羊群效应的新测度指数及其对我国股市波动的预测作用研究

羊群效应作为一种典型的市场投资者行为异象，极易对股票价格波动形成影响.因此，在已有羊群效应测度方法基础上，提出了一种新的羊群效应测度指数，并以我国上证综指为样本，运用广义自回归条件异方差混频数据（GARCH-MIDAS）模型实证检验了该指数对上证综指波动率的影响及预测作用，并与多种常用的同频数据GARCH族模型和纳入经济政策不确定性指数（EPU）的波动率模型进行比较分析.实证结果表明，相对于EPU指数，纳入新羊群效应指数的GARCH-MIDAS模型具有更显著的样本内参数估计结果，同时可以更好地解释上证综指波动的长期成分.进一步，模型信度集合（MCS）检验和预测方向准确性（DoC）检验结果表明，纳入新羊群效应指数能够显著提高模型对我国股市波动率的样本外预测精度.最后，采用不同样本外预测天数、不同损失函数、不同滞后期预测以及基于深证成指样本的各种实证结果进一步证实上述结论的稳健性.

A new herd index and volatility forecasting of China's stock market

Herd behavior is a typical market anomaly, which tends to have a significant impact on stock price volatility. This paper introduced a new measurement of herd behavior (HERDI) based on previous measuring methods and analyzed the effect and forecasting ability of herd behavior (HERDI) on the SSEC volatility using GARCH-MIDAS model. Out-of-sample volatility forecasting comparison was also made between the GARCH-MIDAS model employed HERDI, the GARCH-MIDAS model employed EPU and several commonly used GARCH-type model. The empirical results show that results of parameter estimated by the GARCH-MIDAS model employed HERDI are higher significance than that by the GARCH-MIDAS model employed EPU and the HERDI can well explain the long-term section of stock market volatility. Meanwhile, the model confidence set (MCS) test and the direction-of-change (DoC) test also indicate that the HERDI significantly improve the out-of-sample forecasting accuracy of SEEC volatility. Lastly, further results of empirical tests using various out-of-sample periods, loss functions, time lags for explanatory variable and data sample from Shenzhen component index confirm the robustness of the conclusion above.