Betas versus characteristics: A practical perspective

• Published date: 01 November 2020
• Author: Gregory Nazaire,Maria Pacurar,Oumar Sy
• Date: 01 November 2020
• DOI: http://doi.org/10.1111/eufm.12263

Eur Financ Manag. 2020;26:1385–1413. wileyonlinelibrary.com/journal/eufm © 2020 John Wiley & Sons Ltd.

|

1385

DOI: 10.1111/eufm.12263

ORIGINAL ARTICLE

Betas versus characteristics: A practical

perspective

Gregory Nazaire |Maria Pacurar |Oumar Sy

Rowe School of Business, Dalhousie

University, Halifax, NS, Canada

Correspondence

Oumar Sy, Rowe School of Business,

Dalhousie University, K.C. Rowe

Management Building, 6100 University

Avenue, PO Box 15000, Halifax,

NS B3H 4R2, Canada.

Email: oumar.sy@dal.ca

Abstract

We apply a new dummy‐variable method to examine

which factor exposures (betas) and characteristics

provide independent information for US stock

returns in the context of the multifactor models of

Hou, Xue, and Zhang and of Fama and French.

We find that betas related to market, size, value,

momentum, investment, and profitability factors are

not priced. In contrast, firm characteristics related to

size, value, investment, and profitability have

significant and independent explanatory power,

suggesting that they are important in determining

expected returns. Finally, the cross‐sectional effect of

momentum is subsumed when the return on equity is

factored in.

KEYWORDS

betas, characteristics, dummy‐variable model, asset allocation,

multifactor models

JEL CLASSIFICATION

G10; G11; G12

EUROPEAN

FINANCIAL MANAGEMENT

We thank the editor (John Doukas) and an anonymous referee for very helpful comments and suggestions. We also thank

Philippe Bertrand, Kais Bouslah, the late Jean‐Guy Degos, Iraj Fooladi, Xiaoji Lin, Mohammad Rahaman, Keke Song, and

conference and seminar participants at the 2016 Financial Management Association Annual Meeting, the 2016 Midwest

Finance Association Annual Meeting, the 2016 Global Finance Association Conference, the IAE Bordeaux, the Rowe School

of Business, and the Sobey School of Business for their comments, as well as Ken French and Lu Zhang for sharing their data

on factors. All errors remain our own. Oumar Sy acknowledges financial support from the Rowe Research Award program.

During this research, Oumar Sy has also taught at the Sobey School of Business, Saint Mary’s University. The authors can be

contacted via email: gregory.nazaire@dal.ca (G. Nazaire), maria.pacurar@dal.ca (M. Pacurar), oumar.sy@dal.ca (O. Sy).

1|INTRODUCTION

Financial economists have put considerable effort into understanding the determinants of

expected stock returns. More than four decades of intensive research have identified many

patterns in stock returns, usually known as anomalies (see Harvey, Liu, & Zhu, 2016; Hou,

Xue, & Zhang, 2019). Traditionally, two major competing stories have been advanced to explain

the anomalies, leading to the well‐known “betas versus characteristics”debate. One strand of

the literature championed by Fama and French (1993,1996,2016) argues that risk explains the

anomalies, and proposes multifactor models in an attempt to capture them. Another branch of

the literature (e.g. Daniel & Titman, 1997; Lakonishok, Shleifer, & Vishny, 1994) maintains that

the anomalies are driven by mispricing.

It is now well established that the two explanations are neither exhaustive nor mutually

exclusive. Berk (1995) argues that size‐related characteristics are likely to help explain expected

returns unless the risk‐based model is perfectly specified. Gomes, Kogan, and Zhang (2003)

build a dynamic general equilibrium economy that links the true conditional market beta to

firm size and book‐to‐market ratio and find that these characteristics can rationally explain the

cross‐section of stock returns. However, Daniel, Hirshleifer, and Subrahmanyam (2005) show

that imperfectly rational asset pricing models have implications qualitatively similar to those of

rational asset pricing models, while Ferson, Sarkissian, and Simin (1999) find that an attribute‐

sorted factor can appear to behave as a genuine risk factor even if the underlying sorting

attribute is completely unrelated to risk. Fama and French (2015) offer that the relation be-

tween average returns and characteristics such as book‐to‐market, investment, and profitability

can be explained in the context of the dividend discount model. Therefore, the relation between

these characteristics and returns will be observed regardless of whether prices are rational or

irrational. Lin and Zhang (2013) challenge the traditional interpretation of characteristic‐based

factors as risk factors and maintain that the domination of characteristics over betas in horse

races does not necessarily mean that financial assets are mispriced. If betas are measured with

errors in an environment where the investment capital asset pricing model (CAPM;

Zhang, 2017) holds, this empirical domination of the characteristics is likely to hold. More

recently, Kozak, Nagel, and Santosh (2018) have pointed out that both betas and characteristics

are likely to explain expected returns whether spreads in expected returns are generated as

rational responses to risk or by mispricing, such that studying empirical relations between

returns and betas versus characteristics may not tell us much about the fundamental

determination of prices.

Although the theoretical demarcation of betas and characteristics is complex and con-

troversial, the question of which betas and characteristics are of practical value for designing

and implementing superior investment strategies remains. In this paper, we examine which of

the popular betas and characteristics have pure independent information about average returns

using a new decomposition method. In doing so, we provide new insights into what Cochrane

(2011) called in his American Finance Association presidential address the “multidimensional

challenge.”In particular, we present some evidence on Cochrane’s first set of questions

(p. 1060): “which characteristics really provide independent information about average returns?

Which are subsumed by others?”

To the best of our knowledge, this is the first study to investigate this challenge from an asset‐

allocation angle. We contribute to the literature by adapting the dummy‐variable regression

method of Heston and Rouwenhorst (1994), which is widely used in the international asset‐

allocation literature to assess the benefits of geographical and industrial diversification strategies

1386

|

EUROPEAN

FINANCIAL MANAGEMENT

NAZAIRE ET AL.