We evaluate the Education Maintenance Allowance, a large conditional cash transfer scheme that paid low-income teenagers in England to remain in education beyond age 16. Using the staggered national roll-out of the programme and linked administrative data tracking education, earnings, welfare payments and criminal convictions to age 31, we find no significant overall effect of the policy on labour market outcomes or criminality. High-attaining students were more likely to attend university but no more likely to graduate. Low-attaining students committed fewer crimes. We estimate the Marginal Value of Public Funds was 0.85 (95% confidence interval 0.52–1.29); even at the upper bound of this interval, benefits barely outweigh costs.