Idiosyncratic momentum and the cross‐section of stock returns: Further evidence

• Author: Qi Lin
• DOI: http://doi.org/10.1111/eufm.12247
• Published date: 01 June 2020
• Date: 01 June 2020

Eur Financ Manag. 2020;26:579–627. wileyonlinelibrary.com/journal/eufm © 2019 John Wiley & Sons Ltd.

|

579

DOI: 10.1111/eufm.12247

ORIGINAL ARTICLE

Idiosyncratic momentum and the cross‐section

of stock returns: Further evidence

Qi Lin

School of Finance, Zhejiang University of

Finance and Economics, Hangzhou,

China

Correspondence

Qi Lin, School of Finance, Zhejiang

University of Finance and Economics,

18 Xueyuan Street, 310018 Hangzhou,

China.

Email: qlin_sf@zufe.edu.cn

Funding information

National Natural Science Foundation of

China, Grant/Award Number: 71703142

Abstract

In this article, we evaluate the profitability and

economic source of the predictive power of the

idiosyncratic momentum effect, by using five popular

asset pricing models to construct the idiosyncratic

momentum. We show that all five idiosyncratic

momentum strategies produce similar return predict-

ability and consistently outperform the conventional

momentum strategy in the cross‐sectional pricing of

equity portfolios and individual stocks. This positive

effect of idiosyncratic momentum on returns is

consistent with the investment capital asset pricing

model (CAPM). Further analysis reveals that the firm‐

level idiosyncratic momentum effect cannot extend to

the aggregate stock market.

KEYWORDS

asset pricing, idiosyncratic momentum, investment CAPM,

market return predictability

JEL CLASSIFICATION

G11; G12; G17

1

|

INTRODUCTION

Momentum, the notion that previous winners in the stock market continue to outperform

previous losers, is a pervasive anomaly in asset prices. Although momentum profits have been

EUROPEAN

FINANCIAL MANAGEMENT

The author is grateful to the editor, John Doukas, and an anonymous referee for detailed and insightful comments and

suggestions. The author also gratefully acknowledges financial support from the National Natural Science Foundation

of China (Grant Number: 71703142).

widely documented in both US and international markets,

1

they exhibit significant exposures to

systematic risk factors (e.g., Grundy & Martin, 2001). Gutierrez and Prinsky (2007) find that a

momentum strategy that defines winners and losers based on firm‐specific abnormal returns

significantly improves the performance of a conventional momentum strategy that ranks stocks

on raw returns.

2

Further, Blitz, Huij, and Martens (2011) show that a momentum strategy

constructed based on idiosyncratic returns estimated using the Fama and French (1993) three‐

factor model almost doubles risk‐adjusted profits compared with a strategy associated with

conventional momentum, due in large part to lower return variability. In their pioneering

study, the authors attribute the success of idiosyncratic momentum to the isolation of dynamic

exposures to Fama and French's (1993) three factors, as momentum loads positively (negatively)

on these factors when they have positive (negative) returns during the formation period of the

strategy. Blitz, Hanauer, and Vidojevic (2017), Chang, Ko, Nakano, and Rhee (2018), and Lin

(2019) document that the idiosyncratic momentum strategy continues to yield substantial

profits beyond the US market, including those markets where conventional momentum is

found to be ineffective.

3

Given the importance of the construction of the idiosyncratic returns, a natural question is

whether the choice of asset pricing models to estimate idiosyncratic returns affects the

profitability of the idiosyncratic momentum. In this study, we first re‐evaluate the relation

between idiosyncratic momentum and cross‐sectional stock returns and explore whether

idiosyncratic momentum has greater predictive power than conventional momentum. We

extend Blitz et al.'s (2011) approach and construct the idiosyncratic momentum measures

(iMOM) by estimating idiosyncratic returns from five popular asset pricing models: the capital

asset pricing model of Sharpe (1964) and Lintner (1965) (CAPM); the three‐factor model of

Fama and French (1993) (FF3); the four‐factor model in Carhart (1997) that augments FF3 with

a momentum factor (C4); the five‐factor model of Fama and French (2015) (FF5); and the

q

‐factor model of Hou, Xue, and Zhang (2015) (Q4). We find three primary results.

