高维平滑转换因子模型的一致估计

Consistent Estimation of Large Dimensional Smooth Transition Factor Models

本文设定高维因子模型的因子载荷服从平滑区制转换结构,模型参数的一致估计可通过两阶段估计方法给出。在第一阶段,通过主成分方法估计因子变量;在第二阶段,估计的因子变量视为已知变量,通过非线性最小二乘法估计因子载荷和平滑转换参数。理论研究和随机模拟表明本文提出的两阶段估计方法具有良好的大样本性质和有限样本表现。在实证部分,基于高维平滑转换因子模型研究了美国股票收益率数据的共变特征和非对称效应,结果表明平滑转换因子模型可以较好地刻画美国股票收益率的共变特征和区制转换行为。

In This paper,we specify the loadings evolve as smooth transition regressions and show that consistent estimates of parameters can be obtained by two-stage estimation.We estimate the factors in the first stage by the principal component,and the loadings and smooth transition parameters by nonlinear least squares in the second stage.We propose model selection criteria and a series of linearity test.Theoretical research and stochastic simulation show that the two-stage estimation method proposed in this paper has good large sample property and good finite-sample performance.In the empirical part,smooth transition factor models are applied to the daily US stock returns.The empirical results provide clear evidence in support of the nature of co-movement and smooth regime shift in daily returns.