[Seminaire CREM] On the stability of social preferences

Abstract:
Altruism is a cornerstone of medical professionalism. Yet, its formation and stability throughout medical education remains largely underexplored. This study investigates the temporal stability and heterogeneity of patient-regarding altruism among German medical students using a longitudi- nal experimental design. A total of 1,413 observations were collected from medical students at the University of Cologne, a major medical school in Germany, across four stages of medical education. From 515 students, we collected multiple observations. Medical students engaged in incentivized decision-making tasks simulating physician-patient trade-o!s, where their choices a!ected both per- sonal earnings and charitable donations for real-world cataract surgeries.
The findings reveal that altruistic behavior remains relatively stable over time, with limited within- subject variation. However, substantial between-subject heterogeneity persists, much of which re- mains unexplained by observable characteristics such as demographics, stated preferences, or person- ality traits. Gender di!erences were initially observed, with women displaying higher altruism, but these e!ects diminished when controlling for other variables. Structural estimations using CES and Fehr–Schmidt utility models confirmed the stability of core altruistic parameters, though elasticity of substitution showed greater variability.
Variance decomposition and endogenous group analysis further demonstrated that individual het- erogeneity cannot be captured by a small number of homogeneous subgroups. The results underscore the complexity of modeling physician altruism and challenge the notion of a representative agent in health economics. These insights have implications for medical education and policy design, emphasizing the need to account for persistent individual di!erences in social preferences.
Co-écrit avec Arthur E. Attema, Matteo M. Galizzi, Olivier L'Haridon, François-Charles Wolff

Source : OpenAgenda