Long-term effect of teenage birth on earnings: evidence from a British cohort study