Social Sharing
Extended Viewer
Speaking in Style: Insights for Managers
Mar 1, 1996
John Rabold
Portfolio managers are responsible for all aspects of the portfolios they manage, and so they must use tools which use asset-level information. While their clients often monitor a portfolio in less detail, accounting systems, trading systems, and risk management systems all take into account the specific securities held and weights within each.
Recently, fund sponsors and investment consultants have adopted returns-based style analysis because it is simple and concise in concept and execution. It is, however, a very different "language" from that normally used by a portfolio manager. The language of style analysis does provide, however, a simple, two-dimensional framework for discussion purposes. Since a manager's clients and their consultants may prefer to use this language, managers will find it useful to learn to cope with it as well.
This article suggests ways in which portfolio managers can translate between the language of returns-based style analysis and that of holdings-based portfolio analysis. Portfolio managers can then:
- Describe the current portfolio in the language of style analysis.
- Evaluate the current portfolio, in detail, relative to the style analysis benchmark which the client has produced.
- Evaluate the portfolio's historical performance, in detail, relative to that same style analysis benchmark.
Style analysis
One of style analysis' most important outputs is the style benchmark for the managed portfolio, as of a certain date. The style benchmark is a weighted portfolio of reference indices whose return was as close as possible to that of the managed portfolio over a specified historical period. It thus expresses the apparent style over that period. A U.S. equity manager's style might be represented on a certain date as:
- 40 percent large growth
- 30 percent large value
- 20 percent medium growth
- 10 percent medium value
Fund sponsors and their consultants may use the style benchmark to represent the managed portfolio's style over the evaluation period, and then to evaluate performance relative to that benchmark after the end of the evaluation period. Therefore, it is only fair that the manager be told in advance the contents of the benchmarkthe weights in each of the reference indices. Because its contents can be replicated passively, the benchmark is an alternative in which the fund sponsor can invest and which will approximate the managed portfolio's recent style.
Portfolio management using Barra's equity risk models
Barra's portfolio management tools, used with an equity risk model, are designed to assess risk in a current portfolio, and to evaluate relative historical performance, both relative to the benchmark. A complete listing of holdings in the portfolio and benchmark and their weights is used. As a result, these tools produce much more accurate and detailed results than style analysis can, including risk and return insights down to the asset level, but the results are not stated in the same language.
While some sort of translation between risk model language and style analysis language is called for to facilitate communication among manager, consultant, and sponsor, some simplistic attempts should be avoided. Portfolio managers approaching this question with risk model-based analyzes should not focus only on the risk model factors which seem on the surface to overlap with style analysis (such as "growth" and "size") or try to translate them into style weights. Trying to do so overlooks the structure of the risk model, ignoring the explanatory power of the other factors and the relationships among them.
Conversely, a manager viewing the question from the style analysis perspective should not assume that the largest reference index position in the style benchmark (U.S. Large Growth in the example above) has any special significance in isolation. For example, a "large-capitalization growth manager" will probably have U.S. Large Growth as the largest position in its style benchmark, but that manager should not assume that stocks not in that reference index would be unacceptable in the portfolio. Likewise, that single reference index by itself is not likely to be a useful benchmark for evaluating the current portfolio.
A synthesis
Let us assume that a fund sponsor and its consultant want to use the style analysis language and that the manager wishes to comply. Here are three ways in which the manager's detailed, Barra risk model-based assessments can be coordinated with style analysis results, or can be translated into style analysis language.
1. Describing the style of the current portfolio.
Style analysis attempts to describe the average style of a managed portfolio over an historical period in terms of weights of passively-constructed reference indices. The portfolio manager may also be interested in discussing the position of the current managed portfolio, but style analysis is not designed to do that; it cannot detect the nature of the portfolio now. Barra risk characterization tools, however, used with a risk model, can determine a mix of the same components which can be expected to perform most like the current managed portfolio, that is, to "track" that portfolio most closely. While the characteristics used in style analysis will be used in the assessment, so will the entire structure of the risk model, yielding a superior result for this specific purpose.
