For the past few years, institutional investors in the Netherlands have been steadily adding international securities to their portfolios.

This has prompted discussion of the benefits of diversification and of the management of the associated currency exposure. Our research shows that optimum diversification outside the Netherlands is far greater than it is today, but only if Dutch investors hedge their currency exposure.

Dutch institutional investors have enjoyed very attractive returns. From 1977-95, based on broad market indices, Dutch equities returned an average 15.5% and bonds an average 8.4% (in guilders). The typical Dutch pension portfolio has total overseas exposure of 25-30%, mostly in equities, and a domestic equity component of around 10%.

This portfolio (30% total equities), with an annual risk of about 8% as measured by standard deviation of returns, is more conservative than comparable institutional portfolios in the US (60% equities) or UK (85%). Assuming Dutch portfolios remain at this level of conservatism, should they continue to diversify internationally?

We analysed historical returns to Dutch investors for 1977-95 from Dutch equities and Dutch bonds, as well as foreign equities and foreign bonds. We also produced currency hedged returns based on one-month forward rates. The results showed Dutch bonds and equities had superior individual combinations of risk and return, with hedged foreign bonds a close second.

We also constructed different efficient frontiers based on increasingly expansive sets of allowable assets. For each we selected a portfolio with the same 8% annualised risk postulated for a typical Dutch institutional portfolio. When the portfolio was restricted to Dutch equities and bonds, it contained 66% bonds and 34% equities. Including foreign bonds and foreign equities, the portfolio again contained only Dutch securities, with unhedged foreign assets excluded from the optimal allocation.

If we project past Dutch returns, and assume no currency hedging, even the current level of international diversification (25-30%) is too much.

A more sensible approach is to create better estimates of expected returns. When these returns are related more to general relationships than to historical patterns, asset allocation studies using the Markowitz mean-variance model can give different results.

We believe that a sound passive model for expected aggregate returns is one that relates expected return differentials both to differences in return variance and to differences in beta coefficients against a reasonable world index of stocks and bonds. These return estimates are then aggregated using index weights to estimate returns for foreign equities and bonds. Expected currency hedged returns are set equal to expected unhedged returns less hedging costs.

The resulting efficient frontiers for Dutch investors, assuming index weights within foreign equities and bonds, and not allowing currency hedging, show a much greater allocation to foreign equities. At the 8% risk level, the ideal Dutch portfolio today would comprise 53% Dutch bonds, 29% foreign equities and 18% Dutch equities.

If currency hedging is allowed, still more foreign diversification is justified, even including conservative annual hedging costs of 0.4%. Hedged foreign bonds would make up 47% 0hedged foreign equities 34%, unhedged foreign equities 10% and Dutch equities 9%. The most significant opportunity with currency hedging is diversification out of Dutch bonds and into a mix of foreign equities and bonds.

When the potential for hedging currency risk is exploited fully, there is a substantial increase in expected returns as well as foreign diversification levels. At the 8% risk level, the initial benefit in terms of expected return from allowing any foreign diversification is only 0.5%. This increases to 1.5% if currency hedging is allowed.