Forecasting Performance of Fundamental Natural Gas Price Models, Hedging Strategies, and the Average Cost of Gas: a Study of the U.s. Natural Gas Market

We propose and estimate fundamental models for natural gas prices. We compare how well these models, as well as univariate statistical time series models of NG prices and the NYMEX futures price for natural gas, forecast spot gas prices. We find that a univariate time series model that incorporates fundamental variables related to production, storage, weather, and aggregate output performs best when NG prices are falling. In contrast, when prices are rising, a VAR specification with multiple fundamental endogenous and exogenous variables gives the best predictions for time horizons of either 6, 9, or 12 months, while the futures price gives the best predictions for a 3-month horizon. We also examine the average gas cost to a user who implemented either an Always Hedge, Never Hedge, or a Mixed Strategy based upon the price forecast. The Mixed Strategy utilizes the spot price forecasts based upon the alternative forecasting models. We find that during the falling price phase, but irrespective of the forecast model utilized, the Mixed Strategy always produces an average cost that is less than or equal to the strategy of Always Hedging. The same is true during a rising price phase for the 9and 12-month time horizons, but during the 3and 6month time horizons the policy of always hedging dominates. However, we also find that during a falling price phase the absolute least cost strategy is to never hedge and that this is also the dominating strategy during the rising price phase for 9and 12-month out horizons.