INFORMATION TO USERS The most advanced technology has been used to photo­ graph and reproduce this manuscript from the microfilm master. UMI film s the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer. The quality of th is reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor q uality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. A lso, if unauthorized copyright m aterial had to be removed, a note will indicate the deletion. Oversize materials (e.g., maps, drawings, charts) are re­ produced by sectioning th e original, beginning at the upper left-hand corner and continuing from left to right in equal sections with sm all overlaps. Each original is also photographed in one exposure and is included in reduced form at the back of the book. These are also available as one exposure on a standard 35mm slide or as a 17" x 23" black and w h ite photographic print for an additional charge. Photographs included in the original m anuscript have been reproduced xerographically in th is copy. H igher quality 6" x 9" black and w hite photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order. UMI University Microfilms International A Bell & Howeil Information Company 300 North Zeeb Road, Ann Arbor, Ml 48106-1346 USA 313/761-4700 800/521-0600 Order Num ber 8912662 Factors in th e noncom pliance decision: A n analysis o f M ichigan ta x am nesty participants Young, James Christian, Ph.D. Michigan State University, 1988 C opyright © 1988 by Youug, Jam es C hristian. A ll rights reserved. UMI 300 N. ZeebRd. Ann Arbor, MI 48106 FACTORS IN THE NONCOMPLIANCE DECISION: AN ANALYSIS OF MICHIGAN TAX AMNESTY PARTICIPANTS By James Christian Young A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Accounting 1988 ABSTRACT FACTORS IN THE NONCOMPLIANCE DECISION: AN ANALYSIS OF MICHIGAN TAX AMNESTY PARTICIPANTS By JAMES CHRISTIAN YOUNG Using data collected from individual tax amnesty participants in Michigan, this study performs a series of regression analyses to explore factors that influence taxpayer compliance. Separate regressions were run on the entire data set, five income stratifications based on adjusted gross income, and two data bases created to examine taxpayers who had no prior contact with the Michigan Department of Treasury. Specifically, this study suggests that noncompliance increases ^s income increases. Taxpayers having some form of Treasury Apartment contact prior to amnesty (withheld taxes, estimated taxes, a W-2, or a letter from Treasury requesting information regarding a state tax return) had substantially less unreported income than those amnesty participants who had no contact prior to amnesty. Amnesty participants with the opportunity to evade taxes [a composite of occupation (self-employed, business, professional, or sales), an income level of $30,000 or more, and access to cash income sources] had substantially more unreported income than other taxpayers. The occupations of amnesty participants were found to be a part of the decision process. Specifically, sales and self-employed taxpayers were be more likely to non-delinquent amnesty participants, while unskilled laborers were more likely to be merely delinquent in filing their tax returns (i.e., these individuals were more likely to have paid in most of their tax liability prior to amnesty). Amnesty participants who chose not to file a return in 1986 had more unreported income than amnesty participants who filed a return in 1986 (i.e., these taxpayers were more likely evaders than delinquent taxpayers). Although unable to generalize geographic findings, this study does support the contention that geographic location noncompliance decision. is a part of the The majority of studies (survey and experimental research) testing the compliance level of males versus other taxpayers have found males less compliant. This analytical study concludes that single males are less compliant than other taxpayers. Using exemptions as a surrogate for family size, this study found that as family size increases, so does the likelihood of evasion (versus delinquency of filing returns). Copyright by JAMES CHRISTIAN YOUNG 1988 To Dad and Mom: for challenging me to always ask why for teaching me to strive to be the best for encouraging me in all that I do for loving me as only parents can To Mary Jean: for her constant support for her ability to keep me focused for her love and encouragement To Christian: may I do all these things for you. v ACKNOWLEDGMENTS I am grateful for the advice and encouragement provided by my dissertation committee. Steven C. Dilley (Chairperson), Edmund Outslay, and John H. Goddeeris all provided invaluable support as I worked through the entire process. The Michigan Department of Treasury provided a unique opportunity by allowing me to assist in the creation of the Michigan Amnesty Data Base. Robert A. Bowman, State Treasurer, and Susan W. Martin, Revenue Commissioner, were the principal individuals involved in the decision to create the data base. Stanley Borawski I appreciate the opportunity they provided. and Eric Krupka, were my key liaisons at the Michigan Department of Treasury. Their efforts in the data collection process were unending, and their assistance to me was indispensable. TABLE OF CONTENTS PAGE LIST OF T A B L E S . . . . . . . . . xi i LIST OF FIGURES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv CHAPTER 1 : INTRODUCTION AND PROBLEM IDENTIFICATION 1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 The Phenomenon of Tax Noncompliance. . . . . . . . . . . . . . . . . 2 1.3 Estimating the Tax G a p . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.4 Tax Amnesty P r o g r a m s . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 1.5 Organization of the Dissertation. . . . . . . . . . . . . . . . . . . 18 CHAPTER 2 : THEORY BACKGROUND AND DEVELOPMENT 2.1 2.2 The Noncompliance Problem. . . . . . . . . . . . . . . . . . . . . . . . . 2.1.1 Economic Modeling of the Noncompliance Decision ... 2.1.2 Key Tax Compliance Variables. . . . . . . . . . . . . . . 28 2.1.3 Discussion of Economic and Demographic Factors. ... 2.1.4 Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 24 32 39 Methodological Considerations - Extension of Analytical Research. . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 CHAPTER 3 : DATA COLLECTION, CREATION OF DATA BASE, AND LIMITATIONS OF DATA 3.1 Data Collection - In General ................ 42 3.2 Individual Income Tax Returns Data Collection. . . . . . . . . . 49 3.3 3.2.1 Determination of Variables. . . . . . . . . . . . . . . . . 3.2.2 Data Collection Procedures and Verification of Data . 53 The Advantages and Limitations of theMichigan Amnesty Data Base. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 vii 49 CHAPTER 4 : CREATION OF THE RESEARCH DATA BASE AND A DESCRIPTIVE ANALYSIS OF THE DATA 4.1 An Overview of the Individual Tax Participant. . . . . . . . . 59 4.2 Creation of the Research Data B a s e . . . . . . . . . . . . . . . . . 62 4.2.1 General Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . 62 4.2.2 Verification of Data and ErrorCorrection Procedures. 63 4.3 4.3.1 4.3.2 Summary Characteristics of the Research Data Base. . . . . . 64 An Overview of the Research Data B a s e . . . . . . . . . . . 64 Distributions of Tax Payments CHAPTER 5 : AN OVERVIEW OF THE RESEARCH STUDY STATISTICAL METHODOLOGY and Reported Incomes. . 93 AND 5.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 5.2 Statistical A n a l y s i s . . . . . . . . . 99 5.3 Description of V a r i a b l e s . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 5.4 5.5 5.3.1 Dependent Variable. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 5.3.2 Independent Variables . . . . . . . . . . . . . . . . . . . . . . . . A Summary of the Data Bases To BeAnalyzed . . . . . . . . 