You are on page 1of 9

A Four Factor Performance Attribution Analysis Model for Equity Portfolios

Fund Analytics - In this paper, we discuss a four-factor performance attribution model for
equity portfolios. The model captures the risk and return characteristics of four elementary
equity investment strategies and can be used to identify and quantify an equity portfolios risk
and style exposures, sources of total return, and sources of value added. It can also be used to
help overcome the risk mismatch problem of benchmarking and peer grouping.
How to start an etf
Benchmarking and peer grouping as investment performance measurement tools, while very
well accepted and as intuitive as they are, are not sophisticated enough to provide adequate
information to help investors and plan sponsors in searching for managers who can add value.
It is not enough to know that a manager beat the benchmark or was above the peer group
median. Investors and plan sponsors must also know how the manager beat the benchmark or
the median manager in the peer group. The sources of value added must be identified and
quantified. Benchmarking and peer grouping have so far failed in this regard.
And, with all the intuitive benefits associated with them, benchmarking and peer grouping can
mislead investors in important ways. For example, because benchmarking and peer grouping do
not appropriately and adequately differentiate managers in terms of risk exposures, investors
seeking diversification may be taking unanticipated risk. In the volatile stock market during the
recent years, many investors have found that their portfolios were not as diversified as they
had expected, resulting in excess risk exposures and significant performance shortfalls.
We believe that plan sponsors need to adopt a factor attribution analysis scheme along with
benchmarking and peer grouping. In this paper, we discuss a four-factor performance
attribution model for equity investments that enables us to identify and accurately quantify a
portfolios risk exposures, sources of excess return, and sources of value added. [The excess
return of a portfolio is the portfolio return in excess of the one-month Treasury-bill return; and
the value added by a portfolio is the portfolio return less the benchmark return.]
The Four-Factor Performance Attribution Model
The four-factor model we discuss in this paper extends the Capital Asset Pricing Model [CAPM]
with three additional factors: the Fama-French size and book-to-market [BTM] factors and the
Carhart momentum factor. Published academic studies show that the four-factor model
represents a significant improvement over the [single market factor] CAPM in explaining equity
portfolio performance [see, for example, Fama and French (1992, 1993, and 1996), Carhart
(1997), Chan, Chen, and Lakonishok (2002)].

As a performance attribution model, the four-factor model captures the risk and return
characteristics of four elementary equity investment strategies:
Investing in high versus low market sensitivity stocks
Investing in small versus large market capitalization stocks
Investing in value versus growth stocks
Investing in momentum versus contrarian stocks
The four-factor performance attribution model can be mathematically represented as
RP RF = P + M(RM RF) + SRS + BRB + ORO
RP - RF = Portfolio excess return
RM - RF = Market factor return
RS = Size factor return
RB = BTM factor return
RO = Momentum factor return
P = Portfolio risk-adjusted return
M = Portfolio market beta
S = Portfolio size beta
B = Portfolio BTM beta
O = Portfolio momentum beta
The portfolio excess return, as defined earlier, is the portfolio return less the one-month
Treasury-bill return. The market factor return is the value-weighted average return of all New
York Stock Exchange, American Stock Exchange and NASDAQ stocks less the one-month
Treasury-bill return. The size, BTM, and momentum factor returns are the return to a portfolio
of small-cap stocks minus the return to a portfolio of large-cap stocks, the return to a portfolio
of value stocks [stocks with a high ratio of book to market value] minus the return to a portfolio
of growth stocks [stocks with a low ratio of book to market value], and the return to a portfolio
of momentum stocks [stocks that outperformed in the recent past] minus the return to a
portfolio of contrarian stocks [stocks that underperformed in the recent past], respectively.
[Refer to Fama and French (1993) and

