Trend-Cycle Correlation, Drift Break and the Estimation of Trend and Cycle in Canadian GDP

This paper argues, using Monte Carlo experiments, that bivariate correlated unobserved components (UC) framework delivers more accurate estimation results for trend and cycle parameters than the univariate framework. The paper estimates stochastic trend and cyclical fluctuations in Canada from a bivariate, correlated UC model with drift breaks. In contrast to the univariate results, bivariate estimation shows a fairly volatile stochastic trend after accounting for the drift break along with a negative trend-cycle shock correlation. The estimated cyclical component is moderately large, persistent and consistent with ECRI denoted Canadian recessions.