Diversification and Portfolio Theory: A Review: Gilles Boevi Koumou

Financial Markets and Portfolio Management (2020) 34:267–312
https://doi.org/10.1007/s11408-020-00352-6
Diversification and portfolio theory: a review
Gilles Boevi Koumou1,2
Published online: 4 June 2020

© Swiss Society for Financial Market Research 2020
Abstract
Diversification is one of the major components of investment decision-making under
risk or uncertainty. However, paradoxically, as the 2007–2009 financial crisis revealed,
the concept remains misunderstood. Our goal in writing this paper is to correct this
issue by reviewing the concept in portfolio theory. The core of our review focuses on the
following diversification principles: law of large numbers, correlation, capital asset
pricing model and risk contribution or risk parity diversification principles. These four
diversification principles are the DNA of the existing portfolio selection rules and asset
pricing theories and are instrumental to the understanding of diversification in portfolio
theory. We review their definition. We also review their optimality, with respect to
expected utility theory, and their usefulness. Finally, we explore their measurement.
Keywords Diversification · Portfolio theory · Law of large numbers · Correlation ·

Capital asset pricing model · Risk contribution · Risk parity · Asset pricing theory ·
Expected utility theory
JEL Classification G1 · G11 · G12
1 Introduction
The 2007–2009 financial crisis has raised a large number of questions about the capa-
bility of diversification to protect well against loss. Critics (see Fabozzi et al. 2014;
Holton 2009, among others) argue that diversification failed to adequately protect
against loss during the 2007–2009 financial crisis, because (Pearson) correlations
tend to peak during bear markets. For example, Thomas Kieselstein, CIO and man-
B Gilles Boevi Koumou

[email protected]; [email protected]
1 Canada Research Chair in Risk Management, Department of Finance, HEC Montréal, 3000,
chemin de la Côte-Sainte-Catherine, Montreal, QC H3T 2A7, Canada
2 Department of Economics and Administration Sciences, Université du Québec à Chicoutimi, 555,
boulevard de l’Université, Chicoutimi, QC G7H 2B1, Canada
123
268 G. B. Koumou
aging partner at Quoniam, a quantitative asset management firm based in Frankfurt,

according to Fabozzi et al. (2014: pp. 28–35), remarks,
The financial crisis has clearly shown that when you need diversification most, it
may not work. Historical correlations may simply be wrong. Different liquidity
of different asset classes may mean that some less risky assets may still be
punished because they are tradable. We need better management of such extreme
situations.
Robert Brown, Ph.D., CFA, with Genworth Financial Asset Management in Encino,
California, according to Holton (2009: pp. 21–22), remarks,
It doesn’t work and it doesn’t really have much of any connection to the real
world...I began to watch it closely starting in the 1980s, and with the passage
of each year I’ve appreciated more and more that it’s really nothing more than
a marvelous academic concept that has been heavily popularized within the
financial planning community. It’s nothing more than an abstraction from reality
that doesn’t have a lot of client-based relevance.
For the proponents (see Ilmanen and Kizer 2012; Statman 2013; Miccolis and
Goodman 2012; Stanton 2015, among others), diversification did not fail during the
2007–2009 financial crisis. Rather, it was misunderstood. For example, Ilmanen and
Kizer (2012: pp. 15) argue that,
Yet, diversification has come under attack after the 2007–2009 financial cri-
sis, when diversification seemed to fail as virtually all long-only asset classes,
other than high-quality sovereign debt, moved in the same direction (down). We
argue that the attacks are undeserved. Instead, we believe that the problem is
“user error”; most investors were never as diversified as they thought they were.
There is ample room for improvement by shifting the focus from asset class
diversification to factor diversification.
On the basis of the works of Statman and Scheid (2005, 2008), Statman (2013: pp.
11) points out the misperception of (Pearson) correlations in terms of diversification
benefits.
Even those who believe that Markowitz is wrong continue to use the language
of mean-variance portfolio theory. We have gained Markowitz’s mathematical
formulation of diversification and adopted its language of correlations, but in the
process many have lost the intuition underlying diversification’s benefits.
Statman and Scheid [2008, 2013] noted that, for two reasons, correlation is
not a good measure. First, we tend to misperceive correlations. Relatively high
correlations bring relatively low diversification benefits, but even correlations
greater than 0.90 provide substantial diversification benefits.
Second, the benefits of diversification depend not only on correlations between
investment returns but also on the standard deviations of each investment’s
returns. Return gaps are better measures of diversification’s benefits than are
correlations, because return gaps account for the effects of both correlations and
123
Diversification and portfolio theory: a review 269
standard deviations, while also providing intuitive measures of diversification’s

benefits.
Moreover, Statman and Scheid found that, although correlations tend to be higher
in bear markets than in bull markets, the benefits of diversification are actually
higher in bear markets than in bull markets. This is because standard deviations
also tend to be higher in bear markets than in bull markets, and higher standard
deviations in bear markets add to the benefits of diversification more than higher
correlations subtract from these benefits.
From our viewpoint, diversification is not dead, but it is misunderstood. It is there-

fore a central issue to better understand the concept in portfolio theory, which is our
purpose in this review.
Diversification is one of the major components of investment decision-making
under risk and uncertainty, even long before the birth of portfolio theory. Mention
of it can be found in the Babylonian Talmud: Tractate Baba Mezia, folio 42a,1 in
Shakespeare and Phelps (c1923, Act I, Scene 1) (see also Markowitz 1999; Rubinstein
2002), and in the studies by Bernoulli (1738/1954) and Leavens (1945, pp. 473).
Roughly, portfolio diversification consists of allocating wealth across a variety of
assets. Its benefit is risk reduction minimizing both the probability and severity of
portfolio loss, through a multilateral insurance in which each asset is insured by the
remaining assets.
In their seminal works published, respectively, in the Journal of Finance and Econo-
metrica, Markowitz (1952) and Roy (1952) offered the first mathematical formulation
of diversification in investment decision-making under risk. This marks the birth of
portfolio theory. Since then, several portfolio selection rules (see among others Alexan-
der and Baptista 2002; Cumova and Nawrocki 2014; De Giorgi 2002; Elton et al.
1976; Fishburn 1977; Holthausen 1981; Kang et al. 1996; Konno and Yamazaki 1991;
Markowitz 1959; Rockafellar and Uryasev 2000; Rockafellar et al. 2006, 2007; Shalit
and Yitzhaki 1984) and theories of asset pricing (see among others Ross 1976; Sharpe
1964) based on diversification were proposed.
However, paradoxically, reviews of the literature on portfolio diversification are
rare. Since Markowitz (1952) and Roy (1952), to our knowledge, only three stud-
ies have exclusively devoted to review the literature on portfolio diversification: two
unpublished working papers (Flint et al. 2015; Fragkiskos 2013) and one book (Lhabi-
tant 2017). Fragkiskos (2013) limits his study to asset diversification and offers a
non-technical and solely descriptive review focusing on the list of the most used port-
folio diversification strategies and measures. See Table 1 for more details.
Soon after, based on the most commonly used portfolio diversification measures,
Flint et al. (2015) provide a taxonomy of diversification measures. Their classi-
fication distinguishes five categories of diversification measures: cardinality-based
measures, weight-based measures, return-based measures, risk-based measures and
higher-moment measures (see Table 1). The authors provide a non-technical descrip-
tion of each measure and compare their performance using real data.
1 http://juchre.org/talmud/babametzia/babametzia.htm.
123
270 G. B. Koumou
Most recently, Lhabitant (2017) provides a more detailed and technical review of
portfolio diversification (see Table 1). His book also covers issues on portfolio diver-
sification, including the household diversification puzzle, the diversification versus
concentration dilemma and the interpreting of the (Pearson) correlation.
Short but interesting reviews on portfolio diversification strategies or measures can
also be found in other studies. For example, Deguest et al. (2013) classify portfolio
diversification measures into two categories: weight-based measures and risk-based
measures (see Table 1). Carli et al. (2014) adopt this classification for their review of
portfolio diversification measures (see Table 1). Sharma (2018) distinguishes factor-
based measures from risk-based measures in his review of portfolio diversification
measures. He thus adopts the following classification: weight-based measures, risk-
based measures and factor-based measures (see Table 1).
Instead of a highly selective review as in the studies by Fragkiskos (2013) and Flint
et al. (2015), or an exhaustive review as Lhabitant (2017) did, this review proposes
another route, that of focusing on the following diversification principles: law of large
numbers, correlation, capital asset pricing model, risk contribution or risk parity
diversification principles. These four diversification principles are the DNA of the
existing portfolio selection rules and asset pricing theories and are instrumental to
the understanding of diversification in portfolio theory. The law of large numbers
diversification principle, which is the combination of portfolio size-based and weight-
based diversification principles, is at the core of asset pricing theories; it is also used
as asset allocation strategy. The correlation diversification principle, which exploits
the interdependence between assets, is at the core of expected utility theory and the
mean-variance model, among others. The capital asset pricing model diversification
principle is a sophisticated combination of the market portfolio diversification (at
the level of individual asset) and the correlation diversification (at the level of asset
class). The risk contribution or risk parity diversification principle is at the core of risk
budgeting approaches.
We review the definition of each diversification principle. We also review their
optimality with respect to expected utility theory and their usefulness for a risk averse
investor. Finally, we explore their measurement.
Our analysis is limited in two respects. First, we focus only on asset diversification.
We do not discuss the time diversification. We refer the readers to the study by Lhabitant
(2017, Chapter 6) for time diversification. Second, we focus on theoretical aspects of
portfolio diversification. We do not cover empirical analysis.
This review is structured as follows. We start by introducing the notation and def-
initions required for our review (Sect. 2). We then review our four diversification
principles (Sects. 3 to 6). Section 7 concludes the paper.
123
Table 1 List of diversification strategies and class of diversification measures covered by the existing
reviews
Fragkiskos (2013) Information theory Herfindahl index (Woerheide and Persson

1993)
Time-varying (Pearson) correlation Butler and Joaquin (2002) and Chua et al.
(2009)
Fund of hedge funds Denvir and Hutson (2006)
Factor diversification Bender et al. (2010) and Page and Taborsky
(2011)
Tail measures Ibragimov and Walden (2007), Adam et al.
(2008), Chollete et al. (2011) and
Brandtner (2013)
Return (i) Diversification return (Booth and Fama
1992; Bouchey et al. 2012; Chambers and
Zdanowicz 2014; Qian 2012; Willenbrock
2011) (ii) Embrechts et al.’s (1999) class
of measures (iii) Return gap (Statman and
Scheid 2005, 2008)
International diversification Driessen and Laeven (2007)
Risk ratio (i) Tasche’s (2006) measure (ii)
Diversification ratio (Choueifaty and
Coignard 2008) (iii) Frahm and Wiechers’s
(2013) measure (iv) Pérignon and Smith’s
(2010) measure
Market portfolio Sharpe (1964)
Number of securities Evans and Archer (1968)
Risk contribution Maillard et al. (2010) and Qian (2011)
Principal portfolios (i) Portfolio diversification index (Rudin and
Morgan 2006) (ii) Effective number of bets
(Meucci 2009; Meucci et al. 2015)
Flint et al. (2015) Weight-based measures (i) Effective number of constituents (Deguest
et al. 2013) (ii) Herfindahl index (Evans
and Archer 1968)
Return-based measures (i) Return gap (Statman and Scheid 2005,
2008) (ii) Index contributions (Deguest
et al. 2013) (iii) Adjusted-R 2 (Flint et al.
2015)
Risk-based measures (i) (Pearson) Correlation interlude (Flint
et al. 2015) (ii) Tasche’s (2006) measure
(iii) Diversification ratio (Choueifaty and
Coignard 2008) (iii) Frahm and Wiechers’s
(2013) measure (iv) Pérignon and Smith’s
(2010) measure (v) Effective number of
correlated bets (Carli et al. 2014) (vi)
Effective number of bets (Meucci 2009;
Meucci et al. 2015) (v) Portfolio
diversification index (Rudin and Morgan
2006)
123
272 G. B. Koumou
Table 1 continued
Higher-moment measures (i) Diversification delta (Vermorken et al.

2012) (ii) Kirchner and Zunckel’s (2011)
measure
Cardinality-based measures Number of stocks (Evans and Archer 1968)
Lhabitant (2017) Size-based diversification Portfolio size
Weight-based diversification (i) Concentration ratio (ii) Herfindahl index
(iii) Gini index (iv) Bouchaud et al.’s
(1997) measure
Entropy-based diversification (i) Shannon entropy (ii) Kullback-Leibler
divergence (iii) Renyi entropy
Naive diversification Evans and Archer (1968) and Elton and
Gruber (1977)
Modern portfolio theory (i) Markowitz’s (1952) mean-variance (ii)
Mean-variance extensions (Jiang et al.
2008; Zheng et al. 2009; Bera and Park
2008; Ke and Zhang 2008; Philippatos and
Wilson 1972; Corvalan 2005; Elton and
Gruber 1973; Ledoit and Wolf 2003, 2004,
2017; DeMiguel et al. 2009a; Li et al.
2006) (iii) Weighted average of all
pairwise (Pearson) correlations (iv)
Implied average (Pearson) correlations (v)
Volatility-based (Pearson) correlation
proxy (vi) Variance-based (Pearson)
correlation proxy
Risk-based portfolios (i) Naive risk parity (ii) Maimonides risk
parity (iii) Equal risk contributions
(Maillard et al. 2010) (iv) Maximum
diversification (Choueifaty and Coignard
2008; Carmichael et al. 2018) (v)
Minimum-variance (Coqueret 2015)
Factor diversification (i) Bender et al. (2010) (ii) Ilmanen and
Kizer (2012) (iii) Rudin and Morgan
(2006) (iv) Effective number of bets
(Meucci 2009; Meucci et al. 2015)
Higher-moment diversification (i) Risk budgeting (Baitinger et al. 2017) (ii)
Diversification delta (Vermorken et al.
2012; Flores et al. 2017) (iii)
Mean-variance with higher moments (Lai
et al. 2006) (iv) Downside risk (Rockafellar
and Uryasev 2000; Chekhlov et al. 2005)
(v) Heavy-tailed risks (Ibragimov et al.
2011) (vi) Conditional (Pearson)
correlation (Longin and Solnik 2001)
Deguest et al. (2013) and Weight-based measures Effective number of constituents
Carli et al. (2014)
Risk-based measures (i) Goetzmann et al.’s (2005) measure (ii)
Effective number of correlated bets
(Roncalli 2014) (iii) Effective number of
bets (Meucci 2009)
123
Table 1 continued
Sharma (2018) Weight-based measures (i) Herfindahl index (Woerheide and Persson
1993) (ii) Shannon entropy (Bera and Park
2008) (iii) Effective number of
constituents (Carli et al. 2014)
Risk-based measures (i) Implied (Pearson) correlation index
(Skintzi and Refenes 2005) (ii)
Diversification ratio (Choueifaty and
Coignard 2008) (iii) Goetzmann et al.’s
(2005) measure (iv) Christoffersen et al.’s
(2012) measure
Factor-based measures (i) Effective number of bets (Meucci 2009;
Meucci et al. 2015) (ii) Portfolio
diversification index (Rudin and Morgan
2006) (iii) Unsystematic risk ratio (Sharma
2018)
2 Preliminaries
2.1 Notation
Throughout the review, vectors and matrices are shown in bold. The term asset refers
to both asset class or individual asset (e.g., stocks, bonds). We consider a one-period
settings. Thissetting rules out the possibilityof time diversification.
Let W = w = (w1 , . . . , w N ) ∈ R N : i=1 N
wi = 1; wi ≥ 0, i = 1, . . . , N be
the set of long-only portfolios, where wi is the weight of asset i in portfolio w, is the
transpose operator and R is the set of real numbers. The market portfolio is denoted
by wm . A portfolio w ∈ W is said to be diversified if at least two assets have strictly
positive weights (i.e., ∃ i, j /wi , w j > 0); if all assets have a strictly positive weights
(i.e., wi > 0, ∀ i = 1, . . . , N ), the portfolio is considered completely diversified. A
single asset portfolio is denoted δ i = (δi1 , . . . , δi N ) , where δi j is Kronecker delta
(i.e., δii = 1 for all i = 1, . . . , N and δi j = 0 for all i = j).
Let Ri ∈ R denote the future return on asset i, where R = L∞ (, E, P) is
the vector space of bounded real-valued random variables on a probability space
(, E, P), with being the set of states of nature, E is the − algebra of events, and
P is a − additive probability measure on (, E). The future return on portfolio w is
R(w) = w R, where R = (R1 , . . . , R N ) . The expected value of Ri is μi = E(Ri ),
its variance σi2 = Var(Ri ), its risk (Ri ), its cumulative function FRi (ri ) and its
probability function f Ri (ri ), where E(.), Var(.) and (.) are the operator of expectation,
the operator of variance and the risk measure, respectively. The covariance between
N
Ri and R j is σi j = Cov(Ri , R j ) and the covariance matrix is = σi j i, j=1 , with
σii ≡ σi2 and Cov(.) the operator of covariance. The (Pearson) correlation between
σ
Ri and R j is ρi j = σi iσj j = Cor(Ri , R j ) and the (Pearson) correlation matrix is ρ =
N
ρi j i, j=1 , with Cor(.) the operator of (Pearson) correlation. The return on the risk-free
asset is denoted R f . The expected value of portfolio w is μ(w) = w μ and its risk
123
274 G. B. Koumou
is (R(w)), with μ = (μ1 , . . . , μ N ) ; in particular, (R(w)) ≡ σ 2 (w) = w w in

