Social media discussion and the market reaction to earnings announcements: evidence from China

ABSTRACT

This study investigates the influence of earnings-related social media discussion on stock returns surrounding the earnings announcement. By using deep learning approach to develop a new measure of market-wide earnings expectations based on Chinese Sina Weibo, we find that the constructed Weibo optimism index is positively related to the upcoming earnings announcement returns within a short window. This positive effect is more pronounced for firms with lower-quality earnings and firms with no traditional media coverage. Additionally, we also find that an increase in our Weibo optimism index indicates stronger post-announcement return reversals over a longer window. These results imply that earnings-related social media discussion is largely driven by sentiments, leading investors to overreact to earnings announcements. Our evidence supports the view that social media impedes stock price efficiency in financial markets.

KEYWORDS:

• Social media
• investor sentiment
• earnings expectation
• earnings announcement
• market reaction

JEL CLASSIFICATION:

• G14
• G30
• M41

Disclosure statement

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

Notes

¹ For instance, in China, there exists a group of ‘stock market black mouths’ who use their influence on the internet to lure investors. See ‘China tightens legal traps to snare stock market rats’ available at http://www.china.org.cn/business/2019-06/29/content_74934293.htm.

² The API code can be found at https://cloud.tencent.com/document/product/271/35497.

³ Tencent AI open platform uses hundreds of billions of large-scale internet corpus and combines deep neural network models (e.g. LSTM, BERT) to continuously improve the accuracy of sentiment analysis results, thus it can more precisely and objectively tell the sentiment type of the social media information content.

⁴ Sina Weibo verifies social media celebrities that are influential and have large number of followers. We identify the verified Weibo users whose usernames are appended with a coloured ‘V’ tag.

⁵ Given that many firms choose to announce earnings at the same date, including time fixed effects in the model would absorb the explanatory power of the market-wide Weibo optimism.

⁶ A possible prediction might be that earnings-related online discussion also contains material information. However, if value-relevant information is dominated in social media, it could help investors recognize accounting information. Hence investors’ overreaction to Weibo optimism should not vary with earnings quality.

Additional information

Funding

This work was supported by the National Social Science Fund of China (Grants No. 19BJY008, 20&ZD105),the Humanities and Social Science Foundation of the Ministry of Education ofChina (Grant No. 18YJA790055), the Graduate Project of Wuhan University (GrantNo. 1201-413100137).