Why Groups Diverge: Latino Support for Trump Across Three Elections

Political science offers a clearer account of how group identity generates cohesion than of how it produces divergence. This paper takes up that limitation through Latino support for Trump in 2016, 2020, and 2024. Using the Collaborative Multiracial Post-Election Survey, I apply random forest models with SHAP interpretation to identify the attitudes that most consistently shape Latino support for Trump once proximate and partisan predictors are removed. A recurring non-partisan structure appears across all three elections, organized around racial justice backlash, immigration restriction, and punitive law-and-order politics. A set of cycle-specific issues attaches to that core in each election without displacing it. The findings show that Latino support for Trump reflects cross-pressures generated within a racialized polity, and they point to racialization as a source of within-group divergence, not only group cohesion.