Forward-Looking Statement and Corporate Future Performance ——Based on Text Mining and Machine Learning Technology

Abstract

This paper analyzes the usefulness of the forward-looking information disclosed in the annual report based on the 2011-2016 financial report of listed companies. We construct a new forward-looking information disclosure index based on the forward-looking information in the annual report environment obtained through the Word2Vec machine learning technology and the word frequency of all the words calculated by the natural language processing and text analysis. Our results show that the level of forward-looking information disclosure is positively correlated with the company's future performance. Further research indicates that the stronger the readability of the company's annual report is, the stronger the relationship between forward-looking information and the company's future performance will be. And the relationship between forward-looking information and the company's future performance is more concentrated on firms with high information asymmetry. Besides, subsequent evidence finds that forward-looking information can effectively inhibit the opacity of the company's financial information, especially the company's positive earnings manipulation, thus further confirming the usefulness of forward-looking information.

Keywords: Forward-looking; future performance; financial information opacity; text mining; Machine learning (Word2Vec)

JEL Classification: M41, G14, D82, D21