Oil demand shocks reconsidered: A cointegrated vector autoregression

We reconsider the conclusions about the importance of oil demand shocks and the unimportance of supply shocks reported by Kilian (2009). We investigate whether the proxy for worldwide real economic activity, dry bulk maritime freight costs, represents anything more than transportation costs by analyzing the relation between these costs and oil prices. The meaning of this variable is critical because transportation costs appear on both sides of the equations estimated by Kilian, directly as dry bulk maritime freight costs and as part of the measure for oil prices. We also investigate the effects of representing oil supply with an aggregate of OPEC and non-OPEC production because they likely use different criteria to chose output. Finally we investigate Kilian's use of the first difference of supply while the other variables in his model are represented as levels. The results suggest that OPEC and nonOPEC nations use different criteria to set output and that reductions (increases) in OPEC production raise (lower) oil prices. The elements of the cointegrating relations, their loadings, and impulse response functions suggest that the positive relation between dry bulk maritime freight costs and oil prices simply represents the effect of higher oil prices on transportation costs. Sensitivity analyses suggest that these differences are caused by including transportation costs in the measure of oil prices, aggregating OPEC and non-OPEC productions, and using a very long lag length to estimate the VAR. Together, these results suggest that conclusions about the importance of demand shocks and the unimportance of supply shocks are not robust to alternative specifications that are consistent with many empirical findings about the world oil market.