factor investing - bpmmagazine.com · factor investing has become a topic of interest as it helps...
TRANSCRIPT
Factor Investing Pure and simple
Equity Quantitative Research
April 2017
For Institutional Investors and Advisor Use Only
Factor Investing
Pure and simple
Quantitative Equity Research
HSBC Pure Factor Strategies
Smart beta investing has become increasingly popular in
recent years. Many index providers now offer a broad suite
of smart beta strategies. The most common tend to be
based on the well-established risk premia factors: value,
small cap, momentum, low volatility and quality. However,
HSBC’s research shows that many of these indices exhibit
unintended exposures to unrelated factors because of
their simplistic construction method. A more sophisticated
approach eliminates these unwanted risks, providing a
‘pure’ factor strategy for smart beta investors.
The aim of this paper is to illustrate how HSBC’s
approach emphasizes the importance of factors
being:
► Precise
► Unbiased
► Robust
► Efficient
We investigate a method of measuring the purity of
smart beta strategies based on the portion of active
risk driven by the targeted factor. We find that HSBC’s
pure beta strategies exhibit higher factor efficiency
than commercially available alternatives.
We also compare HSBC’s strategies with conventional
factor implementation to demonstrate the effectiveness
of HSBC’s construction method and the advantages of
factor neutralisation.
Finally, we discuss practical applications to portfolio
management and the value of HSBC’s strategies as a
tool to enhance investment performance.
2 | HSBC Global Asset Management For Institutional Investors and Advisor Use Only
Introduction to Factors
The appeal of smart beta strategies is that they are
systematic and transparent, and thus easy to construct and
rebalance. They can also be an inexpensive way for investors
to obtain exposure to factors they might be lacking within their
portfolios. Many smart beta strategies are constructed with an
emphasis on simplicity, often using simple sorting and
weighting techniques. These are usually based on either a
single factor (e.g. book-to-price) or a composite score (e.g.
value). Other smart beta strategies are put together to
maximise investability, with factor tilts combined with market
cap weighting. Although both these approaches result in
higher exposures to the targeted factor, there is little
restriction on exposures to other factors. This can lead to
unintended factor exposure and taking on undesired risk.
Factor investing has become a topic of interest as it helps
answer a fundamental question: is the concept of diversification
still alive? The financial crisis saw the synchronised movement
of traditionally uncorrelated assets. Supposedly diversified
strategies proved to be less diversified than thought, leading to
dramatic underperformance.
Andrew Ang uses the following analogy to describe factors:
‘factors are to assets what nutrients are to food’. According
to Ang, assets earn risk premia because they are exposed to
the underlying factor risks. Over time a growing proportion of
investment performance has been explained in terms of
factor exposures. Outperformance previously understood as
‘alpha’ is increasingly described as ‘beta’. Beta can come
from equity exposure, style, exotic factors, etc.
Historically, factor investing was considered an active strategy.
Following the recent rise in investor demand for factor
exposures, new cost efficient and highly accessible factor
indices have been introduced. This new dimension in product
design has opened up a set of opportunities designed to
maximise convenience for investors.
Time
Alpha
Alpha
Equity Beta
Alpha
S t y l e
B e t a
Equity Beta
Alpha
Exotic Beta
S t y l e
B e t a
Equity Beta
For Institutional Investors and Advisor Use Only3 | HSBC Global Asset Management
Source: HSBC Global Asset Management
For illustrative purposes only.
Theory behind factor based-investing
CAPM
Returns from a single
systematic risk
APT
Returns from multiple
sources
Fama-French
Value – Size
Cahart
Value – Size –
Momentum
Value – Size –
Momentum -
BAB - QMJ
2014 Frazzini et al.
1970
1976
1993
1997
The Capital Asset Pricing Model (CAPM) was first introduced in the 1960s by
Treynor, Sharpe, Lintner and Mossin. It was the first formal model to capture the
notion of factors being the driving force behind returns. This one-factor model
implies that asset returns can be explained by just a single factor: the sensitivity
of the asset’s excess return to the excess return of the market. This sensitivity is
referred to as the beta of the asset. The intuition behind CAPM is that the
expected return of an asset, which is required to compensate for its
undiversifiable risk, should be a function of its correlated volatility with the
market (ß).
Arbitrage Pricing Theory (APT) was first introduced in 1976 as an alternative
to CAPM and was one of the earliest multi-factor models. Its premise is that
expected returns can be decomposed into a linear combination of factors.
These can be chosen either through economic intuition or through factor
analysis to identify the drivers of returns (a common method is principle
components analysis). The appeal of APT is that it imposes fewer assumptions
and requires less economic structure than CAPM.
One of the best known multi-factor models was introduced by Fama and
French. Using a 50-year dataset between 1941 and 1990, they found that the
link between market beta and average return had been weak. They proposed
adding two factors (size and book-to-market) to the single factor CAPM model to
better explain the cross-section of security returns.
Building on Fama-French legacy Cahart extended the three factor model to
include a momentum factor. The addition of the MOM factor, as it is commonly
known, improved the explanatory power of the model. Until recently it was
considered to be the reference evaluation framework for active management
and mutual funds.
Recently Frazzini et al. introduced a quality factor (QMJ) and a low beta factor
(BAB). This followed the same methodology as Fama-French, extending further
the range of potential valid factors. In addition Novy-Marx introduced a different
quality factor, claiming that it captures alpha.
For Institutional Investors and Advisor Use Only4 | HSBC Global Asset Management
Building Factor Strategies A plethora of factor construction methods have been proposed
in the academic literature, some of which have been
implemented by industry practitioners. In this expanding
ecosystem of factor based products, there is a common
misconception that factor investing is very simple, providing
superior results to traditional funds (e.g. cap-weighted indices,
active management, strategic asset allocation). ‘Raw’
strategies approach factor construction by overweighting
stocks that exhibit a particular characteristic (e.g. Price-to-
Book). To respond to the challenge of transforming academic
risk factors into investable portfolios we focus on four key
qualities: our products must be Precise, Unbiased, Robust and
Efficient.
