The effectiveness of class size reduction on student performance has been of great research interest and policy debate worldwide. Nevertheless, the evidence has been largely inconclusive partially due to potential bias arising from non-random placement of students and teachers in classrooms of different sizes. This study applied instrumental variable (IV) methods and used a regression discontinuity design (RDD) to conduct analyses of TIMSS data in 2003, 2007 and 2011. The purpose was to investigate over time the effects of class size on eighth grade students' cognitive and non-cognitive outcomes on five mathematics and science subjects in four European countries (i.e. Hungary, Lithuania, Romania and Slovenia). IV estimates indicate that in Romania in 2003 smaller classes had significant and positive effects on academic scores in mathematics, physics, chemistry and earth science and in 2007 on enjoyment to learn mathematics. In Lithuania, in 2011 small classes had significant and positive effects on enjoyment to learn biology and chemistry and learning chemistry well. In 2007 however, the effects were reversed on some of the biology related non-cognitive outcomes. Overall, the class size effects that were significant were also substantial in magnitude and typically much larger than the effects reported in Project STAR.