Out‐of‐sample predictability of firm‐specific stock price crashes: A machine learning approach

IF 2.2 3区管理学 Q2 BUSINESS, FINANCE Journal of Business Finance & Accounting Pub Date : 2024-09-17 DOI:10.1111/jbfa.12831

Devrimi Kaya, Doron Reichmann, Milan Reichmann

引用次数: 0

Abstract

We use machine learning methods to predict firm‐specific stock price crashes and evaluate the out‐of‐sample prediction performance of various methods, compared to traditional regression approaches. Using financial and textual data from 10‐K filings, our results show that a logistic regression with financial data inputs performs reasonably well and sometimes outperforms newer classifiers such as random forests and neural networks. However, we find that a stochastic gradient boosting model systematically outperforms the logistic regression, and forecasts using suitable combinations of financial and textual data inputs yield significantly higher prediction performance. Overall, the evidence suggests that machine learning methods can help predict stock price crashes.

查看原文

微信好友朋友圈 QQ好友复制链接

本刊更多论文

特定公司股价暴跌的样本外可预测性：机器学习方法

我们使用机器学习方法预测特定公司的股价暴跌，并评估各种方法与传统回归方法相比的样本外预测性能。我们使用 10-K 申报文件中的财务和文本数据，结果表明，使用财务数据输入的逻辑回归表现相当不错，有时甚至优于随机森林和神经网络等较新的分类器。不过，我们发现随机梯度提升模型的表现明显优于逻辑回归，而且使用财务数据和文本数据输入的适当组合进行预测，其预测性能也明显高于逻辑回归。总体而言，证据表明机器学习方法有助于预测股价暴跌。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文去求助

来源期刊

Journal of Business Finance & Accounting BUSINESS, FINANCE-

CiteScore

4.40

自引率

17.20%

发文量

期刊介绍： Journal of Business Finance and Accounting exists to publish high quality research papers in accounting, corporate finance, corporate governance and their interfaces. The interfaces are relevant in many areas such as financial reporting and communication, valuation, financial performance measurement and managerial reward and control structures. A feature of JBFA is that it recognises that informational problems are pervasive in financial markets and business organisations, and that accounting plays an important role in resolving such problems. JBFA welcomes both theoretical and empirical contributions. Nonetheless, theoretical papers should yield novel testable implications, and empirical papers should be theoretically well-motivated. The Editors view accounting and finance as being closely related to economics and, as a consequence, papers submitted will often have theoretical motivations that are grounded in economics. JBFA, however, also seeks papers that complement economics-based theorising with theoretical developments originating in other social science disciplines or traditions. While many papers in JBFA use econometric or related empirical methods, the Editors also welcome contributions that use other empirical research methods. Although the scope of JBFA is broad, it is not a suitable outlet for highly abstract mathematical papers, or empirical papers with inadequate theoretical motivation. Also, papers that study asset pricing, or the operations of financial markets, should have direct implications for one or more of preparers, regulators, users of financial statements, and corporate financial decision makers, or at least should have implications for the development of future research relevant to such users.

期刊最新文献

Issue Information Out‐of‐sample predictability of firm‐specific stock price crashes: A machine learning approach Verifiable content in social media stock‐analysis articles: The long and short of it Top management team incentive dispersion and management earnings forecasts Coarse performance evaluation for envious agents