Carhart (1997) for construction details for the Fama-French size and BTM factors and the
Carhart momentum factor.]
The portfolio betas are the sensitivities of the portfolio excess return to the factor returns and
hence the models measures of the risk exposures of the portfolio. In other words, the betas of
a portfolio are measures of the extent to which the portfolio return varies with the factor
returns. Thus, for example, the realization of the size factor return will impact a portfolio with a
size beta of 0.50 twice as much as it will impact a portfolio with a size beta of 0.25.
There are two well-known approaches for estimating betas for a portfolio: one is return-based
and the other holdings-based. The return-based approach estimates betas by an ordinary least
squares regression of the portfolio excess returns on the factor returns. With the holdingsbased approach, beta estimates for a portfolio are obtained by first estimating the betas of the
constituent stocks and then taking the value-weighted average betas of the stocks as the betas
of the portfolio. The betas of the constituent stocks can be estimated by mapping them to
constructed portfolios with similar characteristics [market capitalizations and book-to-market
ratios, for example]. In the analysis that follows, the betas are return-based estimates.
Specifically, they were estimated by regressing monthly portfolio excess returns on monthly
factor returns over forty-eight month periods.
The four-factor performance attribution model thus decomposes the portfolio excess return
into five components:
Risk-adjusted return *P+
Return attributable to market factor exposure *M(RM RF)]
Return attributable to size factor exposure *SRS+
Return attributable to BTM factor exposure *BRB+
Return attributable to momentum factor exposure *ORO+
The risk-adjusted return is also referred to as the alpha. It is the return after controlling for
general market movements and other risk factor exposures and hence a measure of the ability
of the manager to generate return by stock selection beyond the reward for taking risk.
Therefore, we believe it should be the dominant criterion for selecting a manager.
Quantifying Risk Exposures
At the heart of the modern portfolio theory is the notion of risk-return trade-off. Too often
investors who do not pay much attention to risk exposures find themselves taking too much
unanticipated risk and suffering significant underperformance. Sound investment strategies

build on a clear understanding of risk exposures. Risk exposure analysis with the four-factor
performance attribution model allows investors to build more accurate risk profiles for the
managers they are interested in than either benchmarking or peer grouping.
MarketSizeBTMMomentum BetaBetaBetaBeta Wilshire 5000 0.990.010.000.00 S&P 500 0.990.150.07-0.03 S&P/BARRA 500 Value 0.98-0.120.330.08 S&P/BARRA 500 Growth 1.01-0.160.16-0.14 FRC 1000 1.00-0.110.05-0.01 FRC 1000 Value 0.87-0.200.300.16 FRC 1000 Growth
1.18-0.02-0.14-0.21 FRC 2000 0.860.700.010.20 FRC 2000 Value 0.680.480.310.33 FRC 2000
Growth 1.110.85-0.190.07 Exhibit 1: Risk Exposures of US Equity Indices
The four-factor performance attribution model identifies and quantifies the risk exposures of a
portfolio by measuring the portfolios betas to the four risk factors. Exhibit 1 shows the beta
estimates for some US equity indices. [The beta estimates for the indices were obtained from
regressing the monthly excess returns of the indices on the monthly factor returns over the
forty-eight month period ending December 31, 2001.]
The close-to-unit market beta and close-to-zero size, BTM, and momentum betas of Wilshire
5000 confirm the robustness of the four-factor model. As a broad market index, Wilshire 5000
behaves very much like the general US stock market, which by definition has a unit beta to the
market factor and a zero beta to each of the other factors.
Examining the betas of the other indices, we notice the following patterns:
Small-cap indices have positive size betas while large-cap indices have negative size betas
Value indices have positive BTM betas while growth indices have negative BTM betas
Small-cap and value indices have positive momentum betas while large-cap growth indices
have negative momentum betas
One of peer groupings problems is the wide range of manager risk exposures even within the
peer group, as illustrated by Exhibit 2. [We estimated the betas for an individual fund by a
regression of the funds monthly excess returns on the monthly factor returns over the fortyeight month period ending December 31, 2001.]
The market betas of the US equity growth funds, as categorized by MorningStar, ranged from
0.39 to 1.70, the size betas ranged from 0.46 to 0.95, and the momentum betas ranged from
0.48 to 0.43. Even worse, within this supposed growth peer group, the BTM betas covered a
range of 1.35 to 0.68. Comparing the BTM betas of the funds in this group to those of the US
equity indices, we find that about 7 percent [84 out of 1258], with a BTM beta greater than
0.35, should definitely be considered as value funds and 45 percent [561 out of 1258], with a

