Returns basedstyleanalysisinexcel mcdermott

JOURNAL OF EDUCATION FOR BUSINESS, 85: 107–113, 2010
Copyright C Heldref Publications
ISSN: 0883-2323
DOI: 10.1080/08832320903258592

Returns-Based Style Analysis: An Excel-Based
Classroom Exercise
John McDermott
Fairfield University, Fairfield, Connecticut, USA

W. Sharpe’s (1988, 1992) returns-based style analysis provides an excellent opportunity to use
a sophisticated portfolio-analysis tool in the classroom to help illustrate important topics in
investment and operations research courses. Students can perform classic returns-based style
analysis by creating a spreadsheet model and using the Solver add-in in Microsoft Excel.
The technique can also be used to supplement classroom units on performance measurement
and style drift in investment courses or to illustrate a finance application in an introductory
operations research course. An example analysis using a popular mutual fund is provided as
well as the accompanying Excel model.

Keywords: Analysis, Fund, Mutual fund, Performance, Style analysis

There are many methods available to financial analysts to are normally hired to manage a particular asset class (e.g.,
conduct a style analysis of a portfolio or mutual fund. Two U.S. Small value) in a portfolio. Style determination is also
potential methods are holdings-based style analysis popular- important in manager performance measurement when the
ized by Morningstar and Sharpe’s returns-based style analy- aim is to separate true alpha from beta. The selection of a
sis (RBSA). RBSA was introduced by Sharpe (1988, 1992) benchmark index is often based on style determination.
and is a powerful analytical tool that is used to capture the
style of a portfolio or mutual fund using only returns data.
Holdings-Based Style Analysis
Specifically, RBSA uses nonlinear optimization to construct
a portfolio of indices to minimize the tracking error with the In holdings-based style analysis, the actual holdings (e.g.,
portfolio being analyzed. RBSA is widely used and accepted individual stocks, mutual funds, exchange-traded funds
in the investment community and forms the basis for the pop- [ETFs]) in the portfolio are disaggregated and then classi-
ular investment analysis software Zephyr StyleADVISOR. fied into general categories based on asset classes, geogra-
The objective of this teaching note and the accompanying phy, or industry classification. Morningstar also classifies the
classroom exercise is to provide students with exposure to a portfolio into one of nine style boxes based on the portfolio’s
powerful analytical technique and also provide some work in general size (large cap, mid cap, small cap) and value–growth
building an Excel model and using the Solver add-in to con- orientation. Kaplan (2003) provided a detailed discussion of
duct nonlinear optimization. I use a well-known mutual fund, Morningstar’s style classification methodology.1
the Dodge and Cox Balanced Fund (DODBX), to illustrate A major drawback of holdings-based analysis is that an
the technique. individual or mutual fund’s holdings may vary, perhaps sig-
nificantly, through time. A snapshot of a portfolio’s holdings
one day, especially in the case of a portfolio comprised of
STYLE ANALYSIS
actively managed funds or individual stocks, may not be at
all representative of the history of the fund or of its holdings
Analysts and researchers may desire to determine the style of
in the future. Portfolio holdings are, in general, self-reported
a particular fund manager. The issue of style is particularly
and therefore subject to manipulation. For example, a port-
relevant in institutional money management when managers
folio manager may engage in window dressing, which is the
practice of altering the holdings just prior to reporting to
Correspondence should be addressed to John McDermott, Fairfield Uni-
make the portfolio look better to investors.2 In contrast with
versity, Finance Department, 1073 North Benson Road, Fairfield, CT 06824, self-reported holdings, realized returns are objective and ex-
USA. E-mail: jmcdermott@mail.fairfield.edu ternal, not subjective and internal.

