Government Warnings and the Limits of Deterrence: TikTok Tax Misinformation and Search Behavior

• In Event: Media and Misinformation

Abstract

Can a government warning aimed at social media misinformation deter engagement with harmful content and reduce downstream interest in the behavior promoted by that content? This paper examines that question through the Internal Revenue Service’s February 27, 2025 “Dirty Dozen” announcement, which for the first time explicitly identified TikTok as a source of dangerous tax advice. The announcement created a rare opportunity to evaluate whether a public enforcement communication can alter behavior in a platform-centered misinformation environment. Although the substantive setting involves tax compliance, the policy question extends more broadly to regulatory responses to misinformation involving finance, health, and other high-stakes domains.

A two-population framework explains why government warnings may generate uneven effects across audiences. Platform-embedded users may never encounter the warning or may discount the signal, while broader publics reached through mainstream media and search channels may update beliefs about enforcement risk and reduce information-seeking related to the flagged behavior. Deterrence theory therefore competes with backfire theory, and the central empirical question concerns not only whether a warning works, but also for whom and through which behavioral channel.

The study combines platform and off-platform measures of response. A regression discontinuity design is applied to 4,702 TikTok videos classified with large language model assistance to estimate whether engagement with tax “hack” content changed after the IRS announcement. In parallel, 149 days of Google Trends data covering 20 aggressive-deduction keywords are used to test whether public search demand for targeted strategies declined after the warning. Compliance-intent searches provide a comparison series that distinguishes targeted deterrence from a general decline in tax-related attention. The design also addresses an important methodological issue: because the announcement occurred near the steepest portion of the filing-season engagement ramp, short regression discontinuity windows risk producing spurious discontinuities when seasonal curvature is not adequately modeled.

Results indicate no measurable reduction in engagement with TikTok hack content following the announcement, suggesting limited penetration into the audience most embedded in the platform-native misinformation ecosystem. In contrast, Google search demand for aggressive deduction strategies declines significantly after the warning, while compliance-oriented searches show no comparable decrease. Additional evidence indicates that TikTok hack content predicts downstream search interest at a short lag. Taken together, the findings support a two-population interpretation: government warnings transmitted through mainstream channels can deter the broader searching public, but such warnings appear substantially less effective for highly platform-committed audiences. The paper contributes to research on regulatory communication, platform governance, tax compliance, and the limits of deterrence in the social media era.

Authors

• Edmund Poku Adu, Arkansas State University
  Presenting Author

• Clement Mensah Damoah, California State University, Fullerton
  Non-Presenting Co-Author