Do firm characteristics matter in explaining the herding effect on returns?

• DOI: http://doi.org/10.1002/rfe.1036
• Date: 01 April 2019
• Published date: 01 April 2019
• Author: Rıza Demirer,Huacheng Zhang

ORIGINAL ARTICLE

Do firm characteristics matter in explaining the herding effect

on returns?

Rıza Demirer

1

|

Huacheng Zhang

2

1

Department of Economics & Finance,

Southern Illinois University Edwardsville,

Edwardsville, Illinois

2

Institute of Financial Studies,

Southwestern University of Finance and

Economics, Chengdu, China

Correspondence

Riza Demirer, Department of Economics

and Finance, Southern Illinois University

Edwardsville, Edwardsville, IL.

Email: rdemire@siue.edu.

Abstract

This paper explores whether firm characteristics matter in determining the effect

of investor herding on asset returns. We find that the level of herding alone does

not command a significant effect on industry returns, implied by insignificant

return spreads between industries that experience high and low degrees of herd-

ing. On the other hand, we observe that herding has a significant interaction with

size and past returns. We find that small firms with high level of herding signifi-

cantly underperform small firms that experience low herding. Similarly, past loser

industries with high level of herding significantly outperform loser industries with

low herding. No significant interactions between book‐to‐market and market beta

with herding are observed. Overall, the findings suggest that the herding effect

presents itself via size and momentum channels with significant investment impli-

cations.

JEL CLASSIFICATION

G14, G15

KEYWORDS

anomalies, asset pricing, industry herding

1

|

INTRODUCTION

A number of studies in the literature suggest a possible link between investor herding and asset returns.

1

Several recent

studies focus on the effect of herding particularly on momentum and reversals and show that the level of herding plays a

significant role in the relationship between past and subsequent returns (e.g., Chen & Demirer, 2018; Demirer, Lien, &

Zhang, 2015). However, despite the extensive literature on the presence of herd behavior in financial markets, the channel

through which herding behavior affects asset prices is still understudied. Clearly, this is not only a matter of interest from

an academic perspective as a herding effect on asset returns has pricing implications but also has significant practical impli-

cations as it would imply the presence of profitable investment opportunities. This paper contributes to the literature from a

novel direction by exploring whether firm characteristics matter in determining the effect of investor herding on asset

returns. More specifically, we examine the possible interaction between herding and well‐known stock market anomalies

and provide evidence on whether the herding effect on returns is channeled via certain firm‐level factors.

In earlier studies, Wermers (2000) and Sias (2004) showed that herding can be rational and deliver high returns. On the

other hand, studies including De Long, Shleifer, Summers, and Waldmann (1990) and Abreu and Brunnermeier (2002)

argued that irrational herding can impose systematic risk on rational investors. Regardless of the theoretical arguments on

herding, if herding is a form of rational or irrational investor behavior driven by informational inefficiencies and is a sys-

tematic phenomenon, a natural research question is whether herding is independent of other systematic effects such as size,

book‐to‐market or momentum. If such an interaction between herding and firm characteristics is indeed presen t, an

Received: 23 May 2018

|

Accepted: 8 June 2018

DOI: 10.1002/rfe.1036

256

|

© 2018 The University of New Orleans wileyonlinelibrary.com/journal/rfe Rev Financ Econ. 2019;37:256–271.

important investment implication would be a hybrid active strategy that exploits this interaction via a conditional invest-

ment approach based on both firm‐level variables and the level of herding in order to generate superior abnorm al returns

beyond which can be obtained via conventional anomalies. Potentially, such an approach can even yield profitable out-

comes even in markets where strategies based on conventional anomalies are not profitable. To that end, the analysis of

herding and stock market anomalies contributes to the literature both from an academic and practical perspective.

Motivated by the findings that herd formation would be more likely to occur at the industry level rather than stock level

(e.g., Bikhchandani & Sharma, 2001; Choi & Sias, 2009; Lang & Lundholm, 1996), we examine the relationship between

the level of herding in an industry and industry returns.

2

The industry focus is further motivated by Chou, Ho, and Ko

(2012) who document the presence of industry‐related patterns driving stock market anomalies, which furt her makes the

analysis of a possible interaction between herding and anomalies meaningful in an industry context. Starting our analysis

by examining whether herding alone can explain the pattern of industry returns unconditionally, we then explore the inter-

action of several well‐known anomalies, such as size and book‐to‐market effects, with herding and examine if the herding

effect presents itself conditional on certain firm‐level characteristics.

Following the popularly utilized herding test originally developed by Chang, Cheng, and Khorana (2000) and assigning

each qualified CRSP stock to one of 49 Fama and French industries, we estimate the herding level for each industry each

month between 1964 and 2016. We then examine the return patterns for industry portfolios based on various firm‐level fac-

tors (size, book‐to‐market ratio, market beta, and past returns) and across high and low degrees of herding. We then evalu-

ate the performance of the industry portfolios in various holding periods to examine the relationship between industry

portfolios sorted on anomaly‐based factors and the level of herding.

The empirical results suggest that the level of herding alone does not command a significant effect on industry returns,

implied by insignificant return spreads between industries that experience high and low degrees of herding. On the other

hand, we observe that herding has a significant interaction with size and past returns. Although small firms within an indus-

try outperform large firms in all subsequent holding periods regardless of their herding levels (confirming the size effect),

we find that small firms with high degree of herding significantly underperform small firms that experience low herding,

suggesting that the herding effect on returns is a small‐firm phenomenon. Considering the argument by Edelen, Ince, and

Kadlec (2016) that institutions tend to buy stocks classified as overvalued, eventually leading to negative abnormal returns,

our findings suggest that institutional herding particularly focused on small firms may be driving the herding and size effect

interaction.

Similarly, we observe a persistent herding effect in the case of past loser industries with loser industries with high level

of herding significantly outperforming loser industries with low herding. The herding effect on loser industries is in fact

consistent with the finding by Brown et al. (2014) that fund managers have a greater tendency to herd on negative stock

information due to reputational concerns and greater litigation risk for holding losing stocks. The herding effect on small

firms and past losers is significant regardless of the formation period used to measure the degree of herding and is robust

at various holding periods up to 12 months. Finally, we find no significant herding effects on industries sorted on book‐

to‐market ratios as well as market betas, implied by insignificant return spreads between high and low d egree of herding

for industries sorted on book‐to‐market ratios and market betas.

These findings extend the evidence by Nofsinger and Sias (1999) and Sias (2004) who examined herding among institu-

tional investors in the United States and find that asset returns follow the direction of the herd. Our findings also suppor t the

recent evidence by Celiker et al. (2015) that industry momentum is related to herding by mutual funds. To that end, the find-

ings provide new insight to the channels in which herding affects asset returns by showing that the herding effect is chan-

neled via size and past performance. An important investment implication would be whether this interacti on can be exploited

to form hybrid investment strategies conditioned on the level of herding and conventional anomalies in order to improve the

profitability of anomaly‐based active management strategies. The findings can also provide a guideline to devise profitable

investment strategies even in markets where conventional anomalies are not present. An outline of the remainder of the paper

is as follows. Section 2 provides data description and the procedure to construct industry portfolios based on the level of

herding as well as firm‐level variables. Section 3 presents empirical results and Se ction 4 concludes the paper.

2

|

DATA AND METHODOLOGY

2.1

|

Data

We collect the data for individual stocks at daily and monthly frequency covering all ordinary shares obtained from the

Center for Research and Security Prices (CRSP) and the accounting informat ion from the quarterly Compustat database.

DEMIRER AND ZHANG

|

257