You are on page 1of 9

RESEARCH

Capturing Value:
Why Less Can Be More

August 2016

Marlena Lee, PhD INTRODUC TION


Vice President
Many valuation ratios have been used to identify value versus growth securities. A partial
Research
list includes book-to-market (B/M), earnings-to-price (E/P), cash flow-to-price (C/P), and
Savina Rizova, PhD sales‑to‑price (S/P). The options expand further when we allow for combinations of variables.
Vice President
How should an investor decide among the many alternatives?
Research

Antonio Picca, PhD Parsimony is an important guiding principle when building a model to understand a real-
Research
world phenomenon. For example, when building asset pricing models to explain differences in
expected returns across stocks, using as few variables as possible makes the model implications
more transparent and can help the researcher glean deeper insights. We believe that using as few
variables as possible is also crucial for portfolio implementation. A streamlined framework allows
us to build portfolios that seek higher expected returns in a more transparent and efficient way.
It also allows for more robust portfolio risk controls. Why are these considerations important for
investors? Greater transparency can reduce monitoring costs. Efficient implementation helps reduce
implementation costs. Building parsimonious models and investment frameworks that can achieve
an investor’s objectives is not easy, but the benefits are worth the effort. As Albert Einstein said,
“Everything should be made as simple as possible, but no simpler.”

Another benefit of a parsimonious approach is that it can reduce the likelihood that an empirical
observation is the result of data mining. The more choices available to a researcher designing an
experiment, the more difficult it becomes to determine if the results are real or due to chance.
We believe this is an important consideration before adding a variable to the investment process,
DIMENSIONAL FUND ADVISORS 2

as one should be reasonably confident the new variable brings cross-sectional regressions using various combinations
additional information about expected returns. Even if a of company size, B/M, E/P, C/P, S/P, and profitability as
variable creates sizable return spreads on its own, we believe the independent variables.3 Exhibit 1 uses US equity data
it is important to understand whether the variable will from January 1964 to December 2015.4 The regressions
bring incremental benefit to an existing portfolio. If the in Panel A show that there have been size and relative
research used to demonstrate the benefit of including price effects in the cross-section of average stock returns.
additional variables uses extreme security weighting schemes Regressions 1 through 4 show that B/M, E/P, C/P, and S/P
or unnecessarily complex combinations of many variables, all contain reliable information about future average stock
care must be taken when interpreting the reliability of such returns, after controlling for size. However, the regressions
results. Remember, there are many benefits to not adding in Panel A do not reveal whether each relative price metric
unnecessary or redundant variables to the investment process. contains unique information about average returns.

Dimensional uses book-to-market equity (B/M) as the We test whether the information contained in one
primary variable to sort securities along the relative price valuation metric is subsumed by the other metrics in
dimension. Research has shown that other price ratios can Panel B of Exhibit 1.5 In regression 5 we include both
be used to identify differences in average returns about as B/M and E/P. In regression 6 we include both B/M and
well as book-to-market but often with higher turnover.1 C/P, and in regression 7 both B/M and S/P. In each of the
More recent research has shown that profitability2 contains three regressions, B/M remains robustly related to returns
additional information about differences in expected while the other relative price measures lose most of their
returns that can be used to improve strategies while explanatory power. These results suggest that E/P, C/P, and
keeping turnover similar. S/P do not contain much additional information about
expected returns beyond what is captured by B/M.
In this paper, we reexamine the evidence on using alternative
price ratios to define value vs. growth. Similar to previous In Panel C of Exhibit 1 we examine the explanatory power
research, we find that many valuation ratios can be used of the relative price metrics when we add profitability to
to identify differences in average returns. However, after the set of characteristics used to explain the cross section
controlling for size, B/M, and profitability, we find that E/P, of average stock returns. Including profitability does not
C/P, and S/P do not appear to contain additional information reduce the reliability of the B/M coefficient, but it reduces
about expected returns. Also, portfolio simulations for further the coefficients and t-statistics on E/P, C/P, and S/P.
US, developed markets ex US, and emerging markets This finding suggests that some of the ability of E/P, C/P,
that use these alternative price ratios individually or in a and S/P to predict differences in average returns over this
combination fail to deliver reliably higher average returns than period is also due to their covariance with profitability.
simulations that use B/M as the sole relative price metric.
These regressions indicate that from 1964 to 2015, E/P, C/P,
CROSS -SEC TIONAL RETURN REGRESSIONS and S/P all contained information about the cross section
A cross-sectional regression helps to identify variables that of returns. These characteristics lost some of their power to
can be observed today and contain reliable information about explain or predict the cross section of stock returns when
differences in future average stock returns. Cross-sectional used in tandem with B/M. When used in tandem with B/M
regressions are particularly useful for testing whether a and profitability, they lost all of their explanatory power. That
variable contains additional information about differences is, the information in E/P, C/P, and S/P about the cross section
in future average returns, controlling for other variables. of stocks returns was subsumed by B/M and profitability.

