RESEARCH INSIGHT

CONSTRUCTING LOW
VOLATILITY
STRATEGIES
Understanding Factor Investing

Mehdi Alighanbari, Stuart Doole, Lokesh Mrig, Durga Shankar

January 2016

JANUARY 2016
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

CONTENTS Executive Summary............................................................................ 3

Introduction ....................................................................................... 4
Low Volatility Investing ...................................................................... 5
Low Volatility Strategies: Heuristic vs Optimization-based ................................ 6
Heuristic Approaches .......................................................................................... 7
Optimization-Based Approaches ........................................................................ 8
Practicalities of Optimization: Use of Constraints .............................. 9
Common Drawbacks of Unconstrained Minimum Volatility .............................. 9
Sector, Country and Style Biases ......................................................................................9
Turnover ........................................................................................................................ 12
Turnover and Path Dependency .................................................................................... 13

Performance Over Market Cycles and Economic Regimes ............... 16

Performance in Bear Markets ........................................................................... 16
Upside Potential ................................................................................................ 17
Behavior During Different Economic Regimes ................................................. 17
MSCI Minimum Volatility Index: Out of Sample Performance ......................... 18
Conclusion ....................................................................................... 20
References ....................................................................................... 21
Appendix 1: Correlation and Volatility ............................................. 22
Correlation Matters .......................................................................................... 22
A Note on Maximum Diversification................................................................. 24
Asset Allocation and Low Volatility Indexes ..................................................... 24
Appendix 2: Currency Hedging Issues .............................................. 26
Local Currency Minimum Volatility Indexes ..................................................... 27
Case Study: A Global “Local Currency” Minimum Volatility Index .................. 27

MSCI.COM | PAGE 2 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

EXECUTIVE SUMMARY
Low volatility is one of the few factors that have historically performed well in turbulent
markets. Moreover, over long periods of time, this defensive strategy has produced a
premium over the market, contravening one of the most basic theories in finance — that
one should not be rewarded with greater returns for taking less than market risk. Since the
global financial crisis hit in 2008, low volatility has garnered increased attention from
institutional investors.
Extensive research has investigated the low volatility anomaly, but the purpose of this paper
is to discuss the practicalities of implementing a low volatility strategy. A low volatility
strategy can be constructed in two key ways, using purely ranking-based (heuristic)
approaches or optimization-based methods. While purely ranking-based approaches are
simpler to understand, we find that optimization-based methods offer greater flexibility in
constructing low volatility strategies. In addition, some purely ranking-based approaches
provide unintended exposures to factors other than low volatility, which can affect the
risk/return profile significantly. Optimization strategies can have shortcomings of their own;
however, constraints can be used to fine-tune the construction methodology.
Using the MSCI World Minimum Volatility Index as an example, we demonstrate how a well-
designed approach can benefit from the advantages and flexibility of an optimization-based
methodology, while incorporating constraints that address the shortcomings of an
unconstrained optimization such as high turnover and large active and unwanted sector and
country bets.
The first step towards an effective minimum volatility index is a robust covariance matrix.
Using a factor model and a fundamental factor model in particular can help reduce the
number of parameters to be estimated and make the resulting covariance matrix more
robust.
We reviewed the behavior of the MSCI World Minimum Volatility Index during various
market regimes since its launch in 2008. The index reduced overall volatility by 30%, holding
up better than the market during downturns. Over the long term, the index outperformed
the market by 20 percentage points as the market itself gained 40%.

MSCI.COM | PAGE 3 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

INTRODUCTION
The low volatility factor, while dating to the 1970s, has experienced renewed interest since
the global financial crisis hit in 2008 as well as due to the growing adoption of factor indexes
(also known as “smart beta”).
Historically, the factor’s performance has declined less than the market during times of
crises and market downturns. When embedded in portfolios, the defensive characteristics of
the factor have tended to protect capital during turbulent markets.
In addition, an extensive body of research shows that low volatility portfolios have
outperformed the market over long periods of time; this outperformance has been
persistent across time and regions. The low volatility factor’s performance is a puzzle
because it is apparently at odds with one of the most basic principles in finance: that higher
volatility is associated with higher returns. According to the Capital Asset Pricing Model
(CAPM), one should not expect a long-term premium for taking less risk than the market as a
whole. The historical return premium has mainly been explained using behavioral finance
arguments, which we summarize in the next section.
Low volatility investing is a broad topic and a vast body of research has been dedicated to
this subject. In this paper, we intend to address the practicalities of constructing low
volatility strategies, responding to common questions that investors raise when evaluating
these strategies.
This paper is the fifth in a series exploring each of the six key factors that have historically
offered long-term excess returns: value, quality, momentum, yield, low volatility and low
size.

MSCI.COM | PAGE 4 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

LOW VOLATILITY INVESTING

While apparently contradicting one of the main principles of finance that higher risk is
associated with higher returns, several empirical studies have demonstrated that lower
volatility stocks have outperformed the market.1
Mostly behavioral arguments have been offered to explain the low volatility premium. Here
are some of the leading explanations:
 Lottery effect. Some observers argue that buying a volatile stock is similar to buying
a lottery ticket where the customer pays a small fee in the hope of winning a large
amount of money — albeit at a very low probability. Therefore, investors often
overpay for high volatility stocks and underpay for low volatility stocks due to the
“irrational” preference for volatile stocks.
 Representativeness. Investors tend to overpay for “glamorous” high volatility stocks
because of the well-publicized success of a handful of such stocks; the speculative
nature of such stocks is often ignored by investors.
 Overconfidence. Investors are overconfident in their ability to forecast the future,
and the extent of their differences in opinions is greater for stocks with more
uncertain outcomes (high volatility stocks). Also, it is tougher for pessimistic
investors to express a negative view via a short sell, resulting in optimists driving up
the prices of high volatility stocks and hence lower expected future returns for high
volatility stocks.
 Agency issue. Asset managers tend to avoid low volatility stocks because there is
less research conducted by brokers and others on these less glamorous stocks.
 Asymmetric behaviors. When the market is on a declining trend, the dispersion of
beta between low volatility and high volatility portfolios has tended to increase (i.e.,
low volatility stocks experienced a much lower beta, or risk, vis-à-vis the market).
Therefore, the low volatility stocks have experienced smaller declines than their
high volatility counterparts. When a bull market occurs, this dispersion has been
smaller and thus low volatility stocks have underperformed only slightly. Net, the
low volatility stocks have performed better over the long term.

1
Black (1972), Haugen and Baker (1991), Chan, et al. (1999), Jagannathan and Ma (2003), Clark, et al.(2006), Ang et al.
(2006), Blitz and Vliet(2007), Nielsen and Subramanian(2008), Sefton, et al. (2011), Baker and Haugen (2012), Frazzini and
Pedersen (2014), Muijsson, et al. (2014), and Stambaugh, et al. (2015).

MSCI.COM | PAGE 5 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

There are also theories that are not based on behavioral biases. For example, Baker, et al.
(2011) observed that low volatility stocks often tend to have lower betas; overweighting
them may lead to higher tracking errors for institutional investors. Such tracking errors need
to be justified by sufficient excess returns (alpha). In other words, institutional investors
cannot buy low volatility stocks wholesale. To a certain extent, the benchmark issue
prohibits institutional investors from fully exploiting the low volatility anomaly.
Separately, Frazzini and Pedersen (2014) argue that the underperformance of higher beta
assets is partly due to leverage constraints and margin requirements faced by many
investors. According to CAPM, all investors invest in the highest Sharpe ratio portfolio and
leverage or de-leverage this portfolio to meet their objectives. However, many investors are
constrained in their use of leverage; instead, they increase their risky holdings. This
increased demand for high beta assets may result in lower long-term risk-adjusted returns
than for low beta assets.
Lastly, Muijsson, et al. (2014) explain the outperformance of low beta stocks based on
interest-rate movements. Their analyses show that the low and high beta portfolios have
been affected differently by changes in the risk-free rate. Returns on low beta portfolios
have increased when the rate decreases and returns on high beta portfolios have risen when
the rate increases. They conclude that the main factor behind low volatility anomaly likely
stems from exogenous macroeconomic factors such as government monetary policies.

