In the September 2005 Investment Advisor, we showed that characteristic (value-growth and market capitalization) constrained investing costs investors almost 300 basis points per year. Then, in the January 2006 issue, we demonstrated that characteristic boxes are not asset classes since they are not compositionally unique, have high correlations with one another, and experience changing membership over time.
Thus we argued that it is a waste of time to use characteristic boxes for "asset" allocation. Instead, the stock market should be thought of as a single asset class rather than as nine unique "asset" classes.
These articles highlight the limitations of value-growth- and market-capitalization-constrained investing, pointing out that this system of investing evolved out of convenience without an empirical basis. We challenge advisors, consultants, and investors who continue to use the system to prove that it is a sensible system. This article constitutes a further challenge to constrained investing in which we will address the purported links between risk management and the use of style boxes.
While we are confident that our research supports our findings, over the last few years at various conferences many advisors and consultants have nevertheless told us of several reasons why they still think constraining managers is good for investors. Two of those reasons involve reducing or controlling risk. One of the arguments is the notion that investing in several characteristic boxes (e.g. small-cap value and large-cap growth) is essential to controlling overall portfolio risk. Furthermore, many advisors believe that it is important to constrain managers to a particular box so that the proper level of diversification can be maintained.
The second argument in support of constrained investing goes beyond diversification and focuses on the relationship among market cap, growth, and volatility. In this argument, the advisor takes on the role of limiting the percentage of small-cap or growth stocks in a portfolio so that the portfolio adapts to the risk tolerance of the investor. This exposure control is currently accomplished by constraining individual managers on the basis of market capitalization and growth. However our research presented below shows that when a portfolio is composed of multiple managers, focusing on an individual manager is an unnecessary and expensive way to control risk. Instead, individual managers should be allowed to pursue their unique style, while risk is controlled at the overall portfolio level.
This, we argue, allows managers, advisors, and investors to move up to the win-win-win situation of unconstrained investing: managers pursuing winning equity styles, advisors efficiently controlling risk at the portfolio level rather than at the individual manager level, and investors taking advantage of superior return opportunities at an acceptable level of risk.
Boxes as a Risk Management Tool
Many advisors use characteristic boxes (CBs) as a way to manage overall portfolio risk. This is done by investing with managers who are categorized in different CBs, such as placing money with a small-cap value fund, a large-cap value fund, and a mid-cap blend fund. The possible choices are represented by the traditional Morningstar characteristic grid shown at right.
The extreme strategy is to invest in nine different managers, each representing one of the nine characteristic boxes. The idea behind this approach is that returns of managers residing in different boxes move independently of one another and thus reduce overall portfolio risk.
In our January 2006 IA article, "Boxes Are Not Asset Classes," we presented evidence that CBs are not unique asset classes and that the average correlation among CB indexes is 0.80. Since risk reduction potential--when risk is measured as the standard deviation of return--is one minus the square root of the average correlation among investments, the risk reduction potential of investing across CBs is slightly more than 10%. This demonstrates that diversifying across CBs provides insignificant risk reduction benefits.
In order to test this proposition with real investment managers rather than with CB indexes, as we did in our January 2006 IA article, we gathered annual return information on 342 mutual funds over the period 1996 through 2002. These represent all non-indexed, U.S. domestic equity funds with complete 1996-02 return data and categorized by Morningstar as either large-cap value (LV), large-cap growth (LG), small-cap value (SV), or small-cap growth (SG) in 2002.
We selected this time period because it falls within the period of our earlier study ("The Problematic Style Grid" research report available at www.athenainvest.com) and thus we are able to compare our results with the mutual fund results. In addition, this period represented a very volatile stock market and challenging risk-management environment. We limited our sample to the corner boxes in the Morningstar characteristic grid, as this gives the CB approach (i.e., investing in managers residing in different CBs) the best opportunity to perform well since the corners have the lowest correlations across CBs (see table below).
The mutual fund results appear in the top portion of the table. The first row shows results for portfolios constructed by investing in four managers, one from each of the four corner boxes. This is the approach preferred by many advisors and is presumed by them, without evidence, to be essential for controlling overall portfolio risk. When portfolios are formed by investing in four fund managers, one from each of the four corner boxes, the average standard deviation is 18.9% and the resulting average information ratio (IR = return/standard deviation) is 43 bps/1% of volatility as shown in columns three and four in the table.
