| The research reported on in this dissertation investigated methods for using an engineering model of energy use to produce prior estimates to help explain the variation in metered natural gas usage. The econometric method for estimating the relationship between energy use and economic, demographic, and building characteristics used a conditional demand analysis (CDA) framework: an approach in which the short term variation in metered energy use is assumed to be conditional on the presence of appliances. This dissertation illustrates that a model that introduces components of an engineering model directly into a conditional demand analysis framework results in estimators that are more efficient, less biased, and easier to interpret than methods that rely on either engineering models or traditional conditional demand analysis models alone.;The analysis was based on data collected from some 1,000 on-site home energy audits conducted in Oregon and Washington in late 1991 and early 1992. This information, along with local weather data, was used to help explain variations in metered natural gas use during 1991. Engineering estimates of energy use were made using a commercial version of the Computerized Instrumented Residential Audit (CIRA) model developed by Lawrence Berkeley Laboratory.;A new method was developed which required several engineering estimates of energy use for each household at different levels of insulation up to and including the actual levels. This improved the model results by allowing the model to isolate the tendency of the engineering model to substantially overestimate the impacts of conservation measures. The resulting model was shown to have improved predictive ability and to provide added value as a policy tool for determining the realized energy savings from conservation measures. |