LOW VOLATILITY STRATEGIES: HEURISTIC VS OPTIMIZATION-BASED

Numerous methodologies have been developed over the years to implement low volatility
strategies. The more recent phenomenon of “smart beta” indexes has sparked interest in
creating investible low volatility indexes. All of these indexes and the underlying approaches
can be categorized into two distinct groups: heuristic and optimization-based.
Heuristic approaches tend to be simple, purely ranking-based indexes. In comparison,
optimization is a more sophisticated approach to creating low volatility indexes. This
sophistication may make the process more complex but it provides significant flexibility that,
if designed properly, can considerably improve the quality of the resulting index, improving
its replicability and avoiding unintended exposures to styles, countries or sectors.
Let’s start with the formula to calculate the volatility of an index:

𝜎𝑝2 = 𝑊 𝑇 . Φ . 𝑊 = ∑𝑁 2 2 𝑁 𝑁
𝑖=1 𝑤𝑖 𝜎𝑖 + ∑𝑖=1 ∑𝑗=1,𝑗<>𝑖 𝜌𝑖𝑗 𝑤𝑖 𝑤𝑗 𝜎𝑖 𝜎𝑗

where 𝜎𝑝 is the volatility of the index returns, 𝜎𝑖 is the volatility of stock (asset) 𝑖 in the
index, 𝑤𝑖 is the weight of stock 𝑖 and 𝜌𝑖𝑗 is the correlation between stock 𝑖 and stock 𝑗.

MSCI.COM | PAGE 6 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

The objective is to find weights (𝑤𝑖 ) that result in the lowest volatility for the index.2 There
are different methodologies to calculate these weights and create an optimal index portfolio
but they all have two stages. The first stage is to estimate volatilities and correlations (the
covariance matrix - Φ in the above equation). The second stage is to use these estimates to
calculate the optimal weights. Therefore, the quality of a low volatility index depends on the
accuracy of the estimations as well as how well the weights are calculated.

HEURISTIC APPROACHES
Many purely ranking-based approaches have been developed to create a low volatility
index. The underlying principle of most of these approaches is to rank the universe of stocks
based on the estimate of their volatility (total volatility, residual volatility or beta), select a
subset of (or in some cases all) the constituents of the universe, and then apply different
weighting schemes. Weighting can be determined by market capitalization, inverse of
volatility, inverse of variance or various other methodologies.
Constraints may be applied to these heuristic approaches to ensure there are acceptable
levels of liquidity and investability, controls for sector and country exposures and limits on
stock weights. The MSCI Risk Weighted Indexes and Volatility Tilt Indexes fit into this
category.
These approaches are simple and transparent and their weighting schemes enjoy a good
degree of flexibility. However, they generally are based on the volatility of individual stocks
and ignore the correlation between stock returns (the second term in the equation), which
can have a major impact on strategy volatility when cross sectional variation between
correlations is high.
Some heuristic approaches also fail to provide a pure exposure to low volatility, implicitly
providing exposure to other factors. Such approaches derive some of their risk/return
behaviors from these residual factors.
Some low volatility strategies explicitly incorporate other factors. For instance, one can sort
and select securities based on their volatility and then weight them based on company
valuations. This type of approach employs multiple factors; while there is a diversification
benefit to combining factors, the risk/return profiles of such strategies can result from their
exposure to, say, value, as much as to volatility.

2
Currency risk is an important consideration in designing low volatility strategies. We discuss the effect of currency on
these strategies in Appendix 2.

MSCI.COM | PAGE 7 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

OPTIMIZATION-BASED APPROACHES
While heuristic approaches reflect the volatility of individual stocks, optimization-based
approaches account for both volatility and correlation effects.3 Optimization can be
performed in various ways, but the differences usually stem from the covariance matrix
estimation (Φ in the equation) and how constraints are applied.
The simplest approach to obtain the covariance matrix is to use the historical returns of
individual stocks and calculate their historical volatilities and pairwise correlations.4 There
are two main issues with this approach: As the number of stocks in the universe increases,
the size of covariance matrix and therefore the number of parameters to be estimated
becomes very large, sometimes requiring estimation of millions of parameters. Also, stock
volatility and the correlation between them can be very unstable; using historical levels may
provide poor estimates of future values (Vangelisti, 1992).
A more common approach to optimization, especially for a large universe of stocks, is using
a factor model, such as a simple statistical model that applies principal component analysis
or a more elaborate fundamental factor model. These models effectively reduce the size of
covariance matrix to be estimated, making calculations less complex and more stable: The
size of covariance matrix remains constant for a fixed number of factors and does not
change even if the number of stocks in the universe varies.
In addition, a fundamental factor model such as a simple 5-factor Carhart or commercial
models take advantage of economic intuition to measure realistic and stable correlations
across the investment universe. Fundamental factor models tend to use current stock
characteristics, resulting in a timelier, stable and robust covariance matrix. The MSCI
Minimum Volatility Indexes, which were introduced in 2008, currently use the Barra GEM2
factor model.5

3
Please see Appendix 1 for a more detailed discussion regarding the effect of correlation and volatility on low volatility
indexes.
4
Correlation between returns of every pair of stocks in the selected universe.
5
MSCI Minimum Volatility Indexes were launched using a previous version of the Barra equity model (GEM). With the
introduction of the more advanced GEM2 Model, these indexes adopted the new model in 2009.

MSCI.COM | PAGE 8 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

PRACTICALITIES OF OPTIMIZATION: USE OF CONSTRAINTS

A naively designed optimization can result in unwanted and extreme exposure to certain
industries, countries or styles. In addition, a poorly designed optimization can result in high
turnover at rebalancing periods. Adding constraints to the optimization, however, can
mitigate these shortcomings without compromising the effectiveness of the optimization.
The design of the optimization and how constraints are incorporated are critical;
manipulating the results by applying constraints after the optimization can undermine the
entire exercise and make the resulting index suboptimal. For instance, optimizing the index
without a sector constraint and later adjusting constituent weightings to impose sector
constraints may impair the quality of the index. Similarly, running an unconstrained
optimization might create an optimal long-short portfolio but the individual long and short
legs can be far from optimal. Thus, constraints should be built into the optimization and the
results should remain untouched.
Selecting constituents prior to optimization can be sensible. For instance, if we want to
exclude certain stocks in the strategy, we can simply remove them before the optimization.

COMMON DRAWBACKS OF UNCONSTRAINED MINIMUM VOLATILITY

The three major issues with an unconstrained minimum volatility portfolio are: 1) biases
towards certain sectors and countries, 2) unwanted style exposures and 3) high turnovers at
rebalancing. In what follows we look at these potential shortcomings and demonstrate how
they can be eradicated using well-designed constraints. We also mention some other ways
to deal with these issues that can be less effective or in cases detrimental to the
optimization process.

SECTOR, COUNTRY AND STYLE BIASES

Stocks from different sectors and countries exhibit varying levels of volatility. For instance,
utility companies tend to be low volatility stocks while information technology stocks tend
to be more volatile. An unconstrained minimum volatility portfolio would often have a
positive and persistent bias towards low volatility sectors and negative exposure to higher
volatility ones. Similarly, some country markets tend to be less volatile while others are
more volatile. Though sometimes these biases are desirable, in many cases investors prefer
to limit active exposures to sectors or countries.
An effective optimization framework incorporates constraints to limit unintended and
unwanted biases. For example, the MSCI Minimum Volatility Indexes uses the Barra Open
Optimizer, which allows for incorporating a wide range of constraints, to calculate the index.