108 J. MCDERMOTT

The reported holdings of a portfolio may also lead to incor- to Minimize TE
rect conclusions about a portfolio’s true asset class exposure subject to
and the extent of its diversification. For example, consider
wi ≥ 0
a position in a large foreign-based multinational company n
i=1 wi = 1
(e.g., Honda Motor) that does a much of its business in the
United States and also trades as an American Depository
Receipt (ADR)on the New York Stock Exchange. Similarly, The constrained quadratic minimization described previ-
what about a large U.S. multinational (e.g., Exxon) that does ously is the essence of RBSA. The minimization can also
a much of its business internationally? What exactly is the be solved by numerical methods in a number of software
nature of these holdings in a portfolio? Holdings-based anal- programs including Solver in Excel.4
ysis is not and cannot be precise on this issue; it most likely To implement RBSA, the selected indices should ideally
would classify Honda as an international and Exxon as a be mutually exclusive (i.e., they do not overlap) and nearly
U.S. holding. The stock returns, however, may be telling a exhaustive5 (i.e., they cover the investable universe). The
very different story. RBSA is purposely agnostic to the de- popular indices I have used successfully are the following:6
bate about how a particular firm should be classified; it is
only interested in how the stock behaves in the context of Fixed Income Indices:
characterizing the overall portfolio’s returns. Citigroup 3-month T-Bill
Lehman U.S. Aggregate Bond
Returns-Based Style Analysis Lehman U.S. Corporate High Yield
Lehman Global (ex U.S.) Treasury
Given the potential shortcomings of holdings-based analysis,
the attention turns to returns-based style analysis. RBSA U.S. Equity Indices:
compares the returns of a stock portfolio or mutual fund Dow Wilshire Large Growth
against the returns of a tracking portfolio of indices that Dow Wilshire Large Value
represent the various equity and fixed income asset classes. Dow Wilshire Small Growth
The analysis defines a mix of the indices that would have best Dow Wilshire Small Value
explained the behavior of the returns of the portfolio being Dow Wilshire Micro Cap
analyzed. Technically speaking, RBSA determines a long- International Equity Indices:
only portfolio of mutually exclusive and exhaustive indices MSCI EAFE
that minimizes the variance of the tracking error relative to MSCI Emerging Markets
the portfolio being analyzed.
Beginning with the historical returns (typically the last 60 Normally, it is preferable to use at least the last 36 months
months) of a portfolio or mutual fund, I denote the client’s of returns data available to conduct the RBSA—obviously
portfolio total return in month t as Rt. Next, I choose a set of the need for a longer time horizon increases with the number
n mutually exclusive and exhaustive indices to create a port- of indices used in the analysis. Monthly returns data for
folio that as closely as possible mimics the client’s portfolio mutual funds and indices are available in many financial
returns. The monthly return on index i in period t is denoted databases and can also be obtained from public websites such
as Ii,t . as Yahoo Finance. Dow Wilshire Style Indices are available
The difference between the client’s portfolio returns and to download at www.wilshire.com.
the return on a mix of the n indices in month t is defined as
as εt . Further, the tracking error (TE) is defined as variance
AN ILLUSTRATION OF RETURNS-BASED
of this difference over the t time periods. Specifically,
STYLE ANALYSIS
εt = Rt − w1 I1,t + w2 I2,t + . . . . . . + wn In,t
T E = V ar [εt ] The DODBX is a very large7 actively managed mutual fund
where wi , equals the weight on index i. The term in brack- that was just recently reopened (in February 2008) to new
ets is return on a portfolio of the indices. The weights are investors.8 An excerpt from the holding-based analysis of
constrained to be nonnegative and to sum to 1.3 DODBX from Morningstar is shown in the Appendix.9 Morn-
ingstar classifies the fund’s U.S. equity holdings as large
wi ≥ 0 value and its fixed income holdings as high quality and inter-
n
i=1 wi = 1 mediate maturity. The DODBX is a large balanced fund with
It is important to choose the set of index weights so that the low turnover; I expect these characteristics to minimize the
tracking error between the portfolio and the mix of indices is differences between returns-based and holding-based analy-
as small as possible. Specifically, the problem is: ses in order to illustrate the legitimacy and power of RBSA
to infer style as well as equity and fixed income from returns
Choose wi for i = 1, n data only.

RETURNS-BASED STYLE ANALYSIS 109

FIGURE 1 Excel spreadsheet model set-up.

