中文摘要：

为了更为有效地探究微观市场结构对股票价格的影响，本文在状态空间模型框架下，同时将交易方向、带方向的交易量、交易间隔、微观噪声以及跳跃因素引入状态方程和观测方程中，建立全面的市场微观结构模型，以反映各变量对交易价格序列产生的暂时性和永久性影响．然后，在采用基于粒子滤波的非参数离群点检测方法检测跳跃并对跳跃进行滤波的基础上，运用贝叶斯最小二乘方法对微观结构模型系数、波动率以及状态变量进行实时更新和估计，在此基础上分析跳跃日度分布、日内分布、个股跳跃和共同跳跃特征，并同时判断不同交易方向、交易量的订单对价格产生的影响．最后选取2015年股灾期间、沪市股票的逐笔数据作为研究样本进行实证研究．实证结果显示：我国股票高频时间序列跳跃和共同跳跃均存在日内季节性，且跳跃频率和大小与市场信息密切相关；买卖订单对交易价格及股票价值的影响并不是完全对称的，交易价格对于微观结构的敏感度、微观结构噪声也随市场波动而发生变化，熊市中都相对更高；考虑跳跃滤波并对参数进行实时估计可以提高价格对于交易的敏感度，能更好地捕捉市场微观结构在价格发现中的作用．

英文摘要：

In order to effectively detect the impact of microstructure on stock prices, this paper develops a general microstructure model based on state space model. It simultaneously introduces size, direction of trades, trading intervals, microstructure noise and thus reflecting the permanent and transitory impact jumps into state equation and observation equation these variables have on the stock price series. Then non-parametric outliers＇ detection method based on particle filters is applied for detecting jumps. In the process of filtering jumps, we can also obtain real-time estimation and updates for the microstructure parameters, volatility and state variables by using Bayesian OLS. The combination of these methods make it possible to analyze intra-day seasonality, daily pattern of jumps ＆ co jumps and simultaneously judge how trade directions and trade volume could influence prices as time goes by. Lastly, we utilize the tick-by-tick high-frequency data for Chinese stock market in 2015 turbulent period to implement empirical analysis. Results demonstrate that the jumps hidden in high-frequency stock price series show both intra-day and daily seasonality. Also the frequency and size of jumps are closely linked to the exogenous macroeconomic information. Buy and sell orders have an asymmetric impact on stock prices and fundamental values. The sensitivity of prices to microstructure variables and noise vary depending on the market conditions, among which is larger in good market circumstances. Therefore, this new model and method can efficiently increase the price sensitivity to transactions, thus better grasping the function of stock market microstructure in price discovery process.