Governing by tweet under a state of exception: an interrupt time-series and text-as-data study of president Bukele’s communication on X (2019–2025)

Abstract

This article investigates how El Salvador’s state of exception—an emergency regime declared in March 2022 that temporarily suspended selected constitutional guarantees in response to a gang-related security crisis—reshaped presidential communication and audience response on X/Twitter. A complete corpus of @nayibbukele’s posts (2019–2025) and daily interactions is analyzed using a quasi-experimental interrupted time-series (ITS) design with calendar controls and HAC errors, complemented by text-as-data: a composite Aggressiveness Index, the Endorsement Ratio (likes + reposts over replies + quotes + 1), topic modeling, lexicon-based framing, sentiment, and hashtag analysis. ITS reveals a discrete post-intervention shift: aggressiveness rises immediately and then gradually attenuates, while endorsement shows no level jump but a steeper positive slope thereafter. Content reconfigures from health/pandemic and local governance to security/anti-gang themes; frames referencing security and maras expand, whereas human rights, media/NGOs, and bitcoin/economy contract. Despite the aggressiveness spike, the sentiment mix becomes less negative and more neutral/positive post-intervention. Placebo cutoffs indicate that the engagement shift is better interpreted as a broader event-window pattern around the state of exception rather than as a discontinuity uniquely tied to a single day. The findings show how an emergency security turn with penal-populist features is performed and amplified through platform communication, aligning discourse, agenda, and audience interaction with punitive policy. The design—ITS coupled with transformer-assisted content measures—offers a replicable template for evaluating executive communication under institutional shocks.