Agent-based Modeling of Tax Evasion E-Book
75,99 €
Mehr erfahren.
- Herausgeber: John Wiley & Sons
- Kategorie: Wissenschaft und neue Technologien
- Serie: Wiley Series in Computational and Quantitative Social Science
- Sprache: Englisch
The only single-source guide to understanding, using, adapting, and designing state-of-the-art agent-based modelling of tax evasion
A computational method for simulating the behavior of individuals or groups and their effects on an entire system, agent-based modeling has proven itself to be a powerful new tool for detecting tax fraud. While interdisciplinary groups and individuals working in the tax domain have published numerous articles in diverse peer-reviewed journals and have presented their findings at international conferences, until Agent-based Modelling of Tax Evasion there was no authoritative, single-source guide to state-of-the-art agent-based tax evasion modeling techniques and technologies.
Featuring contributions from distinguished experts in the field from around the globe, Agent-Based Modelling of Tax Evasion provides in-depth coverage of an array of field tested agent-based tax evasion models. Models are presented in a unified format so as to enable readers to systematically work their way through the various modeling alternatives available to them. Three main components of each agent-based model are explored in accordance with the Overview, Design Concepts, and Details (ODD) protocol, each section of which contains several sub elements that help to illustrate the model clearly and that assist readers in replicating the modeling results described.
- Presents models in a unified and structured manner to provide a point of reference for readers interested in agent-based modelling of tax evasion
- Explores the theoretical aspects and diversity of agent-based modeling through the example of tax evasion
- Provides an overview of the characteristics of more than thirty agent-based tax evasion frameworks
- Functions as a solid foundation for lectures and seminars on agent-based modelling of tax evasion
The only comprehensive treatment of agent-based tax evasion models and their applications, this book is an indispensable working resource for practitioners and tax evasion modelers both in the agent-based computational domain and using other methodologies. It is also an excellent pedagogical resource for teaching tax evasion modeling and/or agent-based modeling generally.
Sie lesen das E-Book in den Legimi-Apps auf:
Seitenzahl: 684
Veröffentlichungsjahr: 2018
Ähnliche
Table of Contents
Cover
Title Page
Copyright
Dedication
Notes on Contributors
Foreword
Preface
Part I: Introduction
Chapter 1: Agent-Based Modeling and Tax Evasion: Theory and Application
1.1 Introduction
1.2 Tax Evasion, Tax Avoidance and Tax Noncompliance
1.3 Standard Theories of Tax Evasion
1.4 Agent-Based Models
1.5 Standard Protocols to Describe Agent-Based Models
1.6 Literature Review of Agent-Based Tax Evasion Models
1.7 Outlook: The Structure and Presentation of the Book
References
Chapter 2: How Should One Study Clandestine Activities: Crimes, Tax Fraud, and Other “Dark” Economic Behavior?
2.1 Introduction
2.2 Why Study Clandestine Behavior At All?
2.3 Tools for Studying Clandestine Activities
2.4 Networks and the Complexity of Clandestine Interactions
2.5 Layers of Analysis
2.6 Research Tools and Clandestine Activities
2.7 Conclusion
Acknowledgment
References
Chapter 3: Taxpayer's Behavior: From the Laboratory to Agent-Based Simulations
3.1 Tax Compliance: Theory and Evidence
3.2 Research on Tax Compliance: A Methodological Analysis
3.3 From Human-Subject to Computational-Agent Experiments
3.4 An Agent-Based Approach to Taxpayers' Behavior
3.5 Conclusions
References
Part II: Agent-Based Tax Evasion Models
Chapter 4: Using Agent-Based Modeling to Analyze Tax Compliance and Auditing
4.1 Introduction
4.2 Agent-Based Model for Tax Compliance and Audit Research
4.3 Modeling Individual Compliance
4.4 Risk-Taking and Income Distribution
4.5 Attitudes, Beliefs, and Network Effects
4.6 Equilibrium with Random and Targeted Audits
4.7 Conclusions
Acknowledgments
References
Appendix 4A
Chapter 5: SIMULFIS: A Simulation Tool to Explore Tax Compliance Behavior
5.1 Introduction
5.2 Model Description
5.3 Some Experimental Results and Conclusions
Acknowledgments
References
Chapter 6: TAXSIM: A Generative Model to Study the Emerging Levels of Tax Compliance
6.1 Introduction
6.2 Model Description
6.3 Results
6.4 Conclusions
Acknowledgments
References
Chapter 7: Development and Calibration of a Large-Scale Agent-Based Model of Individual Tax Reporting Compliance
7.1 Introduction
7.2 Model Validation and Calibration
7.3 Hypothetical Simulation: Size of the “Gig” Economy and Taxpayer Compliance
7.4 Conclusion and Future Research
Acknowledgments
References
7A.1 Purpose
7A.2 Entities, State Variables, and Scales
7A.3 Process Overview and Scheduling
7A.4 Design Concepts
7A.5 Initialization
7A.6 Input Data
7A.7 Submodels
Chapter 8: Investigating the Effects of Network Structures in Massive Agent-Based Models of Tax Evasion
8.1 Introduction
8.2 Networks and Scale
8.3 The Model
8.4 The Experiment
8.5 Results
8.6 Conclusion
References
Chapter 9: Agent-Based Simulations of Tax Evasion: Dynamics by Lapse of Time, Social Norms, Age Heterogeneity, Subjective Audit Probability, Public Goods Provision, and Pareto-Optimality
9.1 Introduction
9.2 The Agent-Based Tax Evasion Model
9.3 Scenarios, Simulation Results, and Discussion
9.4 Conclusions and Outlook
Acknowledgments
References
Appendix 9A
Chapter 10: Modeling the Co-evolution of Tax Shelters and Audit Priorities*
10.1 Introduction
10.2 Overview
10.3 Design Concepts
10.4 Details
10.5 Experiments
10.6 Discussion
References
Chapter 11: From Spins to Agents: An Econophysics Approach to Tax Evasion
11.1 Introduction
11.2 The Ising Model
11.3 Application to Tax Evasion
11.4 Heterogeneous Agents
11.5 Relation to Binary Choice Model
11.6 Summary and Outlook
References
Index
End User License Agreement
Pages
1
3
xiii
xiv
xv
xvi
xvii
xviii
xix
xxi
xxii
xxiii
xxiv
xxv
xxvii
xxviii
1
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
Guide
Cover
Table of Contents
Foreword
Preface
Part I: Introduction
Begin Reading
List of Illustrations
Chapter 2: How Should One Study Clandestine Activities: Crimes, Tax Fraud, and Other “Dark” Economic Behavior?