To investigate what variables contain reliable information In contrast, the information in B/M remained robust across
about differences in next month’s stock returns, we run all regression specifications. The B/M regression coefficients

1. See Davis and Lee (2008).


2. See O’Reilly and Rizova (2013). Profitability is measured as operating profits divided by book value of equity.
3. In this paper, we build on the cross-sectional regression approach from Fama and French (1992 and 2008.)
4. Compustat data on firm fundamentals offer an extensive coverage of the US market since the early 1960s.
5. We do not include all four price ratios in a single regression because of the increasing multicollinearity concerns.
DIMENSIONAL FUND ADVISORS 3

Exhibit 1: Cross-Sectional US Equity Return Regressions (January 1964–December 2015)

Size B/M Prof E/P C/P S/P R2


Panel A. Regressions controlling for size and one valuation ratio
−0.092 0.292 — — — — —
(1)
(−2.22) (4.52) — — — — 0.019
−0.139 — — 1.171 — — —
(2)
(−3.96) — — (1.97) — — 0.018
−0.141 — — — 1.408 — —
(3)
(−3.96) — — — (2.84) — 0.018
−0.116 — — — — 0.036 —
(4)
(−2.99) — — — — (2.54) 0.015
Panel B. Regressions controlling for size, B/M, and an additional valuation ratio
−0.102 0.277 — 0.619 — — —
(5)
(−2.77) (4.65) — (1.21) — — 0.023
−0.104 0.261 — — 0.815 — —
(6)
(−2.80) (4.45) — — (2.00) — 0.023
−0.094 0.297 — — — 0.001 —
(7)
(−2.29) (4.65) — — — (0.10) 0.020
Panel C. Regressions controlling for size, B/M, profitability, and an additional valuation ratio
−0.115 0.276 0.371 — — — —
(8)
(−3.12) (4.81) (2.99) — — — 0.022
−0.116 0.285 0.484 −0.256 — — —
(9)
(−3.23) (5.04) (5.07) (−0.55) — — 0.025
−0.117 0.268 0.398 — 0.136 — —
(10)
(−3.25) (4.72) (3.81) — (0.39) — 0.024
−0.118 0.283 0.371 — — −0.003 —
(11)
(−3.22) (4.95) (2.96) — — (−0.22) 0.024

Source: CRSP and Compustat. Methodology: Monthly Fama-Macbeth regressions, following the approach of Fama and French (1992 and 2008). CRSP data provided by
the Center for Research in Security Prices, University of Chicago.

(and their t-statistics) were virtually unchanged when B/M keep the analysis simple by focusing on the US equity
was used as the sole relative price variable, or in tandem market, a well-diversified market with a long history.
with one of E/P, C/P, and S/P, or in tandem with profitability, Specifically, we divide the US equity market in large and
or in tandem with profitability and one of E/P, C/P, and S/P. small caps and form large and small value strategies that
target the top 30% of market capitalization by B/M, E/P,
These regressions suggest that combining multiple variables C/P, and S/P. We also use an equally-weighted composite
to define value vs. growth would not increase expected measure (EW) where each stock’s rank is the average
returns. In the following section, we test this hypothesis of its rank on B/M, E/P, C/P, and S/P. The strategies are
more directly. rebalanced at the end of each December and stocks are
weighted according to their market capitalization. This
MEASURING THE VALUE PREMIUM approach allows us to reduce the number of discretionary
The cross-sectional regression results in Exhibit 1 showed choices about country weighting, stock weighting, and
that B/M, E/P, C/P, and S/P contain reliable information rebalancing frequencies.
about differences in average stock returns. We can further
compare the informational content of each measure by Exhibits 2A and 2B present summary statistics for large
forming value strategies that seek higher-than-market and small value strategies, respectively, over the period
expected returns. Similar to the regression approach, we from January 1964 to December 2015. As suggested by the
DIMENSIONAL FUND ADVISORS 4

