ARFIMA Models and the Hurst Measures: An Investigation of Commodity Daily Index and Futures Prices
29 Pages Posted: 17 Feb 2016
Date Written: December 14, 2015
In this paper, we study some important stylized facts of commodity daily index and future prices. Our study shows that, except few, all price processes have unit roots with stationary, skewed and kurtotic increments whose R/S and generalized Hurst exponents are greater than 1/2. As a result, ARFIMA(p-1,1 d,q) models with skewed and kurtotic shocks can be good candidates for modelling commodity daily index and future prices. On the other hand, since non-linear models are practically very difficult to implement, we study linear models with fat tail shock processes. We have observed that, even though linear models with fat tail shock processes cannot generate long memory time series, they can generate time series with generalized Hurst exponents similar to ones from the commodity daily index and future prices. This suggests that with regards to the generalized Hurst exponent, linear ARIMA(p-1,1,q) models with alpha-stable shock processes are good models with statistical characteristics very similar to commodity daily prices.
JEL Classification: C32
Suggested Citation: Suggested Citation