We examine the power of media narratives to counter prejudice, a dimension often overlooked in a literature that predominantly documents how media reinforce it. We study an unprecedented experiment from post-WWII United States: in 1946, amid entrenched racial divisions, the producers of the popular children’s radio series The Adventures of Superman suddenly replaced its usual fantasy plots with stories confronting intolerance in American society, designed to elicit moral and emotional engagement among millions of young listeners. The shift generated exogenous variation in exposure to anti-prejudice narratives across cohorts and locations, a feature that we verify using historical listenership records. We show that childhood exposure produces lasting changes in social and political preferences, including greater racial tolerance, stronger support for civil rights, and higher civic mobilization later in life. These individual-level effects translate into societal shifts in favor of civil rights. Our results confirm that countering prejudice requires deeper and more enduring psychological processes compared to persuasion through information or framing.

How does violent state repression affect social capital and political participation in mature democracies? We study the 2001 G8 summit in Genova, Italy, where police violently repressed authorized protests, causing one death, hundreds of injuries, and documented abuses. We proxy local exposure to repression through pre-summit associational networks: municipalities more connected to the Genoa Social Forum (GSF), the main mobilization and logistics channel, were more likely to send participants and, in the aftermath, circulate direct testimony of the facts. Our strategy relies on a market-access style exposure index based on proximity to GSF associations and estimate a difference-in-differences design with continuous treatment using municipal election panels and repeated cross-sectional survey data. We find that higher exposure persistently depresses turnout in subsequent national elections, with no evidence of systematic left-right realignment; instead, exposure increases transitions from voting to abstention across blocs. Additional survey evidence is consistent with a mechanism of “vertical” distrust: exposure predicts lower trust in the police, while generalized trust and sense of local belonging rise.