The axioms of expected utility and discounted utility theory have been tested extensively. In contrast, the axioms of social welfare functions have only been tested in a few questionnaire studies involving choices between hypothetical income distributions. In a controlled experiment with 100 subjects placed in the role of social planners, we test five fundamental properties of social welfare functions to determine the efficacy of traditional social choice models in predicting social planner allocations when presented with choice sets designed to test the axioms of the theory. We find that three properties of the standard social welfare functions tested are systematically violated, producing an Allais paradox, a common ratio effect, and a framing effect in social choice. We find support for scale invariance and a preference for tail-increasing transfers. Our experiment also enables us to test a model of salience-based social choice which predicts the systematic deviations and highlights the close relationship between these anomalies and the classical paradoxes for risk and time.

期望效用公理和折现效用理论已被广泛测试。相反，社会福利功能的公理仅在一些涉及假设收入分配之间选择的问卷调查中得到检验。在一项由100名受试者担任社会计划者角色的对照实验中，我们测试了社会福利功能的五个基本属性，以确定传统的社会选择模型在预测社会计划者分配时的效用，该模型带有旨在测试社会公理的公理。理论。我们发现，被测试的标准社会福利功能的三个属性被系统地破坏，从而产生了阿莱思悖论，共同比率效应和社会选择中的框架效应。我们发现支持音阶不变性并偏爱增加尾部的传输。我们的实验还使我们能够测试基于显着性的社会选择模型，该模型可以预测系统偏差并突出显示这些异常与经典悖论之间在风险和时间方面的紧密关系。