BTM beta greater or equal to 0.00, should be considered as value-growth neutral or value
funds.
Needless to say, peer grouping with such wide ranges of risk exposures does not provide
especially meaningful comparisons. Rather it misleads investors and in many instances is unfair
to the managers being evaluated. Meaningful peer grouping requires that we first quantify
managers risk exposures and then form peer groups of managers with similar risk exposures.
Similarly, meaningful benchmarking requires that there is no severe risk exposure mismatch
between the manager and the benchmark. The main purpose of benchmarking is to measure
the managers ability to generate value on the top of the return from assuming the risk of the
benchmark. [This is perhaps why many investors identify the differential return between the
manager and the benchmark with the alpha, that is, the risk-adjusted return]. Benchmarking
thus
MarketSizeBTMMomentumBetaBetaBetaBetaMedian1.00-0.020.030.00Maximum1.700.950.680.43Minimum0.39-0.46-1.35-0.48Exhibit 2: Risk Exposures of
MorningStar US Equity Growth Mutual Funds
serves its main purpose well as long as the risk exposures of the manager are close to those of
the benchmark. When there is severe risk exposure mismatch, however, benchmarking
becomes very misleading. Exhibit 3 illustrates this phenomenon.
anager As risk exposures are very close to those of the benchmark, S&P 500. As a result, the
o accurately measure the ability of a manager to generate value in addition to the return from
uantifying Sources of Excess Return
eedless to say, investors and plan sponsors are ultimately concerned about the total return a
portfolio manager delivers. They want to select managers who can consistently deliver good
total
Panel A: Excess Returns and Differential Returns between Manager and Benchmark
differential return between Manager A and the benchmark, 2.18%, is very close to the true
alpha [risk-adjusted return] of Manager A, 2.11%. However, Manager B has a substantially
higher exposure to the BTM factor than the benchmark. Consequently the differential return
between Manager B and the benchmark, 2.23%, represents a truly remarkable overestimate of
Manager Bs true alpha, 0.03%.

assuming risk, we need to use a robust factor attribution model, such as the one discussed
here. If one wants to keep benchmarking as an alternative tool because investors are still more
used to it, one must make sure that the benchmark matches the manager well in terms of risk
exposures. Factor analysis can be used to quantify the managers risk exposures and the risk
exposures of available indices to find or construct a benchmark that results in minimum risk
mismatch.
Excess ReturnS&P 50010.15Manager A12.33Manager B12.38Panel B: Risk Exposures and RiskAdjusted ReturnsMarketSizeBTMMomentumS&P 5000.99-0.150.07-0.03Manager A1.000.160.05-0.02Manager
B0.99-0.140.95-0.04Risk-Adjusted
ReturnMarketSizeBTMMomentumS&P 500-0.0511.85-1.650.17-0.17Manager A returns. In
evaluating and selecting managers, it is important to know what returns a manager has
delivered so far. However, in order to succeed in the future, it is even more important to know
how the manager achieved those returns. Benchmarking and peer grouping are mainly
concerned about how a manager has fared relative to
e benchmark and peers. They do not aim to explain the managers return. By decomposing the
ition. Here the mangers excess return, 25.51% is
ecomposed into a risk-adjusted return or alpha, -0.28%, a return component from exposure to
e first quarter of 2000, the manager in the exhibit performed extremely well relative to the
managers return, however, the four-factor performance attribution model presents a clear
picture of its sources and their relative importance. Exhibit 4 illustrates such a return decompos
the market factor, 3.46%, a return component from exposure to the size factor, 5.83%, a return
component from exposure to the BTM factor, 13.13%, and a return component from exposure
to the momentum factor, 3.37%.
Why is it so important to quantify the sources of a managers return? Exhibit 5 explains why.
For
broad market: the manager delivered an excess return of 25.51% while the excess return to the
general market was only 3.46%. Investors in the fund might get exited with such a superb
performance. But what they really ought to do was to carefully look into the sources of the
managers return.
Benchmarking measures the extent by which the manager outperformed the benchmark but
does not aim to explain it. If a risk exposure mismatch existed between the manager and the
benchmark, as in most cases of active managers, it would be difficult to determine whether the

outperformance was by stock selection, by deviation in risk exposures, or by both without