Exhibit 2A: US Large Value Strategy Simulations: Lowest 30% of Market Capitalization by Relative Price

January 1964–December 2015 Large B/M E/P C/P S/P EW


Annualized Compound Return 9.97 12.16 12.11 11.79 12.24 12.35
Annualized Standard Deviation 15.03 15.93 15.81 15.28 16.55 16.00
Average Monthly Return 0.89 1.07 1.06 1.03 1.08 1.08
Difference from Large — 0.18 0.17 0.14 0.19 0.19
t-Statistic of Difference from Large — 2.45 2.32 1.95 2.96 2.57
Difference from B/M — — −0.01 −0.04 0.01 0.01
t-Statistic of Difference from B/M — — −0.15 −0.95 0.32 0.50
Fama/French Five-Factor Model Loadings
Rm−Rf (market) 1.00 1.06 1.05 1.02 1.10 1.07
SMB (size) −0.09 −0.05 −0.07 −0.08 0.07 −0.04
HML (value) 0.00 0.51 0.46 0.42 0.37 0.49
RMW (profitability) 0.02 −0.01 0.12 0.09 0.18 0.09
CMA (investment) 0.02 0.03 0.03 0.07 0.06 0.04
R2 1.00 0.95 0.92 0.92 0.95 0.93
Increase in % of Names Turnover vs. Large — 15% 30% 32% 10% 24%
Increase in % of Names Turnover vs. B/M — — 15% 17% −5% 10%

regression results, all the strategies outperform plain large Most importantly for this study, in large caps the returns
or small indices with a t-statistic close to or above 2. The on value strategies formed on E/P, C/P, S/P, and EW are not
outperformance is similar across different variables, around reliably different from the returns on the strategy formed
2% within large caps and 3.5% within small caps. on B/M. For small caps, the small value strategy that uses

Exhibit 2B: US Small Value Strategy Simulations: Lowest 30% of Market Capitalization by Relative Price

January 1964–December 2015 Small B/M E/P C/P S/P EW


Annualized Compound Return 12.71 16.41 15.08 15.78 16.08 16.07
Annualized Standard Deviation 20.84 21.01 19.77 20.42 21.71 20.34
Average Monthly Return 1.19 1.46 1.34 1.40 1.45 1.42
Difference from Small — 0.27 0.15 0.22 0.26 0.23
t-Statistic of Difference from Small — 3.62 1.95 2.86 3.51 2.97
Difference from B/M — — −0.12 −0.06 −0.01 −0.04
t-Statistic of Difference from B/M — — −2.82 −1.45 −0.26 −1.06
Fama/French Five-Factor Model Loadings
Rm−Rf (market) 1.04 1.08 1.03 1.06 1.11 1.06
SMB (size) 0.90 0.90 0.87 0.90 0.99 0.90
HML (value) 0.14 0.64 0.59 0.59 0.58 0.62
RMW (profitability) −0.02 0.11 0.34 0.32 0.28 0.30
CMA (investment) −0.08 −0.05 −0.06 −0.07 −0.03 −0.05
R2 0.99 0.94 0.94 0.94 0.94 0.94
Increase in % of Names Turnover vs. Small — 19% 38% 38% 14% 32%
Increase in % of Names Turnover vs. B/M — — 19% 19% −5% 13%

Source: CRSP and Compustat. Portfolios are rebalanced annually at the end of December. Large is defined as top 1,000 names. Small is defined as names 1,001 and
below. Utilities are excluded from value breaks and from portfolios. CRSP data provided by the Center for Research in Security Prices, University of Chicago. The five
factors are defined as in Fama and French (2015) and are available at http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html. Past performance is
no guarantee of future results. Filters were applied to data retroactively and with the benefit of hindsight. Returns are not representative of indices or actual strategies
and do not reflect costs and fees associated with an actual investment. Actual returns may be lower.
DIMENSIONAL FUND ADVISORS 5

