Heterogeneous treatment effects using an LPM and Specification (6)
We estimate one model per subheading and include a treatment effect (Post) per subgroup instead of the overall treatment effect. The figure plots the point estimates and the 95 percent confidence intervals. See Supplementary Table 5 for exact statistics and the number of observations of population subgroups. Source data are provided as a Source Data file.
We find no conclusive evidence
We find no statistically significant evidence that climate protests negatively impact climate change concerns in different subpopulations (see Fig. 4 for the plotted coefficients). We test effects depending on age (above median: coef. = 1.59pp, p = 0.003; below median: coef. = 0.81pp, p = 0.212), sex (female: coef. = 1.51pp, p = 0.026; male: coef. = 0.83pp, p = 0.165), income (above median: coef. = 0.82pp, p = 0.012; below median: coef. = 1.54pp, p = 0.052), education (above median: coef. = 0.66pp, p = 0.364; below median: coef. = 1.98pp, p = 0.025), attitudes towards the future (optimistic: coef. = 1.56pp, p = 0.056; pessimistic: coef. = 2.18pp, p = 0.041), political orientation (right-leaning: coef. = 3.02pp, p = 0.007; center: coef. = 2.05pp, p = 0.069; left-leaning: coef. = 1.69pp, p = 0.118), and interest in politics ((very) strong: coef. = 2.09pp, p = 0.001; weak or none: coef. = 0.49pp, p = 0.478). A rich body of literature demonstrates that individual-level factors, ranging from socioeconomic characteristics to values and worldviews, are associated with people’s beliefs and concerns about climate change 33,34 . Read more