107 109 5.4.1 Primary Research Data Base. . . . . . . . . . . . . . . . . . . . . . 109 5.4.2 Data Bases Constructed to Evaluate Taxpayers Without Treasury Department Contact ............ 109 5.4.3 Stratified AG I Data B a s e s . . . . . . . . . . . . . . . . . . . . . . Ill 5.4.4 Missing Information . . . . . . . . . . . . . . . . . . . . . . . . . 112 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 CHAPTER 6 : RESULTS AND ANALYSIS 6.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 6.2 An Overview of theMultiple Regression Analyses. . . . . . . . . . 116 6.2.1 In General. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 6.2.2 A General Discussion About the Assumptions of the Regression Model as it Applies to this S t u d y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 vi i i 6.3 Regression Analysis of the Research Data Base. . . . . . . . . . . 123 6.3.1 In General. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 6.3.2 Goodness of F i t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 6.3.3 Analysis of Regression Coefficients and Confidence Intervals. . . . . . . . . . . . . . . . . . . . . . . . . 126 6.3.4 Residual Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.5 S u m m a r y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141 6.4 Regression Analyses of 140 the Non-Contact Data Bases. . . . . . . . 145 6.4.1 In General. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 6.4.2 Analysis of Contact/No Contact Data Base (CONNCON.SYS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4.3 6.5 145 6.4.2.1 Goodness of Fit. . . . . . . . . . . . . . . . . . . . 147 6.4.2.2 Analysis of Regression Coefficients and Confidence Intervals .............. 149 6.4.2.3 Residual Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . 158 6.4.2.4 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 Analysis of Contact/No Contact Data Base Excluding the Retired/Student Occupational Group (C0NNC0N2.SYS). . . . . . . . . . . . . . . . . . . . . . . . . 162 6.4.3.1 Goodness of Fit. . . . . . . . . . . . . . . . . . . . . . . . . . . 162 6.4.3.2 Analysis of Regression Coefficients and Confidence Intervals .............. 165 6.4.3.3 Residual Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . 169 6.4.3.4 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170 Regression Analyses of the Stratified AGI Data Bases . . . 174 6.5.1 In General. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174 6.5.2 Analysis of Taxpayers With AGI LessThan $7,500 . . . 177 6.5.2.1 In G e n e r a l . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 6.5.2.2 Goodness of Fit. . . . . . . . . . . . . . . . . . . . . . . . . . . 177 6.5.2.3 Analysis of Regression Coefficients and Confidence Intervals .............. ix 179 6.5.2.4 6.5.3 Analysis of Taxpayers With AGI From $7,500 to $14,999. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . In G e n e r a l . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 6.5.3.2 Goodness of. . . . . . Fit. . . . . . . . . . . . . . . . . . . . . 183 6.5.3.3 Analysis ofRegression Coefficients and Confidence Intervals . . . . . . . . . . . . . . . . . . In G e n e r a l . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189 6.5.4.2 Goodness of. . . . . . Fit. . . . . . . . . . . . . . . . . . . . . 189 6.5.4.3 Analysis ofRegression Coefficients and Confidence Intervals . . . . . . . . . . . . . . . . . . 191 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194 Analysis of Taxpayers With AGI From $25,000 to $49,999. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 6.5.5.1 In G e n e r a l . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 6.5.5.2 Goodness of 6.5.5.3 Analysis ofRegression Coefficients and Confidence Intervals . . . . . . . . . . . . . . . . . . Fit. . . . . . . . . . . . . . . . . . . . . 195 197 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201 Analysis of Taxpayers With AGI More Than $50,000. . . 202 6.5.6.1 In G e n e r a l . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202 6.5.6.2 Goodness of 6.5.6.3 Analysis of Regression Coefficients and Confidence Intervals . . . . . . . . . . . . . . . . . . 6.5.6.4 6.5.7 189 6.5.4.1 6.5.5.4 6.5.6 185 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188 Analysis of Taxpayers With AGI From $15,000 to $24,999. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.5.4.4 6.5.5 183 6.5.3.1 6.5.3.4 6.5.4 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182 Fit. . . . . . . . . . . . . . . . . . . . . 202 204 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210 Comparison of the Strata Results. . . . . . . . . . . . . . . . . . 211 x CHAPTER 7 ; IMPLICATIONS OF FINDINGS AND EXTENSIONS OF THE RESEARCH EFFORT 7.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215 7.2 Implications of F i n d i n g s . . . . . . . . . . . . . . . . . . . . . . . . . . . 215 7.3 7.2.1 Academic Noncompliance Research . . . . . . . . . . . . . . . . 215 7.2.2 The Amnesty Participant in Michigan . . . . . . . . . . . . . 222 7.2.3 Designing An Amnesty Program. . . . . . . . . . . . . . . . . . . . . 224 7.2.4 Implications for Enforcement Efforts. . . . . . . . . . . . . . . 232 Extensions of the R e s e a r c h . . . . . . . . . . . . . . . . . . . . . . . . . 233 APPENDIX A : DATA COLLECTION DOCUMENTS . . . . . . . . . . . . . . . . . . . . . 236 BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274 xi LIST OF TABLES PAGE Table 1-1: Comparison of Methods of Estimating U.S. Underground Economy in 1976. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Table 1-2: Income Tax Gap, 1973-1981. . . . . . . . . . . . . . . . . . . . . . . . 9 Table 1-3: Unreported Legal Source Income of Individual Income Tax Filers and Nonfilers, 1973-1981. . . . . . . . . . . . . . . . 10 Table 1-4: Overstated Subtractions from Income Reported by Filers on Individual Income Tax Returns by Type of Subtraction, 1973-1981 . . . . . . . . . 11 Table 1-5: Unreported Legal Source Income and Overstated Offsets to Income, Individual Income Tax, 1981. . . . . . . . . . . . . . 12 Table 1-6: Partial Tax Gap Estimates for the Illegal Sector, 1973-1981. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Table 1-7: Summary of State Tax A m n e s t i e s . . . . . . . . . . . . . . . . . . . 16 Table 1-8: Overview of Amnesty Participation and Collections. . . . 19 Table 1-9: Amnesty Participation by Tax, Excluding Accounts R e c e i v a b l e . . . . . . . . . . . . 19 Table 1-10: Amnesty Collections by Tax, Excluding Accounts Receivable . . . . . . . . . . . . . . . . . . . . . . . . . 20 Table 2-1: Tax Compliance F a c t o r s . . . . . . . . . . . . . . . . . . . . . . . . . 29 Table 2-2: Relationship of Key Variables to Tax Compliance. . . . . . . 30 Table 3-1: Amnesty Participation by Tax, Excluding Accounts Receivable. . . . . . . . . . . . . . . . . . . . . . . . . . 45 Table 3-2: Amnesty Collections by Tax, Excluding Accounts Receivable. . . . . . . . . . . . . . . . . . . . . . . . . . 45 Table 3-3: Individual Amnesty Income Tax Participants: Comparison of Data Base to Total Population. . . . . . . . 47 xii Table 4-1: Summary Characteristics of Michigan Individual Income Tax Amnesty Participants (by Taxpayer). . . . . . 60 Table 4-2: Summary Characteristics of the Research Data Base. . . . 65 Table 4-3: Summary Characteristics of Returns . . . . . . . . . . . . . . . . 73 Table 4-4: Percentage Distribution of Occupations Reported by Michigan Income Tax Amnesty Participants. . . . . . . . 79 Table 4-5: Summary Characteristics of Returns by Occupation Category: Professional. . . . . . . . . . . . . . . . . . . . . . . 82 Table 4-6: Summary Characteristics of Returns by Occupation Category: Professional Support. . . . . . . . . . . . 