Research On The Evolutionary Process And Mechanism Of Herd Behavior In China’s Securities Market

Research on the evolutionary process of herd behavior in the stock market is one of the hot issues in the field of behavioral finance.In this article,the research content is carried out from two aspects:empirical research and mechanism research.In terms of empirical research,this paper focuses on the evolutionary process and characteristics of herd behavior in China’s securities market.In terms of mechanism research,this paper refers to the sequence decision model to construct a parameterized sequence decision model including market evolution pattern,market information structure,and transmission mode.In this paper,I take social learning patterns such as herd learning pattern,herd-anti-herd learning pattern,and Bayesian learning pattern as the research objects,and study the expected benefits and characteristics of various social learning patterns.Then,I discuss the intrinsic relationship between the expected returns of various social learning patterns and the market evolution pattern,the structure,and the transmission of market information.Finally,I explain the behavioral motives of investors choosing various learning patterns including the herd learning pattern.First,I empirically test the evolutionary process and characteristics of herd behaviors of institutional investors in China using the modified model based on LSV and Shi Donghui.Considering that the panel data distribution of market stock returns does not obey the normal distribution,I refer to the HS model proposed by Hwang and Salmon(2002)and combine the non-parametric method to construct a non-parametric HS model.And I studied the herd behavior of the three financial markets,namely the Shanghai A-share Market,the Growth Enterprise Market,and the SSE STAR Market.Then I discuss the characteristics of market herd behavior grouped by market ups and downs,the returns,and market capitalizations of stocks invested by market investors.Research shows that Chinese institutional investors do not have significant overall herd behavior,buying herd behavior,and selling herd behavior during the entire sample period.The degree of herd behavior of institutional investors is related to factors such as positive and negative gains of stocks,market capitalization,and specific industries.The Shanghai A-share Market,the Growth Enterprise Market,and the SSE STAR Market have significant reverse herd behavior throughout the sample period.And the degree of herd behavior in these three markets is related to the rising and falling periods of the market,the return rate,and the market value of the stocks invested by market investors.Secondly,I take the SSE STAR Market as the research object,refer to the method of Hu and Chi(2012)to construct rational market sentiment and irrational market sentiment indicators,and construct indicators such as market volatility and market liquidity.I include the above indicators and herd behavior indicator into the TVP-VAR model,and discuss the evolutionary process of herd behavior under the influence of the above-market factors based on different lag periods and different time points.The main results are as follows:(1)The evolutionary process of herd behavior in the SSE STAR Market under the influence of rational and irrational market sentiments is:positive rational market sentiment inhibits the generation of herd behavior,while positive irrational market sentiment promotes herd behavior most of the time;the short-term impact of two market sentiments on herd behavior is significantly stronger than the long-term impact,but the impact of different types of market sentiment on herd behavior show heterogeneity and time-varying characteristics in the early stage of the market.(2)The evolutionary process of herd behavior in the SSE STAR Market under the influence of market volatility is as follows:the positive impact of market volatility on herd behavior will vary significantly over time.On the one hand,the positive impact of market volatility on herd behavior will accumulate over time;on the other hand,the positive impact of market volatility on herd behavior is more obvious in the initial stage,which shows that the response of herd behavior to asset price fluctuations is lagged.(3)The evolutionary process of herd behavior in the SSE STAR Market under the influence of market liquidity is as follows:the feedback effect of market liquidity on herd behavior has obvious heterogeneity over time.On the one hand,in the early stage of the market,market liquidity has a significantly positive effect on herd behavior,and the short-term and long-term effects are almost the same.In the early stage of the outbreak of the new crown pneumonia in China,the positive impact of market liquidity was significantly reduced and even turned into a suppressive effect.Subsequently,its positive impact showed an oscillatory rebound.In the middle and late stages,there are obvious differences between the short-term and long-term effects of market liquidity on herd behavior,which is reflected in that it has a greater stimulating effect on shortterm herd behavior and a greater inhibitory effect on long-term herd behavior.Finally,referring to the sequence decision model constructed by Banerjee(1992),I designed the factor parameters to describe the structure,cost,and transmission mode of market information.Then,I construct a parameterized sequence decision model under market evolution patterns including information-driven,hybrid-driven based on the market-neutral state,and hybrid-driven based on the market non-neutral state.And I have studied the expected benefits and characteristics of various social learning patterns,discussed the intrinsic relationship between the expected benefits of various social learning patterns and the market evolution patterns,the structure,and the transmission of market information.I have verified that none of the three social learning patterns is constant and optimal under the specific market information structure,transmission model,and market evolution patterns.They all have their relative advantages and existential value.The main results are as follows:(1)The expected return obtained by an uninformed agent after choosing the Bayesian learning pattern has nothing to do with the structure and transmission mode of market information but is only related to whether the market state is neutral.(2)In different market evolution patterns,there are partial differences in the expected returns of uninformed agents choosing the herd learning pattern.The expected returns are related to factors such as market endowment,information structure,cost,and transmission mode.(3)In different market evolution patterns,since the herd-anti-herd learning pattern has a rectification mechanism,excess returns can be obtained,and the excess returns are only related to the cost of market information.(4)Under the assumption of a market-neutral state,uninformed agents choose the herd learning pattern significantly better than the Bayesian learning pattern.When choosing between the herd learning pattern and the herd-anti-herd learning pattern,uninformed agents need to consider factors such as market endowments,the structure,and the transmission of market information.(5)Under the assumption of a market non-neutral state,uninformed agents can obtain excess returns by choosing the Bayesian learning pattern,and choosing the Bayesian learning pattern is better than the herd-anti-herd learning pattern.When choosing between the herd learning pattern and the Bayesian learning pattern,uninformed agents need to consider factors such as market endowments,and the structure and transmission of market information.Especially in the case of reverse transmission of market information,choosing the Bayesian learning pattern is optimal.(6)In the case of nonlinear market information transmission,if the information transmission is positive,no matter how the market information structure changes,the non-linear characteristics of market information transmission will cause the market state to no longer have neutral characteristics,and the market will appear "1" state.The probability and the probability of appearing "0" state are only related to the nonlinear characteristics of information transmission;and the nonlinear characteristics of market information transmission will make the asymptotic expected return of the uninformed agent when choosing the Bayesian learning mode greater than 50%.