Abstract:
Time-varying prices are thought to be critical for increasing economic efficiency of the power system. However, a rigorous assessment of the evidence from field trials that use time-of-use pricing, critical peak pricing, etc. in households is missing. This machine learning-assisted systematic review compiles the largest dataset till date of results from pricing pilots reported in both academic publications and electricity utility reports. This unique dataset enables us to deduce the presence of publication bias in peer-reviewed publications. Employing a multilevel meta-analysis, we estimate an average reduction of 8.7%-10.6% in peak consumption, 1.2%-1.5% in total consumption and no change in off-peak consumption across trials. Our heterogeneity analysis, using Bayesian Model Averaging, finds that a 10% increase in the peak-to-baseline price ratio is associated with a 0.47% reduction in peak consumption with marginal reduction in effects suggesting “scope effect” in household behavior. Overall consumption is not responsive to price ratio. Dynamic pricing thus seems to be effective in managing electricity demand but with limits.