The Excel spreadsheet model set-up and Solver dialog The key to interpreting RBSA is to understand that the
box used to conduct the RBSA are presented in Figures 1 and provided mix of the indices creates a portfolio that most
2.10 Normally, I would provide the monthly returns data to closely tracks the analyzed portfolio; the technique, despite
my students in an Excel spreadsheet and have them structure its pedigree and sophistication, does not reveal the true allo-
the model, deﬁne the optimization in Solver, and interpret the cations! First, it is important to recognize that the precision of
results. The results of conducting the RBSA for the 60-month neither holdings-based analysis nor RBSA is perfect. RBSA
period January 2002 through December 2007 are presented looks at the average asset exposures over the course of the
in Figure 3.11 entire period of the analysis, whereas holdings-based pro-
The portfolio of indices generated by the RBSA is referred vides only a snapshot in time. In fact, we should not be
to as the custom benchmark. How well does the custom surprised to see that returns based in some cases provides a
benchmark track the DODBX? The answer is actually quite very different picture than holdings based. Differences be-
well. Figure 4 provides a comparison of growth of a dollar tween the returns and holdings-based style analyses can be
for the DODBX and the custom benchmark. Further, the R2 attributed to a variety of factors. Consider that the RBSA
of a simple linear regression of the fund’s monthly returns
on the custom benchmark’s monthly returns is 0.93.

FIGURE 3 Dodge and Cox Balanced Fund returns-based style analysis,
FIGURE 2 Solver dialog box. January 2002 to December 2007.

110 J. MCDERMOTT

FIGURE 4 Growth of $1, January 2002 to December 2007.

FIGURE 5 Returns-based style analysis, a 36-month rolling analysis of the Dodge and Cox Balanced Fund.


has no prior knowledge of the allocation by name; it is infer- International Equity
ring an allocation for the historical returns of the portfolio. In
In contrast to international equity, RBSA indicates the
some cases, the RBSA may attribute the influence of some of
U.S.equity position is much greater (i.e., 53.8% U.S.vs. 7.6%
the holdings in one category into another category, if the re-
International); the mutual fund exhibits zero exposure to
turns of those holdings behave more like the second category.
emerging markets and only 2.5% to international developed
Considering a portfolio comprised of entirely U.S. stocks, it
markets. Although the fund does hold some large interna-
might be surprising to see RBSA indicates an exposure to
tional developed market stocks (e.g., Sony, Sanofi-Aventis,
emerging markets. But on closer inspection, it is noticeable
Novartis) that trade as ADRs, RBSA evidences the fund as
that the U.S.holdings consist of several large multinational
overwhelmingly an investment in U.S.equity.
corporations with a significant presence in emerging markets
economies. For example, a RBSA of Exxon allocates 51% to
MSCI Emerging Markets Index in its custom benchmark.12 Performance Measurement
RBSA can provide insights into the true nature of a portfolio,
Sharpe (1988; 1992) suggested that the mean of the track-
insight that holdings-based analysis cannot. I now provide a
ing error can be used as a performance measure. Ter Horst,
brief interpretation of the RBSA of the DODBX by major
Nijman, and de Roon (2004) suggested the use of RBSA to
asset class.
predict future performance. Alternatively, it is possible to use
tracking portfolio as a customized benchmark and compute
Fixed Income alpha relative to the benchmark. Ben Dor, Budinger, Dynkin,
The DODBX’s fact sheetat www.dodgeandcox.com states: and Leech (2008) suggested the use of RBSA in construct-
ing performance benchmarks. In this case, the mean of the
The Fund primarily invests in a diversified portfolio of pri- tracking error is 4.62 bps (SD = 46.21 bps) per month (55.6
marily investment-grade fixed income securities, including: bps annualized). In comparison, the alpha of the fund relative
U.S. government obligations, mortgage and asset-backed se- to the customized benchmark is 60 bps (t = 0.73, df = 59).
curities, corporate bonds, collateralized mortgage obligations As previously mentioned, the R2 of the fund regressed on the
and others. To a lesser extent, the fund may also invest in be- customized benchmark is 0.93.
low investment-grade fixed income securities.

