Authors' note: Subsequent to completing this Working Paper, we realized that the way we constructed the weighted relative prices, ri, as described on page 8, is not invariant to the rate of inflation and this introduces a bias in favour of the menu-cost hypothesis. Preliminary results with a correction to this problem reveal that it affects the quantitative results. We are currently revisiting our empirical analysis more completely to consider the quantitative and qualitative implications of removing this bias.

The menu-cost models of price adjustment developed by Ball and Mankiw (1994;1995) predict that short-run movements in inflation should be positively related to the skewness and the variance of the distribution of disaggregated relative-price shocks in each period. We test these predictions on Canadian data using the distribution of changes in disaggregated producer prices to measure the skewness and standard deviation of relative-price shocks. We find the Canadian data, both in the context of partial correlations and standard price Phillips curve equations, are highly supportive of the predictions that arise from the menu-cost models. Indeed, we find that the positive relationship between inflation and the skewness of the distribution of relative-price shocks is one of the most robust features of the Canadian Phillips curve and significantly improves our ability to explain inflation dynamics.