research

Cross-Border Political Communication and Online Discourse

1. Borrowed Battlegrounds: The Formation and Reorganization of Foreign-Language Event-Publics around U.S. Political Violence

   Muzhi Liu

   Submitted

   Abs Bib

   When political violence strikes an American politician, how does it change the discussion mode of foreign publics on social media, and does it generate more attention to U.S. partisan politics, or more interest in discussing American politics online at all? To answer this question, I study 29 U.S. political-violence events from 2010–2025 in five foreign-language streams on X. My data collection combines four query families, covering the discourse associated with each event, its target, and U.S. political parties. From complete daily counts of posts in each query family, probability-sampled text with LLM labels, and a 1.3-million-post targeted collection around the 2024 assassination attempt on Donald Trump and the 2025 assassination of Charlie Kirk, I find the following. First, event-sparked discussion is uneven across events, with lethality being the dominant predictor of discussion volume, while visibility of the target before the event is not a significant predictor. Second, event discussion dilutes standalone party discussion by 21.9 percentage points and non-U.S. political recontextualization by 4 percentage points, pooling nine events, and these estimates are homogeneous across language groups. Their raw counts do not consistently increase. Third, sentiment toward the attacked person moves toward sympathy, with no change in affect toward that person’s party. The Trump and Kirk case studies confirm that the pattern holds even in outlier cases that trigger massive global attention: participants are overwhelmingly first-time entrants whose activity decays within days, and changes in semantic outcomes match the nine-event analysis. Finally, topic modeling reveals that discussion is contingent on event details rather than timeless topics about politicians and parties. These results suggest that even the most provocative event class—assassination—produces only short-lived global attention instead of drastically shaping how foreign publics perceive U.S. politics and parties.

   @unpublished{liu2026partisananchors,
     title = {Borrowed Battlegrounds: The Formation and Reorganization of Foreign-Language Event-Publics around U.S. Political Violence},
     author = {Liu, Muzhi},
     note = {Submitted},
   }

2. Estimand-First Comparative Text Analysis: Measuring Discourse Change Across Co-Exposed Platforms

   Muzhi Liu

   Submitted

   Abs Bib

   This project develops an estimand-first framework for comparative text analysis, in which cross-context comparisons of political discourse are organized around explicitly defined target quantities rather than off-the-shelf measurement pipelines. I illustrate the framework with an application to electoral spectatorship — how Chinese-speaking audiences followed and discussed the 2024 U.S. presidential election — combining LLM-assisted labeling of more than 14,000 posts and 379,000 comments from Weibo and Rednote with difference-in-differences designs.

   @unpublished{liu2026estimandfirst,
     title = {Estimand-First Comparative Text Analysis: Measuring Discourse Change Across Co-Exposed Platforms},
     author = {Liu, Muzhi},
     note = {Submitted},
   }

Political Identity, Group Attachments, and Voting Behavior

1. When Group Issues and Core Issues Collide: Evidence from Tailored Conjoint Experiments on Latino Voters

   Yamil Velez and Muzhi Liu

   Revise and Resubmit at Political Behavior

   Abs Bib

   This project examines how Latino voters respond when candidates take different positions on group-related issues and broader ideological or policy issues. Using tailored conjoint experiments, we study how voters weigh group attachments against other political considerations.

   @unpublished{velezliu2026groupissues,
     title = {When Group Issues and Core Issues Collide: Evidence from Tailored Conjoint Experiments on Latino Voters},
     author = {Velez, Yamil and Liu, Muzhi},
     note = {Revise and Resubmit at Political Behavior},
   }

2. Democracy Out of Sync: Circadian Misalignment and the Cost of Voting

   John B. Holbein, Xuan Li, and Muzhi Liu

   Under review

   Abs Bib

   This project studies how time-zone-induced circadian misalignment affects the cost of voting. We examine whether biological and temporal constraints influence electoral participation.

   @unpublished{liliu2026democracysync,
     title = {Democracy Out of Sync: Circadian Misalignment and the Cost of Voting},
     author = {Holbein, John B. and Li, Xuan and Liu, Muzhi},
     note = {Under review},
   }

3. Primary Election and Intra-Party Factions

   Shigeo Hirano, Sam Frederick, and Muzhi Liu

   Abs Bib

   This project studies factional competition within primary elections. Drawing on a newly assembled dataset of precinct-level primary election returns spanning dozens of U.S. states, we examine how intra-party divisions structure electoral competition and candidate behavior.

   @unpublished{hiranofrederickliu2026primary,
     title = {Primary Election and Intra-Party Factions},
     author = {Hirano, Shigeo and Frederick, Sam and Liu, Muzhi},
   }

4. Beyond Partisan Mean Gaps: Governing Roles and the Dispersion of Economic Perceptions

   Muzhi Liu

   Working paper

   Abs Bib

   Research on partisan economic perception compares the two camps’ average evaluations. We study their dispersion: whether the governing party’s identifiers share an economic perception. A heterogeneous motivated-weighting model predicts dispersion for camps facing incongruent evidence and convergence under congenial or overwhelming evidence. In the ANES 2016–2020–2024 panel, the same individuals sat farther from their party’s perception distribution in the waves when it held the presidency — across nominal-disagreement, ordinal-distance, and mean- and median-centered measures, three party-classification estimands, and both parties. The pattern is roughly four times weaker for household finances, and about nine-tenths reflects redistribution within reported personal-finance categories rather than compositional change. Across twenty calendar years, the governing-minus-opposition dispersion gap is wider in worse years, net of stable partisan dispersion differences — driven by opposition convergence, not a detectable governing-camp rise. Discordant governing-camp raters were more likely to defect; redistribution at a fixed mean does not detectably shield incumbents.

   @unpublished{liu2026withinparty,
     title = {Beyond Partisan Mean Gaps: Governing Roles and the Dispersion of Economic Perceptions},
     author = {Liu, Muzhi},
     note = {Working paper},
   }

Campaigns, Participation, and Everyday Political Behavior

1. Beyond the Ballot Box: Campaigns and the Politicization of Everyday Consumption

   Muzhi Liu

   Abs Bib

   This project examines whether campaigns and brand-politicization events generate politically patterned changes in everyday consumer behavior. Using Nielsen scanner data, local partisan context, and campaign-intensity measures, I study whether partisan cues shape consumer choice beyond traditional forms of political participation.

   @unpublished{liu2026beyondballot,
     title = {Beyond the Ballot Box: Campaigns and the Politicization of Everyday Consumption},
     author = {Liu, Muzhi},
   }

2. Campaign Giving Under Employment Threat: Federal Employees and the DOGE Workforce Reduction

   Muzhi Liu

   Revise and Resubmit at Political Behavior

   Abs Bib

   This project examines whether politically salient workforce reductions affect campaign donations and political backlash. Leveraging a measure of cut severity constructed from FedScope federal personnel records, I study how policy shocks and employment-related disruptions shape political behavior.

   @unpublished{liu2026demobilization,
     title = {Campaign Giving Under Employment Threat: Federal Employees and the DOGE Workforce Reduction},
     author = {Liu, Muzhi},
     note = {Revise and Resubmit at Political Behavior},
   }