the case of the variance risk measure, μ(δ i ) ≡ μi and σ 2 (δ i ) ≡ σi2 .
We denote the von Neumann-Morgenstern (VNM) utility function by u : R
→ R.
Its associated preference relation is denoted over R such that for all R1 , R2 ∈ R
R1 R2 ⇐⇒ E (u(R1 )) ≥ E (u(R2 )) . (1)
Since the investors are risk averse, u(.) is assumed to be concave. We will denote
E (u(R1 )) by Eu(R1 ).
The mean-variance expected quadratic utility function is
τ
U M V (R(w)) = w μ − w w, (2)
2
where τ is the coefficient of risk aversion. The mean-variance efficient portfolio,

which is obtained by maximizing U M V (R(w)) over W, is denoted wmv and the
global minimum-variance portfolio, which is obtained by minimizing portfolio vari-
ance, is particularly denoted wgmv . Let M V be the preference relation induced by
U M V (R(w)) such that
R1 M V R2 ⇐⇒ U M V (R1 ) ≥ U M V (R2 ). (3)
We recall that M V is equivalent to if each asset’s returns are normally distributed

and utility function u(.) is negative exponential (u(x) = − exp (−τ x)).
2.2 Definitions
In this subsection, we recall some concepts related to portfolio diversification required

for our review.
2.2.1 Exchangeable random variables
The concept of exchangeable random variables plays an important role in the opti-
mality, in terms of expected utility theory, of the naive diversification principle, a
particular case of the law of large numbers diversification principle.
Definition 1 (Exchangeability) The random variables R1 , . . . , R N are said to be

exchangeable if and only if their joint distribution FR (r) is symmetric.
A well-known example of an exchangeable sequence of random variables is an inde-

pendent and identically distributed sequence of random variables. Another example
is a normally distributed sequence of random variables with a constant covariance
matrix. For more details on exchangeable random variables, see the study by Aldous
(1985).
123
2.2.2 Correlation aversion
The notion of correlation aversion was introduced by Epstein and Tanny (1980). It is
equivalent to the notion of multivariate risk aversion of Richard (1975) and De Finetti
(1952) (see Dorfleitner and Krapp 2007).
Consider a decision maker (here an investor) whose preference is represented by a
bivariate utility function v(y, z), y, z ∈ R. Assume that v(y, z) is twice continuously
differentiable in y and z.
Definition 2 (From Richard (1975) Theorem 1) A decision maker is correlation averse
if and only if ∀ (y, z) ∈ R × R,
∂ 2 v(y, z)
≤ 0, (4)
∂ y ∂z
with strict inequality for at least one pair (y, z) ∈ R × R.
A decision maker is said to be a correlation lover if ∂ ∂v(y,z)

2
y ∂z ≥ 0 with strict inequality
for at least one pair (y, z) ∈ R × R. His/her is said to be correlation neutral if
∂ 2 v(y,z)
∂ y ∂z = 0.
In the case where the decision maker preference is represented by a multivariate
utility function v(y1 , . . . , y N ), yi ∈ R, the condition (4) is replaced by the following
one (see Richard 1975)
∂ 2 v(y)
≤ 0. (5)
∂ yi ∂ y j
When v(y, z) is not twice continuously differentiable, condition (4) is replaced by

the submodularity condition
v(y ∨ z) + v(y ∧ z) ≤ v(y) + v(z), ∀ (y, z) ∈ R × R (6)
where y ∨ z is the componentwise maximum of y and z, y ∧ z is the componentwise

minimum of y and z; see Eeckhoudt et al. (2007) for more details.
Consider the following two lotteries

(y − y, z − z) with probability 0.5
L1 =
(y, z) with probability 0.5

(y − y, z) with probability 0.5
L2 =
(y, z − z) with probability 0.5
where y, z are positive. Richard (1975) and recently Eeckhoudt et al. (2007) show that
a decision maker exhibits correlation aversion if he/she prefers L 2 to L 1 . This implies
that correlation averse decision makers prefer diversification (of course correlation
diversification) to concentration (see Epstein and Tanny 1980).
123
276 G. B. Koumou
In this review, we will use the notion of correlation aversion to prove that the
correlation diversification is at the core of expected utility theory.
3 Law of large numbers diversification principle
In this section, we review our first diversification principle, which is the law of large
numbers (LLN) principle.
The LLN is a theorem in probability and statistics first formulated by Cardano in
the 16th century and later proved by Bernoulli in the 18th century (see Seneta 2013). It
states that the average of a large number of independent identically distributed random
variables should be close to the expected value and will tend to become closer as the
number of random variables increases.
The applications of the LLN are not confined to the domain of probability or
statistics. The principle is also used in domains such as economics (including insurance
and finance) for risk management (see Samuelson 1963; Artstein and Hart 1981; Arrow
and Radner 1979; Judd 1985; Al-Najjar 2004; Uhlig 1996; Ross 1999). In this review,
we focus on the application of the LLN in portfolio theory as a diversification principle.
The LLN is applied in portfolio theory both as asset and time diversification princi-
ple. In this review, we focus on the LLN as an asset diversification principle. Readers
are referred among others to the works of Samuelson (1969), Samuelson (1971), Lloyd
and Haney (1980), McEnally (1985), Stangeland and Turtle (1999) and Bianchi et al.
(2016) for more details on the LLN as a time diversification principle.
3.1 Definition
The LLN diversification principle is a well-established risk management technique in

economics long before the birth of portfolio theory. Mention of it can be found in the
study by Bernoulli (1738/1954) (see also Sullivan 2011, pp. 74) as follows
It is advisable to divide goods which are exposed to some normalsize danger

into several normalsize portions rather than to risk them all together.
More specifically, the LLN diversification principle recommends investing a small

fraction of wealth in each of a large number of assets such that the resulting portfolio
weights are almost balanced (see Ross 1976; Ingersoll 1984). In other words, the
LLN principle recommends diversification in terms of portfolio weights and size. The
third dimension of diversification, which is portfolio risk, is not explicitly taken into
account.
Following Ingersoll (1987), the set of LLN well-diversified portfolios can be char-
acterized as follows

1
WLLN = w ∈ W wi = O as N −→ ∞ , (7)
N
123

where O(.) refers to the big-o function. wi = O N1 as N −→ ∞ means that
there is a positive constant c such as wi ≤ c N1 as N −→ ∞ equivalently wi =

O N1 as N −→ ∞ means that wi −→ 0 if N −→ ∞.
How large portfolio size N and balanced portfolio weights w will be in order to
be considered as a LLN diversified portfolio? In the context when the LLN is used to
reduce idiosyncratic risks, several minimum portfolio sizes were suggested; see Evans
and Archer (1968), Tang (2004), Alexeev and Dungey (2015) and Lhabitant (2017,
Table 3.2, pp. 100), among others. However, to the best of our knowledge, the literature
was silent on how balanced portfolio weights must be in order to be considered as a
LLN diversified portfolio. No specific degree of balance was suggested. To implement
a LLN diversification principle, the literature generally adopts the perfect balance
portfolio, also known as the naive or equal weight portfolio diversification principle,
which consists in investing the same amount of wealth in each available asset; more

formally w = N1 , . . . , N1 ≡ N1 .
3.2 Optimality and usefulness
The LLN diversification principle is central to asset pricing theories. It plays an impor-
tant role in the condition of no arbitrage in the arbitrage pricing theory (see Ross 1976;
Chamberlain 1983a, b; Ingersoll 1984), and in idiosyncratic risk pricing in both the
arbitrage pricing theory and the capital asset pricing model (see Sharpe 1964).
The LLN diversification principle, particularly the naive diversification principle, is
also used as an alternative stocks’ weighting scheme (see Clarke et al. 2013) and asset
allocation diversification strategy (see DeMiguel et al. 2009a; Pflug et al. 2012; Jacobs
et al. 2014; Hsu et al. 2018), even if asset classes are less exposed to idiosyncratic
risks.
Samuelson (1967) was the first who analyzed the optimality of the naive diversi-
fication in terms of expected utility theory. He (Samuelson 1967, Theorem II) shows
that the naive diversification is optimal in terms of expected utility theory when the
joint distribution of asset returns is exchangeable.
Green and Hollifield (1992) extend this result to the LLN diversification principle,
but under the restrictive conditions that each asset’s returns are normally distributed
and the utility function is negative exponential. More specifically, (Green and Hollifield
1992, Theorem 1) show that the mean-variance portfolio with mean different to the
mean return on the global minimum-variance portfolio is LLN well-diversified if and
only if the weights w of every portfolio satisfies
N

w μ + λmv ≤ K N |Cov(w R, Ri )| (8)
γmv γmv
i=1
and the payoff, = ω R, on every hedge position ω such that ω 1 = 0 satisfies

N
KN
|E()| ≤ |Cov(, Ri )|, (9)
γmv
i=1
123
278 G. B. Koumou

1 −1 1 wmv μ− 1 −1 μ μ −1 μ − 1 −1 μ wmv μ
where γmv = 2 , λmv = 2 and
μ −1 μ 1 −1 1 − 1 −1 μ μ −1 μ 1 −1 1 − 1 −1 μ
K N is a positive constant used in the study by Green and Hollifield (1992) to define
the LLN diversification as follows |wi | ≤ K N for each i = 1, . . . , N .
Green and Hollifield (1992, Corollary 1) also show that the global minimum-
variance portfolio, wgmv , is LLN well-diversified if and only if the weights w of
every portfolio satisfy

N

wgmv wgmv ≤ K N |Cov(w R, Ri )|. (10)
i=1
In sum, when (i) full information is available, (ii) the joint distribution of asset
returns is not exchangeable and (iii) conditions 8 and 9 or condition 10 are not satisfied,
the LLN diversification principle becomes a heuristic strategy, which is risky for a
risk averse investor whose preference can be represented by expected utility theory.
Diversification strategies taking into account risk information are then preferable.
However, the LLN diversification principle is still useful as a heuristic approach
(particularly the naive diversification) in the presence of estimation errors (see
DeMiguel et al. 2009a; Bouchaud et al. 1997; Bera and Park 2008; Coqueret 2015;
Carrasco and Noumon 2011; Tu and Zhou 2011), or in the presence of a sufficiently
high degree of model uncertainty in the form of ambiguous loss distributions (see
Pflug et al. 2012).
3.3 Measurement
There is no measure to capture the effect of LLN diversification in the literature, except
in the special case of naive diversification. In this review, we present the effective num-
ber of constituents (ENC), a class of measures of naive diversification. We also present
an example of application of the divergence measure (Kullback-Leibler divergence) as
a measure of naive diversification. Finally, we suggest a measure to capture the effect
of LLN diversification based on Kullback-Leibler divergence. For others measures of
the naive diversification, we refer readers to Yu et al. (2014), De Giorgi and Mahmoud
(2016) and Lhabitant (2017, Chapter 1).
3.3.1 Effective number of constituents
The ENC is a class of measures of naive diversification analyzed by Deguest et al.

(2013), Carli et al. (2014) and Lhabitant (2017, Chapter 1, pp. 15) under the name of
Hannah’s and Kay’s (1977) Index. It is defined as
N 1−υ
1
υ
ENCυ (w) = wυ 1−υ
= wiυ , υ ≥ 0, υ = 1, (11)
i=1
1
N υ υ
where wυ = i=1 wi is the L υ norm of w.
123
The ENC can also beviewed as the generalized or power mean of the inverse of
portfolio weight w−1 = w11 , . . . , w1N , where the weight vector is w
N
1−υ 1−υ
1
1
ENCυ (w) = wi , υ ≥ 0, υ = 1. (12)
wi
i=1
The interpretation of ENCυ (w) is straightforward: the higher the value of ENCυ (w)
is, the more diversified w is in terms of naive diversification. ENCυ reaches a minimum
equal to 1 if the portfolio w is fully concentrated in a single asset, w = δ i , and a
maximum equal to N if the portfolio w coincides with the naive portfolio, w = N1 .
Following Hannah and Kay (1977, Chapter 4, pp. 55–56), one can interpret υ as a
parameter capable of variation to reflect alternative views of naive diversification. It
captures investors’ degree of preference for LLN diversification: a high υ emphasizes
high degree of preference, while a low υ emphasizes low degree of preference.
If υ = 2, then ENC2 reduces to the inverse of the familiar Herfindahl or Herfindahl-
Hirschman index, introduced in portfolio theory as a portfolio diversification measure
by Woerheide and Persson (1993) and much used in the literature (see Zhou et al.
2013; DeMiguel et al. 2009a; Lhabitant 2017); more formally
1
ENC2 (w) = , (13)
HI(w)
N
where HI(w) = i=1 wi2 is the Herfindahl index.
By interpreting w as a vector of probability distribution of a random variable (where
N is the number of states) following Yu et al. (2014) and Lhabitant (2017, Chapter
1, pp. 20), it can also be shown that when υ converges to 1, ENC1 reduces to the
exponential of the Shannon index or entropy
ENC1 (w) = exp (SH(w)) (14)

N
with ln(.) the natural logarithm function and SH(w) = − i=1 wi ln (wi ) the Shan-
non entropy, another popular measure of naive diversification (see Zhou et al. 2013;
Yu et al. 2014; Lhabitant 2017).
ENCυ is also related to Bouchaud et al.’s (1997) class of measures of naive diver-
sification (BMυ ) as follows
1 − (ENCυ (w))1−υ
BMυ (w) = , υ ≥ 0, υ = 1. (15)
υ −1
In the case where υ converges to 1, the relation (15) becomes
BM1 (w) = ln (ENC1 (w)) . (16)
123
280 G. B. Koumou
ENCυ is also related to Renyi’s (1961) entropy (RE), another class of measures of
naive diversification, as follows
REυ (w) = ln (ENCυ (w)) . (17)
In practice, the choice of the value of υ remains an open question. Carli et al. (2014)
recommend to use ENC2 when dealing with long-short portfolios and both ENC1 and
ENC2 when dealing with long-only portfolios.
3.3.2 Kullback–Leibler divergence
Measures of divergence between probability distributions are also used to capture the
effect of naive diversification (see Bera and Park 2008; Zhou et al. 2013; Lhabitant
2017). In this review, we focus on Kullback-Leibler divergence.
Kullback-Leibler divergence of a discrete probability distribution p with respect to
a discrete probability distribution q is defined as follows

N

pi
D(p|q) = pi ln . (18)
qi
i=1
D(p|q) quantifies the difference between probability distributions p and q.

Applied to portfolio diversification by interpreting w as a vector of probability
distribution of a random variable, one can measure the degree of naive diversification
of a portfolio w by the Kullback-Leibler divergence of w with respect to the naive
portfolio N1

N

1 wi
D w = wi ln . (19)
N 1/N
i=1

The interpretation of D w N1 is straightforward: the lower the value of D w N1
1
is, the closer w is to and the more diversified w is in terms of naive diversification.

