The role of time‐varying return forecasts for improving international diversification benefits

• Published date: 01 July 2017
• Date: 01 July 2017
• DOI: http://doi.org/10.1002/ijfe.1578

RESEARCH ARTICLE

The role of time‐varying return forecasts for improving

international diversification benefits

Maria del Mar Miralles‐Quiros

1

| Jose Luis Miralles‐Quiros

2

1

Financial Economics, Faculty of

Economics, University of Extremaduras, Av.

Elvas s/n, Badajoz 06071, Spain

2

Financial Economics, Faculty of

Economics, University of Extremaduras,

Badajoz, Spain

Correspondence

Maria del Mar Miralles‐Quiros, Financial

Economics, Faculty of Economics,

University of Extremadura, Av. Elvas s/n,

Badajoz 06071, Spain.

Email: marmiralles@unex.es

JEL Classification: G10; G11; G14

Abstract

The aim of this study is to provide empirical evidence of the international diversifi-

cation benefits obtained employing not only time‐varying volatility forecasts but

also time‐varying return forecasts from a multivariate approach that considers the

dynamic relationships in return series as well as in volatilities and correlations. To

that end, instead of using market indexes from different investment areas, we

employ exchange trade funds actively traded on the New York Stock Exchange in

recent years. It avoids nonsynchronous problems as well as allowing us to allocate

internationally on a daily basis for which this approach is especially appropriate.

Our overall results show that using this technique, it is possible to obtain economic

gains and outperform the common benchmark strategies, even when the costs asso-

ciated with the daily rebalance of each portfolio are taken into account.

KEYWORDS

developed markets , emerging markets, out‐of‐sample evaluation, returnpredictability, volatilitytiming

1|INTRODUCTION

The classical theory of optimal portfolio selection proposed

by Markowitz (1952) states that the correlation structure

across assets is a key feature of the optimization problem

because it determines the riskiness of the investment position.

For that reason, investors usually choose to diversify interna-

tionally and further reduce their portfolio risk assumptions.

However, it is well known that the correlation structure

across assets varies over time and researchers have proposed

various multivariate GARCH specifications to model the

dynamic dependence structure of multivariate time series.

One of the most popular multivariate specifications is the

dynamic conditional correlation (DCC) model proposed by

Engle (2002), which describes the correlation dynamics

among different asset classes and markets.

This line of research has important implications for

investors because it gives them the opportunity to solve the

classical allocation problem with forward‐looking correlation

forecasts obtained from this dynamic correlation model.

Some examples of these empirical applications for

international diversification purposes are those of Jondeau

and Rockinger (2012) and Christofersen et al. (2014). How-

ever, these previous works have only analysed the implica-

tions of volatility timing on international diversification

decisions and have avoided the prediction of expected

returns. The reason for this choice is that there is little infor-

mation content in past variables regarding future returns in

the short term. More precisely, Jondeau and Rockinger

(2012) argue that for such horizons, estimation uncertainty

of predictive regressions may also do more harm than good.

Nevertheless, some authors report evidence that stock

returns are to some extent predictable (DeMiguel, Nogales,

& Uppal, 2014; Marquering & Verbeek, 2004; Pesaran &

Timmerman, 1995; Pesaran & Timmernan, 2000; Solnik,

1993). Moreover, information transmission across markets

has been widely studied (In, Kim, Yoon, & Viney, 2001;

Liu & Pan, 1997; Shin & Soydemir, 2010). More precisely,

these studies have focused on the short‐term interdependence

among stock markets employing vector autoregressive (VAR)

models to capture serial dependence across markets. This

model allows expected returns over every market index to

Received: 13 October 2015 Revised: 12 April 2016 Accepted: 14 March 2017

DOI: 10.1002/ijfe.1578

Int J Fin Econ. 2017;22:201–215. Copyright © 2017 John Wiley & Sons, Ltd.wileyonlinelibrary.com/journal/ijfe 201