they suggest that, within each management style, managers are not loading up on the same risks. In other words, managers seem to be expressing different views or using different portfolio construction techniques (or both!) in their active decisions. The figures in Table 14.1 provide evidence on the ex post performance of individual managers. On the basis of this evidence, investors may wonder whether it makes sense to include traditional managers in the mix at all. The reason for including traditional managers is straightforward: Most institutional investors hold portfolios of managers. Thus, the choice is not between a structured manager and a traditional manager, but between alternative portfolios of managers. What happens if we view the historical experience in this light? To assess the differences between structured and traditional strategies at the portfolio level, we created composites of structured and traditional active managers for the period between 1992 and 2001. As with our earlier analysis, we distinguished between the different manager types using realized tracking errors-but this time we used the prior three years to classify managers for the next three-year holding period (i.e., an investable strategy). We continue to measure performance against the S&P 500. For each three-year time period, we filtered the data into two groups: structured equity managers (1 to 3 percent tracking error) and traditional active managers (5 to 15 percent tracking error). Within each group, we next created 100 randomly selected composite portfolios of two and four managers (equally weighted), and then calculated average buy-and-hold returns for each subsequent three-year period. In Table 14.2, we show the active returns, tracking errors, and information ratios for various cutoff points in the sample. For example, the top quartile represents the 25th best portfolio of managers in the sample according to the indicated statistic. Thus, we can think of these cutoff points as representing an investor's skill level in developing a portfolio of managers. The results in Table 14.2 are consistent with those in Table 14.1: Compared to portfolios of traditional managers, the portfolios of structured managers have higher median excess returns (with less risk), and higher information ratios at all levels. For example, comparing the results with four managers, the median information ratio for portfolios of structured managers is 0.24 compared to -0.12 for portfolios of traditional managers. Not surprisingly, the portfolios of structured managers also have lower average tracking errors and less dispersion in tracking errors and excess returns. Thus, skill at manager selection is much more important when developing a portfolio of traditional managers. While this is a compelling first cut at an investable strategy, comparing core S&P structured and traditional managers may be a naive way of approaching the issue of optimal manager combinations. Many institutional investors choose traditional managers on the basis of a particular expertise: for example, growth and value. How would the results look if we created portfolios of growth and value managers? Table 14.3 shows the results achieved by composite portfolios of growth and value managers over the period from 1992 through 2001, where active returns, tracking errors, and information ratios are measured relative to the S&P 500.