DISCLAIMER This report was prepared as a n account of work sponsored by a n agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, make any warranty, express or implied, or assumes any legal liability or mponn'bility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any

more »

... c commercial product, process, or sem'ce by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necesSarily state or reflect those of the United States Government or any agency thereof. EXECUTIVE SUMMARY This report supplements Lee et al. (1995) , which presents the findings of an impact evaluation of the Manufactured Housing Acquisition Program (MAP). Pacific Northwest Laboratory conducted the evaluation and prepared both reports. This report presents detailed technical information relevant to the MAP impact evaluation. It ,is intended to provide the interested reader with enough information to answer most technical questions about the analysis and results. TIERED ANALYSIS APPROACH We used a three-tiered process to analyze the energy consumption of both MAP and baseline homes. The information from each analysis was useful for designing our final analysis. \ The first approach was a comparison of annual billing data, followed'by a simplified regression analysis to adjust for major home characteristics. We compared the mean annual kWh consumption of MAP and baseline homes, and used the'difference to estimate energy savings. We made no adjustments for long-term weather. This first analysis showed that MAP homes consumed less electricity than the baseline homes used in our analysis, but the differences were'less than the pre-program estimates suggested. We identified several factors to examine further. First, nonelectric supplemental heating was more common in baseline homes than in MAP homes. Second, in some cases heat pumps were more common in baseline homes than MAP homes and they tended to reduce energy consumption. Third, the average baseline home in our sample was smaller than the average MAP home, thus reducing the difference between electricity use in MAP and baseline homes. Fourth, we found that the distribution of total electricity consumption and consumption per square foot in MAP homes exhibited less variance than in, baseline homes. The second-tier approach was an application of the PRlnceton Scorekeeping Method (PRISM). This methodology uses monthly billing data to estimate coefficients that can be used to predict the non-temperature-and temperature-sensitive portions of energy consumption. We used PRISM to estimate electricity consumption for a "normal" weather year. iii We used PRISM to analyze several samples. We applied it to billing data for the entire sample of homes, all baseline homes only, and all MAP homes.only. We then screened the sample to eliminate observations that could not be modeled well by PRISM, and repeated the analyses. The results showed that the standard errors of the savings estimates declined, in most cases, after the billing data were screened. Screening the data, however, considerably reduced the sample sizes and this tended to diminish the accuracy and precision of all estimates. The PRISM approach, however, could not be applied effectively to produce the energy savings estimates needed in this analysis. This was because of the confounding effects of non-electric heat, differences in the efficiency baselines of interest, and other factors not addressed by the PRISM approach. REGRESSION MODEL We used a detailed'regression analysis to control for a wide range of possible energy consumption determinants such occupant demographics, appliance inventories, and weather. This allowed us to estimate energy savings attributable to the MAP features under different conditions. This approach was applied to all sample homes'for which we obtained billing data. Billing-period (usually monthly) data were used. Our model used appliance inventories to explain total kWh consumption like a conditional demand analysis (CDA), but was formulated around the anticipated thermal-physical relationships. It included several appliances and the effects of demographic and behavioral variables that were found to be significant. Because differences in heating performance were the primary anticipated effect of MAP, we focused on coefficients for different types of heating systems and combinations of systems in formulating the model. Several potential limitations of the model are discussed. These include potential , difficulties modeling the space-heating response to temperature and the effect of ventilation. Statistical and econometric,details of the model are presented. The technique for developing confidence intervals with this model is also discussed. iv --.