Better luck next time: Learning through retaking

•We estimate cumulative learning among repeat-takers of the Turkish college entrance exam.•We control for selection into retaking in terms of observables and unobservables.•We overcome lack of panel data using a model and rich performance data.•We find large improvements in exam scores, especially among disadvantaged students.•The Turkish system reduces effect of background on school placement.

In this paper we provide some evidence that repeat taking of competitive exams may reduce the impact of background disadvantages on educational outcomes. Using administrative data on the university entrance exam in Turkey we estimate cumulative learning between the first and the nth attempt while controlling for selection into retaking in terms of observed and unobserved characteristics. We find large learning gains measured in terms of improvements in the exam scores, especially among less advantaged students.