关键词: 社交媒体, 投资者行为, 行业波动, 溢出效应, 金融实践

Abstract:     Preventing and resolving financial risks is a key topic of financial work. Given the frequent price synchronization and turnover of industry indices in Chinese stock markets, prevention and control of inter-industry risk is a top priority. At the same time, the active behavior of investors in social media may accelerate and intensify the risk contagion among industries in the stock market. Therefore, clarifying the impact of investor behavior in social media on the risk of stock market industries and its inherent mechanism has important theoretical and practical significance for investors′ asset management, regulators′ policy optimization, and financial risk prevention and control.
   Research on the spillover of investor behavior mainly focus on three types: first, the spillover of investor behavior on price indicators; second, the spillover of investor behavior among markets; and third, the intermediary role played by investor behavior in other spillover relationships. Lu and Chen (2015) put forward theoretical hypothesis of the "investor sentiment spillover effects", but no subsequent studies have explored it in depth. Which investor behavior is the main factor influencing industry volatility in the social media context? Is there an investor behavior spillover effect? If so, what are the specific manifestations? 
   Based on the above problems, this paper empirically examines the impact of investor sentiment and investor attention in social media on realized volatility at the industry level, explores the existence of industry sentiment spillover effects and their propagation paths, and discusses the spillover effects after the COVID-19 epidemic. This paper collected a sample of 153 listed companies under 19 industries in China′s A-shares from 2019 to 2020. Specifically, we use a panel fixed effects model to study the impact of investor behaviors on industry volatility, while the directed acyclic graphs (DAG), the structural vector autoregression (SVAR) model, and the information spillover model of Diebold and Yilmaz (2012) are used in studies related to spillover effects. 
    We find in this study that: (1) From the social media perspective, investor sentiment is the main factor driving industry volatility in all investor behaviors under the absolute, relative, and differential three dimensions. In absolute and differential dimensions, higher investor sentiment and attention and wider investor disagreement significantly exacerbate industry volatility. While in the relative dimension, large changes in investor sentiment and attention dampen industry volatility. (2) Overall, there is inter-industry risk contagion via investor behavior in the stock market. Theoretical hypothesis of investor sentiment spillover effects is confirmed by 7 pairs of significant sentiment spillover relationships with a single clear direction of spillover. In the post-epidemic period, the number of industry pairs with sentiment spillovers decreases to 5, but the intensity of spillovers becomes stronger, and the scope of impact narrows and appears a continuous path.
    Based on the above findings, we argue that investors should choose a top-down asset allocation approach from sectors to individual stocks in the post-epidemic period. To uncover valuable new information for financial practices such as underlying selection, risk diversification, and asset pricing, the investors should not only take inter-industry linkages and the policies on industries and industrial chains into account but also the investor irrational behavior factors while making an investment decision.
    The main features of this paper are as follows: first, examining the impact of investor behaviors in social media on volatility from the industry level, and further revealing the existence of relatively hidden cross-industry sentiment spillover effects and their propagation paths. Second, portraying investor behaviors in social media by constructing diversified indicators into absolute, relative, and differential dimensions from two behavioral facets of investor sentiment and attention. Finally, the work of this paper provides a novel addition to the empirical investigation of spillovers on investor behavior and also sheds new light on investor asset management, regulatory policy optimization, and the prevention and control of financial risk in the post-epidemic period. In future research, we will consider examining differences in investor behaviors and their effects based on multi-platform information or continue exploring the characteristics and performance of institutional investors as well as other financial markets.

Key words:  social media, investor behavior, industry volatility, spillover effect, financial practice