We estimated the effects of an intervention which provided information about graduate wages to 5593 students in England, using a blinded cluster randomised controlled trial in 50 schools (registration: AEARCTR-0000468). Our primary outcome was students' choice of A-level subjects at age 16. We also recorded the students' expectations of future wages and the A-level subjects they intended to take before and after the intervention, and linked their data into national administrative school examination records. We found that an hour-long lesson on information about graduate wages affected students' beliefs about and choice of subject. They were more likely to take mathematics and less likely to take biology and computing. We found strong evidence that mediating factors such as their beliefs about average graduate salaries and their own likely salary in each subject were affected by the intervention. This suggests that providing accessible and credible information on labour market consequences of school choices may influence students' decisions. In the light of concerns about the quality of careers guidance for school students and expectations that educational choices should be well-informed, the study has clear implications for policy and practice.