This study examines strategic behavior in crowdfunding using a large-scale online experiment. Building on the model of Arieli et. al 2023, we test predictions about risk aversion (i.e., opting out despite seeing a positive private signal) and mutual insurance (i.e., opting in despite seeing a negative private signal) in a static, single-shot crowdfunding game, focusing on informational incentives rather than dynamic effects. Our results validate key theoretical predictions: crowdfunding mechanisms induce distinct strategic behaviors compared to voting, where participants are more likely to follow private signals (odds ratio: 0.139, $p < 0.001$). Additionally, the study demonstrates that higher signal accuracy (85\% vs. 55\%) decreases risk aversion (odds ratio: 0.414, $p = 0.024$) but increases reliance on mutual insurance (odds ratio: 2.532, $p = 0.026$). However, contrary to theory, increasing the required participation threshold (50\% to 80\%) amplifies risk aversion (odds ratio: 3.251, $p = 0.005$), which, pending further investigation, may indicate cognitive constraints. Furthermore, we show that while mutual insurance supports participation, it may hinder information aggregation, particularly as signal accuracy increases. These findings advance crowdfunding theory by confirming the impact of informational incentives and identifying behavioral deviations that challenge standard models, offering insights for platform design and mechanism refinement.

翻译：本研究通过一项大规模在线实验考察众筹中的策略行为。基于Arieli等人（2023）的模型，我们在静态单次众筹博弈中检验了关于风险规避（即在观察到正向私有信号时仍选择退出）与互助保险（即在观察到负向私有信号时仍选择参与）的理论预测，重点关注信息激励而非动态效应。实验结果验证了关键理论预测：与投票机制相比，众筹机制会诱发显著不同的策略行为，参与者更倾向于遵循私有信号（优势比：0.139，$p < 0.001$）。此外，研究证明更高的信号准确率（85% vs. 55%）会降低风险规避倾向（优势比：0.414，$p = 0.024$），但同时增强对互助保险的依赖（优势比：2.532，$p = 0.026$）。然而与理论预测相悖的是，提高参与门槛要求（50%至80%）反而加剧了风险规避（优势比：3.251，$p = 0.005$），这可能暗示认知约束的存在，有待进一步探究。研究还发现，虽然互助保险能促进参与，但可能阻碍信息聚合，这种效应在信号准确率提升时尤为明显。这些发现通过证实信息激励的影响并识别出挑战标准模型的行为偏差，推动了众筹理论的发展，为平台设计与机制优化提供了重要启示。