Tweeting cheap talk: Elites' communication strategies during corruption scandals

Tweeting cheap talk: Elites' communication strategies during corruption scandals — a co-authored paper with Jason Gainous, Mayra Vélez-Serrano, Kevin M. Wagner, Jialong Wang, You Wu, and Weiheng "Mark" Liu. Presented at the 80th Annual Meeting of the Midwest Political Science Association, Chicago IL, 2023. Published OnlineFirst in International Political Science Review on May 9, 2026. DOI: 10.1177/01925121261429933.

The question

Politicians say they will fight corruption all the time. They rarely push substantive reform. Why? When a corruption scandal breaks, who actually engages with it on social media, and in what terms? Are they offering proposals (reform-committed talk), or are they performing concern without committing to anything (cheap talk)?

The case

Puerto Rico, 2014 to 2021. Two main parties (PNP, PPD) trade power; three minor parties (PIP, MVC, PD) hold a handful of seats but no governing majority. Across this window we identified 43 corruption scandals — including the 2019 mass-protest ouster of Governor Ricardo Rosselló and the 2019–2020 arrest of ex-Governor Wanda Vázquez for receiving bribes — and pulled the universe of Twitter posts from every legislator and governor across both major and minor parties.

The method

• Data: scraped via Python from every active legislator and governor account; subsetted to 8,697 tweets from 143 legislators within a 3-day window of each of the 43 events (PNP 5,312, PPD 2,812, PIP 303, MVC 153, PD 22, Independents 95)
• Topic models: Latent Dirichlet Allocation (LDA), 16 data subsets crossed by party × corruption-keyword presence × event window, 10 topics each, top-20 words per topic, coherence-tuned
• Dictionary: a custom Spanish corruption dictionary built from newspaper coverage of the 43 events plus the top corruption-related words extracted from the LDA output, separating "cheap talk" vocabulary from "reform-committed" vocabulary
• Index: each legislator scored on a continuous index from −1 (pure cheap talk) to +1 (pure reform-committed), then a linear regression on majority-party status with controls for NRC sentiment (positive + negative), electoral competitiveness, at-large vs. district seat, seniority, governor/senator dummies, age, and gender

What we found

Two findings, one stark and one telling.

1. Majority parties barely discussed corruption — and when they did, it was vague. Restricting the data to tweets from major-party members within three days of an arrest involving either major party, the result was almost embarrassing: only 1–9 tweets per subset across four such cuts. Major-party members aren't avoiding corruption because they have nothing to say. They're avoiding the topic outright, keeping their own party's vulnerabilities off the timeline.

2. The reform-committed–cheap-talk index confirms it at the legislator level. Holding a majority-party seat predicts significantly less reform-committed language (β = −0.001, p = 0.01), even after controlling for sentiment, electoral context, and institutional position. The sentiment controls are not significant, which matters: what distinguishes majority from minority rhetoric is not emotional intensity but substantive focus on reform. Cheap talk isn't a partisan tactic — it's a structural strategy of those in power.

Minor parties show the opposite pattern. They discuss corruption frequently and substantively, with reform proposals embedded in the discourse. We read this through standard incumbency logic: those benefiting from the status quo have everything to lose from substantive reform, and the rhetoric tracks the incentive.

The paper reads like communication science with a political-economy bone. The fun of it — the reason we wrote it — is that a small institutional system observed across eight years of scandals turns out to recover a rhetorical asymmetry between who holds power and who doesn't, with no need to be told to look for it.