Factor Investing

Introduction & Research

The document is intended only for Professional Clients and Financial Advisers in Continental Europe (as defined in the
Important Information); for Qualified Investors in Switzerland; for Professional Clients in Dubai, Ireland, the Isle of Man,
Jersey and Guernsey, and the UK. It is not intended for and should not be distributed to, or relied upon, by the public or
retail investors. Please do not redistribute.
In recent years, interest in factor-based investing has
increased meaningfully as investors seek precise and
systematic solutions to achieve their investment objectives.

Invesco has been a leader in factor investing for more than

30 years and currently manages more than USD160 billion in
assets within factor-based strategies for clients around the
world at end of June 2017. These strategies aim to deliver
client outcomes that go beyond the limitations of traditional
benchmark-centric approaches.

This booklet consists of a series of papers written by several

Invesco experts on factor investing approach.
Contents
4 Factor investing: an introduction
Jay Raol, Jason Stoneberg and Andrew Waisburd

9 Factor investing: building balanced factor portfolios

Edward Leung and Andrew Waisburd

14 Factor investing: complementing portfolios with customized

factor solutions
Michael Abata, Georg Elsaesser, Brad Smith and Jason Stoneberg

18 How macro factors can aid asset allocation

Jay Raol, Ph.D.

25 Investing in a multi-asset multi-factor world

Alexandar Cherkezov, Dr. Harald Lohre, Sergey Protchenko and
Jay Raol, Ph.D.

33 “Multi-asset-multi-factor-strategies employ the full set of

tools in our toolkit.”
Roundtable with Alexandar Cherkezov, Dr. Harald Lohre, Stephen
Quance and Jay Raol, Ph.D.

36 Why should investors consider credit factors in fixed income?

Jay Raol, Senior Macro Analyst Fixed Income & Shawn Pope,
Macro Quantitative Analyst Invesco Fixed Income

3
Factor investing: an introduction
by Jay Raol, Jason Stoneberg and Andrew Waisburd

In brief
The performance of individual securities and
asset classes can largely be explained by
their systematic exposure to quantifiable
investment themes. These “factors” include
value, momentum, quality, and size, among
others. The rapidly growing space of factor
investing is based on the approach of
explicitly allocating directly into a portfolio
of these factors using tradable securities
merely as instruments to achieve broad
and diversified exposure. Depending on
investors’ preferences, they might choose
an active or a passive approach, based on
a single-factor or a multi-factor strategy.
Regardless of the implementation chosen,
holding a diversified, well-balanced portfolio
of factors aims to reduce risk and deliver a
smoother return stream.

Risk & Reward, Q4/2016 4

In recent years, factor-based investing has become
ever more popular among investors seeking
precise and systematic solutions. We give an
overview of the concept, describe its history and
highlight popular factors. Finally, we contrast
active with passive, as well as multi-factor with
single-factor concepts.
A substantial body of academic research, coupled
with advancements in data collection and
processing, has expanded investors’ understanding
of the key factors historically affecting risk and
return. Moreover, the introduction of specific
indexes and ETFs, alongside the offerings of active
quantitative managers, now provide an array of
options for investors to implement these factors
in their portfolios. Estimates put the current total
assets under management in “factor” or “smart
beta” strategies, as they are often referred to, at
USD 1.2 trillion.1
Style factors, macro factors…
At the most fundamental level, “factors” can be
described as quantifiable characteristics of assets.
They include: value, size, momentum, volatility and
quality. Some researchers distinguish between risk
and return factors, with return factors explaining
long-term returns and risk factors explaining their
variability. However, we prefer to view risk-return on
a continuum. Consequently, we refer to both risk
and return factors as “style factors” (figure 1).
While style factors are often discussed in the context
of equity portfolios, they can also be used for other
asset classes. For example, value corresponds to
Figure 1
What is a factor? Macro and style factors
Macro factors
Economic Inflation Political
Style factors
Value Low size Momentum
Source: Invesco. For illustrative purposes only.
Risk & Reward, Q4/2016 5
assets trading attractively relative to intrinsic value
as measured by price to book in equities and term
premium (the current yield versus future expected
yield) in bonds.
In addition, there are “macro factors”, such as growth
and inflation. These are especially well-suited for
spanning asset classes, as different asset classes
have different macro factor sensitivities. For example,
investors often associate lower average returns with
bonds as compared to equities. But that is not
necessarily true. A factor investor would say that
bonds have a lower exposure to the growth factor,
which often drives equity returns.
…and factor investing
Essentially, factor investing means allocating a
portfolio to style and macro factors in an effort
to achieve particular investment objectives. Similar
to more traditional investment processes, factor
investing involves taking positions in individual
assets. But, unlike more traditional approaches
focusing on security selection, a factor approach
makes use of tradable securities, such as stocks
and bonds, to achieve broad and diversified exposure
to specific investment themes.
A look back in time
The origins of factor investing are largely academic
in nature. The first milestone is the seminal work of
Sharpe (1964). His “market model” separates the
market factor beta from the stock-specific alpha.
The “three factor model” of Fama and French (1993)
extends this approach to include both size and value
(defined as price-to-book ratio) as additional
Currencies Credit Real rates Liquidity
$€£¥
Low volatility Dividend yield Quality
Figure 2
The origins of factor investing

Separation of beta Low volatility Size Invesco Quantative Size and value 2003
and alpha 1972 1981 Strategies 1993 First smart beta ETF
1964 Haugen and Heinz showed Banz finds that small cap 1983 Fama and French launched
Building on Markowitz's that low volatility stocks stocks outperformed large Launch of first developed 3-factor
mean variance analysis realized extra risk–adjusted cap stocks quantitative strategies model by adding size 2008
Sharpe, Lintner and returns and value to the market
factor Norges Bank Investment
Mossin developed the Management review
Capital Asset Pricing approach to active
Model (CAPM) Value Momentum management (Ang,
1976 1981 1993 Goetzman & Schaefer)
Launch of the first index Basu finds that low PE Jagadeesh and Titman
mutual fund stocks generated higher find that buying past
returns relative to high winners and selling
PE stocks past losers was highly
profitable

1993
First exchange traded
fund (ETF) launched

1997
Carhart developed
4-factor model

1960 1970 1980 1990 2000

Source: Invesco. For illustrative purposes only.

explanatory variables. A few years later, Carhart Factors vs. fundamentals

(1997) introduced the momentum factor, to form The factor-based approach is often set in contrast to
what is now known as “the four factor model”. It a “fundamental” approach, which implies that factor
explains stock returns with the four factors: market, investing is not fundamentally based – something of
size, value and momentum. In practice, quantitative a misconception. Many, if not most, widely used
portfolio managers have used variants of the four- factors, such as: value, momentum or quality, rely
factor model to manage money for quite some time.
In fact, Invesco is one of the pioneers in this space
and has been investing via factor models since
1983. Figure 3
Why would we expect to earn a factor premium?
More recently, in a study for the Norwegian
Government Pension Fund (GPFG), Ang, Goetzmann Risk premiums Behavioural rationales Market structure
and Schaefer (2008) give an idea of different For bearing additional risk over Markets are inefficient due to Markets may be inefficient
factors used in the multi-asset space. Ang, Goetzmann the broad equity market e.g. behavioural biases of because of restrictions and
an undesirable return pattern participants limitations
and Schaefer also point out that over two-thirds of
Norway’s sovereign wealth fund’s performance was
driven by exposure to systematic factors. This suggests
a paradigm shift away from allocation by asset class,
toward allocation by factor.

While it may be some time before investors broadly

and holistically reframe their investment problem to
focus on factors similar to GPFG, the study has
created renewed interest in the idea that portfolio
return and risk can be largely explained by factor
exposures, whether intended or not. And, given that
Action Loss
investors are exposed to factors, they would benefit Anchoring
bias aversion
from a better understanding of them. Once this
achieved, they may consider actively investing in
factors for two reasons: to generate factor return Source: Invesco. For illustrative purposes only.
(figure 3) and to manage factor risk.

Risk & Reward, Q4/2016 6

on the same fundamental investment themes used well-established in academic literature. Investors in
by more traditional asset managers. Some would passive strategies expect consistency of the
argue that these drivers take advantage of behavioural methodology and complete transparency with
anomalies, creating exploitable market inefficiencies. respect to both index construction and holdings.
Others would counter that factor returns reflect
premia for additional risk over the broad market. Alternatively, active factor-based strategies are
In either case, similar to most traditional asset frequently offered by quantitative arms of asset
management concepts, factor models require a management firms. They rely heavily on teams of
strong investment rationale. So, the real difference researchers and portfolio managers, who leverage
between a factor-based approach and a more proprietary factors and sophisticated portfolio
traditional one is not the nature of the investment construction techniques evolving over time. Clients
themes, but the way they are implemented in a are often large, institutional investors who are willing
portfolio. Whereas traditional or “fundamental” to forego complete transparency in return for a more
managers typically rely on bottom-up selection and customized and sophisticated approach to delivering
careful investigation into the current state of each factor exposures.
company, factor investing delivers transparent,
structured and disciplined operationalization of Overall, passive factor investing allows investors to
traditional investment themes. access well-established factors in an efficient,
relatively low cost and transparent manner, while
Similar to a more traditional/fundamental approach, active factor investing offers exposure to dynamic,
the factor-based approach is also highly research proprietary factors, which are carefully combined to
intensive, but the research tends to be longer term, seek alpha and diversify risk. Both methods represent
focused on identifying the underlying return and risk  different – but effective – ways to implement factors
drivers and heavily reliant on statistical evidence. By within a portfolio. The specific implementation will
focusing on factor exposures, rather than individual depend on the client.
names, portfolios are built to systematically harvest
factor premia. Single-factor vs. multi-factor?
Both active and passive approaches can be
A few examples implemented by combining factors into multi-factor
To better understand how this works, let us consider solutions, or by targeting single factors such as low
specific examples of style and macro factors. volatility or value. However, passive approaches have

“Momentum” is a common style factor: it refers to

the phenomenon that assets with positive (negative)
returns in the past, tend to also have positive
(negative) returns in the future. There is a very
good reason for this: Peter Lynch, the legendary
“fundamental” manager of Fidelity’s Magellan Fund,
has said that investors tend to “trim the flowers and
water the weeds”. In other words, they sell winners
too early and hold onto losers for too long. Such
behaviour leads to incomplete price discovery and
– ultimately – price trends, i.e. “momentum”.

“Growth”, on the other hand, is an important macro

factor: since World War II, in periods with increasing
GDP growth, stocks have had a significantly higher
Sharpe ratio than bonds. Therefore, a portfolio
with a large equity allocation relative to bonds is
significantly exposed to growth factor risk, and has
typically been rewarded with higher returns in these
periods.

Other macro factors include inflation, currency

appreciation/depreciation or even policy rate changes,
all of which help to explain how multi-asset portfolios
performed in various economic environments.

Active vs. passive?

An important question is whether implementation of
factor investing should be passive or active. Exposure
to factors can be achieved either way, and the
implementation is largely a function of investors’
objectives, preferences and budgets. Passive factor-
based strategies are often labelled “smart beta”,
and are commonly implemented via exchange-traded
funds (ETFs). These strategies seek an enhanced
risk-return profile by using factor-based indexes,
instead of a traditional cap-weighted one. These
“smart” indexes tend to use factor definitions that
are standard, stable and have been widely used and

Risk & Reward, Q4/2016 7

tended to utilize a single-factor framework while About the authors
active approaches are more commonly implemented
as multi-factor. Single-factor exposures provide Jay Raol, Ph.D.
targeted building blocks for investors to create Senior Macro Analyst, Invesco Fixed
custom factor blends, while multi-factor solutions Income
attempt to balance factor exposures holistically on Jay Raol is a member of the macro
the investor’s behalf. research team for Invesco Fixed
Income. In his role, he works on macro
In principle, for investors with absolutely no exposures economic based models for asset
to style and macro factors, the solution could be to allocation and fixed income investing.
employ a multi-factor model with sophisticated
portfolio construction techniques and careful risk Jason Stoneberg
control. In practice, many clients will already have Director, Research & Product
some exposure to factors, but unless their factor Development, Invesco PowerShares
investments have been intentional, their portfolios In his role, Jason oversees the
are unlikely to have an ideal factor balance. research and development of
exchange-traded funds. He also
Simply adding a multi-factor-based strategy to an oversees the firm’s research efforts
existing portfolio with inappropriate factor tilts may related to Invesco PowerShares and
not be the perfect solution. A process of “factor the ETF industry.
completion” could be more appropriate. Various
techniques are available to complete one’s factor
Andrew Waisburd, Ph.D.
exposures, from a simple combination of multiple
Global Head of Portfolio Management,
single-factor portfolios, to a sophisticated, holistic
Invesco Quantitative Strategies
implementation providing optimal factor blends.
He was promoted to Director of
Such portfolio rebalancing can be performed within
Research in 2011 and to his current
a single asset class or across multiple asset classes.
position of Global Head of Portfolio
In the extreme, an investor will change his/her view
Management in 2016. He also serves
of the world – from one in which investments are
as a member of Invesco Quantitative
allocated across assets and securities to one in which
Strategie´s management team
he/she allocates into the underlying factors that
responsible for strategic planning and
drive security and asset class returns.
direction.
Conclusion
Factors are investments, and, as with other
investments, holding a diversified, well-balanced
portfolio of factors can reduce risk and deliver a
smoother, positive return stream. After understanding
the philosophy of balanced factor investing, the next
question is how to practically implement a factor
strategy. In forthcoming articles of Risk & Reward
we plan to investigate these approaches in greater
detail.

Note
1 “Enhanced index” and “smart beta” strategies as defined by
eVestment, Preqin, The Economist Intelligence, as at April
2016.

Risk & Reward, Q4/2016 8

Factor investing: building balanced
factor portfolios
by Edward Leung and Andrew Waisburd
In brief
We examine a sample of US equities over
the past 24 years and find that a simple,
balanced factor portfolio of value and
momentum outperforms a cap-weighted
benchmark. This is true whether the
balanced factor portfolio is formed from
a combination of two individual factor
portfolios or implemented via a single
portfolio built from a multi-factor model
(the “multi-factor portfolio”). We also find
that, relative to a combination of single
factor portfolios, the multi-factor portfolio
more effectively accounts for the
relationship between factors. As a result,
it tends to have higher exposure to the
intended factors so that, ultimately, the
multi-factor portfolio outperforms the
combination. Importantly, we find that there
are ways to construct single factor portfolios
such that their combination delivers both
factor exposure and performance that is
similar in magnitude to that of the multi-
factor approach.
Risk & Reward, #1/2017 9
As factors become an increasingly more important
part of the way in which we invest, there are many
critical questions to be considered. Based on the
premise that factors are investments with risk
and return properties, the relevant decision is how
to allocate between factors to appropriately trade-
off risk and return. Specifically, what are the
factors in which to invest? What is the appropriate
balance between these factors? And, what is the
best method for implementing the balanced
approach?
In recent years, we have observed growing demand
for factor-based approaches to investing. According
to the Invesco Global Factor Investing Study (2016)
conducted by NMG consulting, 70% of the investors
surveyed currently use factors in portfolio
construction, and 71% of respondents expect to
increase factor product allocations in the future.
Several drivers have likely led to this growth. Among
these is an increased awareness of factor investing
thanks to a well-established and growing body of
research on factors such as value, size, momentum,
volatility and quality. Another contributor to this
growth is better access to factor-based products via
quantitative asset managers and exchange traded
funds (ETFs) focused on smart beta. Perhaps most
importantly, the growth in factor investing stems
from an increasing appreciation by members of the
investment community that a meaningful proportion
of their portfolios’ performance is explained by
exposure to factors as systematic drivers of risk and
return.
Investment managers have responded to the growth
in demand. For decades, quantitative asset managers Factor and model performance
have been creating multi-factor portfolios that take
into account the relationship between various
factors, from both a risk and return perspective.
In recent years, we have also seen the introduction
of single-factor “smart beta” portfolios (often in the
form of ETFs) offering exposure to individual factors.
These single-factor portfolios can be combined to
produce a balanced factor allocation as well. While
both multi-factor portfolios and combinations of
single-factor portfolios generate balanced exposure
to multiple factors, the portfolios can differ in
fundamental ways.
In this article, we consider both approaches. First,
we combine two individual portfolios where each
is formed from a single factor only. We compare
this combination to a single portfolio built from
a multi-factor model, “the multi-factor portfolio”.
Intuitively, if all information on the factors is applied coefficients of momentum and value are 3.1%
simultaneously, as is the case with the multi-factor
portfolio, the decision-making process tends to be
more informed and outcomes are improved. As a
result, the multi-factor portfolio outperforms the
combination of single factors. This finding is
consistent with Bender and Wang (2016), Fitzgibbons, These are stocks that have been increasing in price
Friedman, Pomorski and Serban (2016), and Clarke, over the past year and are still trading at attractive
de Silva and Thorley (2016), all of whom find that
“the whole is worth more than the sum of the
parts.” But, unlike these studies, this article also
shows that there are ways to construct single factor
portfolios in such a way that their combination
delivers both factor exposure and performance that
is similar in magnitude to that of the multi-factor
portfolio.
“The whole is worth more
than the sum of the parts.”
Factors and models
In the first section, we consider two commonly used
factors: value and momentum. Both factors are used
by practitioners and have been shown in academic
literature to have forecasting power in the cross-
section. High value stocks tend to outperform low
value (or expensive) stocks, and high momentum
stocks, or stocks with high positive returns in the
past, tend to outperform low momentum stocks.
A large body of literature follows the early work on
value by Basu (1977) and momentum by Jegadeesh
and Titman (1993).
We consider simple, easy to understand, and
commonly used definitions of value and momentum.
Momentum is computed as the cumulative return
over the past 12 months excluding the most recent
month. Value is measured using earnings yield, or
earnings over price, where earnings is the average
over the past four quarters. Each month, both
factors are computed and standardized over a large/
midcap universe of approximately 1,300 US equities.
Finally, we define a model as an equally weighted
combination of momentum and value. The factors
and the model are estimated over the 25-year period
beginning April 1991 and ending October 2016.
Risk & Reward, #1/2017 10
Table 1
1-month information coefficients Value Momentum Model
Average 0.021 0.031 0.040
Standard error 0.008 0.011 0.009
t-statistic 2.72 2.87 4.65
Source: Invesco calculations.
Table 1 reports the performance of each factor
and the overall model. Factor performance is
measured using one-month information coefficients,
or the correlation of factor readings with realized
returns over the subsequent month. Both momentum
and value factors are significantly positively
correlated with subsequent returns. The information
(t-statistic = 2.9) and 2.1% (t-statistic = 2.7),
respectively.
Stocks with strong model readings tend to possess
the properties of both value and momentum stocks.
multiples. As might be expected, this is not especially
common. All other things being equal, stocks that
tend to increase in price are not necessarily those
that look most attractive from a valuation perspective.
In fact, the cross-sectional correlation between
momentum and value is consistently negative, at
-13% on average over the sample period. Combining
drivers of return that are uncorrelated with one
another to create balanced factor exposures is a key
to successful factor investing. In table 1, we see that
the combination of momentum and value outperforms
each factor individually, with an information
coefficient of 4% (t-statistic = 4.7). As we will now
discuss, not all methods of building balanced factor
portfolios take advantage of these correlation
structures as effectively as others.
Multi-factor portfolios versus combinations of
single-factor portfolios
In this second section, we explore different ways of
achieving balanced exposure to multiple factors in
tradable portfolios. We consider two common
approaches to building balanced factor portfolios.
First, we build a single portfolio using the multi-
factor model described in the previous section. This
approach is a common one that has been used by
quantitative asset managers for decades. The multi-
factor forecast simultaneously considers the
information contained in both the momentum and
value factors, and the portfolio formed from this
joint forecast incorporates the inverse relationship
between the two factors.
We also consider a portfolio of momentum and
value formed from two single-factor portfolios,
a momentum portfolio and a value portfolio. This
is a very practical approach. There are a large
and increasing number of single-factor portfolios
available in the marketplace that can be used as
potential building blocks for this type of exercise.
These smart beta portfolios offer the consumer a
wide array of choices regarding provider, factor
definition and portfolio construction methodology.
They provide the ability to combine factors at Figure 1
customized weightings, and all of this flexibility often Active exposure to momentum and value in single-factor portfolios
comes at highly competitive prices. One potential
drawback of using combinations of independently Panel A: Single-factor momentum portfolio
formed, single-factor portfolios is that the approach Momentum Value
may be less effective at capturing relationships Active exposure
between factors. 1.5