Precise: The factors we seek exposure to are precisely
defined, guided by empirical research.
Unbiased: Our indices are constructed to remove hidden
bias towards untargeted factors.
Robust: Strong technological infrastructure, proprietary risk
models and the conceptual clarity of our mathematical
formulation ensure robust implementation.
Efficient: Our indices deliver strong factor purity ratios,
exhibiting a high proportion of targeted risk per unit of active
risk.
We refer to this product range as our “pure” factor
strategy.
One of the challenges of factor investing is
determining which factors really drive returns.
Cochrane (2011) referred to a ‘zoo of new factors’
and Harvey et al. (2014) counted over 300 factors,
showing a dramatic increase in recent years. In this
‘zoo’ it is essential to focus only on factors that are
strongly supported by empirical evidence with solid
economic justifications. From this perspective the
value, size, momentum, low volatility and quality
factors seem a natural choice.
1Minimum-torsion refers to a mathematical technique which applies a linear transformation to the original factors in order to find the closest orthogonal (uncorrelated) ones. For more information, see Meucci, Santangelo and Deguest (2013) –Risk Budgeting and Diversification Based on Optimized Uncorrelated Factors.
A transparent and intuitive construction process
Objective: We try to give investors maximum exposure to
a factor, capturing as much of the premium as possible.
The Challenge: Unfortunately this isn’t enough if we want to
focus on multiple ‘independent’ sources of risk/return.
Moving from the theoretical and impractical long-short
portfolios of Fama-French to long only investable solutions
requires an understanding of the correlations and exposures
between factors. There are three ways to tackle this:
► We could impose risk contribution constraints.
► We could apply a transformation algorithm
such as ‘minimum-torsion’1 to approximate the
closest orthogonal (uncorrelated) factors.
► We could apply neutralisation constraints
to unwanted factor exposures.
The first two approaches require parameterisation of the
factor model. This limits transparency when interpreting
individual stock factor exposures.
The Solution: We take the third approach, following our
emphasis on transparency. We also incorporate an active
weight constraint to improve diversification, a capacity
constraint to avoid illiquid names and a turnover constraint
to control costs.
Robust
PreciseSingle target factor
EfficientUnbiasedPURE Factor Purity Ratio
= risk from desired factor
total active risk
Clear Ex–post Return Attribution
Clear Ex–anteRisk Decomposition
For Institutional Investors and Advisor Use Only5 | HSBC Global Asset Management
Source: HSBC Global Asset Management
For illustrative purposes only.
Style Neutrality: A focus on premia purity and approximate independence from other sources of risk/return is essential to building
efficient strategies. Factor neutralisation relative to the benchmark ensures low correlation with other styles and better risk adjusted
excess returns (IR). Consider the active factor exposure of our pure momentum index against a simple raw momentum index:
What makes a robust smart beta product? Amenc et al. argue for the importance of robustness in smart beta portfolio construction. ‘Factor Fishing’, ‘Model-Mining’,
‘Non-Robust Weighting Scheme’ and ‘Dependency on Individual Factor Exposures’ are common pitfalls to avoid. Of the
precautionary steps taken in our pure strategy, two stand out as being particularly effective in the creation of resilient factor
portfolios:
Turnover Control: A turnover constraint helps control costs and enhances portfolio stability. For example, momentum
strategies naturally exhibit high turnover. With no turnover constraint, momentum has ~300% average annual turnover,
imposing significant transaction costs on the portfolio.
Active exposures
against MSCI World
Index (MXWO).
“Raw” style indices
refer to the equally
weighted first quintile
of the desirable
style.
0%
10%
20%
30%
40%
50%
60%
Jun
-01
May
-02
Ap
r-0
3
Mar
-04
Feb
-05
Jan
-06
Dec
-06
No
v-0
7
Oct
-08
Sep
-09
Au
g-1
0
Jul-
11
Jun
-12
May
-13
Ap
r-1
4
Mar
-15
Feb
-16
Raw Momentum Monthly Turnover
0%
10%
20%
30%
40%
50%
60%
Jun
-01
May
-02
Ap
r-0
3
Mar
-04
Feb
-05
Jan
-06
Dec
-06
No
v-0
7
Oct
-08
Sep
-09
Au
g-1
0
Jul-
11
Jun
-12
May
-13
Ap
r-1
4
Mar
-15
Feb
-16
Pure Momentum Monthly Turnover
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Raw Momentum
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Pure Momentum
VALUE
VOLATILITY
MOMENTUM
QUALITY
SIZE
LEVERAGE
TRADING.ACTIVITY
GROWTH
EARNINGS.VARIABILITY
For Institutional Investors and Advisor Use Only6 | HSBC Global Asset Management
Source: Factset, Thomson Reuters, December 31, 2016
Source: HSBC Global Asset Management, raw data from Thompson Reuters, Factset, IBES, Worldscope and MSCI as of November 30, 2016.
Raw momentum exhibits significant bias to high volatility stocks.
This unintended exposure could prove problematic for performance
and risk. HSBC’s Pure Momentum strategy is by construction
immunised from such exposure.
Furthermore, style neutrality translates into lower correlations
among factor excess returns:
Raw Factors
Raw average pairwise correlation: 10%
Raw average absolute pairwise correlation: 32%
Pure Factors
Pure average pairwise correlation: 3%
Pure average absolute pairwise correlation: 17.5%
Correlations of factor excess total returns over MSCI World (USD), monthly returns 07-2001 to 12-2016.
Source: HSBC Global Asset Management, raw data from Thompson Reuters, Factset, IBES, Worldscope and MSCI as of December 31, 2016.