First, although the conventional momentum strategy produces a slightly greater, yet

insignificant, winner‐minus‐loser spread return on average than strategies sorted by

idiosyncratic momentum, including the strategy formed on

i

MOMFF3, idiosyncratic momentum

strategies have significantly lower volatilities and yield significantly greater Sharpe ratios than

the conventional momentum strategy. Second, we provide new evidence that the performance

of the idiosyncratic momentum strategies is insensitive to the asset pricing models used to

construct them. For all five idiosyncratic momentum winner‐minus‐loser portfolios, the average

excess returns and alphas are consistently positive and the associated

t

‐statistics pass the higher

hurdle rate of three, as suggested by Harvey, Liu, and Zhu (2016), which indicates that the

idiosyncratic momentum effect is unlikely to be due to chance or data snooping. The

mean–variance spanning test of Kan and Zhou (2012) also confirms that idiosyncratic

momentum is not simply a manifestation of conventional momentum and well‐known risk

factors. More importantly, all three tests from Andrews (1991), Andrews and Monahan (1992),

and Ledoit and Wolf (2008) cannot reject the null that there is no difference in Sharpe ratios

between idiosyncratic momentum spread portfolios. Third, model Q4 outperforms the

competing models in capturing different versions of iMOM, resulting in the smallest alphas,

1

See, for example, Jegadeesh and Titman (1993, 2001), Rouwenhorst (1998), Moskowitz and Grinblatt (1999), Johnson (2002), Griffin, Ji, and Martin (2003),

Chui, Titman, and Wei (2010), Fama and French (2012), Asness, Moskowitz, and Pedersen (2013), Goyal and Wahal (2015), and Li (2017).

2

Throughout the article, we refer to Jegadeesh and Titman's (1993) momentum strategy as conventional momentum strategy.

3

Blitz et al. (2011) also refer to idiosyncratic momentum as residual momentum.

580

|

EUROPEAN

FINANCIAL MANAGEMENT

LIN

information ratios, and aggregate pricing errors and yielding the highest

p

‐values in the

Gibbons, Ross, and Shanken (1989) test. These findings echo Hou et al. (2015, 2017), who show

that model Q4 outperforms models FF3 and FF5 in accommodating a wide range of anomalies

at the portfolio level. We also find that our results are insensitive to the “scaling”of

idiosyncratic momentum.

Given the similar return predictability among iMOM and the strong power of idiosyncratic

momentum in predicting the cross‐section of stock returns, we further investigate several

potential explanations for the economic source of its predictability.

The similar performance among idiosyncratic momentum strategies appears to be striking,

as Blitz et al. (2011) argue that the success of idiosyncratic momentum is due to the isolation of

dynamic factor exposures. While we find consistent evidence that conventional momentum has

substantial time‐varying exposures not only to the market factor but also to the size and value

factors, as in Blitz et al. (2011), the similar profitability across idiosyncratic momentum

strategies suggests that hedging out time‐varying exposures to the latter two factors cannot

improve the strategy's performance significantly compared with a strategy that isolates only the

time‐varying exposure to the market factor. In addition, we find that conventional momentum

exhibits time‐varying exposures to factors other than those in the Fama and French (1993)

model. To evaluate the dynamic risk explanation deeply, we further examine the role of time‐

varying factor exposures in the profitability of idiosyncratic momentum strategies by using

conditional asset pricing models in the spirit of Grundy and Martin (2001), which enables us to

neutralize the ‘dynamic’factor exposures directly in all models. Idiosyncratic momentum

strategies constructed using conditional models yield average excess returns that are only

≈

50–80% as high as those constructed using Blitz et al.'s (2011) approach. More surprisingly,

except for the strategy constructed using the conditional CAPM, the studentized time‐series

bootstrap test of Ledoit and Wolf (2008) indicates that none of these strategies yields greater

Sharpe ratios than the conventional momentum strategy, even at the 10% significance level.

Taken together, our results suggest that the superior performance of idiosyncratic momentum

seems not to be driven by the isolation of dynamic factor risk.

Another potential explanation of the idiosyncratic momentum effect is that it captures

common sources of mispricing, such as market‐wide investor sentiment. Miller (1977) argues

that the presence of short‐sale constraints implies that overpricing is more likely than

underpricing, as the former is more difficult to eliminate due to costly shorting. Stambaugh, Yu,

and Yuan (2012) further show that the combination of Miller's (1977) argument about the effect

of short‐sale impediments and market‐wide sentiment suggests that anomalies should be

stronger during periods of high sentiment, to the extent that the anomalies reflect mispricing.

However, we find that idiosyncratic momentum spread returns do not display such patterns

and that their average excess returns and alphas are indistinguishable from zero between high‐

and low‐sentiment periods, particularly after controlling for well‐known risk factors. Moreover,

there is no significant relation between sentiment and average excess returns and alphas on the

long legs, which appears to be consistent with the prediction of Stambaugh et al. (2012).

However, we also find no significantly lower average excess returns and alphas on the short legs

following high sentiment than following low sentiment. Hence, the idiosyncratic momentum is

unrelated to market‐wide investor sentiment.

To gain further insight into the economic source of the idiosyncratic momentum effect

documented in this article, we then explore whether this effect is consistent with the investment

CAPM, which is built on the neoclassical

q

‐theory of investment (Cochrane, 1991; Liu, Whited, &

Zhang, 2009; Zhang, 2017). George, Hwang, and Li (2018) argue that the test of the investment

LIN EUROPEAN

FINANCIAL MANAGEMENT

|

581