Impact measurement of tariff changes when experimentation is not an option-A case study of Ontario, Canada

This paper presents the first year results of a three-year study that was designed to measure the impact of time-of-use rates in the Canadian province of Ontario. Unlike prior studies, which have analyzed such impacts in an experimental setting featuring the randomized allocation of customers to treatment and control groups, this study deals with a situation where the deployment of time-of-use rates was not experimental in nature. It was part of a full-scale program. By carefully drawing samples across four local distribution companies, we were able to interpret the study as a natural experiment and construct control groups and treatment groups. We used the Addilog model of consumer demand to estimate elasticities of substitution between peak, mid-peak and off-peak periods. The Addilog model includes as a special case the constant elasticity of substitution model that has been widely used in the literature on customer response to time-of-use rates. The estimated elasticities are well behaved and consistent with those found in the literature.