Throughout, we utilize the monthly total returns computed by Micropal. Each such return takes into account changes in net asset values and any distributions (dividends, capital gains, etc.). Each is net of any expenses charged to the fund and, of course, all transactions costs. However, load charges -- selling costs not charged to the fund but paid directly by the investor -- are not taken into account. Some of the LS100 funds levy such charges on at least some investors; the amount charged depends on the size of the investment. It also may be waived if investment is made through an intermediary or for certain types of savings plans (e.g. Individual Retirement Accounts). In some cases a fund will offer two or more classes of shares. For example, the A shares may require a "front-end" load charge, while the B shares pay a continuing amount from the fund to compensate sales staffs. For the former, sales costs are not taken into account when total returns are computed; for the latter they are.
In terms of total returns, (so computed) the decade from January 1985 through December 1994 was a good one for the LS100 funds and their investors. The average annual return during the period was 11.95%. However, there was considerable variation from year to year. The standard deviation of annual returns was 9.69%. The figure below shows the total returns for each year.
While total returns represent money that can, eventually, be "taken to the bank", it is also important to compare such returns with those that could have been earned by putting money in a bank. We represent the latter by the return on 1 to 3 month U.S. Treasury Bills. The difference between the return on the LS100 funds and that on a strategy of investing in such Bills is termed the excess return on the LS100. It averaged 5.84% over the decade, with a standard deviation of 9.33%. While the latter was broadly typical of the variation in returns for bond/stock mixes similar to those of the LS100, the former was higher than usual. The year-by-year excess returns are shown below.
While the LS100 funds now control less than half of the assets invested in bond, stock and international U.S. mutual funds, they still represent a significant portion of the market: 44% at the end of 1994. To compute the style of the index as a whole, we value-weight the styles of the individual funds. The figures below show the composite style for 1994, based on returns through November 1993 and net asset values at the end of November 1993.
Not surprisingly, the aggregate of all 100 funds provided significant exposures to all of the asset classes. The Non-US dollar exposure was approximately 5% -- somewhat smaller than the exposure to Non-US Securities, suggesting a modest amount of currency hedging. Among these funds there appears to be a preference for value (low price/book) among large-capitalization stocks but for growth (low price/book) among medium-capitalization stocks.
The managers of most of the LS100 funds can be considered active: they attempt to select securities that will, over time, "beat the market", or "do better than average", given their type. To give meaning to such aspects, of course, one has to define the relevant market for each fund. Equivalently, one has to specify the "types" of securities in which each fund invests.
For this purpose we use the Style Analysis method described in W.F. Sharpe, "Asset allocation: Management style and performance measurement," Journal of Portfolio Management, Winter 1992, pp. 7-19, with some variations. The style of a fund, determined in December of the prior year is taken as a benchmark against which the fund's performance is to be compared. In effect, a style mix portfolio is formed for each fund at the end of December of each year, and the performance of the mix computed, month by month during the following year. In doing so, it is assumed that the returns on the component asset classes can be obtained without cost. In effect, we assume the use of index funds that track their indices perfectly and do so for no charge. Clearly, this is unreasonable. A typical expense ratio for a large no-load index fund is approximately 0.20% ("20 basis points") per year. Hence as a practical matter, even an efficient index fund is likely to underperform its style benchmark by this amount.
To assess the performance of the LS100 as a whole, we form a aggregate style mix at the end of each year, using the relative values of the underlying funds and their individual style mixes as components. We then compute the performance of this mix over the subsequent year. For example, the style mix return for each month of 1994 was determined by computing the total return on a mix that had the composition shown in the previous section at the end of December 1993. Due to market movements among the asset classes, the composition by value of this "buy-and-hold" style mix changed from month to month during the year. The mix also changed in each of the prior Decembers, based on the results of the annual style analyses.
How different were the performances of the LS100 funds and the style mixes designed to represent a similar set of passive investments? The figure below tells the story. Each point represents one month in the decade. The horizontal axis indicates the return on the aggregate style mix, while the vertical axis shows the overall return on the LS100 funds.
The similarities are striking. At this level, at least, our estimates of funds' asset allocations account for the vast majority of the variations in return from month to month. This shows both that asset allocation is of great importance and that our procedures provide good estimates of funds' actual asset allocations.
The next figure shows total returns for each of the ten years for both the LS100 fund mix and the aggregate style mix. As can be seen, the latter outperformed the former in eight of the ten years.
We define a fund's selection return as the difference between its return and that of its style. For the LS100 funds as a whole, the selection return can be computed by subtracting the return on the aggregate style mix from that on the mix of the funds. The figure below shows the results of doing so.
The average difference between the returns on the LS100 funds and those of a strategy with a similar style was -0.64% per year. The standard deviation of the difference was 1.28% per year. The average underperformance was somewhat greater than that likely to be achieved with a truly passive strategy (e.g. -0.20%). However, given the variation in the differences, the average cannot be said to be statistically significantly different from either -0.20% or zero. However, the data provide little support for those who believe that the performance of a typical actively-managed fund is likely to beat a passive alternative (i.e. index fund) with the same style.
The ratio of the variance of the aggregate selection return to that of the LS100 fund mix gives an indication of the importance of the effects of active management on a strategy involving diversification across many funds. It was 1.75%. Thus active management added relatively little risk at the aggregate level. However, for a more typical portfolio concentrated in a relatively few funds, the impact of active management on risk would have been considerably greater. For a portfolio invested in only one fund it would have been greater yet: the average ratio of selection variance to fund variance for a single fund was 20.6% over the three-year period from 1991 through 1993.
These results suggest that active management is likely to account for approximately 20% of the risk of a typical large seasoned fund. For portfolios that include between more than one but fewer than 100 funds, the risk associated with active management is likely to account for from 20% to 2% of total risk, with the actual amount closer to the latter, the greater the diversification across funds.
Data Source Micropal, Inc. © For more
information, please call 617-451-1585