Compared to raw factor implementation, only volatility seems
to demonstrate stronger correlation with other factors, but this is still
within acceptable levels. As we will discuss later, correlations follow
time varying patterns so a static calculation reveals little about
their structure.
VALUE MOMENTUM VOLATILITY SIZE
VALUE 100% -1% -36% 19%
MOMENTUM -1% 100% 23% 3%
VOLATILITY -36% 23% 100% 23%
SIZE 19% 3% 23% 100%
VALUE MOMENTUM VOLATILITY SIZE
VALUE 100% -16% -35% 63%
MOMENTUM -16% 100% 44% 18%
VOLATILITY -35% 44% 100% -14%
SIZE 63% 18% -14% 100%
For Institutional Investors and Advisor Use Only7 | HSBC Global Asset Management
The risk of time-varying correlations A popular factor blending approach is to combine value and
momentum. This is primarily because these factors exhibit
low return correlations. However, in extreme circumstances,
these correlations can break down.
A raw factor implementation of value and momentum depends
on their correlation remaining small and stable. This is often
assumed to be constant and negative. The graph below
demonstrates that this is not the case - the correlation varies
over time, depending on the economic environment:
Correlations calculated using daily excess (against MSCI World Index) total returns (2 years rolling) in USD from 30/06/2003 – 31/12/2016. Source: HSBC Global Asset Management, raw data from Thompson Reuters, Factset, IBES, Worldscope and MSCI as of December 31, 2016.
The historic correlation between raw value and momentum has
been unstable, repeatedly oscillating between large positive
and negative values. In contrast, the pure correlations have shown a
greater tendency to remain close to zero. A major exception is
the period around the ‘quant crisis’ in 2007 when factor
payoffs became unstable, and even then the correlations
were close.
The long only nature of the portfolio construction process
also impacts the combination of value and momentum.
Academic studies that refer to a consistent negative
correlation usually point to Fama-French factors based on
long-short portfolios incorporating illiquid securities.
-1.0
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Pure Value - Momentum Correlation Raw Value - Momentum Correlation
Near zero
range
Less stable
For Institutional Investors and Advisor Use Only8 | HSBC Global Asset Management
How can we measure the
purity of a factor strategy?
As we discussed earlier, most smart beta indices have
unintended exposures to non-targeted factors. This usually
stems from the requirements of transparency, simplicity or
investability. There is evidence that simple minimum
variance optimisation, a common smart beta strategy,
results in time-varying factor exposures. Goldberg et al.
suggest that it is important to be aware of these exposures
and highlight the benefits of targeting pure exposures when
building such strategies.
For this reason, we need to assess how well a given strategy
allocates its risk budget to its target factor. Our factor purity
ratio (FPR) is defined as the ratio of tracking error from the
desired factor(s) to the total tracking error. This quantitative
metric facilitates rapid comparison between different
providers’ factor products to assess their relative purity.
Factor Purity Ratio = 𝐴𝑅𝐷
𝑇𝐴𝑅
∑ARD is the sum of active risk contributions of the desired
factors while TAR is the total active risk of the portfolio.
Total
Return/Risk
Systematic Contributions
Style Industry Country Currency
Stock Specific Contributions
The contributions to total active risk can be estimated
using a factor risk model.
In the chart below we show the factor purity ratios for
HSBC’s Pure strategies against those for MSCI’s
developed world style indices. (i.e MSCI Enhanced Value,
MSCI Momentum Tilt, MSCI Volatility Tilt, MSCI Size Tilt).
A strategy with high exposure to a particular factor will not
necessarily have high FPR. For example, it is well known that
a strategy based simply on the ranking of value stocks often
has significant small-cap exposure. If we were to buy the top
quintile of value names, we would anticipate a high exposure
to both value and small cap factors. We would prefer a pure
beta to have a large proportion of active risk driven by the
style factor of interest and minimal active risk contributions
from other factors.
Decomposition of total active risk in a cross-sectional factor risk model
FPR ratios across HSBC Pure Factors and MSCI Factor Indices
MSCI indices used: MSCI World Enhanced Value, MSCI WorldMomentum Tilt, MSCI World Quality Tilt, MSCI Wold VolatilityTilt. Data source for MSCI Weights: MSCI. Other data: HSBCGlobal Asset Management, raw data from Thompson Reuters,Factset, IBES, Worldscope and MSCI. Data as of December31, 2016.
0
10
20
30
40
50
60
70
80
VALUE MOMENTUM VOLATILITY SIZE
HSBC Pure Factors MSCI Indices
For Institutional Investors and Advisor Use Only9 | HSBC Global Asset Management
Pure Factor Strategies
Source: HSBC Global Asset Management, raw data from MSCI, Thompson Reuters, Factset, IBES and Worldscope.
Data as at 31/12/2016.
As can be seen, factor exposure does not directly correlate with factor purity. For example, HSBC’s Pure Momentum strategy
has a low factor exposure at 0.67 but a high factor purity of 73.9%. This makes sense as momentum is a more volatile factor
with a higher active risk contribution. Looking at the desired factor’s exposures alone might be misleading. Factor exposures fail
to take into account the risks contributed by other potentially undesirable factors.
Concentrating on active risk contribution also connects back to the general debate on risk premia factors. There is a degree
of risk in investing in factors and their returns are time-varying. Note that strategies with the same factor exposures may
have different active risks based on the nature of the factor. A strategy that is pure has less contribution from undesired
risks. The key point is that we are only taking a risk on the factors that we choose to invest in.
Comparing this to the MSCI World Enhanced Value Index, we can see how different HSBC’s Pure Value strategy is in terms
of factor purity. The FPR ratio of the MSCI index at the same point in time is 34.9%, compared to HSBC’s of 68.6%. This
implies that HSBC’s index allocates twice as much of its active risk budget to the desired factor than MSCI.
