Abstract: Research Objectives:This paper tests Fisher Effect in China.Research Methods:The Markov regime switch structure is applied to modeling of co-integration rank, and the time variation of co-integration parameter is considered.The time varying rank and time varying parameter VECM model is established, and the parameters of the model are estimated by using the Bayesian method.Research Findings:By using the data about nominal interest and inflation from 1992 to 2016, the results show that there is a transition between the nominal interest rate and the inflation rate in the time varying parameter VECM model and the first order single integer I (1) data time varying parameter VAR model.During the whole sample period, the two elements of stochastic processes, which are composed of nominal interest rate and inflation rate, are mainly time varying co-integrated.Estimated normalized co-integration vector shows that existing Fisher Effect is dominant in China during the entire sample period.In economic new normal period, non-existent Fisher Effect is in a dominant position.There exists weak Fisher Effect in China since July 2015.Research Innovations:The Markov regime switch structure is applied to modeling of co-integration rank, and the time varying rank and time varying parameter VECM model is established.Research Value:The conclusion helps to re-understand Fisher Puzzle.

Key Words: Fisher Effect; VECM; Time Varying Co-integration Rank; Singular Value Decomposition; Markov-switching