Figure 2.1 Scientific goals of analysis. Source: Own depiction. See for description, explanation, and prediction Bunge (1967b, Part III, Chapters 9 and 10)
Figure 2.2 Input, throughput, and output of clandestine activities. Source: Own depiction
Figure 2.3 Levels of activity and levels of analysis. Source: Own depiction
Figure 2.4 Disregarded abstractness versus misplaced concreteness. Source: Own depiction
Figure 2.5 Difficulties in countering clandestine activities: the complexity of concreteness. Source: Own depiction
Chapter 3: Taxpayer's Behavior: From the Laboratory to Agent-Based Simulations
Figure 3.1 Experimental systems as mediators between theoretical models and economic phenomena
Figure 3.2 ACE simulations and experiments
Figure 3.3 System structure diagram representing the system by Mittone and Patelli (2000)
Figure 3.4 System structure diagram representing the system by Mittone and Jesi (2016)
Chapter 4: Using Agent-Based Modeling to Analyze Tax Compliance and Auditing
Figure 4.1 Social equilibrium
Figure 4.2 Occupational choice with honest tax payment
Figure 4.3 Occupational choice with noncompliance
Figure 4.4 Lorenz curves for honesty (solid) and noncompliance (dashed)
Figure 4.5 Histogram of effective tax rates
Figure 4.6 Cumulative distribution functions for tax and fine revenues (random: solid; targeted: dashed)
Figure 4.7 Compliance levels (occupation 1: solid; occupation 2: dashed
Figure 4.8 Proportion of audited in occupation by percentile of declaration
Figure 4.9 Proportion of audited in occupation by declaration percentile
Figure 4.10 Probability of audit and compliance levels
Chapter 5: SIMULFIS: A Simulation Tool to Explore Tax Compliance Behavior
Figure 5.1 An example of the decision algorithm
Chapter 6: TAXSIM: A Generative Model to Study the Emerging Levels of Tax Compliance
Figure 6.1 (a) Impact of the persistent improvement in the quality of governmental services (improvement at every 500th time steps). (b) Introduction of companies paying preferential taxes (from the 500th time step). In both panels, the horizontal axis shows simulated time, while the vertical one is the number of a) Employed b) Legal employments c) Mixed employments c) Hidden employments. The symbols mark the average and error bars mark the standard deviation of 256 runs.
Figure 6.2 Aggregated tax income. (a) Shows the base scenario (reaching equilibrium in about 350 time steps), while (b) displays the impact of the persistent improvement in the quality of governmental services (improvement at every 500th time steps). The symbols mark the average and error bars mark the standard deviation of 256 runs.
Figure 6.3 Top eight factors with the largest effects in case of the main response variables. Results from a two-level factorial analysis (only two-way interactions shown). The values for the response variables are snapshots after 6000 time steps and averaged over 10 runs. The abbreviation govt stands for government.
Figure 6.4 Percentage of nonlegal contracts (hidden and mixed employments combined), relative to the total level of employment, in case of various combinations of the audit probability and accuracy of audits. The values shown are snapshots taken after 6000 time steps and averaged over 10 independent runs. The rest of the parameters are fixed according to Tables 6.5 and 6.6. (a) Pictures the original market of Section 6.3.1, while (b) shows the case of the competitive labor market of this section. (Note the different horizontal scales.)
Figure 6.5 Transition from a hidden to a legal economy as a function of
audit probability
. Cross sections of the phase space shown in Figure 6.4
Figure 6.6 Transition from a hidden to a legal economy as a function of
audit accuracy
. Cross sections of the phase space shown in Figure 6.4
Figure 6.7 Percentage of aggregate tax income (top right panel) and employment types as a function of the quality of governmental services to employees (horizontal axes) and employers (vertical axes). The percentage is of the maximum observed value in case of tax income and of the total number of employment contracts for employment types. The values shown are from snapshots taken after 6000 time steps and averaged over 10 runs. The rest of the parameters are fixed according to Tables 6.5 and 6.6 with
audit probability
at 25% and
audit accuracy
at 0.45
Figure 6.8 The effect of an adaptive audit strategy. The number of hidden (a), mixed (b), and legal (c) employment contracts for various ratios of random and adaptive audit strategies (horizontal axes: 0 stands for all-adaptive selection, and 1 for all-random selection). The panels in the different rows depict results for different levels of
audit frequency
and
audit accuracy
combinations. The values plotted are snapshots after 6000 time steps and are averaged over 10 independent runs
Figure 6.9 The effect of minimum wage policies. The panel columns represent a minimum tax content of 0%, 16%, 30%, and 44% from left to right. The first row of panels plots the percentage of illegal employment (hidden and mixed contracts combined), while middle and bottom rows separate the two types. The horizontal axes show the value of
audit frequency
, while the vertical axes plot
accuracy
. The grayscale-coded values shown are snapshots after 6000 time steps and show averaged values of 10 independent runs. The rest of the parameters are fixed according to their levels in Tables 6.5 and 6.8
Chapter 7: Development and Calibration of a Large-Scale Agent-Based Model of Individual Tax Reporting Compliance
Figure 7.1 IRCM agent hierarchy
Figure 7.2 Interaction between filer and tax agency
Figure 7.3 Tax year 2006 individual income tax underreporting gap
Figure 7.4 IRCM information reporting parameters screen
Figure 7.5 IRCM filer parameters screen
Figure 7.6 POMDP of the filer's response to the tax audit environment
Figure 7.7 IRCM execution sequence: top-level view
Figure 7.8 Model time series of tax NMPs for alternative increases in share of gig economy workers
Chapter 8: Investigating the Effects of Network Structures in Massive Agent-Based Models of Tax Evasion
Figure 8.1 Closeness centrality as a function of network size
Figure 8.2 Sample network structures used in the simulation
Figure 8.3 Degree centrality distributions for the networks, labels are: (a) random network, (b) ring network, (c) small world network, (d) von Neumann network, (e) power law network, (f) preferential attachment network
Figure 8.4 Closeness centrality distributions for the networks, labels are: (a) random network, (b) ring network, (c) small world network, (d) von Neumann network, (e) power law network, (f) preferential attachment network
Figure 8.5 End of run distributions of VMTR by network type and size
Figure 8.6 End of run distributions of VMTR by network size and type
Figure 8.7 Mean ratio of declared income to actual income over time
Figure 8.8 Variance in the ratio of declared income to actual income over time
Figure 8.9 Simulation run time (in log scale) by scale, for all runs of the simulation
Figure 8.10 Simulation run time (in log scale) by number of processors, for all runs of the simulation
Chapter 9: Agent-Based Simulations of Tax Evasion: Dynamics by Lapse of Time, Social Norms, Age Heterogeneity, Subjective Audit Probability, Public Goods Provision, and Pareto-Optimality
Figure 9.1 Scenario inspired by Hokamp (2014): age heterogeneity (AH) and social norm updating (SNU); lapse of time
LOT
= 0; (a) behaviorally heterogeneous society, (b) neoclassical A-Types, (c) social interacting B-Types.
