Investment Research
An Industry Selection Model
Does Industry Selection Add Up?
Ciprian Marin, Director, Portfolio Manager/Analyst
An industry selection model can be defined as a systematic tool that tilts a portfolio towards industries that are predicted
to outperform and underweights those that are predicted to underperform. If the process can discern winning/losing
industry groups, it can enhance the value added from a stock-selection methodology. While this is a relatively simple
concept, it has historically been difficult to add value through this method without increasing the volatility of the returns.
In this paper, we present an industry selection model and its return sources. We also show that such a model can
potentially add consistent returns at reasonable risk levels. We find that momentum is the most effective factor for
selecting industries but, when used on its own, it does materially increase volatility. The use of other factors diversifies
the model and provides some insurance against momentum swings. Our results indicate that industry selection can play
a role in a portfolio but, as we will detail, we believe implementation with other models is crucial to maximize Sharpe ratios.
2
Introduction
Methodology and Research Results
The high volatility of individual industry returns is a wellestablished characteristic of equity markets. As a result, we believe
industry selection is important in achieving consistent returns and,
in many environments (e.g., 2015 when oil exposure dominated
return outcomes), industry selection can account for a large
component of relative performance. Yet, industry selection is
difficult and models that implement it are prone to volatile
return patterns.
Stock selection models show that certain factors are better than
others at selecting sectors or industries. In order to build our model
for this research, we analyzed the performance of fundamental and
technical factors at an industry level in order to build an industry
selection model.
On the other hand, industry-neutral style returns can have very
low volatility. The outcomes of a successful industry-neutral stock
selection model should be consistent outperformance and high
information ratios. The drawback with this approach, however,
is that the outperformance may be lower than some investors’
objectives. This can be exacerbated by periods of low or decreasing
risk—in fact there are periods when the relative risk (tracking
error) of active portfolios decreases to 50% of the previous level.1
For a systematic investor, generating high relative returns on low
tracking errors may prove difficult.
Is there an approach to industry selection that adds returns to an
industry-neutral stock model without increasing risk? Or, in other
terms, can an investor increase alpha through controlled industry
bets? In this paper, we explore these questions and explain how we
believe our research shows that it is possible to create an industry
selection model based on diversified return sources. This means
that industry allocations can enhance stock selection models,
enabling systematic investors to potentially deliver higher returns
without sacrificing the consistency that is the hallmark of many
systematic approaches.
Data
For our research, we have built the industry model from the bottom
up, starting with all the stocks in the S&P World Developed Broad
Market Index between 1994 and 2015. In the exploratory phase of
the analysis, we report data through 2014 (Exhibits 1–3) and for the
final model we added data through March 2015.2
We use free-float adjusted market capitalization (greater than $100
million) from the S&P/DJ Index database and stock returns from
EXshare/FactSet. For fundamental items we use both FactSet
Fundamentals and Worldscope databases. We use the GICS
definition when classifying sectors and industries. The GICS
structure consists of 10 sectors, 24 industry groups, 67 industries,
and 156 sub-industries.
Momentum
Momentum at the sector level was a natural starting point for our
research. We have long been aware of the power of momentum
(the measure of market sentiment) at a sector level.
However, we used sub-industries (and industries for the final
model), not sectors, because more groups make for a statistically
stronger signal. To clarify, we define momentum as past
performance of the sector/industry. For example, by three-month
industry momentum we mean the performance of a portfolio
which buys the top 20% of industries with the best performance
over the previous three months and sells the bottom 20% of worstperforming industries over the same period.
First, we noticed that sub-industry momentum is very strong
(Exhibit 1). One-month sub-industry momentum contributed, on
average, nearly 70 basis points (bps) per month. At three-months
momentum, returns rose to 87 bps. This impact was persistent
over different time periods, and even when momentum was
calculated over three years, it still contributed 47 bps. However,
it is important to note that standard deviation was very high,
meaning the Sharpe ratio was low.
