Exclusion is a disciplinary practice used by headteachers which removes misbehaving pupils from the classroom or from the school, either temporarily or permanently. Its growing use has led to increased concern about potential negative effects on excluded pupils, including on their educational achievement. This paper estimates the effect of being excluded on subject test scores and teacher assessment outcomes using detailed administrative data on an entire cohort of pupils in the English state school system. To mitigate selection bias, we use a novel empirical approach for this literature which compares excluded pupils with pupils who experienced exclusion after outcomes were measured but not before. We find that excluded pupils perform worse in subsequent tests and teacher assessments, with 0.03-0.07 standard deviation lower standardised test scores and 2.5-3.6 percentage point higher probability of not reaching the expected level in teacher assessments. We assess the extent to which these estimated associations might reflect a negative causal impact of exclusion.