We simulate equal investments in separate momentum

and value portfolios, and compare the properties 1.0
and performance of a combined portfolio to that
of a multi-factor momentum-value portfolio of equal
0.5
total value over the same period. Each of the
portfolios has been built using a mean-variance
optimization framework, in which return forecasts 0.0
are either one of the single-factor forecasts or the
multi-factor forecast described earlier. Risk is
estimated using a fundamental risk model that -0.5
includes value and momentum factors, as previously
defined. We constrain active exposure to all style -1.0
factors other than value or momentum. Since 1/93 1/96 1/99 1/02 1/05 1/08 1/11 1/14
we use a large/midcap U.S. investment universe,
we optimize against the Russell 1000 Index. The Panel B: Single-factor value portfolio
maximum active weight in any individual security
Momentum Value
is constrained to be within two percent of the
Active exposure
benchmark, and GICS industries and sectors are
limited to be within three percent of the benchmark. 1.5
The active risk level is calibrated to be approximately Momentum Value
three percent for both the combination of single- 1.0
factor portfolios and the multi-factor portfolio.1

Importantly, in this article we do not address the 0.5

issue of which portfolio construction methodology
to use when building factor portfolios. Instead, we 0.0
attempt to compare combinations of single factor
portfolios with a multi-factor portfolio where the
portfolio construction methodology is held constant. -0.5
For convenience, we choose mean variance
optimization against a benchmark.
-1.0
1/93 1/96 1/99 1/02 1/05 1/08 1/11 1/14

Based on data from April 1991 onwards; however, due to the calculation methodology,

The momentum portfolio exposure data is only available from January 1993 onwards.
Source: Invesco calculations.

has consistently negative
exposure to value, and
the value portfolio has
consistently negative
exposure to momentum.
Figure 1 shows the active exposure to momentum
and value through time in the single-factor portfolios.
Panel A describes the momentum portfolio, and
panel B describes the value portfolio. We observe
that the momentum portfolio and the value portfolio
each have high positive exposure to their respective
factors. However, we also note that the momentum
portfolio has consistently negative exposure to
value, and the value portfolio has consistently
negative exposure to momentum. This second
finding is critical. It follows from the fact that
momentum and value are negatively correlated
with one another, and the fact that each of the
single-factor portfolios was constructed with the
intention of capitalizing exclusively on the factor
of relevance.
Figure 2 shows the active exposure to momentum
Momentum Value
and value through time in the multi-factor portfolio
(panel A) and in the combined single-factor portfolios
(panel B). In these graphs, active exposure to both
momentum and value are positive through time. This
follows from the fact that in both cases we are building
portfolios that allocate to assets with high exposure
to each of the factors individually. Importantly, we
observe that the level of active exposure for both
momentum and value is substantially higher for the
multi-factor portfolio than for the combination of
single-factor portfolios. This is because the multi-
factor portfolio is not building a portfolio of assets
merely having high individual exposure to momentum
and value – the assets also have high exposure to
momentum and value jointly. The combination of
single-factor portfolios, on the other hand, has its
positive momentum exposure offset by the negative
exposure in the value portfolio, and has its positive
value exposure reduced by the negative exposure in
the momentum portfolio.
The increased exposure to factors with the ability
to forecast return translates directly into portfolio
performance. The first two columns of table 2 report

Risk & Reward, #1/2017 11

the active performance of the multi-factor portfolio Figure 2
and the combination of two single-factor portfolios, Active exposure to momentum and value in balanced factor portfolios
respectively. By construction, both portfolios have
approximately 280bp of active risk, but the multi- Panel A: Multi-factor portfolio
factor portfolio offers an annual return of 222bp. Momentum Value
This is almost 70bp more per year than the 154bp Active exposure
of active return delivered by the portfolio formed 1.0
from a combination of single factors. Ultimately, the
information ratios for both portfolios are positive and
significant, but the multi-factor portfolio is notably 0.8
stronger (0.78 vs. 0.55).
0.6

Multi-factor portfolios 0.4

outperform combinations 0.2

of single factors.
0.0
1/93 1/96 1/99 1/02 1/05 1/08 1/11 1/14

Portfolio construction matters Panel B: Combination of single-factor portfolio

Consistent with Bender and Wang (2016), Fitzgibbons, Momentum Value
Friedman, Pomorski and Serban (2016), and Clarke,
Active exposure
de Silva and Thorley (2016), we find that multi-factor
portfolios outperform combinations of single factors. 1.0
This occurs because the multi-factor portfolios are Momentum Value
built to more effectively account for the correlation 0.8
between factors and, as a result, relevant exposures in
the multi-factor portfolio are higher than those in the
equivalent combination of single factors. Given the 0.6
ubiquity of single-factor options for delivering factor
exposures, it might be worthwhile to construct 0.4
combinations of single-factor portfolios that offer
benefits similar to the multi-factor return forecast.
Let us explore this possibility next. 0.2

Recall that our single-factor momentum portfolio

0.0
has negative value exposure, and our single-factor 1/93 1/96 1/99 1/02 1/05 1/08 1/11 1/14
value portfolio has negative momentum exposure –
both of which lead to diminished exposures in the Based on data from April 1991 onwards; however, due to the calculation methodology,
exposure data is only available from January 1993 onwards.
combined portfolio. As a practical matter, if we Source: Invesco calculations.
were able to create single-factor portfolios based
on factors that were negatively correlated with
one another but were not negatively exposed to
the complementary factor, we might be able to portfolio has zero value exposure, and the value
Momentum Value
mitigate the issue and generate a combined portfolio has zero momentum exposure. In this
portfolio of single factors that performs similarly way, we avoid creating single-factor portfolios
to the multi-factor portfolio. that have deleterious effects on the contributions
of other factors when held in combination.
We build the single-factor portfolios identical Column 3 of table 2 shows the performance of
to those in the second section, except for the the combination of these “enhanced” single-factor
additional requirement that the momentum portfolios.

Table 2
Portfolio performance
Multi-factor portfolio Combination of Enhanced combination
single factors of single factors
Active return 2.22% 1.54% 2.12%
Active risk 2.83% 2.80% 2.88%
Information ratio 0.78 0.55 0.73
t-statistic 3.81 2.68 3.57
Source: Invesco calculations in USD.

Risk & Reward, #1/2017 12

The average value exposure (not tabulated) in the References
combination of single-factor portfolios increases by • Basu, S., “Investment Performance of Common
32%, from 0.32 to 0.42, and the average momentum Stocks in Relation to their Price-Earnings Ratios:
exposure increases by 29%, from 0.43 to 0.55. A Test of the Efficient Market Hypothesis”,
These increases in exposure to factors with positive Journal of Finance, volume 12, number 3, 1977.
returns lead to increases in portfolio return at similar
levels of risk. The combination of single-factor • Bender, J. and Wang, T., “Can the Whole Be More
portfolios now has an information ratio of 0.73, an Than the Sum of the Parts? Bottom-Up versus
increase of 33% over the 0.55 information ratio for Top-Down Multifactor Portfolio Construction”,
the previous combination of single factors. This risk- Journal of Portfolio Management, volume 42,
adjusted return is hardly distinguishable from the number 5, 2016.
0.78 information ratio associated with the multi-
factor portfolio. • Clarke, R., de Silva, H., and Thorley, S.,
“Fundamentals of Efficient Factor Investing”,
Conclusion Financial Analyst Journal, volume 72, number 6,
We have provided empirical evidence for two well- 2016.
established factors: value and momentum. We
have demonstrated the efficacy of each factor for • Fitzgibbons, S., Friedman, J., Pomorski, L., and
forecasting US equity returns and shown that a Serban, L., “Long-Only Style Investing: Don’t Just
multi-factor model capturing a balanced combination Mix, Integrate,” AQR Paper, June 2016.
of uncorrelated factors has been beneficial. The
main focus of this article is on how to implement • Jegadeesh, N., and Titman, S., “Returns to
the model as a portfolio of balanced factor Buying Winners and Selling Losers: Implications
exposures. We examined implementations via for Market Efficiency”, Journal of Finance, volume
combinations of single-factor portfolios and via one 48, number 1, 1993.
multi-factor portfolio. Regardless of the approach
chosen, we found that a simple, balanced factor • NMG Consulting, “Invesco Global Factor Investing
portfolio of value and momentum outperforms a Study”, 2016.
cap-weighted benchmark. Similar to other research,
we also found that single-factor portfolios, when
combined, have lower exposures to the intended
factors, and, as a result, inferior performance
compared to an analogous multi-factor implementation.
However, we also found that, if single-factor portfolios
are built in specific ways, it is possible to combine
them to achieve many of the benefits of the multi-
factor approach. Ultimately, the way in which a
balanced portfolio of factors is constructed should
reflect the preceding points, but it should also take
into account practical concerns, including, but not
limited to, existing factor exposures in a portfolio
and intended factor allocations. Such considerations
would likely lead to use cases for both single factor
and multi-factor portfolios.
About the authors

Edward Leung, Ph.D.

Quantitative Research Analyst,
Invesco Quantitative Strategies
Edward joined Invesco Quantitative Strategies in
November 2007 and is currently responsible for
conducting research on various areas such as asset
selection, factor efficacy, and alpha generation.

Andrew Waisburd, Ph.D.

Global Head of Portfolio Management,
Invesco Quantitative Strategies
Andrew was promoted to Director of Research in
2011 and to his current position as Global Head of
Portfolio Management in 2016. He also serves as a
member of Invesco Quantitative Strategies’
management team responsible for strategic planning
and direction.

Note
1 None of the calculations in this article takes into account trading costs, management
charges and other fees.

Risk & Reward, #1/2017 13

Factor investing: complementing
portfolios with customized factor
solutions
By Michael Abata, Georg Elsaesser, Brad Smith and Jason Stoneberg
In brief
We examine two approaches to
complementing existing portfolios with
customized factor solutions, both of which
are aimed at improving their risk/return
characteristics. After determining the factor
tilts in existing portfolios, the first approach
uses the broad range of available factor-
based ETFs to achieve the desired risk-
adjusted return. The second approach is
actively managed: we develop a highly
customized factor completion portfolio that
reflects a client’s individual risk/return
targets as fully as possible. We find that
both approaches lead to meaningful
improvements, as confirmed by a wide
range of statistics such as expected alpha,
tracking error and information ratio.
Risk & Reward, #2/2017 14
When a portfolio has unwanted factor biases, there
are several ways to deal with this. One possibility
is a factor-based completion portfolio, which we
will look at in this article as part of our series on
factor investing.
It is well established that a meaningful proportion of
a portfolio’s performance is explained by exposure to
factors that drive risk and return. Further research
has shown that certain factors have historically been
more apt at delivering risk-adjusted excess returns.1
This has long been understood by quantitative asset
managers, who build multi-factor portfolios in order
to harvest such premiums. While these portfolios
generally provide well-balanced factor exposure in
isolation, they often fail to account for an investor’s
existing investments.
As with any portfolio, these existing investments also
have factor exposures. For example, we looked at
the factor tilts of more than six hundred actively
managed US large cap funds. On average, relative
to the S&P 500 Index, these funds were underweight
on low volatility and dividend yield, while being
neutral on momentum, quality and value and
overweight on small size. Needless to say, an
individual portfolio or subset of portfolios would
likely see more substantial factor tilts than the
average.
To construct a portfolio with balanced factor exposure,
it is therefore important to understand the factor
tilts implicit in an initial portfolio. Once these tilts
Figure 1
How the completion portfolio works

1: Scanning initial portfolio 2: Choice of completion 3: Balanced portfolio

Source: Invesco. For illustrative purposes only.

have been measured, a completion portfolio can be
constructed. Taken together, the two portfolios
should exhibit the desired factor exposures (figure 1).
Completion portfolios can be implemented in various
ways, ranging from blends of passive factor ETFs to
custom-built actively managed portfolios. Depending
on scale, customization and cost, investors may
choose to have their completion portfolios actively
managed or to implement their own completion
portfolios using ETFs.
Completion portfolios using factor ETFs
ETFs are liquid, transparent and tradable vehicles
that can provide targeted exposure to individual
factors, including: value, momentum, low volatility,
quality, small size and dividend yield. They can
provide a vast array of options to build custom factor
blends, including those which are needed to build
custom completion portfolios in a low-cost, efficient
manner. For instance, if an investor needed 70%
momentum and 30% quality to balance an existing
portfolio’s factor exposure, this could easily be
achieved by blending together two ETFs.
To study the efficacy of this approach, we built
completion portfolios for 642 actively managed
US large cap funds, which have at least ten years
of history and are categorized as US Large Cap
Growth, Large Cap Blend or Large Cap Value by
Morningstar.2 Our long-only completion portfolios
were formed from the S&P 500 low volatility,
momentum, enhanced value and quality factor
indices. In all cases, an allocation of 30% was given
to the completion portfolio with 70% remaining in
the active mutual fund.
Each fund’s completion portfolio was created by
solving for the blend of factor indices that had the
lowest correlation of excess returns to the fund,
based on five years of monthly returns from 2007
to 2011. The completion portfolios were permitted
to include up to four of the factor indices, depending
on the appropriate blend. Performance of the initial
fund, plus the completion portfolio, was measured
from 2012 to 2016. According to our results, adding
and tracking error was reduced for 91%. The averages
of all 642 funds also improved: annualized total
returns in USD increased by 50 bps, volatility (i.e.
standard deviation of returns) decreased by 71 bps
and tracking error decreased by 132 bps.
Actively managed completion portfolios
The second approach uses actively managed
completion portfolios. Our aim was to increase
portfolio risk-adjusted return and diversify specific
(idiosyncratic) risk, while mitigating exposure to
known common risk factors and increasing exposure
to desired alpha signals. We show results for two
portfolios, a US portfolio benchmarked to the S&P
500 and a non-US portfolio benchmarked to the
MSCI EAFE index. In each case, the completion
sleeve is allocated 35% of total portfolio capital.
Simulations were run for the time period from
January 2006 to December 2016.3
The objective was to reduce risk to a desired range
relative to a benchmark: in both the US and the non-
US case, we were looking to target tracking error
against the respective benchmark at between 275
and 325 bps.
Figure 2
An ETF-based completion portfolio has led to better metrics for the
majority of funds
% with improvement
98% 96% 98%
93%
91%
81% 80% 81%
76%
60%
45%
Alpha to
Annualized
return

S&P 500

Risk adjusted
return

Sortino ratio

return

Volatility

drawdown

Tracking error
Worst 12 month

Maximum

Information
ratio

Up capture

Down capture

the completion portfolio led to an improvement of

multiple performance and risk statistics.