MSCI.COM | PAGE 9 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

Exhibit 1 and Exhibit 26 compare the sector and country exposures of a constrained
minimum volatility index to those for an unconstrained index. We have used the MSCI World
Index as of November 25, 2014 as the universe and we have constrained the optimization to
be long only. The only additional constraint applied in this example is to keep sector and
country exposures of the minimum volatility portfolio within 5 percentage points of the
parent index weightings.
Without these constraints, we would see large biases in consumer staples and utilities as
well as in Japanese and Hong Kong equities. Adding these constraints creates an optimized
index while avoiding large and unwanted bets on any sector or country.

Exhibit 1: Active Sector Exposures Constrained vs. Unconstrained Min Vol Strategies
40%

20%

0%

-20%
Health Care Inf. Tech Industrials Materials Energy Cons. Stpls. Cons. Disc. Financials Utilities Telecom.
Unconstrained With Active Sector Exposure limited to 5%

Data as of November 25, 2014

Exhibit 2: Active Country Exposures Constrained vs. Unconstrained Min Vol Strategies
20%

10%

0%

-10%
Japan Hong Kong USA UK Israel Canada France Australia Germany Singapore

Unconstrained With Active Country Exposure limited to 5%

Data as of November 25, 2014

Style factors such as value, leverage and size are important in designing a strategy.
Sometimes, exposures to these factors are intended to capture historical long-term premia
and are explicitly part of the design and optimization process. But often these exposures
emerge unknowingly and randomly. When creating a strategy, whether through an
optimization-based or heuristic approach, the factor exposures will deviate from the market.

6
Only countries with significant exposure have been included.

MSCI.COM | PAGE 10 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

For instance, a low volatility portfolio may result in unintentional low size and high value
exposures.
These unintended active style exposures may be tolerated to a certain extent, as a long-only
strategy often includes some level of residual active style exposure. But with the right tool
and design, these unintended style exposures can be restricted to specified levels.
Fundamental-based models such as Barra factor models (used in conjunction with the Barra
Open Optimizer) explicitly allow limits on unwanted style exposures.
In Exhibit 3, we contrast the effect of an optimized minimum volatility index with no style
constraints with one that limits exposure to style factors. In the unconstrained index, while
the large negative exposure to volatility is intended, the large active residual exposures to
growth and size are accidental and may be unwanted. In the MSCI Minimum Volatility Index,
we constrain all the style factors excluding volatility to within 0.25 standard deviation of
their parent index. The constraints keep all the styles within range while having minimal
impact on the desired volatility exposure.
The constraint on the value factor in the MSCI Minimum Volatility Indexes also implicitly
prevents the index from over-weighting richly valued companies (crowded stocks). While
low volatility stocks tend to be high quality stocks and show higher valuations, these
constraints ensure that the resulting index limits exposure to over-valued stocks compared
to the relevant equity market or the parent index.

Exhibit 3: Active Style Exposures – Constrained vs. Unconstrained Min Vol Strategies
0.50
0.25
0.00
-0.25
-0.50
-0.75
-1.00
Financial Growth Liquidity Momentum Size Size Value Volatility
Leverage Nonlinearity

No Constraints With all Active Styles except Volatility Constrained to be within 0.25 of Benchmark

Adding any constraint to the minimum volatility optimization results in an index with greater
expected volatility, but for a well-designed optimization the increase is minimal. Exhibit 4
illustrates the considerable reduction in expected volatility achieved by moving from the
market cap index (the MSCI World Index) to a minimum volatility index, with varying levels
of constraints. It also reveals that a small increase in expected volatility occurs when
different constraints are added to the unconstrained optimization.

MSCI.COM | PAGE 11 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

Note: Applying the constraints after the optimization may cause a considerable deviation
from optimality and in many cases a feasible solution may not be possible.

Exhibit 4: Expected Volatility for Different Levels of Constraints

MSCI World
With Sector, Factor and Country Constraints
With Country and Factor Constraints
With Sector and Factor Constraints
With Sector Constraints only
With Country Constraints only
With Factor Constraints only
No Constraint Global Optimum

0% 5% 10% 15%
Expected Volatility

TURNOVER
A poorly designed minimum volatility index may also experience high turnover. One can
reduce the rebalancing frequency to limit turnover to a certain extent while still maintaining
the desired level of minimum volatility exposure. To further reduce the turnover to the
desired level, we can explicitly apply turnover constraints to the optimization.
There is a clear trade-off between the level of turnover and the reduction in expected
volatility. However, this relationship is not linear. The marginal benefit of incurring
additional turnover to reduce volatility decreases as turnover increases.
To illustrate this point, first we examine the effect of going from a market cap index to a
minimum volatility index. We run optimizations on the MSCI World Index while changing the
constraint on turnover (illustrated by the blue line in Exhibit 5). Allowing 50% turnover
relative to the market cap index would have reduced the volatility to 9.6 % from 13.7%. In
comparison, the index with no turnover constraint would have achieved volatility of 9.1%
with 76% turnover. This means 90% of possible risk reduction would have been achieved by
allowing 50% turnover relative to the MSCI World Index.
More important is to examine turnover of the minimum volatility index at the rebalancing
dates, when stocks are added to and subtracted from the index. In this example, starting
from the MSCI Minimum Volatility Index (before rebalancing), the volatility level of 9.6% can
be achieved by allowing only 10% turnover, also resulting in a 90% of possible risk reduction
compared to the parent index. The yellow line illustrates the risk reduction that is achieved
by different turnover constraints when we start from a minimum volatility index just before
rebalancing.

MSCI.COM | PAGE 12 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

For the MSCI Minimum Volatility Indexes, we have chosen to rebalance semi-annually and
constrain the turnover to 10% on each rebalancing, resulting in 20% annual one-way
turnover.

Exhibit 5: Effect of Turnover Constraint on Expected Volatility

14%
MSCI World
13%
MSCI Min Vol before Optimal without Turnover
12% Rebalancing Constraints
Risk

11%

10%

9%

8%
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Turnover
From MSCI World From MSCI World Minimum Volatility

Data as of November 25, 2014

TURNOVER AND PATH DEPENDENCY

While a constraint is necessary to avoid excessive turnover and associated costs, it can
create path dependency — that is, the index constituents and their weighting can depend on
the index’s launch date. This can become problematic when the index deviates significantly
from the optimal index. Careful design can mitigate sub-optimality due to path dependency.
To see if path dependency has affected the MSCI Minimum Volatility Indexes,
Exhibit 6 and Exhibit 7 compare ex-post risk/return behavior of indexes that are constructed
in exactly the same way, except for their starting dates. The figures show considerable ex-
post risk reduction and increases in the returns of minimum volatility indexes compared to
the market cap index. Moreover, the differences in annualized volatility between the five
different minimum volatility indexes are negligible and there is no evidence of better
performance for the newer indexes. Thus, the respective start dates have not affected either
riskiness or returns.

MSCI.COM | PAGE 13 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

We also looked at the difference between the MSCI Minimum Volatility Index and its
turnover-relaxed counterpart over time. We look at two measurements, ex-ante risk spread
and active share.7
Exhibit 8 shows the ex-ante risk for the MSCI World Index as well as two versions of the
minimum volatility index, with and without turnover constraints. The risk for the two
versions of minimum volatility indexes is almost identical; it is hard to see the difference. On
the right axis, we measured this small difference (shaded line). Not only is the difference
small (10-30 basis points), there does not seem to be any obvious upward trend that might
indicate the turnover constraint effect is accumulating over time.
In Exhibit 9, we used active share to show how different the two indexes (with and without
turnover constraints) are and whether they diverged over time. As we expected, the
turnover constraint resulted in differences in the two indexes. While there was an increase
in active share initially, it seems that this parameter stabilized, meaning that the constrained
index does not vary much from the optimal approach over time.
An unconstrained optimized index is created by considering only the main objective —
reducing its volatility — ignoring other important considerations such as capacity and
concentration, liquidity, turnover and unintended exposures. However, alternative indexes
can achieve very similar levels of volatility reduction while also accounting for these other
considerations. Through use of constraints, the optimization process can create an
alternative index that not only achieves the main objective of reducing volatility but also
addresses these other investment considerations.

