Dataset on MD&A Texts and Financial Ratios for Predicting Financial Distress in Chinese A-share Listed Companies (2013–2020)

The dataset comprises two types of data: structured financial data sourced from the Guotai-An CSMAR database, which has been truncated to the 1st percentile and standardised, ultimately retaining 21 core financial indicators across five major categories; MD&A text data was crawled from publicly available annual reports on the Juchao Information Network. Following text cleaning, expansion using the FinBert domain-specific vocabulary, and chi-squared feature selection, it was quantified into four features: text sentiment, innovation/risk information content, and text similarity.

All corpus statistics were derived solely from the training set to strictly avoid the disclosure of future information. The dataset is uniquely identified by ‘stock code – year’; each row corresponds to a single annual observation for a company, whilst each column contains the company identifier, 25 feature indicators and a binary ‘ST/*ST’ label indicating financial distress.