This paper examines the ability of linear and nonlinear models to replicate features of real Canadian GDP. We evaluate the models using various business-cycle metrics. From the 9 data generating processes designed, none can completely accommodate every business-cycle metric under consideration. Richness and complexity do not guarantee a close match with Canadian data. Our findings for Canada are consistent with Piger and Morley's (2005) study of the United States data and confirms the contradiction of their results with those reported by Engel, Haugh, and Pagan (2005): nonlinear models do provide an improvement in matching business-cycle features. Lastly, the empirical results suggest that investigating the merits of forecast combination would be worthwhile.