We examine how within-group inequality, by influencing the group bias of state institutions, affects the cost-benefit calculus of individuals engaging in identity-assertive behaviour, that results in police complaints regarding hate crimes. We develop a two-stage contest model of between-group conflict, where the relative influence of a group over institutions, determined by an initial contest, affects subsequent hierarchy-establishing interaction between individuals belonging to opposing groups. Applying this model to caste conflict in India, we find that greater inequality among non-Scheduled Caste (non-SC) Hindus reduces the registered rate of crimes against SCs by non-SC Hindus, as well as the conviction rate for these crimes. Greater inequality among SCs increases both rates. Using state-level annual crime and household consumption data over 2005-2021, we find empirical support for these hypotheses. Between-group inequality does not appear to matter for either the rate of crimes against SCs or the conviction rate. Our analysis suggests that greater inequality within marginalized groups might increase reporting and punishment of aggression against them, thereby serving a protective function.