Research on Applying Probabilistic Approach to Understanding and Forecasting Business Cycles

本课题是在调研最新研究趋势的基础上，对商业波动的概率方法进行研究。特别地，我不仅使用通常的正态分布，而且还使用稳健分布（如t分布）来检查状态切换模型。此外，使用日本实际经济周期指标，如扩散指数（DI）、综合指数（CI）、累计DI、鲁棒化经济周期指标，考虑了使用各种分布的差异和使用鲁棒化分布的效果。使用稳健分布的效果在CI中尤为显著。此外，虽然对模型进行了估计，并对各个样本时期的性能进行了检查，但我无法得出一般的结论，这将被视为未来的研究。此外，利用向量自回归（VAR）模型，分析了与经济周期波动密切相关的变量之间的相关性。从单位根检验开始，对协整分析进行检验，对ECM （Error Correction Model）和VAR模型进行估计，然后进行格兰杰因果检验，检验变量之间的相互关系。特别是，这些程序是在分开的样本期间计算的，通过观察不同的相关结构，提供了各种经济影响。

This project is a research on a probabilistic method for business fluctuations on the basis of surveying the latest study trend. In particular, I examined the Regime Switching Model by using not only usual normal distribution, but also robust distributions such as t distribution. Moreover, the difference between using various distribution and the effect of using robust distributions are considered using real Japanese business cycle indicators such as Diffusion Index (DI), Composite Index (CI), cumulated DI, robustized business cycle indicators. An effect using the robust distribution was particularly striking in CI. In addition, though the models are estimated and the performance was checked our for various sample periods, I could not reached to the general conclusion, which would be considered as a future study.In addition, utilizing the VAR (Vector Autoregressive) model, correlation between the variables which closely related to business cycle fluctuation was analyzed. Starting from the tests of unit roots, cointegrateion analysis was tested, ECM (Error Correction Model) and VAR model are estimated, and then Granger causality tests are conducted for examining interrelationship between variables. Particularly, these procedures were calculated for divided sample periods, which, by looking at the different correlation structure, provided various economic implications.