TIME SERIES INVESTIGATION OF TURKISH BUSINESS CYCLES USING REGIME SWITCHING MODELS

Zeynep ŞENYÜZ (Marmara University)

Last fifteen years have witnessed an explosion of interest among researchers in using nonlinear models. The idea that time series are subject to change over time gave rise to the development of regime switching models in which dynamic behaviour is state dependent. This paper employs various models of this class to analyze the cyclical pattern of the Turkish economy. Following Hansen (1997) and Hamilton (1989), Threshold Autoregression (TAR), Self-Exciting Threshold Autoregression (SETAR) and Markov-switching (MS) models are fitted to post-1987 quarterly GNP and post-1986 monthly IPI data. The method to determine endogenous structural breaks suggested by Hansen (2001) is also applied to assess the estimation results. In a highly non-standard testing environment, specification tests of the models are conducted by taking advantage of the simulation methods. We capture the asymmetry over the different phases of the business cycle in which recessions are short and abrupt while expansions are moderate and highly persistent. We also discuss the striking events of the time in the light of our results. Most importantly, we find strong evidence that the Turkish economy has experienced structural breaks in the considered period, of the type presented in the literature.