Natural Gas Price Forecasting in a Changing World

Abstract

Natural gas price forecasting is critical to many macroeconomic projection models and plays a vital role in state budget planning in energy-producing states. We examine the performances of various individual and composite forecasting models when predicting natural gas prices in the United States. We find the forecast generated by the Energy Information Administration Short-Term Energy Outlook provides a more accurate price prediction at longer forecasting horizons while futures-based forecasts perform better in the short run. Projections based on time-series models perform well at longer forecast horizons when price volatility is relatively low. Further, the Hotelling model performs well for 1- and 3-month ahead forecast horizons. Our findings further support the additional benefit of composite forecasts based on individual methods for more accurate predictions; however, the performance is not uniform at different forecasting horizons.

Keywords: natural gas, price forecast, vector autoregressions, futures prices, forecast evaluation, composite forecast