Twitter investor sentiment and corporate earnings announcements
| Published date | 01 June 2023 |
| Author | Nikolaos Karampatsas,Soheila Malekpour,Andrew Mason,Christos P. Mavis |
| Date | 01 June 2023 |
| DOI | http://doi.org/10.1111/eufm.12384 |
DOI: 10.1111/eufm.12384
ORIGINAL ARTICLE
Twitter investor sentiment and corporate
earnings announcements
Nikolaos Karampatsas
1
|Soheila Malekpour
2
|
Andrew Mason
3
|Christos P. Mavis
4
1
Independent Researcher, London, UK
2
King's Business School, King's College
London, London, UK
3
Independent Researcher, Cardiff, UK
4
Surrey Business School, University of
Surrey, Guildford, UK
Correspondence
Soheila Malekpour, King's Business
School, King's College London, Bush
House, 30 Aldwych, London, WC2B 4BG,
UK.
Email: soheila.malekpour@kcl.ac.uk
Abstract
We examine the impact of firm‐specific investor sentiment
(FSIS) on stock returns for negative and positive earnings
surprises. Using a measure constructed from firm‐specific
tweets, we find that FSIS has a greater impact on stock
returns for negative relative to positive earnings surprises.
We further show that the impact of FSIS is greater
for firms whose valuation is uncertain and difficult
to arbitrage. Moreover, we provide evidence of return
reversals over post‐announcement periods. Our results
highlight the importance of FSIS around earnings
announcements.
KEYWORDS
earnings surprises, investor sentiment, social media, StockTwits,
Twitter
JEL CLASSIFICATION
G02, G11, G12, G14
Eur Financ Manag. 2023;29:953–986. wileyonlinelibrary.com/journal/eufm
|
953
EUROPEAN
FINANCIAL MANAGEMENT
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and
reproduction in any medium, provided the original work is properly cited.
© 2022 The Authors. European Financial Management published by John Wiley & Sons Ltd.
We are grateful to John A. Doukas (the editor) and an anonymous referee for their helpful feedback. We also thank
PsychSignal for providing us with the sentiment data. In addition, we thank Yakov Amihud, Malcolm Baker, Bonnie
Buchanan, Elroy Dimson, Paul Guest, Krishna Paudyal, Dimitris Petmezas, Srinivasan Sankaraguruswamy, Frank
Skinner, Richard Taffler, Nikolaos Tessaromatis, Nickolaos Travlos, Evangelos Vagenas‐Nanos, Wenzhao Wang, Jeffrey
Wurgler, Holly Yang, participants at the European Financial Management Association (EFMA) Annual Conference
(2017), the World Finance Conference (2017), The Behavioural Finance Working Group Conference (2017), the 2nd
Annual Conference of Brunel Studies in Economics and Finance (2016), the Portsmouth‐Fordham Conference on
Banking and Finance (2016) and the Surrey Business School Finance and Accounting Seminar Series (2016) for
comments. All errors and omissions remain the responsibility of the authors.
1|INTRODUCTION
The role of market‐wide investor sentiment in investment decisions is clearly established in the
finance and accounting literature in recent years (e.g., Baker & Wurgler, 2006). Similarly, the
fact that Internet stock messages and other social media platforms may contain information or
sentiment which influences price formation is also documented (e.g., Antweiler & Frank, 2004).
However, less is known about the impact of firm‐specific investor sentiment (FSIS) as expressed
on social media platforms on individual firms’stock prices around corporate events, even
though there is abundant anecdotal evidence suggesting that FSIS affects stock prices. For
example, in January 2021, GameStop's share price jumped to more than 60% in after‐hours
trading after a positive tweet from Tesla's CEO, Mr Elon Musk, regarding the video game firm.
GameStop's share price closed up 92.7% that day before plunging nearly 90% about a week
later.
1
Along these lines, Bartov et al. (2018) explore the effect of the aggregate Twitter opinion
around earnings announcements. By examining the unconditioned impact of Twitter
information on announcement returns, they provide evidence that the aggregate opinion from
individual tweets can be used to predict firms’announcement returns. There is, however, a
large body of work in social sciences that finds asymmetric responses towards negative and
positive information. More specifically, there is a propensity for negative information to be
given more importance than positive information during valuations, which can be attributed to
a‘negativity effect’.