Exhibit 1
Sub-Industry Momentum Is a Strong Signal
Sub-Industry Momentum 1997–2014
Mean Return
(Monthly; %)
Standard Deviation
(Monthly; %)
Sharpe Ratio
(Annualized)
1-Month
0.69
4.41
0.54
3-Month
0.87
4.54
0.67
12-Month
0.84
5.28
0.55
3-Year
0.47
4.91
0.33
Momentum Factor
As of 31 December 2014
The performance quoted represents past performance. Past performance is not a
reliable indicator of future results. This information is for illustrative purposes only and
does not represent any product or strategy managed by Lazard.
Source: FactSet, Lazard, Standard & Poor’s, Worldscope
3
In our view, these data are very powerful, but what is driving
performance—common factors across the entire market or
industry-specific drivers?
To answer this question, one needs to separate how much is
coming from the systematic component (i.e., the market and
underlying trends in the market) and how much is industryspecific (e.g., earnings in an industry having a bull run). We
accomplish this by decomposing returns into its systematic and
residual parts using a statistical model.
The statistical model we use is a principal component analysis
on sub-industry returns. It is similar to a Barra or Axioma factor
model with the difference that factors are not specified in advance
but are selected by the process in such a way that the factors
selected explain the highest amount of variance in the data.3 We
initially used a three-factor statistical model.
The first factor is usually the market, as in most statistical models
the market accounts for most of the volatility in returns. The
second factor is the dominant factor in the respective period. For
example last year, it was exposure to the oil price. In 2011, it was
exposure to Greece and, in 2008, it was credit exposure. The third
factor, which can vary, is the next biggest contributor to volatility.
Once this process was completed we separated sub-industry returns:
• Systematic returns: explained by the three factors.
• Residual returns: unexplained by the statistical model.
Based on our research, systematic-driven momentum returns are
much higher than the residual returns, and they seem to drive
total momentum returns (Exhibit 2). The systematic component is
responsible for 74 bps and the residual component is responsible
for 20 bps of one-month momentum.4 However, the residual
returns do have far less volatility, resulting in comparable Sharpe
ratios to those of the systematic returns.
The question that now emerges is how much of these systematic
returns are purely market driven, i.e., is industry selection mostly
about selecting high-beta industries when the market is doing
well? If that were the case, industry selection would really be about
market timing—which is a difficult and even dangerous practice,
in our view.
To answer this question, we reduced the three factors to one—the
market. The results show that the one-factor systematic momentum
(just the market) contributes less to momentum returns than when
we used three factors, but importantly also accounts for most of
the volatility (Exhibit 3). In other words, the market (or market
timing) is not purely driving momentum returns. This finding
encouraged us to further investigate the source of the returns.
However, we were unable to completely separate the systematic
and residual components. Our hope was that we could just use the
residual component of the returns because of its lower volatility.
This is possible in a stock selection model, which can therefore
potentially generate consistent outperformance with low risk.
In contrast, sub-industry momentum returns are driven by both
systematic and residual factors. Using just the residual returns
would imply losing some power in the model. Retaining systematic
returns does increase volatility, but we believe the higher returns
make it a worthwhile trade-off.
