Dismiss the gap? A real-time assessment of the usefulness of Canadian output gaps in forecasting inflation

We use a new real-time database for Canada to study various output gap measures. This includes recently developed measures based on models incorporating many variables as inputs (and therefore requiring real-time data for many variables). We analyze output gap revisions and assess the usefulness of these gaps in forecasting total CPI inflation and three newly developed measures of core CPI inflation: CPI-median, CPI-trim and CPI-common. We also study whether labour-input gaps, projected output gaps, and simple combinations of output gaps can add useful information for forecasting inflation. We find that estimates of excess capacity (the extent to which the economy is below potential) were probably too large around the 2008-2009 recession, as they subsequently tended to be revised down. In addition, we find that, when forecasting CPI-common and CPI-trim, some gaps appear to provide information that reduces forecast errors when compared with models that use only lags of inflation. However, forecast improvements are rarely statistically significant. In addition, we find little evidence of the usefulness of output gaps for forecasting inflation measured by total CPI and CPI-median.