Abstract:

Credit rating is a basic mechanism for the capital market， but the series of default events in recent years have raised the challenge for the ratings agencies. Since the credit rating is based both on quantitative and qualitative information， whether the rating agencies can interpret risk information in the qualitative text or not is examined. A text mining on the content of annual reports， which are the most reliable text information disclosed by the corporations. A textual content risk information factor of annual reports is built on the basis of text tone and complexity， which includes the portion of negative words， uncertainty words， professional words and complex sentences. With an empirical study on the credit rating of corporate bonds and the textual content risk information factor， the main findings indicate：（1） the textual content risk information factor of the annual reports has a significantly negative relation with the credit rating of corporate bonds， which means the rating agency can interpret the risk information in the text content in an efficient way；（2） the risk information in the annual report will have greater value for corporate bonds with no guarantee， issuers with lower market and financial risk；（3） the breakup of the rigid payment has brought greater pressure for the rating agencies and encourage the rating agencies to upgrade the ability to interpret the risk information， which can evidence the effectiveness of the credit rating in China to some extent.What’s more， the textual content risk information factor of the annual reports reflects the earnings quality and future operating performance， and has incremental information.

Key words: credit rating, textual analysis, risk information of annual report text content