Exhibit 2
Systematic Returns Drive Momentum
Exhibit 3
The Market Is Not the Only Factor Driving Momentum Returns
Decomposing Sub-Industry Momentum Returns Into Systematic and
Residual Using a 3-Factor Statistical Model: 1997–2014
Decomposing Sub-Industry Momentum Returns Into Systematic and
Residual Using a 1-Factor Statistical Model: 1997-2014
Mean Return
(Monthly; %)
Standard Deviation
(Monthly; %)
Sharpe Ratio
(Annualized)
Systematic Momentum
1 Month
0.74
4.99
0.51
Systematic Momentum
12 Months
0.74
5.31
Residual Momentum
1 Month
0.20
Residual Momentum
12 Months
0.37
Factors
Mean Return
(Monthly; %)
Standard Deviation
(Monthly; %)
Sharpe Ratio
(Annualized)
Systematic Momentum
1 Month
0.44
5.31
0.29
0.48
Systematic Momentum
12 Months
0.36
5.30
0.23
1.96
0.36
Residual Momentum
1 Month
0.63
3.02
0.73
2.39
0.53
Residual Momentum
12 Months
0.84
3.85
0.76
Factors
As of 31 December 2014
As of 31 December 2014
The performance quoted represents past performance. Past performance is not a
reliable indicator of future results. This information is for illustrative purposes only and
does not represent any product or strategy managed by Lazard.
The performance quoted represents past performance. Past performance is not a
reliable indicator of future results. This information is for illustrative purposes only and
does not represent any product or strategy managed by Lazard.
Source: FactSet, Lazard, Standard & Poor’s, Worldscope
Source: FactSet, Lazard, Standard & Poor’s, Worldscope
4
When implementing this model, we tried to maintain the power
of a stock selection model by using industry selection only as an
overlay strategy. This means that when ranking stocks, if any stocks
have a high stock-specific alpha and high sub-industry alpha we
would then take a bigger position in the stock.
However, it proved challenging to keep a high exposure to both
our stock selection model and sub-industry model—mostly
because of the limited number of stocks in some sub-industries.
We decided to change our implementation by 1) Using a broader
classification: industries instead of sub-industries, and 2) Adding
other factors to momentum.
Other Factors
We turned to study other factors, as momentum is, of course,
not the only investment style and we did not want our industry
selection model to become momentum biased. With this in
mind, we included a score based on fundamental factors to the
momentum model, and we reviewed a group of fundamental
measures to help select industries.
Interestingly, some of the strongest traditional value measures—
earnings yield, cash flow yield, and book to price—were the
weakest signals and generated negative monthly returns.
On the other hand, we found four factors were helpful for industry
selection: cash flow to assets, operating margin, return on equity
(ROE), and earnings per share (EPS) growth for the last five years.
Just as we did for momentum, the top 20% of industries scored
on each factor was bought and the bottom 20% was sold. The
combination of these four factors contributed 58 bps per month
(Exhibit 4).
Final Model
To construct the final model, we first added an adjustment to soften
the momentum swings. We included the (negative) correlation
with the market in such a way that we penalized industries in
which the market accounted for a large part of the previous
twelve-month returns.5 In simple terms, we wanted to exclude
industries that outperformed because of the market, so we included
a negative correlation term (i.e., the industry correlation with the
market multiplied by -1) as an additional control. This adjustment
smoothed some of the volatility in the momentum model.
The final industry selection model was composed of the following
weights: 60% momentum, 20% (negative) correlations, and 20%
fundamental score.6
While adding the negative correlation and fundamentals to the
overall model generated weaker results (1.30%) than did the standalone momentum factor (1.37%), it diversified the model and
provided some protection against momentum swings (Exhibit 5).
Exhibit 4
Selected Fundamental Factors Add Value to the Model
All Sample Data: 1995–2015
Mean Return
(Monthly; %)
Standard Deviation
(Monthly; %)
Sharpe Ratio
(Annualized)
Earnings Yield
-0.20
4.37
-0.16
Cash Flow Yield
-0.23
4.20
-0.19
Book to Price
-0.64
3.80
-0.58
Cash Flow to Assets
0.37
2.17
0.60
Operating Margin
0.47
2.12
0.76
ROE
0.32
3.36
0.33
EPS Growth – 5 Years
0.34
2.37
0.49
The 4 Selected Factors
0.58
2.75
0.73
Factors
As of 31 March 2015
The performance quoted represents past performance. Past performance is not a
reliable indicator of future results. This information is for illustrative purposes only and
does not represent any product or strategy managed by Lazard.