A wide range of measures have improved

Figure 2 shows that, by adding a completion portfolio,
the portfolio statistics improved for the majority of
Source: Invesco. Percentage of 642 selected US large cap funds where completion portfolios
the funds. For 80% of the funds, annualized returns led to improved statistics for the years 2012 to 2016 (based on total return in USD).
increased; volatility decreased for 81% of the funds

Risk & Reward, #2/2017 15

For each portfolio simulation exercise, conducted via Figure 3
a mean variance optimization, we started with an US case: lower tracking error to S&P 500 with completion portfolio
alpha forecasting process that uses a multi-factor
approach. Broadly speaking, our alpha signal has Without With
equal representation in factors associated with %
commonly used drivers of return (such as value, 6
momentum and quality). As in the academic
literature,4 our high value (i.e. inexpensive) stocks 5
tend to outperform low value (expensive) stocks,
high momentum stocks (those with high positive 4
returns or earnings forecast changes in the past)
tend to outperform low momentum stocks and high 3
quality tends to outperform low quality. Each of
these alpha concepts, in turn, is represented by 2
multiple sub-factors. As an example, value can be
represented by cash flow statement and balance 1
sheet data. Through diversification across concepts,
we looked to take advantage of factors with low or 0
negative historic correlations amongst one another, 2/06 2/08 2/10 2/12 2/14 2/16
while at the same time exhibiting positive correlations
with subsequent returns. Similarly, within each Source: Invesco. Data as at 31 December 2016. The figures are based on simulations of past
performance, which is no reliable indicator for the future. Tracking error based on total return
concept, diversification of factors can provide more in USD.
robust and stable exposure to returns from the US Client Completion
overall concept.

Our portfolio simulations also took into account our Figure 4

risk forecast, using a fundamental risk model that Non-US case: lower tracking error to MSCI EAFE with completion
includes our proprietary alpha signals, individual portfolio
asset position limits (+/- 250 bps relative to the
benchmark, or less based on estimated available Without With
liquidity) and risk factor constraints at portfolio %
level. These constraints include maximum active 4
positions for sectors/industries and countries/regions
(+/- 300 bps), as well as limits on common risk
factor exposures. 3

Possible risk reduction, higher information ratio

Figures 3 and 4 show that the primary objective 2
(reducing risk relative to the benchmark) has been
achieved for both the US and the non-US portfolio.
The portfolios, before inclusion of the completion 1
portfolio sleeve, had average tracking errors of
3.81% (US) and 3.6% (non-US). With the completion
portfolio, average tracking error fell to 2.6% (US) 0
and 2.25% (non-US). 2/06 2/08 2/10 2/12 2/14 2/16

Since the initial portfolios are concentrated Source: Invesco. Data as at 31 December 2016. The figures are based on simulations of past
performance, which is no reliable indicator for the future. Tracking error based on total return
(83 positions in the US portfolio, 80 in the non-US), in USD.
consist of relatively liquid securities and are not Completion
International Client

Table 1
Impact of the completion portfolio (in percentage points)

US: Initial US: Initial portfolio with Non-US: Initial Non-US: Initial portfolio
portfolio completion portfolio portfolio with completion portfolio
Currency -0.07 -0.05 -0.04 -0.07
Growth 0.12 -0.01 0.01 0.00
Leverage -0.15 -0.07 0.00 -0.02
Liquidity 0.07 0.11 0.05 0.03
Medium-term momentum 0.08 0.11 0.05 0.07
Short-term momentum 0.01 0.01 0.02 0.01
Size -0.08 -0.18 0.00 0.03
Value -0.14 0.03 -0.08 0.02
Volatility 0.10 0.05 0.11 0.04

Source: Invesco.

Risk & Reward, #2/2017 16

Table 2
Information ratios in comparison (in %)
US: Initial
portfolio
Active return (%) 0.27
Active risk (%) 3.41
Information ratio 0.08
Source: Invesco. Total return in USD. Active risk = tracking error.
overly burdened with constraints, reducing risk
was fairly straightforward. But, the litmus test for
a completion portfolio, beyond its ability to lower
tracking error, is whether the portfolio’s information
ratio (IR),5 i.e. its active return divided by its active
risk or tracking error, has increased through
lessened exposure to undesired risk factors and
increased exposure to longer-term alpha factors.
Table 1 shows the exposure to various risk and alpha
factors for the US and non-US simulations. In both
cases, the completion portfolio reduced the impact
of the volatility factor (which is typically the biggest
contributor to portfolio risk). In the US case, we also
observed a reduction in exposure to leverage. The
alpha factors, value and medium-term momentum,
both experienced increased exposure, notably in
the case of value, which moved from negative to
positive.
Finally, table 2 shows annualized return and risk
characteristics and IRs over the full simulation
period. Since the completion sleeve for both the
US and non-US simulation has lead to higher active
returns and lower active risk, the information ratio
has increased.
Conclusion
Each equity portfolio, and in fact every single stock,
exhibits certain factor characteristics. Analyzing
factor biases in existing portfolios is a first step
towards determining potential improvements to
their risk/return characteristics. By complementing
existing portfolios with bespoke factor completion
portfolios that account for certain gaps in terms of
(factor-based) diversification, we find that risk/return
profiles can be improved. The possibilities range
from highly liquid low-cost solutions using the broad
set of available factor ETFs, all the way through to
developing actively managed, highly customized
solutions that best reflect a client’s desired risk/
return targets. Our analysis shows that both ways
have the potential to meaningfully improve a wide
range of portfolio statistics, including the information
ratio.
Risk & Reward, #2/2017 17
US: Initial portfolio with Non-US: Initial Non-US: Initial portfolio
completion portfolio portfolio with completion portfolio
0.58 1.88 2.08
2.50 4.08 3.09
0.23 0.46 0.67
About the authors
Michael Abata
Director of Research,
Invesco Quantitative Strategies
In his role, Michael Abata is responsible for managing
the US based research team members. He oversees
the research of return and risk forecasting processes
used in managing our Invesco Quantitative Strategies
equity portfolios.
Georg Elsaesser
Senior Portfolio Manager,
Invesco Quantitative Strategies
Georg Elsaesser is involved in the management of
Invesco Quantitative Strategies equity portfolios and
has long-standing experience in the implementation
of factor-based investment solutions.
Brad Smith
ETF Research Analyst,
PowerShares by Invesco
Brad Smith works as an analyst within PowerShares’
ETF research group providing research and analysis
on PowerShares ETFs and the ETF industry.
Jason Stoneberg
Director, Research & Product Development,
PowerShares by Invesco
In his role, Jason Stoneberg oversees the research
and development of exchange-traded funds. He also
oversees the firm’s research efforts related to
PowerShares and the ETF industry.
Notes
1 E.g. Eugene F. Fama, Kenneth R. French (1992), “The Cross-Section of Expected Stock
Returns”, The Journal of Finance 47(2); M.M. Carhart (1997), “On Persistence in Mutual
Fund Performance”, Journal of Finance 52(1).
2 Morningstar, as at 31 December 2016.
3 The simulation presented here was created to consider the possible results using factor
based completion portfolios. These performance results are hypothetical (not real) and
were conducted via a mean variance optimization. The hypothetical results were derived
by back-testing using a simulated portfolio. It may not be possible to replicate these
results. There can be no assurance that the simulated results can be achieved in the
future. While the factor based completion portfolios were used to analyze the effect on
risk and return and to reduce the risk to a desired range relative to a benchmark, it does
not factor in all the economic and market conditions that can impact results. The
simulated performance results do not reflect the deduction of any fees. Returns would
be reduced by any applicable fees associated with the management of a portfolio.
4 Ibid.
5 The information ratio is sometimes referred to as portfolio utility.
How macro factors can aid asset
allocation
By Jay Raol, Ph.D.
In brief
In this paper, we establish our set of macro
factors – growth, inflation and financial
conditions, which display quite stable
correlations to the returns of various asset
classes – irrespective of their countries of
issuance. Furthermore, we analyze how
asset class volatility moves with the macro
factors. We believe that, by looking at the
sensitivities of asset class returns and
volatilities to changes in macro factors,
allocation within global multi-asset portfolios
can be improved.
Risk & Reward, #2/2017 18
Often, portfolios are built around the correlations
between asset classes. But, such an approach is
not without its shortcomings – especially since
the familiar correlations of the past changed
during the financial crisis. In this paper, we
present an alternative approach to portfolio
construction, one that is based on correlations:
but here the focus is on co-movements of asset
classes with various macro factors.
A primary aim of portfolio allocation is to balance
returns versus risk by adjusting an investment’s size
within an overall portfolio. Typically, an investor must
take into account his or her own risk tolerance,
investment goals and investment timeframe when
making allocation decisions. This makes correctly
measuring risk a central problem for the asset
allocator.
Traditionally, risk has been measured by examining
asset class volatilities and correlations between asset
classes. Investors typically examine the long-run
return, correlation and volatility of each asset class
to determine its size in the portfolio.
Figure 1
Correlations between major asset classes
(Each bar represents the average correlation between one and remaining
asset classes)
• Pre-crisis period (1 January 1997 to 30 June 2007)
• Post-crisis period (1 January 2010 to 31 December 2014)
Correlation
0.72 0.73
0.71 0.71
0.63 0.63
0.50
0.45 0.44 0.46
0.40
0.17
MSCI EM US Treasuries EMBI Global GBI-EM Broad US High Yield Commodities
local currency
Source: International Monetary Fund, “Global Financial Stability Report,” April 2015. Figure 1.20,
p. 34. Data as at 31 December 2014. Cross-asset correlation is measured as the median of the
absolute values of pair-wise correlations between the daily Sharpe ratios of the asset classes (i.e.
of all correlations between the daily Sharpe ratios of any two of the six asset classes, which is 15
correlation coefficients altogether) in the chart over a 60-day window. Asset classes are
represented by MSCI EM = MSCI Emerging Markets Equity Index; US Treasuries = 7–10-year US
Treasury Index; EMBI Global = JPMorgan Emerging Markets Bond Index Global; GBI-EM Broad
Pre-crisis Post-crisis
local currency = JPMorgan Government Bond Index-Emerging Markets in local currency; US HY =
US High-Yield Index; Commodities = Credit Suisse Index. Except for the GBI-EM Broad Local
Currency Index, all indices are in US dollars.
Cross-asset class correlations have risen … Figure 2
The global financial crisis and subsequent response Macro cycles in the United States
of policy makers to stabilize asset prices through
quantitative easing upended the traditional asset Financial cycle Business cycle
allocation model by changing historical correlations Deviation from trend, in %
and volatilities, making them less meaningful in 15
allocation decisions. Simply put, post-crisis
diversification across assets no longer provided 10
investors with their intended risk diversification. This
is because cross-asset class correlations have risen 5
significantly since 2008, making traditional
diversification strategies more challenging (figure 1). 0

… and asset class time series may be too short -5

Another problem with traditional asset allocation is
its dependence on historical data, which may not -10
encompass a full macroeconomic cycle. For example,
most data samples used in asset allocation only -15
include periods of declining interest rates, moderate 71 74 77 80 83 86 89 92 95 98 01 04 07 10 13
inflation and benign business cycle fluctuations. This Source: Borio, Claudio. “The financial cycle, the debt trap and secular stagnation.” Presentation at
is because most financial indices were created over the 84th Annual General Meeting, Bank of International Settlements, 29 June 2014. Data as at
the past few decades, whereas macroeconomic 29 June 2014. According to Borio, the financial cycle comprises the medium-term cycles in the
total non-financial debt-to-GDP ratio and real house prices. The business cycle is the fluctuation in
cycles may have long preceded them.1 real GDP.

From growth and inflation to risk and return

Seeking an alternative to overcome some of these Financial Cycle Business Cycle
challenges, many investors have turned to Figure 3
macroeconomic factors to better explain the risks Scenario analysis of US risk-adjusted returns
and returns in their portfolios. Although conventional
wisdom would suggest that growth and inflation • Bonds • Equities
have the largest effects on investment returns, they Sharpe ratio
are not directly investable. Therefore, it is difficult to 0.40
draw a concrete link between macroeconomic factors
and returns. As a result, many investors have turned
to a scenario-based framework, where they examine
the performance of asset returns in varying 0.13 0.12 0.11 0.04 0.03
economic environments.
-0.09
For example, the risk-adjusted returns of US equities
show a strong positive correlation to changes in
economic growth, irrespective of the inflation
backdrop (figure 3). Conversely, US bonds (and -0.37
commodities, not shown) seem to be more affected
by changes in the inflation rate. Down Up Down Up
Inflation Growth
The scenario analysis approach provides us with the
first clues toward understanding at least one
Sources: Bloomberg L.P., Invesco. Data from 1 January 1973 to 31 March 2017. Sharpe ratios
dimension of risk – correlation between asset are calculated on the excess returns of the Standard and Poor’s 500 equity price index (equities)
classes. When we look at the correlation between and the Bloomberg Barclays US Treasury Index (bonds). Growth and inflation are measured using
bonds and equities (figure 4, light blue line), two the Federal Reserve Bank of Philadelphia’s Survey of Professional Forecasters. Up and down
scenarios represent the average Sharpe ratios during periods of rising and falling growth and
different regimes can be clearly seen over the past inflation, respectively. Bonds Equities
four decades. In the period 1973 – 1998, bonds and

Risk & Reward, #2/2017 19

equities had a slightly positive correlation. Since Figure 4
then, however, the correlation has become much Macro versus asset correlation in the United States
more negative.
Growth/inflation Stocks/bonds
As for the correlation between the macro factors, Correlation
growth and inflation, there are also two clearly 0.8
different regimes (figure 4, dark blue line). In the
0.6
period 1973 – 1998, growth and inflation were
negatively correlated, while they have been 0.4
positively correlated since then.
0.2
And this is the point: the analysis indicates that asset 0.0
class (bond and equity) correlations are driven by
macro factors (growth and inflation). For example, in -0.2
periods when inflation is on the rise, intuition would -0.4
suggest that a fixed return asset (such as a bond)
would have an inferior return relative to a flexible -0.6
return asset (such as a stock). -0.8
1/73 1/78 1/83 1/88 1/93 1/98 1/03 1/08 1/13
Sources: Bloomberg L.P., Invesco. Correlations beginning in 1983 are based on 10-year rolling
data from 1 January 1973 to 31 March 2017. Growth and inflation are measured using the
The analysis can be used Federal Reserve Bank of Philadelphia’s Survey of Professional Forecasters. The correlation is
between the quarterly change in one-year-ahead real GDP growth and the growth in the GDP

to examine the portfolio deflator (measure of the level of prices of all new, domestically produced, final goods and
services). The correlation between stocks and bonds is measured by the correlation between

allocation problem through

excess price returns in the Standard and Poor’s 500 Index and the Bloomberg Barclays US
Treasury Index.
Growth/Inflation Stock/Bonds

a macroeconomic lens.
Figure 5
Macro environment affects diversification benefits
Putting macro factors to work in portfolio
allocation • Stocks • Bonds • Optimized portfolio
The above analysis can be used to examine the Sharpe ratio
portfolio allocation problem through a macroeconomic 0.82
lens. For instance, to answer the question of how
an investor should consider allocating between
stocks and bonds, we first develop a forward-
looking view of growth and inflation. These
0.53
forecasts allow us to construct a “macro factor
framework” to predict how various asset classes
will likely behave in each environment. 0.34 0.35

Figure 5 shows the “optimized portfolio” in each

0.18 0.19
regime – in other words, the allocation of stocks
and bonds that produced the maximum risk-
adjusted returns in each time period. It is possible
to see that the diversification benefits of holding 1973-1998 1998-2016
stocks and bonds is highly dependent upon the Sources: Bloomberg L.P., Invesco. Data from 1 January 1973 to 31 March 2017. The stock and
correlation between growth and inflation. For bond returns are derived from the Standard and Poor’s 500 Index and the Bloomberg Barclays US
Treasury Index. The optimized portfolios are the stock and bond weights that generate the largest
example, during the 1973-1998 regime, there was Sharpe ratios in each of the two periods.
essentially no benefit to owning both stock and

Stocks Bonds Optimized Portfolio

Data and methodology
To calculate risk-adjusted returns, we examined a large sample of global equity indices, credit spreads, 10-year government bond
yields, currencies, commodities, inflation-linked bonds and implied volatilities (see the appendix for the indices used to represent
each asset class). For bond yields, spreads and implied volatilities, we looked at monthly yield differences, and assumed a
portfolio with a one-year duration for ease of computation. For the remaining assets, we simply took their monthly price changes.
Finally, we converted all of the asset returns into Sharpe ratios by subtracting the risk-free return and dividing by the in-sample
volatility.
To this set of time series, we applied principle components analysis (PCA), which allows us to decompose the drivers of returns
into their “orthogonal”, or principle factors. Although PCA does not directly identify the factors, it can be used to infer them. To
solve the problem of time series with different data ranges, we used the “soft-impute” method to estimate the factors across the
whole sample.* Finally, to isolate only those factors that were stable during the sample period, we employed the “bootstrap”
method.** For robustness, we applied this analysis at weekly, monthly and quarterly frequencies. Simply put: we looked for factors
that explained returns in both random time samples and across random asset samples. We believe this helps to ensure the
stability and applicability of the factors.
* Mazumder, R., Hastie, T. and Tibshirani, R.: “Spectral Regularization Algorithms for Learning Large Incomplete Matrices”. J Machine Learning Research, 11 (2010),
p. 2287 – 2322. ** Efron, B.: Bootstrap Methods: “Another Look at the Jackknife”. Annals of Statistics, vol. 7, no. 1, p. 1–26.
All foreign currency equity, bond and volatility returns represent returns hedged into US dollars.