83 Table 4-7: Summary Characteristics of Returns by Occupation Category: Sales . . . . . . . . . . . . . . . . . . . . . . . . . 84 Table 4-8: Summary Characteristics of Returns by Occupation Category: Skilled Labor . . . . . . . . . . . . . . . . . . . . . . 85 Table 4-9: Summary Characteristics of Returns by Occupation Category: Unskilled Labor . . . . . . . . . . . . . . . . . . . 86 Table 4-10: Summary Characteristics of Returns by Occupation Category: Self-Employed. . . . . . . . . . . . . . . . . . . . . . 87 Summary Characteristics of Returns by Occupation Category: Other (Retired, Student) . . . . . . . . . . . . 88 Table 4-12: Summary Characteristics of Returns by Gender. . . . . . . . 90 Table 4-13: Summary Characteristics of Returns by Filing Status . . 91 Table 4-14: Summary Characteristics of Returns: Non-Filers in 1986. . . . . . . . . . . . . . . . . . . . . . . . . . 92 Summary Characteristics of Returns: O pp ortu nity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 Percentage Distributions and Means of Amnesty Tax Paid by Various Categories. . . . . . . . . . . . . . . . . 95 Percentage Distributions and Means of Adjusted Gross Income by Various Categories. . . . . . . . . . . . . . 97 Table 4-11: Table 4-15: Table 4-16: Table 4-17: Table 6-1: Multiple Regression Summary. . . . . . . . . . . . . . . . . . . . . . . 117 Table 6-2: Regression to Predict Noncompliance Using Research Data Base . . . . . . . . . . . . . . . . . . . . . . . . . . 124 Table 6-3: Confidence Intervals of Regression Coefficients for Research Data Base . . . . . . . . . . . . . . . . . 127 xiii Table 6-4: Regression to Predict Noncompliance Using CONNCON Data Base. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146 Table 6-5: Confidence Intervals of Regression Coefficients for CONNCON Data Base. . . . . . . . . . . . . . . . . . . . . . . . . . 150 Table 6-6: Regression to Predict Noncompliance Using C0NNC0N2 Data Base . . . . . . . . . . . . . . . . . . . . 163 Table 6-7: Confidence Intervals of Regression Coefficients for C0NNC0N2 Data B a s e . . . . . . . . . . . . . . . . . . . . . . . . . 166 Table 6-8: A Summary of Significant Variables in the Stratified AGI Regressions . . . . . . . . . . . . . . . . . . . . Table 6-9: Table 6-10: 176 Regression to Predict Noncompliance Using A G O Data B a s e . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178 Confidence Intervals of Regression Coefficients for AGI1 Data Base. . . . . . . . . . . . . . . . . . . . 180 Table 6-11: Regression to Predict Noncompliance Using AGI2 Data Base. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184 Table 6-12: Confidence Intervals of Regression Coefficients for AGI2 Data Base. . . . . . . . . . . . . . . . . . . . . . . . . . . . 186 Table 6-13: Regression to Predict Noncompliance Using AGI3 Data Base. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190 Table 6-14: Confidence Intervals of Regression Coefficients for AGI3 Data Base. . . . . . . . . . . . . . . . . . . . . . . . . . . . 192 Table 6-15: Regression to Predict Noncompliance Using AGI4 Data Base. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196 Table 6-16: Confidence Intervals of Regression Coefficients for AGI4 Data Base. . . . . . . . . . . . . . . . . . . . . . . . . . . . 198 Table 6-17: Regression to Predict Noncompliance Using AGI5 Data Base. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203 Table 6-18: Confidence Intervals of Regression Coefficients for AGI5 Data Base. . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 Table 7-1: Comparison of Results to Other Analytical Noncompliance Studies . . . . . . . . . . . . . . . . . . . . . . . xiv 218 LIST OF FIGURES PAGE Figure 3-1: Michigan Amnesty Data Base: Individual Income Tax Return Variables. . . . . . . . . . . . . . . . . . . . . . . . . 51 Figure 4-1: A Summary of the Michigan Amnesty Data B a s e . . . . . . . . 69 Figure 5-1: Research Variables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 Figure 6-1: Research Data Base: Scatterplot of Standardized Residuals Against Standardized Predicted Scores . . . 142 Figure 6-2: Research Data Base: Histogram of Standardized R e s i d u a l s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143 Figure 6-3: Research Data Base: Normal Probability Plot of Standardized Residuals. . . . . . . . . . . . . . . . . . . . . . . . . 144 Figure 6-4: CONNCON Data Base: Scatterplot of Standardized Residuals Against Standardized Predicted Scores . . . 159 Figure 6-5: CONNCON Data Base: Histogram of Standardized Res i d u a l s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160 Figure 6-6: CONNCON Data Base: Normal Probability Plot of Standardized Residuals. . . . . . . . . 161 Figure 6-7: C0NNC0N2 Data Base: Scatterplot of Standardized Residuals Against Standardized Predicted Scores . . . 171 Figure 6-8: C0NNC0N2 Data Base: Histogram of Standardized Res i d u a l s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172 Figure 6-9: C0NNC0N2 Data Base: Normal Probability Plot of Standardized Residuals. . . . . . . . . . . . . . . . . . . . . . . . . 173 xv CHAPTER 1 INTRODUCTION AND PROBLEM IDENTIFICATION 1st 2nd 3rd 4th Robin: All: Ghost: Ghost: Ghost: Ghost: On Tuesday, I made a false income tax return. Ha! Ha! That’s nothing. Nothing at all. Everybody does that. It’s expected of you. Ruddigore, Act II, Gilbert and Sullivan 1.1 Introduction Tax compliance is a continuing national concern, and an issue that has gained deficits. increasing attention given the size of current budget Recent tax legislation has explicitly attempted to reduce noncompliance and recover uncollected taxes. Uncollected taxes from all sources of noncompliance are estimated by the Internal Revenue Service (IRS) to be $80 to $100 billion a year. In addition, the IRS has expressed great concern over an overall trend of declining voluntary compliance [IRS, 1983]. Because of these and other factors, there is great interest in establishing the factors contributing to compliance decisions. The increase in compliance research has exposed the paucity of empirically verified knowledge (i.e., analytic research) compliance. Graetz and Wilde [1985] have criticized the current status of tax compliance research on the following grounds: 1. Measures of compliance are weak and knowledge of major compliance problems is underdeveloped; 2. Theoretical approaches have been based on unrealistic assumptions and have not recognized the institutional constraints of enforcement policies; and about 2 3. Searches for "magic bullets" that would "solve" the compliance problem (e.g., lov/er tax rates or alternative tax systems) are misplaced, since compliance encompasses a complex range of behaviors. This study attempts to add to the analytic research base already established and further the understanding of the noncompliant taxpayer by examining certain economic and demographic factors available through data collected during the Michigan Tax Amnesty Program. The balance of this chapter is organized as follows: ■ Discussion of tax noncompliance; a Current estimates of the tax gap; a Tax amnesty programs; a Organization of the paper. 