Fig. 2: Heterogeneity assessment using Bayesian model averaging.

The columns in the figure depict the best 26,435 estimated meta-regressions, with each column showing the outcome of one estimated meta-regression model. The dependent variable for each of the meta-regression models is the percentage change in emissions. The possible explanatory variables are depicted in the rows (ordered by their PIP in descending order) and the explanatory variables included in a respective meta-regression model of the column is indicated by the colours. Red colour indicates the variable was included with a negative sign (larger emission reductions). Blue colour indicates a positive sign (smaller emission reductions). No colour indicates that the variable was not included in the meta-regression model represented by that column. The horizontal axis indicates the cumulative posterior model probabilities across all models. The models are ranked by their posterior model probability with the model on the left accounting for the largest posterior model probability. The definitions of the explanatory variables are provided in the Supplementary Information.