The politics of enforcement after a focusing event: A Wave 1–Wave 2 panel study of legitimacy, moral boundaries and affective polarisation in U.S. immigration enforcement and policing

This pre-registration covers Wave 2 of a two-wave study of legitimacy, polarisation, and identity in U.S. immigration enforcement and policing. A central question is whether legitimacy judgments update in the same direction following a focusing event (parallel shifts, reflecting responsiveness to a shared information environment) or in opposite directions (asymmetric updating, reflecting crystallised identity-based frameworks that shape not only baseline attitudes but how new events are interpreted). A second set of questions concerns the mechanisms of updating, including affective polarisation and information-processing filters that determine when incident information generalises to institutional legitimacy. Wave 1 (N = 3,253) established that ICE legitimacy is sharply polarised and organised by a compact psychological bundle: structural awareness, oppositional identity, and intergroup emotions. This bundle accounts for an extraordinary share of variance. When included in models, the progressive coefficient fell by 99% and became statistically indistinguishable from zero. Wave 1 also included an experimental component that produced null results: neither a structural frame (presenting ICE as violating due process) nor an oppositional frame (presenting ICE as under partisan attack) shifted legitimacy judgments. We interpret these null effects as evidence that attitudes were already crystallising: brief abstract frames largely reiterated propositions that respondents had already incorporated into stable interpretive orientations. Wave 2 re-interviews a panel subset of Wave 1 respondents in the wake of a major focusing event: the killings of U.S. citizens Renee Good and Alex Pretti during ICE operations in Minneapolis in January 2026. The primary descriptive test adjudicates between the crystallisation hypothesis (progressives decline while conservatives hold steady or dig in, widening the ideology gap) and the parallel-shift alternative (all groups move in the same direction, even if at different rates). We characterise change using mean change, dispersion, proportions shifting upward versus downward, and change in the progressive–conservative gap. A puzzle motivates our mechanistic hypotheses: if crystallised attitudes resist change because people already hold the relevant beliefs, why would a focusing event produce updating when a brief experimental frame did not? Drawing on construal level theory (Trope & Liberman, 2010), we consider whether the experiment operated at the wrong construal level: psychologically distant, abstract frames are filtered by settled cognitive frameworks, whereas concrete, psychologically proximate events engage experiential-affective processing. In this account, Wave 1 orientations set up affective potential; the focusing event actualises that potential; and acute emotional response is the proximal driver of updating. We evaluate an affective-primacy account using joint models that compare affective pathways (intergroup and event-evoked emotions) against cognitive construal pathways (e.g., justification and typicality) when both are modelled together. Wave 2 extends Wave 1 through multiple pre-specified analytic components. First, it tests crystallisation and selective updating: whether the Wave 1 bundle replicates and tightens and whether Wave 1 orientations predict the direction of change. Second, it tests dynamic boundary processes (scope of justice): whether moral inclusion and critic legitimation shift after the focusing event and whether such revision is norm-conditioned. Third, it tests crimmigration spillover: whether incident-activated contamination of police legitimacy occurs, conditional on perceived ICE–police entanglement. Fourth, it tests affective polarisation: asymmetric emotional dynamics and identity-centrality amplification. Finally, it includes an exploratory mechanisms module on information processing, modelling four filters (acquisition/engagement, perceived elite cue environment, attributional generalisation versus exception, and normative translation) that can attenuate or amplify the path from incident exposure to institutional legitimacy updating. Analytically, we pre-specify families of confirmatory hypotheses (with Holm correction within each family) and treat additional tests as exploratory. For repeated constructs we assess longitudinal measurement invariance. Primary analyses use change-score regression with baseline adjustment; robustness checks use latent change score models. Exposure, elite cues, and perceived norms are treated as endogenous features of the incident environment. Attrition diagnostics and inverse-probability weighting corrections are pre-specified.