1.2 The Phenomenon of Tax Noncompliance Definition of Noncompliance. The Internal Revenue Service regards the appropriate measure of noncompliance to be "all federal income taxes that are owed but not paid" [IRS (1983], Noncompliance also has been defined as the amount of unreported taxable income -- the difference between the total amount of taxable income which is voluntarily reported during a given tax year and the correct amount of taxable income for that year (under the Internal Revenue Code), given all taxable events which occurred during the year. The latter definition will be used in this study to measure noncompliance. Noncompliance from an individual income tax perspective-can be — categorized into several major components, as follows: Tax on Legal-Source Income ■ Tax associated with filed returns ■ Tax associated with nonfiled returns Tax on Illegal-Source Income ■ Illegal gambling ■ Illegal drugs ■ Prostitution Tax Gap v. "Underground Economy". Although the tax gap is frequently thought to be the tax due on income earned in the "underground economy" (i.e., illegal source income) this metaphor is not technically descriptive of the tax gap. It is important to emphasize that the tax gap relates to a large variety of errors and misrepresentations including overstatement of personal and business deductions, personal exemptions and statutory adjustments, as well as understatement of income. While significant amounts of the tax gap are related to cash businesses (e.g., "off-the-books" businesses such as moonlighting workers, sidewalk vendors, etc.), other more "typical" forms of "above­ ground" income share in its make-up dividends, capital gains, etc.). (e.g., unreported interest and In addition, economic activity results in tax evasion. not all underground For example, a person with very low income deriving marginal amounts from underground activities may still have such a low total of income that no tax is owed. Although the term "underground economy" is not completely descriptive of the noncompliance problem, it will continue to be used here and discussed within the context of tax compliance. 1.3 Estimating the Tax Gap Gutmann’s Estimation of the Tax Gap. One event which sparked public interest in the underground economy was Gutmann’s analysis [1977] of the tax gap. Gutmann estimated that the subterranean economy (his chosen label) was in the range of $176 billion in 1976 -- over 10 percent of the gross national product. The basis for Gutmann’s estimate was an analysis of the amounts of cash in circulation compared with demand deposits held by banks. He analyzed these relationships and their trends since World War II (he assumes little in the way of subterranean activity prior to that time). His research concluded that there is far more cash in circulation than can be accounted for by bank deposits and reasonable allowances for cash held by businesses and individuals. Gutmann’s estimates, and particularly his methods, were the subject of considerable initial skepticism. Feige [1979] and Tanzi [1980]) But, as other researchers (e.g., have reported their results, it is becoming clearer that Gutmann’s figures at least reflect a measure of moderation and balance within the context of developing theories on the tax gap. Table 1-1 displays four of the most often referenced estimates of the tax gap. It is notable that the estimates of Gutmann, Feige, and Tanzi all are derived from analyses of indirect indicators of underground activity. The IRS figures have a more empirical basis. Feige’s estimates, developed for the Netherlands Institute for Advanced Study, are viewed dubiously by a number of commentators. The use of the money supply as a basis for estimating the underground economy is generally considered less reliable than Gutmann’s basic approach. Tanzi revised Gutmann’s method by considering a number of other variables interest rates, share of wages and salaries in personal capita income, and average tax rates). (e.g., income, per 5 TABLE 1-1 COMPARISON OF METHODS OF ESTIMATING U.S. UNDERGROUND ECONOMY IN 1976 Author/Source Size of Underground Economy As A Percent of GNP Method of Estimation Feige 22 - 33 % Analysis of total money supply relative to GNP Gutmann 10 - 14 % Ratios of currency in circulation to demand deposits 6 - 8 % Analysis of income tax returns and other measures of noncompliance IRS 4 % Tanzi Ratios of currency to demand deposits, adjusted to consider currency demand, interest rates, tax rates, and other variables Internal Revenue Service Research. Most of the economists and other social scientists who have studied the underground economy have directed their investigations toward obtaining broadly stated estimates of its aggregate size or determining the causative forces and motivations for noncompliance. Although almost all acknowledge the primacy of tax related issues as an underlying cause (and the loss of tax revenues as a principal consequence) there has been little independent research specifically directed at the tax gap. The Internal Revenue Service has made a comprehensive effort to associate measures of underground economic activity with their related tax consequences. Their first report, Estimates of Income Unreported on Income lax Returns [1979] was the result of a research effort stimulated at least in part by the wide interest and public questioning generated by Gutmann’s conclusions. The report’s bottom line was an estimate that between $75 billion and $100 billion in legal sector income, and between $25 billion and $35 billion in illegal sector income, went unreported and untaxed in 1976. This implied losses of federal tax revenues in the range of $19 billion to $26 billion, well over ten percent of actual federal tax revenues of $142 billion for that year. This IRS report in turn generated further public interest, especially as various Congressional committees turned their attention to the problem. Certainly the report and its implications, supported by independent estimates and commentary by economists and others, played a role in pushing the Tax Equity and Fiseal Responsibility Act of 1982 (TEFRA) through to passage. TEFRA contained numerous tax compliance measures designed to encourage self-compliance and deter tax evasion. More recently IRS researchers have published a comprehensive study, Income Tax Compliance Research [1983]. This report reflected a broader view of the tax gap, estimating all its components, not just unreported income as had the 1979 report. In contrast to other researchers who use various indirect methods to estimate the size of the tax gap, the IRS has employed a more direct approach. The estimates of underreported income and overstated items of filers were derived from returns examined under the IRS’ Taxpayer Compliance Measurement Program (TCMP). The estimates developed were then adjusted on the basis of a further comparative study with the IRS’ Information Returns Program (IRP), which pointed out inadequacies in the ability of the TCMP surveys to fully detect underreporting. The IRS points out that the estimates have been notably cautious in estimating illegal sector income from drug sales and in estimating the size of the nonfiling problem. The report’s key findings (as related to the present study) are summarized in Tables 1-2 through 1-6 [IRS, 1983]. Table 1-2 presents estimates of the total federal tax gap for the years 1973, 1976, 1979 and 1981. The total tax gap for 1981 was estimated at about $91.5 billion, with $81.5 billion (or about 89 percent) from the legal sector. billion (about eleven percent) is from the illegal sector. Only $9 The table illustrates the dramatic growth in the aggregate tax gap, from about $30.9 billion in 1973 to $91.5 billion in 1981. due to increases in This growth is not all cheating, other factors arealso at work (e.g., inflation). Table 1-3 provides estimates of unreported legal again for 1973 through 1981. highest amounts source income, As an example, note that in 1981, the of noncompliance derive from unreported wages and salaries (about $94.6 billion) and unreported business income ($58.4 billion). These two categories comprise about 61.3 percent of the total unreported legal source income items. overstated deductions and Table 1-4 shows estimates of exemptions. Overstatements of ordinary business expenses represent about 49 percent of the total for 1981. Table 1-5 provides a different view of the various 1981 estimates, showing a breakdown of unreported income between filers and nonfilers, in addition to exemptions. restating the amounts of overstated deductions and In addition, it shows how the individual income tax gap amount was computed ($68.5 billion, as shown in Table 1-2). Table 1-6 provides a breakdown of the tax gap estimates for the illegal sector. No other direct estimates of the tax gap exist. As a result, the relative merits of the IRS study are somewhat difficult to determine. However, since the IRS has access to a wealth of data in this area, and their tax gap estimates show up in the midrange of opinions (based on Table 1-1), it would appear that the results are objective. In July 1987, the American Bar Association released the results of a four-year study on taxpayer compliance Compliance [1987]). of improving (ABA Commission on Taxpayer The Commission was formed in 1983 to recommend ways compliance with the federal income tax laws. The Commission’s report further corroborates the findings of the 1983 IRS study, noting that the current tax gap approximates $100 billion. According to the report, the factors contributing to the noncompliance decision include (but are not limited to): 1. Opportunity (a combination of income source and occupation), 2. Complexity, 3. Perceived unfairness, 4. Tax rates, and 5. Lack of contact by the IRS. 9 TABLE 1-2 INCOME TAX GAP, 1973-1981 (In Billions of Dollars) 1973 1976 1979 1981 28.8 39.2 62.3 81.5 Corporation tax gap, total 3.5 4.6 6.4 6.2 Individual tax gap, total 25.3 34.6 55.9 75.3 23.8 32.2 50.6 68.5 0.9 1.4 2.0 2.9 22.9 30.8 48.6 65.6 17.3 2.1 3.4 0.1 24.2 3.4 3.0 0.2 38.4 4.7 5.0 0.5 52.2 6.3 6.6 0.5 1.5 2.4 5.3 6.8 Employer underdeposit of withholding3 1.1 0.9 1.8 2.4 Individual balance due after remittance 0.4 1.5 3.5 4.4 2.1 3.4 6.3 9.0 Legal sector tax gap, total Individual income tax liability reporting gap, total Nonfilers’ income tax liability (Net of prepayments and credits) Filers’ income tax liability: Unreported income Overstated business expenses Overstated personal deductions^ Net math error Individual income tax remittance gap, total Illegal sector tax gap (partial)3 * Includes itemized deductions, personal exemptions, and statutory adjustments. 3 Also includes a small amount for underreported withholding by employees and a small negative amount for underclaimed withholding by individuals 3 Includes income from illegal drugs, illegal gambling, and prostitution only. Source; "Income Tax Compliance Research," Internal Revenue Service [1983]. 10 TABLE 1-3 UNREPORTED LEGAL SOURCE INCOME OF INDIVIDUAL INCOME TAX FILERS AND NONFILERS, 1973-1981 (In Millions of Dollars) 1973 1976 1979 1981 Wages and salaries Dividends Interest Capital gains Nonfarm proprietor income and partnership and small business corporation income1 Farm proprietor income Informal supplier income Pensions and annuities Rents Royalties Estate and trust income State income tax refunds, alimony, and other income 33,304 1,920 4,440 5,015 46,274 3,638 6,763 9,935 71,076 5,528 11,548 16,283 94,581 8,747 20,479 17,727 23,906 5,742 10,346 3,123 1,335 312 487 32,565 4,542 12,721 4,067 2,390 1,088 695 47,246 7,832 16,995 6,258 2,711 1,672 1,140 58,400 9,547 17,080 8,799 3,049 2,770 1,330 3,990 6.857 6,260 7.166 Total Unreported Income2 93,919 131,535 194,548 249,675 1 Does not include informal supplier income. 2 Total may not equal sum of components due to rounding. Source; "Income Tax Compliance Research," Internal Revenue Service [1983]. 11 TABLE 1-4 OVERSTATED SUBTRACTIONS FROM INCOME REPORTED BY FILERS ON INDIVIDUAL INCOME TAX RETURNS BY TYPE OF SUBTRACTION, 1973-1981 (In Millions of Dollars) 1973 1976 1979 1981 business expenses* statutory adjustments personal deductions exemptions 7,229 1,368 5,759 4,269 10,887 448 4,737 4,693 13,250 1,113 5,595 7,435 16,179 1,803 6,958 8.060 Total overstatement offsets to income 18,625 20,765 27,393 33,000 Overstated Overstated Overstated Overstated * Consists of overstated expenses of farm and nonfarm proprietors, partnerships and small business corporations, rental properties owned by individuals, and estates and trusts paying income to individuals. Source; "Income Tax Compliance Research," Internal Revenue Service [1983]. 12 TABLE 1-5 UNREPGRTED LEGAL SOURCE INCOME AND OVERSTATED OFFSETS TO INCOME, INDIVIDUAL INCOME TAX, 19B1 (In Millions of Dollars) Tvoe of Income of Offset to Income Filers Nonfilers Wages and salaries Dividends Interest Capital gains Nonfarm proprietor income (except informal supplier income) Farm proprietor income Partnership and small,business corporation income Informal supplier income Pensions and annuities Rents Royalties Estate and trust income State income tax refunds, alimony. and other income 18,881 6,596 12,120 15,241 75,700 2,151 8,359 2,486 94,581 8,747 20,479 17,727 33,615 8,499 10,561 1,048 44,176 9,547 10,786 13,848 4,131 2,637 1,866 646 3,439 3,232 4,688 412 904 684 14,225 17,080 8,799 3,049 2,770 1,330 Total Unreported Income Overstated Overstated Overstated Overstated business expenses statutory adjustments personal deductions exemptions Total Overstated Offsets^ Total Misreporting-* Gross Tax Gap^ 4.975 Note: 7.166 133,840 115,8353 16,179 1,803 6,958 8.060 NA NA NA NA 16,179 1,803 6,958 8.060 33,000 NA 33,000 166,840 115,835 282,675 65,600 5,042 70,642 NA 2,185 2,185 65,600 2,857 68,457 Unclaimed Prepayments and Credits Net Tax Gap 2.191 Total 249,675 Sum of components may not add to totals due to rounding. NA indicates not applicable 1 Such income, which for tax purposes is treated as partnership income, is taxable to stockholders as ordinary income whether or not distributed. •p For a definition of informal suppliers, see Appendix D, Income Tax Compliance Research [IRS, 1983]. 3 Includes business income on a net income basis. * Excludes credits which are offsets to tax liability. 3 This is the sum of "Total llnreported Income" and "Total Overstated Offsets". ^ Tax liability based on total misreporting. Source: "Income Tax Compliance Research," Internal Revenue Service [1983]. 13 TABLE 1-6 PARTIAL TAX GAP ESTIMATES FOR THE ILLEGAL SECTOR, 1973-1981 (In Billions of Dollars) Tax Gap 1973 1976 1979 1981 Illegal drugs (Standard error) 1,2 (0.6) 1.9 (0.9) 4.1 (1.7) 6.1 (2.6) Illegal gambling (Standard error) 0.4 (0.1) 0.6 (0.2) 0.7 (0.3) 0.9 (0.3) Prostitution^ (Standard error) 0.6 (0.6) 1.0 (0.9) 1.5 (1.5) 1.9 (1.9) 2.1 (0.8) 3.4 (1.3) 6.3 (2.2) 9.0 (3.2) Total3 (Standard error)4 1 The drugs included were limited to heroin, cocaine, and marijuana. 2 Female prostitution only. 3 Sum of components may not add to totals due to rounding. 4 Sum of components will not add to totals due to offsetting errors in the calculations of total error. Source; "Income Tax Compliance Research," Internal Revenue Service [1983]. 14 1.4 Tax Amnesty Programs The perception that tax evasion is increasing has been a powerful factor in motivating many state revenue officials to back amnesties and strengthen tax enforcement, which they see as complementary. Typically, these programs provide taxpayers with a one-time opportunity to clear their accounts by paying back taxes and interest without being subject to civil or criminal penalties. Mikesell [1984] states that at least considerations are important in the use of amnesty. three theoretical These include: 1. The perceived fairness of such programs in societies whose tax systems depend largely on voluntary compliance, 2. The necessity that the amnesty program be viewed by taxpayers as a one time chance rather than a recurring opportunity, and 3. The coupling of the amnesty program with tax reforms designed to discourage back sliding on the part of former amnesty participants. Mikeselrs theoretical considerations are weakly supported by the rather thin empirical 1iterature on tax amnesty. States typically hope to accomplish three goals through the use of tax amnesty programs: 1. To collect outstanding tax revenues inexpensively, including revenue otherwise uncollectible due to limited enforcement resources. 