Fig. 2.

A) The likelihood of engaging in sanctioning behaviors based on whether group norms are asked before (red circles) or after (black circles) individual actions are shown. B) The likelihood of sanctioning, based on whether the respondent needs to be the first mover (black circles) or can “pile on” to someone else’s reply (red circles) is shown. In both panels, 95% CI are shown as lines. All models include tweet issue types as fixed effects, and the standard errors are clustered by respondent ID. Logistic regression models are used to analyze binary outcomes (all individual sanctioning behaviors), and OLS models are used when the outcome variable is continuous (average of all possible sanctioning behaviors).