Abstract
The zero-correlation restriction for trend and cycle innovations is empirically examined using an international data set. Unlike the result obtained by Morley et al. (Citation2003), a model with zero-correlation restriction is selected for all nine countries. Since the adequacy of zero-correlation restriction depends on the autoregressive (AR) order assumed, an information-criterion-based model-selection method is applied. In this method, first, possible alternative models are considered by changing the AR order and by including or excluding the zero-correlation restriction, and then the best model is selected from among them. Consequently, it is determined on the basis of the selected model whether the zero-correlation restriction should be imposed.