Derived Annual Estimates of Manufacturing Energy Consumption, 1974-1988

The actual data on purchased energy comes from two sources, the U.S. Department of Commerce Bureau of the Census's Annual Survey of Manufactures (ASM) and EIA's Manufacturing Energy Consumption Survey (MECS). The ASM provides annual estimates for the years 1974 to 1981. However, in 1982 (and subsequent years) the scope of the ASM energy data was reduced to collect only electricity consumption and expenditures and total expenditures for other purchased energy. In 1985, EIA initiated the triennial MECS collecting complete energy data. The series equivalent to the ASM is referred to in the MECS as "offsite-produced fuels." The completed annual series for 1974 to 1988 developed in this report links the ASM and MECS "offsite" series, estimating for the missing years. Estimates are provided for the manufacturing sector as a whole and at the two-digit Standard Industrial Classification (SIC) level for total energy consumption and for the consumption of individual fuels. There are no direct sources of data for the missing years (1982-1984 and 1986-1987). To derive consumption estimates, a comparison was made between the ASM, MECS, and other economic series to see whether there were any good predictors for the missing data. Various estimation schemes were analyzed to fill in the gaps in data after 1981 by trying to match known data for the 1974 to 1981 period.

The most accurate methodology for deriving estimates for the missing years turned out to be a relatively simple process. The chosen procedure first estimates the total consumption of an industry for years for which there are no data available. The method of estimation uses available consumption data from either the MECS or the ASM for endpoint years that bracket years with missing data. The relationship between the Federal Reserve Board (FRB) production index (published in the Statistical Abstract of the United States ) for the endpoint years and the intervening years is applied to the known consumption totals for the endpoint years to derive consumption totals for the missing years. Fuel-specific estimates are obtained by using a linear interpolation on the fuel shares for the endpoint years to determine shares for the intervening years, and applying these shares to the newly-derived consumption totals for the intervening years.

• FRB indices of production are generally good indicators of changes in total energy consumption for the manufacturing sector as a whole and by individual industries (two-digit SIC's). Figure ES1 displays the ratios of total consumption and the FRB manufacturing production index relative to 1974 consumption and production, respectively. The year-to-year changes in production and consumption are very similar, although the trends are different.

• Fuel shares for each industry did not change very much from year to year, so that intervening years' shares could be interpolated linearly. This means that given an estimate of energy consumption by the industry, individual fuels can be obtained fairly accurately by allocating the total.

• Several methods of allocating the consumption among individual fuels were tested. The alternate methods attempted to take into consideration the possibility of fuel-switching, either among all fuels or among gas and oil alone. However, problems were encountered because tests showed that energy prices were not the only determining factor in whether or not fuel switching occurred; and energy prices were not known accurately enough at the level of an individual industry to make the SIC-level estimates reliable.

Analytic Applications of the Derived Series

While this report is designed to make existing information more useful for tabulations and graphical displays, it provides no new analytic information relative to the original ASM/MECS data series. In fact, analysts who attempt to derive further trend and regression relationships from the completed series will discover the estimation technique reinforces some of the trends in the original data.

For any individual year, the derived estimates of natural gas, coal, and total consumption appear to be nearly as reliable as the corresponding published MECS estimates. However, the derived estimates for other fuels, including distillate and residual fuel oil, are much less reliable. The results highlight the potential for bias due to the derivation methodology. This bias appears to dominate the estimates of potential error on oil products in a few volatile industry groups. Users are cautioned to take these potential errors into account before using the derived estimates for analysis.

File Last Modified:  08/14/98

Contact:

robert.adler@eia.doe.gov

Robert Adler

Project Manager

Phone: (202) 586-1134

Fax: (202) 586-0018