Risk & Reward, #2/2017 20

bonds - the optimized portfolio performed no better Figure 6
than either asset class. During this period, stocks Risk-adjusted returns per incremental changes in macro factors (ie.
and bonds were highly correlated, which we would decrease in growth, tightening of financial conditions and rising
expect since growth and inflation were negatively inflation)
correlated. In contrast, during the 1998-2016 Growth Financial conditions Inflation
regime, when growth and inflation were positively US.10Y
correlated, diversification produced tremendous GB.10Y
benefit – the optimized portfolio outperformed each CH.10Y
SE.10Y
asset class. KR.10Y
NO.10Y
Going beyond growth and inflation, bonds and NZ.10Y
JP.10Y
equities: volatility, currencies and a third macro IN.10Y
factor – “financial conditions” HK.10Y
This growth and inflation analysis is very appealing, DE.10Y
CZ.10Y
as it is simple, easy to visualize, intuitive and helps

Government bonds
CL.10Y
to explain past changes in bond/equity correlations. CA.10Y
AU.10Y
While this framework helps us better understand the CO.10Y
distant past, it is not as useful in explaining the more MY.10Y
recent (post-2008) world. Typically, periods around TH.10Y
TR.10Y
major shifts in monetary policy do not fit neatly ZA.10Y
within the growth and inflation factor framework. RU.10Y
Moreover, while growth and inflation shed light on BR.10Y
ID.10Y
the correlation between asset classes, they are less HU.10Y
helpful in predicting volatility – the other important PL.10Y
MX.10Y
dimension of risk. Additionally, much of the analysis AU.Equity
done around the macro factor framework has been CN.Equity
focused on US history and assets. In a globally HK.Equity
JP.Equity
integrated economy and capital markets, however, DE.Equity
we consider currency risk to be an equally important FR.Equity
Equities

UK.Equity
dimension of risk. By including currencies in our EU.Equity
study, we aim to extend the macro factor framework CO.Equity
to inform asset allocation for global portfolios. CL.Equity
BR.Equity
MX.Equity
To address these shortcomings, we’ve reinvestigated CA.Equity
correlation and volatility across a broader range of US.Equity
SA.10Y.BE
asset classes using a multivariate statistical study. BR.10Y.BE
Inflation-linked

We hope to show (1) that the global macroeconomic CA.10Y.BE

AU.10Y.BE
environment dominates risks in a global portfolio; (2) JY.5Y5Y.CPI
that global financial conditions are an important UK.10Y.BE
bonds
Risk assets

driver of risk and return; (3) that volatility is driven DE.10Y.BE

MX.10Y.BE
by macro factors. US.10Y.BE
Copper
Comm.

The global dimension of asset returns Aluminum

Gold
The first observation from this analysis is the Oil
consistency of returns among asset classes across DB.Move
G20.Curr.Vol
Volatility

different geographic regions and factors. By NKY.VIX

grouping those assets with similar signs under the HIS.VIX
three macro factors: growth, financial conditions and SX53.VIX
VIX
inflation, we identify three main asset clusters: TU.Sov.Sprd
Credit EM Sovereign

government bonds, risk assets (equities, duration- RU.Sov.Sprd

hedged inflation-linked bonds, commodities, implied ID.Sov.Sprd
SA.Sov.Sprd
volatilities, duration-hedged emerging market US BR.Sov.Sprd
dollar-denominated sovereign debt, duration-hedged PL.Sov.Sprd
MX.Sov.Sprd
developed market credit) and currencies. Assets EU.IG
within these three clusters tended to behave similarly US.HY
to each other in different macro environments, US.IG
USDJPY
regardless of geographic location. We believe this USDEUR
consistency provides further support for the USDGBP
influence of macro factors on asset class behaviour. USDSEK
USDKRW
Currencies

USDZAR
Once we identified these three clusters, the asset USDSGD
USDCLP
class allocation problem was significantly reduced. USDAUD
Although the investment universe comprises USDNZD
numerous individual assets, by taking correlations USDCOP
USDCAD
into account, investable assets may be grouped USDBRL
into as few as eight categories: global equities and
-0,2 0 0,2 -0,2 0 0,2 -0,2 0 0,2
credit, global developed market government bonds,
global emerging markets government bonds, global Sources: Bloomberg L.P., Invesco. Data from 1 January 2003 to 1 July 2016. The sign and size
inflation-linked bonds, commodities, currencies and of the bars indicate the direction and strength, respectively, of the relationship between the factor
and the asset. Where assets tended to share similar signs across all three macro environments,
volatilities. The output from the PCA is shown in they were grouped into clusters indicated by the boxes on the left.
figure 6.

Risk & Reward, #2/2017 21

Figure 6 shows the average risk-adjusted return Figure 7
(Sharpe ratio) of each asset in response to the Factor portfolios follow macro fundamentals
macro factor over the time period. For example,
a decrease in the growth factor (recession) led Chicago Fed Adjusted Financial Conditions Index
Average Sharpe ratio (RHS)
to positive Sharpe ratios for government bonds,
implied volatilities and the US dollar against other 2 0.5
currencies.
1 0.3
Part of our statistical approach was to avoid
predetermining what factors might be at work in
driving asset returns – for example, based on
0 0.0
economic theory. Instead we looked for consistent
relationships to emerge out of the data using PCA
and then confirmed whether the relationships had -1 -0.3
an economic basis before we identified drivers as
macro factors. We found three stable factors in our
analysis: growth, inflation and financial conditions, -2 -0.5
which we discuss further below. Moreover, the 1/03 1/05 1/07 1/09 1/11 1/13 1/15
relationships implied by our study matched the
scenario results shown in figure 3. Federal Reserve Bank of Chicago National Activity Index
Average Sharpe ratio (RHS)
Figures 7a-c show the Sharpe ratio of the “factor
5.0 0.70
portfolio” at each point in time.2 We construct the
factor portfolio by going long or short for each Chicago Fed Adjusted Financial Conditions Index
individual asset according to the signs and strengths 2.5 0.35
Average Sharpe Ratio RHS
of the Sharpe ratio returns as shown in figure 6.
Figures 7a-c show that each factor portfolio
demonstrates a close relationship with the underlying 0.0 0.00
macro fundamental index. For example, the average
Sharpe ratio of the portfolio in figure 7a follows the
Chicago Fed Adjusted Financial Conditions Index -2.5 -0.35
closely, suggesting that portfolio returns are sensitive
to global financial conditions. Furthermore, figures
7a-c reinforce our view that financial conditions have -5.0
1/03 1/05 1/07 1/09 1/11 1/13 1/15
-0.70
had an outsized impact on portfolio returns in the
post-crisis period.
Citi Global Inflation Surprise Index
As mentioned above, an important outcome of Average Sharpe ratio (RHS)
our analysis was the identification of a third factor, 24 0.24
distinct from the widely accepted factors of growth
and inflation. This third factor seemed to be 16 Federal Reserve Bank of Chicago National Activity Index 0.16
correlated with several proxies for financial Average Sharpe Ratio RHS
conditions. We believe that this “financial conditions” 8 0.08
factor, or this “policy factor”, corresponds to the
effect of monetary and fiscal policy on asset prices. 0 0.00
Because financial conditions affect the discount
-8 -0.08
rate that investors use to determine the net present
value of any asset, any tightening of financial
-16 -0.16
conditions should theoretically prove negative for
all asset classes. We believe this factor provides -24 -0.24
the missing link in the post-2008 world, where 1/03 1/05 1/07 1/09 1/11 1/13 1/15
equity and bond returns have been positive despite
anaemic growth and inflation. It would seem that Source: Bloomberg L.P., Invesco. Data from 1 January 2003 to 1 July 2016. The factor portfolios
unconventional monetary policy (loose financial are graphed along with their corresponding macro factor. For financial conditions, the six-month
conditions) can be considered the primary driver change in the Federal Reserve Bank of Chicago Adjusted Financial Conditions Index is graphed
alongside the six-month average Sharpe ratio for the financial conditions factor. Growth and
of returns. inflation are measured using the three-month change in the Federal Reserve Bank of Chicago
National Activity Index and the three-month change in the CITI Inflation Surprise Index.
Volatility and macro factors
Citi Global Inflation Surprise Index Average Sharpe Ratio
Finally, we incorporated implied volatilities in our
study. As shown by figure 6, volatilities of equities,
interest rates and currencies all tended to respond Conclusion: building portfolios around macro
similarly to the macro factors, i.e. they too acted factors
as a cluster. Our study shows that volatility increased Our study indicates that assets around the world
when growth fell, financial conditions tightened or move according to their asset classification – and not
inflation rose. We believe that it is important to their geographic location. This finding, along with
understand this asymmetry in the reaction of the findings of our scenario and correlation analyses,
volatility to macro factors when sizing risk in has helped establish our set of macro factors. We
portfolio allocation. For example, the same bond believe the three macro factors identified here
allocation may pose different levels of risk in determine the main correlations of asset classes to
different macro environments due to different macro drivers. Lastly, we addressed the matter of
levels of bond market volatility. volatility, and how it moves with macro factors.

Risk & Reward, #2/2017 22

Figure 8
Asset class weights based on macro factor changes
Macro factor shock Developed Emerging Commodities Global Implied Duration Duration Long US dollar Long US dollar
market market equities volatilities hedged hedged versus versus
government government inflation- corporate developed emerging
bonds bonds linked bonds bonds currencies market
currencies

Growth down
Financial conditions
tightener
Inflation up

Source: Invesco, as at 12 April 2017. Red is underweight, green is overweight, yellow is neutral weight.

Together, we can use the sensitivities of asset class About the author
correlations and volatilities to better allocate within
our global portfolios. Jay Raol, Ph.D.
Senior Macro Analyst, Invesco Fixed Income
Applying our framework to the portfolio construction Jay Raol is a member of the macro research team at
problem of investing in a rising inflation Invesco Fixed Income. In his role, he works on macro
environment, we would expect global bonds and economic based models for asset allocation and
equities to underperform based on historical fixed income investing.
correlations to macro factors, while commodities,
inflation-linked bonds and developed market
currencies and volatilities should outperform. We
would thus seek to position our global portfolio
according to the weights illustrated in figure 8.

In future papers, we will discuss how Invesco Fixed Notes

Income utilizes this macro factor framework to 1 M. Drehmann, C. Borio and K. Tsatsaronis (2012): Characterizing the financial cycle: don’t
lose sight of the medium term!, Bank for International Settlements Working Papers, no. 380,
inform our investment process and aid portfolio June 2012, graph 3, p.19.
construction. 2 All calculations are gross of possible fees that might apply to investors.

Appendix
The below indices represent the range of asset classes used in the PCA analysis.

Ticker Name Ticker Name

USDAUD Curncy USDAUD Spot Exchange Rate - USDPLN Curncy USDPLN Spot Exchange Rate -
Price of 1 USD in AUD Price of 1 USD in PLN
USDBRL Curncy USDBRL Spot Exchange Rate - USDRUB Curncy USDRUB Spot T+1 (TOM) Exchange Rate -
Price of 1 USD in BRL Price of 1 USD in RUB
USDCAD Curncy USDCAD Spot Exchange Rate - USDSGD Curncy USDSGD Spot Exchange Rate -
Price of 1 USD in CAD Price of 1 USD in SGD
USDCLP Curncy USDCLP Spot Exchange Rate - USDZAR Curncy USDZAR Spot Exchange Rate -
Price of 1 USD in CLP Price of 1 USD in ZAR
USDCOP Curncy USDCOP Spot Exchange Rate - USDKRW Curncy USDKRW Spot Exchange Rate -
Price of 1 USD in COP Price of 1 USD in KRW
USDCZK Curncy USDCZK Spot Exchange Rate - USDSEK Curncy USDSEK Spot Exchange Rate -
Price of 1 USD in CZK Price of 1 USD in SEK
USDHUF Curncy USDHUF Spot Exchange Rate - USDTWD Curncy USDTWD Spot Exchange Rate -
Price of 1 USD in HUF Price of 1 USD in TWD
USDINR Curncy USDINR Spot Exchange Rate - USDTHB Curncy USDTHB Spot Exchange Rate -
Price of 1 USD in INR Price of 1 USD in THB
USDIDR Curncy USDIDR Spot Exchange Rate - USDTRY Curncy USDTRY Spot Exchange Rate -
Price of 1 USD in IDR Price of 1 USD in TRY
USDJPY Curncy USDJPY Spot Exchange Rate - USDGBP Curncy USDGBP Spot Exchange Rate -
Price of 1 USD in JPY Price of 1 USD in GBP
USDMYR Curncy USDMYR Spot Exchange Rate - USDEUR Curncy USDEUR Spot Exchange Rate -
Price of 1 USD in MYR Price of 1 USD in EUR
USDMXN Curncy USDMXN Spot Exchange Rate - LUCROAS Index Bloomberg Barclays US Agg Credit
Price of 1 USD in MXN Avg OAS
USDNZD Curncy USDNZD Spot Exchange Rate - LF98OAS Index Bloomberg Barclays US Corporate
Price of 1 USD in NZD High Yield Average OAS
USDNOK Curncy USDNOK Spot Exchange Rate - LECPOAS Index Bloomberg Barclays EuroAgg
Price of 1 USD in NOK Corporate Average OAS

Risk & Reward, #2/2017 23

Ticker Name Ticker Name
JPSSEMME Index J.P. Morgan EMBI Plus Mexico DAX Index Deutsche Boerse AG German Stock
Sovereign Spread Index DAX
JPSSGDPO Index J.P. Morgan EMBIG Diversified NKY Index Nikkei 225
Poland Sovereign Spread
JPSSEMBR Index J.P. Morgan EMBI Plus Brazil HSI Index Hong Kong Hang Seng Index
Sovereign Spread
JPSSEMSA Index J.P. Morgan EMBI Plus South Africa SHSZ300 Index Shanghai Shenzhen CSI 300 Index
Sovereign Spread
JPSSEMID Index J.P. Morgan EMBI Plus Indonesia AS51 Index S&P/ASX 200
Sovereign Spread
JPSSEMRU Index J.P. Morgan EMBI Plus Russia XAU Curncy XAUUSD Spot Exchange Rate -
Sovereign Spread Price of 1 XAU in USD
JPSSEMTU Index J.P. Morgan EMBI Plus Turkey LMAHDS03 LME LME Aluminum 3 Month Rolling
Sovereign Spread Comdty Forward
VIX Index Chicago Board Options Exchange LMCADS03 LME LME Copper 3 Month Rolling Forward
SPX Volatility Index Comdty
V2X Index EURO STOXX 50 Volatility Index GACGB10 Index Australia Govt Bonds Generic Yield
VSTOXX 10 Year
VHSI Index HSI Volatility Index GEBR10Y Index Brazil Government Generic Bond
10 Year
VNKY Index Nikkei Stock Average Volatility Index GCAN10YR Index Canadian Govt Bonds 10 Year Note

JPMVXYGL Index J.P. Morgan Global FX Volatility CHSWP10 CMPN CLP SW PESO v CAMARA 10Y
Index Curncy
USCRWTIC Index Bloomberg West Texas Intermediate COGR10Y Index Colombia Government Generic Bond
(WTI) Cushing Crude Oil Spot Price 10 Year Yield
BFCIUS Index Bloomberg United States Financial CZGB10YR Index Czech Republic Governments Bonds
Conditions Index 10 Year Note Generic Bid Yield
GSERMUS Index Goldman Sachs MAP Economic GDBR10 Index Germany Generic Govt 10Y Yield
Surprise Index - US
USGGBE10 Index US Breakeven 10 Year HKGG10Y Index Hong Kong Generic 10 Year

MXGGBE10 Index Mexico Breakeven 10 Year GHGB10YR Index GDMA Hungarian Govt Bond 10 Year

DEGGBE10 Index Germany Breakeven 10 Year GIND10YR Index India Govt Bond Generic Bid Yield
10 Year
UKGGBE10 Index UK Breakeven 10 Year GIDN10YR Index Indonesia Govt Bond Generic Bid Yield
10 Year
FWISJY55 Index JPY Inflation Swap Forward 5Y5Y GJGB10 Index Japan Generic Govt 10Y Yield

ADGGBE10 Index Australia Breakeven 10 Year MAGY10YR Index Malaysia Govt Bonds 10 Year Yield

CDGGBE10 Index Canada Breakeven 10 Year GMXN10YR Index Mexico Generic 10 Year

BRGGBE10 Index Brazil Breakeven 10 Year NDSW10 Curncy NZD SWAP 10YR

SAGGBE10 Index South Africa Breakeven 10 Year NKSW10 CMPN NOK SWAP 10YR
Curncy
MOVE Index Merrill Lynch Option Volatility POGB10YR Index Poland Government 10 Year Note
Estimate MOVE Index Generic Bid Yield
SPX Index S&P 500 Index RRSWM10 Curncy RUB SWAP VS MOSPRIME 10Y

SPTSX Index S&P/TSX Composite Index GSAB10YR Index South Africa Govt Bonds 10 Year Note
Generic Bid Yield
MEXBOL Index Mexican Stock Exchange Mexican GVSK10YR Index KCMP South Korea Treasury Bond
Bolsa IPC Index 10 Year
IBOV Index Ibovespa Brasil Sao Paulo Stock GSGB10YR Index SWEDISH GOVERNMENT BOND
Exchange Index 10 YR NOTE
IPSA Index Santiago Stock Exchange IPSA GSWISS10 Index Switzerland Govt Bonds 10 Year Note
Index Generic Bid Yield
COLCAP Index Colombia COLCAP Index GVTL10YR Index Thailand Govt Bond 10 Year Note

SX5E Index EURO STOXX 50 Price EUR GTRU10YR Index USD Turkey Govt Bond Generic Bid Yield
10 Year
UKX Index FTSE 100 Index GUKG10 Index UK Govt Bonds 10 Year Note Generic
Bid Yield
CAC Index CAC 40 Index USGG10YR Index US Generic Govt 10 Year Yield

Risk & Reward, #2/2017 24

Investing in a
multi-asset multi-factor world
By Alexandar Cherkezov, Dr. Harald Lohre, Sergey Protchenko and Jay Raol, Ph.D.

