| The annual report(10-K Files)is one of the most important channels of communication between listed companies and investors.The traditional readability index of annual reports measures the level of difficulty for readers to understand the content,and research on the readability of annual reports is already well-established.However,with the continuous advancement of technology,the way investors obtain information has changed,and the audience of annual reports has gradually shifted from humans to automated programs.Thus,traditional readability indexes are no longer suitable as a measure of the difficulty of obtaining information for investors.Therefore,this paper starts from the underlying language of the annual report HTML files and considers it from the perspective of program reading to construct a new AI readability index to measure the difficulty of obtaining and processing annual report information.The paper can be divided into three parts.In the first part,the paper uses web crawling technology to obtain annual reports from 20,590 US-listed companies from the year2000 to September 2022,totaling more than 70,000 papers.Based on the underlying language and structure of the annual report files,the paper constructs an AI readability index by comprehensively analyzing aspects such as data identification,table extraction,external links,and file size.The second part is a series of empirical tests conducted on the constructed index.The empirical data used in this paper is the US stock market data for the same period as the annual reports mentioned above,including traditional readability indexes,company fundamental information,and stock price information.In this part,the paper proves through empirical analysis that there is no significant correlation between the AI readability index and the traditional readability index.Furthermore,the paper finds a significant positive relationship between the AI readability index and company fundamental information.In other words,larger companies’ annual reports usually have higher AI readability.Additionally,the paper finds that annual reports that have been audited and approved also have significantly higher AI readability.The final part of this paper analyzes the relationship between the AI readability index and the information content of stock prices.The information content of stock prices is an indicator that measures the market’s reaction to information.Since the AI readability index can reflect information processing costs,this paper aims to verify whether lower information processing costs can help stock prices better reflect market information.The paper finds that an increase in the AI readability index does lead to an increase in the information content of stock prices.Moreover,when using an indicator based on the price drift phenomenon in earnings announcements(PEAD)constructed based on analyst consensus forecasts to measure the information content of stock prices,the empirical results are consistent with expectations.That is,companies with high AI readability have significantly higher information content in their stock prices. |
