Factor Investing Revisited David Blitz
Factor Investing Revisited David Blitz
David Blitz *
Abstract
This paper takes another look at the recommendation of Blitz [2012] to allocate
strategically to the value, momentum and low-volatility factor premiums in the equity
market. Five years of fresh data shows that such a factor investing strategy continued to
deliver out-of-sample. The potential added value of the two new factors in the Fama-
French 5-factor model, operating profitability and investment, is investigated and found
to depend critically on the performance metric that is considered most important. The
paper also reviews the role of small-cap stocks, factor timing, long-only versus long-short
portfolio construction, international evidence and factor investing beyond equities.
* David Blitz, Ph.D. is the Head of Quantitative Equity Research at Robeco Asset Management and can be
contacted at [email protected] and +31 (0)10 - 224 2079. The views expressed in this paper are solely
those of the author and not necessarily shared by Robeco or its subsidiaries.
Electroniccopy
Electronic copy available
available at:
at:https://ssrn.com/abstract=2626336
http://ssrn.com/abstract=2626336
The literature on asset pricing and mutual fund performance distinguishes between three
different sources of return: (i) exposure to the market risk premium, (ii) exposure to
known factor premiums, such as the value premium, and (iii) alpha, or managerial skill.
Because active investing is a zero-sum game before costs and the evidence for alpha is
weak, investors increasingly choose to focus on passively replicating the capitalization-
weighted market index at minimal costs. Although the market risk premium may be
harvested efficiently in this way, it does mean missing out on factor premiums, for which
the evidence is also strong.
Blitz [2012] argues that investors should allocate strategically not only to the risk
premiums offered by traditional asset classes, but also to factor premiums. More
specifically, the study identifies value, momentum, and low-volatility as the three key
factor premiums in the equity market, and argues for a sizable and well-diversified
allocation to these three factors, based on data spanning the 1963-2009 period.
This paper takes another look at such a factor-investing approach. Five years of
out-of-sample data show that the proposed factor mix continued to deliver over the 2010-
2014 period. The potential added value of the two new factors proposed by Fama and
French [2015] to augment their [1993] 3-factor model, operating profitability and
investment, is examined and found to depend critically on whether one’s objective is to
maximize absolute or market-relative performance. Finally, the role of small-cap stocks,
factor timing, long-only versus long-short portfolio construction, international evidence
and factor investing beyond equities are discussed.
DATA
Data is obtained primarily from the online data library of professor Kenneth French. The
market portfolio represents the value-weighted return of the entire CRSP universe at each
point in time. The value, momentum, operating profitability and investment stock
portfolios are each top 30% portfolios based on the stocks in the CRSP universe with an
above NYSE-median market capitalization, i.e. big caps. By ignoring the small-cap
segment of the market, where factor premiums are actually known to be bigger (see, e.g.,
Fama and French [2012]), we aim to ensure that the strategies can be implemented on a
large scale, and that the impact of transaction costs that would be incurred in a real-life
Electroniccopy
Electronic copy available
available at:
at:https://ssrn.com/abstract=2626336
http://ssrn.com/abstract=2626336
application is limited. The value strategy selects stocks on their book-to-market ratio,
momentum considers past 12-1 month return, operating profitability is defined as annual
revenues minus cost of goods sold, interest expense, and selling, general, and
administrative expenses divided by book equity, and investment takes the change in total
assets (here low is actually better). The low-volatility strategy is not provided by Kenneth
French and therefore self-constructed, following a similar methodology. Specifically, it
selects the 30% stocks with the lowest past 36-month total-volatility in the CRSP
universe of stocks with a market capitalization above the NYSE median. The factor
strategies are considered on an equally-weighted basis as well as on a value-weighted
basis. All returns are taken in excess of the risk-free return provided by Kenneth French.
The sample covers the period from July 1963 until December 2014.
RESULTS
This section discusses the main empirical results, looking first at performance over time,
and next at whether the two new Fama-French factors might add value.
