发表机构
Anglo-American University; Institute of Economic Studies, Charles University; Centre for Economic Policy Research; Meta-Research Innovation Center, Stanford（英美大学; 查理大学经济研究所; 经济政策研究中心; 斯坦福大学元研究创新中心）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究通过元分析发现，法治越强反而对应更大的规模溢价，而信任仅在法治薄弱时削弱溢价，揭示了正式与非正式制度对规模溢价异质性的解释作用。
AI 中文摘要
关于规模溢价（即小公司平均回报率高于大公司的趋势）的已报告估计值在研究、国家、时期和设计之间差异很大。我们考察了普遍信任和法治是否有助于解释这种异质性。小公司更加不透明，且更依赖外部融资，因此执法和信息环境对它们的影响应大于对大公司的影响。我们研究了来自105项研究和31个国家的1,613个已报告规模斜率估计值。元回归控制了研究设计、规格、精度、发表背景以及市场和宏观金融条件；贝叶斯模型平均法评估了控制集的不确定性。与法治的关联更为稳定，且与更好的法律机构会缩小溢价的直觉预期相反：更强的法治与更负的已报告规模斜率相关，因此对应更大的传统规模溢价。信任的关联是有条件的且估计精度较低：在法治薄弱的地方，较高的普遍信任与较不负的已报告斜率（从而较弱的溢价）相关，而这种关联随着法治的加强而减弱。因此，正式和非正式制度有助于组织该文献中的部分分歧，尽管分析涉及已报告估计值的变化，并未识别因果效应。
英文摘要
Reported estimates of the size premium, the tendency of smaller firms to earn higher average returns than larger firms, vary widely across studies, countries, periods, and designs. We examine whether generalized trust and rule of law help account for that heterogeneity. Small firms are more opaque and more dependent on outside finance, so the enforcement and information environment should matter more for them than for large firms. We study 1,613 reported size-slope estimates from 105 studies and 31 countries. The meta-regressions control for study design, specification, precision, publication context, and market and macro-financial conditions; Bayesian model averaging assesses uncertainty over the control set. The more stable association is with rule of law, and it runs against the intuitive expectation that better legal institutions shrink the premium: stronger rule of law is associated with more negative reported size slopes, hence larger conventional size premia. The trust association is conditional and less precisely estimated: where rule of law is weak, higher generalized trust is linked to less negative reported slopes (and thus a weaker premium), and this link fades as rule of law strengthens. Formal and informal institutions thus help organize part of the disagreement in this literature, although the analysis concerns variation in reported estimates and does not identify causal effects.
Comments54 pages, 5 figures, 11 tables. Replication package and online appendix: https://meta-analysis.cz/trust