We now compare these results to several other portfolio formation methods. Here we are addressing the question of whether it is necessary to diversify based on CBs, as many advisors advocate, or if there are other selection methods that work as well, if not better.
On the next line in the table, the four managers are randomly selected without regard to the characteristic box into which they fall. The results are virtually identical to those of the CB selection approach. As shown in columns three and four, the average portfolio volatility is 19.6% and the investor earns an average of 42 bps in return for every 1% of volatility, versus 18.9% and 43 bps for the CB method. Thus, knowing the CB in which the manager resides does not help in making the portfolio diversification decision. Random selection, on average, is as good. This calls into question the need for mutual fund categories from services such as Morningstar, when the objective is choosing managers for the purposes of risk control.
The next results provide further evidence that CBs are an unnecessary risk management tool. For this approach, four managers are selected from the same characteristic box. This approach is the polar opposite of the currently preferred approach of selecting managers from different boxes. However, the performance is similar to that of the previous two approaches, with an average volatility of 21.3% and an average return per unit of volatility of 41 bps.
Examining the individual boxes on the next four lines, we see that two CBs--LV and SV--outperform the two previous approaches while the other two--LG and SG--underperform. So staying within a CB can produce superior returns to diversifying across CBs! However, the two growth box results have substantially higher volatility and lower IRs than do the previous results. The issue of being overexposed to small cap or growth is an important one and will be addressed in the next section.
Consistent Style but Characteristic Drift
The final results in the table are based on our earlier study in which we created four style-consistent managers. Our four managers were style consistent, but experienced significant characteristic drift. Examination of the four style results reveals superior performance when compared to the previous mutual fund approaches with respect to average return (11.0%) and information ratio (55), while having a portfolio volatility (19.9%) that falls in the middle of the mutual fund results. These results clearly show that it is not necessary to diversify across characteristic boxes and keep managers in a particular CB in order to control risk and produce superior returns.
Finally, results for the equally weighted S&P 1500 index returns are reported. The S&P 1500 outperforms the mutual fund results (with the exception of the information ratio for the SV portfolio) while, on the other hand, our hypothetical four-style portfolio outperforms the index.
In summary, we find that randomly selecting managers for inclusion in a portfolio without regard to CBs produces virtually identical performance, on average, to that obtained by selecting a manager from each of the corner CBs. Even the polar opposite approach of selecting managers from the same CB performs, on average, about the same as does the currently popular approach of investing in managers from different CBs. Finally, our style-consistent manager portfolio outperforms, in terms of information ratio, the three mutual fund approaches as well as the equally weighted S&P 1500.
The average correlation of funds from different CBs is 0.46 while the average within-CB correlation is a much higher 0.73. Why is it, then, that the three mutual fund results are so similar? The reason is that, even within a particular CB, each manager has a unique equity style--the way the manager analyzes, buys, and sells stocks--that leads to lower correlations among managers. As we have stated in prior articles, style and characteristics are two very distinct aspects of equity portfolio management. Portfolio characteristics do not define equity style.
When investing in four managers, one from each of the corner CBs, versus investing in a single manager, less than half of the risk reduction is the result of imperfect CB index correlations (average correlation is 0.8), while more than half is the result of imperfect manager correlations (average correlation among managers in the same CB is 0.73). This imperfect-across-equity-styles correlation is the major reason that the three mutual fund results are so similar. The benefit of diversifying across manager styles outweighs the benefit of diversifying across CBs.
We conclude that CBs are an unnecessary risk management tool. In fact we go further and argue that constraining managers to fit within a particular CB hurts performance while adding virtually nothing to risk reduction. But what about the widely held concern that unconstrained managers might hold too many small-cap or growth stocks and thus produce an unacceptable level of portfolio volatility?