Exhibit 6: Analyzing Path Dependency: Ex-Post Volatility

20%
Annualized Volatility
16%

12%

8%

4%

0%
Since May 2002 Since May 2006 Since May 2010 Since May 2014
MSCI World Index MV - Starting May 1998 MV - Starting May 2002
MV - Starting May 2006 MV - Starting May 2010 MV - Starting May 2014

7
Active share measures how two indexes (or portfolios) differ from each other by comparing the weight of each stock in
the indexes. It is calculated as half of the sum of the absolute difference between the weights of each stock in the two
indexes.

MSCI.COM | PAGE 14 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

Exhibit 7: Analyzing Path Dependency: Ex-Post Return

15%
Annualized Performance
10%

5%

0%
Since May 2002 Since May 2006 Since May 2010 Since May 2014
MSCI World Index MV - Starting May 1998 MV - Starting May 2002
MV - Starting May 2006 MV - Starting May 2010 MV - Starting May 2014

Exhibit 8: Analyzing Path Dependency: Ex-Ante Volatility

30 1

Increase in Risk level due to

25 0.8

turnover constraint
20
Risk level

0.6
15
0.4
10
5 0.2

0 0

Effect of TO constraint (rhs) MSCI Min Vol Min Vol w/o TO Constraint MSCI World

Exhibit 9: Analyzing Path Dependency: Active Share

40%
Active share vs. No TO constraint

30%

20%
index

10%

0%
Dec-09 Jun-10 Dec-10 Jun-11 Dec-11 Jun-12 Dec-12 Jun-13 Dec-13 Jun-14 Dec-14 Jun-15
Standard (20% Annual TO) 4% Annual TO

MSCI.COM | PAGE 15 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

PERFORMANCE OVER MARKET CYCLES AND ECONOMIC REGIMES

While risk is often measured by the level of volatility, there are other measures of risk that
are important, such as drawdowns during bear markets and market turmoil. In this section,
we present several analyses to show how the MSCI Minimum Volatility Index has historically
performed during different market regimes and during rising and falling markets.

PERFORMANCE IN BEAR MARKETS

We start by analyzing the performance of MSCI World Minimum Volatility Index during
different market downturns (bear markets) over the past 27 years8 (Alighanbari, et al. 2014).
We define bear market as a decline of 20% or more in the MSCI World Index for a period
lasting at least two months. There were four bear markets during this time frame (gray
shaded areas in Exhibit 10). The MSCI World Minimum Volatility Index (blue line)
outperformed the market (the MSCI World Index) across all four bear market periods,
demonstrating its strong defensive characteristics.

Exhibit 10: Minimum Volatility Behavior in Bear Markets

Relative Performance of MSCI World Min Vol Index in Bear Markets
150 800

140 700

600
130
500
120
400
110
300
100
200
90 100

80 0

Bear Markets MSCI World MinVol/MSCI World (lhs) MSCI World (rhs)

Exhibit 11 shows that the MSCI World Minimum Volatility Index experienced substantially
smaller drawdowns during these market downturns when protecting wealth was most

8
MSCI Minimum Volatility historical data starts at May 31, 1988. Please refer to the disclaimers at the end of this
document regarding use of simulated or backtested data.

MSCI.COM | PAGE 16 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

important. In addition, the Minimum Volatility Index also produced lower realized volatility
during the turbulent periods.

Exhibit 11: MSCI Minimum Volatility Index Risk/Return over Bear Markets
Absolute Returns (Gross USD) Realized Volatility
MSCI World MSCI World
Bear Market Periods MSCI World Minimum Volatility Active Return MSCI World Minimum Volatility
Dec 89 - Sep 90 -24.0% -20.2% 3.8% 21.8% 19.7%
Mar 00 - Sep 02 -46.3% -19.8% 26.5% 16.5% 11.0%
Oct 07 - Feb 09 -53.7% -43.0% 10.6% 21.9% 17.1%
Apr 11 - Sep 11 -19.4% -5.1% 14.3% 15.9% 8.9%

UPSIDE POTENTIAL
Not surprisingly, the MSCI World Minimum Volatility Index has outperformed the market
when the market has declined overall. In Exhibit 12, we see that the index outperformed a
negative market 88% of the time with an average outperformance of 8.8 percentage points
based on one-year rolling periods. Where the index underperformed the market, the
average shortfall was only 1.24 percentage points.
In the years where market return exceeded 10%, the Minimum Volatility Index
underperformed; the level of underperformance increased as the market return rose.
However, for moderate positive return periods (0%-10%), the index outperformed the
market 67% of the time.
Exhibit 12: Performance Comparison: MSCI World Minimum Volatility Index vs Market
MSCI World Rolling 1-Year Return <0 0-10% 10%-20% 20%-30% >30%
Hit rate of Outperformance 88.6% 67.3% 39.8% 21.2% 0.0%
Average Outperformance 8.8% 4.5% 3.1% 1.3% 0.0%
Average Underperformance -1.24% -2.6% -3.8% -6.0% -9.3%
No. of Observations 79 52 118 52 16

BEHAVIOR DURING DIFFERENT ECONOMIC REGIMES

The minimum volatility factor has performed well in defensive markets. Here we examine its
behavior against various market/macro signals, extending previous MSCI research on how
various equity factors behaved in changing economic environments (Gupta et al., 2014).
Exhibit 13 extends the bivariate regime analysis introduced by Gupta et al. to include market
risk indicators such as VIX and credit spreads. The blue bars in this exhibit show the average
monthly active return of the MSCI World Minimum Volatility Index when the indicated
market/macro measure is decreasing, while the yellow bars indicate when the measure is
rising. On average, the Index outperformed the market during periods of economic

MSCI.COM | PAGE 17 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

contraction (defined as a decrease in the OECD’s Composite Leading Indicators), high

volatility (increasing VIX), widening of credit spread and rising inflation, as well as falling
interest rates. The pattern of outperformance re-emphasizes the defensive behavior of low
volatility strategies.
Exhibit 13: MSCI World Minimum Volatility Behavior during Market/Macro Regimes
Average Monthly Active Returns
1.0%

0.5%

0.0%

-0.5%

-1.0%
OECD CLI VIX Credit Spread Inflation Interest Rates

Decreasing Increasing

MSCI MINIMUM VOLATILITY INDEX: OUT OF SAMPLE PERFORMANCE

Since the MSCI World Minimum Volatility Index family was launched in 2008,9 it has
returned about 60% while the market has gained about 40% (Exhibit 14). The index’s
volatility over this out-of-sample period (calculated using monthly returns) has been 12.8%
compared to 18% for the parent index, a roughly 30% reduction.
In Exhibit 15, the shaded area shows the performance of the MSCI World Index over this
period and the yellow line the performance of the Minimum Volatility Index relative to the
MSCI World Index. The MSCI Minimum Volatility Index was launched as the market plunged
about 50% in 2008; the index outperformed the market by about 20 percentage points.
Similarly, in 2011, the market declined about 20% and the Minimum Volatility Index avoided
most of that drawdown. More recently, the Minimum Volatility Index held up well during
market turbulence in August 2015.
The MSCI World Minimum Volatility Index has exhibited considerably lower volatility than
the broad market index since its launch. It has also demonstrated strong defensive
characteristics, significantly outperforming the market during market downturns. Finally, it
has outperformed the market over the seven-year period, providing a long-term premium.
This outperformance was achieved over a period where the market itself gained 40%.

