Digging Into Uncertainty in Energy Consumption Data
Energy consumption data are central to a range of economic, infrastructure planning, and environmental analyses. Uncertainty in energy consumption data, however, has rarely been assessed, with uncertainty estimates often based on expert judgment. Researchers at the U.S. Department of Energy’s Pacific Northwest National Laboratory developed a novel method to examine the accuracy of energy consumption data. Analyzing 17 consecutive editions of a widely used global data set, they quantified how fossil energy consumption estimates changed over time as countries updated previously reported data. The size and frequency of revisions varied by country, with most energy consumption trends subsequently revised. Among fuels, coal data points had larger, less frequent revisions, oil data points had smaller, more frequent revisions, and natural gas was in between.
For the first time, researchers conducted a global quantitative analysis of one indicator of uncertainty in energy consumption data: how reported fossil fuel consumption changed over time as previous data were revised. This analysis provides a quantitative basis for informing estimates of uncertainty in energy data, useful for numerous planning and analysis activities.
Uncertainty should be a central part of any scientific analysis; however, uncertainty in energy consumption data is difficult to estimate. In part, this is due to the nature of socioeconomic data sets. Unlike a laboratory measurement, these data are not estimated through repeated observations, which means that conventional uncertainty techniques often do not apply. One indicator of the uncertainty in these data is the extent to which past estimates—for example, natural gas consumption for the United States in 2008—are subsequently revised in later years.
Researchers studied 17 editions of the BP Statistical Review of World Energy (2001−2017) to examine how reported energy consumption was revised over time in aggregate coal, oil, and natural gas consumption data. In the year after initial publication, they found that 70 percent of fossil energy consumption data were revised by an average of 1.3 percent of a country’s total fossil fuel consumption. The rate of data revisions drops off with time, with revisions in 44 percent of historical trends, defined as any data at least five years old. While most revisions were minor adjustments, the average maximum adjustment was 2.7 percent of fossil fuel use. (Note that, when expressed as percentage changes in a specific fuel, rather than as changes in total fossil fuel use, the numerical changes are larger.)
Across countries, coal data showed larger adjustments, with the 95th percentile of all adjustments resulting in a nearly 20 percent change in coal consumption (for countries in which coal was at least 5 percent of the total fossil energy supply). Researchers found no evidence that confidence in energy consumption data increased over time. The scientists noted that they quantified the energy consumption data changes that were found and corrected. Uncertainties that had not been found or corrected would also be present in the data.