factor attribution analysis. The four-factor perfori immediately clear how the manager added
the value. Exhibit 6 provides an illustration.
Return Attributable to Risk Factor Exposure
Manager A and Manager B had the same benchmark, S&P 500, and they both beat it and added
value by about the same amount, 2.18% and 2.23%, respectively. Taking a closer look by
quantifying the sources, we find that the values were added by the two managers through very
different avenues. Manager A took almost identical risk exposures to those of the benchmark,
and added value almost entirely by stock selection: the differential risk-adjusted return, 2.16%,
between Manager A and the benchmark accounts for almost the entire value added by
Manager A, 2.18%. Manager B, on the other hand, showed little stock selection ability, and
generated the value added, 2.23%, mainly by taking an excess exposure to the BTM risk factor
relative to the benchmark *Manager Bs BTM beta of 0.95 versus the benchmarks 0.07+, which
added 2.09%.
Exhibit 7 graphically decomposes Manager Bs value added.
Performance Attribution at the Composite Level
Most plan sponsors enlist a number of managers within different categories [large-cap value,
mid-cap core, and small-cap growth, for example]. Given the diverse investment styles and risk
characteristics of the manager team, peer grouping is not practically feasible for performance
measurement at the composite level. [By composites, we mean aggregates of managers
enlisted -202468101214Exhibit 7: Components of Value AddedManager B12.380.0311.851.542.26-0.22S&P
50010.15-0.0511.85-1.650.17-0.17Difference2.230.080.000.112.090.06Excess Return =Alpha +Market +Size +BTM +Momentum by a plan sponsor; for example,
the US equity composite consists of all US equity managers in the plan.] While benchmarking is
still feasible, it will not give us a clear risk picture of the composite. Nor will it give us a clear
view of the benefit from diversification, which is the main reason for hiring managers of
different investment styles in the first place. Moreover, performance measures from
benchmarking normally do not provide an especially meaningful comparison between
managers in the composite because, as we discussed before, the sources of the values added
by the managers can be very different. The economically meaningful way to compare managers
is to compare them by risk-adjusted return, which is the return after controlling for risk
exposures and which measures the ability to generate return by selecting stocks undervalued
relative to their risk characteristics.
The four-factor performance attribution model provides a unified approach to performance
measurement. It uses the same set of risk factors in measuring performance of managers,

whether they are large-cap, mid-cap, or small-cap managers and whether they are value, core,
or growth managers. As a result, the four-factor performance attribution model provides
performance measures that are perfectly comparable across managers in the composite, and it
quantifies the risk exposures and decomposes the return for the composite just as for any
individual manager in the composite. Exhibit 8 is an illustration of performance attribution at
the composite level. The risk exposures of the composite are simply the value-weighted
average risk exposures of the managers constituting the composite.
REFERENCES
Panel A: Risk Exposures of the CompositeWeightCategoryMarketSizeBTMMomentumManager
A0.40Large-Cap
Growth1.18-0.22-0.19-0.21Manager
B0.30Mid-Cap
Core0.950.180.050.18Manager
C0.30Small-Cap
Value0.980.480.310.33Composite1.050.110.030.07Panel
B:
Risk
Factor
ReturnsMarketSizeBTMMomentum11.9710.972.385.61Panel C: Attribution of the Composite
ReturnExcessRisk-Adjusted Return =Return +Market +Size +BTM +MomentumManager A9.520.5614.12-2.41-0.45-1.18Manager
B14.490.0211.371.970.121.01Manager
C20.821.2311.735.270.741.85Composite14.400.1512.581.210.080.39Exhibit 8: Performance
Attribution at the Composite LevelRisk Factor ReturnReturn Attributable to Risk Factor
ExposureBeta

[1] Carhart, Mark, 1997, On Persistence in Mutual Fund Performance, Journal of Finance, 5781.
[2] Chan, Louis, Hsiu-Lang Chen, and Josef Lakonishok, 2002, On Mutual Fund Investment
Styles, Review of Financial Studies, 1407-1437.
[3] Chan, Louis, Narasimhan Jegadeesh, and Josef Lakonishok, 1999, The Profitability of
Momentum Strategies, Financial Analysts Journal, 80-90.
[4] Daniel, Kent, Mark Grinblatt, Sheridan Titman, and Russ Wermers, 1997, Measuring Mutual
Fund Performance with Characteristic-Based Benchmarks, Journal of Finance, 1035-1058.
[5] Fama, Eugene, and Kenneth French, 1992, The Cross-Section of Expected Stock Returns,
Journal of Finance, 427-465.
[6] Fama, Eugene, and Kenneth French, 1993, Common Risk Factors in the Returns on Stocks
and Bonds, Journal of Financial Economics, 3-56.

[7] Fama, Eugene, and Kenneth French, 1996, Multifactor Explanations of Asset Pricing
Anomalies, Journal of Finance, 55-84.
[8] Grinblatt, Mark, Sheridan Titman, and Russ Wermers, 1995, Momentum Investment
Strategies, Portfolio Performance, and Herding: A Study of Mutual Fund Behavior, American
Economic Review, 1088-1105.
[9] Jegadeesh, Narasimhan, and Sheridan Titman, 1993, Returns to Buying Winners and Selling
Losers: Implications for Stock Market Efficiency, Journal of Finance, 48-91.
[10] Rouwenhorst, K. Geert, 1998, International Momentum Strategies, Journal of Finance, 267284.

You might also like