发表机构
University of British Columbia; RALP Technologies（不列颠哥伦比亚大学; RALP科技有限公司）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文识别出异步观察共同马尔可夫状态的理性主体因信息年龄差异产生信息不对称的内生机制，在高斯AR(1)环境中分析其后果，相关框架对多领域有启示意义。
AI 中文摘要
标准理论将信息不对称归因于私人持有信息、异质性信号或信息获取成本。本文识别出一种独特机制：在异步时间观察共同马尔可夫状态的其他相同理性主体，会因信息年龄不同而持有不同的预测分布。当转移核分离信息年龄时，独立伯努利观察到达会使这种散度几乎必然无限次重现，即使到达概率相同。当主体选择成本高昂的更新强度时，独立更新实现会在事前相同的主体中再生异质性信息年龄。在平稳高斯AR(1)环境中，结果取决于共同陈旧性和年龄差距：固定差距时，更旧的信息会增加个体不确定性，同时降低预期平方预测分歧和证明相对新鲜度的价值，这产生了陈旧共识：高不确定性下仍低分歧。在多个有序年龄下，成本高昂的证明支持部分瓦解均衡，其中更新鲜的类型进行证明，足够陈旧的类型形成聚合。在几何信息年龄下，只要存在正证明均衡，更频繁的更新会弱化最小和最大均衡披露阈值。该框架对信贷、保险、金融市场和多智能体AI具有启示意义。
英文摘要
Standard theories trace information asymmetry to privately held information, heterogeneous signals, or costly information acquisition. We identify a distinct mechanism: otherwise identical rational agents observing a common Markov state at asynchronous times can hold different predictive distributions because their information differs in age. When the transition kernel separates information ages, independent Bernoulli observation arrivals make such divergence recur infinitely often almost surely, even with identical arrival probabilities. When agents instead choose costly updating intensities, independent update realizations can regenerate heterogeneous information ages among ex ante identical agents. In a stationary Gaussian AR(1) environment, the consequences depend on common staleness as well as the age gap: holding the gap fixed, older information raises individual uncertainty while reducing expected squared forecast disagreement and the value of certifying relative freshness. This generates stale consensus: low disagreement despite high uncertainty. With multiple ordered ages, costly certification supports partial-unraveling equilibria in which fresher types certify and sufficiently stale types pool. Under geometric information ages, more frequent updating weakly lowers the smallest and largest equilibrium disclosure cutoffs wherever positive-certification equilibria exist. The framework has implications for credit, insurance, financial markets, and multi-agent AI.
Comments30 pages
