Long Range Dependence and Breaks in Energy Prices

Abstract

This study analyzes the degree of time persistence in United States energy prices (natural gas, crude oil, and bituminous coal) using innovative fractional integration models. Annual price data from 1919 to 2006 are used for natural gas and from 1870 to 2006 for crude oil and bituminous coal. The results indicate that energy prices are explained in terms of a long memory model that incorporates persistence components with autocorrelation. The degree of integration is above 1 for natural gas prices, around 1 for coal prices, but significantly below 1 for oil prices. These results are verified for parametric and nonparametric methods and a single break in the data is found in 1973 for natural gas and oil prices and in 1945 for coal prices.

Keywords:

• energy prices
• fractional integration
• persistence
• structural breaks

Notes

¹Models pertaining to the time path of exhaustible resource prices have postulated various factors: technical change, resource abundance, extraction costs, environmental constraints, speculative motives, etc. (see Hotelling, 1931; Barnett and Morse, 1963; Heal, 1976; Pindyck, 1980; and Slade, 1982, among others).

²The choice of the bandwidth clearly deals with the classical trade-off bias/asymptotic variance: The asymptotic variance is decreasing with m while the bias is increasing with m. Some authors use an interval value for m (Lobato and Savin, 1998).

³Using a shorter time horizon and higher frequency data, Barros and Gil-Alana (in press) provide similar results on the persistence of oil prices.

⁴Studies by Villar and Joutz (2006) and Ramberg and Parsons (2010) provide excellent reviews of the literature (see citations therein) on the long-run relationship between crude oil and natural gas prices.