9
MSCI World Minimum Volatility (USD) was launched on April 14, 2008

MSCI.COM | PAGE 18 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

Exhibit 14: MSCI World Minimum Volatility Index Performance since Launch
180
160
140
120
100
80
60
40
Apr-08 Apr-09 Apr-10 Apr-11 Apr-12 Apr-13 Apr-14 Apr-15
World Min Vol MSCI World

Exhibit 15: Defensive Behavior of MSCI World Minimum Volatility Index

160 130

140 120

110
120
100
100
90
80
80
60 70

40 60
Apr-08 Apr-09 Apr-10 Apr-11 Apr-12 Apr-13 Apr-14 Apr-15
MSCI World Min Vol/World(rhs)

MSCI.COM | PAGE 19 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

CONCLUSION
The minimum volatility factor is one of the few factors that have performed well during
turbulent markets, providing capital preservation when it is needed most. Yet it remains an
anomaly as it has produced better-than-market returns over long time periods while
offering less risk.
In this paper, we looked at some of the practicalities of designing a low volatility strategy.
There are two main ways to design these strategies: heuristic (purely ranking-based) and
optimization-based approaches. While heuristic approaches tend to be simpler,
optimization-based approaches provide a more flexible framework to incorporate different
types of constraints. Moreover, only optimization-based approaches can take full advantage
of the correlation between stocks, a key component in designing a low volatility strategy.
Low volatility indexes, whether created using a purely ranking-based approach or by
optimization, can result in large unintended tilt towards other style factors. While combining
multiple factors can be a sensible approach for diversification purposes, sometimes these
residual factors can have more effect on the risk/return profile of the strategy than the
volatility factor itself.
Optimization-based approaches have their pitfalls. Estimating the full covariance matrix can
be cumbersome as the number of stocks increases. Use of a factor model (especially a
fundamental factor model), can help reduce the number of parameters to be estimated and
make the resulting covariance matrix more stable.
Constraints also are important in designing a minimum volatility index. Implementing
constraints directly in the optimization can help achieve the desired level of sector, country
and style exposures and limit turnover without compromising much in risk reduction.
Applying constraints after optimization can defeat the whole purpose of the optimization
and can result in an inferior index.
Finally, we examined the behavior of MSCI World Minimum Volatility Index during different
market regimes to show the different characteristics of the index. Using 27 years of available
data as well as the seven years of history since the index has been live, the index has
produced considerably lower volatility than the market, has behaved defensively in market
downturns and has outperformed the market during these periods. Over the long term, this
index has generated superior performance, benefiting from the low volatility premium.

MSCI.COM | PAGE 20 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

REFERENCES
M. Alighanbari,, R.A. Subramanian and P. Kulkarni. (2014). "Factor Indexes in Perspective: Insights from 40
Years of Data." MSCI Research Insight.

A. Ang, R. Hodrick, Y. Xing and X. Zhang. (2006). “The Cross-Section of Volatility and Expected Returns.”
Journal of Finance 61(1), 259-299.

M. Baker, B. Bradley and J. Wurgler. (2011). “Benchmarks as Limits to Arbitrage: Understanding the Low-
Volatility Anomaly.” Financial Analysts Journal 67(1), 40-54.

N.L. Baker and R.A. Haugen. (2012). “Low Risk Stocks Outperform within All Observable Markets of the
World.” available at SSRN: http://ssrn.com/abstract=2055431.

F. Black. (1972). “Capital Market Equilibrium with Restricted Borrowing.” Journal of Business 45(3), 444-455.

D. Blitz and P. Van Vliet. (2007). “The Volatility Effect: Lower Risk without Lower Return.” Journal of
Portfolio Management 34(1), 102-113.

L.K.C. Chan, J. Karceski and J. Lakonishok. (1999). “On Portfolio Optimization: Forecasting Covariances and
Choosing the Risk Model.” The Review of Financial Studies 12(5), 937-974.

Y. Choueifaty and Y. Coignard. (2008). “Towards Maximum Diversification.” Journal of Portfolio

Management 34(4), 40-51.

R. Clarke, H. De Silva and S. Thorley. (2006). “Minimum-Variance Portfolios in the US Equity Market.”
Journal of Portfolio Management 33, 10-24.

A. Frazzini and L.H. Pedersen. (2014). "Betting Against Beta." Journal of Financial Economics 111(1), 1-25.

A. Gupta, A. Kassam, R. Suryanarayanan and K. Varga. (2014). “Index Performance in Changing Economic
Environments.” MSCI Research Insight.

R. Haugen and N. Baker. (1991). “The Efficient Market Inefficiency of Capitalization-Weighted Stock
Portfolios.” Journal of Portfolio Management (Spring 1991), 35-40.

R. Jagannathan and T. Ma. (2003). “Risk Reduction in Large Portfolios: Why Imposing the Wrong Constraints
Helps.” Journal of Finance 58(4), 1651-1684.

C. Muijsson, E. Fishwick and S. Satchell. (2014). “Taking the Art out of Smart Beta.”
http://sydney.edu.au/business/__data/assets/pdf_file/0003/214356/DP-2014-
008_Taking_The_Art_out_of_Smart_Beta.pdf. Discussion paper.

F. Nielsen and R.A. Subramanian. (2008). “Far From the Madding Crowd – Volatility Efficient indexes.” MSCI
Barra Research Insights.

J. Sefton, D. Jessop, G. De Rossi, C. Jones and H. Zhang. (2011). “Low-risk investing,” UBS Research.

R.F. Stambaugh, J. Yu and Y. Yuan. (2015). “Arbitrage Asymmetry and the Idiosyncratic Volatility Puzzle.”
Journal of Finance 70(5), 1903-1948.

M. Vangelisti. (1992). “Minimum-Variance Strategies: Do They Work?” Barra Newsletter.

MSCI.COM | PAGE 21 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

APPENDIX 1: CORRELATION AND VOLATILITY

CORRELATION MATTERS
An optimization-based strategy relies on both correlation and volatility to create the
minimum volatility index. In this section, we look at the effect of correlation on the overall
volatility reduction of the minimum volatility index.
While we know the reduction of volatility (minimum volatility index compared to the market
cap index) comes from both selecting lower volatility stocks and selecting lowly correlated
stocks, it is not easy to separate the two effects. Below, we create a proxy for the effect of
selecting lower volatility stocks versus the correlation reduction.
For these analyses we use the MSCI World and MSCI World Minimum Volatility indexes as of
November 26, 2014. The risk levels for stocks as well as indexes are taken from the Barra
GEM2 model.
Exhibit 16 demonstrates the distribution of volatility of the stocks in each index. The MSCI
World had a relatively symmetric distribution around 25% volatility (with a slight tail for the
higher volatilities). The MSCI World Minimum Volatility Index, as expected, picked more of
the lower volatility stocks and therefore is skewed towards the left side of the graph. (Note:
We are ignoring the weight of each stock in the index; this graph shows only the percentage
of the stocks that are in each volatility segment.)
Exhibit 16: Volatility Distribution: MSCI World Minimum Volatility Index vs Market Cap
30%

25%

20%
Frequency

15%

10%

5%

0%
14% 20% 26% 32% 38% 44% 50% 56%
World Minimum Volatility World Volatility

Clearly, some of the reduction that we see in the expected volatility of the MSCI World
Minimum Volatility Index comes from selecting lower volatility stocks. This can be confirmed
by looking at the average and weighted average of the constituents’ volatility in the MSCI
World Minimum Volatility Index versus its parent market cap index (Exhibit 17).

