Time-Varying Causality between Oil and Commodity Prices in the Presence of Structural Breaks and Nonlinearity

The recent commodity price boom has spurred interest to understand determinants of commodity price movements. This paper investigates the causal relationship between oil prices and the prices of 25 other commodities, which include both metals and agricultural products, in the presence of instability and nonlinearity. For this purpose, we make use of a long annual time series dataset spanning from 1900 to 2011, and analyze time-varying Granger causality test, since the inference drawn based on linear Granger causality tests could be invalid due to structural breaks and nonlinearity – which we show are present in the relationship between the variables of interest. We find that, under the case of time-varying causality there are fewer rejections of the null, than under the standard linear Granger causality test, thus highlighting the importance of accounting for instability and nonlinearity. Relying on the time-varying causality test, we observe stronger evidence of other commodity prices in predicting (in-sample) oil prices (15 cases) than the other way around (7 cases).