System Structure of Agent-Based Model Responsible for Reproducing Business Cycles and the Effect of Tax Reduction on GDP
条件 , 那么 “ 复制每种宏观行为 ” 的必要条件是有可能被定 义的 。 通过一系列计算机实验来验证该想法 , 本研究分析了 商业周期 、 和减税对国内生产总值 ( GDP ) 产生的效果 , 将二者作为经济体系的重要宏观行为实例 。 研究结果表明 , 复制商业周期和减税效果 , 必需的模型架构包括投资信用产 生 、 和分别与政府开支低效 、 家庭开支低效 、 企业开支低效 有关的各因素 。
关键词 : 基于 agent 建模 ; 模型架构 ; 商业周期 ; 减税 ; 低效 政府开支 ; 模型有效性
1 . Introduction
Agent-based modeling ( ABM ) is a bottom-up modeling method in which we view artificial , computer-generated societies as laboratories where we attempt to grow specific social structures ( Epstein & Axtell , 1996 ). The purpose of these models is to discover the fundamental local or micro-mechanisms that generate macroscopic social structures and collective behaviors ( Epstein & Axtell , 1996 ). Although ABM is a promising methodology that can deal with heterogeneity , individual agents ’ bounded rationality , and non-equilibrium dynamics in social systems , validation still proves to be a significant issue . As pointed out in the literature ( Ormerod & Rosewell , 2009 ), one common criticism by economists could be stated as follows , “ you have presented one set of behavioral rules to explain your chosen phenomenon , but there must be many such sets which produce the same result , so how do you know yours is correct ?” ( p . 10 ). Some economists even go so far as to imply that it is excessively easy to construct an agent-based model that produces desired phenomena . As argued by Marks ( 2007 ), the problem behind this criticism is the functional complexity inherent in ABM . It has also been argued that macro behaviors may be insensitive to many micro variables ; and , as a result , it would be difficult to derive the necessary conditions for the model to exhibit specific macro behaviors ( Marks , 2007 ). The severity of this problem increases when the model is described with detail and realism , as this requires more variables and higher degrees of freedom ( Marks , 2007 ). For this reason , the model should be as simple as possible , and even then , it would be difficult to achieve quantitative predictions .
When input conditions are expressed by specific values of micro variables or parameters , there is a great deal of freedom , as pointed out in the
39