MSCI.COM | PAGE 22 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

Exhibit 17: Volatility Reduction of MSCI World Minimum Volatility Index of MSCI Worlndex
MSCI World MSCI World Minimum Volatility
Volatility Estimate 13.66% 9.62%
Average Volatility of Constituents 24.69% 20.36%
Wt. Avg Volatility of Constituents 22.16% 18.83%

But the question remains how much of the volatility reduction comes from selecting lowly
correlated stocks. To answer this question, let’s try separating the correlation effect. To do
this, we need to make several assumptions and approximations.
For each index we have:
σ2p = ∑ wi2 σ2i + ∑i<>𝑗 ρij wi wj σi σj
Now we assume that the correlation between all the stocks in the portfolio is equal to an
average correlation:
∀i, j; ρij = ρP
With this assumption, we can then calculate the average correlation level as:
σ2p − ∑ wi2 σ2i
ρp =
∑i<>𝑗 wi wj σi σj

Applying this formula to the MSCI World Index and the MSCI Minimum Volatility Index, we
find: ρ𝑊𝑜𝑟𝑙𝑑 = 0.38
ρMin Vol = 0.26
The numbers clearly show that the MSCI Minimum Volatility Index is benefiting from a lower
correlation between stocks selected.
To calculate a proxy for the effect of correlation on volatility reduction, we insert the
correlation of the MSCI World Index (ρWorld ) into the above equation for the MSCI World
Minimum Volatility Index:

σ2Min Vol = ∑ wi2 σ2i + ∑ ρWorld wi wj σi σj = (11.6%)2

̂
i<>𝑗

This means that if the correlation had stayed the same, the effect of selecting (and
overweighting) lower volatility stocks in the minimum volatility index would have been a
reduction in volatility from 13.7% to 11.6%. In other words, we can argue that from the 4.0%
reduction of the volatility (Exhibit 17), approximately 2% comes from selecting lower

MSCI.COM | PAGE 23 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

volatility stocks and 2% comes from the correlation effect. While a number of assumptions
are used in this calculation, the results demonstrate that selecting lower volatility stocks and
giving them higher weights than the parent index as well as selecting stocks with lower
correlations contributes to the reduction in volatility of the index.

A NOTE ON MAXIMUM DIVERSIFICATION

Discussing the maximum diversification approach to portfolio construction is out of the
scope of this report. In this section, we briefly discuss the basic difference between the
maximum diversification and minimum volatility approaches. The “Most Diversified
Portfolio” is defined (Choueifaty and Coignard 2008) as the portfolio with the maximum
diversification ratio, defined as:
𝑊′∅
𝐷(𝑃) =
√W′ΦW
Where W is the vector of weights of individual assets in the portfolio, Φ is the covariance
matrix and ∅ is the vector of asset volatilities. It is clear from the formula that the
denominator is the volatility of the portfolio and the numerator is the weighted sum of the
volatilities of the portfolio constituents. If we remove the numerator, the problem becomes
a minimum variance problem. Including the numerator, maximization would try to minimize
the volatility of the portfolio (the denominator) while selecting and overweighting higher
volatility stocks. By doing so, the approach tries to achieve a reduction of portfolio volatility
through selecting more volatile but lowly correlated assets. This process achieves the
objective of having a highly diversified portfolio (assets with low correlation to each other)
but at the same time targets high volatility stocks. In a sense, this approach is in line with
CAPM theory and contrasts with the low volatility anomaly.
Comparing a maximum diversification portfolio with a purely ranking-based low volatility
portfolio and minimum volatility portfolio, we find different treatments of volatility and
correlations. In the purely ranking-based portfolio, the emphasis is on selecting and
overweighting low volatility stocks while ignoring the correlation effect on reducing the
overall volatility of the portfolio. This approach contrasts with maximum diversification,
which selects high volatility stocks while trying to achieve portfolio volatility reduction
through the correlation effect. The minimum volatility approach fits somewhere in between:
It benefits from the correlation effect on portfolio volatility reduction, but at the same time
implicitly selects and overweights lower volatility stocks.

ASSET ALLOCATION AND LOW VOLATILITY INDEXES

Even ignoring the long-term premium that has been achieved by low volatility indexes, their
lower volatility puts them somewhere between equities and fixed income. Therefore, low

MSCI.COM | PAGE 24 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

volatility indexes can be used to construct a separate asset class in the allocation process.
Incorporating low volatility indexes could have been especially helpful in times of low rates
where investors often struggled to generate targeted rates of returns from their bond
portfolios.
In Exhibit 18, we look at the effect of incorporating low volatility indexes into the asset
allocation process. Replacing a market cap-weighted equity allocation with a low volatility
index-based investment enhanced the return while reducing overall portfolio risk during the
study period. Substituting a 60% allocation to market-cap-weighted equity with a low
volatility index-based portfolio in a 60%/40% equity/fixed-income portfolio reduced risk by
roughly 20% while enhancing return by an annualized 60 basis points during the study
period.
In addition, incorporating the use of low volatility indexes in the asset mix in this way would
have allowed investors to increase their allocation to equities without increasing risk. An
80%/20% allocation mix to low volatility equity and fixed income resulted in similar risk
levels to those seen in the traditional 60%/40% allocation but with a higher return due to
the premia from equities in general as well as from low volatility stocks.

Exhibit 18: Effect of Using Low Volatility Index Portfolios in Asset Allocation
Equity Allocation Fixed Income Portfolio Portfolio Risk
Tota l Equi ty MSCI World Min Vol Allocation* Return** Risk** Reduction***
60% 60% 0% 40% 6.69% 9.92%
60% 40% 20% 40% 6.92% 9.11% 8.1%
60% 20% 40% 40% 7.12% 8.44% 14.9%
60% 0% 60% 40% 7.30% 7.94% 19.92%
80% 0% 80% 20% 7.63% 9.60% 3.18%
*Barclays Capital Global Aggregate
** Statistics determined over the period from Jan 1990 - Sep 2015
*** Compared to 60% MSCI World / 40% Fixed Income allocation

MSCI.COM | PAGE 25 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

APPENDIX 2: CURRENCY HEDGING ISSUES

Currency treatment is an important consideration when it comes to low volatility indexes,
regardless of whether constructed through optimization or heuristically. When designing a
low volatility portfolio, it is important to specify the investor’s local (reference) currency.
Let’s suppose an investor wants to create a one-stock portfolio with the lowest historical
volatility of daily returns from a universe of two Japanese stocks. A Japanese investor can
simply pick the stock with lower historical volatility calculated using yen returns. For a U.S.
investor, however, the answer may be different. The historical volatilities now should be
based on the performance of each stock in USD. Depending on the correlation between the
currency movement and stock movements, the lower volatility stock may be different for
the Japanese investor than for the U.S. investor.
Similar considerations apply for the creation of low volatility indexes. For the MSCI Minimum
Volatility Indexes, the base currency is defined in the equity risk model and within the
optimization; different indexes can be generated that are ex-ante optimal for different
investor reference currencies.
The volatility of an equity index depends on two things: 1) the volatility and correlation of
the underlying stocks and 2) the behavior of the currencies of those securities and how the
currency returns have related to stock returns. This equity-currency relationship is captured
by the chosen global Barra Equity Model and hence in the optimization process where
overall (forecast) index volatility is minimized. It is a little more complicated for a minimum
volatility portfolio based on an international market that uses a different currency (for
example, a U.S.-based investor looking at an MSCI Japan Minimum Volatility Index).
The choice of currency affects the composition of the low volatility index because the index
will have some home bias for the investor as it reflects stock selections by the optimizer to
reduce overall index risk. Is there a “correct” currency to use as a base for this optimization
process? The theoretical answer depends on whether the index will be currency-hedged.
Suppose we have a Japanese investor interested in a minimum volatility index based on
MSCI Japan. The investor and index currencies are identical. Yen is the base currency and
currency plays no role. Effectively, we ignore the currency-related segments of the global
Barra factor model (by setting currency exposures to zero). For a U.S. investor, the optimal
(lowest forecast volatility) index is achieved with a base currency of the U.S. dollar. The sign
of the correlation between stock returns and currency returns helps drive stock selection
and weighting. To pick between two Japanese stocks with the same local currency volatility,
we look for the one with the most negative (or least positive) correlation to the currency
pair.