Managers using the Barra Aegis System, Risk Manager and Optimizer can easily perform this kind of analysis on their own portfolios. For a managed U.S. equity portfolio, identify the five component portfolio files SAPGRO, SAPVAL, MIDGRO, MIDVAL, and SC600, which are supplied by Barra. Use the Aegis Risk Manager to define these style components and your current managed portfolio as composite portfolios (System|Composites...).Next, create a new portfolio (File|NewPortfolio...) and insert the five style components, preceding their file names with a plus sign (+), as in "+SAPGRO." You need not insert any weights for these names. Save the portfolio (File|Save Portfolio), perhaps using the name "STYLUNIV" to remind yourself that it is the universe of assets available for use in your style analysis tracking portfolio. Finally, run the Aegis Optimizer. Use portfolio STYLUNIV as the universe, your managed portfolio as the benchmark (Settings | Risk Settings...), and a positive cash flow of at least $100,000. Move the general asset constraints upper bound (Settings | Constraints...) up to 100%. This will determine the weighted mix of the five available assets which is predicted to "track" your managed portfolio most closely (Figure 1). Save this output portfolio as STYLENOW.
Figure 1
STYLENOW represents the style of the current portfolio for short-term, prospective purposes. It utilizes the best of both worldsthe robustness of a risk model-based risk characterization of a current portfolio, and the simple language and reasonable benchmarks of style analysis. Users of this approach should not, however, confuse this result with the style benchmark. The first is a snapshot that includes the effects of short-term tactics; the latter is derived very differently, represents an extended timespan, and may capture long-term style more closely.
2. Characterizing risk relative to a style analysis benchmark.
A portfolio manager can use the style benchmark provided by its client as a benchmark in a risk model-based analysis of the current managed portfolio. The fund sponsor must wait for each month's performance result to be calculated in order to update its style analysis review, but the manager can be constantly assessing the portfolio's detailed risks relative to the current style benchmark.
With the Risk Manager in the Aegis system, the process overlaps that of creating a benchmark portfolio to track the managed portfolio. Load the unweighted universe list STYLUNIV and enter the weights provided by your client's style analysis. Save this portfolio with a name like STYLBNCH (File | Save Portfolio As...). Define STYLBNCH as a composite, then open the managed portfolio and select STYLBNCH as its benchmark and "none" as its market. Look at the risk decomposition of the managed portfolio (View | Risk Decomposition), as in Figure 2.
Figure 2
Aegis Risk Manager, when set in this way, answers the central questions about active risk: How far might the performance of the managed portfolio be from that of the benchmark? Which bets am I making relative to the benchmark? How big are they? Am I taking enough risk in the areas where I expect to earn a healthy active return? Have I hedged my bets sufficiently in the areas where I have little knowledge or confidence?
Style analysis also reports the "closeness of fit" or R2 of the style benchmark to the managed portfolio at a specified point in time. Managers can obtain a similar statistic, though one based upon current portfolio structure rather than historical monthly returns, from Aegis Risk Manager.
While viewing the managed portfolio's risk decomposition as described above, divide predicted benchmark variance by predicted total variance to obtain the approximate proportion of predicted total variance quot;explained" by the style benchmark. This is a measure of closeness of fit, or R2. In Figure 2, this predicted R2 is about 89 percent. The predicted active variance divided by the predicted total variance will be approximately one minus this R2 value. The relationship is approximate because when the portfolio beta is not 1, active variance is not independent of benchmark variance.
This exercise shows the relationship between R2 and active risk. Active risk is the more useful measure to the portfolio manager for two reasons. It expresses the potential performance difference between the portfolio and the benchmark directly, as a standard deviation of annual return. It also recognizes that a component of risk which is perfectly correlated with the benchmark can cause active returns and should therefore be monitored along with other active risks.
3. Evaluating performance relative to a style analysis benchmark.
STYLBNCH, the weighted mix of equity asset classes, can also be used as the benchmark in a risk-model-based performance analysis. Style analysis is blind to actual portfolio composition, but can still produce a benchmark which is reasonable and fair for use by all parties. Knowing the portfolio's actual holdings, the manager can evaluate the portfolio's relative performance each month, determining the precise reasons for the performance differences.
Style analysis benchmarks change gradually over time. To be accurate, the analysis should reflect the changing of the benchmark. The use of the most recent version of STYLBNCH to evaluate the entire history would introduce a bias, because it could not have been determined until the end of the period being evaluated. Likewise, the user should not be tempted to use STYLENOW as the performance benchmark because it includes current (and transitory) tactical decisions.
Conclusion
Returns-based style analysis and holdings-based risk modeling use very different approaches, but can be used productively together. Since delivering useful information and improving communication are goals of each, sponsors, consultants, and managers can integrate these approaches to produce superior results.