Fig. 1: Average emissions changes by scheme.

Panel (a, d, g) present weighted mean effect sizes together with their 95% confidence intervals based on multilevel random and mixed effects models and represent the effect of the policy observed in each period after its introduction in comparison to the counterfactual emissions without the policy. The estimates are ordered according to the number of studies they comprise (depicted on the left). The average treatment effect for the Chinese ETS pilots comprises the effects of all eight regional pilot schemes. Cross-country collects the evidence from studies assessing countries with and without carbon pricing, not focusing on a specific carbon pricing scheme. Panel (a, d, g) comprise, respectively n = 470, n = 253, and n = 142 effect sizes clustered on the study level. Panels (b, e, h) show the distribution of assigned risks of bias (RoB). Panel (c, f, i) show the distribution of statistical power. Power above 80% is considered adequate. For synthetic control designs no statistical power was derived, thus presented as “NA".