In brief
In this article, we advance the use of factor
investing across multiple asset classes. It
turns out that style factors well established
in the equity domain – such as value,
momentum or quality – do extend to other
asset classes as well. Even more so, multi-
asset multi-factors significantly expand the
investment opportunity set relative to a
traditional multi-asset universe. Seeking
to exploit this potential, we put forward
an innovative diversified risk parity strategy
that is designed to strive for maximum
diversification in the multi-asset multi-factor
world. To illustrate the strategy’s merits, we
investigate its stylized facts vis-à-vis more
standard allocation approaches.

Risk & Reward, #3/2017 25

An increasing number of investors see value in
using factors in an attempt to control risk or
improve returns – or both. While broad adoption of
factor investing is still well in the future, a multi-
asset multi-factor strategy permits full utilization
of all potential benefits of factor investing, and
therefore represents a natural evolution. In this
article, we develop such a strategy – and use the
opportunity for a thorough discussion into various
aspects of factor investing.
Diversification is a powerful risk control tool. In order
to reduce uncompensated risk, diversification seeks
to spread investments across different asset classes.
Combining asset classes that do not move in tandem
is meant to help smooth the overall investment
experience, maximizing the benefits of diversification.
In other words, losses in one asset class might be
mitigated, or even completely offset, by gains in
another. Provided that all asset classes genuinely
offer a sound risk-return rationale, one would expect
such a multi-asset portfolio to yield more attractive
risk-adjusted returns than any of the contributing
asset classes on its own.
Diversification is a powerful
risk control tool.
The limits of this basic logic were severely challenged factors. Figure 2 gives a quick overview of the most
when multi-asset diversification failed to smooth the
investment experience during the 2008 financial
crisis. With the exception of high-quality government
bonds, most risky assets declined simultaneously,
which led to larger than expected drawdowns in
seemingly diversified portfolios. While one may
argue that the financial market crisis was a rather
rare event, it was not the first time markets have
exhibited such a risk on/risk off pattern, i.e. risky
assets being bought or sold across the board.
Thus, two vital questions emerge. First: are there
additional factors to aid traditional asset classes in
navigating turbulent times? Second: given a viable
set of factors, how can a diversified portfolio strategy
be constructed using factors as building blocks?
Advancing better building blocks for portfolio
construction
From asset classes to factors
The systematic search for reasonable and uncorrelated
building blocks to complement traditional asset
allocation suggests turning to factor strategies.
Factor strategies have exploded in popularity – often
referred to as smart beta, alternative risk premia or
style investing. Factor investing offers great promise,
but investors should be cautious. Just because a
potential factor looked promising in the past, doesn’t
necessarily mean that the same factor will be useful
going forward. How, then, can relevant factors be
Factor investing offers great
promise.
Risk & Reward, #3/2017 26
identified? Typically, a given factor provides a
premium for bearing a certain type of risk. Often,
the list of factors is joined by market anomalies
that may not be attributable to risk, but exhibit a
consistent return profile explainable by a behavioural
or structural rationale (figure 1).
Establishing a solid rationale is the first step to
identifying relevant factors, and helps safeguard
against data-snooping biases that hamper analysis
of the cross-section of asset returns. In addition,
any given factor should comply with a strict set of
criteria: the factor must offer an attractive risk and/
or return profile established using sound econometric
techniques, the factor must be unique in the sense
that it complements what is already in place, the
factor rationale must be persistent through time
and the factor must be investable as opposed to
just delivering profits on paper.
Anatomy of factors
A common feature of traditional asset classes is their
directional nature; that is, an investment in a traditional
asset class is exposed to market risk. When markets
appreciate or depreciate broadly, traditional asset
classes generally follow suit. Conversely, factors
often perform differently, particularly if we control
for market movement. For instance, the value factor
might be defined within a traditional asset class as
the performance of value securities compared to
growth securities. Typically, factors represent a
certain investment style within a given asset class –
and these factors are thus referred to as ‘style’
Figure 1
Rationalizing factors
Risk-based explanation
The factor compensates
for systematic or
genuine risk.
Behavioural explanation Structural explanation
The factor is rooted The factor premium
in persistent, but not potentially results
necessarily rational from the structure of
investor behaviour. the industry, market
constraints, etc.
Data snooping? Safeguard against data-snooping biases
when testing for the relevance of given factors.
Source: Invesco. For illustrative purposes only.
Figure 2
Major style factors
Carry
Value Momentum
Market
Quality Defensive
Source: Invesco. For illustrative purposes only.
salient style factors – carry, value, momentum,
quality and defensive. Interestingly, investigation
shows these general factors apply across asset
classes. Thus, one could think of these factors as
likely approximations of latent risk factors in an
asset pricing context. Managing relevant factors in
conjunction with multiple asset classes therefore
both expands the opportunity set and improves risk
controls.
Relevant factors and asset classes
To aid more detailed understanding, we present here
a set of relevant factors across asset classes.
• Equity factors: We manage four factors within
global equities (value, momentum, quality and
defensive). These factors are not based on just
one defining characteristic. We utilize a wide-
ranging set of characteristics capturing each of
the four factor concepts. The momentum factor
consists of both price and earnings momentum
characteristics. Value consists of a differentiated
set of indicators that aim to capture a premium
more reliably than any single indicator. The
quality concept is based on metrics that measure
operational efficiency of firms, as well as financial
health. The defensive concept captures low
volatility characteristics of stocks.
In constructing the corresponding factor portfolios,
genuine factors are optimized in a way that
minimizes exposure to other unwanted factors.
As a result, the equity factors represent a clear-
cut set of factors when compared to a naive
sorting of firm characteristics to build factor
portfolios. Construction takes into account
Genuine factors are optimized
in a way that minimizes
exposure to other unwanted
factors.
Risk & Reward, #3/2017 27
 Carry: High yield assets tend to outperform low
yield assets – provided all else is equal
 M
 omentum: Recent relative price winners tend
to outperform recent relative price losers
 M
 arket: Asset classes tend to exhibit time series
predictability that can be exploited systematically
 Defensive: Low risk assets tend to have higher
risk-adjusted returns than high risk assets
 Q
 uality: High quality assets tend to have higher
risk-adjusted returns than low quality assets
 Value: Cheaper assets (according to a given
valuation metric) tend to outperform expensive
ones
portfolio turnover, and includes transaction costs
such that the associated factor returns correspond
to investable portfolios.
• Currency factors: To build FX (foreign exchange)
factors, we consider a universe of liquid currencies
from both developed and emerging markets.
The selection of liquid currencies is based on the
Triennial Central Bank Survey of FX turnover.1
The number thus increases over time, and currently
comprises 23 currencies. Large increases occur
after 1998 and 2004 as more EM currencies had
sufficient liquidity and were de-pegged from the
US dollar. Any pegged currency is not included
in the universe.
To construct FX style portfolios, the universe is
sorted into terciles based on factor-defining
criteria. With FX carry, for instance, currencies
are sorted according to 1-month forward yield,
and the carry trade is long the highest carry
tercile and short the lowest carry tercile. Both
sides are equally weighted, and weights are
rebalanced monthly. Besides FX carry, we cover
three further FX styles: (1) FX value is long the
cheapest currencies and short the most expensive
currencies based on purchasing power parity;2
(2) FX momentum is long/short the best/worst
performing currencies based on the currencies’
3-month price momentum; (3) FX quality is long
currencies with falling structural inflation, and
short currencies with increasing structural inflation.3
Currencies with falling structural inflation are
likely to appreciate nominally through purchasing
power parity. In addition, falling structural inflation
is associated with productivity increases, which
are associated with increases in the real effective
exchange rate.4
• Commodity factors: For commodities, we consider
three factors (carry, curve and momentum) within
the 24 commodities of the S&P GSCI universe.
Due to the unique nature of commodities, their
prices are primarily driven by inventory conditions.
The commodity carry strategy longs commodities
with tighter time-spreads. When supply and
 demand for commodities are persistent over time,
this strategy tends to outperform, as it captures
the price appreciation driven by low inventories.
To capture the correlation between time-spreads
and price movements, the commodity curve
strategy goes long deferred future contracts, and
shorts the nearest-dated contracts of each basket
commodity. The short position is beta-adjusted
and rebalanced on a monthly basis. Lastly, we
include the momentum strategy, which captures
the market’s under- or overreaction to changes
in commodity demand. It especially outperforms
when supply takes longer to adjust.5
• Interest rates factors: To build rates factors, we
consider a universe of liquid 10-year OTC swaps
from both developed and emerging markets.
The selection of liquid swaps is based on the
Triennial Central Bank Survey of OTC interest rate
derivatives turnover.6 The number thus increases
over time.
To construct rates style portfolios, the universe is
sorted into terciles based on the factor-defining
criteria. Building rates value portfolios, we first
compute the real yield for each country by
subtracting inflation from the 10 year yield and
then go long the highest real yield tercile and
short the lowest real yield tercile. Both sides are
equally weighted, and weights are rebalanced
monthly. Besides value, we consider three further
rates styles: (1) rates momentum is long/short
the swaps with the highest/lowest yield changes
over the last 12 months excluding the latest
month; (2) rates quality is long swaps with a low
beta to the universe average monthly change in
yield and short swaps with a high beta to the
universe average monthly change in yield using
a rolling 5 year window; (3) carry is long a
combination of long/short highest/lowest 2 year
yields, the slope (the difference between 10 and
2 year yields), and the butterfly (the difference
between the 5 and the average of 2 and 10 year
yields).
• Traditional asset classes: To manage exposure
relative to traditional asset classes, we consider
equity index futures for the S&P 500, Nikkei 225,
FTSE 100, EuroSTOXX 50 and MSCI Emerging
Markets, as well as bond futures for US 10Y
Treasuries, EuroBund, JGB 10Y and Gilt. The credit
risk premium is represented by the Bloomberg
Barclays US Corporate Investment Grade and High
Yield Indices, both duration-hedged.7
Mean-variance spanning of style factors vis-à-vis
traditional asset classes
To help gauge the contribution of the investment
factors, it is instructive to examine the corresponding
factor returns and volatilities with a mean-variance
diagram alongside traditional asset classes. Figure 3
charts all factors and asset classes in terms of their
mean and standard deviation, pertaining to the
sample period from 2001 to 2016. To visualize the
investment opportunity set of a classical multi-asset
investor, we compute the efficient frontier, based on
the above international bond and equity indices
labelled MA (multi-asset).8 Then, we consecutively
add factor sets by asset class – first, we include
equity factors, which significantly boost the return

Figure 3
Mean-variance spanning of asset classes and factors
MA MA + EQ MA + EQ + FX
MA + EQ + FX + Cmdty MA + EQ + FX + Cmdty + Rates
Expected return
0.10
FX Carry

Cmdty Carry MSCI EM

0.08

Equity Defensive
Cmdty Curve
Equity Momentum
Equity.Value FX Momentum
0.06
FX Quality Credit HY Nikkei 225
Equity Quality Cmdty Momentum
Bund
US 10Y S&P 500

0.04 Gilt
Rates Value FTSE 100
Rates Carry FX Value
Rates Quality
Credit IG EuroSTOXX 50
JGB 10Y
0.02 Rates Momentum

0.00 0.00 0.05 0.10 0.15 0.20 0.25

Standard deviation
The figure depicts efficient frontiers based on different sets of underlying asset classes and/or factors. Starting with the traditional asset
classes: equity, government bonds and duration-hedged corporate bonds, we compute the multi-asset (MA) efficient frontier (dark blue
line). The underlying mean-variance optimizations are subject to full investment and short-sale constraints. Mean-variance inputs are
derived from monthly return data over the sample period from 31 January 2001 to 31 December 2016. Next, we sequentially add
further style factors by asset class to compute efficient frontiers labelled MA + EQ (light blue line for multi-asset and equity factors),
MA + EQ + FX (purple line for adding foeign exchange style factors), MA + EQ + FX + Cmdty (orange line for adding commodity styles
factors), and finally MA + EQ + FX + Cmdty + Rates (green line representing the frontier based on all traditional and style factors).
Sources: Bloomberg,
MA Invesco, Goldman Sachs. MA+EQ MA+EQ+FX MA+EQ+FX+Cmdty
MA+EQ+FX+Cmdty+FI SP500 EuroSTOXX50 FTSE100
Nikkei225 MSCI_EM US10Y Bund
JGB10Y Gilt Credit.IG Credit.HY
Risk & Reward, #3/2017 28
Equity.Quality Equity.Defensive Equity.Value Equity.Momentum
FX.Carry FX.Value FX.Momentum FX.Quality
perspective and testify to favourable equity factor
performance. Next, we additionally include the four
FX factors, shifting the efficient frontier further to
the northwest. The same can be observed when
including commodity factors. Finally, we add the
four rates factors to the mix. Note, that these factors
have historically shown sub-par return performance.
Still, they help shift the frontier to the left by
offering diversification benefits. In addition, the rates
factors are likely to perform well in a stable to rising
interest rate environment. Given the low level of
current yields, we expect their forward-looking
return to be higher than their past returns. While
this shift seems rather marginal when compared to
the other shifts, it is statistically relevant in terms
of ameliorating the properties of the minimum-
variance portfolio.9 Stated differently: the four rates
factors represent meaningful tail hedges. Of course,
these results all have to be taken with a grain of
salt, given the input sensitivity of any mean-variance
optimization. Still, we find in unreported results
that these general conclusions continue to hold
when investigating shorter and more recent time
periods.
Maximum diversification in a multi-asset
multi-factor world
Having selected a viable set of assets and factors,
the final challenge is optimally combining them into
one coherent portfolio. While the classical mean-
variance paradigm of Markowitz (1952), shown in
figure 3, is a classic method for optimally balancing
the trade-off of expected portfolio return and risk,
it often suggests highly concentrated portfolio
weights that – experience tells us – are likely to
disappoint ­ex-post. Obviously, these observations
are rather unsettling given that diversification is at
the heart of mean-variance portfolio theory.
How to define and achieve a diversified portfolio
allocation
In the framework of Markowitz, diversification benefits
are usually considered as increases in expected
portfolio return at a given level of risk and/or decreases
in expected portfolio risk at a given level of return.
However, diversification as such is not explicitly
defined. Notably, Meucci (2009) has advanced
a framework to manage the degree of portfolio
diversification, which happens to resonate well with
the general intuition that ‘a portfolio is well-diversified
if it is not heavily exposed to individual shocks.’ To
this end, Meucci suggests extracting uncorrelated
risk sources from the underlying assets. Maximum
diversification obtains for a risk parity strategy along
these uncorrelated risk sources. That is, the portfolio
is allocated in such a way that each uncorrelated risk
source contributes equally to overall portfolio risk.
We dub this approach ‘diversified risk parity’ (DRP)
as in Lohre, Opfer and Ország (2014).
Extending to the multi-asset multi-factor universe
The diversification rationale naturally extends to
the case of multi-asset multi-factor investing. At the
heart of a maximum diversification strategy is the
choice of risk model and the corresponding factors
along which to diversify. In general, there are three
viable options:
1. ‘Kitchen sink’: Consider every asset class
and factor as a unique source of risk. This
operationally simple approach does not rely
Risk & Reward, #3/2017 29
on clustering asset classes or factors, and
can therefore become unstable because of
dimensionality concerns.
2. Cluster factors by asset class: This approach
would resonate with the organizational structure
of a typical asset management organization.
However, it fails to account for factor risks that
cross asset classes.
3. Cluster factors across asset classes: Clustering in
this way is a natural choice when considering the
world from a pure factor investing view, as it more
closely ties the return and risk objectives to the
diversification process.
Therefore, we will focus on option 3. Naturally, a
complete risk model should jointly consider major
traditional asset classes, and aggregate style factors.
In this vein, figure 4 depicts a correlation matrix of
seven components – the three asset classes and the
four relevant factors aggregated across asset classes.
As for asset classes, we consider a global equity market
factor, a global duration risk factor and a credit risk
factor. As for style factors, we consider aggregate
carry, value, momentum and quality factors.10
As a result, we obtain building blocks for portfolio
construction that generally tend to be uncorrelated.
Note, that the highest correlations can be found
among asset classes, e.g. government bonds
negatively correlated to equities and credit. The
latter two display the highest correlation: 0.71.
Across factors, the highest correlation amounts
to 0.40 (momentum vs. quality), suggesting ample
room for diversification.
Figure 4
Clustering by styles – correlation of traditional
asset classes and style factors
Equity
-0.32 Duration
0.71 -0.39 Credit
0.35 -0.01 0.35 Carry
0.12 -0.14 0.16 0.30 Value
-0.31 0.24 -0.20 0.18 0.02
Mo­men­
tum
-0.67 0.36 -0.57 -0.23 -0.14 0.40 Quality
The chart plots the correlation of aggregate asset class risk
factors and style factors over the whole sample period from
31 January 2001 to 31 December 2016. Above the diagonale
we visualize the assets’ and factors’ linear relationships in terms
of ellipsoids. Blue (red) ellipsoids represent positive (negative)
correlations that collapse into a straight line for a correlation of
+/- 1. Below the diagonale we give the corresponding correlation
coefficients.
Sources: Bloomberg, Invesco, Goldman Sachs.
Maximum diversification benefits Figure 5
To accord with the diversification framework of Meucci Risk decomposition and portfolio weights
(2009), the set of seven asset classes and style factors
must be transformed into one of uncorrelated risk • Equity • Duration • Credit
• Carry • Value • Momentum
sources. However, to achieve our ultimate objective, • Quality
we must manage each of the seven components Risk decomposition, %
such that they remain true to their definition while 100
interacting with the other components in a predictable
way. A traditional academic route would consider a 90
‘principal component analysis’ (PCA). A PCA is nothing 80
but a data reduction technique that extracts as much
of the underlying factors’ variation as possible, while 70
forcing zero correlation across components. While
60
this procedure results in uncorrelated risk sources,
these would be rather cumbersome in the context of 50
portfolio management: principal components tend to
be hard to interpret in economic terms, and are fairly 40
unstable. The latter characteristic would unduly 30
increase portfolio turnover, invalidating any associated
portfolio strategy. 20