Chapter 10: Modeling the Co-evolution of Tax Shelters and Audit Priorities*
Figure 10.1 Design framework for our genetic algorithm. The framework for the evolution of audit scorecards is entirely analogous.
Figure 10.2 Example of ownership network and transactions between partners. (a) Example of a network of interconnected entities. The number in parenthesis denotes each owner's basis (usually just the purchase price of their original contribution). P1 corresponds to the partnership entity created by the partners Alice, Bob, and Cathy. (b) Example of a transaction between entities. The dotted line denotes the transfer of assets. Emma purchases Alice's 50% share of P1 in exchange for $50, and the outside basis of the share is increased to $50
Figure 10.3 STEALTH tax ecosystem simulator.
Figure 10.4 Concurrent optimization of high likelihood audit scores and low-risk task schemes.
Figure 10.5 The steps in the IBOB abusive tax avoidance scheme. The basis of an asset is artificially stepped up and tax is avoided by using “pass-through” entities. IBOB step 1.
Figure 10.6 The steps in the IBOB abusive tax avoidance scheme. The basis of an asset is artificially stepped up and tax is avoided by using “pass-through” entities. (a) IBOB step 2, (b) IBOB step 3.
Figure 10.7 Example of mapping a list of integers (genotype) into a list of transactions (phenotype) by using grammatical evolution.
Figure 10.8 Design framework for our genetic algorithm. The framework for the evolution of audit scorecards is entirely analogous.
Figure 10.9 Evolution of IBOB in STEALTH experiment 1 (LimitedAudit). (a) Best fitness for one run (b) Distribution of audit points for one run (c) Distribution of audit points averaged over runs.
Figure 10.10 Evolution of IBOB in STEALTH experiment 2 (EffectiveAudit). (a) Best fitness for one run (b) Distribution of audit points for one run (c) Distribution of audit points averaged over runs.
Figure 10.11 Evolution of IBOB in STEALTH experiment 3 (CoEvolution).
Chapter 11: From Spins to Agents: An Econophysics Approach to Tax Evasion
Figure 11.1 Ground state spin configurations of the Ising model on a square lattice with nearest-neighbor interaction . At temperature Eq. (11.1) yields two degenerate states where either all spins point in the “upward” (a) or “downward” (b) direction. Some of the nearest neighbor interactions are indicated by dashed curvy lines. In addition, one often implements periodic boundary conditions where also sites of the left and right (upper and lower) boundary are connected by , for example, sites and or sites and
Figure 11.2 Main panel: The range of reported income is divided into bins and the height of each bin corresponds to the frequency of declarations within the corresponding -interval. The Ising distribution (open bins) is obtained by assigning all declarations with to and all declarations with to . Inset: Upon setting (indicated by the dashed line) the Ising data give an excellent approximation to the average reported income as obtained from the experimental data (Bazart and Bonein, 2014) for all periods of the experiment.
Figure 11.3 Social network of a tax payer on lattice site for a square lattice with nearest-neighbor interactions. Shown are the situations of a honest (, (a)) and cheating (, (b)) agent, respectively, connected to noncompliant neighbors
Figure 11.4 Tax evasion dynamics on a square lattice with nearest-neighbor interactions . (a, b) Results for agents with a social temperature of and audit probabilities and , respectively. (c, d) The corresponding results for a social temperature of . The scale of temperature is set by . Tax audits enforce compliance over time steps. Initial condition is full compliance for all agents () except for (a) where the dashed line corresponds to an initial condition with full evasion ()
Figure 11.5 Ising model with heterogeneous agent types. Each site is coupled to a heat bath with temperature as indicated by boxes. Moreover, A-Type agents are coupled to a
negative
magnetic field (e.g., site ) whereas C-Type agents are coupled to a
positive
magnetic field (e.g., site ). Nearest neighbors are coupled by the exchange constant
Figure 11.6 Field-temperature parameter space of probabilities for the transition of an agent from compliance to noncompliance. (a) All nearest neighbors are compliant (), (b) All nearest neighbors are noncompliant () as illustrated in Figure 11.3. -field values are in units of the exchange constant whereas temperature values are given in units of
Figure 11.7 Tax evasion dynamics for the multi-agent Ising model. In each panel the percentage of two agent-types is fixed to as indicated in the heading. The fraction of the other two agent-types is then varied in steps of . System size: square lattice with nearest-neighbor coupling . Enforced compliance period after audit: time steps. Agents are specified by the following parameters: A-Types (, ), B-Types (, ), C-Types (, ), D-Types (, ). Temperature values are in units of and field values in units of
List of Tables
Chapter 1: Agent-Based Modeling and Tax Evasion: Theory and Application
Table 1.1 Mathematical syntax for standard theories of tax evasion
Table 1.2 Overview of agent-based tax evasion settings, research groups A–C
Table 1.4 Overview of agent-based tax evasion settings, research groups MB–Z. See Table 1.2 for a brief description
Table 1.3 Overview of agent-based tax evasion settings, research groups D–MA
Chapter 2: How Should One Study Clandestine Activities: Crimes, Tax Fraud, and Other “Dark” Economic Behavior?
