A now-casting model for Canada: Do US variables matter?

We propose a dynamic factor model for now-casting the growth rate of the quarterly real Canadian gross domestic product. We select a set of variables that are monitored by market participants, track their release calendar, and use vintages of real-time data to reproduce the information sets that were available at the time when the forecasts would have been made. The accuracy of the forecasts produced by the model is comparable to those of the forecasts made by the market participants and institutional forecasters. We show that including U.S. data in a now-casting model for Canada improves its predictive accuracy dramatically, mainly because of the absence of timely production data for Canada. Moreover, Statistics Canada produces a monthly real GDP measure along with the quarterly one, and we show how the state space representation of our model can be modified to link the monthly GDP properly with its quarterly counterpart. Published by Elsevier B.V. on behalf of International Institute of Forecasters.