B/M outperformed the other small value strategies. The and S/P. Our analysis focuses on mechanical indices that are
difference in average returns was generally small. rebalanced once a year and have no hold ranges. In a real-
world strategy turnover could be managed and reduced, but
The similar performance across the different relative price these numbers imply that E/P, C/P, and EW value strategies
measures might lead us to conclude that they all target the would likely have higher turnover for a similar level of
same sources of higher expected returns. However, time- expected outperformance vs. the market.
series regression results using the Fama/French five-factor
model show that their performance can be attributed to Besides comparing the average returns generated by the
different sources. The strategies that use B/M to define value alternative value measures, investors might also consider
have a higher HML coefficient relative to the other metrics, other characteristics of the distribution of returns such
which is expected given that the HML factor is also formed as volatility, tracking error, and minimum relative
using B/M sorts. The portfolios that use E/P, C/P, S/P, and performance. Exhibits 2A and 2B show the volatility of
EW tradeoff lower HML loadings with higher profitability each of the large value or small value strategies were similar.
(RMW) loadings. This result highlights the importance of From Exhibits 2C and 2D we observe that the large and
understanding the interaction between premiums and the small value strategies had similar tracking error (TE) to the
risk of inadvertent exposures for investors who prefer to market, but EW has the highest point estimate in both large
pursue multiple premiums in a single strategy or who prefer and small. A similar picture is painted by the maximum
to blend value and quality strategies. and minimum rolling five-year return differences. The
higher maximum outperformance offered by EW is more
We believe investors who seek to maximize their expected than offset by its lower minimum underperformance. Thus,
net returns should also consider the turnover associated the value strategies considered here that combine multiple
with different valuation ratios, since higher turnover relative price metrics did not “smooth out” the ride when
generally leads to higher trading costs. Exhibits 2A and measured on an absolute or relative basis.
2B present turnover statistics for large and small value
strategies relative to plain large or small indices. Turnover A final consideration should be made about the differences
is computed as the number of new names in the portfolio in year-on-year or decade-on-decade realized returns
as a percentage of the total number of names in the across value strategies. Even though on expectation all
portfolio. The tables show that all the value strategies had of the value measures have similar returns, their realized
higher turnover than the relevant market strategies. The returns can be substantially different over short periods
B/M and S/P value strategies had similar turnover. The E/P, of time. Indeed, all of the alternative measures have high
C/P, and EW value strategies had higher turnover than B/M correlation to B/M, but their tracking error to B/M is close

Exhibit 2C: US Large Value Strategy Simulations: Lowest 30% of Market Capitalization by Relative Price

January 1964–December 2015 Large B/M E/P C/P S/P EW


Annualized TE with Large — 6.28 6.40 6.23 5.63 6.46
Max. Rolling 5-Year Relative Performance (Ann.) — 12.55 11.09 12.23 13.99 12.91
Min. Rolling 5-Year Relative Performance (Ann.) — −5.50 −5.20 −7.45 −5.21 −7.40
Monthly Correlation with B/M — — 0.98 0.98 0.97 0.99
Annualized TE with B/M — — 3.50 3.35 3.96 2.56
Compound Returns: 1964–1969 7.14 8.16 8.44 8.23 9.55 8.55
Compound Returns: 1970–1979 5.73 11.73 11.37 11.56 9.26 11.76
Compound Returns: 1980–1989 17.43 20.25 18.13 17.95 19.75 19.21
Compound Returns: 1990–1999 18.25 17.42 17.05 15.59 17.43 15.96
Compound Returns: 2000–2009 −0.47 2.65 3.96 3.92 4.65 4.92
Compound Returns: 2010–2015 12.99 11.91 13.33 13.15 12.51 12.95
DIMENSIONAL FUND ADVISORS 6

Exhibit 2D: US Small Value Strategy Simulations: Lowest 30% of Market Capitalization by Relative Price

January 1964–December 2015 Small B/M E/P C/P S/P EW


Annualized TE with Small — 6.49 6.80 6.53 6.44 6.83
Max. Rolling 5-Year Relative Performance (Ann.) — 17.79 17.52 18.17 18.67 18.89
Min. Rolling 5-Year Relative Performance (Ann.) — −2.94 −5.80 −7.78 −7.72 −6.95
Monthly Correlation with B/M — — 0.99 0.99 0.99 0.99
Annualized TE with B/M — — 3.61 3.29 3.27 2.99
Compound Returns: 1964–1969 19.29 20.43 16.75 20.00 21.69 19.25
Compound Returns: 1970–1979 11.02 15.10 12.92 14.30 12.80 13.90
Compound Returns: 1980–1989 15.65 19.42 19.95 20.85 21.29 22.02
Compound Returns: 1990–1999 15.25 18.13 16.15 14.96 14.82 15.42
Compound Returns: 2000–2009 5.47 13.84 12.20 13.47 14.15 13.89
Compound Returns: 2010–2015 12.64 11.34 12.23 11.22 13.05 11.73