N
1
In the extreme case of D w N = 0, w coincides with N1 . Note that instead of using

the Kullback-Leibler divergence of w with respect to N1 , D w N1 , one can also use

the Kullback-Leibler divergence of N1 with respect to w, D N1 w .
The Kullback-Leibler divergence can also be adapted to capture the effect of LLN
diversification. Given a positive infinitesimal quantity > 0, one can consider that
the portfolio w is LLN well-diversified at level if and only if

1
D w ≤ . (20)
N
123
4 Correlation diversification principle
Our second diversification principle is the correlation diversification principle. The

term correlation refers here to any dependence measure including measure of similarity
or dissimilarity.
4.1 Definition
The correlation diversification principle is also a well-established risk management

technique in economics and finance long before the birth of portfolio theory. Mention
of it, implicitly, can be found in the Babylonian Talmud: Tractate Baba Mezia, folio
42a2 (see DeMiguel et al. 2009b, pp. 1914; Sullivan 2011, pp. 1914)
A man should always keep his wealth in three forms: one-third in real estate,
another in merchandise, and the remainder in liquid assets.
in Shakespeare and Phelps (c1923, Act I, Scene 1) (see also Markowitz 1999; Rubin-
stein 2002)
My ventures are not in one bottom trusted, Nor to one place; nor is my whole
estate Upon the fortune of this present year; Therefore, my merchandise makes
me not sad.
and more recently in Leavens (1945, pp. 473)
An important question is the extent of diversification that is desirable. Table 1
indicates that diversification among 10 items greatly narrows the range of prob-
able results as compared with 1 item. Adding another 10 will not give as much
relative improvement. In fact, it may be shown that in general the improvement
by diversification, in narrowing the spread between probable losses and gains,
varies as the square root of the number of items. For example, it would take 40
issues to give results twice as good (that is, with half the spread) as 10 issues.
Thus, after a reasonable diversification, which naturally can (and must, because
of inadequate supplies) be larger for a portfolio measured in millions of dollars
than for one measured in thousands of dollars, the advantage of any further
diversification hardly balances the difficulty of choosing (and watching) a great
variety of additional securities.
The assumption, mentioned earlier, that each security is acted upon by inde-
pendent causes, is important, although it cannot always be fully met in practice.
Diversification among companies in one industry cannot protect against unfavor-
able factors that may affect the whole industry; additional diversification among
industries is needed for that purpose. Nor can diversification among industries
protect against cyclical factors that may depress all industries at the same time.
Diversification is only one principle of investment management; it is primarily
to offset lack of full knowledge. The investor must still use all the information
2 http://juchre.org/talmud/babametzia/babametzia.htm.
123
282 G. B. Koumou
and judgment that he can muster in choosing specific securities and in timing
his purchases and sales.
Roy (1952) and Markowitz (1952) were the first to provide investment theories
that covered the effect of correlation diversification. For example, Markowitz (1952)
describes correlation diversification as follows
The adequacy of diversification is not thought by investors to depend solely on

the number of different securities held. A portfolio with sixty different railway
securities, for example, would not be as well diversified as the same size portfolio
with some railroad, some public utility, mining, various sort of manufacturing,
etc. The reason is that it is generally more likely for firms within the same
industry to do poorly at the same time than for firms in dissimilar industries.
(Markowitz 1952, pp. 89)
More specifically, the correlation diversification principle consists of allocating

wealth across a variety of assets, but exploiting the correlation between asset returns in
order to reduce portfolio risk. The underlying philosophy is as follows: diversification
helps to minimize both the probability of portfolio loss and its severity, through a
multilateral insurance in which each asset is insured by the remaining assets. The
key to the success of this multilateral insurance lies in the correlation between assets.
The less positively correlated the assets, the lower the probability that the assets do
poorly at the same time in the same proportion, and the better the protection offered
by this multilateral insurance, namely diversification. Therefore, in the presence of
correlation, it becomes sub-optimal to follow the LLN diversification principle ex-
ante.
As we can observe, the definition of the correlation diversification principle remains
vague and not unique. It does not provide any indication on how to allocate the wealth in
order to obtain a correlation well-diversified portfolio. Then how to obtain a correlation
well-diversified portfolio? Let G(w) be an objective function of investors that covers
the correlation diversification principle. Assume that the optimal portfolio of G(w)
can be obtained by maximization. The correlation diversification generated by G(w)
can be quantified using the following measure

N
N
G(w) = G(w) − wi G(δ i ) = wi (G(w) − G(δ i )) . (21)
i=1 i=1
The term in parentheses, G(w) − G(δ i ), measures the benefit of diversification to

hold portfolio w instead of to concentrate on asset i. It follows that G(w) measures
the average benefit of diversification to hold portfolio w instead of holding a single
asset portfolio. The set of correlation well-diversified portfolios associated with G(w)
can therefore be characterized as follows

N

WG = w ∈ W w ∈ arg Max G(w) − wi G(δ i ) . (22)
W i=1
123
The correlation diversification is at the core of the mean-variance model (also known
as the efficient diversification) and its extensions; see Markowitz (1959), Konno and
Yamazaki (1991), Fishburn (1977), Kang et al. (1996), Shalit and Yitzhaki (1984),
Alexander and Baptista (2002), De Giorgi (2002), Rockafellar and Uryasev (2000),
Krokhmal et al. (2001), Rockafellar et al. (2006), Rockafellar et al. (2007), Holthausen
(1981) and Cumova and Nawrocki (2014) among others.
It is also central to expected utility theory in general. To verify this let us first show
that a risk averse investor whose preference can be represented by expected utility
theory exhibits correlation aversion. Without a loss of generality, assume that the
investor’s initial wealth W0 = 1. The optimization problem of a risk averse investor
whose preference can be represented with expected utility theory is

N
max Eu 1 + wi Ri . (23)
w∈W
i=1
The function u(.) can also be viewed as a multivariate utility function ũ : R N

→ R
such that

N
ũ (R1 , . . . , R N ) = u 1 + wi Ri . (24)
i=1
The partial derivative of ũ (R1 , . . . , R N ) with respect to Ri and R j gives

N
∂ 2 ũ(R ∂ 2 u 1 + i=1 wi Ri
1, . . . , RN )
= wi w j ≤ 0, ∀ i, j = 1, . . . , N (25)
∂ Ri ∂ R j ∂ Ri ∂ R j
From (5) and (25), we can observe that a risk averse investor whose preference can be
represented by expected utility theory exhibits correlation aversion. As a result, the
diversification principle in expected utility theory follows the correlation diversifica-
tion principle and is guided by the concavity of u(.). This result remains valid in the
presence of the risk-free asset. Readers are also referred to Samuelson (1967), Scheff-
man (1973), Brumelle (1974) and MacMinn (1984) for the role of asset correlations
in portfolio diversification in expected utility theory.
In accordance with (21), the correlation diversification in expected utility theory
can be quantified by

N
Eu(w) = Eu (R (w)) − wi Eu(Ri ). (26)
i=1
As we can remark, the optimal portfolio of Eu(w) coincide with expected utility
optimal portfolio if and only if assets have the same expected utility, Eu(Ri ) = u, i =
123
284 G. B. Koumou
1, . . . , N . As a result, the correlation diversification is optimal in terms of expected

utility theory if and only if investors are indifferent between assets. This is in line with
the notion of preference for diversification introduced by Dekel (1989) and extended
later by Chateauneuf and Tallon (2002) and Chateauneuf and Lakhnati (2007).
In practice, one of the key challenges is the choice of the correlation measure to
identify the presence of correlation diversification benefit. In general, the (Pearson)
correlation measure is considered, and it is generally argued that correlation diversifi-
cation vanished when all (Pearson) correlations go to one (Holton 2009). This view is
very restrictive. First of all, (Pearson) correlation is a valid dependence measure only
if the joint distribution of asset returns is normally distributed, which is not generally
the case (see Mandelbrot 1967; Officer 1972; Hinich and Patterson 1985; Cont 2001;
Fama 1965). As a consequence, alternative dependence measures were suggested (see
Embrechts et al. 1999; Mashal and Zeevi 2002; Dhaene et al. 2002b, a; Malevergne
and Sornette 2003; Kole et al. 2007; Hennessy and Lapan 2002; Alink et al. 2004;
Gatfaoui 2018). Second, even when the joint distribution of asset returns is normally
distributed, as pointed out by Statman (2013), the benefit of correlation diversification
is not only dependent on the (Pearson) correlation. Using the return gaps, a measure of
correlation diversification introduced by Statman and Scheid (2008, 2005), Statman
(2013) provides the following example of international diversification to support his
argument
The correlation between U.S. stocks’ monthly returns, as measured by the S&P
500 Index, and international stocks, as measured by the MSCI EAFE index, was
0.88 during the 13-year period from January 2000 through December 2012. The
12-month gap between the returns of U.S. and international stocks during the
period ranged from a low of 0.2% to a high of 23.0%.
The benefits of diversification between U.S. and international stocks were neg-
ligible during the 12 months ending in October 2001, where the return gap was
0.2%, but huge during the 12 months ending in March 2004, where the return
gap was 23.0%. Investors do not know future return gaps, but they do know that
diversification would place them inside the gap, whether that gap is 0.2% or
23.0%.
U.S. stocks, as measured by the S&P 500, lost 37.00% in 2008; international
stocks, measured by the MSCI EAFE, lost 43.06%. The return gap that year
was 6.06%. An investor who diversified her portfolio equally between U.S. and
international stocks in 2008 lost 40.03% of her portfolio. That loss is midway
inside the gap between the loss of U.S. stocks’ loss and international stocks’
loss.
A 40.03% loss is very sad, but not as sad as the 43.06% loss of the investor
who placed his entire portfolio in international stocks. The top of the gap does
not necessarily provide a positive return. The entire gap can be in the region
of losses, as in 2008. Diversification does not eliminate the risk of loss-it only
mitigates that risk. Only the risk-free rate eliminates losses, and that rate is low.
123
The lack of (Pearson) correlation to identify the presence of correlation diversi-

fication was also highlighted in the mean-variance model by Li and Ziemba (1990)
and Wright (1987), and in expected utility theory by Samuelson (1967), Scheffman
(1973), Scheffman (1975), Brumelle (1974) and Wright (1987).
Another important issue is that asset correlations are time-varying (see Longin and
Solnik 1995; Campbell et al. 2002; Hiang Liow 2012; Solnik et al. 1996; Goetzmann
et al. 2005; Ehling and Heyerdahl-Larsen 2016; Evans et al. 2017; Andersson et al.
2008) and diversification is useful only in down markets. Thus, only downside cor-
relation will be considered when performing or evaluating correlation diversification
as suggested by Ang and Chen (2002), Campbell et al. (2002), Chollete et al. (2009),
Chua et al. (2009) and Page and Panariello (2018), among others.
Another important issue is that correlation diversification tends to be highly concen-
trated on asset’s common risk factors when it is applied in high correlated environments
(this is the case in asset class allocation). To avoid such an undesirable and risky con-
centration, the literature recommends risk factor allocation; see Bender et al. (2010),
Page and Taborsky (2011), Asl and Etula (2012), Ilmanen and Kizer (2012), Koedijk
et al. (2016a), Koedijk et al. (2016b), Blyth et al. (2016), Greenberg et al. (2016), Bass
et al. (2017) and Vaucher and Medvedev (2017), among others, for more details about
risk factor allocation. In Sect. 6, we will present one example of risk factor allocation
based on the risk budgeting approach.
Another issue is that asset correlations are estimated using historical data. How-
ever, it is well-known that past performance does not guarantee future performance.
Moreover, historical estimates of correlation suffer from estimation errors. Thus,
sophisticated approaches were developed to more accurately forecast or estimate of
asset correlations; see Engle (2002), Tse and Tsui (2002), Creal et al. (2011), Bailey
et al. (2019), Ledoit and Wolf (2003), Fan et al. (2008), Cappiello et al. (2006),
Poignard and Fermanian (2019), Van der Weide (2002), Creal and Tsay (2015),
Christoffersen et al. (2017), Nakagawa et al. (2018), Kim and Jung (2018), among
others.
4.3 Measurement
Several measures were designed to quantify the effect of portfolio correlation diversifi-
cation. In this review, we focus on the most commonly used measures. These measures
are: diversification returns (Booth and Fama 1992; Willenbrock 2011; Chambers and
Zdanowicz 2014; Bouchey et al. 2012; Qian 2012), Embrechts et al.’s (1999) class of
measures and diversification ratio (Choueifaty and Coignard 2008; Choueifaty et al.
2013). We refer the readers to Statman and Scheid (2005, 2008), Mitton and Vorkink
(2007, pp. 1269, Equation 7), Skintzi and Refenes (2005) and Carmichael et al. (2018)
for other measures of portfolio correlation diversification.
4.3.1 Diversification returns
In the particular case where u(.) is the logarithmic function, Eu(w) from (26) coin-
cides with diversification returns, a measure of portfolio correlation diversification
123
286 G. B. Koumou
introduced by Booth and Fama (1992),
E ln(w) = E ln(1 + w R) − w E ln(1 + R), (27)
where ln(1 + R) = (ln(1 + R1 ), . . . , ln(1 + R N )) is the vector column of assets’

compound returns and ln(1 + w R) is portfolio compound return. Booth and Fama
(1992) focus on the approximation of E ln(w) obtained using the second-order Taylor
series expansion at the mean return.
Two other approximations of E ln(w) have also received considerable attention in
the literature. Let μG be the geometric mean. Since
ln(1 + μG ) = E ln(1 + R), (28)
E ln(w) can be rewritten in terms of μG
E ln(w) = ln(1 + μG (w)) − w ln(1 + μG ), (29)
where μG = (μG 1 , . . . , μG N ) is the vector of assets’ geometric means and μG (w)

is portfolio geometric mean. The approximation of (29) using the first-order Taylor
series expansion at zero leads to excess growth rate
E ln(w) = μG (w) − w μG , (30)
a measure of portfolio correlation diversification analyzed by Fernholz and Shay

(1982) and Fernholz (2010) in stochastic portfolio theory.
Taking the second-order Taylor series expansion of (28) at zero and neglecting the
terms (μG )2 and (E(R))2 lead to
Var(R)
μG ≈ E(R) − . (31)
2
Combining (30) and (31), one obtains another approximation of E ln(w)
E ln(w) ≈ w σ 2 − w w, (32)

where σ 2 = σ12 , . . . , σ N2 . This approximation of E ln(w) was analyzed by Willen-
brock (2011), Chambers and Zdanowicz (2014), Bouchey et al. (2012), Qian (2012)
and Fernholz (2010) but under the name of excess growth rate, and has been shown by
Carmichael et al. (2015) to be the specific portfolio diversification in the mean-variance
model . Its normalized version
w σ 2
GLR(w) = (33)
w w
was analyzed by Goetzmann et al. (2005) and Goetzmann and Kumar (2008).
123
4.3.2 Embrechts et al.’s (1999) class of measures
Let R1 , . . . , R N be a sequence of random variables. Embrechts et al.’s (2009) class of

correlation diversification measures is defined as
N

N
EFK (R) = (Ri ) − Ri . (34)
i=1 i=1
Applied to portfolio w, the following class of measures is obtained

N
EFK (w) = (wi Ri ) − w R . (35)
i=1
An intuitive interpretation can be provided to EFK (w) when (.) is homogeneous

of degree one, i.e., (b R) = b(R) for all b ∈ R+ . In that case, EFK (w) can be
rewritten as follows

N
EFK (w) = wi ( Ri ) − w R . (36)
i=1
Similarly to Eu(w), EFK (w) measures the average benefit of diversification, in terms
of risk reduction, of holding portfolio w instead of holding a single asset portfolio.
Tasche (2006) analyzes the normalized version of EFK which applied to portfolio
w is defined as follows

w R
DF (w) = N . (37)
i=1 (wi Ri )
In general, EFK (w) and DF (w) are viewed as adequate portfolio correlation
diversification measures when (.) is a coherent (in the sense of Artzner et al. (1999))
and comonotonic risk measure (see Emmer et al. 2015).
When (.) is different from the volatility or the variance risk measure, EFK and
DF take higher moments into account. An example is the diversification delta (DD)
introduced by Vermorken et al. (2012)
DD(w) = 1 − exp (EFK H (w)) , (38)
where (.) is Shannon entropy denoted by H (.); more formally

(R) ≡ H (R) = − f R (r ) ln ( f R (r )) dr . (39)
Flores et al. (2017) show that the DD is an inadequate measure of diversifica-

tion since H (.) is not positive homogeneous, sub-addtitive and comonotonic additive.
123
288 G. B. Koumou
Moreover, the entropy risk measure H (.) is not left-bounded. As an alternative, Flores
et al. (2017) suggest replacing H (.) by the exponential entropy exp(H (.))

exp(H (R)) = exp − f R (r ) ln ( f R (r )) dr . (40)
exp(H (.)) satisfies desirable properties, and defines a new diversification measure
which can be viewed as a special case of DF
DD∗ (w) = 1 − DFexp(H ) (w). (41)
The principal advantage of Embrechts et al.’s (2009) or Tasche’s (2006) class mea-
sures is they consider the joint distribution of asset returns. Their major limit is the
computational cost. In addition, they are also only applicable to portfolios of risky
assets. Furthermore, they require the specification of the risk measure (.) and do not
explicitly take into account asset interdependence measures.
4.3.3 Diversification ratio
Another measure designed to capture the effect of portfolio correlation diversification

is the diversification ratio (DR) developed by Choueifaty and Coignard (2008). It is
defined as the ratio of the weighted average of assets’ volatilities to portfolio volatility,
i.e.,
w σ
DR(w) = √ . (42)
w w
It can be related to Tasche’s (2006) measure when (.) is the volatility risk measure
1
DR(w) = . (43)
DF (w)
By definition, DR(w) is greater than 1, i.e., DR(w) ≥ 1. If w is full concentrated,

w = δ i , then DR(w) = 1. The optimal portfolio of DR is obtained by maximization
and is called the most diversified portfolio (MDP).
According to Choueifaty et al. (2013), the MDP can also be obtained by solving
the following optimization problem with weights rescaled to sum to 1 afterwards
max w w (44)
w∈R+
s.t. w σ = 1. (45)
From (44)-(45), the MDP can also be obtained in two steps. In the first step, the
following optimization problem is solved
max w ρ w. (46)
w∈W
123
Let w∗ denote the solution of the problem (46). The second step is to rescale w∗ using
assets’ volatilities
wi∗ wi∗
N
wmd p,i = , ∀ i = 1, . . . , N (47)
σi σi
i=1
where wmd p is the weight vector of the MDP.

