Are Smart Beta strategies suitable for hedge fund portfolios?

• Author: Giovanni Zambruno,Asmerilda Hitaj
• Date: 01 April 2016
• Published date: 01 April 2016
• DOI: http://doi.org/10.1016/j.rfe.2016.03.001

Are Smart Beta strategies suitable for hedge fund portfolios?

Asmerilda Hitaj ⁎, Giovanni Zambruno

Dipartimentodi Statistica e MetodiQuantitativi, Universityof Milano Bicocca, Via Bicoccadegli Arcimboldi 8, 20126Milano, Italy

abstractarticle info

Articlehistory:

Received29 May 2015

Accepted21 March 2016

Availableonline 14 April 2016

In the equity context differentSmart Beta strategies (such as the equally weighted, globalminimum variance,

equal risk contribution and maximum diversified ratio) have been propos ed as alternatives to the ca p-

weighted index. These newapproaches have attracted the attention of equity managers as differentempirical

analyses demonstrate the superiority of these strategies with respect to cap-weighted and to strategies that

consideronly mean and variance.In this paperwe focus our attentionto hedge fund index portfoliosand analyze

if theresults reported in theequity frameworkare still valid. We considerhedge fund index andequity portfolios,

the approaches used for p ortfolio selection are the four ‘Smart Beta’strategie s, mean–variance and mean–

variance–skewness.In the two latter approaches the Taylor approximationof a CARA expected utility function

and the Polynomial Goal Prog raming (PGP) have been used. Th e obtained portfolios are an alyzed in the

in-sampleas well as in the out-of-sample perspectives.

© 2016 ElsevierInc. All rights reserved.

Keywords:

Highermoment portfolio selection

Smart Betastrategies

Expectedutility

PolynomialGoal Programming

1. Introduction

Since the pioneering work of Markowitz (1952), based on mean–

variance framework, several po rtfolio selection models hav e been

proposed. Especially in the equity worlddifferent strategies have been

suggested in substitution to cap italization-weighted or value-

weighted indexes. Some of these are: equally weighted (EW), global

minimumvariance (GMV),equal risk contribution(ERC) and maximum

diversified (MD) which have raised great interest both from market

practitioners and academic researchers. These approaches are known

as risk-based or ‘Smart Beta’strategies and are assumed to be robust

since they rely at most on the estima tion of the covariance matrix

ignoring theexpected returns.

Different empiricalanalyses have been performed in caseof equity

portfolios, concluding the superiority of these strategies with respect

to those based on mean and variance. DeMiguel, Garlappi, and Uppal

(2009) have demonstrated through emp irical analysis that the EW

portfolio has never been outperformed systematically by the other 13

strategies considered in their paper. Choueifaty and Coignard (2008)

proposed the maximum diversified ra tio and empirically found

that the MDP is more efficient ex post compared to the market

capitalization-weighted b enchmark, the MV and the EW. Maillard ,

Roncalli, and Teiletche (201 0) compared the GMV, the EW and

the ERC in an asset/sector alloca tion context and concluded that the

ERC approachmight be considered a good trade-offbetween the other

two methods in terms of absolute ri sk level, risk budgeting and

diversification.

Accordingto these empiricalworks the risk-basedstrategies seem to

work well in the equity context. But equity is just one class of invest-

ments, and one may wonder how these risk-based strategiesperform

if we consider a different type of inve stment vehicle. An important

investment instrument are the hedge funds, which are often included

in investors' portfolios, as they are characterized by high returns and

low correlation with market re turns. Therefore the objecti ve of

this paper is to analyze how these risk-based strate gies perform in

case of hedgefund portfolios. To our knowledge, no empiricalresearch

exists addressing this question. It is well kn own that hedge funds

arecharacterized byhigh skewness and kurtosis,and differentempirical

analysis have demonstrated th at a three or four-moment portfo lio

allocation is better than a two-mom ent portfolio allocation (see

Athayde & Flores, 2002; Jondeau, Poon, & Rockinger, 2007; Martellini

& Ziemann, 2010; Hitaj, Martellini, & Zambruno, 20 12; Hitaj &

Mercuri, 2013b,etc.).

Different methods have been proposed to incorp orate higher

momentsin portfolio allocation.Athayde and Flores(2002) constructed

the efficient frontier based on the first four moments of the portfolio

return distribution. Jondeau et al. (2007)worked on the Taylor expan-

sion of the expected utilityfunction truncated at the fourth order (see

among others Martellini & Zieman n, 2010; Hitaj & Mercuri, 2013a ,

etc.).In the empiricalpart of this work we considerthe Taylorexpansion

stopped at the second order (referringto the mean–variance (EU-MV)

portfolio) or at the third order (refe rring to the mean–variance–

skewness (EU-MVS) portfolio) for different levels of riskaversion.

Davies, Kat, and Lu (2009) provid e a financial application of the

PolynomialGoal Programing (PGP) approach.This concept generalizes

Reviewof Financial Economics 29 (2016)37–51

⁎Correspondingauthor.

E-mailaddresses:asmerilda.hitaj1@unimib.it(A.Hitaj), giovanni.zambruno@unimib.it

(G. Zambruno).

http://dx.doi.org/10.1016/j.rfe.2016.03.001

1058-3300/©2016 Elsevier Inc. All rightsreserved.

Contents listsavailable at ScienceDirect

Review of Financial Economics

journal homepage: www.elsevier.com/locate/rfe