Table 1.
Established explanations for public support and political mobilization.
Table 2.
Dependent variable question wording and answer categories.
The survey data and question wording was translated from a conjoint survey experiment in Switzerland [95].
Table 3.
Conjoint attributes and randomized attribute levels and their relation to the proposed CO2 act on which the public voted.
The survey data and question wording were translated from a conjoint survey experiment in Switzerland [95].
Fig 1.
Model performance using ROC area under the curve for public support and political mobilization using experimental and non-experimental data in forest and XGBoost machine-learning models with and without PCA dimensionality reduction.
Dots with error bars represent out-of-fold performance on the training data, and triangles represent out-of-sample performance.
Fig 2.
Permutation-based variable importance scores using the tuned random forest model with PCA.
The box and violin plots display the distribution of cross-entropy loss for a given variable for 50 permuted samples from the observed data to account for uncertainty associated with predictions.
Fig 3.
Permutation-based variable importance scores using the tuned random forest model with all individual predictors.
Dots represent the mean cross-entropy loss for a given variable for 50 permuted samples for the observed data to account for the uncertainty associated with the importance of predictions.