10

0
One needs to transform the
2/06 2/08 2/10 2/12 2/14 2/16

set of seven asset classes •

•
S&P 500
MSCI EM
• EuroSTOXX 50 •
• Nikkei 225 •
FTSE 100
US 10Y
and style factors into one of •
•
Bund
Credit Equity
IG
• JGB 10Y
• Credit HY Credit •
Duration
• Gilt
Carry
FX Carry
• • Rates Carry
Quality •
uncorrelated risk sources.
CmdtyValue
Carry Momentum FX Value
• Equity Value • Cmdty Curve • Rates Value
• FX Momentum • Equity Momentum • Cmdty Momentum
• Rates Momentum • FX Quality • Equity Quality
• Equity Defensive • Rates Quality
Portfolio weights, %
A more useful route is an approach similar to PCA, 100
but one that defines diversification in a way more 90
closely tied to our asset classes and factors. In that
regard, Deguest, Meucci and Santangelo (2015) 80
suggest a more appropriate alternative to decompose
70
the investment opportunity set. In particular, their
methodology can be used to de-correlate the original 60
factors at minimum tracking error. As a consequence,
50
the corresponding portfolio is more stable than that
derived from a PCA, and better positioned to harvest 40
factor premia when and if they occur; see Bernardi,
Leippold and Lohre (2017) for an application in the 30
realm of commodity investing. 20

Applying this methodology to the case of multi-asset 10

multi-factor investing, we demonstrate that the 0
findings indeed translate to a stable and investable 2/06 2/08 2/10 2/12 2/14 2/16
portfolio. Back testing the DRP allocation from
The upper chart decomposes the systematic volatility of the
2006—2016, we maintain a balanced exposure SP500
diversified risk parity strategyEuroSTOXX50 FTSE100
by relevant asset classes and
relative to the set of seven asset classes and factors. factors.Nikkei225
The lower chart plots MSCI_EM
the single factor andUS10Y
asset
Bund
allocation JGB10Y
of the diversified risk parity strategy. Gilt
Slightly deviating from the unconstrained DRP Credit.IG Credit.HY FX.Carry
Sources: Bloomberg, Invesco, Goldman Sachs. Data as at
portfolio we introduce long-only constraints to only Cmdty.Carry
31 December 2016. Rates.Carry FX.Value
consider positive asset and factor weights. Judging Equity.Value Cmdty.Value Rates.Value
FX.Momentum Equity.Momentum Cmdty.Momentum
by the overall risk decomposition along the 3 asset Rates.Momentum FX.Quality Equity.Quality
classes and 4 factors, this constraint does not really Equity.Defensive Rates.Quality
result in a considerable decrease in diversification (equities and bonds) do not dominate the risk
(see upper chart in figure 5). While, by design, the budget. On average, the DRP portfolio has a 8.5%
unconstrained portfolio maintains 7.0 bets position in equities and 17.6% in government bonds,
throughout time, the constrained DRP portfolio which corresponds to an average portfolio duration
averages 6.76 bets over time (table 1). of approximately 1.5 years.

The lower chart in figure 5 illustrates the corresponding
portfolio weights when the strategy’s portfolio
construction is based on a rolling 60 months
estimation window. All in all, one observes quite a
stable allocation, yet one that sometimes actively
responds to changes in risk structure. It is also
noteworthy that the traditional asset classes
Table 1 presents performance statistics for the DRP
portfolio based on either rolling (Panel A) or expanding
window (Panel B) estimation. Over the whole sample
period, the DRP (long-only) returned 4.20% at a
volatility of 1.96% – corresponding to a Sharpe ratio
of 1.57. In contrast, note that the classic 60/40
strategy that invests 60% in the underlying equity

Risk & Reward, #3/2017 30

Table 1
Benchmarking the diversified risk parity strategy

Method DRP DRP 1/N MVP ERC 60/40

optimal long-only
Panel A: Return p.a. 4.63% 4.20% 4.20% 3.29% 3.76% 4.45%
Rolling window
(60 months) Volatility p.a. 2.22% 1.96% 3.39% 1.32% 1.36% 8.96%
Sharpe ratio 1.58 1.57 0.92 1.66 1.95 0.41
Maximum drawdown -2.00% -2.60% -7.70% -1.15% -1.49% -33.94%
# bets 7.00 6.76 3.05 5.78 6.62 2.45

Panel B: Return p.a. 4.31% 4.39% 4.20% 3.21% 3.77% 4.45%

Expanding
window Volatility p.a. 2.28% 2.17% 3.39% 1.27% 1.39% 8.96%
Sharpe ratio 1.40 1.50 0.92 1.66 1.91 0.41
Maximum drawdown -3.58% -3.75% -7.70% -1.32% -1.49% -33.94%
# bets 7.00 6.93 2.82 5.44 6.71 2.28
The table gives performance figures for the two diversified risk parity strategies from the perspective of an USD investor: the optimal DRP and the constrained DRP
with long-only positions. The latter constraints do hardly come at the cost of less return or less diversification. For comparison, the table gives 3 standard risk allocation
techniques, such as 1/N (equal weight), minimum-variance (MVP) and standard risk parity (ERC, equal contribution to portfolio risk by any single factor). In addition,
we give the results for a 60/40 benchmark that allocates 60% to the underlying equity markets and 40% to the underlying bond markets.
Sources: Bloomberg, Invesco, Goldman Sachs. Period: 31 January 2006 to 31 December 2016.

markets and 40% in the underlying bond markets,
has a much lower risk-adjusted performance (0.41),
even though the return is similar to the one of the
DRP portfolio. Given the highly asymmetric risk profile
of the 60/40 strategy one would have suffered a
hefty drawdown over the course of the financial
market crisis. For further benchmarking of the DRP
results, we ran alternative risk-based allocation
strategies such as 1/N (equal weight), minimum-
variance (MVP) and standard risk parity (ERC, any
single factor contributes equally to portfolio risk).
Interestingly, the simplest allocation technique, 1/N,
which simply applies equal weight to all single factors,
has a fairly high return as well. However, its highly
diversified weight allocation does not necessarily
translate into a diversified risk allocation, rendering
1/N the most risky strategy with a maximum drawdown

Figure 6
Performance comparison and decomposition over time
Equal weight • Equity • Duration • Credit
Minimum-variance (MVP) • Carry • Value • Momentum
Standard risk parity (ERC) • Quality • Idiosyncratic
Diversified risk parity (DRP) Performance
Performance Cumulative performance in %
1.6 60

50
1.5

40
1.4

30
1.3
20

1.2
10

1.1
0

1.0 -10 1/06 1/08 1/10 1/12 1/14 1/16

1/06 1/08 1/10 1/12 1/14 1/16

The left chart shows the performance of the long-only DRP strategy from the perspective of an USD investor. In comparison, the
performance of alternative risk-based allocation techniques, such as 1/N (equal weight), minimum-variance (MVP) and standard risk
parity (ERC, equal contribution to portfolio risk by any single factor). The right chart decomposes the cumulative performance of the
long-only DRP strategy in terms of the underlying aggregate asset classes and Equity.Negative
style factors fromDuration.Negative Credit.Negative
the perspective of an USD investor.
equal
Sources: mvp Invesco,
Bloomberg, erc drpGoldman Sachs. Data as at 31 December 2016; Carry.Negative
31 January 2006Value.Negative
= 100. Momentum.Negative
Quality.Negative Idiosyncratic.Negative Equity
Duration Credit Carry
Value Momentum Quality
Idiosyncratic Performance

Risk & Reward, #3/2017 31

of -7.70%. Conversely, minimum-variance strategies
often exhibit concentrated weight and risk allocations
when tilting towards low risk assets. In the present
use-case, the minimum-variance portfolio lags in
terms of return performance – figure 6, left chart
shows that the performance drag of the minimum-
variance portfolio is especially severe in the last
5 years of the sample period. This verdict does not
apply to the standard risk parity strategy. Notably,
the DRP allocation appears quite consistent, in that
its performance can be attributed to the underlying
portfolio construction, which is designed not to
minimize risk but to take risk in the most balanced
fashion. As a result, its cumulative performance
decomposition (figure 6, right chart) is also fairly
balanced in terms of underlying style factors. In
particular, its performance is not dominated by
traditional asset class premia.
Conclusion
To embrace the full potential benefits of factor
investing, investors need to not only single out
relevant factors (a non-trivial matter), but also find
ways to create a diversified portfolio thereof, which
exhibits improved outcomes. We have presented
the diversified risk parity strategy, which seeks to
accomplish precisely this along the major asset
classes and relevant factors within a single portfolio.
In many ways, the strategy represents the culmination
of advancements in understanding of asset pricing,
diversification and risk/return trade-offs. This strategy
can serve as a stand-alone solution, or it can be
easily blended into a classical multi-asset multi-factor
solution setup.
References
• Bernardi, S., M. Leippold and H. Lohre (2017):
Maximum Diversification Strategies along
Commodity Risk Factors, European Financial
Management, forthcoming.
• Deguest, R., A. Meucci, and A. Santangelo (2015):
Risk Budgeting and Diversification Based on
Optimized Uncorrelated Factors, Risk, 11, Issue 29,
70–75.
• Kan, R. and G. Zhou (2012): Tests of Mean-
Variance Spanning. Annals of Economics and
Finance 13, 139–187.
• Lohre, H., H. Opfer and G. Ország (2014):
Diversifying Risk Parity, Journal of Risk, 16,
53–79.
• Markowitz, H. M. (1952): Portfolio Selection,
Journal of Finance, 7, 77–91.
• Meucci, A. (2009): Managing Diversification, Risk,
22, 74–79.
Risk & Reward, #3/2017 32
About the authors
Alexandar Cherkezov
Portfolio Manager,
Invesco Quantitative Strategies
In his role, Alexandar Cherkezov manages multi-
asset portfolios that include the elements factor
based investing, active asset allocation and downside
risk management.
Dr. Harald Lohre
Senior Research Analyst,
Invesco Quantitative Strategies
Dr. Harald Lohre develops quantitative models to
forecast risk and return used in the management
of multi-asset strategies.
Sergey Protchenko
Senior Quantitative Research Analyst,
Invesco Quantitative Strategies
Sergey Protchenko is responsible for statistical
testing, factor analysis, coefficient derivation and
performance attribution for the quantitative models
used in managing large, mid- and small cap equity
and tactical asset allocation portfolios.
Jay Raol, Ph.D.
Senior Macro Analyst,
Invesco Fixed Income
Jay Raol is a member of the macro research team
at Invesco Fixed Income. In his role, he works on
macro-economic based models for asset allocation
and fixed income investing.
Notes
1 http://www.bis.org/publ/rpfx13.htm
2 World Bank (2014) “Purchasing Power Parities and Real Expenditures of World Economics.”
3 Structural inflation is calculated by looking at the change in the intercept from regressing
changes in realized inflation on changes in the unemployment rate on rolling 10-year basis.
3-month lags are applied to inflation and unemployment data to account for reporting
delays.
4 Obstfeld, M. and K. Rogoff (1996): Foundations of International Macroeconomics, MIT Press.
5 The return time series for the three commodity factors pertain to the Goldman Sachs indices
Commod Carry Risk Premium Basket Index RP16, Commod Curve RP09 and Commod
Momentum Risk Premium Basket Index RP17.
6 http://www.bis.org/publ/rpfx13.htm
7 Calculations are from the perspective of a USD investor; all returns are either in local
currency or USD-hedged.
8 The efficient frontier collects the risk and return combinations to be achieved in the
optimum of a standard mean-variance portfolio optimization. That is: the efficient frontier
gives the highest attainable return at a given level of risk or (vice versa) the lowest
attainable variance at a given level of return.
9 In particular, we employ statistical tests for mean-variance spanning as outlined in Kan and
Zhou (2012). These tests reject spanning for any of the sequential expansion exercises,
supporting the statistical relevance of the choice of style factor sets.
10 Note that the defensive equity factor enters the overall quality style factor, together with
equity quality, rates quality and FX quality. In a similar vein, the commodity curve factor
joins commodity carry and FX carry to constitute the aggregate carry factor. To obtain risk-
balanced aggregate asset class and factor returns, the seven components’ return time series
derive from a risk parity weighting of the underlying constituents.
“Multi-asset-multi-factor-strategies
employ the full set of tools in our
toolkit.”
Roundtable with Alexandar Cherkezov, Dr. Harald Lohre, Stephen Quance and Jay Raol, Ph.D.
Alexandar Cherkezov Dr. Harald Lohre
Portfolio Manager, Senior Research Analyst,
Invesco Quantitative Strategies Invesco Quantitative Strategies
At Invesco, investment professionals across
different investment centers and regions work
alongside one another to establish a meaningful
multi-asset multi-factor investment process.
We asked four of them to join us for a roundtable
discussion: Dr. Harald Lohre and Alexandar
Cherkezov from Invesco Quantitative Strategies
in Frankfurt, Jay Raol, Ph.D, from Invesco Fixed
Income in Atlanta and Stephen Quance, Director
of Factor Investing in Singapore.
Risk & Reward, #3/2017 33
Stephen Quance Jay Raol, Ph.D.
Director of Factor Investing Asia-Pacific Senior Macro Analyst,
Invesco Fixed Income
Risk & Reward
What is your main motivation for investigating ­
multi-asset multi-factor strategies?
Dr. Harald Lohre
At Invesco, we have a long history managing multi-
factor equity strategies, often as part of multi-asset
mandates. Thus, expanding the concept to other
asset classes seems only natural. In fact, there is a
substantial amount of academic research showing
how useful it is to extend factor investing beyond
traditional equity.
Alexandar Cherkezov
Factor investing is highly transparent, and it is
structured along the relevant drivers of risk and
return. For bonds, this is probably more important
than ever in the current environment: with bond
yields hovering at record-lows, there is no place else
to turn when it comes to delivering attractive levels
of return.
Stephen Quance
I fully agree with Alexandar. Nevertheless, let me
also stress that multi-asset multi-factor investing
should not be confined to periods of low interest
rates. Striving for maximum diversification to pursue
high risk-adjusted returns does not go out of fashion
when rates are higher. Rather, a multi-asset multi-
factor strategy is in many ways the final frontier
on the broad spectrum of factor utilization.
Risk & Reward
How does your strategy differ from other multi-asset
portfolios?
Jay Raol, Ph.D.
In our approach, securities are selected for their
expected exposure to quantitative and evidence-
based factors. At the core sits a body of research,
much of it taken from academia, about asset pricing
and market dynamics.
Stephen Quance
This approach is different from picking a security
that you think is undervalued. It is more systematic
in nature, more scalable. And, it becomes easier to
maintain broad diversification at the security level.
Factor investing should not be seen as a replacement
for traditional alpha seeking strategies; there will
always be a place in the market ecosystem for them.
But, because the mentality, process and sources of
returns come from different places, the return stream
will likely be dispersed as well.
Risk & Reward
What are the potential advantages of including
multiple asset classes in a factor strategy? In
particular, what makes you confident that factors can
work in realms other than equities?
Jay Raol, Ph.D.
Why should risk premia and behavioral biases be
limited to equity investors only? The data confirms
that the rationale underpinning equity factors also
apply to other asset classes. Three important equity to stay at the forefront.
style factors – value, quality and momentum – apply
equally to fixed income. There is a strong economic
rationale for all three, and they display statistical
characteristics similar to the equity space. Together, research agenda.
this makes us confident in building our multi-asset
factor library.
Dr. Harald Lohre
Factors often span traditional asset classes. For
instance, equities and high-yield bonds often move
together, or show positive correlations. If we identify new algorithms, like machine learning. We are also
a factor that helps capture this correlation, the
resulting portfolio might exhibit superior risk and
return characteristics compared to a standard asset
class allocation that ignores common factors.
Stephen Quance
Yes, it’s like an extra degree of diversification that is
harder to manage piecemeal. And, we all know true
diversification is a powerful element of risk control.
Risk & Reward
Diversification seems to permeate your work on
multi-asset multi-factor strategies. But, isn’t purely
risk-based allocation to assets and factors a rather
passive approach to investing?
Alexandar Cherkezov
In fact, quite the opposite is true. For one thing,
the overarching risk model requires a decent amount other words: is long-short the only way?
of active decision making. To start, we need to
define the major asset classes and factors. Then,
the underlying factor strategies themselves are
active strategies that might significantly deviate
from standard market investments.
Risk & Reward, #3/2017 34
Dr. Harald Lohre
And even though there is plenty of empirical
evidence supporting the usefulness of style factors,
there is only little evidence that these factors can
be timed ex-ante. Thus, a maximum diversification
approach is a prudent choice.
Risk & Reward
Let’s move on to a different topic. With factor
strategies gaining ever more popularity, aren’t you
afraid that some of these factors might become too
crowded?
Dr. Harald Lohre
We put a lot of effort into identifying economically
meaningful factors that have a sound basis for
adding value. We then consider the most effective
way to capture them over the long-run. For instance,
take value: we observe a value premium in the data
using a number of different definitions for value,
including book-to-market, earnings yield, cash flow
yield etc. Equally important considerations are
scalability, concentration and liquidity. With solid
rationales and thoughtful execution, we maximize
the likelihood of realizing positive premiums over
time.
Jay Raol, Ph.D.
We are constantly iterating on existing factors to
better refine their risk and return profiles. Markets
are always evolving and adapting to innovations,
including factor research. Therefore, we must
continue investing in our own research process
Risk & Reward
Speaking of research: tell us about your current
Jay Raol, Ph.D.
We have a formal process to review promising
developments that could lead to improvements in
our process. For instance, we are currently looking
into using new data sets, like satellite imagery, and
looking to include macroeconomic factors, such as
growth or inflation, which can complement style
factors. Factors are an iterative research-driven
process. Most of the work we do fails to bear fruit,
but occasionally we find new breakthroughs that
advance our capabilities.
Dr. Harald Lohre
Likewise, it is always important to research the
interaction between factors. Understanding the
drivers of a single factor might be interesting, but
the full power of the approach is unlocked by
combining investable factors that efficiently deliver
genuine return streams in a single portfolio.
Risk & Reward
Turning to instruments: can a multi-asset multi-factor
strategy only be implemented with derivatives? In
Alexandar Cherkezov
Yes – and no. With long only positions, you can
indeed build meaningful factor exposures, but they
will also contain a considerable level of traditional
market risk.
Jay Raol, Ph.D.
If you approach factor investing from a long/short
perspective, the building blocks for portfolio
construction are much less correlated, and you
will be able to exploit the full potential of the
strategy. Using derivatives is the most efficient
way to implement such a strategy.

