Fig. 2: Sensitivity analysis.
From: Community-based fact-checking reduces the spread of misleading posts on X (formerly Twitter)
a The estimated ATTs (circles) across different response times from post creation to note display (grouped within 4-h windows). Full regression results and sample sizes are reported in Tables S17 and S18. b The estimated ATTs (circles) of community notes across the months following the roll-out of the “community notes” program in October 2022. Full regression results and sample sizes are reported in Tables S19 and S20. c The estimated ATTs (circles) of community notes depending on helpfulness scores. Full regression results and sample sizes are reported in Tables S25 and S26. d The estimated ATTs (circles) within subgroups separated by poster and post characteristics. Full regression results and sample sizes are reported in Tables S27–S32. In all plots, the error bars represent 99% CIs derived using the bootstrap method for 1000 resamples; the grey bands visualize the ATT (with 99% CIs) estimated via the two-period DiD model from our main analysis.