Raw Factor Strategies
Total Active Risk (%) FPR (%) Active Exposure
VALUE 3.18 68.6 1.1
MOMENTUM 2.42 73.9 0.7
VOLATILITY 3.18 52.1 0.6
SIZE 2.61 53.1 1.5
Total Active Risk (%) FPR (%) Active Exposure
VALUE 5.36 34.91 1.1
MOMENTUM 3.98 50.20 0.8
VOLATILITY 5.41 27.23 0.6
SIZE 3.09 12.55 1.1
For Institutional Investors and Advisor Use Only10 | HSBC Global Asset Management
The charts below are decompositions of active risk for the HSBC Pure Value Strategy and MSCI Enhanced Value Index as at
end of December 2016. This is a common risk attribution output from portfolio attribution packages.
Decomposition of total active risk
HSBC Pure Value MSCI Enhanced Value
Source: HSBC Global Asset Management, raw data from MSCI, Thompson Reuters, IBES and Worldscope.
Data as at December 31, 2016.
Decomposition of style active risk
HSBC Pure Value MSCI Enhanced Value
Looking at the decomposition of active risk for the MSCI index above, we see that a significant portion comes from country
active risk. A closer look at the breakdown shows that the majority of this comes from active exposure to Japan. Furthermore
we can see that even though the biggest component of style active risk is indeed value, there are still significant contributions
from volatility and momentum This does not appear to be consistent with a simple ‘index’ strategy – the ex-ante performance
becomes too dependent on a single risk which is not immediately associated with the strategy.
0%
20%
40%
60%
80%
100%
Style Countries Industries Currencies Non-Factor
Positive Contribution
Negative Contribution
0%
10%
20%
30%
40%
50%
60%
70%
80%
Val
ue
Mo
me
ntu
m
Pro
fita
bili
ty
Trad
ing
Act
ivit
y
Earn
ings
Var
iab
ility
Leve
rage
Size
Gro
wth
Vo
lati
lity
Positive Contribution
Negative Contribution
0%
20%
40%
60%
80%
100%
Style Countries Industries Currencies Non-Factor
Positive Contribution
Negative Contribution
0%
10%
20%
30%
40%
50%
60%
70%
80%
Val
ue
Mo
me
ntu
m
Vo
lati
lity
Pro
fita
bili
ty
Leve
rage
Trad
ing
Act
ivit
y
Size
Gro
wth
Earn
ings
Var
iab
ility
Positive Contribution
Negative Contribution
For Institutional Investors and Advisor Use Only11 | HSBC Global Asset Management
The Importance of a Pure StrategyNon-target exposures can have a positive contribution to return. The style-only return attribution of the raw value strategy below
shows that momentum exposure has enhanced performance significantly. However, it is hard to justify its classification as a
value strategy when almost as great a share of return originates from momentum exposure.
2016 saw a fundamental shift in factor performance as defensive styles retreated (e.g. quality and low volatility) and cyclical
factors (e.g. value and size) outperformed. Unintentional systematic exposures are most likely to attract attention when the target
style itself delivers a weak or negative return. In the example of raw low volatility below, the targeted style underperforms the
benchmark by around 1%. Significant uncontrolled negative contributions from value and momentum accentuate this negative
return, producing an overall style performance of -1.8%. The net result is that the client receives a worse return than his target
style ought to deliver.
In both cases of value and volatility, the pure strategies demonstrate the benefit of a factor construction process that purposely
constrains non-target exposures, leaving the advertised style as a dominant driver of return even when its return is weak.
Total active return and total active style return attributions versus MSCI World for HSBC Pure strategies and corresponding MSCI
factor indices. The attributions consider returns for the period 30/11/2015 to 31/12/2016.
-4%
-3%
-2%
-1%
0%
1%
2%
3%
4%
5%
6%
7%
Mar
ket
Styl
e
Co
un
trie
s
Ind
ust
ries
Cu
rren
cies
Res
idu
al
Co
ntr
ibu
tio
n t
o A
ctiv
e R
etu
rn
HSBC Pure Volatility MSCI Volatility Tilt
-4%
-3%
-2%
-1%
0%
1%
2%
3%
4%
5%
6%
7%
Mar
ket
Styl
e
Co
un
trie
s
Ind
ust
ries
Cu
rren
cies
Res
idu
al
Co
ntr
ibu
tio
n t
o A
ctiv
e R
etu
rn
HSBC Pure Value MSCI Enhanced Value
-2%
-1%
0%
1%
2%
3%
4%
5%
VA
LUE
PR
OFI
TAB
ILIT
Y
MO
MEN
TUM
SIZE
VO
LATI
LITY
GR
OW
TH
LEV
ERA
GE
EAR
NIN
GS.
VA
RIA
BIL
ITY
TRA
DIN
G.A
CTI
VIT
Y
Co
ntr
ibu
tio
n t
o A
ctiv
e St
yle
Ret
urn
MSCI Enhanced Value HSBC Pure Value
-2%
-1%
0%
1%
2%
3%
4%
5%
VA
LUE
PR
OFI
TAB
ILIT
Y
MO
MEN
TUM
SIZE
VO
LATI
LITY
GR
OW
TH
LEV
ERA
GE
EAR
NIN
GS.
VA
RIA
BIL
ITY
TRA
DIN
G.A
CTI
VIT
Y
Co
ntr
ibu
tio
n t
o A
ctiv
e St
yle
Ret
urn
MSCI Volatility Tilt HSBC Pure Volatility
For Institutional Investors and Advisor Use Only12 | HSBC Global Asset Management
Analysis: Value + Momentum Investing
We have already illustrated the time varying relationship
between value and momentum. The combination of value
and momentum as an investment strategy has been well
studied in academic literature and is popular among
investment practitioners.
