We examine the role of teachers and students in the formation of test scores at the higher secondary level (grade 12) in public schools in Delhi, India. Using the value added approach, we find substantial variation in teacher and student quality within schools: over the period spanning grades 11 and 12, being taught by a one standard deviation better than average teacher raises test scores by 0.373 standard deviation; and being a one standard deviation better than average student raises it by 0.799 standard deviation. Being permanent (tenured) positively predicts teacher effectiveness, while educational qualifications, training, experience and personality traits have no predictive power. Relative to families where only fathers earn, those where both parents earn negatively predict student effectiveness, while religion, caste and parents' education have no predictive power.