Exhibit 2 shows performance statistics over the most recent 5-year period, from
January 2010 until December 2014. Over this out-of-sample period the market itself
already showed a very high excess return of 15.6% per annum, arguably reducing the
need for a more sophisticated approach. Still, the (equally-weighted) 3x1/3 portfolio was
able to boost return by an additional 1.9%, and improve the Sharpe ratio from 1.16 to
1.27. Looking at the performance of the individual factors we see that low-volatility was
particularly strong, with a Sharpe ratio of 1.76 and a CAPM alpha of 6.9%. This was
sufficient to offset the relatively weak performance of value and momentum, which did
manage to generate a higher raw return than the market portfolio, but not a superior risk-
adjusted return (lower Sharpe ratio and negative CAPM alpha). The 3x1/3 portfolio based
on value-weighted factor strategies also shows a higher Sharpe ratio and a positive
CAPM alpha, but the improvement is only about half that of the equally-weighted factor
strategies. Once more, low-volatility saves the day.
Although the diversified factor investing strategy continued to deliver in the most
recent 5-year period, one might be concerned that the added value is below the long-term
historical average of Exhibit 1. In order to provide a better feeling for the short- and
medium-term performance of the 3x1/3 strategy, Exhibit 3 plots its 1-, 3- and 5-year
rolling market-relative return. Based on these graphs it is clear that the recent
performance of the 3x1/3 strategy is well within the historically observed range of
outcomes. It also shows that even periods of underperformance are not unusual. Exhibit 4
shows the full-sample estimated probabilities of underperformance over 1-, 3- and 5-year
periods for the 3x1/3 strategy. For the equally-weighted factor investing strategy, the
probability of underperformance is 28% on a 1-year basis, 13% on a 3-year basis and 7%
on a 5-year basis. So even though the strategy added considerable value over the 50+ year
In this case the two new Fama-French factors suddenly become a lot more
interesting, receiving combined weights of no less than 30.2% and 53.5% (for equally-
weighted and value-weighted factor portfolios respectively), mainly at the expense of the
weight given to the value and low-volatility factors. Compared to the 3x1/3 portfolio this
improves the information ratios by amounts that cannot be dismissed as negligible, from
0.75 to 0.90 and from 0.46 to 0.73 respectively. These improvements are primarily driven
by a reduction of tracking error. Thus, the two new Fama-French factors appear to be
much more attractive from an information-ratio perspective than from a Sharpe-ratio
perspective.
The finding that a low-volatility strategy is less attractive from an information-
ratio perspective is not surprising, because benchmark-driven investing has been
identified as one of the reasons this anomaly exists in the first place; see, for instance,
Blitz and van Vliet [2007] and Baker, Bradley and Wurgler [2011]. The finding that little
weight is given to value in the portfolios optimized for information ratio may relate to the
observation of Fama and Fench [2015] that value seems to have become redundant in
their 5-factor model. However, Asness, Frazzini, Israel and Moskowitz [2015] show that
a modified value factor, based on book-to-market ratios that assume a publication lag for
book values but not for market values, is not subsumed by other factors.
EXTENSIONS
This section discusses various extensions of the basic value, momentum and low-
volatility factor mix for U.S. equities discussed in the previous section.
Small-cap
Next to the value, momentum and low-volatility factor strategies Blitz [2012] also
considers small-cap stocks as a potential fourth factor strategy. Although the return
premium offered by the small-cap strategy is found to be positive, it is markedly lower
than the premium for the value, momentum and low-volatility strategies. As a result,
small-cap does not make it to the optimal factor mix.
Factor timing
Potentially, the performance of a diversified factor mix, such as the 3x1/3 strategy
discussed in the previous section, could be improved significantly by tactically varying
10
International evidence
The results presented in Blitz [2012] and this paper are entirely based on U.S. equity
market data. However, value, momentum and low-volatility have also been shown to be
strongly present in international equity markets. Fama and French [1998] provide
evidence for value in Europe and Japan, Rouwenhorst [1998] provides evidence for
momentum in Europe, and Fama and French [2012] reconfirm the effectiveness of value
and momentum in global developed equity markets. Evidence for value and momentum
in emerging markets is provided by Van der Hart, Slagter and van Dijk [2003]. The low-
volatility effect is documented by Blitz and van Vliet [2007] for the U.S., Europe and
Japan, and by Blitz, Pang and van Vliet [2013] for emerging markets. The only exception
seems to be the absence of a clear momentum effect in Japan, although Asness [2011]
argues that after adjusting for the strong negative interaction between momentum and
value strategies, a strong momentum effect is, in fact, also present in Japan. Altogether
these results imply that the value, momentum and low-volatility factor premiums are
quite robust and that factor investing can be applied to the entire global equity portfolio.
