Publication Type:Conference Proceedings
Source:2016 ACEEE Summer Study on Energy Efficiency in Buildings (2016)
Randomized controlled trials (RCTs) are widely viewed as the “gold standard” of evaluation. However, analysis of the effect of energy pricing has largely been conducted through quasi-experimental methods. Using a rare set of large-scale randomized field experiments of time-based electricity pricing, we compare the estimates obtained from commonly used non-experimental methods against RCT estimates. We demonstrate empirical evidence in favor of four stylized facts that highlight the importance of understanding two important sources of bias in this context: selection bias and spillover effects. First, difference-in-difference and propensity score methods tend to underestimate the true average treatment effect. Second, regression discontinuity methods tend to overestimate the effect. Third, selection biases in quasi-experimental methods tend to be more pronounced in opt-in treatments relative to opt-out treatments. Fourth, the three-in-five day baseline with an additive adjustment recomm ended by KEMA (2011) tends to underestimate the impact of the intervention, a pattern we attribute to intertemporal spillover effects.