One of the major motives for an electric utility to implement a restructuring of rates is to level out daily load profiles. In undertaking such reform, a utility does not want to cause any deterioration of its customers' load factors.' Significant deterioration of customer load factors at the local distribution level may provide signals for expanded investment in distribution facilities. Thus, before implementing such a program, forecasts of changes in load factors will often be required to aid in future distribution and system planning, and rate setting. Having no experience with this policy, where does a utility acquire such information? One approach is to conduct extensive experiments with various price ratios but because of the associated expense, experimentation has generally been confined to smaller customers (e.g., residential [2; 3; 8]). A second approach is to simulate responses with the aid of process or linear programming models [11; 21; 15]. A third approach is to make use of empirical estimates of responses observed in other jurisdictions [1; 10; 13; 20; 23]. The second and third approaches do offer much potential and should be fully explored. A fourth approach is to make use of historical data generated under uniform prices along with some basic principles which characterize neoclassical production theory [19; 24]. It is this latter approach which this paper adopts. The major focus of this paper is to forecast the change in load factor caused by moving to time-of-use rates. This goal is not as ambitious as that pursued by Panzar and Willig [19] and Tishler [24], which is to identify changes between work-shifts. Thus, although the lack of historical time-of-use data and this paper's modeling structure do not permit the identification of such movements of electricity between shifts, one is able to project load factors changes with less restrictive assumptions and less data. Essentially, as outlined in section II, the assumptions of separability of electricity, Shephard's lemma, and Young's theorem enable one to take advantage of available historical data on electricity usage by shift for forecasting impacts on load factors. Based on up to 15 years of experience with the Hopkinson rate structure, the most appropriate dual unit cost function is