Impact of news sentiment and topics on IPO underpricing: US evidence

• DOI: https://doi.org/10.1108/IJAIM-06-2021-0117
• Published date: 16 November 2021
• Date: 16 November 2021
• Pages: 73-94
• Subject Matter: Accounting & finance,Accounting/accountancy,Accounting methods/systems
• Author: Elena Fedorova,Sergei Druchok,Pavel Drogovoz

Impact of news sentiment and

topics on IPO underpricing:

US evidence

Elena Fedorova

Financial University under the Government of the Russian Federation,

Moscow, Russian Federation

Sergei Druchok

National Research University Higher School of Economics,

Moscow, Russian Federation, and

Pavel Drogovoz

Bauman Moscow State Technical University, Moscow, Russian Federation

Abstract

Purpose –The goal of the study is to examine the effectsof news sentiment and topics dominating in the

news f‌ieldprior to the initial public offering (IPO) on the IPO underpricing.

Design/methodology/approach –The authors’approachhas several steps. The f‌irst is textual analysis.

To detect the dominating topics in the news, the authors use Latent Dirichlet allocation. The authors use

bidirectional encoder representations from transformers (BERT) pretrained on f‌inancial news corpus to

evaluate the tonality of articles. The second is evaluation of feature importance. To this end, a linear

regression with robust estimatorsand Classif‌ication and Regression Tree and Random Forest are used. The

third is data. The text dataconsists of 345,731 news articles from Thomson Reuters relatedto the USA in the

date range from 1 January 2011 to 31 May 2018. The datacontains all the possible topics from the website,

excluding anything relatedto sports. The sample of 386 initial public offerings completed in the USA from1

January 2011to 31 May 2018 was collected from Bloomberg Database.

Findings –The authors found that sentiment of the media regarding the companies going public inf‌luences IPO

underpricing. Some topics, namely, the climate change and environmental policies and the trade war betwee nt he

US and China, have inf‌luence on IPO underpricing if they appear in the media prior to the IPO day.

Originality/value –The puzzle of IPO underpricing is studied from the point of Narrative Economics

theory for the f‌irst time. While most of the works cover only some specif‌ic newssegment, we use Thomson

Reuters news aggregator,which uses such sources The New YorkPost, CNN, Fox, Atlantic, The Washington

Post ? Buzzfeed. To evaluate the sentiment of the articles, a state-of-the-art approach BERT is used. The

hypothesis that some common narratives or topics in the public discussion may impose inf‌luence on such

example ofbiased behaviour like IPO underpricing hasalso found conf‌irmation.

Keywords IPO underpricing, Textual analysis, News sentiment, News topics, Narrative economics,

BERT, CART, USA market

Paper type Research paper

1. Introduction

Initial public offering is a very complicated process. Sell side or the issuers are striving to earn

more, buyers or institutional investors would prefer to pay less, professional intermediators of the

The authors would like to thank Professor Robert Shiller for inspiration for this study with his idea of

narrative economics.

Impact of news

sentiment and

topics

73

Received1 June 2021

Revised26 September 2021

Accepted16 October 2021

InternationalJournal of

Accounting& Information

Management

Vol.30 No. 1, 2022

pp. 73-94

© Emerald Publishing Limited

1834-7649

DOI 10.1108/IJAIM-06-2021-0117

The current issue and full text archive of this journal is available on Emerald Insight at:

https://www.emerald.com/insight/1834-7649.htm

process (underwriters) use their expertise in pursuit of high fees and brilliant reputation, and

regulatory machinery overlooks the whole process and strictly punishes the violators. Despite all

the above-mentioned factors, money left on the table can be observed systematically and since the

long in the past (Chambers and Dimson, 2009;Ibbotson, 1975).

The role of the media has been gaining attention by the researchers in the f‌ield of f‌inance

and corporate f‌inance in particular (Ahmad et al., 2016;Bajo and Raimondo, 2017;Liu et al.,

2014;Seng et al., 2017;Tetlock, 2014;Comiran et al., 2018), the attention has been primarily paid

to the effect of media on retail investors (Bushee et al., 2020). However, one of the recent

advances in the f‌ield of behavioral economics, the theory of Narrative Economics (Davis, 2020;

Shiller, 2017), assumes that the activity of all economic agents is inf‌luenced by narratives, but

not plain facts. The media represents a good source to trace the dominating beliefs and

narratives among economic agents, including professional and retail investors, and to evaluate

its inf‌luence on initial public offering (IPO) underpricing. We would like to test the

aforementioned theory of information distribution among economic agents via popular

narratives and thus set the following goal of the study: to examine the effects of news sentiment

and topics dominating in the news f‌ield prior to the IPO on the IPO underpricing.

This study contributes to the existing research by application of state-of-the-art approaches to

quantitative assessment of textual information, by addressing the puzzle of IPO underpricing

from the point of view of narrative economics (Shiller, 2017), and by using the textual data sample

covering all the news f‌ield (345,731 news articles) to explain IPO underpricing. To detect the

dominating topics in the news and evaluate the tonality of articles, we use textual analysis

algorithms. More specif‌ically, we group texts from the corpus by topics with the help of Latent

Dirichlet allocation (LDA; Blei et al., 2003), which is a three-level hierarchical Bayesian model, in

which each item of a collection is modelled as a f‌inite mixture over an underlying set of topics.

This model helped us to detect 67 themes, as the analysis of U

mass

metric and average topic

overlap measured by Jaccard similarity suggested. The tone of the text is evaluated with the help

of bidirectional encoder representations from transformers (BERT) (Devlin et al., 2018)–a state-of-

the-art language representation neural network model, which is pretrained for the purposes of

sentiment analysis of f‌inancial texts. Using BERT is preferable to other sentiment analysis

techniques, such as those based on a vocabulary (Bajo and Raimondo, 2017;Boulton et al., 2017;

Brau et al., 2016;Loughran and McDonald, 2013;Tetlock, 2007;Zou et al.,2020), as it helps to

avoid subjectivity when choosing a vocabulary and sees the text as a whole, as opposed to bag-of-

words approach.

As a setting for the empirical research, the US market is chosen for several reasons. First,

f‌irms that decide to go public in the American market are thoroughly scrutinized by the

regulatory organs, and they are being imposed with quiet period restrictions. This implies that

all the substantial information is revealed at once in prospectus available for everyone through

the internet. The news in this case may not contain any new information, being “noise”and

containing only the sentiment. Second, most of the most modern natural language processing

algorithms are adopted solely to English language, which makes sentiment evaluation handier.

The rest of the study is organized as follows. In Section 1, the review of the major theories

and ideas explaining the IPO underpricing is presented and the hypotheses are introduced.

Section 2 contains data description andmethodology presentation. Section 3 introduces the

empirical results and Section 4 discusses the main results and ideas. Section 5 puts for the

discussion and conclusionof this study.

2. Literature review and hypotheses

The systematic initial public offerings underpricing, characterized by abnormal positive

return in the f‌irst day of trading along the with aftermarket eff‌iciency, has been discussed

IJAIM

30,1

74