Source: CRSP and Compustat. Portfolios are rebalanced annually at the end of December. Large is defined as top 1,000 names. Small is defined as names 1,001 and
below. Utilities are excluded from value breaks and from portfolios. CRSP data provided by the Center for Research in Security Prices, University of Chicago. Past
performance is no guarantee of future results. Filters were applied to data retroactively and with the benefit of hindsight. Returns are not representative of indices or
actual strategies and do not reflect costs and fees associated with an actual investment. Actual returns may be lower.

to 4%. Unfortunately, it is not possible to say in advance on these dimensions. In this section we study this issue in
which will be the highest for any given period. For example, more detail by analyzing the performance of different value
looking at rolling five-year returns, in large caps B/M had metrics in strategies that integrate multiple premiums. To do
the highest return in 29% of the rolling periods, E/P 15%, this we simulate large cap strategies that emphasize securities
C/P 17%, S/P 31%, and EW 8%. In small caps, B/M had with higher profitability and lower relative price, where
the highest return in 37% of the periods, E/P 4%, C/P 20%, relative price is measured using B/M, E/P, C/P, S/P, and the
S/P 24%, and EW 16%. Similarly, the strategy with the equally weighted average of the ranks EW.
highest return varied decade-by-decade. Given the tracking
error between the strategies, and that the average long-term Each of the strategies overweights the firms with lower
returns of the different value strategies are similar, this is relative price and higher profitability within the large cap
expected. We cannot reliably predict which variable will universe. On average each strategy allocates approximately
yield higher returns over any time period. 70% of its weight to the stocks that are in the high 50%
relative price group and 70% of its weight to the stocks in
What’s the take away? In a well-diversified market, over the the high 50% profitability group.
past 52 years, value strategies formed using B/M, E/P, C/P,
S/P, or EW had similar average returns after controlling Exhibit 3 shows the results for US large cap portfolios. Over
for company size. They also had similar volatilities, similar the period from January 1964 to December 2015, the annual
tracking error, and similar max underperformance relative outperformance of all the tilted portfolios relative to the
to plain large or plain small strategy. Based on this evidence, market cap weighted large cap portfolio is reliably positive
we would declare a tie. How might the tie be broken? and close to 1%. Return differences among all the tilted
Expected turnover (lower is better) and how each variable portfolios are small and not reliably different from zero.
interacts with other sources of expected return can help
break this tie. Regression results using the Fama/French five-factor model
show differences in HML and RMW coefficients. Similarly
CAPTURING THE VALUE PREMIUM to what we observed in the previous section, the portfolio
In the previous sections we saw that the premiums associated that uses B/M to define value has a higher HML coefficient
with E/P, C/P, and S/P were due to both value and profitability relative to the other tilted portfolios, while the portfolios
exposure. As a consequence these measures are unlikely to that use E/P, C/P, S/P, and EW tradeoff lower HML loadings
have incremental benefit for a portfolio that already focuses
DIMENSIONAL FUND ADVISORS 7

Exhibit 3: Emphasizing Lower Relative Price and Higher Profitability within US Large Cap Stocks

January 1964–December 2015 Large B/M E/P C/P S/P EW


Annualized Compound Return (Simulated) 9.97 11.20 10.95 10.92 11.04 11.02
Annualized Standard Deviation 15.03 15.18 15.03 14.93 15.33 15.10
Average Monthly Return 0.89 0.99 0.96 0.96 0.98 0.97
Difference from Large — 0.10 0.07 0.07 0.09 0.08
t-Statistic of Difference from Large — 2.75 2.03 2.01 2.39 2.17
Difference from B/M — — −0.02 −0.02 −0.01 −0.01
t-Statistic of Difference from B/M — — −1.17 −1.41 −0.58 −1.17
Fama/French Five-Factor Model Loadings
Rm−Rf (market) — 1.03 1.02 1.01 1.03 1.02
SMB (size) — −0.01 −0.03 −0.02 0.03 −0.01
HML (value) — 0.20 0.14 0.12 0.13 0.17
RMW (profitability) — 0.17 0.24 0.24 0.26 0.23
CMA (investment) — 0.04 0.07 0.08 0.08 0.06
R2 — 0.98 0.98 0.98 0.98 0.98

