Modelling the business cycle in South Africa : a non-linear approach

In this paper the South African business cycle is modeled, using a simple linear method and comparing it to non-linear methods. This is useful to address the debate between the Classical and Keynesian economists regarding their views on the business cycle. They believe in a stable economy with exogenous shocks and an unstable economy with an endogenous business cycle respectively. Linear models are usually associated with the Classical view and non-linear models with the Keynesian view. A detailed discussion on the non-linear model-building process, with particular emphasis on the family of STAR models is done. The South African GDP is used and AR, TAR, LSTAR and ESTAR models are fitted and compared. It finds that a parameterized nonlinear model (such as the family of STAR models) outperforms the simple regression model. This is due to asymmetric behaviour in the GDP data and the possibility of a threshold between a recession and an expansion. The results in this paper support the structural or institutional view of business cycles, which states that economic fluctuations are caused by various structural or institutional changes