Managing Risk at the Portfolio Level
We have argued that CBs are an unnecessary risk management tool since simply selecting managers at random produces similar performance results. In addition, we have argued both here and elsewhere that managers should not be CB-constrained since this leads to underperformance. So our recommendation is to hire unconstrained managers, each pursuing a unique equity style. This will lead to characteristic drift--for example, movement toward more small or growth stocks--and a perceived increase in volatility. So how does the advisor go about controlling risk exposures when managers are free to roam the stock universe?
In an unconstrained world, asset allocation is done among styles rather than CBs. In this regard, the number of managers selected may be the same in both situations. Even though the individual unconstrained managers may experience considerable drift, combining several managers, particularly those with low drift correlation over time, leads to much more stable characteristics at the overall portfolio level.
This is demonstrated in the figures in the "Unmanaged Tilts" charts (below) that show the unmanaged size and value tilts of our four styles described in the aforementioned "Problematic Style Grid." We define tilts as holdings-based standard normal deviates, so a tilt of 1.0 means that the portfolio is one standard deviation above the average for the stock universe. Notice that the tilt of each style manager (Graham, Neff, O'Neil, and T. Rowe Price) varies considerably from year to year, sometimes changing by as much as 0.5 standard normal deviates in a year.
However, if an equally weighted portfolio of all four styles is constructed, the overall portfolio tilt is much more stable; in particular, note that the portfolio size tilt is around 0 for most of the sample period. Thus it is possible to allow individual manager characteristic drift while at the same time maintaining fairly stable overall portfolio tilts. The overall portfolio result, then, is superior performance combined with stable characteristics even as the individual managers roam the stock universe as they see fit.
Tilts are measured as market value standard normal deviates and as price-to-sales standard normal deviates from the S&P 1500 average.
Our evidence shows that combining a few characteristic-unconstrained style managers does not lead to extreme characteristic tilts, e.g., too many small-cap stocks. In fact, our evidence suggests that diversifying across styles rather than across CBs leads to stable characteristic tilts over time despite significant characteristic drift by individual managers.
However, this may not always be the case. In such situations, there are a number of ways an advisor can control size and value tilts as well as volatility at the portfolio level.
- Take into consideration the history of tilts and volatility when deciding which style managers to include in a portfolio. We refer to this as identifying and investing in complementary styles.
- Consider investing in a manager who offers a range of risk-return portfolios while pursuing the same style in each portfolio. This can be accomplished through the use of written calls and convertible securities, among other tools.
- Vary the amount invested in money market securities to offset risk changes in the equity portion of the portfolio.
- Use ETFs to offset undesirable equity portfolio tilts. Our example portfolio above was size neutral but had an average 0.2 growth tilt. This could be offset by investing a portion of portfolio assets in a pure value ETF. The ETF position could be adjusted over time to offset the changing portfolio tilt resulting from manager decisions.
- Consider hiring a master manager who will oversee and manage using an overlay portfolio.
We argue that any time managers are constrained by characteristics, the portfolio is very likely to underperform. Constraining a manager to stay within a CB is a very expensive way to control risk. Put another way, CB-constraining managers is the poor man's way of controlling risk. There are more efficient ways, such as those mentioned above. Let the equity portion of your portfolio be managed by unconstrained managers and control risk at the portfolio level, not at the individual manager level.
Set Them Free and Profit
Many advisors and consultants use characteristic boxes for controlling portfolio risk and constraining managers, two techniques that play a major role in defining their practices. This article offers evidence that they are instead employing an unnecessary risk management tool and are hurting investment performance. Perhaps the advisor's practice is defined incorrectly. Let equity managers have the freedom to invest according to their stated style and allow characteristic--not style--drift. For portfolios comprising multiple managers, control risk and characteristic tilts at the portfolio level, not at the individual manager level.
This allows all parties to move up to the win-win-win situation of unconstrained investing: managers pursuing winning equity styles; advisors efficiently controlling risk at the portfolio level rather than at the individual manager level; and investors taking advantage of superior return opportunities at an acceptable level of risk.
Craig Callahan is founder and president of ICON Advisers in Denver and can be reached at email@example.com. Mr. Callahan is also a principal at Athena Investment Services. C. Thomas Howard is a Professor of Finance at the Daniels College of Business in Denver. He can be reached at firstname.lastname@example.org.