Extended Data Table 1 Demographics of Twitter Panel and Associated Subgroups
From: Post-January 6th deplatforming reduced the reach of misinformation on Twitter
- The panel of active users has previously been linked to a proprietary voter file. This allows for comparison of the demographic attributes of users in our sample. The panel’s demographics are broadly representative of the population of Twitter users, although it skews somewhat white and female5. Partisanship is derived from a modeled propensity to identify as one party or the other, not party registration; validation of this measure is presented in Hughes et al4. Other authors have noted that misinformation sharers tend to be older and more Republican8. Extended Data Table 1 presents the demographic composition of our subgroups. We note that this descriptive pattern holds here, and we expand this further to note that this tendency is present across our subgroups of interest, ranging from relatively anodyne behaviors such as Trump following up to suspension (and QAnon support among non-suspended users). Relative to non-suspended users, suspended users are significantly older (suspended mean 48.1, non-suspended mean 39.2, t = 22.8, p < 1e-16), significantly more white (suspended 92.0%, non-suspended 85.1%, t = 7.0, p < 1e-11), significantly more female (suspended 56.3%, non-suspended 50.6%, t = 29.6, p < 1e-16), and significantly more Republican (suspended 55.1%, non-suspended 30%, t = 20.0, p < 1e-16).
