Extended Data Fig. 1: Replication of the DID results varying the number of deplatformed accounts.
From: Post-January 6th deplatforming reduced the reach of misinformation on Twitter
DID estimates where the intervention depends on the number of deplatformed users that were followed by the not-deplatformed misinformation sharers. Results are two-way fixed effect point estimates (dots) and 95% confidence intervals (bars) of the difference-in-differences for all activity levels combined. Estimates use ordinary least squares with clustered standard errors at user-level. The Figure shows results including and excluding Trump followers (color code). The x-axis shows the minimum number of deplatformed accounts the user followed from at least one (1+) to at least ten (10+). Total sample sizes for each dosage level: Follow Trump (No): 1: 625,865; 2: 538,460; 3: 495,723; 4: 470,380; 5: 451,468; 6: 437,574; 7: 426,772; 8: 417,200; 9: 408,672; 10: 401,467; Follow Trump (Yes): 1: 688,174; 2: 570,637; 3: 514,352; 4: 481,684; 5: 460,676; 6: 444,656; 7: 432,659; 8: 421,924; 9: 413,241; 10: 405,766.
