中文摘要：

在传统ARCH中引入体制服从2个状态Markov过程的SWARCH-t（2，3）模型，并与传统GARCH模型中误差项服从的正态、t、GED分布相比，SWARCH-t（2，3）模型较大的提高上证指数的拟合能力，较好的改善了估计精度和对回报序列的描述，解决了传统GARCH高的持续性与较差的预测能力之间的矛盾。同时发现体制25I起的波动是体制1的4．23倍，波动在分解为ARCH波动和体制波动之后，其中的ARCH波动持续性较传统GARCH模型各种分布相比大幅减少，而体制波动的持续性却较高，但这种持续性会因不断出现的政策发生切换。

英文摘要：

The volatility estimated by the traditional ARCH or GARCH shows high persistence. The reason is that the regime structural change abrupt in variance progress is not captured. In this paper, the regime-switching ARCH is introduced and the persistence of volatility is investigated to China stock. The result shows that, after the volatility is separated into the volatility of ARCH progress and regime, the persistence of ARCH progress is dramatically decreasing compared with the traditional GARCH models of the distribution of Gaussian, t and GED while there exists a quite high persistence of regime. But the persistence switches frequently with the policy.