2. To promote improved future citizen compliance with the tax laws, and 3. To bring individuals into the state’s revenue system who are unknown to the state and not easily detectable. Other factors must be evaluated in designing an amnesty program. Prior research in the area of amnesty programs suggests that the programs will work only if used infrequently and randomly. Another factor is that of citizen percept ion of the program’s fairness or equity (potentially the most controversial aspect of amnesty). A further consideration is 15 the relationship of the amnesty program to the more general issue of revenue system reform. If amnesty participants face the same incentives to engage in tax delinquency after the program as they did before, little or no permanent change may occur in their behavior. As a result, most states have coupled increased penalties and enforcement activities with the amnesty program. State Amnesty Results. amnesty program in 1982, Illinois apparently sponsored the first raising less than $100,000. Larger scale programs began in 1983, when four states sponsored programs. only one state (Arizona) collected more than $1 mill ion. ran programs in 1984, collecting over $250 mill ion, collecting $152 mill ion of that in a second program. However, Seven states with 111inois Six states followed in 1985, with California collecting $146 mill ion of the $190 mill ion total. New York and Michigan offered amnesties in 1986 and had two of the most successful programs -- raising $363 mill ion and $110 mill ion, respectively. Table 1-7 provides a summary of state tax amnesty programs. The Michigan Tax Amnesty Program. In April 1985, a tax amnesty program was proposed as part of a larger tax package. In addition to the amnesty, for the final enforcement effort computation of legislation included funding an (audit and discovery divisions), changes interest on underpayments of tax, penalties that could be assessed. and enhanced in the increases in Subsequent to amnesty, interest rate computations were tied to the prime rate and adjusted semiannually. In addition, most previous penalties were increased and their coverage expanded. New penalties were added for frivolous filings and for 16 TABLE 1-7 SUMMARY OF STATE TAX AMNESTIES Year State $ (millions) Collections 1982 1983 111inois Arizona Idaho Missouri North Dakota A1abama 111inois Kansas Massachusetts Minnesota Oklahoma Texas California Colorado Louisiana New Mexico South Carolina Wisconsin Michigan New York 0.1 6.0 0.3 0.9 0.2 3.2 152.4 0.6 84.5 11.9 13.9 0.5 146.5 6.0 1.2 13.9 7.1 20.0 103.9 363.2 1984 1985 1986 % of State Taxes - - 0.3 - - 0.1 1.8 1.5 0.2 0.5 0.5 0.3 1.0 0.2 0.4 1.1 1.7 Returns 400 10,600 900 241 630 10,150 22,456 750 52,000 10,400 56,500 85 164,000 7,000* 400* 48,000* 7,600 30,000* 121,491 145,000* Increased Enforcement Yes Yes Yes No No Yes Yes Yes Yes No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes *Incomplete data. Source: Individual state departments of revenue, National Conference of State Legislatures. 17 taxpayers discovered subsequent to amnesty whose liability was covered by the amnesty program. Under the amnesty program, the Department of Treasury waived all criminal and civil penalties for the failure to file a return or pay any state tax if an amnesty application was filed along with appropriate returns and payment of any state tax and related interest due. The legislation excluded from participation all taxpayers under criminal investigation or civil or criminal prosecution, but it allowed tax and interest amounts on outstanding assessments (i.e., part of the accounts receivable system) to qualify for the amnesty. Under the legislation, all taxes due on or before September 30, 1985 were eligible for amnesty. ended on June 30, 1986. The amnesty program began on May 12, 1986 and During the program, $109.8 mill ion in revenue was generated from interest and tax payments. This revenue covered over 128,000 filings by about 75,000 taxpayers. Table 1-8 provides a summary of the amnesty program. Of the $109.8 million generated, about $65.2 mill ion was related to receivables known to the Treasury Department prior to amnesty. These taxpayers filed approximately 81,000 returns (63 percent of the total), and paid 59 percent of the total tax and interest payments under the amnesty. Taxpayers unknown to the Treasury Department prior to amnesty paid taxes and interest of about $44.6 mill ion (41 percent of the total) and filed about 47,000 returns (about 37 percent of the total). Tables 1-9 and 1-10 present additional information regarding the filings of taxpayers unknown to the Treasury Department prior to amnesty. Table 1-9 provides information relating to returns filed as part of the amnesty program. As can be seen from the table, individual income tax 18 applicants accounted for about 69 percent of these new filings, with the majority of these individuals filing only one return. Only in the intangibles and single business taxes do the returns per applicant exceed two. Table 1-10 presents information regarding the revenues collected from these same taxpayers. revenues Not surprisingly, the largest portion of raised came from individual income tax participants ($13.6 million or about 30.5 percent). of the returns, interest Although accounting for only 11 percent single business tax participants paid in taxes and of about $12.4 mill ion (about 27.8 percent of the total). difference between returns and payments The is even more dramatic for use tax participants, who filed only 1.4 percent of the returns but generated 18.9 percent of the total revenue ($8.4 mill ion, and about $12,800 in taxes and interest per return). 1.5 Organization of the Dissertation The balance of this dissertation is organized as follows. Chapter 2 provides a detailed discussion of the development of noncompliance theory as it relates to this study. Chapter 3 discusses the data used in this study, including collection and construction of the Michigan Amnesty Data Base. Chapter 4 reviews the construction of the primary research data base used in the study and analyzes descriptive statistics related to the data. A description of the research and related statistical methodology considerations are discussed in Chapter 5. Chapter 6 presents the results of the analyses, and Chapter 7 summarizes the findings, discusses the contributions and implications of the results, potential extensions to the research. and comments on 19 TABLE 1-8 OVERVIEW OF AMNESTY PARTICIPATION AND COLLECTIONS (Amounts in Thousands of Dollars) Tax Case Class Number of Returns Tax Amount Interest Amount Total Amount Original Cases* Accounts Receivable Total Amnesty Denied** Amnesty Accepted 47,175 81.043 128,218 6,727 121.491 $37,458.2 56.980.4 94,438.6 5,180.1 $89,258.5 $ 7,149.3 8,206.9 15,356.2 714.1 $14,642.1 $ 44,607.1 65.187.5 109,794.6 5,894.2 $103,900.4 * ** "Original Cases" refers to 1iabilities other than those included in accounts receivable, that is, those which were not previously known or assessed by the Treasury Department. A request may have been denied because of failure to pay the full required amount or because the case was not eligible for amnesty. Source: Michigan Department of Treasury, Fisher, and Goddeeris [1987] TABLE 1-9 AMNESTY PARTICIPATION BY TAX, EXCLUDING ACCOUNTS RECEIVABLE (Data Through January 13, 1987) Tax Type Individual Income Withholding Intangibles Sales Use Single Business Diesel Fuel Gasoline Cigarette Miscellaneous Total Source: Number of Applicants 20,496 606 2,311 1,037 358 2,487 35 5 8 - - Number of Tax