Tracking Error Percentage Decomposition
Portfolio 1: TE 1.58%
Consider the following two portfolios:
► Portfolio 1: 50% Pure Value strategy +
50% Pure Momentum strategy
► Portfolio 2: 50% raw value strategy +
50% raw momentum strategy
Portfolio 2: TE 4.88%
Percentage contribution (PCR) to tracking error (TE) for the cross-section of portfolio/benchmark weights on 31/12/2016.
Source: HSBC Global Asset Management, raw data from MSCI, Thompson Reuters, Factset, IBES and Worldscope as of December 31, 2016.
If combined factors are not purified beforehand, Portfolio 2 shows that the volatility factor contributes significantly to tracking error. This
is to be expected: as discussed before, raw momentum exhibits high exposure to volatility. In contrast, Portfolio 1 appears to have a
more balanced risk profile and 59.3% of the tracking error for Portfolio 1 is attributable to the targeted styles (value-momentum).
-15
-5
5
15
25
35
45
55
Per
cen
tage
Co
ntr
ibu
tio
n t
o A
ctiv
e R
isk
(%)
0
20
40
60
80
Market Style Countries Industries Currencies NonFactor
Per
cen
tage
Co
ntr
ibu
tio
n t
o A
ctiv
e R
isk
(%)
0
20
40
60
80
Market Style Countries Industries Currencies NonFactor
Per
cen
tage
Co
ntr
ibu
tio
n t
o A
ctiv
e R
isk
(%)
-5
5
15
25
35
45
55
Per
cen
tage
Co
ntr
ibu
tio
n t
o A
ctiv
e R
isk
(%)
Unintended Exposure
For Institutional Investors and Advisor Use Only13 | HSBC Global Asset Management
Applications to Portfolio Management
Diversifying a Passive Holding
Factor tilts can be incorporated into portfolio management as an overlay to reduce risk, improve performance and enhance risk
adjusted returns.
In order to demonstrate this we consider three scenarios where different objectives require different factor tilts.
Using the MSCI World Index as a base portfolio, we consider three test cases:
Case 1: Add 10% Pure Size tilt to improve performance
Case 2: Add 20% Pure Low Volatility tilt to reduce risk
Case 3: Combine 10% Pure Size and 20% Pure Low Volatility tilt for better risk adjusted returns
Case 1: MXWO + 10% size tilt Case 2: MXWO + 20% low vol tilt Case 3: MXWO + 10%
size + 20% low vol tilt
Annualised performance numbers from internal backtests covering 07/2001 to 12/2016. Source: Factset, Thomson Reuters, MSCI,
IBES, WorldscopeSimulated data is shown for illustrative purposes only, and should not be relied on as indication for future returns. As with any mathematical
model that calculates results from inputs, results may vary significantly according to the values inputted. Prospective investors should
understand the assumptions and evaluate whether they are appropriate for their purposes. Some relevant events or conditions may not have
been considered in the assumptions. Actual events or conditions may differ materially from assumptions.
HSBC Global Asset Management | 14
14.2%
14.4%
14.6%
14.8%
15.0%
15.2%
15.4%
MXWO MXWO+ 10 %Size Tilt
MXWO+ 20%
Low VolTilt
MXWO+ 10%Size +
20% VolTilt
Annualized Volatility
5.2%
5.4%
5.6%
5.8%
6.0%
6.2%
6.4%
6.6%
6.8%
MXWO MXWO+ 10 %Size Tilt
MXWO+ 20%
Low VolTilt
MXWO+ 10%Size +20%
Vol Tilt
Annualized Return
0.30
0.32
0.34
0.36
0.38
0.40
0.42
0.44
0.46
MXWO MXWO+ 10 %Size Tilt
MXWO+ 20%
Low VolTilt
MXWO+ 10%Size +
20% VolTilt
Sharpe Ratio
-0.2 -0.1 0
VALUE
QUALITY
MOMENTUM
VOLATILITY
SIZE
-0.2 -0.1 0 -0.2 -0.1 0
For Institutional Investors and Advisor Use Only
Completion of an Active Portfolio
Building portfolios with a bottom-up approach can sometimes result in a collection of securities that exhibit unwanted factor
exposures. It is important to manage these biases in order to improve a portfolio’s risk profile.
A value oriented portfolio is used as the base case in this section. Such a portfolio will exhibit a natural bias to the value factor.
After performing factor analysis we can identify any significant unintended negative exposure to momentum. Using the relevant
pure factor strategy we are able to mitigate the unwanted exposure to momentum and keep the factor profile of the portfolio close
to benchmark. The wealth curve below demonstrates how the momentum bias correction provides a slight uplift to performance.
Active factor exposures versus MSCI World Index
Portfolio 1 with unwanted momentum exposure Portfolio 2 with tilt correction – adding a 10%
momentum tilt
The charts above show active average exposures and the cumulative wealth curve of hypothetical strategies for illustrative purposes only.
Data period: 31/07/2007 – 31/12/2016.
Source: Factset, Thomson Reuters, MSCI, IBES, Worldscope, cumulative total returns net of trading cost.
As with any mathematical model that calculates results from inputs, results may vary significantly according to the values inputted.
Prospective investors should understand the assumptions and evaluate whether they are appropriate for their purposes. Some relevant
events or conditions may not have been considered in the assumptions. Actual events or conditions may differ materially from
assumptions.
