Corporate Hedging and the High Idiosyncratic Volatility Low Return Puzzle

• DOI: http://doi.org/10.1111/irfi.12109
• Published date: 01 September 2017
• Author: Victor Fang,Hong Feng Zhang,Michael T. Chng,Vincent Xiang
• Date: 01 September 2017

Corporate Hedging and the High

Idiosyncratic Volatility Low Return

Puzzle*

MICHAEL T. CHNG

†,‡

,VICTOR FANG

‡

,VINCENT XIANG

‡

AND

HONG FENG ZHANG

‡

†

International Business School Suzhou, Xiˈan Jiaotong-Liverpool University, Suzhou,

Jiangsu, China and

‡

Department of Finance, Deakin Business School, Deakin University, Burwood,

Australia

ABSTRACT

The literature offers various explanations to either support or refute the Ang

et al. (2006) high idiosyncratic volatility low return puzzle. Fu (2009) finds a

significantly positive contemporaneous relation between return and exponen-

tial generalized autoregressive conditional heteroskedastic idiosyncratic vola-

tility. We use corporate hedging to shed light on this puzzle. Conceptually,

idiosyncratic volatility matters to investors who face limits to diversification.

But limits to diversification become less relevant for firms that consistently

hedge. We confirm the main finding in Fu (2009), but only for firms that do

not consistently hedge. For firms that adopt a consistent hedging policy,

idiosyncratic volatility, whether contemporaneous or lagged, is insignificant

in Fama–MacBeth regressions, controlling for size, book-to-market, momen-

tum, liquidity, and industry effects.

JEL Codes: G12; G32

I. INTRODUCTION

In portfolio theory, firm-specific risk is diversifiable, hence irrelevant to expected

return. In the capital asset pricing model, only systematic risk matters, and

firm-specific risk cannot explain cross-sectional stock return. Firm-specific risk

is commonly measured as the idiosyncratic volatility σ

εi

of residual stock return

ε

it

from a factor-return specification. In recent years, the idiosyncratic volatility

puzzle has drawn substantial research attention. This puzzle exists at two levels.

First is an empirically nontrivial relation between σ

εi

and stock return r

i

. The

literature offers various explanations based on limits to diversification, for

* We are grateful to the editor Allaudeen Hameed, John Wei, Yuelan Chen and two anonymous ref-

erees for comments on earlier drafts. Earlier versions were presented at LaTrobe University,University

of Adelaide, Fudan University, Taiwan National University ofScience and Technology, AsianFA meet-

ing, and SFM conference. The authors retain fullproperty rights to all errors.

© 2017 International Review of Finance Ltd. 2017

International Review of Finance, 17:3, 2017: pp. 395–425

DOI: 10.1111/irfi.12109

example, transaction cost in Brennan (1975), search cost in Merton (1987), over-

confidence in Odean (1999), home or geographical biases in Grinblatt and

Keloharju (2001) and Huberman (2001), and behavioral biases in Barberis and

Huang (2001). Under-diversified portfolios contain nontrivial σ

εi

, for which

investors require a return to bear. However, Campbell et al. (2001) and Goyal

and Santa-Clara (2003) argue it is not firm-specific risk per se that is priced. Rather,

σ

εi

is a proxy for latent risks that are not captured by Fama–French factors.

Regardless of the cause, the various explanations posit a positive cross-sectional

relation between σ

εi

and r

i

.

Second is the high idiosyncratic volatility low return puzzle documented in

Ang et al. (2006). They find that stocks with high σ

εi

in month texhibit low

returns in month t+ 1, which cannot be explained by Fama–French factors,

momentum, or liquidity. This finding contradicts both risk-based and behavioral

explanations of a positive relation between σ

εi

and r

i

. The Ang et al. (2006) puzzle

hints at a negative first-order cross-serial covariance betwee n σεit1and r

it

. Huang

et al. (2010) attribute the significance of σεit1to the omission of the relevant

variable problem in the Fama–MacBeth regressions. The highest σ

εi

portfolio

contains the largest proportion of both recent winner and loser stocks. These

stocks have the greatest tendency to exhibit short-term return reversal.

1

The

return reversal effect is being absorbed by σεit1. By including r

it 1

, Huang et al.

(2010) show that σεit1is no longer significant.

Other papers have offered various plausible explanations for the Ang et al.

(2006) puzzle. Peterson and Smedema (2011) find a robust positive relation

between return and expected idiosyncratic volatility after controlling for realized

idiosyncratic volatility. Chabi-Yo (2011) utilize higher moments to explain the

idiosyncratic volatility puzzle. The paper finds that, after controlling for a firm

ˈs co-skewness, there is no significant relation between idiosyncratic volatility

and expected stock return. Rubin and Smith (2011) dissect the explanatory power

of idiosyncratic volatility into cross-sectional and time-series dimensions. They

argue that the mixed findings in the literature are because of the lack of a suitable

proxy for σ

εi

that can adequately capture both dimensions.

In a follow-up study, Ang et al. (2009) document the idiosyncratic volatility

puzzle in 22 developed markets other than the USA. However, studies on

Asia-Pacific markets yield mixed findings. Eun and Huang (2007) document a sig-

nificantly negative relation between idiosyncratic volatility and average stock

return. Nartea et al. (2011); Nartea and Wu (2013) find a positive relationship

between idiosyncratic volatility and stock returns for Malaysia, Singapore,

Thailand, and Indonesia, but no significant relationship for the Philippines and

Hong Kong. They caution that the idiosyncratic volatility puzzle documented

in developed markets may not apply to emerging stock markets. In a recent study,

Liu and Di Lorio (2015) find a positive relation between idiosyncratic volatility

1 In value-weighted portfolios, winner stocks receive heavier weighting than loser stocks, such

that return reversal by winner stocks overshadow that of loser stocks. This causes low subse-

quent value-weighted return.

International Review of Finance

© 2017 International Review of Finance Ltd. 2017396