Source: CRSP and Compustat. Portfolios are rebalanced annually at the end of December. Large is defined as top 1,000 names. Utilities are excluded from value breaks
and profitability breaks but included in the portfolios. CRSP data provided by the Center for Research in Security Prices, University of Chicago. The five factors are
defined as in Fama and French (2015) and are available at http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html. Past performance is no
guarantee of future results. Filters were applied to data retroactively and with the benefit of hindsight. Returns are not representative of indices or actual strategies and
do not reflect costs and fees associated with an actual investment. Actual returns may be lower.

with higher RMW loadings. These two effects tend to net Exhibit 4 displays similar results for large cap strategies in
out yielding very similar average returns across simulations. developed ex US markets over the period from January 1990

Exhibit 4: Emphasizing Lower Relative Price and Higher Profitability within Developed ex US and Emerging
Markets Large Cap Stocks

Large B/M E/P C/P S/P EW


Panel A: Developed ex US (January 1990–December 2015)
Annualized Compound Return (Simulated) 4.40 5.94 6.07 6.15 5.84 6.01
Annualized Standard Deviation 17.09 17.36 17.08 17.15 17.16 17.22
Average Monthly Return 0.48 0.61 0.61 0.62 0.60 0.61
Difference from Large — 0.13 0.13 0.14 0.12 0.13
t-Statistic of Difference from Large — 3.56 3.64 3.86 3.44 3.57
Difference from B/M — — 0.01 0.01 −0.01 0.00
t-Statistic of Difference from B/M — — 0.25 0.49 −0.59 0.25
Panel B: Emerging Markets (January 1994–December 2015)
Annualized Compound Return (Simulated) 4.78 6.60 6.71 6.58 6.99 6.99
Annualized Standard Deviation 22.56 22.95 22.79 22.60 22.82 22.86
Average Monthly Return 0.61 0.76 0.76 0.75 0.79 0.79
Difference from Large — 0.15 0.16 0.14 0.18 0.18
t-Statistic of Difference from Large — 4.04 3.58 3.54 4.34 4.48
Difference from B/M — — 0.01 −0.01 0.03 0.03
t-Statistic of Difference from B/M — — 0.15 −0.25 1.01 1.55

Source: Bloomberg. Portfolios are rebalanced annually at the end of December. Past performance is no guarantee of future results. Filters were applied to data
retroactively and with the benefit of hindsight. Returns are not representative of indices or actual strategies and do not reflect costs and fees associated with an actual
investment. Actual returns may be lower.
DIMENSIONAL FUND ADVISORS 8

to December 2015, and emerging markets from January 1994 We believe to effectively pursue higher expected returns
to December 2015.6 These strategies have a similar emphasis requires expertise in every step of the investment process:
on lower relative price and higher profitability as the US from identifying reliable drivers of expected returns to
strategies in Exhibit 3. In both developed and emerging designing, managing, and trading portfolios. All steps in the
markets, the tilted portfolios using E/P, C/P, S/P, and the process need to work together. So when evaluating different
EW composite variable have similar performance to the metrics for defining a premium, we are careful about adding
portfolio formed using B/M. Differences across strategies are constraints that can reduce flexibility and increase execution
small and not reliably different from zero. Thus, there is no costs. This approach has enabled Dimensional to capture
compelling evidence that using additional valuation ratios the value premium for our clients for almost 25 years.
can help investors further improve returns in portfolios that
already emphasize size, profitability, and B/M. REFERENCES
Davis, Jim and Inmoo Lee. 2008. “Defining Value
CONCLUSIONS and Growth: Implications for Returns and Turnover.”
It is sensible to incorporate additional variables if they Dimensional Fund Advisors.
enhance our understanding of returns. The better we
understand the drivers of returns, the better we can design Fama, Eugene F. and Kenneth R. French. 1992. “The Cross-
strategies that seek to deliver systematic outperformance. Section of Expected Stock Returns.” The Journal of Finance
However, additional variables also have the potential to add 47. 427–65.
constraints to a portfolio or can make it more difficult to
control turnover. We believe it is important to consider this Fama, Eugene F. and Kenneth R. French. 2008. “Dissecting
tradeoff when making decisions about what premiums to Anomalies.” The Journal of Finance 63. 1653–78.
pursue and how to define them.
Fama, Eugene F. and Kenneth R. French. 2015. “A Five-
In the case of value, there are many different variables that Factor Asset Pricing Model.” Journal of Financial Economics
have been suggested as measures of relative price. However, 116. 1: 1–22.
the evidence in this paper suggests that earnings-to-price,
cash flow-to-price, sales-to-price or a blend of these metrics Fama, Eugene F. and Kenneth R. French. 2015. “Incremental
do not contain additional information about expected returns Variables and the Investment Opportunity Set.” Journal of
beyond that contained in book-to-market and profitability. Financial Economics 117. 3: 470–488.

