ABSTRACT
This registered report details a longitudinal experiment testing political self-effects – changes in issue-related attitudes and indicators of political self-concept – resulting from political expression on social media. We also tested whether such effects were due to the composition or release of political messages on social media over time. American adults (N = 576), logged on to a fictional social media platform and engaged with discussions involving one of two political issues at three time points over a week. We examined whether participants who wrote a comment about the issue publicly (on the social media platform) or privately (in a text box) experienced changes in their issue-related attitudes or political self-concepts (compared to a control condition). Overall, we did not find consistent main effects of message composition or release at individual timepoints. Instead, message composition sustained issue-related attitudes over time (vs. control). Interaction analyses found that among those with high baseline issue interest, expression (vs. control) led to decreased issue interest and importance over time. Among those high in baseline political interest, message release (vs. control) increased positive issue-related attitudes. We outline an agenda for studying political self-effects that addresses, a) their cumulative nature, b) conditional effects, and c) the challenge of endogeneity.
Acknowledgement
The authors would like to sincerely thank the editor and anonymous reviewers, whose excellent feedback strengthened this study throughout the research process. Much gratitude goes to Justice Quick who made contributions to earlier forms of this work and to research assistants Ben Diaz and Sydney Yamanishi, whose assistance was invaluable.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1. Pre-registration specified a fixed 2-minute exposure to the platform. We removed this fixed exposure over concerns of attention loss and drop-outs during out pre-testing.
2. Pre-registration instructed participants to write a paragraph. Upon pre-testing we deemed this was unrealistically lengthy and reduced the request to “2–3 sentences.”
3. W3 manipulation check also confirmed successful manipulation.
4. Pre-registration specified that time would be entered as a random effect, but this led to models that would not converge. Therefore, time was only modeled as a fixed-effect.