Multifractal Hurst analysis of crude oil prices

Daily records of international crude oil prices are studied using multitractal analysis methods. Rescaled range Hurst analysis provides evidence that the crude oil market is a persistent process with long-run memory effects. On the other hand, height-height correlation analysis reveals evidence of multifractal structures in the sense that the crude oil dynamics displays mixing of (rough) Hurst exponents. The existence of two characteristic time scales in the order of weeks and quarters is discovered and the corresponding prices dynamics are extracted using moving-average-based filtering. These results seem to demonstrate that the crude oil market is consistent with the random-walk assumption only at time scales of the order of days to weeks. A plausible oil price formation mechanism is discussed in terms of the market dynamics at three different time scales. (C) 2002 Elsevier Science B.V. All rights reserved.