本文首先利用基于 分布的 模型来估计股票和债券收益率的动态相关系数，随后采用 模型刻画了我国股债相关系数的非线性动态特征，并进行了样本外预测，结果表明 模型的预测效果要优于 模型，这说明我国股债联动是非线性的，而且从低区制过渡到高区制的速度较缓慢。最后，本文分析了我国股债联动非线性特征产生的主要原因，并在此基础上提出了相关的政策建议。

At first the dynamic correlation of returns between stock market and bond market is calculated by using DCC-MVGARCH model based on distribution in this article, and subsequently the nonlinearity of stock-bond correlation is characterized by using LSTAR model, then the out-of-sample forecast is conducted, the results show that LSTAR model forecasts better than AR(1) model, which indicates that stock-bond co-movement in China is nonlinear. Furthermore, it shifts slowly from lower regime to upper regime. Lastly the main reasons causing nonlinearity of stock-bond co-movement are analyzed based on which some relative advices are proposed. Key words: