PISA is one of the most widely used instruments to assess student performance in the world. This paper investigates how changes in testing dates and students' dates of birth (eligibility) between 2000 and 2018 have affected results in Brazil. Controlling for changes in students' grade distribution over time, we show that changes in grade composition are correlated with increases in average score. We estimate that about 60 % of improvements in math are unrelated to student improvement. Since changes in testing date and eligibility are unrelated to the school system, performance increases over time might not be a result of a country's educational policy.