The US Department of Energy’s Energy Information Administration (EIA) provides public forecasts of energy prices through its Short-Term Energy Outlook. The importance of accurate public outlook information is apparent in the energy sector, especially considering the systematic impact of the energy sector on the entire economy.

New research by Dwight R. Sanders et al. from Southern Illinois University and other institutions in the US assesses the performance of energy price forecasts at longer forecast horizons. The group examined both the rationality and information content of the EIA’s multiple-horizon forecasts for several important energy commodities.

Price forecasts for crude oil, retail gasoline, retail diesel fuel, natural gas, coal and electricity were tested for unique information at the one-, two-, three- and four-quarter horizons. In particular, the Vuchelen & Gutierrez direct test considers the marginal information content of the k ahead price forecast relative to the k–1 ahead forecast.

For instance, if the two-quarter ahead forecast is shown to provide no incremental information relative to the one-quarter ahead forecast, then only the one-quarter forecast need be used. Paralleling the work of Manfredo & Sanders, the work also determines if a composite forecast using the k and k-1 ahead forecasts will yield a more accurate forecast.

Generally, the results show that the EIA’s energy price forecasts do contain unique information at multiple horizons. Specifically, the forecasts for petroleum-based product prices (crude oil, gasoline, diesel) contain unique information out to the three-quarter horizon. The price forecasts for natural gas and electricity are informative at all horizons out to four quarters.

In stark contrast, coal price forecasts contain some unique information at the one-quarter horizon, but beyond this horizon they provide no unique information to the forecast user.

Across the energy commodities and forecast horizons examined, there are some cases where the energy price forecasts are not rational, and there are a number of cases where a composite forecast of the k and k-1 ahead forecasts are actually preferred. In these cases, forecast users should make the appropriate adjustments or create the recommended composite forecasts.

On average, however, the EIA appears to be doing an admirable job at providing multiple-horizon price forecasts which contain useful information. The only major deficiency is in coal, where the DOE may want to re-examine its forecasting procedures.

Energy Economics, Volume 31, Issue 2, March 2009, Pages 189–196.
DOI: 10.1016/j.eneco.2008.08.010