15 | HSBC Global Asset Management
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
Au
g-0
7
Dec
-07
Ap
r-0
8
Au
g-0
8
Dec
-08
Ap
r-0
9
Au
g-0
9
Dec
-09
Ap
r-1
0
Au
g-1
0
Dec
-10
Ap
r-1
1
Au
g-11
Dec
-11
Ap
r-1
2
Au
g-12
Dec
-12
Ap
r-1
3
Au
g-1
3
Dec
-13
Ap
r-1
4
Au
g-1
4
Dec
-14
Ap
r-1
5
Au
g-1
5
Dec
-15
Ap
r-1
6
Au
g-1
6
Dec
-16
Cu
mu
lati
ve R
etu
rns
Portfolio 1 Portfolio 1 + Momentum Tilt
-0.2 0 0.2 0.4 0.6
VALUE
QUALITY
MOMENTUM
VOLATILITY
SIZE
-0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
For Institutional Investors and Advisor Use Only
Multi-Factor Investing
There is a strong motivation to invest in multiple factors simultaneously: whilst we can expect validated risk premia to perform well
in the long term, each factor may undergo its own short term investment cycle. Given sufficiently low correlations between the
individual factors, investing in a multi-factor composite should provide diversification against short term underperformance whilst
harvesting long term outperformance.
Multifactor construction methodologies vary widely across the industry. Simultaneous investment in multiple single factor indices is
commonplace; this “top-down” approach is attractive in its simplicity and an intuitive concept to communicate. However, it is a sub-
optimal way to harvest factor premia and has been shown to provide inferior risk adjusted returns both in theory and in practice.
Doubling up of holdings in the different indices can also result in high turnover and concentrated stock-specific positions.
In recent years, a large volume of white papers has been published by various asset management houses demonstrating the
virtues of using a “bottom-up” approach. This strategy selects stocks according how well they cater for all targeted factors,
improving trading efficiency and ensuring sufficient multi-factor exposure through a minimum number of holdings.
Our HSBC Multi-Factor capability focuses on five factors that are broadly compatible but capture different elements of the economic
cycle:
► Cyclical (value, size)
► Defensive (quality, low risk)
► Dynamic (group momentum)
The contrasting exposures presented by these factors are evident in the Global Developed long-short cumulative return chart
below:
Source: HSBC Global Asset Management, raw data from MSCI, Thompson Reuters, IBES and Worldscope as of December
31, 2016. Data range from 31/12/2005 to 31/12/2016.
-40%
-20%
0%
20%
40%
60%
80%
100%
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Long-Short Single Factor Cumulative Total Returns
Value Quality Group Momentum Low Risk Size
For Institutional Investors and Advisor Use Only16 | HSBC Global Asset Management
Unbiased
Correlation between factor exposures can be detrimental to bottom-up portfolio construction regardless of whether they are positive
or negative. Secondary exposures from one factor score can act to accentuate or dilute other targeted alphas, producing uneven
aggregate alpha sensitivities in the portfolio.
We can measure quantitatively the extent to which multi-factor scores are biased towards individual factors using the transfer
coefficient (TC). The TC is a metric that calculates the cross-sectional correlation between factor z-score and combined score to
assess how well the relative differences in stocks’ sensitivities to each factor are reflected in the output.
The table below considers the correlation between individual factor’s exposures and the TCs for the combined score (arithmetic
mean factor score). The starkest observation is the weak representation of value exposure in the combined score (top right hand
corner). We can see that value shares negative correlation with every other factor except Group Momentum. The negative value
components within the other factors act to erode its influence, with the result that its transfer coefficient is much lower.
Raw factor exposure Pearson correlations for MSCI World constituents as at December 31, 2016. Source: HSBC Global Asset
Management, raw data from MSCI, Thompson Reuters, IBES and Worldscope.
*Combined score calculated as the equal weighting of individual factor scores.
†Transfer coefficient measured as the correlation of combined score with each individual stock.
Historic value
Current value
Risk adjusted
value
Forward value
Value
Profitability
Leverage
Earnings quality
Quality
Group momentum
Multiple
specification
periods
Momentum
Predicted beta
Uses estimated
volatilities and
correlations
Low Risk
Market
capitalisation
Sales
Total assets
Size
Multi-Factor Purity
The complicated nature of multi-factor portfolio construction means that it is even more important to focus on the delivery of
a product that satisfies the four elements of purity.
Precise
Our core multi-factor methodology focuses on the pursuit of five styles that are widely supported in the academic and professional
community: value, quality, group momentum, low risk and size. The factor definitions are the product of rigorous theoretical and
empirical research. We choose to express each factor as composites of related metrics which capture the various dimensions of
each factor. We can therefore pursue these factors in a precise manner without suffering the limitations of any single metric.
VALUE QUALITYGROUP
MOMENTUMLOW BETA SIZE
Average cross-correlation
Combined Score* Transfer Coefficient†
VALUE 1.00 -0.13 0.13 -0.40 -0.33 -0.18 0.14
QUALITY -0.13 1.00 0.04 0.17 0.14 0.05 0.62
GROUPMOMENTUM
0.13 0.04 1.00 -0.30 0.03 -0.03 0.43
LOW RISK -0.40 0.17 -0.30 1.00 0.02 -0.13 0.27
SIZE -0.33 0.14 0.03 0.02 1.00 -0.04 0.42
Average TC 0.37
For Institutional Investors and Advisor Use Only17 | HSBC Global Asset Management
Efficient
Unbiased representation of individual factors in the combined factor score is key. However, these efforts are futile if the
transformation of final scores into stock weights reintroduces additional uncontrolled factor exposures. Several major multi-factor
providers employ a simplistic weighting scheme that multiplies the combined scores by the stock’s market cap. Market cap indices
are known to exhibit significant value and size biases and will weaken the stock-level link between weight and target factors
exposure. In the left diagram below, we plot stocks by value and quality score and colour the top and second quintile stocks under
a typical ranking scheme used by other providers. In the second chart, we colour the top and second quintiles by weights after the
scores and market cap weights have been combined. The latter chart is a very muddled and bears limited resemblance to the
original score quintiles.
Source: HSBC Global Asset Management. For illustrative purposes only.
