|
Abstract:
|
The U.S. Energy Information Administration (EIA) publishes both monthly short-term and annual long-term (to 2050) projections of energy production, consumption, prices, and related statistics. For consistency and explainability, EIA calibrates the first one to two years of the long-term projected series to agree with the short-term projections. Because the historical data series for many energy-related statistics are volatile, the short-term projections often change substantially from year to year. The adjustment factors used to calibrate the long-term series are therefore sensitive to the choice of calibration year. We examine methods of calibrating projected series to the overall trend of the historical and/or short-term projected series, estimated by Holt-Winters smoothing with a trend coefficient. The methods, which have been implemented in the R software, allow users to customize the calibration process for particular series through a set of user-selected parameter values. The parameters include the Holt-Winters smoothing and trend parameters, a maximum year-to-year change tolerance factor, and maximum and minimum values for the calibrated series.
|