Returns 32,614 1,045 5,361 1,751 658 5,130 47 7 8 554 47,175 % of Total Returns 69.13% 2.22 11.36 3.71 1.39 10.87 .10 .01 .02 1.17 100.00% Returns Per Applicant 1.59 1.72 2.32 1.69 1.84 2.06 1.34 1.40 1.00 - - Michigan Department of Treasury, Fisher, and Goddeeris [1987] TABLE 1-10 AMiESTY COLLECTIONS BY TAX, EXCLUDING ACCOUNTS RECEIVABLE (As of Jine 19, 1987) Tax Type Tax Amoint (Audi ted) Interest Amount (Audited) Total Amount (Audi ted) Individual Income Withholding Intangibles Sales Use Single Business Diesel Fuel Gasoline Cigarette Miscellaneous $11,566.4* 1,084.0 5,009.5 2,494.7 6,733.2 10,395.5 87.5 11.4 11.6 64.3 $ 2,041.6* 134.7 865.8 379.5 1,689.3 2,013.4 9.8 2.3 1.8 11.1 $13,608.0* 1,218.7 5,875.3 2,874.2 8,422.5 12,408.9 97.3 13.7 13.4 75.4 Total $37.458.2 $ 7.149.3 $44.607.1 * Amounts in Thousands of Dollars Source: Michigan Department of Treasury, Fisher, anti Goddeeris [1987] Percent of Total Amount 30.51% 2.73 13.17 6.44 18.88 27.82 .22 .03 .03 .17 100.00% Total Amount Per Applicant $ 663.93 2,011.06 2,542.32 2,771.65 23,526.54 4,989.51 2,780.00 2,740.00 1,675.00 -- Total Amoint Per Return $ 417.24 1,166.22 1,095.93 1,641.46 12,800.15 2,418.89 2,070.21 1,957.14 1,675.00 136.10 $ 945.57 CHAPTER 2 THEORY BACKGROUND AND DEVELOPMENT 2.1 The Noncompliance Problem Academic analysis of the noncompliance problem has been performed on two distinct levels -- (1) determining the size of the problem and (2) determining the causes of noncompliance. Estimating the Size of Noncompliance. If we accept the fact that there is a tax gap (as defined in Chapter 1), determining its size is a logical question. Attempts by economists to estimate aggregate noncompliance using macroeconomic data date to the late fifties and early sixties (e.g., Groves [1958]). One method developed initially in the United States involves making inferences about tax evasion on the basis of changes in money holdings in the economy over time (Guttman [1977], Feige [1979], Tanzi [1982]). The results of these studies and others were discussed in Chapter 1. Although these studies are open to error, they do provide an indirect measure of the extent of tax evasion. the different approaches noncompliance, it is not used to measure surprising the that aggregate the Given level estimates of differ considerably. Determining the Causes of Noncompliance. The noncompliance decision is an intricate and multidimensional problem. The decision process itself has been the subject of various forms of theoretical modeling. In fact, research in the noncompliance area began with the development of economic models to try and shed 1ight on the factors which affect the tax evasion decision. This research, although problematic on several counts, provided a basis for the decision making process. 21 Proper models, of 22 course, are predicated on the identification of the salient factors that are a part of this decision making process. More recent research has concentrated on identifying the factors that are important in this process. Factors so identified in research to date can be broken into three major attitudinal. groups -- Various economic, demographic, methodological approaches and psychological have been used or to identify the important factors within these classifications. Studies which have explored the psychological or attitudinal factors of the noncompliance decision have re Iied on indirect measures of evasion to identify these factors. The studies have principally used surveys asking about past evasion or about attitudes toward evasion. The validity of these studies depends on the degree to which the pattern of survey responses corresponds to actual behavior. Studies examining economic and/or demographic factors of the noncompliance decision have focused on using data from the Internal Revenue Service's Taxpayer Compliance Measurement Program (TCMP). TCMP data base has been used extensively for this purpose. The Although there are significant weaknesses involved in using this data, it has been the best available data for identifying these factors. An Overview of the TCMP. Approximately every four years, the IRS selects a sample of returns for intensive examination. The samples for the TCMP individual surveys are probability samples of taxpayers who filed Form 1040 for the survey year. The samples are stratified by return type (based on level of reported income and whether business or non-business). sample. Approximately 50,000 taxpayers are included in each The returns are selected after math error verification has been completed, but before any other operational enforcement has occurred. 23 The TCMP surveys consist of intensive audits of the returns.Theses audits differ several respects. First, sampled tax from ordinary operational only experienced returns.Second, every item on audits in IRS examiners audit the the returns is examined. Third, the results of the audits are recorded on special "checksheets" which show the original corrected entries entries made as by the taxpayers determined in the side-by-side audit. With with the appropriate weighting, the sums of the differences for a given Form 1040 1 ine may be considered estimates from the TCMP of noncompliance with respect to the corresponding provisions of the Internal Revenue Code. The estimates based on TCMP individual survey to surveys assume that the survey results are not affected by changes in techniques. The TCMP examinations are conducted examination under detailed instructions which have not materially changed from survey to survey. The TCMP results may be changed in the IRS internal appeals process, or in the courts. Such changes are not reflected in the TCMP data. Goals of Compliance Research. There is no doubt that compliance is a serious problem under the current income tax. The problem demands attention, both in theory development and empirical analysis, in order to achieve a refinements better understanding of the problem. Hopefully, theoretical and empirical study will enable us to design better compl iance mechanisms and to improve tni/se Loat already exist.. It is apparent from the prior research in this area that the decision making process is so complex that no one study can expect to properly model the noncompliance decision, or identify all of the salient factors involved in that decision. Instead, proper theory development 24 depends on numerous studies to come up with a "composite" of this decision making process. Denzin [1978, p. 48] writes: I conclude that no single method will ever permit an investigator to develop causal propositions free of rival interpretations. . . multiple methods of observations must be employed. This is termed triangulation, and I now offer as a final methodological rule the principle that multiple methods must be used in every investigation, since no method is ever free of rival causal factors. Abdel-khalik and Ajinkya [1979, p. 21] concur, stating "[a]11 approaches to research are desirable, although they may have different degrees of strength and reliability." The discussion that follows provides an overview of the economic modeling research to date and the various factors that have been identified as potentially a part of the noncompliance process. 