While pay for performance has been implemented in the health care sector to increase treatment quality in many countries, Germany has just recently laid out the legal foundations for a pay for performance component with bonuses and fines for hospital services. Although the existing theoretical and empirical results of analyzing similar pay for performance systems in other markets show that a bonus for good quality does not necessarily have a positive effect on the latter, there is no empirical evidence on the effects of bonuses and fines.  We introduce a controlled online experiment in which we analyze the treatment behavior of medical students and real hospital physicians. We impose a sequential design varying the reimbursement. In part 1, participants are remunerated according to the current German DRG system. In part 2, we then introduce the new system with the pay for performance component including bonuses and fines. In the experiment, participants have to decide between two treatment options for each patient generating a conflict between maximal compensation for the hospital and optimal treatment for the patient. Treatment cases are real simplified cases for which treatment options are based on the respective evidence-based guideline. We find that hospital physicians do not alter their decisions significantly with the introduction of bonuses and fines. This is in contrast to the existing laboratory experiments using a student subject pool for performance pay with bonuses. The majority behaves altruistically both in the DRG and in the pay for performance system. Moreover, we find significant differences between hospital physicians and medical students. Firstly, in line with previous experimental studies hospital physicians behave more altruistically. Secondly, medical students do react significantly to bonuses and fines by making more patient-optimal decisions.  Our results for real hospital physicians are in line with previous empirical evidence in the sense that we find only limited impact of performance pay in the form of bonuses and fines on physician provision behavior. Our experimental design thus serves as a wind tunnel study and proposes that more research such as varying the level of bonuses is needed to determine effective performance incentives.