2017/10/06 by Datong P. Zhou, Zhou, Datong P., Maximilian Balandat +3
Economics, Econometrics and Finance · Engineering · #FOS: Electrical engineering #FOS: Physical sciences #Housing Market and Economics #Physics and Society (physics.soc-ph) #Smart Grid Energy Management #Systems and Control (eess.SY) #Water resources management and optimization #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.1710.03190
openalex publication_date 2017/10/06 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
We evaluate the causal effect of hour-ahead price interventions on the\nreduction in residential electricity consumption using a data set from a\nlarge-scale experiment on 7,000 households in California. By estimating\nuser-level counterfactuals using time-series prediction, we estimate an average\ntreatment effect of ~0.10 kWh (11%) per intervention and household. Next, we\nleverage causal decision trees to detect treatment effect heterogeneity across\nusers by incorporating census data. These decision trees depart from\nclassification and regression trees, as we intend to estimate a causal effect\nbetween treated and control units rather than perform outcome regression. We\ncompare the performance of causal decision trees with a simpler, yet more\ninaccurate k-means clustering approach that naively detects heterogeneity in\nthe feature space, confirming the superiority of causal decision trees. Lastly,\nwe comment on how our methods to detect heterogeneity can be used for targeting\nhouseholds to improve cost efficiency.\n