1980-1990   MSCI World MSCI Europe MSCI U.S. MSCI Japan Average return 19.2% 18.1% 16.9% Volatility 14.4 17.7 15.9 24.3% 21.7 1991-2001   MSCI World MSCI Europe MSCI U.S. MSCI Japan Average return 6.5% 8.7% 12.5% Volatility 14.6 15.3 14.5 -5.6% 25.2 1990s. Also shown in the table are the historical standard deviations of returns. Both statistics (historical average return and historical volatility) were calculated using monthly excess return data. As the table clearly illustrates, the historical average returns are quite sensitive to the choice of time period. For example, in the 1980s the best-performing of these three equity markets was the Japanese market, while in the 1990s the U.S. market showed the best performance. Notice that while historical averages seem to be quite sensitive to the choice of time period, the historical volatilities appear to be less so. This is an important point to which we will return. Now, suppose that an investor decided to construct optimal portfolios on the basis of the average returns shown in Table 9.1. In other words, suppose that an investor used the average returns (and risk characteristics) of the 1980s and built an optimal portfolio, and then did the same using the data from the 1990s. How would these portfolios compare? Figure 9.2 shows the two sets of optimal asset allocations, with the very loose constraint that the portfolio weights must sum to 100 percent. As we can see, the choice of time period used for estimating returns has dramatic consequences for the portfolio weights. Using the data from the 1980s, the optimal portfolio has a long position in Japanese equity. By contrast, a short position in Japanese equity is implied when the sample is restricted to the data from the 1990s. In any event, the portfolio weights are so extreme that no prudent investor would actually implement them as a strategic asset allocation. The technical issues associated with standard approaches to strategic asset allocation give rise to two practical issues. First, because of the potential for extreme portfolio positions, practitioners often find it hard to develop an intuition behind the portfolio. Second, because it is unlikely that investors will implement the extreme portfolio positions, it is hard to develop an approach to portfolio advice that can be used across clientele types: Each clientele type is likely to need their own set of constraints. Thus, the standard approach to strategic asset allocation fails on two grounds: It gives extreme portfolios, and does not allow for consistent advice giving. Each of these issues can be addressed by using an equilibrium approach.