Can the segregation of administrators and citizens based on ethnicity improve justice delivery for hate-crimes? I examine caste-specific police stations in India, i.e., staffed by and that cater solely to an ethnic category (Scheduled Caste/Tribe) to pursue charges against out-groups. Using an original individual-level dataset of roughly half a million cases from 2017-2022, I conduct difference-in-differences analyses to estimate the bodies’ impact on case filings. I apply machine learning on citizen testimonies to uncover the content of hate-crime, and employ qualitative research to explore how co-ethnic institutions operate on-the-ground. To probe final outcomes, I analyze 1.3 million judicial rulings. Marginalized groups can face hate-crime, including sexual violence perpetrated against minority women by out-groups. Creating caste-only police enclaves generates mixed results; while there is evidence of increased state recognition of hate-crimes, and more complaints addressed by co-ethnic officers, the effects are geographically limited. The institutions are not associated with change in judicial outcomes for victims that pursue punishment against perpetrators. Ethnicity-based segregation could offer some benefits for complainants at specific stages of seeking restitution, but falls short of meaningfully altering the final justice that they receive.