Controlling for Nonprogram Effects in a Statistical Engineering Analysis

This study examined the extent to which engineering estimates of kilowatt-hour (kWh) savings from National Grid USA’s Small C&I Program in 2001 are realized in customer billing data. Separate realization rates were estimated by business type to assist in future program design. The analysis focused on lighting measures. Studies of this type are commonly referred to as “realization rate” studies. Measured changes in energy usage from customer billing data are commonly compared to preliminary savings estimates to develop realization rates for energy conservation programs. An important part of these analyses is controlling for nonprogram-induced changes in energy consumption due to factors such as weather, business fluctuations, or changes in customers’ building characteristics. However, limited customer-level data are typically available for program participants’ (and nonparticipants’) behavioral or building changes that affect changes in kWh usage over the window of analysis. This paper presents a set of innovative statistical techniques used to leverage survey information from a sample of participants and nonparticipants to control for nonprogram usage changes in the analysis population as a whole.