Does Prospective Physicians’ Behavior Depend on the Payment System? Results from a Laboratory Experiment Applying a Within-Subject-Design

Knowledge on how health care providers respond to incentives is necessary for the implementation of successful health policy reforms. However, researchers are often confronted with methodological challenges arising because they do not control the environment where data is generated: Ceteris paribus changes are rarely implemented in field settings, and the potential presence of unobservable variables or mechanisms such as self-selection makes it difficult to achieve causal inference with regard to strength and direction of effects.

In a few recent studies, researchers have applied experimental economic methods to address these methodological challenges.  An example is the study by Hennig-Schmidt, Selten and Wiesen (2011, Journal of Health Economics, 30, 637-646), who designed a controlled laboratory experiment to investigate how physician behavior is influenced by capitation (CAP) and fee-for-service (FFS) payment systems. They found that patients are under-served under CAP payment and over-served under FFS payment.

The novel feature of the design by Hennig-Schmidt et al. (2011) is an incentive to care for the patient benefit resulting from participants’ decisions. Even though no real patients are present in the lab, participants’ choices have consequences for real patients outside the lab: The money corresponding to the aggregated patient benefits is transferred by a trustworthy mechanism to a charity caring for real patients.

The aim of this paper is twofold: First we aim to confront the research question of Hennig-Schmidt et al. 2011 about how the two payment systems influence physician behavior to an experimental design that is robust with regard to selection mechanisms.  Second we aim at testing whether there is a carry-over effect of the previous payment mechanism when a new payment mechanism is introduced.

To achieve a design which is robust with regard to sample selection, the experimental design of Hennig-Schmidt et al. is modified by applying a within-subject-design where decision makers are confronted with both the CAP and the FFS treatment regimes. This design also enable the possibility of testing whether decision makers’ behavior are dependent on which payment system they experience first and second. The Experimental data is evaluated by means of non-parametric statistics, i.e., Wilcoxon rank-sum test and sign tests are applied when testing hypotheses.

We find that physicians provide significantly more services under FFS than under CAP, confirming previous experimental findings. Moreover, the patient benefit deviates significantly from the patient optimum under both payment systems. Further, we observe that in order to anticipate the effects of a payment reform on patient benefits, it is important to take the status quo payment system into account. Decision makers who first experience CAP and later are confronted with FFS provide higher patient benefits under FFS than students who first decide under FFS and then under CAP. Conversely, physicians provide a smaller patient benefit under CAP when CAP is the first payment system compared with a situation where CAP is the second payment system.