A Forecasting Model for Inventory Investments in Canada

The authors present an empirical model to forecast short-run inventory investment behaviour for Canada. As with other recent studies that examine this series, they adopt an error-correction framework. Estimations using non-linear least squares and quarterly data yield both a good model fit and good out-of-sample forecasts. Given the debate in the United States on whether the adoption by firms of new information-technology-based methods of inventory management led to a decline in the volatility of U.S. output growth, the authors examine this issue for Canada. Results of the heteroscedasticity-robust Quandt likelihood ratio test advocated by Stock and Watson (2002) reveal very different dates for structural breaks in the volatilities of the growth contribution of inventory investment and of Canadian output growth: 1984Q1 and 1991Q2, respectively. Thus, the authors conclude that the "inventory hypothesis" is likely not an important explanation for the decline in the volatility of Canadian GDP growth.