Another alternative optimization problem to obtain the MDP can also be found in
the study by Carmichael et al. (2018). The portfolio construction based on the DR
is known as the maximum diversification strategy, which is one of the most popular
risk-based or smart beta investment strategies. It is currently used in the industry to
manage billions of dollars.3 One limitation of the DR is that it is only valid when risk
can be measured by variance or volatility.
5 CAPM diversification principle
Our third diversification principle is the capital asset pricing model (CAPM) diversi-
fication principle. It was introduced by Sharpe (1964).
5.1 Definition
The CAPM distinguishes between individual asset and asset classes diversification.
5.1.1 Individual asset diversification
The CAPM individual asset diversification principle is based on the implication of the
CAPM which states that individual asset risk, measured by variance, can be decom-
posed into two components: systematic risk and idiosyncratic risk. The diversification
strategy that can totally eliminate individual asset idiosyncratic risk is the CAPM
individual asset diversification. In other words, as highlighted by Ingersoll (1984), a
portfolio is well-diversified in the sense of CAPM individual asset diversification prin-
ciple only if it is perfectly correlated with the market portfolio. As a consequence, the
set of well-diversified portfolios in terms of the CAPM individual asset diversification
can be characterized as follows

W M P = w ∈ W | Cor(R(w), R(wm )) = 1 (48)
A trivial example of an element of W M P is the market portfolio.
5.1.2 Asset class diversification
The CAPM asset class diversification principle is based on the one-fund theorem,
another implication of the CAPM, first established by Tobin (1958). The theorem states
3 See http://www.tobam.fr/about-us/ for more details.
123
290 G. B. Koumou
that, in the presence of the risk-free asset, every mean-variance efficient portfolio is a
linear combination
of the risk-free asset and the market portfolio (risky fund); more
formally w = w f , (1 − w f )wm , where w f , the weight of the risk-free asset, is
determined by the investor’s risk aversion coefficient
μ(wm ) − R f
1 − wf = . (49)
τ σ 2 (wm )

In the literature, the mix w = w f , (1 − w f )wm is interpreted as a correlation diver-
sification between two asset classes, the risky fund and the risk-free asset.
In sum, the CAPM diversification principle is a sophisticated combination of the
market portfolio diversification principle and the correlation diversification principle.
The difference between CAPM and correlation diversification principles is that, at
individual asset level, the former concerns only idiosyncratic risk reduction and the
latter concerns both systematic and idiosyncratic risk reduction.
The CAPM predicts the mean-variance

optimality of the market portfolio and the mix
portfolio w f , (1 − w f )wm at equilibrium (see Roll 1977). The problem is that the
CAPM is based on many highly sensitive and unrealistic assumptions (see Markowitz
2005; Goltz and Le Sourd 2011). In addition, the CAPM cannot be justified on positive
grounds. Several empirical studies (see Black et al. 1972a; Miller and Scholes 1972;
Malkiel and Xu 1997; Malkiel and Yexiao 2006; Levy 1978; Friend and Blume 1970;
Fu 2009; Ang et al. 2006; Boyer et al. 2009; Mitton and Vorkink 2007; Ang et al. 2009;
Blume and Friend 1973; Goyal and Santa-Clara 2003; Fama and French 1992, 1993;
Lehmann 1990; Kraus and Litzenberger 1976; Tinic and West 1986, 1984, among
others) report evidence that the CAPM fails to explain the price behavior of risky
assets as well as might be expected. Roll (1977) has gone further by arguing that there
is practically no possibility to justify the CAPM on positive grounds, since the true
market portfolio is unobservable.
Despite its theoretical and empirical limitations, the CAPM diversification is one
of the most used diversification principles. In practice, the market portfolio is often
approximated by a stock market (capitalization or equally-weighted index). However,
a stock market index is not a good proxy for the market portfolio (Goltz and Le
Sourd 2011). Several studies have tested the mean-variance optimality of stock market
indexes. There is no consensus about the result. For example, Fama and MacBeth
(1973), Gibbons (1982) and recently Levy and Roll (2010) show that stock market
indexes are mean-variance optimal. However, Zhou (1991), Gibbons et al. (1989),
Harvey and Zhou (1990), Grinold (1992) and recently Brière et al. (2013) find that
stock market indexes are not mean-variance optimal.
The risk-free asset is often approximated by bonds, and its weight w f is set generally
equal to 0.4. This provides the popular 60/40 portfolio, i.e., 60 percent of capital on
stocks (generally stock market index) and 40 percent on the bonds, promoted by
financial industries.
123
Table 2 Selected studies on alternative asset classes

Alternative asset classes References
Commodities Estrada (2016), Hoevenaars et al. (2008), Gorton and Rouwenhorst

(2006), Emmrich and McGroarty (2013), Baur and Lucey (2010),
Bessler and Wolff (2015), Gao and Nardari (2018) and Froot
(1995)
Real state Fugazza et al. (2007), Hoevenaars et al. (2008), Froot (1995),
Rehring (2012), Garcia-Feijoo et al. (2012), Stelk et al. (2017)
and Boudry et al. (2020)
Hedge funds Hoevenaars et al. (2008), Bekkers et al. (2009), Bessler et al.
(2012) and Banz and De Planta (2002)
Cryptocurrencies Bouri et al. (2017b), Corbet et al. (2018), Bouri et al. (2017a),
Milunovich (2018) and Briere et al. (2015)
Art Renneboog and Spaenjers (2013), Mei and Moses (2002),
Goetzmann (1993), Korteweg et al. (2015), Anderson (1974),
Campbell (2008), Taylor and Coleman (2011), Tucker et al.
(1995), Pesando (1993), Mandel (2009), Oosterlinck (2017),
Skinner and Jackson (2018) and Masset and Weisskopf (2018)
Fine wine Fogarty (2010), Chu (2014), Fogarty and Sadler (2014), Aytaç and
Mandou (2016), Bouri (2015), Bouri et al. (2018) and Le Fur and
Outreville (2019)
The meltdown of 2001–2002 and the 2007–2009 financial crisis amply demon-
strated the lack of sufficient diversification of the 60/40 portfolio during periods of
distress. The reasons for this lack of diversification are twofold. The first reason is
that stocks and bonds do not have the identical or similar risk profile as pointed out by
Qian (2011) and Geczy (2016). Stocks are more volatile (i.e., risky) than bonds, so the
60/40 portfolio risk profile is similar to that of stocks. The 40% proportion of bonds
fails to provide adequate buffer against loss in stocks, even when correlation between
the two asset classes is low. The second reason is that stock market portfolios are not
well-diversified as might be expected (see Amenc et al. 2012; Choueifaty et al. 2013).
As a consequence, several alternatives to the 60/40 portfolio were suggested. The
first alternative is to improve on stock market portfolios by considering alternative
weighting schemes, including the fundamental weighting scheme of Arnott et al.
(2005), the risk contribution weighting scheme (see Maillard et al. 2010), the naive
portfolio weighting scheme (see Chow et al. 2011), the minimum variance weighting
scheme (see Haugen and Baker, Spring 1991; Clarke et al. 2006) and the maximum
diversification weighting scheme of Choueifaty and Coignard (2008).
The second alternative is to improve on the 60/40 portfolio by considering addi-
tional asset classes such as cash, commodities, real estate, private equity, hedge funds,
managed futures, cryptocurrencies, art and fine wine. Several studies explore the bene-
fits of considering these alternative asset classes in asset allocation; see Table 2, among
others.
The third alternative is to replace the 60/40 allocation by another (1 − w f )/w f
allocation. For example, Qian (2011), using the risk contribution diversification prin-
ciple, suggests replacing the 60/40 portfolio by the 25/75 portfolio; see Asness et al.
123
292 G. B. Koumou
(2012) and Anderson et al. (2012) for more alternatives using the risk contribution
diversification principle. The (100+X)/X long-short weighting schemes, including the
popular 130/30 long-short portfolio are also suggested (see Lo and Patel 2008; Gilli
et al. 2011; Johnson et al. 2007; Jacobs and Levy 2007; Krusen et al. 2008).
Despite the fact that all alternatives to the 60/40 portfolio have attracted substantial
interest from both institutional and individual investors over the last decade, according
to the 2018 Trends in Investing Survey,4 conducted by the Journal of Financial Planning
and the FPA Research and Practice Institute, the 60/40 portfolio is still viable.
5.3 Measurement
The effect of the CAPM asset class diversification can be quantified using the corre-
lation diversification measures presented in Sect. 4.3. In this section, we present the
measures designed to capture the CAPM individual asset diversification.
5.3.1 Portfolio size
The first measure designed to capture the effect of the CAPM individual asset diver-
sification is portfolio size; more formally

N
PS(w) = 1{wi >0} (wi ), (50)
i=1
where 1{wi >0} (wi ) is the indicator function. This measure was introduced implicitly
by Evans and Archer (1968), who were inspired by the CAPM. More specifically,
the CAPM implies that portfolio risk, measured by variance, can be decomposed as
follows
σ 2 (w) = β(w)2 σ 2 (wm ) + σε2 (w), (51)
where β(w)2 σ 2 (wm ) is the systematic risk with β(w) = Cov(R(wσm2 ),R(w)) =
N m
i=1 wi βi and σε (w) is the idiosyncratic risk. When portfolio w coincides with the
2
market portfolio wm , all the idiosyncratic risk is eliminated or diversified, or equiva-

lently β(w) = 1 and σ 2 (w) = σ 2 (wm ).
From (51), Evans and Archer (1968) tried to find if there is a minimum number
of securities (N ∗ ) inferior to the size of the market (Nm ) such that for all N ≥ N ∗
the idiosyncratic risk is almost diversified, or equivalently β(w) ≈ 1 and σ 2 (w) ≈
σ 2 (wm ). Using individual assets listed in the Standard and Poor’s Index for the year
1958, setting portfolio w to the naive portfolio (w = 1/N ) and selecting portfolios
at random, the authors show that 10 individual assets are sufficient to eliminate all
the idiosyncratic risk. This minimum number of assets was updated several times in
the literature (see Alexeev and Dungey 2015; Tang 2004; Lhabitant 2017). However,
4 For more details see https://www.onefpa.org/business-success/ResearchandPracticeInstitute/Pages/
2018TrendsInInvesting.aspx.
123
little has been done to identify which N ∗ of the available individual assets should be
included. To the best of our knowledge, only Mao (1970) provided theoretical selection
criteria to identify which N ∗ individual assets should be included.
Portfolio size is the most used measure of portfolio diversification in the literature,
because it is straightforward to implement and to interpret. However, it remains a
crude and naive measure of the CAPM individual asset diversification as highlighted
in Sharpe (1972), because two portfolios with same size can have different level of
idiosyncratic risk regardless of whether the portfolio is equally weighted or market
weighted. Worse, a small portfolio can be more diversified in terms of the CAPM
individual asset diversification than a large portfolio.
5.3.2 Sharpe’s (1972) measure
Another measure designed to capture the effect of the CAPM individual asset diver-
sification is the measure developed by Sharpe (1972).
Assume that individual assets’ idiosyncratic shocks εi , i = 1, . . . , N are uncorre-
lated. Then the decomposition (51) becomes

N
σ 2 (w) = β (w)2 σ 2 (wm ) + wi2 σε2i . (52)
i=1
Consider the individual asset k as a reference asset. Let σε2i|k be the relative asset i’
idiosyncratic risk such that
σε2i
σε2i|k = . (53)
σε2k
Sharpe’s (1972) measure is defined as follows
1
SH(w) = N . (54)
i=1 wi σεi|k
2 2
When individual assets have the same idiosyncratic risk level, σε2i = σε20 , ∀ i =
1, . . . , N , SH(w) is proportional or equivalent to the inverse of the Herfindhal index
1
SH(w) = σε20 N . (55)
i=1 wi
2
The major limitation of SH(w) is that it fails to identify the market portfolio as a
well-diversified portfolio.
5.3.3 Barnea and Logue’s (1973) measures
Another measure designed to capture the effect of the CAPM individual asset diver-
sification is the coefficient of determination introduced by Barnea and Logue (1973).
123
294 G. B. Koumou
It is defined as a ratio of portfolio systematic risk to portfolio risk
β (w)2 σ 2 (wm )
CD(w) = . (56)
σ 2 (w)
By definition, CD(w) is in a range between 0 and 1, i.e., 0 ≤ CD(w) ≤ 1. If w is

well-diversified in terms of the CAPM individual asset diversification, then CD(w) =
1. Contrary to Sharpe’s measure, CD(w) identifies the market portfolio as a well-
diversified portfolio (CD(wm ) = 1). However, it can generate counter intuitive results.
To see this, consider two portfolios w1 and w2 which have the same idiosyncratic risk
but different total risk. Without the loss of generality assume that σ 2 (w1 ) > σ 2 (w2 ).
According to CD(w), w1 is more diversified than w2 .
Barnea and Logue (1973) also developed the tracking error, another measure cap-
turing the effect of the CAPM individual assets diversification, but which identifies
the market portfolio as a well-diversified portfolio and corrects the lack of CD(w). It
is defined as follows

σε(w) = E(ε(w)2 ). (57)
If w is well-diversified in terms of the CAPM individual asset diversification, then

σε(w) = 0.
The two measures CD(w) and σε(w) are more used in corporate finance (Datta et al.
1991; Amihud and Baruch 1999; Demsetz and Strahan 1997; Amihud and Baruch
1981; Jahera et al. 1987) than in portfolio selection (Cresson 2002; Byrne and Lee
2003).
6 Risk contribution diversification principle
Our last diversification principle is the risk contribution (RC) or the risk parity diver-
sification principle. This diversification principle was suggested after the 2007–2009
financial crisis as an alternative to the market portfolio and the 60/40 portfolio. Since
then, it has become a popular investment strategy such that it is featured in the Wall
Street Journal (see Dagher 2012). Despite the huge losses observed in risk parity port-
folios in 2015, this strategy continues to be used by the investors. According to 2016
Risk Parity Investment Survey,5 of the 102 investors surveyed, only 12% said they
would reduce their current allocations, 72% said they would maintain their current
allocations, and 16% said they would reduce their current allocations.
6.1 Definition
The RC diversification principle is based on risk allocation rather than wealth or

capital allocation. It consists of equalizing assets’ absolute risk contributions, so it is
also called equal risk contribution.
5 http://www.ai-cio.com/2016-Risk-Parity-Investment-Survey/.
123
Formally, assume that portfolio risk (w R) can be additively decomposed

N

(w R) = i w R , (58)
i=1

where i w R is the absolute risk contribution of asset i. Then portfolio w is diversi-
fied in terms of the RC diversification if and only if (see Maillard et al. 2010; Bhansali
et al. 2012; Roncalli 2014; Roncalli and Weisang 2016)

i w R = j w R ∀ i, j = 1, . . . , N . (59)
When the risk measure (.) is assumed to be homogeneous of degree κ ∈ R, the

decomposition (58) can be obtained using the Euler principle (see Tasche 2000, 2008)
w ∂ w (w R)
i (w R) = , ∀ i = 1, . . . , N (60)
κ
R)

∂(w R)
where = (1 , . . . , N ) and ∂ w (w R) = ∂(w
∂w1 , . . . , ∂w N is the vector
of assets’ marginal risk contributions and ∂ w is the partial derivative operator with
respect to w.
From (59), the set of the RC well-diversified portfolios can therefore be character-
ized as follows

