Recent work on policy rules under uncertainty have highlighted the impact of output gap measurement errors on economic outcomes and their importance in the formulation of appropriate policy rules. This paper investigates the reliability of current estimates of the output gap in Canada. We begin by assembling a new data base of quarterly realtime output estimates which spans the post-WWII period and contains data vintages dating back to 1972. We use this with a broad range of univariate and multivariate output gap models to recreate ?contemporary? estimates of the output gap and then study how these estimates are revised over time. The nature and sources of these revisions are used to draw conclusions about the overall measurement errors associated with current estimates of the output gap. Relative to similar recent work with US realtime data, we find that revisions in Canadian output gaps are more important and that the role of data revision is less innocuous than previously indicated. We also show that using the change rather than the level of the output gap may only modestly reduce the measurement problem, and we investigate the relative importance of model uncertainty to overall measurement uncertainty.
