An Examination of Energy Intensity in the U.S. Manufacturing Sector

 

 

Gavin Pickenpaugh

Economist

National Energy Technology Laboratory (Morgantown, WV)

and

Peter Balash

Senior Economist

National Energy Technology Laboratory (Pittsburgh, PA)

 

 

United States energy intensity, measured as primary energy consumption per dollar of real gross domestic product (GDP), fell approximately 45 percent from 1980 to 2009 (EIA, n.d.). Research concerning the driving forces of improvements in energy intensity dates back to the late 1970s, and a lot of the research attempts to tease out the contributions of efficiency and structural change (Myers and Nakamura, 1978).

This study implemented decomposition and regression analysis to examine energy intensity of the U.S. manufacturing sector over the 1998 to 2006 period, which declined almost 29 percent.  The first analysis section decomposed changes in energy intensity into efficiency changes and structural changes, while the second analysis section used regression analysis to examine the association between energy intensity with other variables, such as energy prices and employment.

The conclusions drawn from the decomposition analysis vary, depending on whether a national or regional scheme was used. The national analysis found that 82 percent of the decline in intensity from 1998 to 2006 was due to efficiency and 18 percent was due to structural change.  The regional analyses results were mixed, finding efficiency improvements as the main driver of the 45 percent decline in intensity in the West Region and the 31 percent decline in the South Region; conversely, structural change was found to be the main driver of the 18 percent intensity decline in the Midwest Region and the 32 percent intensity decline in the Northeast Region.  The structural change component’s dominance in two of the regions points to factors such as international competition from countries with competitive advantages, such as lower wage rates.

The findings of the regression models differ, depending on the type of model used and whether analysis was conducted at the national or regional level.  One general conclusion drawn from the regression analysis is that findings may change widely, based on factors such as model choice, aggregation level of the data, and variables and time periods used.

In the national analyses, the random-effects and between models estimate that industries with higher energy prices tend to have lower energy intensities; however, the fixed-effects model did not find a significant price effect.  This is consistent with the weak correlations between the first differences of energy price and intensity, contrary to the relatively large cross-sectional correlations between the levels of the two variables. In general, the fixed-effects models find most covariates to be insignificant at both the national and regional level, with the only consistent finding being that changes in average wage are significantly (negatively) associated with changes in energy intensity.  One plausible explanation for the lack of significance in the other covariates in the fixed-effects models is that all variables displayed small amount of variation within industries relative to the amount of variation between industries; since the fixed-effects model omits between-industry variation, its coefficient estimates tend to have higher standard errors.

The national random-effects and between models found the following (the term “associated” means a statistically significant relationship):

  • Higher participation rates in efficiency programs are associated with higher energy intensities. This may be due to a selection bias, in which industries where higher energy intensities are drawn to such programs more so than industries where energy intensity is relatively low.
  • Larger average floorspace is associated with lower energy intensities (significant only in between model).
  • Higher wages are associated with higher energy intensities, contrary to the fixed-effects model.
  • The random effects model estimates a significant negative time trend.
  • Higher employment levels are associated with lower energy intensities.

A significant negative price effect is estimated in the regional models in all of the between models, but only one of the random-effects models. This lack of statistical significance of the price variable in the random effects could potentially be due to factors such as omitted variable bias (not all variables in the national model were available at the regional model), missing data for certain regions/industries, and the fact that the analysis was only for the 1998, 2002, and 2006 periods, whereas regression analyses in the literature tended to cover a larger time period and show significant price effects on intensity.

In three of the four regions, the between and random-effects models each find that higher average wages are associated with higher intensities.  All of the random-effects regional models and two of the between-regional models find higher employment levels are associated with lower energy intensities. Additionally, each of the random-effects regional models found a statistically significant negative time trend for energy intensity.

An area of potential near-term future research is a similar analysis that appends 2010 data to the 1998, 2002, and 2006 data, once all of the 2010 MECS data are released.  This would provide a longer period of analysis from which to draw conclusions for both the decomposition and regression analyses.  Additionally, a parallel analysis examining electricity intensity is another possible future area of analysis.  A related area of potential future research dealing with energy efficiency is the rebound effect—the idea that improvements in efficiency may lead to increases in demand, which negate some of the energy savings (e.g., more fuel-efficient vehicles may encourage individuals to travel more miles).

 

References

Energy Information Administration (EIA). (n.d.). “International Energy Statistics. Energy Intensity - Total Primary Energy Consumption per Dollar of GDP (Btu per Year 2005 U.S. Dollars (Market Exchange Rates)).”

Myers, J., Nakamura, L. (1978). “Saving Energy in Manufacturing.” Cambridge, MA: Ballinger.

Wooldridge, J. (2002). “Econometric Analysis of Cross Section and Panel Data.” Cambridge, MA: The MIT Press.

Click to view a printable version of this article.