MSCI.COM | PAGE 26 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

If the U.S. investor intends to hedge her equity currency exposure, then she would want to
see the same returns in U.S. dollars as a Japanese investor sees in yen. Hence, the
optimization currency is yen (local currency) and the USD-JPY FX hedge is applied to the
resulting minimum volatility index.

LOCAL CURRENCY MINIMUM VOLATILITY INDEXES

When a currency exposure is “perfectly hedged” in an international portfolio, the investor
sees (in her own local currency) the same return as a local investor in those same securities
in their local currency. If the parent index has multiple currencies, the idea is the same for
each currency in turn. The impact of the currency-related covariance in the risk model
should be zero: in practice, we can set all the currency elements of the exposure matrix to
zero. The optimizer will then to build an MSCI “local currency” minimum volatility index: the
same for all investor perspectives.
Formally, of course, a minimum volatility index with any choice of base currency can be
hedged to a target (investor) currency, but in the medium term we will be effectively
(economically) over- or under-hedging with this “mixed” approach (so named as we are
mixing hedging of currency risk via optimization with currency risk management with FX
forwards). Local currency indexes offer a “cleaner” base for investors employing a hedging
strategy with strong currency views. The view might be directional on the currency, or
directional with respect to currency volatility, or even driven by non-market considerations.
For example, the latter view might be driven by an accounting-based aversion to seeing
currency returns influence profits in an investment portfolio subject to gain/loss constraints.
The local currency approach may also be convenient as the benchmark for active portfolio
managers who run a single low volatility strategy portfolio but have investors or sub-funds
with different FX hedging requirements. Other investors may stick with a mixed approach
for operational simplicity (existing AUM, ease of establishing a new fund or share classes)
after considering carefully the tracking error and hedge differences in different market
conditions. The choice of a reserve currency outside the parent index’s currency region as
the base currency is sometimes a surprisingly close proxy to the local currency solution (in a
way that a commodity currency chosen as base currency is unlikely to be). Ultimately, the
hedging strategy will depend on the investment horizon of the asset owner and her
tolerance of currency-related impacts and capacity to directly hedge currency risk.

CASE STUDY: A GLOBAL “LOCAL CURRENCY” MINIMUM VOLATILITY INDEX

We now consider a USD-denominated investor looking for an MSCI ACWI Minimum Volatility
index. We look at two cases: first, an index which takes account of all sources of risk; and
second, the local currency version, which does not incorporate estimates of currency

MSCI.COM | PAGE 27 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

volatilities and correlations. In relation to risk minimization, the former approach does a
better job of medium-term risk reduction for unhedged returns. However, we have already
noted that the investor use cases for local currency construction place greater weight on
other investment considerations.
We looked at the influence of the base currency on index characteristics over a full (12-year)
simulation period and then, given the rise in FX volatility, over the five years to the end of
2014. For factor exposures, there are only minor differences between the approaches.
The sector weights of the local currency minimum volatility index relative to those in the
standard index are more revealing. On average, in both simulations, the local currency index
is overweight materials and the relative position in technology shows wide variation. In
Exhibit 19, we see the “home bias” that favored U.S. information technology because of
currency risk reduction, even if the sector is not associated with a low volatility strategy. The
local currency version almost always has a maximum underweight in technology.

Exhibit 19: Active Sector Exposures over Five-Year Simulation

With a different parent index, the average sector difference can be more marked. When we
compare an MSCI Japan Minimum Volatility index in yen to one optimized in U.S. dollars, the
latter has a much higher average weight to consumer discretionary stocks. This overweight
is plausible, given the sector’s sensitivity to dollar-yen currency correlations. Moreover, the
level of sector differences can be dwarfed by stock-level differences in holdings: One-way
turnover between minimum volatility indexes optimized in different currencies can be 20%-
30%.
Relative country and regional exposures showed wider variation than sectors for both
simulation periods. Without the impact of currency volatility and correlations, the country
positioning for the local currency index is more stable, with active country constraints clearly
binding and the “home bias” reversed.

MSCI.COM | PAGE 28 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

Exhibit 20: Active Country Exposures – Over Five Years

We also aggregated the percentage country weights into currency blocks and reviewed the
time-series of exposures in turn for a range of reserve, developed and emerging market
currencies (Exhibit 21). We see the home bias in the U.S. dollar exposure when the dollar is
the base currency and the increase in yen exposure as the negative currency-equity
correlation increases. Exposures for the euro (and sterling) are, however, similar. The
commodity currencies generally see more stable currency positions in the local currency
minimum volatility index.

Exhibit 21: Active Currency Exposures

Simulation period: 11/29/2002 to 12/31/2014

The key performance characteristics of the minimum volatility indexes are those linked to
the risk-driven index objectives, e.g., how do total risk, beta and tail-risk properties such as

MSCI.COM | PAGE 29 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

maximum drawdown change? The difference in sector, country and factor exposures (and
for the unhedged index, the currency weights) already indicates some of the exposure to tail
events. We also looked at the slippage between the returns on the two indexes (hedged and
unhedged) because short-run drawdowns versus a benchmark or an alternative strategy can
lead to investor regret. In Exhibit 22, we show the relative risk reduction view for U.S. dollar
returns, while in Exhibit 23, we show the comparison for hedged returns (using the local
currency return as a proxy for a “perfect hedge” overlay). To aid comparability, we show in
each bar the percentage reduction in that risk measure compared to the parent ACWI index.

Exhibit 22: Percentage Risk Reduction in Simulation for Alternative MV Indexes

(Left-hand chart: 12-Year simulation; Right-hand chart: 5-year simulation)

Exhibit 23: Percentage Risk Reduction in Simulation for Alternative Hedged MV Indexes

(Left-hand chart: 12-year simulation; Right-hand chart: 5-year simulation)

The performance difference coming from the varying exposures of a local currency approach
can be episodic. Recently, as FX volatility has risen, the divergence has generally been at its
greatest – the full-period (unhedged) tracking error of 1.4% is low compared with the most

MSCI.COM | PAGE 30 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

recent period and is a weak guide to performance dispersion (see Exhibit 24). While for
unhedged returns the USD-optimized minimum volatility index offered the greater risk
reduction, once hedged indexes are considered, it is the local currency optimized version
with the hedge overlay that has shown greater relative risk reduction (especially over the
last five years). But such differences are not so apparent in every parent index minimum
volatilty simulation.