W RC = w ∈ W|i w R = j w R , ∀ i, j = 1, . . . , N , (61)
or as follows (see Maillard et al. 2010)
N
2
W RC = arg Min i w R − j w R . (62)
w∈W i, j=1
Assume that (.) is the volatility or variance risk measure. Then W RC becomes
(see Maillard et al. 2010)
β −1
W RC = w ∈ W|wi = N i −1 , ∀ i, j = 1, . . . , N (63)
i=1 βi
i ,R(w))
with βi = Cov(R
σ 2 (w)
.
Assume further that assets’ (Pearson) correlations are identical. Then W RC becomes
(see Maillard et al. 2010)
σ −1
W RC = w ∈ W|wi = N i −1 , ∀ i, j = 1, . . . , N (64)
i=1 σi
123
296 G. B. Koumou
and the RC diversification principle is referred to as the naive risk parity (see Bhansali
et al. 2012; Fisher et al. 2015; Asness et al. 2012; Anderson et al. 2012; Chaves et al.
2012), or the inverse volatility strategy (see Chaves et al. 2011; Neffelli 2018).
In the case where (.) is the volatility risk measure and assets’ (Pearson) correlations
are not identical, or (.) is not the volatility risk measure, the exact solution of (62)
no longer exists. Several algorithms were proposed to find the RC well-diversified
portfolios numerically; see Maillard et al. (2010), Chaves et al. (2012), Spinu (2013),
Mausser and Romanko (2014), Hochreiter (2015), Feng and Palomar (2015), Bai and
Scheinberg (2015), Bai et al. (2016) in the case where (.) is the volatility risk measure,
and Mausser and Romanko (2018), Boudt et al. (2013), Cesarone and Colucci (2018)
otherwise.
Three intuitive economic justifications were provided to the RC diversification
principle. The first intuitive economic justification is based on the risk minimization.
Indeed, from Maillard et al. (2010), it is straightforward to show that the RC well-
diversified portfolios can be obtained by solving the following optimization problem
min (w R) (65)

w∈W

1 1
s.t D w ≤ − (η + N ln(N )) , (66)
N N
where the left-hand side of the diversification constraint (66) is the Kullback-Leibler
divergence of the naive portfolio N1 with respect to portfolio w, and η is an arbitrary
constant such that η ∈ (−∞, −N ln(N )). In light of formulation (65)-(66), it is clear
that the RC well-diversified portfolio is a shrinkage of the minimum risk portfolio
towards the naive portfolio. Therefore, the RC diversification principle is useful for a
risk minimization investor in the presence of estimation errors or model uncertainty
(see Fisher et al. 2015). The degree of uncertainty or estimation errors is captured by the
constant η. In addition, from (66) and Sect. 3.3.2, the RC diversification principle can
also be viewed as a LLN diversification principle at level = − N1 (η + N ln(N )), or
as a combination of the correlation diversification principle covered by the objective
function in (65) (risk measure) and the LLN diversification principle at level =
− N1 (η + N ln(N )) covered by the constraint (66).
The second intuitive economic justification is based on the leverage aversion theory.
Pioneered by Black (1972) and extended recently by Frazzini and Pedersen (2014),
the leverage aversion theory breaks the CAPM, and predicts that the highest risk-
adjusted return is not achieved by the market portfolio, but rather by a portfolio that
over-weights safer assets. As a result, an investor who is less leverage averse than
the average investor can benefit by over-weighting low-beta assets, under-weighting
high-beta assets, and applying some leverage to the resulting portfolio. Black et al.
(1972b) and Frazzini and Pedersen (2014) confirm this prediction across a range of
markets and major asset classes. Asness et al. (2012) show that this theory constitutes
a theoretical underpinning for the naive risk parity by providing empirically evidence
that the out-sample performance of the naive unleveraged and leveraged risk parity
are consistent with the leverage aversion theory. However, Anderson et al.’s (2012)
empirical results do not support this consistency.
123
The third intuitive economic justification is provided by Kaya and Lee (2012)
and Kaya (2014) based on Bayesian rule. Under the conditions that (i) asset returns
are normally distributed; (ii) the investor believes a priori that each of the short
sale-constrained minimum-variance portfolio weights, wi , is independently and iden-
tically distributed as a Beta distribution with probability density function π(wi ) =
γ
(γ +2) wi
(γ +1) wγ +1 with w > 0, γ > 0 and (.) is the gamma function; and (iii) the investor
believes a priori that the variance of the short sale-constrained minimum-variance
portfolio, denoted by σ 2 , has an independent prior distribution π(σ 2 ), Kaya (2014)
shows that the weights of the risk parity portfolio are the mode of the posterior distri-
bution of the short sale-constrained minimum-variance portfolio weights and can be
obtained by solving
γσ2
N
min w w − ln(wi ) (67)
0≤w≤w T −1
i=1
with w large enough to make the upper constraint w < w unbinding and T the number
of observations of asset returns. More specifically
w∗
wr c = (68)
1 w∗
with w∗ a solution of (67). In other words, the risk parity portfolio is the portfolio with
the highest probability of being the minimum-variance portfolio given the investor’s
prior distribution on portfolio weights and the observed asset returns.
It is proved, in the study by Maillard et al. (2010), that the RC diversification principle
is mean-variance optimal when (.) is the volatility risk measure, but under very
restrictive (if not impossible) conditions, namely that assets have identical Sharpe
ratio and identical (Pearson) correlations.
Kaya (2014) extends this result by showing that the RC diversification principle is
mean-variance optimal if and only if
μ μ
dg ρ −1 = c1, (69)
σ σ

where μ σ = μ1
σ1 , . . . , μN
σN , dg(.) is a diagonal matrix operator and c is a constant.
The condition (69) is less restrictive than the conditions that assets have identical
Sharpe ratio and identical (Pearson) correlations, but remains restrictive in general.
When (.) is the volatility risk measure, Fisher et al. (2015) also show that the RC
diversification principle is optimal, but in the case where assets’ (Pearson) correlations
are identical and with respect to minimax or maximin portfolio optimization under
some particular conditions. More specifically, the authors show that if assets’ expected
123
298 G. B. Koumou
returns are nonnegative and the sum of assets’ Sharpe ratios is greater than a certain
(unknown) value, then the naive risk parity portfolio is the only minimax portfolio
among all portfolios without short sales
min wrc μ = max min w μ, (70)

μ∈Mσ w∈W μ∈Mσ
N μi
μ μ
where Mσ = μ : i=1 σi ≥ σ , μi ≥ 0, i = 1, . . . , N with σ a constant.
The authors also demonstrate that if each assets’ expected return is positive and all
(Pearson) correlations are less than one, then the naive risk parity is the only maximin
portfolio with respect to the Sharpe ratio among all portfolios without short sales
wrc μ w μ
min = max min ,
∗
μ∈Mσ , ρ∈Mρ wrc (dg(σ )ρ dg(σ )) wr c w∈W μ∈M∗σ , ρ∈Mρ w (dg(σ )ρ dg(σ )) w
(71)

where Mρ = ρ : ρi j, i= j ≤ ρ, 0 < ρ < 1, ρii = 1, i = 1, . . . , N and M∗σ =

μ : μσii ≥ μ
σ > 0, i = 1, . . . , N .
Finally, the authors show that if the sum of all assets’ Sharpe ratios is positive, then
the naive risk parity is the only maximin portfolio with respect to the Sharpe ratio,
among all portfolios without short sales
wrc μ w μ
min∗∗ = max min∗∗ √ , (72)
μ∈Mσ wrc wr c w∈W μ∈Mσ w w
N μi
μ
where M∗∗σ = μ: i=1 σi ≥ σ > 0 .
In sum, the RC diversification principle is a heuristic approach, except in unrealistic
circumstances as highlighted in Agapova et al. (2017)
The discussions about Risk Parity portfolios found in the literature are suffi-
ciently arcane so as to be viewed as profound by many investors. However,
except in unrealistic circumstances, a Risk Parity portfolio is not an optimal
portfolio and hence has no special attraction. It does not maximize the informa-
tion ratio, does not maximize the Sharpe ratio, does not maximize reward subject
to a specified level of risk, does not minimize risk at a specified level of reward,
does not minimize variance, is not mean-variance efficient, etc. So, what’s the
big deal about Risk Parity?
In practice, the RC diversification principle is applied both to asset classes (Qian
2011; Chaves et al. 2011) and individual assets (Clarke et al. 2013), and it is often
leveraged to match a desirable expected return target (Anderson et al. 2012; Asness
et al. 2012).
When (.) is not the volatility risk measure, the key challenge is to compute the RC
diversification portfolio. Mausser and Romanko (2018), Boudt et al. (2013), Cesarone
and Colucci (2018) offer some solutions for some particular risk measures.
123
Another issue related to the RC diversification principle is the estimation of the

covariance matrix when (.) is the volatility risk measure. In general, the sam-
ple covariance matrix is used. However, for high-dimensional datasets, the sample
covariance matrix is not an accurate or reliable estimator. As a consequence, more
sophisticated covariance matrix estimators have been proposed; see the studies by
Engle (2002), Bailey et al. (2019), Ledoit and Wolf (2003), Fan et al. (2008), Cap-
piello et al. (2006), Van der Weide (2002), among others, and Ardia et al. (2017),
Neffelli (2018) and Nakagawa et al. (2018) for their application to the RC diversifica-
tion principle. To mitigate the estimation error of the covariance matrix, Kapsos et al.
(2018), following Tutuncu and Koenig (2004), and Costa and Kwon (2019), follow-
ing Kapsos et al. (2018) and Goldfarb and Iyengar (2003), also suggest the robust RC
diversification approach.
Another major issue is that the RC diversification principle can still be highly
concentrated in only one or two true risk factors, in particular in high correlated
environments (see Bhansali et al. 2012, pp. 103). Thus, in the case where (.) is the
volatility risk measure, as an alternative, the factor-based RC diversification principle
was suggested. In the factor-based RC diversification principle, asset returns are first
decomposed as the sum of k risk factors (independent or lowly correlated). Next,
assets’ absolute risk contributions are computed. Meucci (2009) and Meucci et al.
(2015) provide an example of risk factors decomposition when (.) is the volatility
risk measure. Meucci (2009) shows that when extracting the risk factors (uncorrelated)
using the principal component analysis, the absolute risk contribution of asset i, i ,
becomes
i (w R) = w
i2 ξi (), ∀ i = 1, . . . , N (73)
−1
where ξ() = (ξ1 (), . . . , ξ N ()) is the vector of eigenvalues of ,
w = E w
with E the eigenvectors matrix of . Meucci et al. (2015) find that when extracting
the risk factors (uncorrelated) using the minimum torsion linear transformation, the
vector of assets’ absolute risk contributions becomes
≡ ((t )−1 w) ⊗ (t w), (74)
where = (1 , . . . , N ) , t = dg(σ )π c−1 dg(σ )−1 where π is a perturbation matrix,

c is the Riccati root of asset returns (Pearson) correlation matrix ρ and ⊗ denotes the
term-by-term product.
One important limitation of the principal component analysis and the minimum
torsion linear transformation risk factors decomposition is that the extracted risk fac-
tors are economically difficult to interpret. Thus, alternative decompositions such as
Hoeffding decomposition (see Rosen and Saunders 2010), Fama-French decompo-
sition (see Carli et al. 2014; Roncalli and Weisang 2016), macroeconomic factors
decomposition (Greenberg et al. 2016; Bass et al. 2017) and style factors decom-
position (Koedijk et al. 2016a, b) were suggested. A limitation of the Fama-French,
macro-factors and style factors decompositions is that risk factors are not uncorrelated.
The factor-based RC diversification can also be sub-optimal in terms of expected
utility theory for a risk averse investor. As an illustration, consider the factor-
123
300 G. B. Koumou
based RC introduced by Meucci (2009) using principal component analysis. This

strategy does not distinguish between negative and positive (Pearson) correlation.
Consider the following extreme case. Suppose that one has a universe of two per-
fectly positive correlated assets. Assume that the factors are extracted using principal
component analysis.
2 In that case, it is straightforward to verify that the distri-
bution = w 1 ξ1 (), 0, . . . , 0 is concentrated. Now, suppose that one has a
universe
2 of two perfectly negative correlated assets. The result remains the same,
= w 1 ξ1 (), 0, . . . , 0 . It follows that the factor-based RC, using principal com-
ponent analysis, violates the correlation diversification principle, which is at the core
of expected utility theory. This result remains true when factors are extracted using
the minimum torsion linear transformation.
However, since the factor-based RC, with uncorrelated factors, is the naive factor-
based risk parity, it can be optimal with respect to minimax or maximin portfolio
optimization of Fisher et al. (2015), consequently, very useful for a risk averse investor.
6.3 Measurement
To measure the effect of the RC diversification principle, the following measure is

suggested inspired from the effective number of constituents
N 1−υ
1
υ
ENCBυ (w) = qi (w) 1−υ = (qi (w))υ , (75)
i=1
where
i (w)
qi (w) = . (76)
(w R)
Roncalli (2014) analyzes ENCBυ (w) where (.) is the volatility risk measure (effec-
tive number of correlated bets (ENCB)) and υ = 2 or υ converges to 1.
If qi (w) is defined at the level of risk factors, then ENCBυ (w) becomes the effective
number of uncorrelated bets (ENBυ (w)). Meucci (2009) analyzes ENBυ (w) where
(.) is the volatility risk measure, υ converges to 1 and factors are extracted using
principal component analysis. Meucci et al. (2015) analyze ENBυ (w) where (.) is
the volatility risk measure, υ converges to 1 and factors are extracted using minimum
torsion linear transformation.
7 Conclusion
Diversification is one of the major components of investment decision-making as well

as risk and return. However, as the 2007–2009 financial crisis revealed, the concept
remains misunderstood. Our goal in writing this paper is to correct this issue by
reviewing the concept in portfolio theory.
123
Table 3 Summary of the definition, optimality and usefulness and measurement of our four diversification
principles
Law of large numbers Definition It recommends investing a small fraction of

wealth in each of a large number of assets
Optimality and Usefulness It is optimal (i) in terms of the
mean-variance model under conditions (8)
and (9) or condition (10) of Green and
Hollifield (1992); (ii) in terms of expected
utility theory when the joint distribution of
asset returns is exchangeable (Samuelson
1967); and (iii) it is useful in the presence
of estimation errors or a high degree of
model uncertainty (see DeMiguel et al.
2009a; Bera and Park 2008; Coqueret
2015; Tu and Zhou 2011)
Measurement (i) Effective numbers of constituents
(Deguest et al. 2013; Carli et al. 2014;
Lhabitant 2017); (ii) Kullback–Leibler
divergence (Bera and Park 2008; Zhou
et al. 2013; Lhabitant 2017)
Correlation Definition It consists of allocating wealth across a
variety of assets, but exploiting the
correlation between asset returns in order
to reduce portfolio risk. The underlying
philosophy is as follows: diversification
helps to minimize both the probability of
portfolio loss and its severity, through a
multilateral insurance in which each asset
is insured by the remaining assets
Optimality and Usefulness It is optimal if and only if assets have
identical expected utility
Measurement (i) Diversification returns (Booth and Fama
1992; Willenbrock 2011; Chambers and
Zdanowicz 2014; Bouchey et al. 2012;
Qian 2012) ; (ii) Excess growth rate
(Fernholz and Shay 1982; Fernholz 2010);
(iii) Embrechts et al.’s (1999) class of
measures ; (iv) Return gap (Statman and
Scheid 2005, 2008); (v) Mitton and
Vorkink’s (2007, Equation 7, pp. 1269)
measure; (vi) Implied correlated index
(Skintzi and Refenes 2005); (vii) Rao’s
Quadratic Entropy (Carmichael et al.
2018); (viii) Tasche’s (2006) class
measures; (ix) Diversification delta
(Vermorken et al. 2012; Flores et al. 2017);
(x) Diversification ratio (Choueifaty and
Coignard 2008; Choueifaty et al. 2013)
123
302 G. B. Koumou
Table 3 continued
Sharpe (Market portfolio) Definition It is a sophisticated combination of the

