Violence and ethnic identity

Can violence erase an ethnic identity? We study this question using the Shining Path insurgency in Peru (1980-1992), which sought to replace competing affiliations with a single class identity and killed roughly 70,000 people, three-quarters of them Indigenous. Combining individual-level data on self-identification and mother tongue with event-level data on Shining Path violence, we implement a difference-in-differences design that exploits variation in exposure across cohorts at different stages of identity formation, within a matched border sample of pre-conflict comparable districts. Individuals exposed during their formative years (ages 0-19) are substantially less likely to identify as Indigenous or to speak an Indigenous language than cohorts whose identity was already formed when violence arrived. The effect is concentrated mostly in early childhood and particularly pronounced for mother tongue; the margin most directly shaped by parental transmission. Violence against non-Indigenous victims has no comparable effect. The mechanism is intra-group violence: because perpetrators were overwhelmingly co-ethnics from the same villages, visible Indigenous identity offered no protection and imposed costs. Consistent with this, the effect is strongest in Indigenous-homogeneous districts.