Abstract
The relationship between class size and school performance has always been ambiguous and the current literature has found no direct connection between them, especially in the Brazilian context. Therefore, this study aimed to verify whether the number of students per class influences school performance. We used Microdata from the Prova Brazil of 2017. Using the propensity score matching statistical model, with the nearest neighbor matching estimator, we grouped the classes into clusters by similarity. The metric used to group the clusters was the Euclidean distance. We attempted to verify adherence to the normal distribution of data using the Kolmogorov Smirnov test and tested the null hypothesis of the medians using the Wilcoxon test. All the statistical analysis were performed using SPSS Statistic version 20. The results showed that the number of students per class has little influence on performance and, when the influence exists, larger classes perform better.