Table 2.1 Research methods and their application in the study of clandestine activities (CLAs)
Chapter 4: Using Agent-Based Modeling to Analyze Tax Compliance and Auditing
Table 4.1 Income distribution in two scenarios: when all taxpayers are honest and when noncompliance is possible
Table 4.2 Tobit regression
Chapter 5: SIMULFIS: A Simulation Tool to Explore Tax Compliance Behavior
Table 5.1 Parameters: no change = Greek
Table 5.2 Parameters: functions = roman
Table 5.3 Agent subtypes (occupational status by income level) and homofily network building criteria
Chapter 6: TAXSIM: A Generative Model to Study the Emerging Levels of Tax Compliance
Table 6.1 The 23 employment types considered in TAXSIM
Table 6.2 Parameters of TAXSIM (part 1)
Table 6.3 Parameters of TAXSIM (part 2)
Table 6.4 Parameters used in the scenario experiments
Table 6.5 Parameters of the two-factorial experiments
Table 6.6 Parameter values for the analysis of the main parameters
Table 6.7 Parameter values for the adaptive audit strategy experiments
Table 6.8 Parameter values for the minimum wage experiments
Chapter 7: Development and Calibration of a Large-Scale Agent-Based Model of Individual Tax Reporting Compliance
Table 7.1 Comparison of actual versus artificial taxpayer data for study region
Table 7.2 Line item net misreporting percentages: IRS versus IRCM
Chapter 8: Investigating the Effects of Network Structures in Massive Agent-Based Models of Tax Evasion
Table 8.1 The design of experiments that was executed
Chapter 9: Agent-Based Simulations of Tax Evasion: Dynamics by Lapse of Time, Social Norms, Age Heterogeneity, Subjective Audit Probability, Public Goods Provision, and Pareto-Optimality
Table 9.1 Model parameters and mathematical symbols
Table 9.2 Social norm updating. Risk aversion changes with respect to age
Table 9.4 Calibration and sensitivity analysis: average extent of tax evasion
Table 9.3 Overview: specification of agent-based simulations
Table 9A.1 Political cycle
Table 9A.2 Public goods provision cycle
Chapter 10: Modeling the Co-evolution of Tax Shelters and Audit Priorities*
Table 10.1 Example of an audit score sheet with three individually observable events, resulting in seven total rows and corresponding audit points
Table 10.2 Initial average distribution of audit points for each experiment
Table 10.3 Parameters for IBOB experiments
Chapter 11: From Spins to Agents: An Econophysics Approach to Tax Evasion
Table 11.1 Summary of parameter ranges that specify the various agent types in the heterogeneous Ising model
Wiley Series in Computational and Quantitative Social Science
Computational Social Science is an interdisciplinary field undergoing rapid growth due to the availability of ever increasing computational power leading to new areas of research. Embracing a spectrum from theoretical foundations to realworld applications, theWiley Series in Computational and Quantitative Social Science is a series of titles ranging from high level student texts, explanation and dissemination of technology and good practice, through to interesting and important research that is immediately relevant to social / scientific development or practice. Books within the series will be of interest to senior undergraduate and graduate students, researchers and practitioners within statistics and social science.
Behavioral Computational Social Science
Riccardo Boero
Tipping Points: Modeling Social Problems and Health
John Bissell (Editor), Camila Caiado (Editor), Sarah Curtis (Editor), Michael Goldstein (Editor), Brian Straughan (Editor)
Understanding Large Temporal Networks and Spatial Networks: Exploration, Pattern Searching, Visualization and Network Evolution
Vladimir Batagelj, Patrick Doreian, Anuska Ferligoj, Natasa Kejzar
Analytical Sociology: Actions and Networks
Gianluca Manzo (Editor)
Computational Approaches to Studying the Co-evolution of Networks and Behavior in Social Dilemmas
Rense Corten
The Visualisation of Spatial Social Structure
Danny Dorling
AGENT-BASED MODELING OF TAX EVASION
THEORETICAL ASPECTS AND COMPUTATIONAL SIMULATIONS
Edited by
Sascha Hokamp
Universität Hamburg
László Gulyás
Eötvös Loránd University
Matthew Koehler
The MITRE Corporation
Sanith Wijesinghe
The MITRE Corporation
This edition first published 2018
© 2018 John Wiley & Sons Ltd
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by law. Advice on how to obtain permission to reuse material from this title is available at http://www.wiley.com/go/permissions.
The right of Sascha Hokamp, László Gulyás, Matthew Koehler and Sanith Wijesinghe to be identified as the authors of the editorial material in this work has been asserted in accordance with law.
Registered Offices
John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, USA
John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, UK
Editorial Office
9600 Garsington Road, Oxford, OX4 2DQ, UK
For details of our global editorial offices, customer services, and more information about Wiley products visit us at www.wiley.com.
Wiley also publishes its books in a variety of electronic formats and by print-on-demand. Some content that appears in standard print versions of this book may not be available in other formats.
Limit of Liability/Disclaimer of Warranty
While the publisher and authors have used their best efforts in preparing this work, they make no representations or warranties with respect to the accuracy or completeness of the contents of this work and specifically disclaim all warranties, including without limitation any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives, written sales materials or promotional statements for this work. The fact that an organization, website, or product is referred to in this work as a citation and/or potential source of further information does not mean that the publisher and authors endorse the information or services the organization, website, or product may provide or recommendations it may make. This work is sold with the understanding that the publisher is not engaged in rendering professional services. The advice and strategies contained herein may not be suitable for your situation. You should consult with a specialist where appropriate. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this work was written and when it is read. Neither the publisher nor authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages.
Library of Congress Cataloging-in-Publication Data applied for
Hardback ISBN: 9781119155683
Cover image: © KrulUA/Gettyimages
To our families
Notes on Contributors
Amanda Andrei is a social scientist and senior artificial intelligence engineer at The MITRE Corporation specializing in social analytics, development of innovative processes and spaces, and application of mixed methods to sociotechnical problems. She received her BA in Anthropology from the College of William & Mary, her Certificate in Computational Social Science from George Mason University, and her MA in Communication, Culture, and Technology from Georgetown University.
Kim M. Bloomquist is an operations research analyst with the U.S. Internal Revenue Service's Taxpayer Advocate Service. He has authored and coauthored numerous papers and book chapters on the economics of taxpayer compliance and agent-based modeling. His research has been cited in Congressional testimony, the Washington Post, Tax Notes, and Tax Notes International. He has received several awards for his research on tax compliance including the Organization for Economic Cooperation and Development's (OECD) Jan Francke Tax Research Award, the Cedric Sandford medal for the best paper at the seventh International Conference on Tax Administration in Sydney, Australia, and the IRS Research Community Award for Research Technical Expertise. Kim received his PhD (Computational Social Science) from George Mason University, Fairfax, Virginia in 2012.
Kevin Comer is a Modeling and Simulation Engineer at the MITRE Corporation, specializing in agent-based modeling design, development and validation. His primary focus is in simulating the dynamics of the individual health insurance market. He also works on modeling processes in cybersecurity and homeland security. He received his B.S. in Systems Engineering and Economics from the University of Virginia, his M.S. in Operations Research from George Mason University Volgenau School of Engineering, and his Ph.D. in Computational Social Science from George Mason University's Department of Computational Social Science. His email address is [email protected].
