IBSS: Agent-Based Model of the Role of Perceptions in Income Tax Evasion

这一跨学科研究项目将通过关注风险和税收公平的认知是如何形成的，这些认知是如何在社会网络中传播的，以及个人行为和总体水平的结果之间复杂的反馈循环所造成的相互作用，来推进对税收遵从行为如何产生的基本理解。该项目还将探讨审计、处罚和税收政策以及这些政策的变化如何影响人们的看法，并最终减少逃税。该项目将提供关于纳税合规行为如何被感知的聚合合规性改变，以及行为和感知的聚合合规性如何依赖于税收感知传播的社交网络结构的新见解。该项目将展示使用基于代理的建模来研究个人与政府之间的互动，以及如何利用兰德公司从美国生活小组收集的数据来解决这样的问题。这项分析还将具有实用价值，因为它将阐明为什么某些高税收水平的国家，如斯堪的纳维亚国家，可以保持高税收合规水平，而其他税收水平较低的国家可能像希腊和意大利那样存在高水平逃税。因此，项目结果应该有助于像美国这样的国家减少逃税发生。所得税逃税是一个给地方、州和联邦各级税务机关和政府带来相当大挑战的问题。其原因和影响既有经济上的，也有社会上的。然而，与逃税相关的认知的作用，以及社交网络如何影响这些认知，人们仍然知之甚少。调查人员将建立一个基于代理人的所得税逃税计算模拟模型。在模拟中，个人的合规行为将通过一个适应过程发生变化，该过程基于他们过去在审计和逃税处罚方面的经验、他们对税率公平的看法，以及他们与社交网络中的人的社交互动。同时，调查人员还将在全国范围内开展税收公平感调查。调查将收集个人层面的数据，这些数据将为模拟模型的行为机制提供信息。这些机制影响了规避倾向，调查为选择特定于模型的参数值提供了经验依据。将使用敏感性分析进一步了解模型假设。模拟模型将经过验证和校准，以再现美国不同税级的全国所得税合规水平。除了美国，该模型还将使用希腊和丹麦的数据集和合规水平进行校准。一旦校准，该模型将被用来理解逃税行为如何根据各种初始假设而不同地演变，例如社交网络是如何构建的，或者什么财政政策到位，以及系统产生临界点动态的条件。然后，该模型将被用来确定在最大限度地减少逃税和恢复合规方面最有效的财政政策。调查人员将采用稳健决策，这是一种改进政策决策的方法，根据政策的潜在表现以及这种表现对关键不确定性来源的稳健程度对政策进行排名。该项目通过NSF跨学科行为和社会科学研究(IBSS)竞赛得到支持。

This interdisciplinary research project will advance basic understanding of how tax compliance behavior emerges by focusing on how perceptions of risk and tax fairness form, how these perceptions spread through a social network, and the interplay caused by the complex feedback loops among individual behaviors and population-level outcomes. The project also will address how audit, penalty, and taxation policies, and changes in those policies, influence perceptions and ultimately reduce tax evasion. The project will provide new insights regarding how tax compliance behavior is modified by the perceived aggregated compliance and on how behavior and the perceived aggregated compliance depend on social network structure over which perceptions of taxation spreads. The project will demonstrate the utility of the use of agent-based modeling for research on interactions between individuals and governments as well as how data from the American Life Panel collected by Rand Corporation can be employed to address problems like this. The analysis also will have practical utility, because it will shed light on why certain countries with high taxation levels, such as Scandinavian nations, can maintain high levels of tax compliance while other countries with lower taxation levels may have high levels of tax evasion as is the case in Greece and Italy. Project findings therefore should help nations like the United States reduce the occurrence of tax evasion. Income tax evasion is a problem that poses considerable challenges for tax authorities and governments at the local, state, and federal levels. Its causes and implications are both economic and social. The role of tax evasion-related perceptions and how social networks influence those perceptions remain poorly understood, however. The investigators will build an agent-based computational simulation model of income tax evasion. Within the simulation, the compliance behavior of individuals will change through an adaptation process based on their past experiences with audits and tax evasion penalties, their perception of the fairness in taxation rates, and social interactions with people in their social networks. In conjunction, the investigators will conduct a national survey on the perceptions of tax fairness. The survey will gather individual-level data that will inform the simulation model's behavioral mechanisms. These mechanisms influence the propensity to evade, and the survey provides an empirical basis for choosing model-specific parameter values. Model assumptions will further be informed using sensitivity analyses. The simulation model will be validated and calibrated to reproduce U.S. national levels of income tax compliance for different tax brackets. In addition to the U.S., the model also will be calibrated using data sets and compliance levels for Greece and Denmark. Once calibrated, the model will be used to understand how tax evasion behaviors evolve differently depending on various starting assumptions, such as how social networks are structured or what fiscal policies are in place and the conditions for the system to produce tipping point dynamics. The model then will be used to identify fiscal policies that are most effective in minimize tax evasion and recovering compliance. The investigators will employ robust decision making, a method for improving policy decision making, to rank policies based both on their potential performance, as well as how robust this performance is to the key sources of uncertainty. This project is supported through the NSF Interdisciplinary Behavioral and Social Sciences Research (IBSS) competition.