Unit roots in real primary commodity prices? a meta-analysis of the Grilli and Yang data set

The long-run behavior of real primary commodity prices, especially whether these series are trend stationary or contain a unit root, has been a topic of major debate in applied economics. In this paper, I perform a meta-analysis and combine the evidence of 12 representative studies on the subject in order to reach a unified conclusion about the presence of unit roots in these prices. The studies use different testing procedures, but share the common null hypothesis of a difference stationary process. Also, they use the individual price indices from the Grilli and Yang data set that is arguably one the most popular sources of long-term data on commodity prices. The combined evidence against unit roots in real primary commodity prices is strong: out of 24 cases, the unit root is not rejected in at most four. This rate means that real primary commodity prices are mean reverting and thus, to some degree, forecastable.