This paper summarizes the results of recent research evaluating the Bank of Canada's Quarterly Projection Model (QPM). Because QPM consists of a steady-state model and a dynamic model, our evaluation work consists of two parts.

The first part assesses the calibration of QPM's core steady-state using a variant of Canova's (1994, 1995) Monte Carlo approach. Using parameter values drawn from prior distributions, we assess QPM's sensitivity to various plausible parameter values. Our approach differs somewhat from the recent literature in that it specifically takes into account the uncertainty that surrounds the estimates of the steady-state values we are trying to evaluate. Instead of attempting to match exactly the desired properties of the data, we calculate confidence intervals around the mean of the variable we wish to match, subsequently discarding parameterizations that result in simulated data falling outside this interval.

The second part of the evaluation uses artificial data, generated stochastically with QPM, to test the dynamic model's ability to replicate key historical moments. Autocorrelations, reduced-form regressions, and temporal bivariate correlations are used to compare historical data with data produced by QPM. We also assess the sensitivity of our results to the structure of the stochastic shocks and the specification of the monetary policy rule.

The results of the two evaluations reveal some strengths and weaknesses in the model. For example, while most of the parameter calibrations in the steady-state model appear reasonable, there are some parameters for which other values may be more appropriate. Similarly, while the dynamic model can replicate most of the key historical moments, some work is required to develop the linkages between foreign and domestic variables.


1. Robert Amano - Also affiliated with the Center for Research on Economic Fluctuations and Employment, Université du Québec à Montréal, Montréal, Québec, H3C 3P8, Canada