Caution is needed when evaluating back-tested simulations Novy-Marx, Robert. 2015. “Backtesting Strategies based on
that use many variables. We believe that including multiple Multiple Signals.” National Bureau of Economic Research,
variables in a model could help explain the specific patterns No. w21329.
in historical data very well without actually improving
the model’s ability to explain systematic differences in O’Reilly, Gerard and Savina Rizova. “Expected Profitability:
expected returns. Novy-Marx (2015) shows that back-tested A New Dimension of Expected Returns.” Dimensional Fund
performance of strategies that use multiple metrics can Advisors’ Quarterly Institutional Review 8, no. 1 (2013): 4–7.
appear to be “highly significant” even when the variables
have no true relation to returns. Thus, the greater the
complexity, the more difficult it is to assess whether a result
is reliable or due to chance.

6. Bloomberg data offer an extensive coverage of the developed ex US markets since 1990 and emerging markets since 1994.
DIMENSIONAL FUND ADVISORS 9

APPENDIX
The Fama-French (2015) factors are constructed using the interest expense, and selling, general, and administrative
portfolios formed on market equity (Size), book-to-market expenses divided by book equity for the last fiscal year end
(B/M), operating profitability (OP), and asset growth in t-1. Inv for June of year t is the change in total assets from
(Inv). These portfolios, which are constructed at the end of the fiscal year ending in year t-2 to the fiscal year ending
each June, are based on the intersections of two portfolios in t-1, divided by t-2 total assets. The B/M, OP, and Inv
formed on Size and three portfolios formed on B/M, OP, breakpoints are the 30th and 70th NYSE percentiles. The
and Inv. The Size breakpoint for year t is the median NYSE portfolios for July of year t to June of t-1 include all NYSE,
market equity at the end of June of year t. B/M for June AMEX, and NASDAQ stocks that have market equity data
of year t is the book equity for the last fiscal year end in for December of t-1 and June of t, and (positive) book
t-1 divided by market equity for December of t-1. OP for equity data for t-1.
June of year t is annual revenues minus cost of goods sold,

This information is provided for registered investment advisors and institutional investors and is not intended for public use.
Dimensional Fund Advisors LP is an investment advisor registered with the Securities and Exchange Commission.
The returns of the simulated portfolios are based on model/back-tested simulations. The performance was achieved with
the retroactive application of models designed with the benefit of hindsight; it does not represent actual investment
performance. Back-tested simulated performance is hypothetical (it does not reflect trading in actual accounts) and is
provided for informational purposes only. Simulated performance may not reflect the impact that economic and market
factors might have had on the advisor’s decision making if the advisor had been actually managing client money. Past
performance is no guarantee of future results. Diversification does not protect against loss in declining markets. There is no
guarantee strategies will be successful. Investing involves risk, including loss of principal.
All expressions of opinion are subject to change. This information is intended for educational purposes, and it is not to be
construed as an offer, solicitation, recommendation, or endorsement of any particular security, products or services.
Eugene Fama and Ken French are members of the Board of Directors for and provide consulting services to Dimensional Fund
Advisors LP. Robert Novy-Marx provides consulting services to Dimensional Fund Advisors LP. The following persons listed
as references are current or past employees of Dimensional Investment LLC, a subsidiary of Dimensional Fund Advisors LP:
Gerard O’Reilly, Jim Davis, and Inmoo Lee.

RM53283 08/16 1056793

You might also like