Media-based investor sentiment and stock returns: a textual analysis based on newspapers

ABSTRACT

This study constructs a media-based investor sentiment index based on a textual analysis of China’s leading financial newspapers. We employ both the Word2Vec technique and the dictionary method to measure the aggregate textual tone in the media news. Based on a sample of publicly listed A-share firms in China, we document that media-based sentiment is positively (negatively) related to the cross-section of stock returns over a short (long) horizon. Further analysis shows that higher information quality, measured by more analyst coverage, better audit opinions and non-governmental ownership could mitigate the effects of sentiment. Overall, our findings imply that media news contains important information for measuring the overall investor sentiment that drives the future stock price up in the short term and then down in the long term.

KEYWORDS:

• Dictionary method
• information quality
• media-based sentiment
• Word2Vec

JEL CLASSIFICATION:

• G12
• G14

Acknowledgments

An important part of the research was done when Yu He was a visiting PhD. student at Nanyang Technology University, Singapore. Yu He gratefully acknowledges research support from Chinese Top University Graduate Students Studying Aboard Program under China Scholarships Council (No. 201906490030). Xuankai Zhao gratefully acknowledges financial support from the National Social Science Foundation of China (No. 18AZD007), Beijing Outstanding Young Scientist Program (No.BJJWZYJH01201910034034), and Beijing Municipal Social Science Foundation (No. 17JDYJB019).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ Market-based sentiment is the first principle component of six proxies for sentiment: closed-ended fund discount, turnover, number of initial public offerings (IPOs), average first-day IPO returns, share of equity issues in total equity and debt issues, and dividend premium.

² The root positive dictionary contains ‘upper price limit (涨停)’, ‘bull (牛市)’ and ‘rising of stock price (上涨)’. The root negative dictionary contains ‘lower price limit (跌停)’, ‘bear (熊市)’ and ‘dropping of stock price (下跌)’.

³ Considering that IPOs are heavily affected by the Chinese government, we do not include IPO-related variables, such as the number of IPOs, average first-day IPO stock returns and proportion of equity issues in the total issue of equity and long-term debt. Instead, we include two new variables: the deposit savings growth rate and the growth rate of new investor accounts (Firth et al., 2015).

⁴ We examine the effect of media-based sentiment on the stock returns for different periods (i.e. h = 1, 2, 3, 4, …, 15) and present the associated results in Appendix D.

⁵ According to Chu, Qin, and Fang (2019), during the period 2010–2016, the scale of margin trading is up to 99% and the trading of short selling is only 1%.

Additional information

Funding

This work was supported by the Beijing Outstanding Young Scientist Program [No. BJJWZYJH01201910034034], National Social Science Foundation of China [No. 18AZD007], Beijing Municipal Social Science Foundation [No. 17JDYJB019], and Chinese Top University Graduate Students Studying Aboard Program [No. 201906490030].