Risk & Reward

If you were to quickly summarize the main appeal
of multi-asset multi factor investing …

Stephen Quance
The concept is clearly designed to be a strategic
holding. It is also expected to have low correlation
to any single asset class, like stocks or government
bonds. This is wonderful news when one asset class
declines sharply. Investors simply need to have
realistic expectations, for instance, if stocks take off.
The strategy shouldn’t be expected to keep up in the
short term. Slow and steady is more the game plan
here.

Dr. Harald Lohre

As multi-factor, bespoke and more complex solutions
become increasingly commonplace, a logical question
is: how far can this go? A multi-asset multi-factor
strategy utilizes all of our knowledge and experience
with factor investing, and combines it with expertise
in managing multi-asset strategies. By capitalizing
on factor exposures across traditional asset classes,
multi-asset-multi-factor strategies employ the full set
of tools in our toolkit.

Risk & Reward

Thank you all very much!

Risk & Reward, #3/2017 35

Why should investors consider credit
factors in fixed income?
Authors: Jay Raol, Senior Macro Analyst Fixed Income & Shawn Pope, Macro Quantitative Analyst Invesco Fixed Income

Introduction to factors complex than equities. While equities of one issuer

A substantial body of academic research, coupled
with a long track record of use in portfolios, has led
to a wider acceptance of factor investing within the
investment community. Most of the academic
research and practical implementation of factors
has been done in the equity asset class, where
factors have been key characteristics used to explain
equity risk and return. In over 50 years of research,
three general reasons have been given for why factors
earn excess returns. First, factors are by-products
of the collective behaviour biases of investors that
result in sub-optimal investing. Second, factors can
earn higher returns for higher risk. And third,
structural differences, such as liquidity differences
between securities, can lead to excess returns.
Often, a single factor’s risk and return encompasses
all three explanations.
are interchangeable, bonds are typically not. For
example, bonds of the same issuer can have
different maturities, liquidity, embedded optionality
and can sit in different parts of the capital structure.
Moreover, bonds have finite lives and usually
disappear from the investment universe after five
years. This added complexity is one of the reasons
that fixed income factor research has been slower
to evolve.
Second, factors help to explain the price changes of
assets. When interest rates were high, many investors
were content to earn returns from coupons, without
much thought of price appreciation. However, as
yields have fallen, factors have naturally become
viewed as more valuable in helping to generate
returns from prices, and not just coupons.

Factors should exist in all asset classes
While factor investing is quite established in equities,
there is much less academic research and a much
shorter track record when it comes to fixed income
portfolios. However, we believe the underlying
reasons for factors are not asset class-specific.
Factors simply connect investor behaviour to
investment returns. As such, there is no reason
to believe they cannot be applied to other asset
classes, such as fixed income.
Factors are only recently being harvested in fixed
income portfolios. What is the reason for this lag
in adoption? First, fixed income is inherently more
Risk premia definitions of factors provide investors
with the most certainty in terms of returns
Many investors have expressed a high degree of
uncertainty about using factors in fixed income.
We believe choosing the right factor definitions can
improve certainty and comfort around the concept
of factors. In our view, however risk premia
definitions are superior, since they are the most
likely to provide certainty of outcomes to investors.
Most importantly, by expecting higher returns for
unwanted risk, risk premia-based definitions offer
a compelling rationale for returns that fits within
an efficient market framework. As a result, they

Figure 1
Three major reasons for excess returns associated with factors

Risk premiums Behavioral rationales Market structure

For bearing additional risk Markets are inefficient due Markets may be infefficient
over the broad market e.g. to behavioral biases of because of restrictions and
an undesirable return pattern participants limitations or by the actions
of policy makers

Return for Over Liquidity

drawdown extrapolation imbalance

Source: Invesco. For illustrative purposes only.

36
should offer more confidence in their potential
risk-reward payoffs. A recent review of the literature
confirms this view. Two new studies utilizing robust
techniques to guard against data mining, confirm
that there are only a few, largely risk premia-based,
definitions that have a high likelihood of existence.1
From another angle, several authors have identified
a striking relationship whereby factor strategies
with high tail risk have higher Sharpe ratios.2
More certainty around risk is another advantage of
risk premia definitions. By pre-identifying the risks
inherent in strategies, and not mistaking them for
pure alpha, investors can better size these factors
in portfolios. For a conservative investor, risk premia
are likely to have fewer unknowns, or unidentified
risks.
Factor definitions in fixed income must be carefully
designed to facilitate their practical implementation
There are some major differences between equity
and fixed income factor investing. The spread of
electronic trading, dedicated pools of factor investors
and deeper shorting liquidity are some of the reasons
that equity and fixed income factor implementations
differ. Fixed income, generally, has higher transaction
costs, lower liquidity and lacks a deep short market,
aside from a few types of government bonds. Higher
transaction costs mean that factor returns need to
be heavily scrutinized to ensure that their returns
are positive and not just trading frictions.
In addition, lower liquidity at the bond level means
that factor definitions must be robustly designed so
that their risk and return characteristics are relatively
independent of the exact number or types of bonds
used. Often, only 60% of the bonds needed for a
factor portfolio are available for trading. There needs
to be some confidence that factor portfolios can be
formed with the available underlying liquidity in the
market. Finally, it is generally difficult to short bonds.
Therefore, practically speaking, long-only portfolios
are the principal way to gain factor exposures in
fixed income.
Figure 2
The timeline of factor research at Invesco Fixed Income
Establishment of
Macro Research Group
Development of
macro factors
Construction of forecasting
framework
2014 2015
Derivative portfolio
management expanded
Research database
completed
Source: Invesco, 1 June 2013 to 30 June 2017. For illustrative purposes only.
Fixed income investors may wish to consider
credit factors first
While we strongly believe that factors can be found
in all asset classes, for fixed income investors, we
think credit offers the best place to start factor
investing. Because corporate bonds offer a larger
cross sectional universe from which to build
portfolios than government bonds or currencies,
investors would likely be better able to form large
diversified portfolios that retain mostly factor
exposures. Second, given the long-only constraint,
we would expect credit beta exposure to be a large
driver of returns. Credit beta has one of the most
consitent Sharpe ratios among all asset classes,
and clear risk-return characteristics, which breed
confidence in the likelihood of future excess returns.
Factor-based Investing at Invesco Fixed Income
History of factors at Invesco Fixed Income (IFI)
IFI began developing our factor-based investing
framework three years ago - starting with macro
factors and expanding to credit factors. The genesis
of our work was the adoption of factors as part of
our active management process. We believe factor-
based investing will likely be the next iteration of
active investment management, and we continue
to refine our factor-based approach with the goal
of being an industry leader.
Invesco Fixed Income’s factor philosophy
IFI’s factor philosophy reflects our goal to provide
the best investment performance for our investors.
First, we think factors must have a strong
fundamental rationale, rooted in economic theory
– backtesting is not sufficient to warrant inclusion
in our portfolios. We believe that all quantitative
processes have embedded assumptions at their
core. By acknowledging this, we believe there is
less likelihood of perpetuating poor assumptions.
Second, while factors can be used for alpha
generation, beta replication and risk hedging, at IFI, 
we emphasize beta replication and risk hedging.
We think the academic literature and investment
Invesco self-
Fundamental/factor
alignment indexing launched
2016 2017
Credit factors
research starts
Currency, rates and credit
factor funds launched
37
professionals have been too focused on alpha. We Factors in action – liquidity, quality, value,
think there are many potential factors that have momentum and the multi-factor approach
been under-researched and underutilized because In credit, our research has focused on adapting key
they are more suited for beta replication or hedging, equity factor definitions to corporate bonds. While
but showed no alpha. corporate bonds have traditionally been broken up
into maturity, rating and industry buckets, we have
Third, we believe factors should represent a trade-off created a four-factor model which includes liquidity,
between risk and return by showing a regime quality, value and momentum. We briefly describe
dependency. We believe that factors that offer return those factors below. In keeping with our factor
for risk are likely to be more consistent over time. philosophy, we describe the fundamental rationale,
In addition, we prefer to identify the risks associated regime dependency of the factor and consistency
with factor strategies. We believe this allows for of performance across investment grade, high yield
more robust ex-post risk assessment by reducing and equities, which we believe indicates robustness.
the number of unknowns. Our definitions build on work in the literature,
although some key details differ.3, 4, 5 Finally, we
Fourth, we believe factors require continuous provide an example of the potential excess return
research. From definition to implementation, provided by a multi-factor credit model.6
we believe factors can always be improved. In rare
cases, risk and reward attitudes in markets can
structurally shift, causing material changes in factors’ Summary of factor risks and returns
expected risk and return profiles. As investors adopt Table 1a-b summarizes the risk and return
factor investing, we believe it will be important to characteristics of the four factors relative to the
constantly monitor and adapt factors. Bloomberg Barclays US Corporate Investment Grade
and High Yield Indices (“IG and HY indices”).
Finally, we seek factor definitions that are robust to All of the Sharpe ratios, except investment grade
security selection. In other words, we seek factors momentum, exceed those of the market weighted
that are likely to perform equally well whether they index.
represent 100% of a factor portfolio or only a portion.
By separating the performance of any one security Credit factor descriptions
from the overall factor portfolio, we are better able Liquidity
to implement factor portfolios in relatively illiquid We start with liquidity and treat it separately because
markets. We think this can facilitate the coherent it is somewhat unique to the fixed income space.
addition of security selection through careful credit “Liquidity” is the excess risk and return associated
analysis to a factor portfolio. Since our portfolios with holding illiquid bonds. In fixed income, illiquid
only need a small percentage of the available bonds are often not marked to market accurately.
securities to provide meaningful factor exposure, As a result, they tend to have a higher yield for a
our team of fundamental credit analysts can select lower beta exposure. From a backtesting perspective,
specific bonds to maximize portfolio returns. there seems to be a higher Sharpe ratio (Table 1a-b)
without any additional drawdown.

Table 1
a) Investment Grade

IG Index Liquidity Quality Value Momentum Multi-factor

Beta 1 0.82 0.48 1.17 0.67 0.63
Alpha (bps) 0 4.1 2.47 5.96 -0.09 5.02
Turnover (annual %) 19 39 57 269 295 209
Tracking Error (bps) 0 129 244 126 246 188
Sharpe Ratio 0.18 0.31 0.29 0.34 0.14 0.38
Drawdown (%) 24 22 14 24 15 14
Correlation to IG Index 1 0.96 0.89 0.98 0.81 0.93

b) High Yield
Beta 1 0.8 0.64 1.4 0.68 0.71
Alpha (bps) – 23.28 11.27 3.51 21.27 8.1
Turnover (annual %) 31.08 85 65 255.12 276.12 192
Tracking Error (bps) – 296 386 561 433 324
Sharpe Ratio 0.31 0.54 0.51 0.32 0.61 0.72
Drawdown (%) 45 38 34 51 33 33
Correlation to HY Index 1 0.96 0.96 0.95 0.9 0.97
Source: Bloomberg Barclays US Corporate Investment Grade and High Yield indices, Invesco calculations. Summary statistics are shown for investment grade and high
yield factors over the period 1 January 1994 to 31 March 2017. “bps” is basis points. The “Market Index” refers to the Bloomberg Barclays US Corporate Investment
Grade Index and Bloomberg Barclays US Corporate High Yield Index for the investment grade and high yield benchmarks, respectively. All of the statistics are in
excess returns, or duration hedged returns. Turnover is calculated as a half of the percentage of the portfolio buys and sells. The drawdown is calculated from the
highest peak to trough over the backtest period.

