基于非线性分位数自回归的股市收益自相关性研究

Stock index return autocorrelations analysis based on nonlinear quantile autoregression

股市收益的自相关性是金融市场研究的热点问题,对股票价格预测及股市风险度量具有重要意义。由于收益序列存在尖峰厚尾及非线性特征,传统的线性均值模型往往难以具有代表性。在分位数回归框架下,以正负向收益作为分段机制,建立非线性分位数自回归模型,用于刻画收益序列的自相关特征。选取上证综指日收益率、周收益率及月收益率数据进行实证研究,结果表明,不同时间频率的收益序列自相关性较为一致,且存在典型的非线性和异质性特征。当前期收益为负向时,收益序列呈现出低分位点处(低迷市场)强正自相关、中位点处(温和市场)弱自相关、高分位点处(积极市场)强负自相关的特点;而当前期收益为正向时,结论与之相反。最后对结论进行分析并给出相应政策性建议。

The stock return autocorrelation is a hot topic in financial markets research,which is particularly important in stock return forecasting and market risk measurement.Because of the return series has a fat tail and nonlinear characteristics,The traditional linear mean regression model is often difficult to representative.In quantile regression framework,we use the positive and negative return as threshold mechanism,and establish nonlinear quantile regression models,which is used to describe auto-correlation characteristics of the return series.The empirical results are based on daily,weekly and monthly returns of Shanghai composite index,we find that different frequency return series are significant autocorrelative and have typical nonlinear and heterogeneous characteristics.while the past return is negative,we find the return series exhibit strong positive autocorrelation at lower quantiles,and weak autocorrelation at the median and strong negative autocorrelation at upper quantiles.while the past return is positive,we get the contrary conclusions.Finally,the conclusion is analyzed and the corresponding policy suggestions are given.