Background: The social gradient in mortality among young adults is well-documented, but poorly understood. This study contrasts the role of parental income and education versus own school performance as predictors of early death. Methods: We estimated the prospective association between grade point average from lower secondary education and mortality between age 16 and 30 for the Norwegian birth cohorts 1985-2002 (N = 986,573). This association was compared to the associations between measures of family socioeconomic status and young adult mortality in the offspring. We also estimated hazard ratios in models including both school performance, parental income, and education. Sibling comparison analyses were also estimated to assess the importance of within-family variation. Last, we used the cause of death register to estimate the hazard ratio of different causes of death among individuals with poor school performance. Findings: During the observational period, we observed 1300 deaths in the lowest quartile of school performance (0.06%) and 413 deaths in the highest quartile of school performance (0.02%). The risk of early death among boys in the lowest quartile of school performance compared to the highest quartile was larger for boys [HR = 3.68, 95% CI 3.54-3.81] than for girls [HR = 2.82, 95% CI 2.62-3.01]. The corresponding hazard ratios across parental income quartiles were 1.79 [95% CI 1.67-1.1.91] for boys, and 1.63 [95% CI 1.43-1.82] for girls. Across parental education level, the hazard ratio was 2.03 [95% CI 1.85-2.21] for boys, and 1.64 [95% CI 1.35-1.93] for girls. When jointly including school performance, parental income and education in the same model, parental income and education were insignificant, while the association school performance remained strong: HR = 3.57 [95% CI 3.44-3.71] for boys, and 2.98 [95% CI 2.78-3.18] for girls. With regards to causes of death, the highest hazard ratio among those in the lowest quartile of school performance was for drug-related poisoning, with 6.47 [95% CI 2.78-3.18] for boys and 7.3 [95% CI 4.93-10.80] for girls. The results were consistent in sibling comparison analyses. Interpretation: School performance is a substantially stronger predictor of early death than common measures of socioeconomic background. School performance absorbs the social gradient in early death.