Andrés M. Cuervo Díaz completed his BSc in Physics from Universidad de los Andes, Bogota, Colombia, in 2014 in the field of quantum optics and is a master's student in the study program “Integrated Climate System Sciences” at the Cluster of Excellence (DFG EXC 177 CliSAP), participating in the research group Climate Change and Security (CLISEC). He is mainly interested in agent-based and integrated assessment modeling for the evaluation of policies' performance and possible strategies to enhance environmental protection and climate change action.
László Gulyás holds a PhD in Computer Science from Eötvös Loránd University, Hungary. He is an assistant professor at Eötvös Loránd University, Budapest, and held the position of head of division at AITIA International, Inc. He has been doing research on agent-based modeling and multiagent systems since 1996. He has been involved in teaching both graduate and undergraduate level courses in agent-based modeling and simulation at Harvard University, at the Central European University and at the Eötvös Loránd University, Hungary. He has authored several book chapters and journal articles. In particular, László has published on agent-based modeling of tax evasion since 2009.
Nigar Hashimzade obtained her PhD in Economics from Cornell University in 2003. Prior to her current position at Durham she has worked at the University of Exeter and the University of Reading. Her research interests are in applied microeconomic theory and in quantitative methods. Since 2012, she has been involved in research on the behavioral approach to tax evasion funded jointly by the ESRC, HMRC, and HMT. Among other projects, she works on developing behavioral models of tax compliance decisions in social networks and on building agent-based models that can be used to assess and compare tax enforcement policies in a complex environment. Nigar has published in leading international academic journals and contributed to a number of monographs, for some of which she has also been a coeditor. Her previous career was in theoretical physics.
Christine Harvey is a high-performance and analytic computing engineer at The MITRE Corporation in Washington, D.C. She specializes in data analysis and high-performance computing in simulations. In addition, her research interests include agent-based modeling and big data analysis, particularly in the field of Healthcare. She completed her masters in Computational Science from Stockton University in 2013 and is expected to finish her PhD in Computational Science and Informatics at George Mason University in 2018.
Erik Hemberg is a postdoctoral researcher in the ALFA group in CSAIL at the Massachusetts Institute of Technology. He performs research regarding scaleable machine learning. He is currently involved in research regarding tax evasion and physiological time series prediction. He received his PhD in Computer Science from the University College Dublin, Ireland. At ICAIL 2015 he co-authored the Peter Jackson Best Innovative Application paper.
Sascha Hokamp obtained his PhD in Public Economics from the Brandenburg University of Technology Cottbus, Germany, in 2013. He was awarded a stipend from the Deutsche Bundesbank via the “Verein für Socialpolitik” in 2011 and 2012. He is involved in organizing the biannual “Shadow Economy” conference series, founded at the Westfälische Wilhelms-Universität Münster, Germany, in 2009. He served in 2015 as a guest editor in Economics of Governance on “The Shadow Economy, Tax Evasion, and Governance.” Sascha is a member of the Research Unit for Sustainability and Global Change (FNU) and of the Center for Earth System Research and Sustainability (CEN), Universität Hamburg, Germany, and he is participating researcher at the Cluster of Excellence “Integrated Climate System Analysis and Prediction” (DFG EXC 177 CliSAP), on the project “Societal Use of Climate Information.” His research topics are integrated assessment modeling of climate change, agent-based modeling of environmental challenges, and climate policy as well as illicit activities (tax evasion and doping in elite-sports) and the shadow economy.
Matthew Koehler is the Applied Complexity Sciences Area Lead for the U.S. Treasury/Internal Revenue Service, U.S. Commerce, and Social Security Administration Program Division at The MITRE Corporation. He has concentrated on decision support using agent-based models and simulations, and the analysis and visualization of large datasets coming from complex systems. He received his AB in Anthropology from Kenyon College, his MPA from Indiana University's School of Public and Environmental Affairs, his JD from George Washington University's Law School, and his PhD in Computational Social Science from George Mason University's Krasnow Institute for Advanced Study, Department of Computational Social Science.
Toni Llacer holds a PhD in Sociology from the Universitat Autónoma de Barcelona (UAB), Spain. His research field is the interdisciplinary study of tax evasion. He obtained a bachelor's degree in Economics from Universitat Pompeu Fabra, a bachelor's degree in Philosophy from Universitat de Barcelona (Academic Excellence Award), and a master of science in Applied Social Research from UAB.
Tamás Máhr has a PhD in Computer Sciences from the Delft University of Technology, The Netherlands, and project manager of the simulation group of AITIA International Inc. He has been the senior software architect of the CRISIS project on agent-based modeling of the macro-financial system, responsible for the CRISIS Game Architecture, as well as for the CRISIS Integrated Simulator. His past research involved agent-based transportation planning and robustness of multiagent logistical planning, but he also published on auction algorithms, which are commonly used mechanisms in multiagent systems. Before the PhD track, Tamás got his MSc from the Budapest University of Technology and Economics, Hungary.
Shaun Michel is a computational sociologist in The MITRE Corporation's Department of Artificial Intelligence and Cognitive Science, specializing in simulation modeling, social behavior, and globalization. He began working on a PhD in Sociology at George Mason University in 2012 and he holds a master's degree in Sociology from East Tennessee State University.
Francisco J. Miguel Quesada is an associate professor at the Universitat Autónoma de Barcelona (UAB) teaching Methodology for the Social Sciences, Sociology of Consumption, and Applied Statistics for Marketing Analysis. He holds a PhD in Sociology from the UAB and a university specialist degree in Sociology of Consumption from the Universidad Complutense de Madrid. He has conducted research in sociology of consumption, social indicators, and school-to-work transitions. At present, he mainly works in the domain of computational sociology as GSADI group member. As Head of the “Laboratory for Socio-Historical Dynamics Simulation” (LSDS), he has been involved in several projects on the use of agent-based social simulation for modeling social networks dynamics and evolution of social behavior.
Luigi Mittone is a full professor of Economics at the University of Trento, Italy. At the University of Trento he is the Director of the Doctoral School of Social Sciences, Director of the Cognitive and Experimental Economics Laboratory, and coordinator of the International Master in Economics (MEC). He is also the coordinator of the research project in Experimental Economics and Nudging at the Bruno Kessler Research Center. He attended his master's degree in Economics at the Bocconi University of Milan (Italy), his master's degree in Social Sciences at the University of Birmingham (UK), and his PhD in Economics at the University of Bristol (UK). Luigi's main research interests and publications are in the field of Experimental and Behavioral Economics: fiscal evasion theory, consumer behavior, mental modeling of uncertain events, intertemporal choices, and cooperation among agents; Computational Economics: fiscal system dynamics with heterogeneous agents; and Public and Health Economics.
