The role of educational experiences in shaping political behaviour has received extensive attention in recent years. However, the mechanisms through which education changes people’s political behaviour in adulthood are often hidden due to poor measures of education. Using new linked administrative and survey data, we try to unpack the causal “black box” of education in the English context, where educational divides were starkly evident during the Brexit referendum and subsequent realignment of the electorate in 2019. Specifically, we link detailed administrative data on adults’ school records to their political views in a large-scale probability household panel survey, and test how subjects studied in adolescence are linked with political attitudes and behaviour in adulthood, net of attainment and other characteristics. We find that studying Economics/Business Studies and, to a lesser extent, Maths is associated with more right-wing economic and social attitudes, and with voting for the Conservative Party. On the contrary, studying English in particular is associated with more socially liberal and economically left-wing positions, and voting for the Labour Party. We argue that subject-specific effects on political outcomes may reflect direct learning from subject content, in addition to the socialisation and human capital effects identified in prior research. To further understand the causal foundation of these effects, we (i) analyse whether just passing a crucial educational hurdle at age 16 changes political behaviour in adulthood, and (ii) study schooling reforms that changed the set of subjects offered to students in different schools. Our results provide evidence for the micro foundations of how educational expansion has reshaped patterns of political conflict.