Macroeconomics from the Bottom Up via Very Large Scale Multi-Agent Systems

通过超大规模多代理系统自下而上的宏观经济学

• 批准号: 0527608
• 负责人: Robert Axtell
• 金额: $ 68.63万
• 依托单位: Brookings Institution
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2006
• 资助国家: 美国
• 起止时间: 2006-01-01 至 2007-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0527608&HistoricalAwards=false
• 关键词: Macroeconomics; Bottom; Up; via; Very

This research concerns the creation of a new kind of macroeconomic model. Conventional macroeconomics uses mathematics to represent how aggregate economic variables, like overall savings and investment, jointly determine one another. Such models are highly aggregate, as if the overall economy consisted of one large firm selling to a single consumer who is also an employee. Due to recent advances in computer technology, the macroeconomy can be represented much more realistically, with all the heterogeneity that is present in the real world also present in models of it.In the proposed research, computational models will be built consisting of large numbers (millions) of software objects known as agents, with each object representing a person. Over time these agents will interact economically, buying and selling artificial goods, consuming and saving, investing and trading. Using such models, aggregate economic relationships will emerge in cross-section as well as over time. The overall goal is to end up with a behaviorally rich model of an economy 'under glass,' so to speak-an artificial economy-that can be calibrated to real economies of today. If successful, this research will move macroeconomics away from highly stylized models and empirically unrealistic behavioral specifications, to models that are more believable. Ultimately, the goal of such efforts is to produce macroeconomic models that have high performance at predicting the near term unfolding of an economy, thus providing more accurate forecasts of economic evolution. These models could aid policy-makers in their decision-making, which in turn could translate into potentially very large increases in the performance of the economy.

本研究涉及一种新的宏观经济模型的创建。 传统的宏观经济学使用数学来表示总的经济变量，如总储蓄和投资，如何共同决定彼此。这些模型是高度聚合的，就好像整个经济由一个大公司向一个同时也是雇员的消费者销售产品组成。由于计算机技术的最新进展，宏观经济可以被更真实地表示，所有的异质性，这是目前在真实的世界也目前在它的模型。在拟议的研究，计算模型将建立包括大量（数百万）的软件对象被称为代理，每个对象代表一个人。随着时间的推移，这些代理人将在经济上相互作用，购买和销售人造商品，消费和储蓄，投资和交易。使用这些模型，总的经济关系将出现在横截面以及随着时间的推移。总的目标是最终得到一个行为丰富的“玻璃下”经济模型，可以这么说，一个人造经济，可以校准到今天的真实的经济。如果成功，这项研究将使宏观经济学从高度程式化的模型和经验上不切实际的行为规范转向更可信的模型。最终，这些努力的目标是产生宏观经济模型，这些模型在预测经济的近期发展方面具有高性能，从而提供更准确的经济演变预测。 这些模型可以帮助决策者进行决策，这反过来又可能转化为经济业绩的巨大增长。