market portfolio (at the level of stocks) and
the correlation diversification principles (at
the level of asset class)
Optimality and Usefulness It is optimal in terms of the mean-variance
model by definition, but under highly
sensitive and unrealistic assumptions
Measurement (i) Portfolio size (Evans and Archer 1968);
(ii) Sharpe’s (1972) measure; (iii) Barnea
and Logue’s (1973) measures
Risk contribution Definition It consists of equalizing assets’ absolute risk
contributions
Optimality and Usefulness It is optimal (i) in the case of the volatility
risk measure, but under very restrictive (if
not impossible) conditions, namely that
assets have identical Sharpe ratio and
identical (Pearson) correlations; (ii) in the
case of the volatility risk measure if and
only if the condition (69) of Kaya (2014) is
satisfied; and (iii) it is useful in risk
measure minimization in presence of
estimation error or model uncertainty
Measurement (i) Effective numbers of correlated bets
(Roncalli 2014); (ii) Effective numbers of
bets (Meucci 2009; Meucci et al. 2015)
The core of our review has focused on the four diversification principles, which are
the DNA of the existing portfolio selection rules and asset pricing theories. These four
diversification principles are: law of large numbers, correlation, capital asset pricing
model, and risk contribution diversification principles.
We have reviewed the definition of each diversification principle. We have also
reviewed their optimality with respect to expected utility theory and their usefulness
for a risk averse investor. Finally, we have explored their measurement. Table 3 sum-
marizes our findings.
We hope that this review will help to better understand the concept of diversification
in portfolio theory and to choose portfolio diversification measures more effectively.
Acknowledgements This paper is based on material from the author’s dissertation in the Department of
Economics at Université Laval. We gratefully acknowledge the financial support from FQRSC (Fonds de
Recherche de Québec-Société et Culture) [grant numbers 176559], and Canada Research Chair in Risk
Management. We also thank the anonymous referee for helpful comments.
References
Adam, A., Houkari, M., Laurent, J.-P.: Spectral risk measures and portfolio selection. J. Bank. Finance 32,
1870–1882 (2008)
123
Agapova, A., Ferguson, R., Leistikow, D., Meidan, D.: What’s the big deal about Risk Parity? J. Asset
Manag. 18, 341–346 (2017)
Al-Najjar, N.I.: Aggregation and the law of large numbers in large economies. Games Econ. Behav. 47,
1–35 (2004)
Aldous, D.J.: Exchangeability and related topics. In: École d’Été de Probabilités de Saint-Flour XIII-1983.
Springer, pp. 1–198 (1985)
Alexander, G.J., Baptista, A.M.: Economic implications of using a mean-VaR model for portfolio selection:
a comparison with mean-variance analysis. J. Econ. Dyn. Control 26, 1159–1193 (2002)
Alexeev, V., Dungey, M.: Equity portfolio diversification with high frequency data. Quant. Finance 15,
1205–1215 (2015)
Alink, S., Löwe, M., Wüthrich, M.V.: Diversification of aggregate dependent risks. Insur. Math. Econ. 35,
77–95 (2004)
Amenc, N., Goltz, F., Lodh, A., Martellini, L.: Diversifying the diversifiers and tracking the tracking error:
outperforming cap-weighted indices with limited risk of underperformance. J. Portf. Manag. 38, 72–88
(2012)
Amihud, Y., Baruch, L.: Risk reduction as a managerial motive for conglomerate mergers. ACA Trans. 12,
605–617 (1981)
Amihud, Y., Baruch, L.: Does corporate ownership structure affect its strategy towards diversification?
Strateg. Manag. J. 20, 1063–1069 (1999)
Anderson, R.C.: Paintings as an investment. Econ. Inq. 12, 13–26 (1974)
Anderson, R.M., Bianchi, S.W., Goldberg, L.R.: Will my risk parity strategy outperform? Financ. Anal. J.
68, 75–93 (2012)
Andersson, M., Krylova, E., Vähämaa, S.: Why does the correlation between stock and bond returns vary
over time? Appl. Financ. Econ. 18, 139–151 (2008)
Ang, A., Chen, J.: Asymmetric correlations of equity portfolios. J. Financ. Econ. 63, 443–494 (2002)
Ang, A., Hodrick, R.J., Xing, Y., Zhang, X.: The cross-section of volatility and expected returns. J. Finance
61, 259–299 (2006)
Ang, A., Hodrick, R.J., Xing, Y., Zhang, X.: High idiosyncratic volatility and low returns: international and
further U.S. evidence. J. Financ. Econ. 91, 1–23 (2009)
Ardia, D., Bolliger, G., Boudt, K., Gagnon-Fleury, J.-P.: The impact of covariance misspecification in
risk-based portfolios. Ann. Oper. Res. 254, 1–16 (2017)
Arnott, R.D., Hsu, J., Moore, P.: Fundamental indexation. Financ. Anal. J. 61, 83–99 (2005)
Arrow, K.J., Radner, R.: Allocation of resources in large teams. Econometrica 47, 361–385 (1979)
Artstein, Z., Hart, S.: Law of large numbers for random sets and allocation processes. Math. Oper. Res. 6,
485–492 (1981)
Artzner, P., Delbaen, F., Eber, J.-M., Heath, D.: Coherent measures of risk. Math. Finance 9, 203–228 (1999)
Asl, F.M., Etula, E.: Advancing strategic asset allocation in a multi-factor world. J. Portf. Manag. 39(59–66),
14 (2012)
Asness, C.S., Frazzini, A., Pedersen, L.H.: Leverage aversion and risk parity. Financ. Anal. J. 68, 47–59
(2012)
Aytaç, B., Mandou, C., et al.: Wine: to drink or invest in? A study of wine as an investment asset in French
portfolios. Res. Int. Bus. Finance 36, 591–614 (2016)
Bai, X., Scheinberg, K.: Alternating direction methods for non convex optimization with applications to
second-order least-squares and risk parity portfolio selection. Available at optimization-online (2015)
Bai, X., Scheinberg, K., Tutuncu, R.: Least-squares approach to risk parity in portfolio selection. Quant.
Finance 16, 357–376 (2016)
Bailey, N., Pesaran, M.H., Smith, L.V.: A multiple testing approach to the regularisation of large sample
correlation matrices. J. Econom. 208, 507–534 (2019)
Baitinger, E., Dragosch, A., Topalova, A.: Extending the risk parity approach to higher moments: is there
any value added? J. Portf. Manag. 43, 24–36 (2017)
Banz, R., De Planta, R.: Hedge funds: all that glitters is not gold-seven questions for prospective investors.
Financ. Mark. Portf. Manag. 16, 316 (2002)
Barnea, A., Logue, D.E.: Stock-market based measures of corporate diversification. J. Ind. Econ. 22, 51–60
(1973)
Bass, R., Gladstone, S., Ang, A.: Total portfolio factor, not just asset, allocation. J. Portf. Manag. 43, 38–53
(2017)
123
304 G. B. Koumou
Baur, D.G., Lucey, B.M.: Is gold a hedge or a safe haven? An analysis of stocks, bonds and gold. Financ.
Rev. 45, 217–229 (2010)
Bekkers, N., Doeswijk, R.Q., Lam, T.W.: Strategic asset allocation: determining the optimal portfolio with
ten asset classes. J. Wealth Manag. 12, 61–77 (2009)
Bender, J., Briand, R., Nielsen, F., Stefek, D.: Portfolio of risk premia: a new approach to diversification. J.
Portf. Manag. 36, 17 (2010)
Bera, A., Park, S.: Optimal portfolio diversification using the maximum entropy principle. Econom. Rev.
27, 484–512 (2008)
Bernoulli, D.: Exposition of a new theory on the measurement of risk. Econometrica 22, 23–36 (1738/1954)
Bessler, W., Holler, J., Kurmann, P.: Hedge funds and optimal asset allocation: Bayesian expectations and
spanning tests. Financ. Mark. Portf. Manag. 26, 109–141 (2012)
Bessler, W., Wolff, D.: Do commodities add value in multi-asset portfolios? An out-of-sample analysis for
different investment strategies. J. Bank. Finance 60, 1–20 (2015)
Bhansali, V., Davis, J., Rennison, G., Hsu, J.C., Li, F.: The risk in risk parity: a factor based analysis of
asset based risk parity. J. Invest. 21, 102–110 (2012)
Bianchi, R.J., Drew, M.E., Walka, A.N.: The time diversification puzzle: a survey. Financ. Plan. Res. J. 2,
12–48 (2016)
Black, F.: Capital market equilibrium with restricted borrowing. J. Bus. 45, 444–455 (1972)
Black, F., Jensen, M.C., Scholes, M., et al.: The capital asset pricing model: some empirical tests. Stud.
Theory Cap. Mark. 81, 79–121 (1972a)
Black, F., Jensen, M.C., Scholes, M., et al.: The capital asset pricing model: some empirical tests. Stud.
Theory Cap. Mark. 81, 79–121 (1972b)
Blume, M.E., Friend, I.: A new look at the capital asset pricing model. J. Finance 28, 19–34 (1973)
Blyth, S., Szigety, M.C., Xia, J.: Flexible indeterminate factor-based asset allocation. J. Portf. Manag. 42,
79–93 (2016)
Booth, D.G., Fama, E.F.: Diversification returns and asset contributions. Financ. Anal. J. 48, 26–32 (1992)
Bouchaud, J.-P., Potters, M., Aguilar, J.-P.: Missing information and asset allocation. Science & Finance
(CFM) working paper archive 500045, Science & Finance, Capital Fund Management (1997)
Bouchey, P., Nemtchinov, V., Paulsen, A., Stein, D.M.: Volatility harvesting: why does diversifying and
rebalancing create portfolio growth? J. Wealth Manag. 15, 26–35 (2012)
Boudry, W.I., Deroos, J.A., Ukhov, A.D.: Diversification benefits of reit preferred and common stock: new
evidence from a utility-based framework. Real Estate Econ. 48, 240–293 (2020)
Boudt, K., Carl, P., Peterson, B.G.: Asset allocation with conditional value-at-risk budgets. J. Risk 15, 39–68
(2013)
Bouri, E., Gupta, R., Wong, W.-K., Zhu, Z.: Is wine a good choice for investment? Pac. Basin Finance J.
51, 171–183 (2018)
Bouri, E., Jalkh, N., Molnár, P., Roubaud, D.: Bitcoin for energy commodities before and after the December
2013 crash: diversifier, hedge or safe haven? Appl. Econ. 49, 5063–5073 (2017a)
Bouri, E., Molnár, P., Azzi, G., Roubaud, D., Hagfors, L.I.: On the hedge and safe haven properties of
Bitcoin: Is it really more than a diversifier? Finance Res. Lett. 20, 192–198 (2017b)
Bouri, E.I.: Fine wine as an alternative investment during equity market downturns. J. Altern. Invest. 17,
46–57 (2015)
Boyer, B., Mitton, T., Vorkink, K.: Expected idiosyncratic skewness. Rev. Financ. Stud. 23, 169–202 (2009)
Brandtner, M.: Conditional value-at-risk, spectral risk measures and (non-) diversification in portfolio
selection problems—a comparison with mean-variance analysis. J. Bank. Finance 37, 5526–5537
(2013)
Brière, M., Drut, B., Mignon, V., Oosterlinck, K., Szafarz, A.: Is the market portfolio efficient? A new test
of mean-variance efficiency when all assets are risky. Finance 34, 7–41 (2013)
Briere, M., Oosterlinck, K., Szafarz, A.: Virtual currency, tangible return: portfolio diversification with
bitcoin. J. Asset Manag. 16, 365–373 (2015)
Brumelle, S.L.: When does diversification between two investments pay? J. Financ. Quant. Anal. 9, 473–483
(1974)
Butler, K.C., Joaquin, D.C.: Are the gains from international portfolio diversification exaggerated? The
influence of downside risk in bear markets. J. Int. Money Finance 21, 981–1011 (2002)
Byrne, P., Lee, S.: An exploration of the relationship between size, diversification and risk in UK real estate
portfolios: 1989–1999. J. Prop. Res. 20, 191–206 (2003)
123
Campbell, R., Koedijk, K., Kofman, P.: Increased correlation in bear markets. Financ. Anal. J. 58, 87–94
(2002)
Campbell, R.A.: Art as a financial investment. J. Altern. Invest. 10, 64 (2008)
Cappiello, L., Engle, R.F., Sheppard, K.: Asymmetric dynamics in the correlations of global equity and
bond returns. J. Financ. Econom. 4, 537–572 (2006)
Carli, T., Deguest, R., Martellini, L.: Improved Risk Reporting with Factor-Based Diversification Measures.
EDHEC-Risk Institute Publications (2014)
Carmichael, B., Koumou, G., Moran, K.: Unifying Portfolio Diversification Measures Using Rao’s Quadratic
Entropy (2015)
Carmichael, B., Koumou, G.B., Moran, K.: Rao’s quadratic entropy and maximum diversification indexa-
tion. Quant. Finance 18, 1017–1031 (2018)
Carrasco, M., Noumon, N.: Optimal portfolio selection using regularization. Tech. Rep., Citeseer (2011)
Cesarone, F., Colucci, S.: Minimum risk versus capital and risk diversification strategies for portfolio
construction. J. Oper. Res. Soc. 69, 183–200 (2018)
Chamberlain, G.: A characterization of the distributions that imply mean-variance utility functions. J. Econ.
Theory 29, 185–201 (1983a)
Chamberlain, G.: Funds, factors, and diversification in arbitrage pricing models. Econometrica 51, 1305–
1323 (1983b)
Chambers, D., Zdanowicz, J.S.: The limitations of diversification return. J. Portf. Manag. 40, 65–76 (2014)
Chateauneuf, A., Lakhnati, G.: From sure to strong diversification. Econ. Theory 32, 511–522 (2007)
Chateauneuf, A., Tallon, J.-M.: Diversification, convex preferences and non-empty core in the Choquet
expected utility model. Econ. Theory 19, 509–523 (2002)
Chaves, D., Hsu, J., Li, F., Shakernia, O.: Risk parity portfolio vs. other asset allocation heuristic portfolios.
J. Invest. 20, 108–118 (2011)
Chaves, D., Hsu, J., Li, F., Shakernia, O.: Efficient algorithms for computing risk parity portfolio weights.
J. Invest. 21, 150–163 (2012)
Chekhlov, A., Uryasev, S., Zabarankin, M.: Drawdown measure in portfolio optimization. Int. J. Theor.
Appl. Finance 8, 13–58 (2005)
Chollete, L., De la Pena, V., Lu, C.-C.: International diversification: a copula approach. J. Bank. Finance
35, 403–417 (2011)
Chollete, L., Heinen, A., Valdesogo, A.: Modeling international financial returns with a multivariate regime-
switching copula. J. Financ. Econom. 7, 437–480 (2009)
Choueifaty, Y., Coignard, Y.: Toward maximum diversification. J. Portf. Manag. 35, 40–51 (2008)
Choueifaty, Y., Froidure, T., Reynier, J.: Properties of the most diversified portfolio. J. Invest. Strateg. 2,
49–70 (2013)
Chow, T.-M., Hsu, J., Kalesnik, V., Little, B.: A survey of alternative equity index strategies. Financ. Anal.
J. 67, 37–57 (2011)
Christoffersen, P., Errunza, V., Jacobs, K., Langlois, H.: Is the potential for international diversification
disappearing? A dynamic copula approach. Rev. Financ. Stud. 25, 3711–3751 (2012)
Christoffersen, P., Jacobs, K., Jin, X., Langlois, H.: Dynamic dependence and diversification in corporate
credit. Rev. Finance 22, 521–560 (2017)
Chu, P.K.K.: Study on the diversification ability of fine wine investment. J. Invest. 23, 123–139 (2014)
Chua, D.B., Kritzman, M., Page, S.: The myth of diversification. J. Portf. Manag. 36, 26 (2009)
Clarke, R., De Silva, H., Thorley, S.: Minimum-variance portfolios in the US equity market. J. Portf. Manag.
33, 10 (2006)
Clarke, R.G., de Silva, H., Thorley, S.: Risk parity, maximum diversification, and minimum variance: an
analytic perspective. J. Portf. Manag. 39, 39–53 (2013)
Cont, R.: Empirical properties of asset returns: stylized facts and statistical issues (2001)
Coqueret, G.: Diversified minimum-variance portfolios. Ann. Finance 11, 221–241 (2015)
Corbet, S., Meegan, A., Larkin, C., Lucey, B., Yarovaya, L.: Exploring the dynamic relationships between
cryptocurrencies and other financial assets. Econ. Lett. 165, 28–34 (2018)
Corvalan, A.: Well Diversified Efficient Portfolios. Working Papers Central Bank of Chile 336, Central
Bank of Chile (2005)
Costa, G., Kwon, R.H.: Risk parity portfolio optimization under a Markov regime-switching framework.
Quant. Finance 19, 453–471 (2019)
Creal, D., Koopman, S.J., Lucas, A.: A dynamic multivariate heavy-tailed model for time-varying volatilities
and correlations. J. Bus. Econ. Stat. 29, 552–563 (2011)
123
306 G. B. Koumou
Creal, D.D., Tsay, R.S.: High dimensional dynamic stochastic copula models. J. Econom. 189, 335–345
(2015)
Cresson, J.E.: R 2 : a market-based measure of portfolio and mutual fund diversification. Q. J. Bus. Econ.
41, 115–143 (2002)
Cumova, D., Nawrocki, D.: Portfolio optimization in an upside potential and downside risk framework. J.
Econ. Bus. 71, 68–89 (2014)
Dagher, V.: New allocation funds redefine idea of ’Balance’. Wall Street J. 6 (2012)
Datta, D.K., Rajagopalan, N., Rasheed, A.M.A.: Diversification and performance: critical review and future
directions*. J. Manag. Stud. 28, 529–558 (1991)
De Finetti, B.: Sulla preferibilita. Giornale degli economisti e Annali di Economia 685–709 (1952)
De Giorgi, E.: A Note on Portfolio Selection under Various Risk Measures. IEW-Working Papers 122,
Institute for Empirical Research in Economics-University of Zurich (2002)
De Giorgi, E.G., Mahmoud, O.: Diversification preferences in the theory of choice. Decis. Econ. Finance
39, 143–174 (2016)
Deguest, R., Martellini, L., Meucci, A.: Risk Parity and Beyond-From Asset Allocation to Risk Allocation
Decisions. SSRN Working Paper (2013)
Dekel, E.: Asset demand without the independence axiom. Econometrica 57, 163–169 (1989)
DeMiguel, V., Garlappi, L., Nogales, F.J., Uppal, R.: A generalized approach to portfolio optimization:
improving performance by constraining portfolio norms. Manag. Sci. 55, 798–812 (2009a)
DeMiguel, V., Garlappi, L., Uppal, R.: Optimal versus naive diversification : how inefficient is the 1/N
portfolio strategy? Rev. Financ. Stud. 22, 1915–1953 (2009b)
Demsetz, R.S., Strahan, P.E.: Diversification, size, and risk at bank holding companies. J. Money Credit
Bank. 29, 300–313 (1997)
Denvir, E., Hutson, E.: The performance and diversification benefits of funds of hedge funds. J. Int. Financ.
Mark. Inst. Money 16, 4–22 (2006)
Dhaene, J., Denuit, M., Goovaerts, M.J., Kaas, R., Vyncke, D.: The concept of comonotonicity in actuarial
science and finance: applications. Insur. Math. Econ. 31, 133–161 (2002a)
Dhaene, J., Denuit, M., Goovaerts, M.J., Kaas, R., Vyncke, D.: The concept of comonotonicity in actuarial
science and finance: theory. Insur. Math. Econ. 31, 3–33 (2002b)
Dorfleitner, G., Krapp, M.: On multiattributive risk aversion: some clarifying results. RMS 1, 47–63 (2007)
Driessen, J., Laeven, L.: International portfolio diversification benefits: cross-country evidence from a local
perspective. J. Bank. Finance 31, 1693–1712 (2007)
Eeckhoudt, L., Rey, B., Schlesinger, H.: A good sign for multivariate risk taking. Manag. Sci. 53, 117–124
(2007)
Ehling, P., Heyerdahl-Larsen, C.: Correlations. Manag. Sci. 63, 1919–1937 (2016)
Elton, E.J., Gruber, M.J.: Estimating the dependence structure of share prices-implications for portfolio
selection. J. Finance 28, 1203–1232 (1973)
Elton, E.J., Gruber, M.J.: Risk reduction and portfolio size: an analytical solution. J. Bus. 50, 415–37 (1977)
Elton, E.J., Gruber, M.J., Padberg, M.W.: Simple criteria for optimal portfolio selection. J. Finance 31,
1341–1357 (1976)
Embrechts, P., Furrer, H., Kaufmann, R.: Handbook of Financial Time Series, chap. Different Kinds of
Risk, pp. 729–751 (2009)
Embrechts, P., McNeil, A., Straumann, D.: Correlation and dependence in risk management: properties and
pitfalls. In: Risk Management: Value at Risk and Beyond. Cambridge University Press, Cambridge,
pp 176–223 (1999)
Emmer, S., Kratz, M., Tasche, D.: What is the best risk measure in practice? A comparison of standard
measures. J. Risk 2, 31–60 (2015)
Emmrich, O., McGroarty, F.J.: Should gold be included in institutional investment portfolios? Appl. Financ.
Econ. 23, 1553–1565 (2013)
Engle, R.: Dynamic conditional correlation: a simple class of multivariate generalized autoregressive con-
ditional heteroskedasticity models. J. Bus. Econ. Stat. 20, 339–350 (2002)
Epstein, L.G., Tanny, S.M.: Increasing generalized correlation: a definition and some economic conse-
quences. Can. J. Econ. / Revue canadienne d’Economique 13, 16–34 (1980)
Estrada, J.: Alternatives: how? How much? why? J. Wealth Manag. 19, 49–61 (2016)
Evans, J.L., Archer, S.H.: Diversification and the reduction of dispersion: an empirical analysis. J. Finance
23, 761–767 (1968)
123
Evans, P., McMillan, D.G., McMillan, F.J.: Time-varying correlations and interrelations: firm-level-based
sector evidence. J. Asset Manag. 18, 209–221 (2017)
Fabozzi, F. J., Focardi, S. M., Jonas, C.: Investment Management: A Science to Teach or an Art to Learn?.
CFA Institute Research Foundation (2014)
Fama, E.F.: The behavior of stock-market prices. Rev. Econ. Stud. 38, 34–105 (1965)
Fama, E.F., French, K.R.: The cross-section of expected stock returns. J. Finance 47, 427–465 (1992)
Fama, E.F., French, K.R.: Common risk factors in the returns on stocks and bonds. J. Financ. Econ. 33,
3–56 (1993)
Fama, E.F., MacBeth, J.D.: Risk, return, and equilibrium: empirical tests. J. Polit. Econ. 81, 607–636 (1973)
Fan, J., Fan, Y., Lv, J.: High dimensional covariance matrix estimation using a factor model. J. Econom.
147, 186–197 (2008)
Feng, Y., Palomar, D.P.: SCRIP: successive convex optimization methods for risk parity portfolio design.
IEEE Trans. Signal Process. 63, 5285–5300 (2015)
Fernholz, R.: Diversification. Wiley, New York (2010)
Fernholz, R., Shay, B.: Stochastic portfolio theory and stock market equilibrium. J. Finance 37, 615–624
(1982)
Fishburn, P.C.: Mean-risk analysis with risk associated with below-target returns. Am. Econ. Rev. 67,
116–26 (1977)
Fisher, G.S., Maymin, P.Z., Maymin, Z.G.: Risk parity optimality. J. Portf. Manag. 41, 42–56 (2015)
Flint, E.J., Chikurunhe, F., Seymour, A.: The cost of a free lunch: Dabbling in diversification. Peregrine
Securities (2015)
Flores, Y.S., Bianchi, R.J., Drew, M.E., Trück, S.: The diversification delta: a different perspective. J. Portf.
Manag. 43, 112–124 (2017)
Fogarty, J.J.: Wine investment and portfolio diversification gains. J. Wine Econ. 5, 119–131 (2010)
Fogarty, J.J., Sadler, R.: To save or savor: a review of approaches for measuring wine as an investment. J.
Wine Econ. 9, 225–248 (2014)
Fragkiskos, A.: What is Portfolio Diversification? SSRN Working Paper (2013)
Frahm, G., Wiechers, C.: A Diversification Measure for Portfolios of Risky Assets, pp. 312–330. Palgrave
Macmillan, London (2013)
Frazzini, A., Pedersen, L.H.: Betting against beta. J. Financ. Econ. 111, 1–25 (2014)
Friend, I., Blume, M.: Measurement of portfolio performance under uncertainty. Am. Econ. Rev. 60, 561–
575 (1970)