The RBSA presented in Figure 4 reveals, as expected, a Style Drift
high-quality and short-term bond strategy as the benchmark The classroom assignment could be extended to determine
portfolio allocates 27% to T-Bills, 9% to the Lehman Aggre- the extent of style drift in the portfolio or mutual fund.14 This
gate Index, 2% to Global Treasury bonds and no weight to could be accomplished by using subperiods and comparing
U.S. high-yield bonds. the results or by conducting rolling period analysis (e.g.,
The RBSA reveals the average equity–fixed income split using rolling 36-month windows) to examine the changes in
of the DODBX as 62–38. A holdings-based analysis or a weights over time. An example of a rolling-period analysis is
review of the fund’s prospectus would be obscure or even in- illustrated in Figure 5. The analysis reveals a fairly consistent
accurate on this point. For example, the fund literature states style over time.
only that “Under normal circumstances the Fund will hold
no more than 75% of its total assets in stocks.”13Although
as of June 30, 2007, the fund reported that it held 69.2% in
equities. CONCLUSION

U.S. Equity My example is a simple one by design. The DODBX strategy
is well known and its equity style has been consistent (i.e.,
The large weight (41.3%) on the DJ Wilshire Large Value U.S. Large Value) through time. The fund is very large, held
index in the custom benchmark reveals the equity side of the 84 different stocks on March 31, 2008, concentrated mostly
DODBX as primarily a U.S.Large Cap Value strategy. A look in the U.S., and its turnover is quite low (27%) for an actively
at the top 10 holdings of the fund as of March 31, 2008, lists managed fund. Therefore, I would expect a holdings-based
large U.S.companies such as Comcast, Wal-Mart, Hewlett- analysis to confirm the RBSA and it does, perhaps convinc-
Packard, confirming this finding. The RBSA also shows ex- ing the students of its validity and power. In contrast, con-
posure to U.S.Small Value (5.0%) and U.S.Microcap (7.5%). sider a complicated portfolio of stocks, bonds, ETFs, and
Note the weight on both U.S. Growth indices (Large and actively managed funds. Conducting a holdings-based anal-
Small Growth) are both zero in the style benchmark suggest- ysis is fraught with potential problems for more complicated
ing no or little exposure to this asset class. In sum, the RBSA and dynamic portfolios. It is precisely this situation in which
suggests the fund’s U.S.equity strategy has been primarily RBSA is most useful as a complement to holdings-based
large value and secondarily small value and microcap. analysis.15

112 J. MCDERMOTT

I believe an introduction to RBSA has a place in upper- through 1000. The Dow Wilshire large-cap indices
level undergraduate course and introductory MBA courses contain the 750 largest stocks in the U.S., whereas
in investments, and certainly in courses in portfolio analy- small-cap indices capture stocks 751 through 2,500
sis. I have used the topic as part of a larger case study on and the microcap has 2,500 through approximately
performance evaluation in a senior-level undergraduate fi- 4,000.
nance capstone course. It could be, however, structured as 6. Thanks to Jim Davis of Dimensional Fund Advisors
a stand alone assignment on style analysis and performance for suggesting this set of indices.
measurement. For example, “Conduct a style analysis of Mu- 7. As of December 31, 2007, DODBX had over $28
tual Fund X using Sharpe’s returns-based style techniques. billion in assets.
Compare and contrast the results with Morningstar’s holding- 8. See www.dodgeandcox.com.
based style analysis” or “Construct a custom benchmark for 9. See www.morningstar.com.
Mutual Fund X using Sharpe’s returns-based style analysis. 10. Please contact the author for the Excel workbook.
How has the fund performed? Compare your result to a sin- 11. See Lobosco and Di Bartolomeo (1997) for calculat-
gle (or multi-) factor model alpha. To what do you attribute ing approximate confidence intervals.
the differences?” 12. A caution is necessary; the style benchmark can only
I believe the topic is also well suited to an introduc- explain about 41% of the variation in Exxon’s monthly
tory course in operations research to illustrate a real-world returns for the past five years ending in December
application of mathematics science tools, Excel modeling, 2007. This highlights the use of RBSA as a tool for
and the use of Solver to conduct nonlinear optimization. analyzing portfolios.
An assignment built on RBSA provides the students with 13. See www.dodgeandcox.com/balancedfund.asp
an opportunity to develop an Excel model to do sophis- 14. Chan, Chen, and Lakonishok (2002) analyze style
ticated quantitative analysis and to explore a widely ac- drift in mutual funds.
cepted tool in industry to complement holdings-based style 15. For example, Brown and Goetzmann (2001) used
analysis. RBSA in analyzing the performance of hedge funds.

