Forecasting Natural Gas Prices

6624 Words27 Pages
Forecasting natural gas prices using cointegration technique Dr Salman Saif Ghouri Abstract This paper uses Augmented Dickey-Fuller and Phillips-Perron technique for determining whether individual crude oil prices (West Texas Intermediate, Brent, Japan crude cocktail) and natural gas prices- Henry Hub (HH), National Balancing Point (NBP), European and Japanese liquefied natural gas (LNG) prices are stationary or non-stationary. It then applies Johansen and Juselius cointegration technique for establishing long-run correlation between respective oil prices and natural gas prices. The paper concludes that all individual series pertaining to oil and natural gas prices are non-stationary and indeed having long-run relationship, despite…show more content…
Both buyers and sellers have a reasonable assurance that prices marked to HH or NBP are competitive. In sharp contrast, Table 1 Regional pipeline natural gas trade (bn cu m) Regions North America Europe Algeria Russian Federation Others 2004 121.78 154.17 34.32 148.44 43.35 December 2006 2006 Organization of the Petroleum Exporting Countries 251 Table 2 LNG trade movement, 2004 (bn cu m) Buyers/sellers Asia Pacific Africa Middle East USA Trinidad & Tobago Total Asia Pacific 81.67 0.78 34.47 1.68 0 118.6 USA 0.99 3.74 0.61 13.13 18.47 40.02 Europe 0.18 34.45 5.39 0.86 0.86 C&S America Total 82.84 38.97 40.47 1.68 13.99 177.95 in a regulated market, there is no competition or transparency, this leads to different prices in the same market. This paper examines the long-run relationship between natural gas/LNG prices in different markets with those of respective crude oil prices using cointegration techniques (Cuthbertson, Hall, 1992). The basic philosophy of cointegration technique is that two or more variables may have long-run relationship if they move closely together in the long run, even though they may have drifted apart in the short-run. This long run association is referred to as a cointegrating vector. If there is a long run correlation between the variables, a regression containing all the variables of a cointegrating vector will have a stationary error term, even if none of the variables taken alone is stationary. Stock (1987) has
Open Document