Do SEC filings indicate any trends? Evidence from the sentiment distribution of forms 10-K and 10-Q with FinBERT
| DOI | https://doi.org/10.1108/DTA-05-2022-0215 |
| Published date | 21 April 2023 |
| Date | 21 April 2023 |
| Pages | 293-312 |
| Subject Matter | Library & information science,Librarianship/library management,Library technology,Information behaviour & retrieval,Metadata,Information & knowledge management,Information & communications technology,Internet |
| Author | Hyogon Kim,Eunmi Lee,Donghee Yoo |
Do SEC filings indicate any trends?
Evidence from the sentiment
distribution of forms 10-K and 10-Q
with FinBERT
Hyogon Kim
Management of Technology, Gyeongsang National University, Jinju,
Republic of Korea
Eunmi Lee
Textile and Apparel Management, University of Missouri, Columbia, Missouri,
USA, and
Donghee Yoo
Management Information Systems, BERI, Gyeongsang National University, Jinju,
Republic of Korea
Abstract
Purpose –This study quantified companies’views on the COVID-19 pandemic with sentiment analysis of US
public companies’disclosures. The study aims to provide timelyinsights to shareholders, investors and consumers
by exploring sentiment trends and changes in the industry and the relationship with stock price indices.
Design/methodology/approach –From more than 50,000 Form 10-K and Form 10-Q published between
2020 and 2021, over one million texts related to the COVID-19 pandemic were extracted. Applying the
FinBERT fine-tuned for this study, the texts were classified into positive, negative and neutral sentiments.
The correlations between sentiment trends, differences in sentiment distribution by industry and stock price
indices were investigated by statistically testing the changes and distribution of quantified sentiments.
Findings –First, there were quantitative changes in texts related to the COVID-19 pandemic in the US
companies’disclosures. In addition, the changes in the trend of positive and negative sentiments were found.
Second, industry patterns of positive and negative sentiment changes were similar, but no similarities were
found in neutral sentiments. Third, in analyzing the relationship between the representative US stock indices
and the sentiment trends, the results indicated a positive relationship with positive sentiments and a negative
relationship with negative sentiments.
Originality/value –Performing sentiment analysis on formal documents like Securities and Exchange
Commission (SEC) filings, this study was differentiated from previous studies by revealing the quantitative
changes of sentiment implied in the documents and the trend over time. Moreover, an appropriate data
preprocessing procedure and analysis method were presented for the time-series analysis of the SEC filings.
Keywords Sentiment analysis, Trend, FinBERT, SEC, Disclosure, COVID-19
Paper type Research paper
1. Introduction
Corporate disclosures, such as annual and quarterly reports, provide useful information to
shareholders, investors and so on, as official documents that contain a company’sviewson
overall management. In the United States, most public companies must prepare Form 10-K
(annual reports) and Form 10-Q (quarterly reports) and submit them regularly to the United
States Securities and Exchange Commission (SEC). The SEC filings are considered valuable data
for corporate analysis because they contain quantitative data such as financial statements and
qualitative data described from companies’perspectives, such as business overview, business
conditions, perceived risks and future plans (Stephany et al.,2020).
ThecurrentissueandfulltextarchiveofthisjournalisavailableonEmeraldInsightat:
https://www.emerald.com/insight/2514-9288.htm
293
Received 24 May 2022
Revised 26 September 2022
4 November 2022
Accepted 10 November 2022
Data Technologies and
Applications
Vol. 57 No. 2, 2023
pp. 293-312
© Emerald Publishing Limited
2514-9288
DOI 10.1108/DTA-05-2022-0215
Sentiment
analysis of
SEC filing with
FinBERT
Companies experiencing major management changes both internally and externally
have an ethical responsibility to provide the latest information to support investors’
correct decision-making (Yuthas et al., 2002). In response to the COVID-19 pandemic that
shocked the entire industry, multiple companies quickly included their views on the
COVID-19 pandemic in corporate disclosures. According to a survey by Larcker et al.
(2020), in the 3,644 SEC filings collected in 2020, COVID-19 mentions rose sharply to
4.8 per cent in January, 35.5 per cent in February, 76.6 per cent in March, 100 per cent in
April and 99.9 per cent in May 2020. Therefore, through the analysis of the SEC filings
released in 2020–2021, it is possible to examine the views of US companies on the COVID-19
pandemic over time.
However, SEC filings are unstructured documents and consist of massive numerical and
textual data.Text mining is gaining insightsand discovering meaningfulinformation such as
patterns and important links from unstructured text data like SEC filings, newsand internet
postings. Sentiment analysis, one of the text mining techniques, aims to systematically
extract, identify, categorize and quantify emotional states, subjective information and so on
from texts. Being popular and widely used in corporate management, sentiment analysis
classifies produced data like consumers’experiences and opinions on social media platforms
into positive and negative (Capuano et al.,2021;Tian et al.,2021). However, sentiment
analysis is also being attempted on documents in which authors express opinions and
sentiments less explicitly, such as SEC filings and corporate social responsibility (CSR)
reports. For example, SEC filings have been utilized in sentiment analysis studies for
corporate evaluation, including performance prediction (Azimi and Agrawal, 2021),
financial risk prediction (Wang et al.,2013) and identification of insolvent companies
(Gandhi et al., 2019). In addition, through sentiment analysis of CSR reports, the values like
return on assets were predicted (Che et al.,2020;Myšková and Hájek, 2018).
Text mining techniques including sentiment analysis are gradually evolving for
practical data analysis. In particular, the development of deep learning and natural
language processing (NLP) has made remarkable achievements in various classification
and prediction tasks. Studies are being conducted to challenge more complex sentiment
classification using a language model such as Bidirectional Encoder Representations from
Transformers (BERT) that considers contextual meanings and bidirectional sentences.
Although sentiment analysis using text mining and NLP techniques is being actively
used, there are relatively few studies on sentiment analysis on formal documents such as
SEC filings. Moreover, few studies reveal how sentiment implied in documents changes
over time. To fill the gap in the literature, this study aims to quantitatively measure the
views and attitudes of US companies to the COVID-19 pandemic related to sentiment and to
explore changes and trends in the companies’sentiment over time. The results of this study
may contribute to various investors, consumers and administrators in exploring corporate
responses and changes in their responses to the uncontrollable events affecting the entire
society like the COVID-19 pandemic and getting a better understanding of the individual
company, as well as all industries.
The detailed tasks of this study to achieve the study purpose are as follows. First,
sentiment analysis of texts related to the COVID-19 pandemic in the SEC filings was
performed. Several significant issues, such as the WHO’s pandemic declaration, increase
in confirmed COVID-19 cases and deaths, vaccine development and vaccination and the
emergence of COVID-19 variants, affected corporate disclosures. This would result in
a quantitative change in texts related to the COVID-19 pandemic and indicate certain
trends. Second, it was examined in the study whether the distribution of sentiment due to
the COVID-19 pandemic differed by industry. The COVID-19 pandemic affected industries
as a whole, but the magnitude and intensity of the impact varied by industry (Szczygielski
et al., 2022). Accordingly, it was predicted that the sentiment distribution and trend would
DTA
57,2
294
To continue reading
Request your trial