Information campaigns for residential energy conservation

This paper evaluates an intervention that randomized information letters about energy efficient investments and behaviors among 120,000 customers of two utilities in Germany. We find that conservation effects differ considerably between both utilities, ranging from a precisely estimated zero effect to 1.4%. By contrast, we do not detect significant framing effects from presenting savings in monetary or ecological terms. Based on random causal forest methods, we show that the effect heterogeneity across utilities cannot be explained by sociodemographic characteristics. Our results demonstrate the importance of site-specific factors for the effectiveness of information campaigns, which has crucial implications for targeting and the ability to infer population-wide effect sizes from pilot studies.