11
12
SUMMARY
This paper takes another look at the recommendation of Blitz [2012] to allocate
strategically to the value, momentum and low-volatility factor premiums in the equity
market. Five years of fresh data shows that such a factor investing strategy continued to
deliver out-of-sample. Although performance is lower than over the full sample, it is well
within the historically observed range of outcomes. The two new factors in the Fama-
French 5-factor model, operating profitability and investment, do not appear to be helpful
for further improving absolute performance, but offer more potential for investors
interested in benchmark-relative performance. Small-cap stocks seem to be more
attractive as a catalyst for unlocking the full potential of value, momentum and low-
volatility, than as a basis for a fourth factor next to these three.
The literature provides little support for the possibility to successfully time the
exposure to different factors, which argues in favor of establishing a well-diversified mix
of lowly correlated factors. Although a long-short approach to factor investing may
appear to be superior in theory, practical considerations such as costs, benchmarking and
existing allocations make a long-only approach more appropriate for most investors. As
factor premiums have also been documented for international equity markets, factor
investing can be applied to the global equity portfolio. In fact, it may even be applied to
the entire portfolio, because factor premiums appear to transcend beyond the equity
market, to asset classes such as corporate bonds and commodity futures.
13
14
15
CAPM beta 1.00 0.98 1.09 0.70 1.09 1.05 0.92 0.98
CAPM alpha 0.0% 4.3% 4.2% 3.2% 1.8% 3.0% 4.1% 3.5%
CAPM beta 1.00 0.91 1.00 0.74 0.95 0.92 0.88 0.91
CAPM alpha 0.0% 2.6% 3.0% 1.5% 1.2% 2.4% 2.6% 2.3%
16
CAPM beta 1.00 1.09 1.18 0.70 1.11 1.19 0.99 1.06
CAPM alpha 0.0% -0.4% -0.8% 6.9% 0.6% 0.9% 2.0% 1.5%
CAPM beta 1.00 1.06 1.08 0.69 0.92 0.96 0.94 0.94
CAPM alpha 0.0% -2.2% -0.2% 4.7% 1.4% 1.2% 0.9% 1.1%
17
30%
20%
10%
0%
-10%
-20%
Jul-63
Jul-66
Jul-69
Jul-72
Jul-75
Jul-78
Jul-81
Jul-84
Jul-87
Jul-90
Jul-93
Jul-96
Jul-99
Jul-02
Jul-05
Jul-08
Jul-11
Jul-14
Equally-weighted Value-weighted
15%
10%
5%
0%
-5%
-10%
Jul-63
Jul-66
Jul-69
Jul-72
Jul-75
Jul-78
Jul-81
Jul-84
Jul-87
Jul-90
Jul-93
Jul-96
Jul-99
Jul-02
Jul-05
Jul-08
Jul-11
Jul-14
Equally-weighted Value-weighted
18
15%
10%
5%
0%
-5%
-10%
Jul-63
Jul-66
Jul-69
Jul-72
Jul-75
Jul-78
Jul-81
Jul-84
Jul-87
Jul-90
Jul-93
Jul-96
Jul-99
Jul-02
Jul-05
Jul-08
Jul-11
Jul-14
Equally-weighted Value-weighted
19
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
1Y 3Y 5Y
Equally-weighted Value-weighted
20
10% MomWin
Value
8%
Excess return
LowVol
6%
Growth HighVol
4% Market
2%
MomLos
0%
0% 5% 10% 15% 20% 25% 30%
Volatility
6%
Excess return
LowVol
5% Market
Growth
4%
HighVol
3%
2%
MomLos
1%
0%
0% 5% 10% 15% 20% 25% 30%
Volatility
21
22