NOTES REFERENCES

1. Available for download at: http://corporate. Ben Dor, A., Budinger, V., Dynkin, L., & K. Leech (2008). Constructing peer
morningstar.com/UK/html/pdf.htm?../documents/M- benchmarks for mutual funds: A style analysis-based approach. Journal
ethodologyDocuments/ResearchPapers/Holdings- of Portfolio Management, 34(2), 65–77.
Brown, S., & Goetzmann, W. (2001). Hedge funds with style (NBER Work-
basedAndReturns-basedStyleModels PK.pdf
ing Paper 8173). Cambridge, MA: National Bureau of Economic Re-
2. There is an extensive literature on window dress- search.
ing. Influential research includes Grinblatt and Tit- Chan, L., Chen H., & Lakonishok, J. (2003). On mutual fund investment
man (1993), Grinblatt, Titman, and Wermers (1995), styles (2002). Review of Financial Studies, 15, 1407–1437.
and Wermers (2002). Grinblatt, M., & Titman, S. (1993). Performance measurement without
benchmarks: An examination of mutual fund returns. Journal of Busi-
3. The nonnegativity constraint on the weights can be ness, 66(1), 47–68.
relaxed when appropriate. For example, when analyz- Grinblatt, M., Titman, S., & Wermers, R. (1995). Momentum investment
ing a portfolio that uses derivatives, short positions, strategies, portfolio performance, and herding: A study of mutual fund
or leverage. behavior. American Economic Review, 85, 1088–1105.
4. I find it instructive to run an ordinary least squares Kaplan, P. (2003, June). Holdings based and returns based style models.
New York: Morningstar Research.
(OLS) regression with the mutual fund’s monthly re- Lobosco, A., & Bartolomeo, D. (1997). Approximating the confidence inter-
turns as the dependant variable and the monthly re- vals for Sharpe style weights. Financial Analysts Journal, 53(4), 80–85.
turns of the indices as the independent variables. The Sharpe, W. (1988). Determining a fund’s effective asset mix. Investment
OLS results can be contrasted with the RBSA results Management Review, 2, 59–69.
specifically highlighting the natural interpretation of Sharpe, W. (1992). Asset allocation: Management style and performance
measurement. Journal of Portfolio Management, 18, 7–19.
the RBSA slopes as portfolio weights. Ter Horst, J., Nijman, T., & de Roon, F. (2004), Evaluating style analysis.
5. You may be wondering about U.S. midcap stocks. Journal of Empirical Finance, 11(1), 29–53.
They are split between the large and small-cap DJ Wermers, R. (2002). Mutual fund herding and the impact on stock prices.
Wilshire indices. S&P midcap indices use stocks 501 Journal of Finance, 54, 581–622.


APPENDIX

Dodge and Cox Balanced (DODBX)

Copyright of Journal of Education for Business is the property of Taylor & Francis Ltd. and its content may not
be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written
permission. However, users may print, download, or email articles for individual use.

Returns basedstyleanalysisinexcel mcdermott

More Related Content

What's hot (16)

Similar to Returns basedstyleanalysisinexcel mcdermott (20)

More from bfmresearch (20)

Recently uploaded (20)

Returns basedstyleanalysisinexcel mcdermott