Froot, K.A.: Hedging portfolios with real assets. J. Portf. Manag. 21, 60–77 (1995)
Fu, F.: Idiosyncratic risk and the cross-section of expected stock returns. J. Financ. Econ. 91, 24–37 (2009)
Fugazza, C., Guidolin, M., Nicodano, G.: Investing for the long-run in European real estate. J. Real Estate
Finance Econ. 34, 35–80 (2007)
Gao, X., Nardari, F.: Do commodities add economic value in asset allocation? New evidence from time-
varying moments. J. Financ. Quant. Anal. 53, 365–393 (2018)
Garcia-Feijoo, L., Jensen, G.R., Johnson, R.R.: The effectiveness of asset classes in hedging risk. J. Portf.
Manag. 38, 40 (2012)
Gatfaoui, H.: Diversifying portfolios of U.S. stocks with crude oil and natural gas: a regime-dependent
optimization with several risk measures. Energy Econ. 80, 132–152 (2018)
Geczy, C.: The new diversification: open your eyes to alternatives. J. Priv. Equity 20, 72–81 (2016)
Gibbons, M.R.: Multivariate tests of financial models. J. Financ. Econ. 10, 3–27 (1982)
Gibbons, M.R., Ross, S.A., Shanken, J.: A test of the efficiency of a given portfolio. Econometrica 57,
1121–52 (1989)
Gilli, M., Schumann, E., Di Tollo, G., Cabej, G.: Constructing 130/30-portfolios with the Omega ratio. J.
Asset Manag. 12, 94–108 (2011)
Goetzmann, W.N.: Accounting for taste: art and the financial markets over three centuries. Am. Econ. Rev.
83, 1370–1376 (1993)
Goetzmann, W.N., Kumar, A.: Equity portfolio diversification. Rev. Finance 12, 433–463 (2008)
Goetzmann, W.N., Li, L., Rouwenhorst, K.G.: Long-term global market correlations. J. Bus. 78, 1–38
(2005)
Goldfarb, D., Iyengar, G.: Robust portfolio selection problems. Math. Oper. Res. 28, 1–38 (2003)
Goltz, F., Le Sourd, V.: Does finance theory make the case forcapitalization-weighted indexing? J. Index
Invest. 2, 59–75 (2011)
123
308 G. B. Koumou
Gorton, G., Rouwenhorst, K.G.: Facts and fantasies about commodity futures. Financ. Anal. J. 62, 47–68
(2006)
Goyal, A., Santa-Clara, P.: Idiosyncratic risk matters!. J. Finance 58, 975–1007 (2003)
Green, R.C., Hollifield, B.: When will mean-variance efficient portfolios be well diversified? J. Finance 47,
1785–1809 (1992)
Greenberg, D., Babu, A., Ang, A.: Factors to assets: mapping factor exposures to asset allocations. J. Portf.
Manag. 42, 18–27 (2016)
Grinold, R.C.: Are benchmark portfolios efficient? J. Portf. Manag. 19, 34–40 (1992)
Hannah, L., Kay, J.A.: Concentration in Modern Industry: Theory, Measurement and the UK Experience.
Springer, Berlin (1977)
Harvey, C.R., Zhou, G.: Bayesian inference in asset pricing tests. J. Financ. Econ. 26, 221–254 (1990)
Haugen, R.A., Baker, N.L.: The efficient market inefficiency of capitalization-weighted stock portfolios. J.
Portf. Manag. 17, 35–40 (1991)
Hennessy, D.A., Lapan, H.E.: The use of Archimedean copulas to model portfolio allocations. Math. Finance
12, 143–154 (2002)
Hiang Liow, K.: Co-movements and correlations across Asian securitized real estate and stock markets.
Real Estate Econ. 40, 97–129 (2012)
Hinich, M.J., Patterson, D.M.: Evidence of nonlinearity in daily stock returns. J. Bus. Econ. Stat. 3, 69–77
(1985)
Hochreiter, R.: An evolutionary optimization approach to risk parity portfolio selection. In: European
Conference on the Applications of Evolutionary Computation. Springer, pp. 279–288 (2015)
Hoevenaars, R.P., Molenaar, R.D., Schotman, P.C., Steenkamp, T.B.: Strategic asset allocation with liabil-
ities: beyond stocks and bonds. J. Econ. Dyn. Control 32, 2939–2970 (2008)
Holthausen, D.M.: A risk-return model with risk and return measured as deviations from a target return.
Am. Econ. Rev. 71, 182–188 (1981)
Holton, L.: Is Markowitz wrong? Market Turmoil fuels nontraditional approaches to managing investment
risk. J. Financ. Plan. 22, 20–26 (2009)
Hsu, P.-H., Han, Q., Wu, W., Cao, Z.: Asset allocation strategies, data snooping, and the 1/N rule. J. Bank.
Finance 97, 257–269 (2018)
Ibragimov, R., Jaffee, D., Walden, J.: Diversification disasters. J. Financ. Econ. 99, 333–348 (2011)
Ibragimov, R., Walden, J.: The limits of diversification when losses may be large. J. Bank. Finance 31,
2551–2569 (2007)
Ilmanen, A., Kizer, J.: The death of diversification has been greatly exaggerated. J. Portf. Manag. 38, 15–27
(2012)
Ingersoll, J.E.: Some results in the theory of arbitrage pricing. J. Finance 39, 1021–1039 (1984)
Ingersoll, J.E.: Theory of Financial Decision Making, vol. 3. Rowman & Littlefield, Lanham (1987)
Jacobs, B.I., Levy, K.N.: 20 myths about enhanced active 120–20 strategies. Financ. Anal. J. 63, 19–26
(2007)
Jacobs, H., Müller, S., Weber, M.: How should individual investors diversify? An empirical evaluation of
alternative asset allocation policies. J. Financ. Mark. 19, 62–85 (2014)
Jahera Jr., J.S., Lloyd, W.P., Page, D.E.: The relationship between financial performance and stock market
based measures of corporate diversification. Financ. Rev. 22, 379–389 (1987)
Jiang, Y., He, S., Li, X.: A maximum entropy model for large-scale portfolio optimization. In: 2008 Inter-
national Conference on Risk Management & Engineering Management. IEEE, pp. 610–615 (2008)
Johnson, G., Ericson, S., Srimurthy, V.: An empirical analysis of 130/30 strategies: domestic and interna-
tional 130/30 strategies add value over long-only strategies. J. Altern. Invest. 10, 31–42 (2007)
Judd, K.L.: The law of large numbers with a continuum of iid random variables. J. Econ. Theory 35, 19–25
(1985)
Kang, T., Brorsen, B.W., Adam, B.D.: A new efficiency criterion: the mean-separated target deviations risk
model. J. Econ. Bus. 48, 47–66 (1996)
Kapsos, M., Christofides, N., Rustem, B.: Robust risk budgeting. Ann. Oper. Res. 266, 199–221 (2018)
Kaya, H.: The Bayesian roots of risk balancing. J. Invest. Strateg. 3, 19–39 (2014)
Kaya, H., Lee, W.: Demystifying risk parity. Available at SSRN 1987770 (2012)
Ke, J., Zhang, C.: Study on the optimization of portfolio based on entropy theory and mean-variance model.
In: 2008 IEEE International Conference on Service Operations and Logistics, and Informatics, vol. 2.
IEEE, pp. 2668–2672 (2008)
123
Kim, J.-M., Jung, H.: Directional time-varying partial correlation with the Gaussian copula-DCC-GARCH
model. Appl. Econ. 50, 4418–4426 (2018)
Kirchner, U., Zunckel, C.: Measuring portfolio diversification. arXiv preprintar. arXiv:1102.4722 (2011)
Koedijk, K., Slager, A., Stork, P.: Investing in systematic factor premiums. Eur. Financ. Manag. 22, 193–234
(2016a)
Koedijk, K., Slager, A., Stork, P.: A Trustee guide to factor investing. J. Portf. Manag. 42, 28 (2016b)
Kole, E., Koedijk, K., Verbeek, M.: Selecting copulas for risk management. J. Bank. Finance 31, 2405–2423
(2007)
Konno, H., Yamazaki, H.: Mean-absolute deviation portfolio optimization model and its applications to
Tokyo Stock Market. Manag. Sci. 37, 519–531 (1991)
Korteweg, A., Kräussl, R., Verwijmeren, P.: Does it pay to invest in art? A selection-corrected returns
perspective. Rev. Financ. Stud. 29, 1007–1038 (2015)
Kraus, A., Litzenberger, R.H.: Skewness preference and the valuation of risk assets. J. Finance 31, 1085–
1100 (1976)
Krokhmal, P., Uryasev, S., Palmquist, J.: Portfolio optimization with conditional value-at-risk objective and
constraints. J. Risk 4, 43–68 (2001)
Krusen, C., Weber, F., Weigand, R.A.: 130/30 Funds. Spec. Issues 2008, 176–185 (2008)
Lai, K.K., Yu, L., Wang, S.: Mean-variance-skewness-kurtosis-based portfolio optimization. In: First Inter-
national Multi-Symposiums on Computer and Computational Sciences (IMSCCS’06), vol. 2. IEEE,
pp. 292–297 (2006)
Le Fur, E., Outreville, J.-F.: Fine wine returns: a review of the literature. J. Asset Manag. 20, 196–214
(2019)
Leavens, D.H.: Diversification of investments. Trusts Estates 80, 469–473 (1945)
Ledoit, O., Wolf, M.: Improved estimation of the covariance matrix of stock returns with an application to
portfolio selection. J. Empir. Finance 10, 603–621 (2003)
Ledoit, O., Wolf, M.: Honey, I shrunk the sample covariance matrix : problems in mean-variance optimiza-
tion. J. Portf. Manag. 30, 110–119 (2004)
Ledoit, O., Wolf, M.: Nonlinear shrinkage of the covariance matrix for portfolio selection: Markowitz meets
Goldilocks. Rev. Financ. Stud. 30, 4349–4388 (2017)
Lehmann, B.N.: Residual risk revisited. J. Econom. 45, 71–97 (1990)
Levy, H.: Equilibrium in an imperfect market: a constraint on the number of securities in the portfolio. Am.
Econ. Rev. 68, 643–658 (1978)
Levy, M., Roll, R.: The market portfolio may be mean/variance efficient after all. Rev. Financ. Stud. 23,
2464–2491 (2010)
Lhabitant, F.-S.: Portfolio Diversification. Elsevier, Saint Louis, available from: ProQuest Ebook Central.
[1 March 2019] (2017)
Li, D., Sun, X., Wang, J.: Optimal lot solution to cardinality constrained mean-variance formulation for
portfolio selection. Math. Finance 16, 83–101 (2006)
Li, Y., Ziemba, W.T.: Rules for diversification for all risk averters. J. Econ. Bus. 42, 165–170 (1990)
Lloyd, W.P., Haney Jr., R.L.: Time diversification: surest route to lower risk. J. Portf. Manag. 6, 5–9 (1980)
Lo, A.W., Patel, P.N.: 130/30. J. Portf. Manag. 34, 12–38 (2008)
Longin, F., Solnik, B.: Is the correlation in international equity returns constant: 1960–1990? J. Int. Money
Finance 14, 3–26 (1995)
Longin, F., Solnik, B.: Extreme correlation of international equity markets. Jo. Finance 56, 649–676 (2001)
MacMinn, R.D.: A general diversification theorem: a note. J. Finance 39, 541–550 (1984)
Maillard, S.E., Roncalli, T., Teiletche, J.: The properties of equally weighted risk contribution portfolios.
J. Portf. Manag. 36, 60–70 (2010)
Malevergne, Y., Sornette, D., et al.: Testing the Gaussian copula hypothesis for financial assets dependences.
Quant. Finance 3, 231–250 (2003)
Malkiel, B.G., Xu, Y.: Risk and return revisited. J. Portf. Manag. 23, 9 (1997)
Malkiel, B.G., Yexiao, X.: Idiosyncratic Risk and Security Returns. Tech. Rep. https://www.utdallas.edu/
~yexiaoxu/IVOT_H.PDF (2006)
Mandel, B.R.: Art as an investment and conspicuous consumption good. Am. Econ. Rev. 99, 1653–63
(2009)
Mandelbrot, B.: The variation of some other speculative prices. J. Bus. 40, 393–413 (1967)
Mao, J.C.: Essentials of portfolio diversification strategy. J. Finance 25, 1109–1121 (1970)
Markowitz, H.: Portfolio selection. J. Finance 7, 77–91 (1952)
123
310 G. B. Koumou
Markowitz, H.: Portfolio Selection: Efficient Diversification of Investments, 2nd edn. Blackwell Publishers,
Inc., Malden, MA (1959)
Markowitz, H.M.: The early history of portfolio theory: 1600–1960. Financ. Anal. J. 55, 5–16 (1999)
Markowitz, H.M.: Market efficiency: a theoretical distinction and so what? Financ. Anal. J. 61, 17–30
(2005)
Mashal, R., Zeevi, A.: Beyond correlation: extreme co-movements between financial assets. Unpublished,
Columbia University (2002)
Masset, P., Weisskopf, J.-P.: When rationality meets passion: on the financial performance of collectibles.
J. Altern. Invest. 21, 66–83 (2018)
Mausser, H., Romanko, O.: Computing equal risk contribution portfolios. IBM J. Res. Dev. 58, 5–1 (2014)
Mausser, H., Romanko, O.: Long-only equal risk contribution portfolios for CVaR under discrete distribu-
tions. Quant. Finance 18, 1927–1945 (2018)
McEnally, R.W.: Time diversification: surest route to lower risk? J. Portf. Manag. 11, 24–26 (1985)
Mei, J., Moses, M.: Art as an investment and the underperformance of masterpieces. Am. Econ. Rev. 92,
1656–1668 (2002)
Meucci, A.: Managing diversification. Risk 22, 74–79 (2009)
Meucci, A., Santangelo, A., Deguest, R.: Risk budgeting and diversification based on optimised uncorrelated
factors. Risk 11, 70–75 (2015)
Miccolis, J.A., Goodman, M.: Next generation investment risk management: putting the ’Modern’ Back in
modern portfolio theory. J. Financ. Plan. 25, 44–51 (2012)
Miller, M. H., Scholes, M.: Rates of return in relation to risk: a reexamination of some recent findings. In:
Studies in the theory of capital markets, 23 (1972)
Milunovich, G.: Cryptocurrencies, mainstream asset classes and risk factors: a study of connectedness.
Aust. Econ. Rev. 51, 551–563 (2018)
Mitton, T., Vorkink, K.: Equilibrium underdiversification and the preference for skewness. Rev. Financ.
Stud. 20, 1255–1288 (2007)
Nakagawa, K., Imamura, M., Yoshida, K.: Risk-based portfolios with large dynamic covariance matrices.
Int. J. Financ. Stud. 6, 52 (2018)
Neffelli, M.: Target matrix estimators in risk-based portfolios. Risks 6, 125 (2018)
Officer, R.R.: The distribution of stock returns. J. Am. Stat. Assoc. 67, 807–812 (1972)
Oosterlinck, K.: Art as a wartime investment: conspicuous consumption and discretion. Econ. J. 127, 2665–
2701 (2017)
Page, S., Panariello, R.A.: When diversification fails. Financ. Anal. J. 74, 19–32 (2018)
Page, S., Taborsky, M.A.: The myth of diversification: risk factors vs. asset classes. J. Portf. Manag. 37,
1–2 (2011)
Pérignon, C., Smith, D.R.: Diversification and value-at-risk. J. Bank. Finance 34, 55–66 (2010)
Pesando, J.E.: Art as an investment: the market for modern prints. Am. Econ. Rev. 83, 1075–1089 (1993)
Pflug, G.C., Pichler, A., Wozabal, D.: The 1/N investment strategy is optimal under high model ambiguity.
J. Bank. Finance 36, 410–417 (2012)
Philippatos, G.C., Wilson, C.J.: Entropy, market risk, and the selection of efficient portfolios. Appl. Econ.
4, 209–220 (1972)
Poignard, B., Fermanian, J.-D.: Dynamic asset correlations based on vines. Econom. Theory 35, 167–197
(2019)
Qian, E.: Risk parity and diversification. J. Invest. 20, 119–127 (2011)
Qian, E.: Diversification return and leveraged portfolios. J. Portf. Manag. 38, 14–25 (2012)
Rehring, C.: Real estate in a mixed-asset portfolio: the role of the investment horizon. Real Estate Econ.
40, 65–95 (2012)
Renneboog, L., Spaenjers, C.: Buying beauty: on prices and returns in the art market. Manag. Sci. 59, 36–53
(2013)
Renyi, A.: On measures of entropy and information. In Proceedings of the Fourth Berkeley Symposium on
Mathematical Statistics and Probability, Volume 1: Contributions to the Theory of Statistics. University
of California Press, Berkeley, CA, pp. 547–561 (1961)
Richard, S.F.: Multivariate risk aversion, utility independence and separable utility functions. Manag. Sci.
22, 12–21 (1975)
Rockafellar, R.T., Uryasev, S.: Optimization of conditional value-at-risk. J. Risk 2, 21–41 (2000)
Rockafellar, R.T., Uryasev, S., Zabarankin, M.: Master funds in portfolio analysis with general deviation
measures. J. Bank. Finance 30, 743–778 (2006)
123
Rockafellar, R.T., Uryasev, S., Zabarankin, M.: Equilibrium with investors using a diversity of deviation
measures. J. Bank. Finance 31, 3251–3268 (2007)
Roll, R.: A critique of the asset pricing theory’s tests Part I: on past and potential testability of the theory.
J. Financ. Econ. 4, 129–176 (1977)
Roncalli, T.: Introduction to risk parity and budgeting. In: Francis, B.R.T. (ed.) CRC Financial Mathematics
Series. Chapman & Hall, Boca Raton (2014)
Roncalli, T., Weisang, G.: Risk parity portfolios with risk factors. Quant. Finance 16, 377–388 (2016)
Rosen, D., Saunders, D.: Risk factor contributions in portfolio credit risk models. J. Bank. Finance 34,
336–349 (2010)
Ross, S.A.: The arbitrage theory of capital asset pricing. J. Econ. Theory 13, 341–360 (1976)
Ross, S.A.: Adding risks: Samuelson’s fallacy of large numbers revisited. J. Financ. Quant. Anal. 34,
323–339 (1999)
Roy, A.D.: Safety first and the holding of assets. Econometrica 20, 431–449 (1952)
Rubinstein, M.: Markowitz’s “Portfolio Selection”: a fifty-year retrospective. J. Finance 57, 1041–1045
(2002)
Rudin, A.M., Morgan, J.: A portfolio diversification index. J. Portf. Manag. 32, 81–89 (2006)
Samuelson, P.A.: Risk and uncertainty: a fallacy of large numbers. Scientia 98, 108–113 (1963)
Samuelson, P.A.: General proof that diversification pays. J. Financ. Quant. Anal. 2, 1–13 (1967)
Samuelson, P.A.: Lifetime portfolio selection by dynamic stochastic programming. Rev. Econ. Stat. 51,
239–246 (1969)
Samuelson, P.A.: The “fallacy” of maximizing the geometric mean in long sequences of investing or
gambling. Proc. Natl. Acad. Sci. 68, 2493–2496 (1971)
Scheffman, D.T.: The Diversification Problem in Portfolio Models. Department of Economics Research
Reports 7318, University of Western Ontario (1973)
Scheffman, D.T.: A definition of generalized correlation and its application for portfolio analysis. Econ.
Inq. 13, 277–86 (1975)
Seneta, E.: A tricentenary history of the law of large numbers. Bernoulli 19, 1088–1121 (2013)
Shakespeare, W., Phelps, W.L.: The merchant of Venice, The Yale Shakespeare, New Haven: Yale University
Press, bibliogr. : p. [114] (c1923)
Shalit, H., Yitzhaki, S.: Mean-gini, portfolio theory, and the pricing of risky assets. J. Finance 39, 1449–1468
(1984)
Sharma, P.: Improving portfolio diversification: identifying the right baskets for putting your eggs. Manag.
Decis. Econ. 39, 698–711 (2018)
Sharpe, W.F.: Capital asset prices : a theory of market equilibrium under conditions of risk. J. Finance 19,
425–442 (1964)
Sharpe, W.F.: Risk, market sensitivity and diversification. Financ. Anal. J. 28, 74–79 (1972)
Skinner, S.J., Jackson, J.D.: American art as an investment: new evidence from an alternative approach. J.
Econ. Finance 43, 367–381 (2018)
Skintzi, V.D., Refenes, A.-P.N.: Implied correlation index: a new measure of diversification. J. Futures
Mark. Futures Options Other Deriv. Prod. 25, 171–197 (2005)
Solnik, B., Boucrelle, C., Le Fur, Y.: International market correlation and volatility. Financ. Anal. J. 52,
17–34 (1996)
Spinu, F.: An algorithm for computing risk parity weights. Available at SSRN 2297383 (2013)
Stangeland, D.A., Turtle, H.J.: Time diversification: fact or fallacy. J. Financ. Educ. 25, 1–13 (1999)
Stanton, C.: The portfolio trade deadline approaches: 60/40 needs an upgrade. Invest. Advisor 35, 18–18
(2015)
Statman, M.: Is Markowitz wrong? Investment lessons from the financial crisis. J. Portf. Manag. 40, 8–11
(2013)
Statman, M., Scheid, J.: Global diversification. J. Invest. Manag. 3, 1–11 (2005)
Statman, M., Scheid, J.: Correlation, return gaps and the benefits of diversification. J. Portf. Manag. 34,
132–139 (2008)
Stelk, S.J., Zhou, J., Anderson, R.I.: REITs in a mixed-asset portfolio: an investigation of extreme risks. J.
Altern. Invest. 20, 81–91 (2017)
Sullivan, E.J.: A.D. Roy: The Forgotten Father of Portfolio Theory, pp. 73–82 (2011)
Tang, G.Y.: How efficient is naive portfolio diversification? An educational note. Omega 32, 155–160 (2004)
Tasche, D.: Risk Contributions and Performance Measurement. Tech. Rep., Research paper, Zentrum Math-
ematik (SCA) (2000)
123
312 G. B. Koumou
Tasche, D.: Measuring sectoral diversification in an asymptotic multifactor framework. J. Credit Risk, 2
(2006)
Tasche, D.: Capital allocation to business units and sub-portfolios: the Euler principle. Papers.
arXiv:0708.2542v3 (2008)
Taylor, D., Coleman, L.: Price determinants of Aboriginal art, and its role as an alternative asset class. J.
Bank. Finance 35, 1519–1529 (2011)
Tinic, S.M., West, R.R.: Risk and return: Janaury vs. the rest of the year. J. Financ. Econ. 13, 561–574
(1984)
Tinic, S.M., West, R.R.: Risk, return, and equilibrium: a revisit. J. Polit. Econ. 94, 126–147 (1986)
Tobin, J.: Liquidity preference as behavior towards risk. Rev. Econ. Stud. 25, 65–86 (1958)
Tse, Y.K., Tsui, A.K.C.: A multivariate generalized autoregressive conditional heteroscedasticity model
with time-varying correlations. J. Bus. Econ. Stat. 20, 351–362 (2002)
Tu, J., Zhou, G.: Markowitz meets Talmud: a combination of sophisticated and naive diversification strate-
gies. J. Financ. Econ. 99, 204–215 (2011)
Tucker, M., Hlawischka, W., Pierne, J.: Art as an investment: a portfolio allocation analysis. Manag. Finance
21, 16–24 (1995)
Tutuncu, R., Koenig, M.: Robust asset allocation. Ann. Oper. Res. 132, 157–187 (2004)
Uhlig, H.: A law of large numbers for large economies. Econ. Theor. 8, 41–50 (1996)
Van der Weide, R.: GO-GARCH: a multivariate generalized orthogonal GARCH model. J. Appl. Econom.
17, 549–564 (2002)
Vaucher, B., Medvedev, A.: Efficient integration of risk premia exposures into equity portfolios. J. Asset
Manag. 18, 538–546 (2017)
Vermorken, M.A., Medda, F.R., Schroder, T.: The diversification delta: a higher-moment measure for port-
folio diversification. J. Portf. Manag. 39, 67–74 (2012)
Willenbrock, S.: Diversification return, portfolio rebalancing, and the commodity return puzzle. Financ.
Anal. J. 67, 42–49 (2011)
Woerheide, W., Persson, D.: An index of portfolio diversification. Financ. Serv. Rev. 2, 73–85 (1993)
Wright, R.: Expectation dependence of random variables, with an application in portfolio theory. Theor.
Decis. 22, 111–124 (1987)
Yu, J.-R., Lee, W.-Y., Chiou, W.-J.P.: Diversified portfolios with different entropy measures. Appl. Math.
Comput. 241, 47–63 (2014)
Zheng, Y., Zhou, M., Li, G.: Information entropy based fuzzy optimization model of electricity purchasing
portfolio. In: 2009 IEEE Power and Energy Society General Meeting. IEEE, pp. 1–6 (2009)
Zhou, G.: Small sample tests of portfolio efficiency. J. Financ. Econ. 30, 165–191 (1991)
Zhou, R., Cai, R., Tong, G.: Applications of entropy in finance: a review. Entropy 15, 4909–4931 (2013)
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.
Gilles Boevi Koumou is currently a Lecturer at Université de Québec à Chicoutimi (Canada). Before, he
was a Post-Doctoral research associate at the Canada Research Chair in Risk Management at the HEC
Montréal (Department of Finance). He received his Ph.D. degree in Financial Economics from Université
Laval (Canada). His research interests are: financial decision theory (particularly portfolio theory, includ-
ing asset pricing) and financial machine learning.
123

