This paper reports on the results of a dynamic pricing experiment that compares the performance of three popular pricing programs–hourly pricing, critical peak pricing, and critical peak-pricing with a rebate–for a representative sample from the population of households in the District of Columbia. The sampled households differ in terms of their income levels, electricityusing appliance holdings and whether they own a smart thermostat. Using a nonparametric conditional mean estimation framework that allows for customer-specific fixed effects and hourof- sample fixed effects, I find that customers on all of the dynamic pricing programs substantially reduce their electricity consumption during high-priced periods. The hourly average treatment effects associated with each of these dynamic pricing plans are larger in absolute value for households with all-electric heating and households with smart thermostats. Low-income households have significantly larger hourly average treatment effects than higher income households on the same dynamic-pricing tariff. The results of these experiments are also used to investigate two hypotheses about differences in the customer-level demand response to the three dynamic pricing tariffs. Specifically, I find that for roughly the same marginal price during a critical peak period, critical peak pricing yields a larger hourly average demand reduction than critical peak pricing with a rebate. I also find that the demand reduction associated with higher hourly prices is very similar to the predicted demand reduction associated with the same price increase under critical peak pricing.
Final Report: An Experimental Comparison of Critical Peak and Hourly Pricing: The PowerCentsDC Program (1.72MB PDF)