In the typical U.S. pension plan, each manager has been hired to deliver performance based on an investment policy.
The essence of this policy is captured by the benchmark against which each manager is compared. Thus, a manager compared to the Russell 1000 Value index is hired to deliver performance in conjunction with the policy implied by that benchmark, large-cap, value oriented with a market beta in line with that index. The assumption is made that the style profile of the manager might not be identical, but is close, to the style profile of the benchmark.
The style profile is the exposure of a portfolio to those persistent factors that explain equity returns. The exposures of the portfolio to size, value/growth and the market matter. Thus the assumption is that style misfit, defined as the style profile difference between the manager and the benchmark at each evaluation point, won't matter that much. This assumption has not been correct for the majority of active managers.
The style misfit and its return impact can be identified by going one level deeper in analyzing a manager and the benchmark. When you do, you will find that over any evaluation period, style misfit return can be large or small, positive or negative, depending on the magnitude of the style misfit differences and the returns to the factors that explain equity returns. Those factors are size (the return of small stocks minus large stocks); value/growth (the return of high book-to-market stocks minus low book-to-market stocks); and the market.
Since mid-1963, the return to the size factor has been 250 basis points per year; that is, small stocks have outperformed large stocks by that amount. A 0.3 exposure difference between a manager and benchmark would have generated a performance difference between the two of 75 basis points (250x0.3) per year. The return to the value/growth factor has been 450 basis points per year; that is, high book-to-market (value) stocks have outperformed low book-to-market (growth) stocks. A 0.3 exposure difference between a manager and benchmark would have generated a performance difference between the two of 135 basis points (450x0.3) per year. A market beta difference of 0.1 produced a 115-basis point return (1,150x0.1) difference, based on the market's return of 1,150 basis points, excluding the risk-free rate.
These style profile differences really matter. Although the factor returns have been positive in the long run, we experienced a three-year period ended in 1996 in which the factor returns for size and value/growth were negative. Indeed, in any given quarter the factor returns can be very large, as demonstrated by value outperforming growth in the first quarter of 1992 by 17% and small-cap underperforming large-cap in the second quarter of 1992 by 10%.
Clearly variability exists in these factor returns, so the style misfit return is not always in the same direction. In addition, the manager's style exposures are constantly changing because of the changing active bets in the portfolio.
Once the manager's return is adjusted for the misfit return, what's left is the manager's stock selection skill in relation to the benchmark. In order to build a superior team of managers, however, you've got to know more.
A domestic equity team also has an investment policy, a target, represented by the benchmark against which the team is compared. In the process of building a team, assets can be allocated to each manager in a number of ways. First, the assets can be allocated in an identical fashion to the representation of the manager's index to the team's target. For example: If the Russell 3000 is the team's target and the Russell 1000 and 2000 make up 90% and 10% of that target respectively, then the small-cap managers are given 10% of the assets and the large-cap managers are given the remainder. Value and growth managers would be given, for example, 50% each, if that were the weighting of value and growth in the indexes.
In this allocation scheme, no consideration is given to the alpha differences between managers or to the correlation of those alphas. In addition, no consideration is given to the style misfit between the manager and the benchmark. Failure to take a manager's specific style profile into account clouds the style profile of the team and makes for less robust, suboptimal alternatives for team construction. The manager of managers has left lots of information on the table, unused. More effective teams can be built when this information is properly formed and properly used.
If a plan sponsor wishes to take into account his expectations of the alpha differences between managers, the alphas of all managers must be comparable regardless of style and regardless of benchmark. The results of benchmarking clearly do not provide manager alphas that are comparable, as they were developed using different benchmarks.
Earlier we identified the need to adjust a manager's alpha for the misfit return caused by the common factors that explain equity returns. This yielded the stock selection skill of a manager vs. his benchmark. But benchmarks have alphas, and the alphas of different benchmarks can be quite dissimilar (Pensions & Investments, April 28). Therefore, it is now essential to further adjust the alpha of the manager to recognize the alpha of his benchmark.
A few words about benchmark alphas are in order. To compare alphas across all manager styles, each manager must be evaluated from the same point, the same perspective. We use the capitalization-weighted universe of all domestic equities as our measuring point. From this perspective, all portfolios, even indexes, have alphas.
All portfolios have an exposure to size, value/growth and the market. Once a portfolio's style profile is estimated at each evaluation point, the returns to the factors in the next evaluation period are used to compute the systematic returns to the portfolio. The alpha of the portfolio is then computed by subtracting the systematic returns from the portfolio's returns. Thus the alpha of each portfolio is identified.
Benchmark alphas appear to be random, but can be large or small, positive or negative, over any evaluation period. The alphas of benchmarks have been economically large in relation to the alphas of the managers.
In a prior exercise, we showed that during the 1991-'95 period the Russell 1000 and 2000 value and growth indexes had alphas that, on average, ranged from a positive 150 basis points to a negative 150 basis points per year (P&I, June 9). The magnitude of these alphas is very large in relation to manager alphas and must be taken into account. If these benchmark alphas are not accounted for, the identification of a manager's skill is contaminated.
Once a manager's misfit-adjusted alpha is in turn adjusted for the benchmark's alpha, the stock selection skill of the manager is identified after all style impacts are removed. These are the manager alphas that can be compared regardless of style. These are the alphas to be considered in the team-building process.
The plan sponsor is now in position to use the information that previously was left on the table. The manager of managers can now develop several alternative manager teams. These alternatives will take into account not only the managers' alphas after all style impacts are removed but also the specific style profile of each manager. These alternatives will have a team alpha and a team style profile.
The style profile of each team then can be compared to the team's target, the team's investment policy objective, and the style misfit can be identified. Each team will have an alpha based on the expected alpha for each manager.
A menu of team alternatives can be developed with different alpha/style misfit tradeoffs. The volatility of the team alphas and the volatility of the team misfit then can be identified either based on history or on sponsor expectations. The investment committee then can decide on the team alpha and misfit alternative it prefers.
Or the committee might decide not to have any misfit. Not having any misfit will affect team alpha considerations, but that's part of the tradeoff between alpha and misfit. Only by understanding a manager's alpha after all style impacts are removed and understanding the time series of his style profile, can an effective tradeoff be made.
In order to turn investment policy fiction into investment policy fact, it's necessary to evaluate each manager's skill after adjusting for misfit vs. benchmark. In order to take into account a manager's alpha in the team-building exercise, his alpha must be further adjusted for the alpha of the benchmark against which he is compared. This yields a manager's skill after all style impacts are removed.
Likewise, to take into account a manager's specific style profile, the style profile of his benchmark can not be used as a surrogate. However, once the managers' alphas after all style impacts are used along with the managers' specific style profiles, more effective team building can proceed. Clearer, better constructed alternatives of team alphas/misfit tradeoffs can be identified for the investment committee. The investment committee then can buy into the misfit risk or leave that to the responsibility of the manager of managers as a way to add return, and volatility.