On the Stationarity of Exhaustible Natural Resource Prices: Misspecification Effects

In this paper we examine whether the real prices of eleven natural resource commodities exhibit stochastic or deterministic trends. A common methodological feature in the relevant empirical literature, most of which published in the Journal of Environmental Economics and Management, has been so far the application of univariate tests for unit roots. In these tests the real price for each commodity is tested for unit roots in isolation from all other natural resource commodity prices. We claim that this approach is likely to produce spurious inferences concerning the true number of unit roots, since it ignores any possible dynamic interactions among the available set of nominal prices. We suggest that the hypothesis of stationarity of real commodity prices should be properly defined and tested within a multivariate error correction model, which explicitly accounts for all possible linear interdependencies among the series involved. In such a framework, the stationarity of the real prices that participate in the system depends on whether the system exhibits sufficient cointegration with a specific cointegration matrix. Our empirical results suggest that within this multivariate framework, eight of eleven real prices of exhaustible natural resource commodities satisfy the restrictions for being stationary. On the contrary, all of these eleven real prices appear to be non-stationary, when the unit root hypothesis is tested in the context of incomplete models.