We estimate the effect on primary school students’ achievement of a Brazilian policy that changed the entry age into first grade from seven to six years old. Before the policy, a typical child spent up to three years in pre-school, and four years in primary school. The reform reorganized the system so that pupils would spend two years in pre-school (at ages 4 and 5) and five years in primary education. The timing of national tests and of the incremental process of implementing the reorganization allowed us to use a DDD approach to make causal estimates of its effect on student achievement. The reform increased 5th grade students’ mathematics scores by 0.10 SDs and reading scores by 0.12 SDs over student gains in untreated schools. These effects are robust across various treated cohorts, and across alternative samples of schools.