Income tax evasion dynamics and social interactions are analyzed with an agent-based model in heterogeneous populations. One novelty is the combined analysis of back auditing and ageing, which allows for incorporating psychological findings with respect to social norm updating over a taxpayer’s life cycle. Another novelty concerns individual’s social behavior regarding a Pareto-optimal provision of public goods. Simulation results support the counterintuitive conclusion drawn elsewhere in the literature that income tax compliance may decrease with raising marginal per capita returns (MPCRs). Yet, back auditing seems to have by far the strongest impact on the dynamics of fiscal fraud and also can help to curb the extent of tax evasion (ETE).
Following the lead of Mittone and Patelli (2000), agent-based tax evasion models are gaining more and more popularity. Bloomquist, 2006 and Pickhardt and Seibold, 2011 provide an overview of recent developments, while Garson, 2009 and Heath et al., 2009 offer more general surveys of agent-based models. One reason for the ongoing attractiveness of agent-based tax evasion frameworks may be the fact that these settings give researchers a higher degree of control than available with human subjects in lab experiments. Another distinguishing feature of agent-based modeling is the opportunity to include rather complex real world situations, for instance, various social interaction procedures in a society of behaviorally heterogeneous agents; aspects which the standard neoclassical approach to tax evasion cannot handle.
In this paper heterogeneous taxpayer agents have to file a tax return in an environment which consists of a government, a tax authority and individual social networks of acquaintances. Income tax evasion dynamics are then investigated according to five aspects, that are, (i) lapse of time effects (or back auditing), (ii) age heterogeneity of agents, (iii) social norm evolution due to age effects, (iv) provision of pure public goods, and (v) behavioral updating with respect to Pareto-optimal provision of pure public goods. Among other things, the agent-based framework permits to examine various aspects of conditional cooperation and social behavior patterns that may be linked to contributions of many other scholars, e.g. Traxler, 2010, Kube and Traxler, 2011, Traxler and Spichtig, 2011 and Traxler and Winter, 2012. Findings include support for the counterintuitive result of Falkinger, 1988, Falkinger, 1995 and Cowell, 1992 that increasing the marginal per capita return (MPCR) may lead to more tax evasion.
Moreover, it is worth emphasizing that even though agent-based tax compliance models allow for a large variety of simulation scenarios, one may observe at least two open issues. First, the relevance of Pareto-optimal allocations has not yet been analyzed with an agent-based computational setting taking into account tax evasion and public goods provision.1 Second, there is a persistent lack of docking and replication studies that verify and validate simulation results in this field of research.2 Therefore, the purpose of this paper is to analyze income tax evasion dynamics within an agent-based model, which addresses both issues by extending and replicating an existing agent-based tax compliance framework.
The paper is organized as follows. The next section presents a mathematical background to augment the neoclassical model of Allingham and Sandmo (1972) by pure public goods provision. The third section describes the agent-based computational model. The fourth section provides results of various simulation scenarios. The final section concludes.
In this paper I have replicated and extended various features of agent-based tax evasion models with a view to analyze interaction dynamics that may result. In particular, I have implemented four extensions, (i) age heterogeneity, (ii) norm updating due to an age effect, (iii) provision of a pure public good and (iv) behavioral updating with respect to the Pareto-optimal provision of pure public goods. The dynamic effects of each extension have been analyzed ceteris paribus and jointly.
Findings include that age heterogeneity leads to fluctuations in income tax evasion, if and only if lapse of time effects (back auditing) or social norm updating are considered. Moral age effects, where older agents tend to be more compliant due to the evolution of social norms over their life cycle, may lead to less tax evasion, which is shown by a substantial vertical downward shift of income tax evasion dynamics. It is remarkable that these results confirm outcomes of Nordblom and Žamac (2012) and other scholars who analyze age effects and social interactions. Moreover, social norm updating may be linked to subject’s audit experience (e.g. see Lipatov, 2012, with respect to social interaction dynamics and tax evasion); a feature that was employed to update agent’s subjective audit probability. In addition, instead of fixed magnitudes tax administrative parameters may vary and depend on tax and penalty rates. However, these amendments rather delineate a future research agenda.
Next, findings with respect to interaction dynamics between the government as a provider of public goods and consumers, i.e. taxpayer agents, support the counterintuitive conclusion of Falkinger, 1988, Falkinger, 1995 and Cowell, 1992 that the provision of public goods leads to more tax evasion, provided that (voluntary) marginal per capita returns exceed a certain threshold. A precise estimation or calculation of turning points is left for future research. Another research task of interest may be an analysis of tax evasion dynamics due to individualizing the effectiveness of the public sector. Nonetheless, it is worth noting that the technical probability for a Pareto-optimal provision of public goods was rather small within the agent-based tax compliance simulation.
To my best knowledge, for the first time the Pareto-optimality concept is incorporated within an agent-based model and it is shown that, in fact, Pareto-optimal allocations of public goods do definitely emerge. Finally, back auditing and/or social norm updating not only dominate income tax compliance dynamics in any simulation scenario, but may also change findings that occur in scenarios in which back auditing and/or social norm updating is not applied. Therefore, future research on tax evasion dynamics should always incorporate lapse of time effects and social behavior patterns.
