基于二分网络演化的市场波动与羊群行为

Market Volatility and Herding BehaviorBased on the Evolution of Bipartite Network

在交易过程中,作为智能A gen t的交易者不但学习他人的交易策略,而且会"创新"性地提出自己的交易策略,从而在引发股票市场上不同程度的波动与羊群行为。本文首先将交易者与股票的"多对多"关系等价转换为"一对多"关系,构建了由股票和交易者组成的二分网络市场模型,并给出了交易与定价规则;然后,利用计算实验的方法对模型进行仿真实验,得到了与真实股价相似的时间序列以及不同"创新"概率下的持股分布;通过理论推导,我们发现了三种不同的持股分布:具有指数截断的幂律分布、二项分布和脉冲分布。最后,给出了研究结论与启示。将计算实验与复杂网络应用于行为金融研究具有较强的理论价值,同时对于投资者和监管方来说都有一定的借鉴和参考意义。

During the process of transaction,the traders as the intellectual agents do not only study other＇s trading strategy,but also present their trading strategy innovatively by themselves,which triggers some different levels of market volatility and herding behavior in stock market.Firstly,we change the ＂many-to-many＂ relationship between traders and stocks to ＂one-to-many＂ relationship equally,construct the bipartite network model of traders and stocks,and present some rules of trading and pricing.Secondly,based on the computational experiments,we obtain the time series which are similar to real stock price very much and the stock-holding distributions of some ＂innovative＂ probabilities.Through the theoretical derivation,we find that there are three types of stock-holding distribution： power-law distribution with exponential cut-off,binomial distribution and fluctuating distribution.At last,we also present some conclusions.It has a strong theoretical value that agent-based computational experiment and complex network are used in the research of behavioral finance.At the same time,the results will give some reference to investors and regulators.