38
Figure 3
Credit returns under different VIX scenarios
• Investment Grade • High Yield
Active excess return (bps)
70
60
50
40
30
20
10
0 beta securities must offer a higher Sharpe ratio to
-10
-20
-30
1 2 3 4 5
Decrease Changes in the VIX Increase
Source: Bloomberg Barclays US Corporate Investment Grade and
High Yield indices, Invesco calculations. The scenerios were during
the period 1 January 1994 – 31 March, 2017. The average return
of the liquidity factor in both high yield and investment grade is
plotted for five different scenerios, or periods, of VIX changes.
Bucket one are the periods with the largest quintile of VIX
changes and represents periods when risk sentiment was the best.
Bucket five are the periods with the smallest quintile of VIX
changes and represents the peiods when risk sentiment was the
worst. The returns are plotted in excess returns, or duration
hedged returns, against the benchmark excess returns. The
benchmarks used were the Bloomberg Barclays US Corporate
Investment Grade and High Yield Indicies for the investment grade
and high yield liquidity factor, respectively. Past simulated
performance is not a guide to future returns. An investment
cannot be made into an index.
Figure 3 shows the return of the liquidity factors for
both high yield and investment grade bonds in
different risk environments. The average return of
the liquidity factor in both high yield and investment
grade is plotted for five different VIX scenarios.
Bucket one represents the periods with the largest
decreases in the VIX and represents periods when
risk sentiment was the best. Bucket five represents
the largest increases in the VIX and represents
periods when risk sentiment was the worst.
The returns are plotted in terms of excess returns
(duration-hedged returns) versus the benchmark
excess returns. The benchmarks used were the
Bloomberg Barclays US Corporate Investment
Grade and High Yield Indicies for the investment
grade and high yield liquidity factor, respectively.
Contrary to the idea of a higher “risk premium”
driving higher returns, the liquidity factor
outperformed during periods of extreme stress
(see bucket 5). Past simulated performance is not
a guide to future returns. An investment cannot be
made into an index. However, in reality the risk is
significant, in that it is extremely likely that selling
an illiquid bond during times of stress would result
in a significant loss. The scenario analysis returns
only accrue to buy-and-hold investors. Therefore,
only investors who can hold illiquid bonds through
market turmoil would be able to harvest higher
Sharpe ratios. The liquidity factor is defined by
those older vintage bonds that are small in issue
size relative to large, newly issued bonds. This
factor definition has been well researched in the
literature.7
Quality
Quality is the excess risk and return associated with
holding low-volatility, or low-beta, bonds.8 The
quality factor is a characteristic of securities that
are good stores of value during times of stress,
since they have low-volatilities. Figures 4a-c show
that the quality factor consistently outperformed
during periods of stress across the three asset
classes. It underperformed, however, during strong
rallies. Table 1 shows that the quality factor earned
risk adjusted alpha and had a higher Sharpe ratio
than the market index. Since most investors prefer
the embedded leverage in high-beta securities, low
compensate. Quality is defined as the return of
those bonds that have relatively short maturities
and low default risk as measured by their ratings.
Value
Value is the excess return obtained by holding assets
that are cheap to their intrinsic long-run estimated
prices. Since a bond’s price is a function of its default
risk, a natural definition is to look for those bonds
that are cheap relative to their intrinsic default rate.
Table 1a-b shows that the value factor earned risk
adjusted alpha and had a higher Sharpe ratio than
the market index. Figures 4a-c show that value
provided strong Sharpe ratios in compensation for
the materially larger tail risk during times of stress.
Value is defined as characteristics of those bonds
that are trading at a lower price relative to bonds
in the same industry with similar default risks and
maturities.
Momentum
Momentum is the return of past winners versus past
losers. As expected momentum produced the weakest
Sharpe ratios in investment grade (Table 1a-b),
especially using definitions most consistent with
traditional equity momentum. This is partly because
bonds can only appreciate by a limited amount ,
especially in investment grade where prices are close
to par. As a result, the time horizon and structure of
momentum are different for bonds than equities.
More speculative bonds have the strongest Sharpe
ratios using the equity-based definition.9 Our
analysis indicates that momentum profits after
transactions costs are not necessarily very positive.
However, momentum offers strong diversification
and manageable trading costs in a multi-factor
portfolio.
Comparing quality, value and momentum factors
in different risk environments
Figures 4a-c show the same five VIX scenarios for
high yield, investment grade and equities across
quality, value and momentum. There is a striking
similarity in conditional correlations, or return
patterns, across VIX scenarios for all of the factors
across the three asset classes. Quality and momentum
were positively correlated, but negatively correlated
with risk sentiment. They had the highest return
periods when risk sentiment was the lowest. Value
was negatively correlated with quality and momentum
and was positively correlated with risk sentiment.
Value tended to have its highest return periods when
the VIX was decreasing the most. We think that this
consistency is a sign that our definitions are capturing
the common behaviour of investors driving these risk
premia in all three asset classes.
39
Figure 4 Figure 5
Credit returns under different VIX scenarios Cumulative total returns of factors by asset class
Large decrease Increase a) Investment grade factors cumulative total returns
Decrease Large increase
Increase IG Index Value
Multi-factor Qualitiy
a) Investment grade VIX scenarios Momentum Liquidity
Active excess return (bps) Index
60 8.0
50 HY Index
40 6.0
30 Multifactor

20 4.0
Momentum
10
0
2.0 Value
-10
-20
0.0 Quality
-30
-40 Liquidity
-2.0
-50
1994
1996

1998

2000

2002

2004

2006

2008

2010

2012

2014

2016
Quality Value Momentum

b) High yield VIX scenarios b) High yield factors cumulative total returns
Active excess return (bps)
HY Index Value
100 Multi-factor Qualitiy
80 Momentum Liquidity
Index
60
4.0
40
3.5 IG Index
20
3.0
0 Multifactor
2.5
-20
2.0 Momentum
-40
1.5
-60 Value
1.0
-80
Quality Value Momentum 0.5 Quality

0.0
c) Equities VIX scenarios Liquidity
-0.5
Excess return (%)
1994
1996

1998

2000

2002

2004

2006

2008

2010

2012

2014

2016

2.5

2.0 Source: Bloomberg Barclays US Corporate Investment Grade and

High Yield indices, Invesco calculations. The returns are calculated
1.5 for the period 1 January 1994 – 31 March 2017. The cumulative
returns for four factors, the multi-factor portfolio and the
1.0 Bloomberg Barclays US Corporate High Yield Index (“HY Index”)
are plotted in Figure 5a. Likewise, the cumulative returns for
0.5 same factors, the multi-factor portfolio and the Bloomberg
Barclays US Coporate Investment Grade Index (“IG Index”) are
0.0 plotted in Figure 5b. Past simulated performance is not a guide to
future returns. An investment cannot be made into an index.
-0.5

-1.0

-1.5
Quality Value Momentum
Source: Source: Bloomberg Barclays US Corporate Investment
Grade and High Yield indices, Invesco calculations. The scenario
returns were calculated from 1 January 1994 – 31 March 2017.
“bps” is basis points. For the equity factor returns, “Quality” is
taken from Frazzini, Andrea and Lasse H Pedersen (2014),
“Betting Against Beta”, Journal of Financial Economics, 111,
1-25. The value factors was taken from Asness and Frazzini
(2013), “The Devis in HML’s Details,” Journal of Portfolio
Management, 29, 29-68. The momentum factor is based on Fama
and French (1996), “Multifactor Explanations of Asset Pricing
Anomalies,” Journal of Finance, 51, 55-84. The returns for the
credit factors are expressed in excess return against the Bloomberg
Barclays US Corporate High Yield Index and the Bloomberg
Barclays US Corporate Investment Grade Index. The darkest bar
represents the dates when the VIX decreased the most and,
represents, periods of very positive risk sentiment. The lightest
bar represents the dates when the VIX increased the most and
represents periods of very negative risk sentiment. Past simulated
performance is not a guide to future returns. An investment
cannot be made into an index.
Benefits of a multi-factor portfolio
If we examine the correlation of our factors to the
IG and HY indices in Table 1a-b, we can see that our
factors helped diversify portfolios while generating
higher Sharpe ratios over the period shown. However,
single factors can experience long periods of under
or outperformance. Therefore, we believe it is vital
to take a balanced, multifactor approach to ensure
consistent outperformance. For simplicity, we show
the return profile and attribution of an equally
weighted multi-factor portfolio. Table 1a-b shows
that, in both high yield and investment grade, the
multi-factor portfolio produced higher Sharpe ratios
without adding a significant amount of downside
risk. Figure 5a-b shows the cumulative returns of
the individual factors, the IG and HY indices and the
multi-factor portfolios over the period.

40
Figure 6 Factors are always evolving and require
S&P 500 implied volatility curve, pre-and post-1987 continuous research and active management
We end our discussion of factors on a word of caution
Implied Volatility
1,6

and the need for continuous research. It is very likely

that factor investing will change the landscape of
more fundamentally based investment strategies.
1,4

1,2

As more players adapt to factor-based investing,

we believe that factor definitions and their risks and
rewards must be continuously considered to ensure
1

0,8

Pre-1987
that they are appropriately used in portfolios. This is
0,6
because market attitudes toward risk and reward can
shift. A striking example of a major shift was the
US equity market crash of 1987. Prior to 1987, there
0,4

0,2
was no difference between the volatility implied in
Post-1987 a put versus a call, or the “skew” (Figure 6). This
meant that investors were indifferent between
0

Out-of-the-money In-the-money Out-of-the-money

put strikes strikes
call strikes
Source: Chicago Board Options Exchange (CBOE), 2010. For
illustrative purposes only. “The CBOE Skew Index”, illustrates the
difference in implied volatilites of options on the S&P 500 Index upside participation. This structural change resulted
before and after the stock market crash on 19 October 1987. The
permanent change in preference for downside protection after the
event caused put options, which protect against large falls in
equity prices, to trade at much higher prices than call options.
potential upside participation in the market and
downside protection. However, after the 1987 crash,
investors seemed to prefer downside protection over
in the birth of a risk premium available to those
investors willing to take unwanted downside equity
risk.

A more recent example is the recent preference for

low-beta stocks. Figure 7 shows the forward price-
Figure 7 earnings (PE) ratios of low- versus high-beta stocks.
Low-versus high-beta ratio of forward PE ratios The ratio of forward PEs is a proxy for investor
attitudes about future expected returns. A higher
Forward PE ratios
PE ratio means that a stock is considered expensive
2.0
and, therefore, likely to offer limited upside. For
Low-beta at a premium this reason low-beta stocks would be less likely to
1.5 outperform high-beta stocks. In the post-crisis
period, for example, low-beta stocks have traded
1.0 at historically rich levels relative to high-beta stocks,
as represented by their forward PE ratios.
0.5
High-beta at a premium This is an important shift, similar to the volatility
skew described above. Furthermore, both shifts
0.0
could be permanent. It is, therefore, important to
constantly re-evaluate risk premia to detect shifts
-0.5
76 80 84 88 92 96 00 04 08 12 14
in investor attitudes toward risk and return and
determine their likely staying power. We believe such
Source: Goldstein, Price and Zhao (2016), “Low-Beta Strategies: continuous research and active management are
Duck and Cover,” Empirical Research Partners, data from
1 January 1976 to 31 May 2016. The figure shows the forward necessary to ensure that investors earn the kind of
PE ratios of large capitilzation stocks of the lowest versus highest returns they expect from their factor portfolios.
quintiles of beta. A higher ratio indicates that low-beta stocks are
expensive relative to high-beta stocks. For illustrative purposes
only.

Conclusion
We believe the adoption of factors in fixed income
Summary of results and implications for the allows investors to better decide which risks and
future returns are appropriate for their portfolios. Ultimately,
For investors seeking to apply the advantages of this may lead to smarter decisions by investors and
equity factor investing to fixed income, we believe more efficient markets. However, by altering investor
our risk premia-based credit factor definitions offer behaviour, factors may also alter the risk-return
a compelling investment profile. Compared to other landscape. At IFI, we are constantly adapting our
definitions, we think risk premia definitions provide factor framework and evolving our investment
investors with more certainty around both risk and processes to stay ahead of these trends to help
return. Since factor-based investing is, necessarily, clients achieve their goals.
long-only in fixed income, we think it makes sense to
concentrate on applying credit factors on top of the
credit risk premium. At IFI, we have narrowed our
credit factors to four: liquidity, quality, value and
momentum. We believe that each factor offers
compelling diversification to benchmarks, higher
Sharpe ratios and robustness in its consistency in
risk and return across credit assets and compared
to their equity counterparts. We believe combining
these four factors in a multi-factor investment
generates a compelling portfolio.

41
Notes
1 Harvey, Liu and Zhu (2015), “… and the Cross-Section of
Expected Returns,” Working Paper; Harvey and Liu (2016),
“Luck Factors,” Working Paper
2 Hamdan, Pavlowsky, Roncalli and Zheng (2012), “A Primer on
Alternative Risk Premia,” Working Paper; Lemperiere,
Deremble, Nguyen, Seager, Potter and Bouchaud (2015), “Risk
Premia: Asymmetric Tail Risks and Excess Returns,” Working
Paper
3 Israel, Palhares and Richardson (2016), “Common Factors in
Corporate Bond and Bond Fund Returns,” Working Paper
4 Houweling and van Zundert (2014), ”Factor Investing in the
Corporate Bond Market,” Working Paper
5 Bai, Bali and Wen (2016), ”Common Risk Factors in the Cross-
Section of Corporate Bond Returns,” Working Paper
6 We constructed factor portfolios by market value weighting the
top quintile of sorted portfolios. The constituents of the
Bloomberg Barclays US Corporate Investment Grade and High
Yield Indices were used in factor construction from the period
January 1, 1994 to March 31, 2017. For the construction of
factors excluding liquidity, bonds were first screened for
liquidity by keeping only the top 60% and 30% in bond size
each month for investment grade and high yield, respectively.
7 Bao, Pan. and Wang (2011), ”,”Liquidity in Corporate Bonds,”
Journal of Finance, 66, 911-946.
8 Frazzini, Andrea and Pedersen (2014), “Betting Against Beta”,
Journal of Financial Economics, 111, 1-25
9 Lin, Wu, and Zhou (2016), “Does Momentum Exist in Bonds of
Different Ratings?” Working Paper.

42
Important information
The document is intended only for Professional Clients and Financial Advisers in Continental Europe (as defined below); for Qualified
Investors in Switzerland; for Professional Clients in Dubai, Ireland, the Isle of Man, Jersey and Guernsey, and the UK. It is not
intended for and should not be distributed to, or relied upon, by the public or retail investors. Data as at 31 July 2017, unless
otherwise stated.

Certain products mentioned are available via other affiliated entities. Not all products are available in all jurisdictions.

All articles in this publication are written, unless otherwise stated, by Invesco professionals. Where individuals or the business have
expressed opinions, they are based on current market conditions, they may differ from those of other investment professionals and are
subject to change without notice. This publication does not form part of any prospectus. This document contains general information only
and does not take into account individual objectives, taxation position or financial needs. Nor does this constitute a recommendation of the
suitability of any investment strategy for a particular investor. While great care has been taken to ensure that the information contained
herein is accurate, no responsibility can be accepted for any errors, mistakes or omissions or for any action taken in reliance thereon.
Opinions and forecasts are subject to change without notice. The value of investments and any income will fluctuate (this may partly be the
result of exchange rate fluctuations) and investors may not get back the full amount invested. Neither Invesco Ltd. nor any of its member
companies guarantee the return of capital, distribution of income or the performance of any fund or strategy. Past performance is not a
guide to future returns. This document is not an invitation to subscribe for shares in a fund nor is it to be construed as an offer to buy, hold
or sell any financial instruments. As with all investments, there are associated inherent risks. This document is by way of information only
and no financial advice. This document has been prepared only for those persons to whom Invesco has provided it. It should not be relied
upon by anyone else and you may only reproduce, circulate and use this document (or any part of it) with the consent of Invesco. Asset
manage­ment services are provided by Invesco in accordance with appropriate local legislation and regulations.

This should not be considered a recommendation to purchase any investment product. This does not constitute a recommendation of any
investment strategy for a particular investor. Investors should consult a financial professional before making any investment decisions if
they are uncertain whether an investment is suitable for them.

For the distribution of this document, Continental Europe is defined as Austria, Belgium, Denmark, Finland, France, Germany, Greece,
Ireland, Italy, Luxembourg,
The Netherlands, Norway, Spain, Sweden and Switzerland.

This publication is issued:

– in Austria by Invesco Asset Management Österreich – Zweigniederlassung der Invesco Asset Management Deutschland GmbH,
Rotenturmstraße 16—18, 1010 Vienna, Austria.
– in Belgium by Invesco Asset Management SA Belgian Branch (France), Avenue Louise 235, 1050 Bruxelles, Belgium.
– in Denmark, Finland, France, Luxembourg and Norway by Invesco Asset Management SA, 16-18 rue de Londres, 75009 Paris, France.
– in Dubai by Invesco Asset Management Limited, PO Box 506599, DIFC Precinct Building No 4, Level 3, Office 305, Dubai, United Arab
Emirates. Regulated by the Dubai Financial Services Authority.
– in Germany by Invesco Asset Management Deutschland GmbH, An der Welle 5, 60322 Frankfurt am Main, Germany.
– in Ireland by Invesco Global Asset Management DAC, Central Quay, Riverside IV, Sir John Rogerson’s Quay, Dublin 2, Ireland. Regulated
in Ireland by the Central Bank of Ireland.
– in the Isle of Man by Invesco Global Asset Management DAC, Central Quay, Riverside IV, Sir John Rogerson’s Quay, Dublin 2, Ireland.
Regulated in Ireland by the Central Bank of Ireland.
– in Italy and Greece by Invesco Asset Management SA, Sede Secondaria, Via Bocchetto 6, 20123 Milan, Italy.
– in Jersey and Guernsey by Invesco International Limited, 2nd Floor, Orviss House, 17a Queen Street, St. Helier, Jersey, JE2 4WD.
Invesco International Limited is regulated by the Jersey Financial Services Commission.
– in The Netherlands by Invesco Asset Management S.A. Dutch Branch, UN Studio Building, Parnassusweg 819, 1082 LZ, Amsterdam,
Netherlands.
– in Switzerland by Invesco Asset Management (Schweiz) AG, Talacker 34, 8001 Zurich, Switzerland.
– in Spain by Invesco Asset Management SA, Sucursal en España, C/ GOYA, 6 - 3°, 28001 Madrid, Spain.
– in Sweden by Invesco Asset Management SA, Swedish Filial, Stureplan 4c, 4th Floor 114 35 Stockholm, Sweden.
– in the UK by Invesco Asset Management Limited, Perpetual Park, Perpetual Park Drive, Henley-on-Thames, Oxfordshire, RG9 1HH, United
Kingdom. Authorised and regulated by the Financial Conduct Authority.
[CEUK 875/2017]