The resource sector has traditionally played an important role in the Canadian economy, especially in the area of foreign trade. Over the past decade, total exports of commodities represent, on average, about forty one per cent of Canada's exports of goods and fifteen per cent of Canada's gross domestic product (GDP).
Consequently, changes in world commodity prices have historically been a key determinant of Canada's terms of trade, which in turn have affected the real income of Canadians.
The staff at the Bank of Canada (BOC) have designed the Bank of Canada Commodity Price Index (BCPI) to track the prices paid for key Canadian commodities. The BCPI is a fixed-weighted index of the spot or transaction prices of 23 commodities produced in Canada and sold in world markets. All components of the BCPI are priced in U.S. dollars. The choice of commodities is determined by their importance in Canadian production, subject to limitations imposed by data availability. For the purpose of this paper, we split the BCPI into three subindices: non energy commodity prices (BCNE), the West Texas Intermediate crude oil price (WTI), and energy prices excluding crude oil. To obtain real commodity prices, we divide by the U.S. GDP deflator.
In this paper, we employ three different empirical approaches to model commodity prices. For real BCNE prices, we use an approach that combines a structural vector autoregressive model (SVAR) with a single equation model. The SVAR is used to give us a historical decomposition of movements in real BCNE prices, and to project the permanent (or long-run) component of prices, while the transitory (or short-run) component of real BCNE prices is forecasted with a single equation model.
We find that this approach successfully captures the strong linkage of real BCNE prices with the world economic activity and the real U.S. effective exchange rate in the short run. A 1% positive shock to world economic activity leads to an approximate 6% peak response of real BCNE prices, while the response to the real U.S. dollar effective exchange rate shock is small but it is statistically significant and exhibits the expected sign. We also find that the variance of the transitory component of real BCNE prices accounts for approximately 60% of total real BCNE price variance. This result is consistent with numerous other studies of commodity prices. In terms of out-of-sample forecasting, our approach outperforms a SVAR model and an autoregressive (AR) model.
Two separate models are used for real crude oil prices and other real energy prices. For real crude oil prices, we use a statistical multiple structural-break approach to identify significant shifts in OPEC behavior. After controlling for these mean shifts, we find a very strong role for world economic activity in the determination of oil prices. We estimate that a 1% positive shock to world economic activity leads to an approximate 12% peak response of real crude oil prices with a lag of two to three quarters. In terms of out-of-sample forecasts, the real oil price model outperforms an AR model and a random walk (RW) model. For real prices of other energy components, we use an error correction representation which exploits the long-run relationship between real energy prices excluding crude oil and real crude oil prices. The U.S. output gap, a proxy for North American economic conditions, is shown to be a key factor explaining the short-run price variations.
Download
PDF Ebook Forecasting and Analyzing World Commodity Prices
