U.S. Energy Information Administration logo
Skip to sub-navigation
‹ Consumption & Efficiency

Commercial Buildings Energy Consumption Survey (CBECS)

Back to Methodology

How We Estimated Energy End-Use Consumption in the 2018 CBECS

Release date: April 12, 2023

The energy end-use consumption tables for the 2018 CBECS provide estimates of the amount of electricity, natural gas, fuel oil, and district heat used for 10 end-use categories:

  • Space heating
  • Cooling
  • Ventilation
  • Water heating
  • Lighting
  • Cooking
  • Refrigeration
  • Computing (including servers)
  • Office equipment
  • Other uses

Although details vary by energy source, the end-use estimation process has three basic steps:

  • Creation of engineering models for each end use
  • Imputation of total energy consumption for cases with missing data
  • Calibration of the end-use estimates to the CBECS total building energy consumption

Creation of engineering end-use models

The end-use estimation procedure begins with a group of engineering end-use models. In general, these models estimate the amount of energy used for each end use in a building based on the size of the building, the type of building, the hours of operation, the expected efficiency of energy using equipment, and (where relevant) the climate of the building’s location. The parameters for equipment efficiency and the expected demand for some end uses came from the American Society of Heating, Refrigerating, and Air-Conditioning Engineers (ASHRAE), Illuminating Engineering Society of North America (IESNA), and other standard engineering handbooks. In addition, some information about the expected demand for energy end uses came from large-scale field studies of commercial buildings such as the California Commercial End-Use Survey (CEUS).

Space heating and cooling

The heating and cooling models begin by estimating the total energy required or expected for heating and cooling in the building. The models account for building heat loss (or gain) as a function of the annual heating (or cooling) degree days in the building’s location, the building’s dimensions, and the average amount of heat conducted into or out of the building, based on the thermal properties of the roof and wall materials. In addition, the models account for ventilation heat loss (or gain) as a function of the volume of external air brought into a building each day, the temperature difference between the outside air and the inside air, and (for cooling) the heat capacity of the air.

After the total energy required for heating and cooling has been estimated, the amount of each specific fuel used for heating and cooling is modeled using CBECS Buildings Survey data for estimated percentage of floorspace heated or cooled by each fuel and equipment type and engineering values for average efficiency of each equipment type. The resulting model predicts the values of electricity, natural gas, fuel oil, and district heat consumption (in British thermal units [Btu]) for each building’s space heating. It also predicts the values of electricity consumption (in Btu) for each building’s space cooling.


The engineering model for ventilation estimates supply and return fan energy use by estimating the volume of air used for ventilating each building per square foot per year. To estimate total ventilation air volume, the model accounts for different ventilation requirements based on building use, operating hours, months of operation, building floorspace, and system type. Typical meteorological year data help develop estimates of variable air-volume energy factors for areas with similar climates. The resulting model predicts the values of electricity consumption (in Btu) for each building’s ventilation.

Water heating

The water heating model uses building activity, months in use, and size measures from the CBECS Buildings Survey to estimate the amount of hot water required in the building. The energy needed to provide this hot water depends on the ground water temperature, the average efficiency of the water heating equipment types, and estimated distribution losses (if any). Additional energy use is estimated in systems with booster water heaters that operate at high temperatures. The resulting model predicts the values of electricity, natural gas, fuel oil, and district heat consumption (in Btu) for each building’s water heating.


The lighting model estimates electricity consumption from lighting for all building types. The model calculates expected lighting use as a function of building use, operating hours, and building floorspace. Information from the CBECS Buildings Survey on percentages of floorspace lit is used to determine the total amount of lighting provided by each lamp type. The model assumes average lamp system efficacy (the ratio of the light output from a light source to the power consumed, measured in lumens per watt) for each lamp type. It also assumes recommended average illuminance levels (the total number of lumens divided by area) by building type. The resulting model predicts the values of electricity consumption (in Btu) for each building’s lighting.


The cooking model estimates cooking energy consumption (in Btu) for natural gas and electricity. Commercial kitchens can be quite varied in their set-ups, and limited information is collected by CBECS on the types and number of cooking equipment used by commercial buildings. The model combines estimated conditional intensities from the CEUS along with CBECS information on floorspace and CBECS Buildings Survey responses about cooking fuel use. Because we did not have the conditional intensities for fuel oil and district heat cooking, we estimated them as a proportion of the electricity and natural gas total.


The refrigeration model calculates commercial refrigeration electricity consumption (in Btu). The CBECS Buildings Survey collects information on the number, but not the size, of refrigerators used by commercial buildings. The model incorporates the CBECS information on the number of refrigerators and building type, and it uses average consumption per unit from engineering handbooks to estimate an annual total. For some building types with a high number of refrigerators of variable size, such as food sales buildings, the model relies predominantly on end-use intensity estimates from the CEUS, per square foot, by building type, to estimate refrigeration consumption within each building.