Source: FactSet, Lazard, Standard & Poor’s, Worldscope
Exhibit 5
Industry Selection Model Adds Value
All Sample Data: 1995–2015
Mean Return
(Monthly; %)
Standard Deviation
(Monthly; %)
Sharpe Ratio
(Annualized )
Momentum
1.37
4.62
1.03
Market Correlation
(Inverse)
0.66
3.72
0.61
Fundamental Score
0.66
2.71
0.85
Industry Alpha Score
1.30
4.51
1.00
Factors
As of 31 March 2015
The performance quoted represents past performance. Past performance is not a
reliable indicator of future results. This information is for illustrative purposes only and
does not represent any product or strategy managed by Lazard.
Source: FactSet, Lazard, Standard & Poor’s, Worldscope
5
Exhibit 6
Consistently Adding Value over Time
Industry Selection Model Factors
(%)
6
4
2
0
-2
1995
Momentum
1999
Correlation Inverse
2003
Fundamental Score
2007
2011
2015
Industry Alpha Score
As of 31 March 2015
The performance quoted represents past performance. Past performance is not a reliable indicator of future results. This information is for illustrative purposes only and does not represent
any product or strategy managed by Lazard.
Source: FactSet, Lazard, Standard & Poor’s, Worldscope
Exhibit 6 shows the performance of the three different components
of the model and the final model over time. In 15 of the 20
back-tested years the industry model added value (denoted by the
industry alpha score).
We believe if a stock is favored by both the stock selection and
industry selection models that security warrants a larger position
size. An industry selection model should not interfere with a stock
selection model but should complement it.
Conclusion
A stock selection model is still the most reliable and effective
way to systematically build a portfolio, in our view. However,
the industry selection model overlay is a compelling choice for
specific asset allocators that understand the trade-offs involved in
the process.
Our research shows that it is possible to create an industry selection
model based on diversified return sources. Although momentum
is the main driver of industry returns and momentum returns are
heavily reliant on macroeconomic conditions, we show that the
market contribution to industry momentum profits has been very
limited and that both common and specific factors have driven
industry momentum returns. We add technical and fundamental
factors to create a diversified and robust industry selection model.
6
This content represents the views of the author(s), and its conclusions may vary from those held elsewhere within Lazard Asset Management.
Lazard is committed to giving our investment professionals the autonomy to develop their own investment views, which are informed by a
robust exchange of ideas throughout the firm.
Notes
1 For example, as a result of monetary easing, market risk dropped significantly between 2011 and 2014.
2 The start dates on the reported results are later than 1994 because items such as three-year momentum would require three extra years of data.
3 Principal Component Analysis uses an iterative procedure which creates linear combinations of the inputs in such a way that the first combination will explain the maximum amount of variance.
4 As additional clarification, the results of the return components do not add up exactly to those shown in Exhibit 1. This follows from the process of first decomposing the returns (into
systematic and residual components) and then calculating the momentum returns. We also are using only 1- and 12-month momentum returns (dropping the 3-month data point) to represent
short- and long-term momentum.
5 Twelve-month momentum made it to the final model because shorter period momentum has higher turnover.
6 Model weights were selected following the in-sample table not presented in this paper.
Important Information
Published on 30 June 2016.
Information and opinions presented have been obtained or derived from sources believed by Lazard to be reliable. Lazard makes no representation as to their accuracy or completeness. All
opinions expressed herein are as of the published date and are subject to change.
Equity securities will fluctuate in price; the value of your investment will thus fluctuate, and this may result in a loss. Securities in certain non-domestic countries may be less liquid, more volatile,
and less subject to governmental supervision than in one’s home market. The values of these securities may be affected by changes in currency rates, application of a country’s specific tax laws,
changes in government administration, and economic and monetary policy.
Hypothetical performance and simulations shown are for illustrative purposes only and do not represent actual performance; they are not a reliable indicator of future results. Hypothetical
performance results may benefit from hindsight, are based on historical data, and do not reflect the impact that certain economic and market factors might have had on the decision-making process.
A quantitative investment strategy relies on quantitative models and quantitative filters, which, if incorrect, may adversely affect performance.
This material is provided by Lazard Asset Management LLC or its affiliates (“Lazard”). There is no guarantee that any projection, forecast, or opinion in this material will be realized. Past performance does not guarantee future results. This document is for informational purposes only and does not constitute an investment agreement or investment advice. References to specific
strategies or securities are provided solely in the context of this document and are not to be considered recommendations by Lazard. Investments in securities and derivatives involve risk,
will fluctuate in price, and may result in losses. Certain securities and derivatives in Lazard’s investment strategies, and alternative strategies in particular, can include high degrees of risk and
volatility, when compared to other securities or strategies. Similarly, certain securities in Lazard’s investment portfolios may trade in less liquid or efficient markets, which can affect investment performance.
Australia: FOR WHOLESALE INVESTORS ONLY. Issued by Lazard Asset Management Pacific Co., ABN 13 064 523 619, AFS License 238432, Level 39 Gateway, 1 Macquarie Place, Sydney
NSW 2000. Dubai: Issued and approved by Lazard Gulf Limited, Gate Village 1, Level 2, Dubai International Financial Centre, PO Box 506644, Dubai, United Arab Emirates. Registered in Dubai
International Financial Centre 0467. Authorised and regulated by the Dubai Financial Services Authority to deal with Professional Clients only. Germany: Issued by Lazard Asset Management
(Deutschland) GmbH, Neue Mainzer Strasse 75, D-60311 Frankfurt am Main. Hong Kong: Issued by Lazard Asset Management (Hong Kong) Limited (AQZ743), Unit 29, Level 8, Two Exchange
Square, 8 Connaught Place, Central, Hong Kong. Lazard Asset Management (Hong Kong) Limited is a corporation licensed by the Hong Kong Securities and Futures Commission to conduct
Type 1 (dealing in securities) and Type 4 (advising on securities) regulated activities. This document is only for “professional investors” as defined under the Hong Kong Securities and Futures
Ordinance (Cap. 571 of the Laws of Hong Kong) and its subsidiary legislation and may not be distributed or otherwise made available to any other person. Japan: Issued by Lazard Japan Asset
Management K.K., ATT Annex 7th Floor, 2-11-7 Akasaka, Minato-ku, Tokyo 107-0052. Korea: Issued by Lazard Korea Asset Management Co. Ltd., 10F Seoul Finance Center, 136 Sejong-daero,
Jung-gu, Seoul, 04520. People’s Republic of China: Issued by Lazard Asset Management. Lazard Asset Management does not carry out business in the P.R.C. and is not a licensed investment
adviser with the China Securities Regulatory Commission or the China Banking Regulatory Commission. This document is for reference only and for intended recipients only. The information
in this document does not constitute any specific investment advice on China capital markets or an offer of securities or investment, tax, legal, or other advice or recommendation or, an offer
to sell or an invitation to apply for any product or service of Lazard Asset Management. Singapore: Issued by Lazard Asset Management (Singapore) Pte. Ltd., 1 Raffles Place, #15-02 One
Raffles Place Tower 1, Singapore 048616. Company Registration Number 201135005W. This document is for “institutional investors” or “accredited investors” as defined under the Securities
and Futures Act, Chapter 289 of Singapore and may not be distributed to any other person. United Kingdom: FOR PROFESSIONAL INVESTORS ONLY. Issued by Lazard Asset Management
Ltd., 50 Stratton Street, London W1J 8LL. Registered in England Number 525667. Authorised and regulated by the Financial Conduct Authority (FCA). United States: Issued by Lazard Asset
Management LLC, 30 Rockefeller Plaza, New York, NY 10112.
LR26889