Our portfolio construction methodology not only focuses on providing unbiased exposure to each of the alpha factors, but it also
controls exposure to non-alpha factors ensuring efficient target exposure.
Orthogonalized factor exposure Pearson correlations for MSCI World constituents as at December 31, 2016.
Source: HSBC Global Asset Management, raw data from MSCI, Thompson Reuters, IBES and Worldscope
Robust
Robust implementation remains just as important in a multi-factor strategy. Proprietary risk models and linear portfolio construction
tools ensure that non-target systematic and stock specific risks can be constrained without losing the intuition behind the weight
allocated to each stock.
For example, consider a portfolio that is currently underweight Technology and French stocks. Our portfolio construction tool
proposes the following changes on the rebalance date. The changes are intuitive; Stock A and Stock D’s trades are clearly
recommended as a result of their multi-factor scores. Stock B is recommended to neutralize the underweight exposure to
Technology, whilst Stock C is bought in spite of its low score to reduce the negative exposure to France.
Stock Sector Country Multi-Factor Score Trade
Stock A Financials USA +3.0 +0.10%
Stock B Technology UK +0.2 +0.10%
Stock C Energy France -0.1 +0.10%
Stock D Industrials Germany -2.7 -0.10%
0
0.5
1
0 0.2 0.4 0.6 0.8 1
Value - QualityScore Quintiles
Other Stocks Q1 by Score Q2 by Score
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
Value - Quality Weight Quintiles
Other Stocks Q1 by Weight Q2 by Weight
VALUE QUALITYGROUP
MOMENTUMLOW BETA SIZE
Average cross-correlation
Correlation between individual factor exposure
and combined score
VALUE 1.00 0.00 0.00 0.00 0.00 0.00 0.44
QUALITY 0.00 1.00 0.00 0.00 0.00 0.00 0.45
GROUP MOMENTUM
0.00 0.00 1.00 0.00 0.00 0.00 0.42
LOW BETA 0.00 0.00 0.00 1.00 0.00 0.00 0.44
SIZE 0.00 0.00 0.00 0.00 1.00 0.00 0.42
Average TC 0.44
Cross-correlation in exposures can be near-eliminated through sequential residualisation of the individual factors. If we
replace the raw factors in the previous table with new orthogonalised versions, we produce a combined score with a much
more consistent set of TCs (see below). The highest – lowest TC spread is reduced from 0.48 to just 0.03 and the average
TC is boosted from 0.37 to 0.44.
Consistent
TCs across
all five
factors
18 | HSBC Global Asset Management For Institutional Investors and Advisor Use Only
For illustrative purposes only.
Multi-Factor Performance
By blending multiple factors in one equity fulfilment vehicle, we can deliver consistent active performance over the long term as
evidenced by the cumulative return chart below. Moreover, our pure approach, with intelligently crafted factor definitions, efficient
risk budgeting and unbiased exposure ensures that our performance is competitive. A backtest of our strategy initiated in January
2004 outperforms FTSE RAFI on both an actual and risk adjusted basis. Our portfolio delivers a much higher IR than MSCI’s
Minimum Volatility strategy, with a similar Sharpe ratio, while being more diversified to the economic cycle.
CORE PASSIVE VALUE LOW VOL
Developed WorldJan 2004 - Dec 2016
HSBC Global Multi-Factor MSCI World Index FTSE RAFIMSCI Minimum
Volatility
Annual Returns 9.37% 6.73% 7.17% 8.31%
Annual Volatility 13.9% 15.0% 17.0% 11.0%
Sharpe Ratio 0.68 0.45 0.42 0.76
Maximum Drawdown -48% -54% -57% -43%
Annual Relative Returns 2.64% n/a 0.44% 1.58%
Tracking Error 2.39% n/a 3.92% 7.07%
Information Ratio 1.10 n/a 0.11 0.22
0%
5%
10%
15%
20%
25%
30%
35%
40%
-50%
0%
50%
100%
150%
200%
250%
Jan
-04
Jun
-04
No
v-0
4
Ap
r-0
5
Sep
-05
Feb
-06
Jul-
06
Dec
-06
May
-07
Oct
-07
Mar
-08
Au
g-0
8
Jan
-09
Jun
-09
No
v-0
9
Ap
r-1
0
Sep
-10
Feb
-11
Jul-
11
Dec
-11
May
-12
Oct
-12
Mar
-13
Au
g-1
3
Jan
-14
Jun
-14
No
v-1
4
Ap
r-1
5
Sep
-15
Feb
-16
Jul-
16
Dec
-16
Turn
over
Cum
ula
tive R
etu
rns
PortTurnover PortWealth BenchWealth
Source: HSBC Global Asset Management, raw data from MSCI, Thompson Reuters, IBES and Worldscope as of December
31, 2016.
Source: Bloomberg, HSBC Global Asset Management, raw data from MSCI, Thompson Reuters, IBES and Worldscope as of
December 31, 2016.
For Institutional Investors and Advisor Use Only19 | HSBC Global Asset Management
Conclusion
Factor strategies represent a highly accessible and efficient
way of investing in factor premia through passive vehicles.
HSBC’s approach to delivering factor exposure focuses on
the integration of four distinct characteristics: Precision,
Unbiasedness, Robustness and Efficiency. Our methodology
incorporates factor neutralisation in order to improve premia
purity and turnover control for stability and cost reduction.
We compared the risk/return profile of HSBC’s suggested
methodology with the conventional (unconstrained) raw
factor implementations. Not controlling for exposures to
unwanted styles leads to the ‘contamination’ of a signal’s
purity and unintended risk exposure. Finally, we
demonstrated some practical applications of using factors in
portfolio management.
HSBC’s approach achieves competitive performance
characteristics avoiding, by design, unintended exposures.
For Institutional Investors and Advisor Use Only20 | HSBC Global Asset Management
For more information: http://www.global.assetmanagement.hsbc.com/canada
References Amenc, N., Goltz, F., Lodh, A. and Sivasubramanian, S.