2
Motivated by this literature, we examine the heterogenous impact of FSIS
on earnings announcement returns when the type of earnings news is jointly accounted and
improve our understanding of the phenomenon.
Earnings announcements are an important source of information for the fundamental
valuation of firms. There is an extensive literature documenting significant stock price
movements around earnings announcements (e.g., Bartov et al., 2002; Kasznik &
McNichols, 2002). Prior studies address the conceptualization of the earnings–return
relationship and concur that firms with positive (negative) earnings outcomes, experience
significantly positive (negative) abnormal stock price performance. These studies assume that
rational investors efficiently impound accounting information into stock prices and
arbitrageurs offset the actions of irrational investors. Our paper acknowledges that investors
may be irrational and arbitrage may be limited and argues that a behavioural aspect is related to
firms’short‐term performance at the time of earnings announcements.
3
We argue investors
may be prone to sentiment and rational investors may face limits to arbitrage, therefore, stock
1
Source: Financial Times, ‘Moment of weakness: Amateur investor left counting GameStop losses’, 5 February 2021.
2
The ‘negativity effect’is a behavioural/psychological idea defined by a larger effect of negative versus positive
environmental stimuli on an organism (Peeters & Czapiński, 1990). There are two main interpretations of this effect:
(1) in decisions under risk, the potential downside risk is more heavily weighted than any potential benefits and (2) in
the formation of overall valuations, negative signals are weighted more heavily than positive ones. Some related studies
dealing with this effect include Kanouse and Hanson (1971), Peeters (1971), Kernell (1977), Kahneman and Tversky
(1979), Ronis and Lipinski (1985), Aragones (1997), Singh and Teoh (2000), Baumeister et al. (2001), Akhtar et al.
(2011) and Chau et al. (2016).
3
Inspired by De Long et al. (1990), we argue that the differences of opinion can be large around earnings
announcements, therefore, investors’responses to earnings surprises may be compatible with their prevailing
sentiment. In addition, we argue if the change in sentiment around earnings announcements is unpredictable, trading
against irrational investors’trades may become costly and risky for arbitrageurs, as a result, arbitrageurs may reduce
the size of the position they take and become unable to drive stock prices back to fundamental values.
954
|
EUROPEAN
FINANCIAL MANAGEMENT
KARAMPATSAS ET AL.
prices may not be at their fundamental values at the arrival of new information about earnings.
Any abnormal price movement which cannot be explained by corporate earnings may be
considered a mispricing, therefore, we investigate the relationship between mispricing and
FSIS in the days preceding the earnings announcement. We examine the impact of FSIS in the
context of earnings announcements as the expected direction of earnings announcements and
the corresponding price reactions can be clearly developed and predicted. This allows us to
formulate and examine our hypotheses in relation to the heterogeneous impact of FSIS.
In our analysis, we use data from a unique and comprehensive Twitter and StockTwits investor
sentiment database; PsychSignal.
4
PsychSignal is a leading provider of real‐time sentiment data to
investors and traders covering more than 10,000 individual securities including all stocks in the
NASDAQ100 and S&P500 indices.
5
PsychSignal investor sentiment database is based on the
linguistic processing of millions of firm‐related messages posted on Twitter and StockTwits
analyzed by a highly specialized natural language processing (NLP) engine. We acknowledge that
there may be certain limitations in using PsychSignal sentiment measures for our research
purposes as the technical details of their algorithm are undisclosed due to confidentiality. However,
using commercial providers' data such as PsychSignal is an improvement over self‐collected and
self‐analyzed social media investor sentiment data.
6
PsychSignal's NLP engine process the language
used in tweets the same way a professional trader would do, therefore, it is able to provide its
subscribers with the most granular investor sentiment data. Furthermore, PsychSignal represents
an accurate investor sentiment data set used by actual investors as it is a commercial provider that
directly feeds analyzed Twitter and StockTwits data to its subscribers. PsychSignal's unique data
set allows its users to account for the near‐instant information flows in social media that might
have a different effect on stock returns relative to information from other sources.
