problems. As a result, the CBECS sample design has undergone numerous revisions over the years, as EIA has attempted to address both the cost and the coverage issues. Generally speaking, most rounds of the CBECS have relied on a combination of an area frame and a list frame, merging existing lists of commercial buildings from a variety of sources and adding them at the second stage of the area frame sample.

The area frame is based on field listings of commercial buildings within specified geographic areas. Over the years, the sample typically included between 100 and 130 counties or groups of counties, which served as primary sampling units (PSUs). Within the PSUs smaller geographic areas were randomly selected, and then all commercial buildings within these areas were listed and stratified. In the final step a sample of buildings was randomly selected from each stratum. The sampling frame is updated from one data collection to the next in order to account for changes, such as new construction. The updates are typically based on information from local sources or databases of commercial projects, and new construction is usually sampled at a higher rate to enable EIA to produce estimates for this subpopulation.

Given the variations in building size and in the intensity of energy use among commercial buildings, the sampling rates for large buildings that use large amounts of energy must be higher than would be feasible to accomplish through the area-probability sampling alone. To assure that these types of buildings are adequately represented in the sample, the area frame is supplemented with a list frame of special buildings, such as very large buildings, hospitals, and government buildings. The two frames are then matched, and duplicates are eliminated.

Over the years, EIA has considered a variety of lists to use as the primary source for the sampling frame, including tax records, mail delivery points, insurance lists, customer information from utilities, Federal Emergency Management Agency records, and aerial and satellite photographs (French, 2007). Some of these alternatives have been evaluated empirically, while others have not. One feasibility study, which was conducted in 1981 in the service areas of Seattle Power and Light Company and Portland Electric, was performed to evaluate the use of electric utility customer information in building a sampling frame. The study concluded that differences between the way that records are kept at different utilities and the utilities’ varying degree of motivation to respond to an energy survey would represent significant challenges (French, 2007).

For the last round of the CBECS in 2007, EIA’s data collection contractor, the National Opinion Research Center (NORC), proposed a new



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement