STAR Models: An Application for GDP Per Capita Growth Rate

The linearity assumption has been valid for a long time but most of the economic time series exhibit asymmetric behavior. The number of studies of nonlinear time series studies have increased in application by removing calculation difficulties. Markov regime switching models, TAR and STAR models are examples of nonlinear time series studies. GDP per capita is an important element of economical growth and has been the subject of many studies. This study covering the annual 1961-2017 data examined the nonlinear structure of GDP per capita growth rate for Turkey with STAR models. The serie is stationary according to the unit root test results and the delay length of the model has been found as two. Transition variables has been created by the delays of the dependent variable one by one and the linearity of the model has been tested against its nonlinearity alternative. The first delay has been selected as the transition variable. At the level of significance of 5%, the linearity hypothesis has been rejected. Findings of the study show that LSTAR model type is appropriate for growth. Since the appropriate model is LSTAR, transition variable has been standardized by its standard error. The parameter indicating the smoothness of transition shows that the transition from one regime to another is quick.