Beyond the numbers: The effect of 10-K tone on firms' performance predictions using text analytics

Abstract

Investors use annual reports that companies mandatorily disclose every fiscal year-end when they make important investment-related decisions. The Securities and Exchange Commission in the United States has implemented and monitors the "Plain-English Rule" to create a transparent and readable annual report. However, only a general guideline exists for the description of annual reports (Plain English Rule), and there are no specific regulations on volume and format. As such, it is risky to interpret business managers' positive expressions because the data can include business managers' subjective opinions and viewpoints from the companies' perspectives. Additionally, the business managers' unrevealed tendencies, such as strategic reporting or cognitive bias, remain unknown. Therefore, this study focuses on the narrative sections of the annual reports that companies need to disclose after every fiscal year-end and uses text mining to determine the relationship between the assessment information of companies created by business managers and the performance of companies. The methodology of this study is grounded in text mining. To analyze these relationships, we collected all the 10-Ks of all public firms in the United States and employed the text-mining method to identify the tones of these 10-K narratives to determine whether they changed in line with current earnings levels. Additionally, we explore factors that could give rise to tone flexibility among reports and examine companies whose tone was more positive relative to their current performance to ascertain how future performance might differ from current performance.