using bonds with hedging as a vehicle for achieving international diversification. Currently, equities are a much more popular instruments for international diversification. 18 FOOTNOTES 1. It is in the currencies of these countries that forward contracts are readily available. 2. French and Poterba (1990) suggest that national investors weight their home country assets more heavily in internationally diversified portfolios out of choice rather then because of institutional constraints. The results presented in Table IT suggest that a large weight in the home country asset(s) is optimal as well. 3. As one would expect, when there are less then perfect positive correlations between bonds and stocks, the combined portfolio of stocks and bonds will have a superior SHP measure. 4. Our analysis in this section draws on our earlier work, Eun and Resnick (1988). 5. It can be seen from comparing (4b) and (5b) that if the last term of (4b), ECRJ (fce;), is small, which is highly likely, then the dollar rate of return will be similar for both hedging methods. 6. In this study, we adopt the 'complete' hedge approach, with a hedge ratio of minus unity. We do not consider cross-hedging because the cross-hedge ratio is quite variable over time. A recent study by Kaplanis and Schaefer (1991) shows that the cross-hedge approach often leads to poorer out-of-sample period results than the unitary approach. A complete hedging strategy is in accord with the empirical findings of Adler and Simon (1986) and the theoretical work of Eaker and Grant (1985). Other recent studies on hedging strategies include Black (1989, 1990) and Celebuski, Hill and Kilgannon (1990). 7. When the expected-return vector is estimated for the MVP, MVP(H), BST and BST(H) strategies of combined investment in stocks and bonds, we found that better results obtained when equations (8) and (9b) were calculated separately for the bond asset class and for the stock asset class rather than together. The results presented in this study are calculated in this manner. Additionally, it is noted that the 'MVP(H) strategy' is actually a misinformer since the ex ante expected returns for all N securities will not be the same after the forward premiums are added according to equation (9b) or in the combined bond and stock version of this strategy when Yo is estimated uniquely for bonds and for stocks. Consequently, the expected-return vector is required as input into the portfolio problem to frnd the ex ante solution weights. The name, however, is retained for simplicity. 8. Confer Alexander and Francis (1986, pp. 64-65). 9. In addition to hedging over the full twelve month investment horizon using twelve-month forward contracts, we also examined a 'roll-over' hedging strategy using the MVP(H) and BST(H) models where we use one-month, three-month and six-month forward contracts that are 'rolled-over' at expiration to allow for hedging the exchange risk over the twelve month investment period. In general, the average SHP values tended to increase with the length of the forward contract used to hedge the exchange risk for both the MVP(H) and BST(H) strategies . Fl BIBLIOGRAPHY Adler, M. and B. Dumas, 1983, International Portfolio Choice and Corporation Finance: A Synthesis, Journal of Finance 38, 925-84. Adler, M. and D. Simon, 1986, Exchange Risk Surprises in International Portfolios, Journal of Portfolio Mana~ement 12, 44-53. Agmon, T. and Y. Amihud, 1981, The Forward Exchange Rate and the Prediction of the Future Spot Rate, Journal of Banking and Finance 5, 425-37. Alexander, G. and J. Francis, 1986, Portfolio Analysis, 3rd ed. (prentice-Hall, Englewood Cliffs, NJ). Black, F., 1989, Universal Hedging: Optimizing Currency Risk and Reward in International Equity Portfolios, Financial Analysts Journal 45, 16-22. __, 1990, Equilibrium Exchange Rate Hedging, Journal of Finance 45, 899-907. Burik, T. and R.M. Eunis, 1990, Foreign Bonds in Diversified Portfolios: A Limited Advantage, Financial Analysts