Three essays on business cycle asymmetrie

In my first essay, I investigate whether business cycle dynamics are characterized by asymmetries. I provide evidence on this issue using a variety of time series models. My approach is fully parametric and testing strategy is robust to any conditional heteroskedasticity, outliers, and/or lone memory that may be present.;My results using data for five industrialized countries indicate fairly strong evidence of nonlinearities in the conditional mean dynamics of the G DP growth rates for Canada, Japan, and the US. For France and the UK, the conditional mean dynamics appear to be largely linear.;This study shows that while the existence of conditional heteroskedasticity and long memory does not have much effect on testing for linearity in the conditional mean, accounting for outliers does reduce the evidence against linearity.;In second essay I am testing for the possible existence of neglected nonlinearities using neural network tests. Neural networks are considered to be robust to nonlinear econometric models. The results for this essay showed that neural network models outperformed linear and nonlinear econometric models. The results from the second essay are the basis for the third essay where I am comparing in sample as well as out of sample forecast performance of neural networks models to linear models.;Results from the first two essays provide a strong evidence to conclude that nonlinearities are prevalent in all the five countries studied. Although the traditional time series models in the first essay failed to detect nonlinearities in France and the UK data series, neural networks did find evidence for such characteristics.;Results from essay three show that although in sample performance of neural network models is superior to linear models, out of sample performance of linear models is better than neural network models.;These studies confirm that business cycles in these 'G5' countries are characterized by nonlinearities. That means we cannot evaluate the impact on any monetary policy shock in these countries based on linear models. Further, we also conclude that neural network models outperform linear and traditional nonlinear econometric models based on in sample forecasting. However, out of sample forecasting performance of linear models is superior to neural network models.