We construct our FSIS measure by using PsychSignal data for individual firms over the period
2011–2015 and analyze its impact on the announcement of 14,423 corporate earnings where actual
earnings diverge from investment analysts’forecasts. We demonstrate that FSIS is heterogeneously
related to abnormal stock returns during earnings surprises. Our results indicate that investors are
more prone to behavioural biases around negative events, as the impact of FSIS on abnormal
returns is larger for negative than for positive earnings surprises. This result is consistent with
evidence that responses to negative and positive information are asymmetric due to a ‘negativity
bias’(e.g., Akhtar et al., 2011;Aragones,1997;Baumeisteretal.,2001;Chauetal.,2016;Kahneman
&Tversky,1979; Kanouse & Hanson, 1971;Kernell,1977;Peeters,1971; Ronis & Lipinski, 1985;
Singh & Teoh, 2000). Our results also indicate that the impact of FSIS is larger for firms that face
uncertainty in valuation and difficulty in arbitrage; small firms, young firms, firms with high return
volatility, growth and value firms and non‐dividend‐paying firms. We also show that the impact of
4
We use the daily data from PsychSignal which is based on social media feeds for the days before earnings
announcements. See the Supporting Information: Internet Appendix Section A for further information about
PsychSignal data.
5
The need for new sources of information is highlighted by Goldstein and Yang (2015) who conclude that the
information available to investors now is so complex that informed traders tend to specialize or have a comparative
advantage in different types of financial information. Many traders seek to gain advantages in technology by
subscribing to commercial services which supply live textual sentiment feeds.
6
There has been a significant increase in the use of sentiment data of commercial providers among both market
participants and academic researchers in recent years. There is prior evidence of secondary data of sentiment and other
features, as provided by commercial providers, having significant effects on the financial markets. Indeed, Behrendt
and Schmidt (2018) validate the relevance of Bloomberg sentiment data to stock markets and Cathcart et al. (2020)
indicate that Thomson Reuters News Analytics media tone explains and predicts CDS returns.
KARAMPATSAS ET AL.EUROPEAN
FINANCIAL MANAGEMENT
|
955
Get this document and AI-powered insights with a free trial of vLex and Vincent AI
Get Started for FreeUnlock full access with a free 7-day trial
Transform your legal research with vLex
-
Complete access to the largest collection of common law case law on one platform
-
Generate AI case summaries that instantly highlight key legal issues
-
Advanced search capabilities with precise filtering and sorting options
-
Comprehensive legal content with documents across 100+ jurisdictions
-
Trusted by 2 million professionals including top global firms
-
Access AI-Powered Research with Vincent AI: Natural language queries with verified citations
Unlock full access with a free 7-day trial
Transform your legal research with vLex
-
Complete access to the largest collection of common law case law on one platform
-
Generate AI case summaries that instantly highlight key legal issues
-
Advanced search capabilities with precise filtering and sorting options
-
Comprehensive legal content with documents across 100+ jurisdictions
-
Trusted by 2 million professionals including top global firms
-
Access AI-Powered Research with Vincent AI: Natural language queries with verified citations
Unlock full access with a free 7-day trial
Transform your legal research with vLex
-
Complete access to the largest collection of common law case law on one platform
-
Generate AI case summaries that instantly highlight key legal issues
-
Advanced search capabilities with precise filtering and sorting options
-
Comprehensive legal content with documents across 100+ jurisdictions
-
Trusted by 2 million professionals including top global firms
-
Access AI-Powered Research with Vincent AI: Natural language queries with verified citations
Unlock full access with a free 7-day trial
Transform your legal research with vLex
-
Complete access to the largest collection of common law case law on one platform
-
Generate AI case summaries that instantly highlight key legal issues
-
Advanced search capabilities with precise filtering and sorting options
-
Comprehensive legal content with documents across 100+ jurisdictions
-
Trusted by 2 million professionals including top global firms
-
Access AI-Powered Research with Vincent AI: Natural language queries with verified citations
Unlock full access with a free 7-day trial
Transform your legal research with vLex
-
Complete access to the largest collection of common law case law on one platform
-
Generate AI case summaries that instantly highlight key legal issues
-
Advanced search capabilities with precise filtering and sorting options
-
Comprehensive legal content with documents across 100+ jurisdictions
-
Trusted by 2 million professionals including top global firms
-
Access AI-Powered Research with Vincent AI: Natural language queries with verified citations