Journal 46, 31-42. Celebuski, M., 1. Hill and J. Kilgannon, 1990, Managing Currency Exposures in International Portfolios, Financial Analysts Journal 46, 16-23. Eaker, M. and D. Grant, 1985, Optimal Hedging of Uncertain and Long-Term Foreign Exchange Exposure, Journal of Banking and Finance 9, 221-321. Eun, C. and B. Resnick, 1984, Estimating the Correlation Structure of International Share Prices, Journal of Finance 39, 1311-24. __, 1985, Currency Factor in International Portfolio Diversification, Columbia Journal of World Business 20, 45-53. __, 1987, International Diversification under Estimation Risk: Actual vs. Potential Gains, in S. Khoury and A. Gosh, eds.: Recent Developments in International Banking and Finance (D.C. Heath, Lexington, MA), 135-47. __, 1988, Exchange Rate Uncertainty, Forward Contracts and International Portfolio Selection, Journal of Finance 43, 197-215.. Eun, C., R. Kolodny and B. Resnick, 1991, U.S. - Based International Mutual Funds: A Performance Evaluation, Journal of Portfolio Mana~ement Spring, 88-94. French, K. and J. Poterba, 1991, Investor Diversification and International Equity Markets, American Economic Review 81 No.2, 222-226. Frenkel, J.,.1981, Flexible Exchange Rates, Prices and the Role of News: Lessons from the 1970's, Journal of Political Economy 89, 665-706. BI __, and R. Levich, 1968, Transactions Costs and Interest Arbitrage: Tranquil versus Turbulent Periods, Journal of Political Economy 85, 1209-26. Grubel, H., 1968, Internationally Diversified Portfolios: Welfare Gains and Capital Flows, American Economic Reviev,: 58, 1299-1314. Jobson, J.D. and B. Korkie, 1980, Improved Estimation and Selection Rules for Markowitz portfolios, Paper presented at the Western Finance Association meeting. __, 1981, Putting Markowitz Theory to Work, Journal of Portfolio Management, 7, 70-74. Jorion, P., 1985, International Portfolio Diversification with Estimation Risk, Journal of Business 58, 259-78. __, 1986, Bayes-Stein Estimation for Portfolio Analysis, Journal of Financial and Ouantitative Analysis 21, 279-92. __, 1987, Why Buy International Bonds?, Investment Management Review, September/October, 19-28. __, 1989, Asset Allocation with Hedged and Unhedged Foreign Stocks and Bonds, Journal of Portfolio Management 16, 49-54. Kaplanis, E. and S. Schaefer, 1991, Exchange Risk and International Diversification in Bond and equity Portfolios. Working Paper, London Business School. Lee, A., 1989, Hedging Decisions Within the Overall Asset Allocation Strategy, exerted from Managing Currency Risk. The Institute of Chartered Financial Analysts, 13-19. Lessard, D., 1976, World, Country and Industry Relationships in Equity Returns: Implications for Risk Reduction through International Diversification, Financial Analysis Journal 32, 2-8. Levy, H. and Z. Lerman, 1988, The Benefits of International Diversification in Bonds, Financial Analysis Journal 44, 56-64. Levy, H. and M. Samat, 1970, International Diversification of Investment Portfolios, American Economic Review 60, 668-75. Madura, J. and W. Reiff, 1985, A Hedge Strategy for International Portfolios, Journal of Portfolio Management 11, 70-74. Markowitz, H., 1952, Portfolio Selection, Journal of Finance 7,77-91. Merton, R., 1972, An Analytic Derivation of the Efficient Portfolio Frontier, Journal of Financial and Ouantitative Analysis 7, 1851-72. Perold, A.F. and E.C. Schulman, 1988, The Free Lunch in Currency Hedging: Implications for Investment Policy and Performance Standards, financial Analysis Journal 44, 45-52. B2 Rosenberg, M.R., 1989, International Fixed Income Investing: Theory and Practice, Merrill Lynch Research Report, May, 1-16. Solnik, B., 1974, Why Not Diversify Internationally?, Financial Analysts Journal 20, 48-54. Thomas, III, L.R., 1987, The Performance of Currency Hedge Foreign Bonds, Goldman Sachs Research Report, November, 1-20. Tobin, J., 1958, Liquidity Preference as Behavior Towards Risk, The Review of Economic Studies 26, 65-86. B3