Persistence, Long Memory, and Unit Roots in Commodity Prices

This article investigates the degree of persistence in several weekly and monthly agricultural prices (corn, soybeans, barrow and gilts, and milk) using long memory (fractional integration) techniques. The results indicate mean reversion (i.e., orders of integration smaller than one) in some of the agricultural prices like corn, milk, and barrow and gilts when the disturbances are autocorrelated. Further, we examine the stability across time in the degree of dependence, and the results indicate that the fractional differencing parameters have not remained constant across time. When we take into account a structural break we find that during the first subsamples, the series are stationary though highly persistent, with orders of integration close to 0 and with large autoregressive coefficients. However, for the periods after the break, the series seem to be nonstationary I(1).