Textual Information and IPO Underpricing: A Machine Learning Approach

Katsafados, Apostolos G.; Androutsopoulos, Ion; Chalkidis, Ilias; Fergadiotis, Emmanouel; Leledakis, George N.; Pyrgiotakis, Emmanouil G.

doi:10.2139/ssrn.3720213

Cited by 6 publications

(8 citation statements)

References 97 publications

(173 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This means that the TF-IDF weighting scheme produces a set of weights for our textual features that enhance the ability of our models to classify bidders from non-involved banks. This result is consistent with previous findings, as the TF-IDF approach tends to perform better in many NLP tasks compared to simple proportional weighting (Loughran and McDonald, 2011;Loughran and McDonald, 2016;Katsafados et al, 2020). Second, the use of our finance word embeddings increases the performance of both the TF and the TF-IDF centroid embedding model, compared to using generic word embeddings.…”

Section: Combination Of Financial Variables With Bag Of Words Textual Featuressupporting

confidence: 91%

Machine learning in U.S. Bank Merger Prediction: A Text-Based Approach

et al. 2021

Self Cite

View full text Add to dashboard Cite

This paper investigates the role of textual information in a U.S. bank merger prediction task. Our intuition behind this approach is that text could reduce bank opacity and allow us to understand better the strategic options of banking firms. We retrieve textual information from bank annual reports using a sample of 9,207 U.S. bank-year observations during the period 1994-2016. To predict bidders and targets, we use textual information along with financial variables as inputs to several machine learning models. Our key findings suggest that: (1) when textual information is used as a single type of input, the predictive accuracy of our models is similar, or even better, compared to the models using only financial variables as inputs, and (2) when we jointly use textual information and financial variables as inputs, the predictive accuracy of our models is substantially improved compared to models using a single type of input. Therefore, our findings highlight the importance of textual information in a bank merger prediction task.

show abstract

Section: Combination Of Financial Variables With Bag Of Words Textual Featuressupporting

confidence: 91%

Machine learning in U.S. Bank Merger Prediction: A Text-Based Approach

et al. 2021

Self Cite

View full text Add to dashboard Cite

show abstract

“…We plan two types of tree-based machine learning models, namely, Random Forest Regressor (Baba and Sevil, 2020; Katsafados et al , 2020; Quintana et al , 2017) and Gradient Boosting Regressor (Muditomo and Broto, 2021). Both methods are very popular prediction algorithms.…”

Section: Methodsmentioning

confidence: 99%

COVID-19: the impact of the pandemic fear on IPO underpricing

Fedorova

Chertsov²,

Kuzmina³

2022

JFRA

View full text Add to dashboard Cite

Purpose The purpose of this study is to assess how the information disclosed in prospectuses impacted the initial public offering (IPO) underpricing at a time of high government interference amid the ongoing pandemic. Design/methodology/approach The design of this study has several tracks, namely, a macro-level track, which is represented by the government measures to halt the pandemic; a micro-level track, which is followed by textual analysis of IPO prospectuses; and, finally, a machine learning track, in which the authors use state-of-the-art tools to improve their linear regression model. Findings The authors found that strict government anti-COVID-19 measures indeed contribute to the reduction of the IPO underpricing. Interestingly, the mere fact of such measures taking place is enough to take effect on financial markets, regardless of the resulting efficiency of such measures. At the micro-level, the authors show that prospectus sentiments and their significance differ across prospectus sections. Using linear regression and machine learning models, the authors find robust evidence that such sections as “Risk factors”, “Prospectus summary”, “Financial Information” and “Business” play a crucial role in explaining the underpricing. Their effect is different, namely, it turns out that the more negative “Risk factors” and “Financial Information” sentiment, the higher the resulting underpricing. Conversely, the more positive “Prospectus summary” and “Business” sentiments appear, the lower the resulting underpricing is. In addition, we used machine learning methods. Consisting of more than 580 IPO prospectuses, the study sample required modern and powerful machine learning tools like Isolation Forest for pre-processing or Random Forest Regressor and Light Gradient Boosting Model for modelling purposes, which enabled the authors to gain better results compared to the classic linear regression model. Originality/value At the micro level, this study is not confined to 2020, but also embraces 2021, the year of the record number of IPOs held. Moreover, in this paper, these were prospectuses that served as a source of management sentiment. In addition, the authors used a tailor-made government stringency index. At the micro level, basing the study on behavioural finance hypotheses, the authors conducted both separate and holistic analysis of prospectuses to assess investors’ reaction to different aspects of IPO companies as well as to the characteristics of the IPOs themselves. Lastly, the authors introduced a few innovations to the research methodology. Textual analysis was conducted on a corpus of prospectuses included in a study sample. However, the authors did not use pre-trained dictionaries, but instead opted for FLAIR, a modern open-source framework for natural language processing.

show abstract

“…For example, Goel and Gangolly (2012), Purda and Skillicorn (2015), and Goel and Uzuner (2016) perform textual analysis on Item 7 to detect corporate fraud. Katsafados et al (2020) combine Item 7 and Item 4 We use Beautiful Soup (https:// beautiful-soup-4.readthedocs.io/en/latest).…”

Section: Data and Toolkitmentioning

confidence: 99%

EDGAR-CORPUS: Billions of Tokens Make The World Go Round

Loukas¹,

Fergadiotis²,

Androutsopoulos³

et al. 2021

Proceedings of the Third Workshop on Economics and Natural Language Processing

Self Cite

View full text Add to dashboard Cite

We release edgar-corpus, a novel corpus comprising annual reports from all the publicly traded companies in the us spanning a period of more than 25 years. To the best of our knowledge, edgar-corpus is the largest financial nlp corpus available to date. All the reports are downloaded, split into their corresponding items (sections), and provided in a clean, easy-to-use json format. We use edgar-corpus to train and release edgar-w2v, which are word2vec embeddings for the financial domain. We employ these embeddings in a battery of financial nlp tasks and showcase their superiority over generic glove embeddings and other existing financial word embeddings. We also open-source edgarcrawler, a toolkit that facilitates downloading and extracting future annual reports.

show abstract

Textual Information and IPO Underpricing: A Machine Learning Approach

Cited by 6 publications

References 97 publications

Machine learning in U.S. Bank Merger Prediction: A Text-Based Approach

Machine learning in U.S. Bank Merger Prediction: A Text-Based Approach

COVID-19: the impact of the pandemic fear on IPO underpricing

EDGAR-CORPUS: Billions of Tokens Make The World Go Round

Contact Info

Product

Resources

About