Gareth Myles is Professor of Economics and a Research Fellow at the Institute for Fiscal Studies. He is an associate editor of the Journal of Public Economic Theory and was managing editor of Fiscal Studies from 1998 to 2013. He obtained his BA from Warwick in 1983, his MSc from the London School of Economics in 1984, and his DPhil in 1987 from Oxford. His major research interest is in Public Economics and his publications include Public Economics (1995), Intermediate Public Economics (2006) and numerous papers in International Tax and Public Finance, the Journal of Public Economic Theory, and the Journal of Public Economics. Gareth is an academic adviser to HM Treasury and HM Revenue and Customs. He has also provided economic advice to international bodies including the European Commission and the OECD.
José A. Noguera is an associate professor (2002) in the Department of Sociology at the Universitat Autónoma de Barcelona, and Director of the Analytical Sociology and Institutional Design Group (GSADI). He holds a PhD in Sociology (1998) from the Universitat Autónoma de Barcelona and has been visiting researcher at the University of California, Berkeley, and at the London School of Economics and Political Science. He has been the principal investigator of several GSADI projects funded by the R&D Spanish National Plan since 2006. His empirical research is currently focused on tax compliance, social influence dynamics, prosocial motivations, and the feasibility of universal welfare policies such as basic income. José's publications also cover sociological theory, philosophy of social science, social policy, and normative social theory. His work in sociological theory aims to demonstrate the explanatory power of analytical sociology and the social mechanisms approach in sociology, with a strong focus on methodological individualism, the theory of rationality, social ontology, and the philosophy of social science. He is editor of PAPERS: Revista de Sociologia, an editorial board member of Basic Income Studies, and has served on the editorial board of Revista Española de Investigaciones Sociológicas. He is a member of the International Network of Analytical Sociologists (INAS) and serves on the Board of the Spanish Basic Income Network (RRB) and on the International Advisory Board of the Basic Income Earth Network (BIEN).
Una-May O'Reilly is the founder of the AnyScale Learning For All (ALFA) group at the Massachusetts Institute of Technology Computer Science and Artificial Intelligence Laboratory. She has expertise in big data, scalable machine learning, evolutionary algorithms, and frameworks for large-scale, automated knowledge mining, prediction, and analytics. She holds a PhD from Carleton University, Ottawa, Canada. She is an elected Fellow of the International Society for Genetic and Evolutionary Computation, holds the EvoStar Award for Outstanding Achievements in Evolutionary Computation, is Vice Chair of ACM SIGEVO, area editor for Data Analytics and Knowledge Discovery for Genetic Programming and Evolvable Machines (Kluwer), editor for Evolutionary Computation (MIT Press), and action editor for the Journal of Machine Learning Research. At ICAIL 2015 she co-authored the Peter Jackson Best Innovative Application paper.
Aloys L. Prinz is a full professor of Economics at Westfälische Wilhelms-Universität Münster, Institute of Public Economics. His work focuses primarily on public economics and taxation. He is the co-organizer of the biannual conferences on the shadow economy and tax evasion. In addition to publishing intensively in scientific journals, he has also coauthored books for general readers on social policy, public debt, monetary policy and taxation. Moreover, Aloys is also an active researcher in economic ethics.
Jacob Rosen is a chief technology officer at Deftr. His research focuses on the abstract representation of financial transactions and the use of artificial intelligence techniques to characterize human behaviors. Jacob received an MSc in Technology and Policy from the Massachusetts Institute of Technology.
Viola L. Saredi is a junior research fellow at the Cognitive and Experimental Economics Laboratory at the University of Trento, Italy. As a junior researcher, she has recently visited the Department of Applied Psychology at the University of Vienna. She holds a PhD degree in Behavioral and Experimental Economics from the University of Trento, and a master's degree in Economics from the Catholic University of the Sacred Heart, Milan, Italy. In keeping with the interdisciplinary nature of her field of interest, Viola's research mainly focuses on the application of innovative experimental methods to the study of psychological drivers and biases affecting consumers' behavior and clandestine activities (such as tax evasion, bribing, and agency dilemmas).
Götz Seibold is a professor at the Brandenburg University of Technology Cottbus-Senftenberg, Germany, where he holds the Chair for Computational Physics. He received his PhD in Physics from the University of Stuttgart, Germany, in 1995. He works in the general area of solid-state theory with a focus on superconductivity and strongly correlated electron systems. He is internationally recognized for his work on electronic inhomogeneities in high-temperature superconductors and the time-dependent Gutzwiller approximation. Besides, Götz is interested in the field of econophysics where he works on agent-based models related to tax evasion and the shadow economy.
David Slater is a lead operations research analyst within The MITRE Corporation's Operations Research Department. He has spent the last 5 years at MITRE applying Applied Mathematics to some of the United States' most critical and challenging Systems Engineering problems. His research interests include complex systems analysis, network science, and complexity science. He received his BS from Michigan Technological University in 2005 and an MS and PhD in Applied Mathematics from Cornell University in 2009 and 2011, respectively.
Eduardo Tapia Tejada holds a PhD in Sociology from the Universitat Autònoma de Barcelona and is a postdoctoral researcher at The Institute for Analytical Sociology (Sweden). He is a graduate in Sociology from the Universidad Federico Villarreal of Peru and he holds a master's degree in Sociological Research from the Universitat Autònoma de Barcelona. He also holds a degree in Design and Evaluation of Social Projects from the Pontificia Universidad Católica del Perú. He was a postdoctoral research fellow at GSADI (2013–2014) in the project Social and environmental transitions: SIMULPAST. His research interests involve two main areas: diffusion dynamics, particularly how certain beliefs spread through social groups, and beliefs evaluation mechanisms. In order to tackle these issues, Eduardo uses tools such as network analysis, statistical modeling, and agent-based simulation. The main theoretical question that motivates his research is to analyze how actions at the micro level generate patterns at the macro level.
István János Tóth is senior research fellow at the Institute of Economics of the Hungarian Academy of Sciences in Budapest and managing director of Corruption Research Center Budapest (CRCB). He graduated at the Corvinus (former Karl Marx) University of Budapest in 1984, and studied at École des Hautes Études en Sciences Sociales (EHESS) in Paris in 1990–1991. He holds a master's degree in Economics and Sociology. István János obtained his PhD at the Hungarian Academy of Sciences in 1998. His main research interests are the problems of corruption and economic institutions, hidden economy, and tax evasion.
