Estimating Turning Point Of Business Cycle On The Basis Of Particle Filtering Algorithm

Generally, the process of economic development always has fluctuations, and thateconomic fluctuation is always cyclical. To identify and examine the turning point of thebusiness cycle has always been one of the key problems in the field of business cycle theory.There are now two models that can be used to identify the business cycle’s turning point. Theone is the Markov Swith Model (MSM), while the other is the smooth transition auto-regression(STAR) model. The MSM would identify the state of the economy by the use of transitionprobability. To get a more precise estimation of the transition probability, the time span of thetime-series data is supposed to be long, and more than one switch of business cycle should becontained. However, as for our country, the data is not so sufficient.This paper uses sequential Monte Carlo methods to estimate the asymmetric effect inthe hidden Markov model so as to detect the turning point of China’s business cycle byEmploying monthly year-on-year growth rate of China’s gross industrial output andChina’s economic boom index. Unlike most Markov switching models, mechanismtransition probability in the model of this paper is time-variant transition probabilitydetermined by the beta distribution, of which the random part is determined by anexogenous variable, so can avoid the situation that the estimation precision is low becauseof the shortage of sample data. This paper estimates the parameters and potential statevariables in the model based on particle filtering method and Bayesian method andaccurately identify all the previous turning points of China’s business cycle.