In this article we present results from a study investigating the impact of three state exit exam systems on teaching and learning in college-preparatory schools. The study compares one state with a traditionally more centralized exam regime, one state that is more de-centralized and one state that has recently switched to more centralized testing. The German Abitur is a cognitively rather complex exam that is largely unstandardized as measured by the standards of international testing regimes. Moreover, performance differences in system monitoring tests between states with different exam regimes can only be found for mathematical literacy. Therefore, centrally regulated topics and grading criteria--as opposed to exams that are locally designed and reviewed centrally--make little difference in softer subjects with a more open canon, but seem to have a stronger impact in mathematics. Against this background and taking an international perspective, we argue that an overall low stakes testing regime might be the first step towards a good compromise between local flexibility for students' interests on one hand, and rigor as well as a healthy dose of performance motivation on the other hand. (Contains 1 figure, 7 tables, and 7 footnotes.)