Office equipment and computing

The office equipment model estimates electricity consumption (in Btu) from office equipment for all building types based on equipment information collected by the CBECS Buildings Survey and average equipment efficiency. The model divides office equipment electricity consumption into two components: computer equipment and other office electric loads. Computer equipment includes PCs, laptops, tablets, monitors, servers, and data centers. The non-computer-based equipment includes multifunction office devices such as printers, copiers, scanners, and fax machines. It also includes cash registers, interactive white boards, and video displays.

Other uses

We estimate consumption for other uses separately for electricity, natural gas, fuel oil, and district heat. These fuels are all measured in Btu. The model for miscellaneous uses for electricity relies on end-use estimates from the previous CBECS survey. Many types of equipment that use electricity are not included in the previously listed end uses, such as medical equipment, air compressors, and other miscellaneous plug loads. CBECS does not explicitly ask for all of the possible other uses of electricity. Therefore, the engineering model estimates other electricity use by applying the average intensities for miscellaneous electricity use to the CBECS floorspace, with some additional adjustments for reported medical equipment. These estimates were then adjusted for the number of months of building operation per year. For natural gas, fuel oil, and district heat, the models for other energy use are based on building characteristics information collected by CBECS and average intensities, adjusted for the number of potential other uses, such as natural gas dryers, electricity generation, and manufacturing.

The sum of the engineering models produces a benchmark that is used during consumption data editing, as described in How We Reviewed Data to Ensure Quality of the 2018 CBECS.

The engineering end-use estimates calculated above were based on building characteristics from the questionnaire but not on any reported energy consumption data.

Imputation of total energy consumption

A key metric that CBECS provides is total consumption for each building for each energy source used in the building. In many cases, we obtain this consumption from either the Buildings Survey or the Energy Supplier Survey. In other cases, where we do not have a value for total consumption for the building, it must be imputed. To perform this imputation, we use a cross-sectional regression model to impute the total consumption of each energy source that is missing for a building. The cross-sectional regression models were fit with consumption as the dependent variable and the engineering estimates for each use as the independent variables. In addition, the independent variables could include building characteristics variables that might affect total consumption such as the presence of energy-intensive equipment or processes, conservation measures, and unusually high or low activity in the building.

Calibration of end-use consumption

Because the individual engineering end-use estimates are modelled without reference to the total building-level energy consumption, they will not sum to the reported or imputed value for total consumption. To ensure the final end-use estimates add up to the total, the preliminary estimates must be adjusted up or down by an appropriate percentage. This percentage will vary from one end use to another based on each estimate’s uncertainty and correlation with other end uses. The expected uncertainty of each end-use estimate is based on the amount of available information that was used to model the estimate. For example, the uncertainty of heating and cooling is expected to be low because the model considers the dimensions of the building, the annual degree days, the efficiency of the heating and cooling equipment, and the heat gain or loss associated with the building construction materials. The cooking estimate is more uncertain because it only considers building size and use. The adjustments are optimized such that the estimates with a lower expected degree of uncertainty are adjusted by a lower percentage than estimates with a high degree of uncertainty.

Changes from 2012 CBECS end-use methodology

The 2018 end-use estimation procedures were similar to those used for the 2012 CBECS, but we made a number of changes to the process between the two surveys.

  • New information from the 2018 Buildings Survey questionnaire was incorporated into the models. In particular, we added detailed questions about the fuel used by each type of heating and cooling equipment present in the building.
  • Where possible, we updated efficiency and intensity factors to reflect how the stock of equipment in commercial buildings is newer on average in 2018.
  • In 2018, the calibrated final end uses were more directly based on the engineering estimates. The calibration methodology was different in 2012, with more than one step between the initial engineering end-use estimates and the final estimates.
    • In 2012, calibration was integrated in the cross-sectional regression models used to impute missing values for total building-level energy consumption. Total consumption was modeled as the sum of a number of sub-models, each of which had an engineering end use as its primary component. These sub-models could include other building characteristics, along with equation parameters estimated by the model. The sum of these sub-models provided an estimate of total consumption, and this estimate was used as an imputed value where needed. In 2018, the regression model used the engineering end-use models to impute missing consumption, but the regression model had no influence on the final end-use estimates.
    • The 2012 sub-models were considered to be intermediate end-use estimates. To ensure that the final end-use estimates added up to the total consumption, each building’s end uses were fully calibrated by multiplying each use’s proportion of the sum of the sub-modelled estimates (in other words, the estimate of total consumption as produced by the regression equation) by total building-level consumption. Unlike 2018, for a single fuel within a building, the same proportion was applied to all intermediate end-use estimates.

We provide further details on the 2012 end-use estimation methodology on our website.