(October 2014): Robustness of Smart Beta Strategies, ERI
Scientific Beta Publication.
Ang, A., 2014, Asset Management: A Systematic Approach
to Factor Investing, Oxford University Press.
Fama, E.F. and French, K.R. (2013), A Four-Factor Model for
the Size, Value and Profitability Patterns in Stock Returns,
working paper.
Frazzini, A. and Pedersen, L.H. (2013) – Betting Against
Beta, Journal of Financial Economics, 111, 1-25.
Goldberg, Leshem and Geddes (2013) – Restoring Value
to Minimum Variance, forthcoming in Journal of Investment
Management.
Harvey, Campbell R., Yan Liu, and Heqing Zhu (2013): ...
and the Cross-Section of Expected Returns, unpublished
working paper, Duke University.
Hunstad and Dekhayser (2014) – Evaluating the Efficiency of
‘Smart Beta’ Indexes, working paper.
Meucci, Santangelo and Deguest (2013) – Risk Budgeting
and Diversification Based on Optimized Uncorrelated
Factors, working paper.
John Cochrane of the University of Chicago coined the
term ‘zoo of factors’ in his 2011 presidential address to the
American Finance Association.
http://faculty.chicagobooth.edu/john.cochrane/research/
papers/discount_rates_jf.pdf
HSBC Global Asset Management | 21 For Institutional Investors and Advisor Use Only
Appendix Our factor strategies are constructed from the active universe for the relevant market cap weighted benchmark. They are
rebalanced monthly with an 8% turnover allowance (~100% per year).
Factor Construction
Each single factor strategy is constructed from several individual factors, which we describe in detail below.
In order to combine these components into one factor we first need to normalise them by subtracting the global mean and
dividing by the global standard deviation.
Xi - CapWeightedMean(X)
Zi = Standard Deviation(X)
This procedure ensures that all the individual components are in the same scale and their combination results in the formation
of meaningful factors.
In addition, extreme normalised values that are outside the range [-3, 3] are set to -3 / 3.
The individual components of each factor are combined dynamically (ie the weights are not static) through a specialised algorithm
► At every point in time (cross-section) we calculate the Spearman rank correlation matrix of components and run a
principal components analysis (PCA)
► We extract the first principal component and normalise to sum to 100%. Unlike the equal weight approach, this
captures more of the information in individual components.
Single Factor Definitions
Source: GLOBAL EQUITY FUNDAMENTAL FACTOR MODEL, Nick Baturin, Sandhya Persad and Ercument Cahan September 2012,
Version 1.1, Bloomberg
22 | HSBC Global Asset Management For Institutional Investors and Advisor Use Only
Portfolio Optimisation
HSBC Global Asset Management | 23 For Institutional Investors and Advisor Use Only
Source: GLOBAL EQUITY FUNDAMENTAL FACTOR MODEL, Nick Baturin, Sandhya Persad and Ercument Cahan September 2012,
Version 1.1, Bloomberg
Important Information:
HSBC Global Asset Management is a group of companies in many countries and territories throughout the world that are
engaged in investment advisory and fund management activities, which are ultimately owned by HSBC Holdings Plc.
HSBC Global Asset Management is the brand name for the asset management business of HSBC Group. HSBC Global
Asset Management (Canada) Limited is a subsidiary of HSBC Bank Canada and provides services in all provinces of
Canada except Prince Edward Island.
This material has been prepared by HSBC Global Asset Management (UK) Ltd and has been approved for distribution in
Canada for informational purposes only and is not a solicitation or an offer to buy or sell any security or instrument or to
participate in any trading or investment strategy. All opinions and assumptions included in this presentation are based
upon current market conditions as of the date of this presentation and are subject to change.
Any portfolio characteristics shown herein, including position sizes and sector allocations among others, are generally
averages and are for illustrative purposes only and do not reflect the investments of an actual portfolio unless otherwise
noted. The investment guidelines of an actual portfolio may permit or restrict investments that are materially different in
size, nature and risk from those shown. The material contained in this document is for general for informational purposes
only, and is not intended to provide and should not be relied on for accounting, legal or tax advice. You should consult your
tax or legal advisor regarding such matters.
The investment processes, research processes or risk processes shown herein are for informational purposes to
demonstrate an overview of the process. Such processes may differ by product, client mandate or market conditions.
We accept no responsibility for the accuracy and/or completeness of any third party information obtained from sources we
believe to be reliable but which have not been independently verified.
Any forecast, projection or target contained in this presentation is for information purposes only and is not guaranteed in
any way. HSBC accepts no liability for any failure to meet such forecasts, projections or targets.
Neither MSCI nor any other party involved in or related to compiling, computing or creating the MSCI data makes any
express or implied warranties or representations with respect to such data (or the results to be obtained by the use
thereof), and all such parties hereby expressly disclaim all warranties of originality, accuracy, completeness,
merchantability or fitness for a particular purpose with respect to any of such data. Without limiting any of the foregoing, in
no event shall MSCI, any of its affiliates or any third party involved in or related to compiling, computing or creating the
data have any liability for any direct, indirect, special, punitive, consequential or any other damages (including lost profits)
even if notified of the possibility of such damages. No further distribution or dissemination of the MSCI data is permitted
without MSCI's express written consent.
Past performance is not indicative of future performance.
No part of this publication may be reproduced, stored in a retrieval system or transmitted, on any form or by any means,
electronic, mechanical, photocopying, recording or otherwise, without the prior written permission of HSBC Global Asset
Management Limited.
© Copyright 2017. HSBC Global Asset Management (Canada) Limited. All rights reserved.
Expiry: December 31, 2017
DK1700181A
For Institutional Investors and Advisor Use Only24 | HSBC Global Asset Management