Geoffrey Warner is a lead information systems engineer at The MITRE Corporation. He received his PhD in theoretical condensed matter physics from the University of Illinois at Urbana-Champaign. His research involves the application of mathematical and computational modeling to a diverse range of problems, from tax evasion to the physics of remote sensing devices.
Sanith Wijesinghe is an Innovation Area Leader at The MITRE Corporation where he oversees research and development efforts to support multiple federal government agencies. His most recent work on detecting tax evasion schemes was awarded the Peter Jackson award by the International Association of Artificial Intelligence and Law and was featured in the New York Times. Prior to joining MITRE, he was Vice President for Project Deployment at MillenniumIT, a software development firm serving the capital markets industry. He received his master's and PhD degrees in Aeronautics/Astronautics from the Massachusetts Institute of Technology.
Foreword
We model social systems to better explain, predict, design, and act. Macroeconomic models explain growth rates and patterns. Financial models predict stock prices. Models of correlated assets help the Federal Communications Commission design their spectrum auctions. And industrial organization models of firm competition inform decisions to intervene by the Justice Department.
The aforementioned macroeconomic models differ in their assumptions and methodologies. Macroeconomists gloss over firm level strategic choices that lie at the core of the industrial organization models. Auction models assume strategically rational actors, whereas some financial models assume actors who buy and hold.
Models include some variables and leave out others. They simplify the world, and as a result, err. We can correct for the biases and faults in any one model by constructing complementary models. To be of use, these other models must capture or include different features or relationships. They can then fill in the gaps missed by the original model. By offering more coverage and encompassing a larger number of processes, an ensemble of models provides deeper and broader knowledge and more refined insights than does a single model.
This book exemplifies this many-model approach. The chapters within explore a single policy domain – tax evasion – through the lenses of a variety of agent-based models (ABMs). While each model makes a substantial stand-alone contribution, the volume in its entirety makes a much larger one.
ABMs consist of situated agents endowed with attributes who follow rules instantiated in computer programs. The modeler defines the agents, their attributes, their physical or social placement, the rules they follow, and the processes by which their behaviors aggregate. She then hits the return key and watches phenomena emerge. Hence, many refer to ABMs as a bottom-up methodology.
An ABM of friendship formation might position people in a space, assign them physical features and rules for making and breaking friendships based on those features, and then allow networks to form. The modeler builds in the assumptions. The friendships emerge.
ABMs offer finer granularity and more flexibility than game theory models. Game theory models often consist of two, three, or an infinity of types – as these make deriving equilibria possible. Game theory models also assume, for the most part, rational actors. An ABM can include any number of types as well as any behavioral rule that can be encoded as an algorithm.
ABMs can also include heterogeneity within types, network structures, learning and adaptation, and (almost) any size population. That flexibility forms an Achilles heel large enough to be hit with an arrow from about 10,000 feet. Almost anyone can program an ABM. Not many people can solve for a Bayesian Perfect Subgame Equilibrium. Thus, there exists an abundance of not so useful, baked in the results, ABMs. It might even be that as time goes to infinity, so too does the ratio of bad to useful ABMs.
Books such as this can reverse that trend. The editors have selected a collection of sound exemplars. And, as if to rein in over-eager newbies, the editors lead with a treatise on the art and practice of writing scientific ABMs. They take the stand that ABMs can assist, aid, and improve, and do not claim that ABMs are the be all or bee's knees. This modesty stems from an awareness that ABMs have been touted (overhyped) as a breakthrough methodology for social science for more than a quarter-century.
ABMs have had successes to date, particularly in domains such as epidemiology, traffic, crowd dispersion, racial segregation, and meme spreading. These are all domains in which the network of interactions or geography plays a large role and behavior is low-dimensional or rule-based. This foray into tax evasion offers an opportunity to push the frontier of impact into economic policy.
At the moment, ABMs occupy an awkward position within economic analysis. On the one hand, they have become ubiquitous and ridiculously broad in scope: The set of ABMs range from elaborate government-funded, million agent models with household level granularity, to laptop models of supply chain dynamics, to models by physicists of spinning magnets. On the other hand, they remain a (somewhat) fringe methodology among the academic elite. Not many ABMs make their way into the top journals. That said, journal editors do serve up prodigious quantities of the near beer of numerical estimations and calibration exercises.
Tax evasion – the singular focus of this work – provides a near ideal context by which to demonstrate the potential of ABMs. That potential can be best seen juxtaposed with neoclassical economic models.
The straw man neoclassical economic model of tax evasion assumes a single representative agent who decides how much to shade reported income. The model also makes assumptions about the tax progressivity, risk-aversion, and auditing policies – do tax authorities look harder at rich people or do they monitor according to an equilibrium strategy?
The amount of evasion predicted by neoclassical models depends on the assumptions. Increasing risk-aversion drives up evasion and progressive taxation biases the income of evaders upward. These models operate and succeed within a small box. They explain why people cheat, they predict who will cheat, they point to better designs for tax policy, and they tell the Internal Revenue Service where to look for evaders.
Though useful as a tool for organizing our thinking about the implications of tax progressivity, changing levels of risk-aversion, and auditing practice, they prove a rather blunt instrument. They provide deeper insights than say “bigger houses sell for more money” and “studying harder improves grades” but not by much.
ABMs allow for a much richer set of assumptions. They also embrace the complexity of the worlds in which evasion takes place. Evasion does not take the form of lying about 5% of your income. Evasion comes in the form of equity swaps, shell companies, misreporting of expenses and gains, and even the now famous Double-Dutch Irish Sandwich used by corporations.
The models in this book emphasize broadening assumptions within four categories: behavior, occupations, networks, and social influence. Each of these four extensions adds value on its own. As an ensemble, they add much more.
By behavioral diversity, I mean differences in how people make decisions about paying taxes. The neoclassical model assumes that each person makes a calculated decision whether or not to cheat. If that model produces boundary decisions, then either people do not cheat, or they all cheat. If it produces interior solutions, everyone cheats a little.
Some of the models discussed in this book include three other types of people: honest people who always pay their taxes (that's me by the way!), socially influenced people who mimic the behaviors of others, and random, idiosyncratic people who evade taxes either on a whim or because they accidentally tossed out a 1099 form thinking it was another American Association of Retired Persons application.
The honest types lower the number of evaders. That alone is a contribution because the economic models suggest far more evasion than can be identified empirically. The socially influenced types create patterns in tax evasion. If evaders avoid the tax authorities, then social influence creates a positive feedback for evasion. The random, idiosyncratic people introduce noise.
