Learning, Credibility and Asset Markets

这个项目的目标是从一般的理论上，以及在具体的和经验相关的经济问题上理解资产市场均衡。在之前的拨款下，编写了计算混合了私人和社会风险的模型中均衡的算法。在本产品中，它们被用来研究一类新的具有有限保险可能性的模型：债务约束资产市场。在这个模型中，流动性约束来自这样一种假设，即不可能迫使借款人偿还或用他们未来的收入偿还债务。交易员可能被拒绝授信，他们的资产可能被扣押以偿还过去的债务。这合理地反映了资产市场的实际运作方式。债务人监狱不复存在，事实上，现代破产法使交易员即使在面临破产法的情况下也有可能保留一些资产。债务受限的市场可以解释为什么保险市场在现实世界的经济体中是不完整的。对资产分配模型的动态研究仍在继续。这个项目放宽了真实经济周期模型中的代表性经济代理人的假设。超级计算机上的数值模拟被用来确定具有私人和社会风险混合的资产分布均衡集的全局图景。该项目开发并测试了准确逼近这些资产分布的方法。将上一个项目中研究声誉和重复博弈中的学习的方法扩展用于研究资产市场中的学习和近似知识。这一点很重要，因为它允许对无法解释的资产市场行为进行严格的分析。大多数目前的理论要么假设常识，要么假设行为如此愚蠢，以至于很难想象一个天真的特工会这样做。总而言之，这项研究开发了强大的新方法工具，用于研究异类代理的市场动力学，并严格建模代理犯错误但从错误中学习的市场。这些工具被用来研究资产市场的现实模型，在这些模型中，某些风险不能投保，或者可以以令人望而却步的费率投保。

The goal of this project is to understand, both in a general theoretical way, and in concrete and empirically relevant economic problems, asset market equilibrium. Algorithms for computing equilibria in models with a mixture of private and social risk were programmed under the previous grant. They are used in this product to study a new class of models with limited insurance possibilities: debt constrained asset markets. Liquidity constraints emerge in this model from the assumption that it is not possible to force borrowers to service or pay off debts from their future income. Traders may be denied credit, and their assets may be attached for the payment of past debts. This is a reasonable representation of the way in which actual asset markets work. Debtor prison no longer exists, and indeed, modern bankruptcy laws make it possible for traders to preserve some assets even in the face of bankruptcy laws. Debt constrained markets could explain why insurance markets are incomplete in real world economies. Research is continued on the dynamics of asset distribution models. This project relaxes the assumption of a representative economic agent made in real business cycle models. Numerical simulation on the supercomputer is used to determine the global picture of the set of equilibria of asset distributions with a mix of private and social risk. The project develops and tests methods of accurately approximating these asset distribution. Methods developed in the previous project to study reputation and learning in repeated games are extended and used to study learning and approximate knowledge in asset markets. This is important because it permits rigorous analysis of asset market behavior that can't otherwise be explained. Most current theories either assume common knowledge or assume behavior that is so stupid that it is hard to imagine even a naive agent behaving this way. In sum, this research develops powerful new methodological tools for studying market dynamics with heterogeneous agents and for rigorously modelling markets in which agents make mistakes but learn from their mistakes. These tools are used to study realistic models of asset markets in which certain risks can not be insured or can be insured at prohibitive rates.