Determinants of the relationship between investor sentiment and analysts’ private information production

• DOI: http://doi.org/10.1111/jbfa.12340
• Author: Jeff J. Wang,Walied Keshk
• Published date: 01 October 2018
• Date: 01 October 2018

DOI: 10.1111/jbfa.12340

Determinants of the relationship between investor

sentiment and analysts’ private information

production

Wali ed Kesh k1Jeff J. Wang2

1CaliforniaState University Fullerton, Fullerton,

CA,USA

2SanDiego State University, San Diego, USA

Correspondence

WaliedKeshk, California State University

Fullerton,800 N. State College Blvd., Fullerton,

CA9283, USA.

Email:wkeshk@fullerton.edu

JELClassification: D82, D83, G41, M41

Abstract

We examine whether and how investor sentiment affects analysts’

private information production and the factors that moderate this

influence. As expected, analysts’ private information production

(measured by the information asymmetry component of analyst

forecast dispersion) is significantly negatively associated with

investor sentiment. Our tests also show that higher institutional

holdings of stock, longer analysts’ earnings forecasting experience,

and higher levels of short selling of stock mitigate this negative

influence of sentiment on analysts’ private information production.

Our findings are informative for researchers, financial analysts, and

market participants who rely on analysts’ forecasts to form earnings

expectations.

KEYWORDS

analysts’ earnings forecasting experience, analysts’ forecasts, ana-

lysts’ private information production, institutional holdings, investor

sentiment, short selling

1INTRODUCTION

We examine the relationship between investor sentiment and analysts’ private information production, measured by

the information asymmetry component of annual earnings forecast dispersion during 1986 to 2014 for the universe

of US firms. Further,we examine whether institutional holdings of the firm's stock, analysts’ earnings forecasting expe-

rience, and the level of short selling of the firm's stock mitigate the relationship between investor sentiment and ana-

lysts’ private information production. Prior studies have focused on whether and how investor sentiment affects ana-

lysts’ stock recommendations and earnings forecasts. Ke and Yu(2009) find that sentiment has a negative effect on the

profitability of analysts’ stock recommendations, while Hribar and McInnis (2012) find that, when sentiment is high,

analysts’ forecasts of one-year-ahead annual earnings are more optimistic for difficult-to-value firms.

We extend this line of research as we investigate a mechanism through which investorsentiment affects analysts’

earnings forecasts. Barron, Kim, Lim, and Stevens(1998) decompose analyst forecast dispersion into two components:

common uncertainty toward future earnings and information asymmetry among analysts, whereby the information

1082 c

2018 John Wiley & Sons Ltd wileyonlinelibrary.com/journal/jbfa JBus Fin Acc. 2018;45:1082–1099.

KESHK ANDWA NG 1083

asymmetry component is considered a proxyfor the amount of private information produced by analysts.1This decom-

position process has been widely used in recent research (e.g., Barron, Byard, & Kim, 2002b; Barron, Stanford, & Yu,

2009; Byard, Li, & Yu,2011). We argue that examining the influence of investor sentiment on the information asymme-

try component of analyst forecast dispersion allows us to draw inferences about analysts’ effort exertedin producing

private information when forecasting firms’ earnings.

We developour predictions based on research on the effect of mood on information processing (e.g., Bless, Bohner,

Schwarz, & Strack, 1990; Bodenhausen, Kramer, & Süsser, 1994). A main finding of that research is that a positive

mood (i.e., high sentiment), compared to a negative mood (i.e., low sentiment), induces individuals to be less critical

of the available information, pay less attention to details, and rely to a greater extent on heuristics to makefaster and

less effortful judgments. Although the prior research findings were obtained using less sophisticated individual par-

ticipants, there is a chance that similar results will hold for the more sophisticated professional analysts. Therefore,

when investor sentiment is high rather than low, analysts are likely to exert less effort in gathering and processing

public information to produce private information when forecasting firms’ earnings. In other words, we expect to find

lower private information production by analysts (i.e., lower information asymmetry among analysts) when investor

sentiment is high rather than low. Consistent with our expectations,our analyses show that investor sentiment is sig-

nificantly negatively correlated with analysts’ production of private information.

In addition, although prior studies focus mainly on documenting the influence of investor sentiment on analysts’

forecasts, the main focus of our study is on the factors that can mitigate the undesirable influence of investor senti-

ment on analysts’ private information production. We investigatewhether the negative relationship between investor

sentiment and analysts’ private information production is moderated by institutional holdings of the firm'sstock,ana-

lysts’ firm-specific earnings forecasting experience, and the levelof short selling of the firm's stock. First, prior studies

suggest that higher levels of institutional holdings of the firm's stock serve as external monitoring that strengthens

analysts’ reputational concerns when producing their forecasts and recommendations (e.g., Huang & Wright, 2015;

Ljungqvist, Marston, Starks, Wei, & Yan,2007; Malmendier & Shanthikumar, 2014). As a result, we predict that higher

levels of institutional holdings of stock will mitigate the reduction in analysts’ private information production during

high investor sentiment periods. Our tests support this prediction.

Second, prior research documents that longer analysts’ firm-specific earnings forecasting experience is associated

with a higher levelof forecast accuracy (e.g., Clement, 1999; Hussain, 2010) and a higher likelihood of issuing bold fore-

casts (Clement& Tse, 2005). Consistent with these prior studies, we hypothesize and find that longer analysts’ earnings

forecasting experience is associated with a reduction in thenegative association between investor sentiment and ana-

lysts’ private information production. Third, psychology and accounting research show that the availability of discon-

firming information that contradicts existing expectationsinduces individuals to expand their information search and

processing (e.g., Burgoon & Burgoon, 2001; Clor-Proell, 2009). We argue that the negative signal inferred from a high

level of short-selling of the firm's stock will contradictthe general positive mood that prevails in the stock market dur-

ing periods of high investor sentiment which will induce analysts to produce more private information. In other words,

we predict that a high level of short-selling will be associated with a reduction in the negative association between

investor sentiment and analysts’ private information production. The results of our tests support this prediction.

Tosummarize, we predict and show that analysts’ private information production is reduced during periods of high

investor sentiment and that this negative association is mitigated byhigher institutional holdings of stock, longer ana-

lysts’ earnings forecasting experience, and higher levelsof short selling of stock. Our paper contributes to research on

the influence of investor sentiment on analysts’ forecasts and recommendations (e.g., Ding, Charoenwong, & Seetoh,

2004; Hribar & McInnis, 2012; Walther & Willis, 2013) in at least two important ways. First, although prior research

(e.g., Hribar & McInnis, 2012) documents the influence of investor sentiment on the characteristics of analysts’ fore-

casts, our paper sheds light on one mechanism through which investor sentiment influences the characteristics of

1Barronet al. (1998) calculate the mean squared error of individual analysts’ forecasts to measure average uncertainty toward earnings numbers and divide

analystforecast dispersion by this mean squared error to measure information asymmetry among analysts. In other words, analyst forecast dispersion, which

is the standard deviation of individual analysts’ forecasts, is equivalent to the product of uncertainty toward earnings numbers and information asymmetry

amonganalysts.