Second, the ABMs include occupational variation. This allows for more realistic assumptions. A school teacher or line worker who earns all her income as salary and receives an employer wage reporting form W2 has less room for evasion than a New York City financial consultant pulling in $10 million per year who works part time out of her home in Connecticut. The consultant can claim any number of business-related expenses (a Rolling Stones concert) that the teacher or line worker cannot. The consultant can also claim to live more than half the year in Connecticut and avoid the nearly 4% New York City income tax. Although, if she reports 183 days in CT and 182 in NYC, she will trigger an audit automatically.
The key observation that evasion requires the opportunity to evade is obvious yet ignored by the economic model. To construct an ABM to capture how opportunity does not spread equally across professions requires categorizing occupations by evasion potential and calibrating the numbers to economic data. For the ABM to be the cat's meow and not the dog's breakfast, the categorizations must be germane, the assumed behaviors within a category must be appropriate, and the calibration must be precise. All are doable. None are easy.
Third, ABMs naturally embed social networks and geography. In the real world, people reside in physical space, and they mix with their friends rather than randomly. The same assumptions can be built into an ABM. Adding the network has no effect unless information or influence spreads over the network. Hence, my earlier comment about the whole exceeding the parts. The neoclassical evaders can be put in any network you like. Nothing will change.
However, when we add social influence (the final category), networks do matter. ABMs typically add social influence in one of two ways, both can be found in this book. As already mentioned, an ABM can include some agents who are socially influenced. Alternatively, an ABM can assume that everyone has some level of susceptibility to social pressure. Much like someone drives 68 mph in a 65 mph limit zone because “everyone else does,” so too might someone claim a home office even though their father-in-law now sleeps in that room.
To be fair, game theoretic models can also include social influence and they can even capture social influence over networks. However, when we add in behavioral types and occupational types, we soon find that our model world contains too many types and cases to perform all those calculations. Computation proves the only practical path.
Computation does more than overcome the combinatorial explosion of cases. It also reveals the class of outcome that obtains. Some systems go to equilibria quickly. Others produce simple patterns. Others walk randomly within a set. And last, others produce complex, often novel patterns.
The potential for patterns, randomness, and complexity in addition to equilibria challenges the standard comparative static evaluative approach, that is, if we turn this policy knob to the right, the equilibrium moves in this direction. We might also turn a knob and transform a well-behaved system into one that is chaotic.
This book should speak, if not sing, to two kinds of audiences. For those interested in the methodology of ABMs, what follows includes a mixture of novel applications of workhorse models along with new, higher fidelity models. For those interested in tax evasion, and the potential efficacy of interventions, it provides a wealth of ideas and insights. For both audiences, the book will expand their thinking about tax evasion.
In sum, although Ben Franklin claimed that, “In this world nothing can be said to be certain, except death and taxes,” in fact, many people do avoid and evade taxes. They have their Double-Dutch Irish Sandwich and eat it too. To explain and predict evasion within the complex financial world, to design systems that reduce it, and to take appropriate actions, we need many models, and we need models that embrace the complexity of the world. This collection of ABMs gets us started on that path.
Scott E. PageUniversity of MichiganSanta Fe Institute
Preface
Tax evasion or tax noncompliance is an age-old behavior of humans. It is probably as old as taxation itself. Taxation is an essential tool of any institutionalized community, a means by which resources are collected for the common causes. However, free-riding also appears to be an age-old, ever-existing phenomenon, leading to actors seeking safe ways to avoid the burdens of taxation. Various rulers and governments developed various taxes and tax systems, adding several layers of interpretations for avoiding paying taxes. Subjects of unjust rulers may feel disassociated with the projects that are financed by their taxes and may also feel that they do not possess the power to influence the decisions on the allocation of the taxes collected. This leads to a very complex net of possible reasons, causes, and effects for tax noncompliance that culminate in the repeated decisions by millions of individuals. Such decisions lead to the numbers individuals put in their tax reporting forms.
Agent-based modeling, on the other hand, is a novel methodology to study complex social systems, made possible by the recent advances in computer technology. Its main tenet is to model the individual decision maker, with all its idiosyncrasies and interactions, with specific goals and concerns. The multi-level, networked interactions of these individually modeled decisions are often too complex to follow with the analytical means provided by traditional mathematics. Therefore, the consequences of the micro-level rules are simulated in a computer instead, and the macro-level patterns that emerge are studied.
The mission of our book is to apply the emerging method of agent-based modeling to address the age-old problem of tax evasion.
While the sparking idea for the book emerged in 2014 in Barcelona, at a dinner during the Social Simulation Conference (The 10th Conference of the European Social Simulation Association (ESSA)), the origins can be traced back to the series of Shadow Conferences on the shadow economy, co-organized by Sascha Hokamp, where all the other editors of this volume had participated. At the dinner in Barcelona, Sascha Hokamp and László Gulyás were discussing the increasing number of agent-based contributions to the field of tax noncompliance. They decided that the field would benefit from a collection of the best works in this domain, especially when presented from a unified perspective. After the book project was accepted by John Wiley & Sons Limited, they decided to strengthen the editorial team by inviting two colleagues from the United States: Matthew Koehler and Sanith Wijesinghe.
The current volume is the result of three-year-long efforts of several people. The editors would like to thank all the authors who accepted our invitations and contributed the best of their work to this volume. We are also indebted to John Wiley & Sons Limited for supporting the production of this book since the very beginning. In particular, we are grateful for the cooperation by Viktoria Hartl-Vida, Debbie Jupe, Heather Kay, Alison Oliver, Blesy Regulas, Jo Taylor, and Liz Wingett. Furthermore, we would like to extend our thanks to the numerous additional people at John Wiley & Sons Ltd who were involved in the production of this book but are not named here explicitly. Special thanks are also due to three distinguished scholars, James Alm, Scott E. Page, and Klaus G. Troitzsch for their help with advertising the book. The valuable comments and contributions of 14 anonymous external reviewers are also gratefully acknowledged. Sascha Hokamp would like to thank Silvija Glodaite and Vanessa Hesping for their assistance in Warsaw and Barcelona, respectively. We, the editors, dedicate this volume to our families.
Sascha HokampUniversität HamburgLászló GulyásEötvös Loránd UniversityMatthew KoehlerThe MITRE CorporationSanith WijesingheThe MITRE Corporation September 2017
Part IIntroduction
Chapter 1Agent-Based Modeling and Tax Evasion: Theory and Application
Sascha Hokamp, László Gulyás, Matthew Koehler and Sanith Wijesinghe
1.1 Introduction
While the formal study of tax evasion began with a seminal paper by M.G. Allingham and A. Sandmo titled “Income Tax
