Modelling nonlinearities in GDP. Some diferences between us and spanish data

Abstract

This paper studies the dynamic behaviour of US and Spanish GDP constructing final form (univariate) models, which can explain the non-linear aspects to be found in the trend and cyclical component of the mentioned aggregates. The main directions of research into switching regime models and the main results obtained in their application to the United States GDP are discussed. The option taken in this paper is to develop on the TAR models followed by Tiao and Tsay (1994) with two useful modifications. The nonlinear model for Spanish GDP distinguishes three different regimes and fits the data and forecast better than lineal models. The nonlinear model for US considers four regimes and fits similar than linear models, is stable and its forecasting performance is better or worse than the one obtained with linear models depending on the presence of recessions in the period of forecasting. But linear models for US GDP must be discarded because are unstable. This, in turn, emphasises the interest for nonlinear models. The US and Spanish business cycles shows similarities in the sense that both economies enter into recession as a result of negative shocks, expansions show short cyclical oscillations, periods of recovery after a recession are abrupt and there is evidence of positive duration dependence in recessions. Nevertheless, while US GDP shows a dynamic behaviour that pushes the economy out of recessions, Spanish GDP requires the help of positive shocks. Net exports and possibly the competitiveness of exports may play a highly important role in the Spanish business cycle and inventory investment could be the main factor in the case of US. The evidence of positive duration dependence is not strong enough for US expansions while in the case of Spanish expansions this effect could have a U-shape.