Estimating Treatment Effects of a Residential Demand Response Program Using Non-experimental Data

Residential Demand Response has emerged as an instrument of the modern smart grid to alleviate supply and demand imbalances of electricity. Utilizing their flexibility of electricity demand, residential households are offered monetary incentives to temporarily reduce energy consumption during times when the grid is strained due to a supply shortage. In this paper, we estimate the magnitude of reductions of 1,025 residential households in California, serviced by the three main utilities PG&E, SDG&E, and SCE, using time-series regression on hourly smart meter data. By predicting the counterfactual consumption during Demand Response periods, which is the hypothetical consumption had there been no intervention, we find a user-averaged reduction of 0.12 kWh (8.6% of the mean consumption) per hour. The reduction is heterogeneous on a temporal and geographic level, as we find the largest reductions to occur in the early afternoon and early evening, as well as a positive correlation with ambient air temperature. Further, our findings suggest that households with automated smart-home devices, which can be automatically shut off during Demand Response periods, show a larger percentage reduction of the mean (12%) compared to households without (8.3%).