Individual risk tolerance and herding behaviors in financial forecasts

• Date: 01 November 2019
• DOI: http://doi.org/10.1111/eufm.12231
• Published date: 01 November 2019

Eur Financ Manag. 2019;25:1348–1377.wileyonlinelibrary.com/journal/eufm1348

|

© 2019 John Wiley & Sons Ltd.

DOI: 10.1111/eufm.12231

ORIGINAL ARTICLE

Individual risk tolerance and herding

behaviors in financial forecasts

Jeppe Christoffersen

|

Simone Stæhr

Department of Accounting and Auditing,

Copenhagen Business School, Solbjerg

Plads 3, Frederiksberg, 2000, Denmark

Email: jc.acc@cbs.dk; sst.acc@cbs.dk

Funding information

Copenhagen Business School

Abstract

Financial analysts tend to demonstrate herding beha-

vior, which sometimes compromises accuracy. A num-

ber of explanations spanning rational economic logic,

cognitive biases, and social forces have been suggested.

Relying on an experimental setting where participants

forecast future earnings from a rich information set, we

posit and obtain support for individual risk tolerance (or

lack thereof) as an explanatory variable for herding

behaviors. Specifically, less risk‐tolerant individuals

forecast with less boldness and instead issue forecasts

in agreement with the consensus forecast. The results

are argued to be at least partially a product of cognitive

biases and an intuitive reaction to uncertainty.

KEYWORDS

boldness, cognitive bias, experiment, intuition, news asymmetry

JEL CLASSIFICATION

G41

EUROPEAN

FINANCIAL MANAGEMENT

We are grateful for discussions and comments from Steffen Andersen, Thomas Plenborg, Melanie Schneider, and

Thomas Riise Johansen at Copenhagen Business School, from Niclas Hellman at Stockholm School of Economics, from

Richard Barker at University of Oxford, from Mitchel Stevens and Brian Knutson at Stanford University, from

participants at the Bernheim Idea Workshop and at the SCANCOR Friday Seminar Series at Stanford University, at the

PhD Seminar Series at Copenhagen Business School, and at the Behavioral Finance Working Group Conference at

Queen Mary University of London. Finally, we are grateful for comments and suggestions from two anonymous referees

and Editor John Doukas. Special thanks go to the graduate students that volunteered to participate in the experiment

and to our colleagues who assisted in pre‐tests of the experiment design. The views expressed in this paper are not

necessarily reflected by the views of others and any possible mistakes are on our own.

1

|

INTRODUCTION

‘Who herds? Who doesn’t?’The questions have been implicitly asked during many years of

research on financial analysts’earnings forecasts and recently Huang, Krishnan, Shon, and

Zhou (2017) very explicitly posed the questions in the title of their article in a leading

accounting journal. Although the questions are explicit, the answers remain elusive, primarily

because explanations have related to the circumstances under which individuals herd more or

less rather than on the characteristics of these individuals. In this paper, we seek to complement

the current literature by expanding the potential explanations of variation in herding behaviors

to also include variation in the personal characteristics of the analysts.

An example of a paper claiming but not aiming to answer the question of who herds is the

above‐cited study by Huang et al. (2017). The authors very explicitly answered the questions in

their title (‘Who Herds? Who Doesn’t?’) with the finding that herding propensity varies across

time, forecast horizon, analyst following, broker/employer size, and forecast order. While these

factors are very specific to the task and content of conducting financial forecasts, they relate

very little to the characteristics of the persons producing these forecasts or their cognitive

processes. A similar lack of focus on individuals is found in other contributions explaining

variation in herding behavior with reputational reasons (Holmes, Kallinterakis, & Ferreira,

2013), releases of macro data (Galariotis, Rong, & Spyrou, 2015), regulation limiting selectively

disclosed information (Hahn & Song, 2013; Mensah & Yang, 2008), opaqueness of firms’

information environment (Leece & White, 2017), analysts’access to information (Christensen,

Mikhail, Walther, & Wellman, 2017), analyst affiliation (Xue, 2017), brokerage size (Clement &

Tse, 2005; Jegadeesh & Kim, 2010), the number of industries followed (Clement & Tse, 2005),

firm earnings uncertainty (Song, Kim, & Won, 2009), dispersion across recommendations

(Jegadeesh & Kim, 2010), forecast revision frequency (Jegadeesh & Kim, 2010), forecast horizon

(De Bondt & Forbes, 1999), analyst’s prior accuracy (Clement & Tse, 2005), fee concerns

(Trueman, 1994), career concerns (Clarke & Subramanian, 2006; Clement & Tse, 2005; Hong,

Kubik, & Solomon, 2000; Nolte, Nolte, & Vasios, 2014), experience (Clement & Tse, 2005; Hong

et al., 2000; Youssef & Rajhi, 2010), and trust building (Kadous, Mercer, & Thayer, 2009).

It is clear from the above that many of the previous studies focus on variables that differ between

each individual, such as prior forecast errors and experience; however, these variables describe the

individuals in their job situation and do not relate to their personalities more generally. There are

two notable exceptions not listed above. One is Kumar (2010), who found that female analysts issue

bolder forecasts. However, the explanation given is that only the best potential female analysts enter

the profession due to a perception of discrimination in the analyst labor market; that is, the results

were argued to be an effect of biased self‐selection rather than traits relating to typical female

personalities. The other exception is Jiang, Kumar, and Law (2016), who found that analysts

supporting the Republican Party are less likely to issue bold forecasts, that is, they are more likely to

exhibit herding behavior. The authors’key theoretical argument was that analysts who are most

strongly aligned with the Republican Party are likely to possess conservative traits and thus prefer

the status quo. This makes them cautious about interpreting new information and about updating

their beliefs and, therefore, they do not stray from the consensus.

Our study is related to that of Jiang et al. (2016), since we focus on a personal characteristic

closely related to conservatism, that is, individuals’general risk tolerance.

1

We hypothesize and find

1

Risk tolerance refers to participants’perceived willingness to take risks based on a 10‐point Likert scale (from not willing to take risks to very willing to take

risks).

CHRISTOFFERSEN AND STÆHR EUROPEAN

FINANCIAL MANAGEMENT

|

1349