马尔可夫状态转换加随机波动瞬时利率模型实证研究

An Empirical Study of Stochastic Volatility Diffusion Model of Short-term Interest Rates with Markov-Switching

文章对中国瞬时利率动态行为进行了实证研究,比较了一类马尔可夫状态转换加随机波动扩散模型。与以往研究不同,文章对模型所有参数采用基于Gibbs抽样的马尔可夫链蒙特卡罗模拟方法进行估计。同时,通过MAE(绝对误差平均值)、MRSE(平方误差均值)、调整R^2、对数损失函数LL以及非参数Wilcoxon检验对各种模型的样本内与样本外预测能力进行了分析与比较,结果表明:中国利率市场确实存在马尔可夫状态转换现象,其中Smith模型更适合刻画国内瞬时利率动态行为。

In this paper we empirically study the dynamic behavior of Chinese short-term rates using stochastic volatility diffusion models with Markov-switching.Unlike previous studies,we estimate all the parameters using a Gibbs sampling-based Markov Chain Monte Carlo algorithm.We compare the in-sample and out-sample forecasting abilities of these models and find that there really exists Markovswitching phenomena in the dynamic behavior of Chinese short-term rates and the Smith model is the best for forecasting.