Diversification and Portfolio Theory: A Review: Gilles Boevi Koumou

Uploaded by

Copyright:

Available Formats

Diversification and Portfolio Theory: A Review: Gilles Boevi Koumou

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Diversification and Portfolio Theory: A Review: Gilles Boevi Koumou

Uploaded by

Copyright:

Available Formats

Financial Markets and Portfolio Management (2020) 34:267–312

Diversification and portfolio theory: a review

Gilles Boevi Koumou1,2

Published online: 4 June 2020

Keywords Diversification · Portfolio theory · Law of large numbers · Correlation ·

JEL Classification G1 · G11 · G12

B Gilles Boevi Koumou

aging partner at Quoniam, a quantitative asset management firm based in Frankfurt,

standard deviations, while also providing intuitive measures of diversification’s

From our viewpoint, diversification is not dead, but it is misunderstood. It is there-

Fragkiskos (2013) Information theory Herfindahl index (Woerheide and Persson

Higher-moment measures (i) Diversification delta (Vermorken et al.

is (R(w)), with μ = (μ1 , . . . , μ N ) ; in particular, (R(w)) ≡ σ 2 (w) = w w in

R1 R2 ⇐⇒ E (u(R1 )) ≥ E (u(R2 )) . (1)

where τ is the coefficient of risk aversion. The mean-variance efficient portfolio,

R1 M V R2 ⇐⇒ U M V (R1 ) ≥ U M V (R2 ). (3)

We recall that M V is equivalent to if each asset’s returns are normally distributed

In this subsection, we recall some concepts related to portfolio diversification required

2.2.1 Exchangeable random variables

Definition 1 (Exchangeability) The random variables R1 , . . . , R N are said to be

A well-known example of an exchangeable sequence of random variables is an inde-

2.2.2 Correlation aversion

with strict inequality for at least one pair (y, z) ∈ R × R.

A decision maker is said to be a correlation lover if ∂ ∂v(y,z)

When v(y, z) is not twice continuously differentiable, condition (4) is replaced by

v(y ∨ z) + v(y ∧ z) ≤ v(y) + v(z), ∀ (y, z) ∈ R × R (6)

where y ∨ z is the componentwise maximum of y and z, y ∧ z is the componentwise

3 Law of large numbers diversification principle

The LLN diversification principle is a well-established risk management technique in

It is advisable to divide goods which are exposed to some normalsize danger

More specifically, the LLN diversification principle recommends investing a small

3.2 Optimality and usefulness

and the payoff, = ω R, on every hedge position ω such that ω 1 = 0 satisfies

3.3.1 Effective number of constituents

The ENC is a class of measures of naive diversification analyzed by Deguest et al.

ENC1 (w) = exp (SH(w)) (14)

In the case where υ converges to 1, the relation (15) becomes

BM1 (w) = ln (ENC1 (w)) . (16)

REυ (w) = ln (ENCυ (w)) . (17)

3.3.2 Kullback–Leibler divergence

D(p|q) quantifies the difference between probability distributions p and q.

4 Correlation diversification principle

Our second diversification principle is the correlation diversification principle. The

The correlation diversification principle is also a well-established risk management

The adequacy of diversification is not thought by investors to depend solely on

More specifically, the correlation diversification principle consists of allocating

The term in parentheses, G(w) − G(δ i ), measures the benefit of diversification to

4.2 Optimality and usefulness

The function u(.) can also be viewed as a multivariate utility function ũ : R N

The partial derivative of ũ (R1 , . . . , R N ) with respect to Ri and R j gives

1, . . . , N . As a result, the correlation diversification is optimal in terms of expected

The lack of (Pearson) correlation to identify the presence of correlation diversi-

4.3.1 Diversification returns

introduced by Booth and Fama (1992),

E ln(w) = E ln(1 + w R) − w E ln(1 + R), (27)

where ln(1 + R) = (ln(1 + R1 ), . . . , ln(1 + R N )) is the vector column of assets’

ln(1 + μG ) = E ln(1 + R), (28)

E ln(w) can be rewritten in terms of μG

E ln(w) = ln(1 + μG (w)) − w ln(1 + μG ), (29)

is (R(w)), with μ = (μ1 , . . . , μ N ) ; in particular, (R(w)) ≡ σ 2 (w) = w w in

An intuitive interpretation can be provided to EFK (w) when (.) is homogeneous

where (.) is Shannon entropy denoted by H (.); more formally

The CAPM predicts the mean-variance

Formally, assume that portfolio risk (w R) can be additively decomposed

When the risk measure (.) is assumed to be homogeneous of degree κ ∈ R, the

min (w R) (65)

where = (1 , . . . , N ) , t = dg(σ )π c−1 dg(σ )−1 where π is a perturbation matrix,