Exhibit 24: Tracking Risk and Return Characteristics for ACWI Local Currency MV Index

ACWI Min Vol ACWI Min Vol ACWI Min Vol ACWI Min Vol
(Loc Ccy), 5y (Loc Ccy) (LC, hedged) 5y (LC, hedged)

Tracking Error (%) 1.8 1.4 1.5 1.2

Ann. act. retn (simulated) -1.0 -0.2 -0.5 -0.1

Historical Beta 1.07 1.05 0.9 0.95

Max Active Drawdown (%) 6.7 6.7 3.7 4.0

Active Drawdown (mths) 44 44 38 69

MSCI.COM | PAGE 31 OF 33
© 2016 MSCI Inc. All rights reserved. Please refer to the disclaimer at the end of this document.
 RESEARCH INSIGHT

CONTACT US AMERICAS ABOUT MSCI

Americas 1 888 588 4567 * For more than 40 years, MSCI’s research-
clientservice@msci.com Atlanta + 1 404 551 3212 based indexes and analytics have helped
Boston + 1 617 532 0920 the world’s leading investors build and
Chicago + 1 312 675 0545 manage better portfolios. Clients rely on
Monterrey + 52 81 1253 4020 our offerings for deeper insights into the
New York + 1 212 804 3901 drivers of performance and risk in their
San Francisco + 1 415 836 8800 portfolios, broad asset class coverage and
Sao Paulo + 55 11 3706 1360 innovative research.
Toronto + 1 416 628 1007
Our line of products and services includes
indexes, analytical models, data, real estate
benchmarks and ESG research.
EUROPE, MIDDLE EAST & AFRICA
MSCI serves 98 of the top 100 largest
Cape Town + 27 21 673 0100 money managers, according to the most
Frankfurt + 49 69 133 859 00 recent P&I ranking.
Geneva + 41 22 817 9777 For more information, visit us at
London + 44 20 7618 2222 www.msci.com.
Milan + 39 02 5849 0415
Paris 0800 91 59 17 *

ASIA PACIFIC

China North 10800 852 1032 *

China South 10800 152 1032 *
Hong Kong + 852 2844 9333
Mumbai + 91 22 6784 9160
Seoul 00798 8521 3392 *
Singapore 800 852 3749 *
Sydney + 61 2 9033 9333
Taipei 008 0112 7513 *
Thailand 0018 0015 6207 7181*
Tokyo + 81 3 5290 1555

* = toll free

JANUARY 2016
 CONSTRUCTING LOW VOLATILITY STRATEGIES | JANUARY 2016

NOTICE AND This document and all of the information contained in it, including without limitation all text, data, graphs, charts (collectively, the “Information”) is
the property of MSCI Inc. or its subsidiaries (collectively, “MSCI”), or MSCI’s licensors, direct or indirect suppliers or any third party involved in making
or compiling any Information (collectively, with MSCI, the “Information Providers”) and is provided for informational purposes only. The Information
DISCLAIMER may not be modified, reverse-engineered, reproduced or redisseminated in whole or in part without prior written permission from MSCI.

The Information may not be used to create derivative works or to verify or correct other data or information. For example (but without limitation),
the Information may not be used to create indexes, databases, risk models, analytics, software, or in connection with the issuing, offering,
sponsoring, managing or marketing of any securities, portfolios, financial products or other investment vehicles utilizing or based on, linked to,
tracking or otherwise derived from the Information or any other MSCI data, information, products or services.

The user of the Information assumes the entire risk of any use it may make or permit to be made of the Information. NONE OF THE INFORMATION
PROVIDERS MAKES ANY EXPRESS OR IMPLIED WARRANTIES OR REPRESENTATIONS WITH RESPECT TO THE INFORMATION (OR THE RESULTS TO BE
OBTAINED BY THE USE THEREOF), AND TO THE MAXIMUM EXTENT PERMITTED BY APPLICABLE LAW, EACH INFORMATION PROVIDER EXPRESSLY
DISCLAIMS ALL IMPLIED WARRANTIES (INCLUDING, WITHOUT LIMITATION, ANY IMPLIED WARRANTIES OF ORIGINALITY, ACCURACY, TIMELINESS,
NON-INFRINGEMENT, COMPLETENESS, MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE) WITH RESPECT TO ANY OF THE
INFORMATION.

Without limiting any of the foregoing and to the maximum extent permitted by applicable law, in no event shall any Information Provider have any
liability regarding any of the Information for any direct, indirect, special, punitive, consequential (including lost profits) or any other damages even if
notified of the possibility of such damages. The foregoing shall not exclude or limit any liability that may not by applicable law be excluded or limited,
including without limitation (as applicable), any liability for death or personal injury to the extent that such injury results from the negligence or
willful default of itself, its servants, agents or sub-contractors.

Information containing any historical information, data or analysis should not be taken as an indication or guarantee of any future performance,
analysis, forecast or prediction. Past performance does not guarantee future results.

The Information should not be relied on and is not a substitute for the skill, judgment and experience of the user, its management, employees,
advisors and/or clients when making investment and other business decisions. All Information is impersonal and not tailored to the needs of any
person, entity or group of persons.

None of the Information constitutes an offer to sell (or a solicitation of an offer to buy), any security, financial product or other investment vehicle or
any trading strategy.

It is not possible to invest directly in an index. Exposure to an asset class or trading strategy or other category represented by an index is only
available through third party investable instruments (if any) based on that index. MSCI does not issue, sponsor, endorse, market, offer, review or
otherwise express any opinion regarding any fund, ETF, derivative or other security, investment, financial product or trading strategy that is based on,
linked to or seeks to provide an investment return related to the performance of any MSCI index (collectively, “Index Linked Investments”). MSCI
makes no assurance that any Index Linked Investments will accurately track index performance or provide positive investment returns. MSCI Inc. is
not an investment adviser or fiduciary and MSCI makes no representation regarding the advisability of investing in any Index Linked Investments.

Index returns do not represent the results of actual trading of investible assets/securities. MSCI maintains and calculates indexes, but does not
manage actual assets. Index returns do not reflect payment of any sales charges or fees an investor may pay to purchase the securities underlying the
index or Index Linked Investments. The imposition of these fees and charges would cause the performance of an Index Linked Investment to be
different than the MSCI index performance.

The Information may contain back tested data. Back-tested performance is not actual performance, but is hypothetical. There are frequently
material differences between back tested performance results and actual results subsequently achieved by any investment strategy.

Constituents of MSCI equity indexes are listed companies, which are included in or excluded from the indexes according to the application of the
relevant index methodologies. Accordingly, constituents in MSCI equity indexes may include MSCI Inc., clients of MSCI or suppliers to MSCI. Inclusion
of a security within an MSCI index is not a recommendation by MSCI to buy, sell, or hold such security, nor is it considered to be investment advice.

Data and information produced by various affiliates of MSCI Inc., including MSCI ESG Research Inc. and Barra LLC, may be used in calculating certain
MSCI indexes. More information can be found in the relevant index methodologies on www.msci.com.

MSCI receives compensation in connection with licensing its indexes to third parties. MSCI Inc.’s revenue includes fees based on assets in Index
Linked Investments. Information can be found in MSCI Inc.’s company filings on the Investor Relations section of www.msci.com.

MSCI ESG Research Inc. is a Registered Investment Adviser under the Investment Advisers Act of 1940 and a subsidiary of MSCI Inc. Except with
respect to any applicable products or services from MSCI ESG Research, neither MSCI nor any of its products or services recommends, endorses,
approves or otherwise expresses any opinion regarding any issuer, securities, financial products or instruments or trading strategies and MSCI’s
products or services are not intended to constitute investment advice or a recommendation to make (or refrain from making) any kind of investment
decision and may not be relied on as such. Issuers mentioned or included in any MSCI ESG Research materials may include MSCI Inc., clients of MSCI
or suppliers to MSCI, and may also purchase research or other products or services from MSCI ESG Research. MSCI ESG Research materials, including
materials utilized in any MSCI ESG Indexes or other products, have not been submitted to, nor received approval from, the United States Securities
and Exchange Commission or any other regulatory body.

Any use of or access to products, services or information of MSCI requires a license from MSCI. MSCI, Barra, RiskMetrics, IPD, FEA, InvestorForce, and
other MSCI brands and product names are the trademarks, service marks, or registered trademarks of MSCI or its subsidiaries in the United States
and other jurisdictions. The Global Industry Classification Standard (GICS) was developed by and is the exclusive property of MSCI and Standard &
Poor’s. “Global Industry Classification Standard (GICS)” is a service mark of MSCI and Standard & Poor’s.

MSCI.COM | PAGE 33 OF 33
© 2015 MSCI Inc. All rights reserved.