2.1.1 Economic Modeling of the Noncompliance Decision There is a great deal compliance by individuals, of work in the area of economics related to much of it devoted to models describing the noncompliance decision. date can be organized into four main the development of The models developed to groups: economic models, uncertainty models, norms of compliance models, and inertia models. Economic Models. Early studies involved analytic models which attempted to model the compliance decision in terms of the tax rate, the probability of being detected, and sanctions applied if discovered. Allingham and Sandmo [1972] pioneered this effort, and Srinivasan [1973], Yitzhaki [1974], and Farrington and Kidd [1977] all provided extensions of the original work. These studies viewed the taxpayer as a utility maximizer, generally risk-neutral, and perfectly amoral. Within this context, the taxpayer will make a choice not to comply whenever the tax savings is greater than the cost of being discovered (given their 25 assessment of the probability of detection). In other words, the taxpayer is viewed as a very thoughtful gambler, who carefully chooses an optimal level of evasion in light of the potential penalty that will be paid, and the likelihood of detection. Although criticized for being too simplistic, (i.e., not taking into account all of the factors involved in the noncompliance decision), these models became the foundation of other model development in the field of economics. Uncertainty Models. Uncertainty models refine economic models by acknowledging that most taxpayers do not know their audit probability or the potential penalties for tax evasion. Based on a variety of studies, it appears that many taxpayers do not even know their marginal tax rate. Without this basic information, it is impossible for taxpayers to determine either the risk or the reward of tax evasion with any degree of accuracy. Given these major uncertainties, the tax evasion decision is not based on maximizing strategies, but instead on rules of thumb or heuristics (i.e., on the perception of risk as opposed to actual risk (Frieland [1982]; Kahneman, Slovic and Tversky [1982]). For example, experimental research indicates that past experience with audits may bias estimates of audit probabilities. Benjamini and Maital [1985] found that compliance was higher among those taxpayers who were previously audited. Norms of Compliance. Considerable empirical evidence suggests that social norms (or role expectations) are important factors in the decision to comply or not comply with the tax laws. Those who comply with the tax laws are more 1 ikely to view tax evasion as wrong and perhaps even immoral (Schwartz and Orleans [1967]; Ekstrand [1980]; Tittle [1980]; Scott and Grasmick [1981]; Warneryd and Walerud [1982]). Alternatively, 26 tax evaders are more likely to think that tax evasion is a "victimless crime" and, in general, socially acceptable. Benjamini and Maital [1985] have noted that tax evasion may be regarded as a free-rider problem. In this context, each taxpayer would prefer that all other taxpayers comply with the tax laws, allowing for the personal evasion of taxes (i.e., taxpayers are only interested in their own position).Commitment to social norms, including the norm of compliance with tax laws, causes the taxpayer to incur a "psychic cost" from evasion and, therefore, reduces the attractiveness of the free-rider option. Public choice theory suggests that the strength of the commitment to comply is influenced by whether taxpayers believe that other taxpayers are also complying cooperation). (referred to as a strategy of conditional The strategy adopted is "if you pay your taxes, I'll pay mine; but if you cheat, I'll cheat, too." (Van den Doel [1978]; Laver [1981]) It appears that a taxpayer's perception of the overall fairness of the tax system also affects their commitment to comply, although the evidence is mixed in this regard (Spicer and Lundstedt [1976]; Mason and Calvin [1978], Song and Yarbrough [1978]; Yankelovich, Skelly and White, Inc. [1984]). Also, commitment to comply survey research indicates that an individual's with tax laws appears to be part of a broader commitment to their community (and/or country). Inertia Model. The inertia model focuses on the taxpayer's behavior over a long period of time (assuming that most individuals are creatures of habit). Inertia theory proposes that if taxpayers begin their tax reporting 1 ife in compl iance with the tax laws, they will remain in 27 compliance out of "inertia" more than anything else. The only way for this pattern of behavior to change is for a significant event to intervene in their life, changing their behavior and getting them to evade (Spicer [1986]). The theoretical grounding for this theory comes from psychology, specifically [1957]). Festinger's theory of cognitive dissonance (Festinger The cognitive dissonance theory states that when an individual holds inconsistent beliefs or acts in a way inconsistent with his beliefs, unpleasant feelings arise. These feelings force the individual to change either beliefs or behavior, so that one is consistent with the other. Applying this theory to tax compliance, when a taxpayer first begins to evade, guilty feelings inconsistent with beliefs. will be prevalent, because actions were However, as taxpayers begin to commit more acts of evasion, over time their commitment to social norms of compliance decreases, leading to a tax evasion belief/behavior result (Spicer [1986]). A Simplified Model of Tax Evasion. Spicer [1986] has formalized a somewhat simple model of tax evasion. An act of tax evasion will be committed only where: ( 1 - jr)t0y - 7rft0y - p c > 0 The taxpayer is viewed as deciding whether or not to commit a particular act of evasion rather than deciding how much to evade. If the taxpayer commits the act of evasion, the taxes evaded will be equal to tfly, where t is the tax rate that would apply on the unreported income, e is the fraction of taxable income not reported, and y is taxable income. If the taxpayer commits the act and is caught, a fine is imposed equal to ftfly, where f is the fine rate imposed on the taxes evaded. The 28 probability of detection of a particular act of evasion by tax authorities is n. Whether or not the act is detected, however, it is assumed that the taxpayer incurs a psychic cost equal to pc. The psychic cost of committing the act of evasion is due to several factors including a desire to avoid risks, feeling guilty, and other demographic and psychological variables. From this simple model, it is obvious that as n and f increase, fewer acts of evasion will be committed. Also, a increase in psychic cost will al so decrease the incidence of evasion. decrease) in t is somewhat less obvious. assumed to be positive, The effect of an increase (or However, if psychic cost is then certain acts of evasion will not be undertaken (since the psychic cost will offset the financial gain from the evasion). As t increases, the expected financial gain from an act of evasion increases, and may then outweigh the psychic cost. an increase in t will lead to more acts of evasion. As a result, Spicer states that further insight into the noncompliance decision will require a greater specification of the determinants of psychic cost. Conclusion. More recent models have adopted relaxed assumptions and utilized more sophisticated techniques of analysis. However, it appears that these advancements (to increase real ism) have increased the ambiguity of the results. Economic modeling appears to provide the best possibility of understanding the noncompliance decision, and accurate economic modeling requires the identification and validation of the key tax compliance variables that are a part of the decision making process. 2.1.2 Key Tax Compliance Variables Much research on the noncompliance of taxpayers has focused on whether fear of punishment is a primary motive for complying with tax 29 laws. This concept is called deterrence. While research findings have been fairly consistent and support the role of deterrence, a theory of noncompliance which focuses solely on punishment is too narrow in scope and is based on an overly optimistic view of the potential of law enforcement. The Internal Revenue Service has detailed 64 potential compliance factors. Table 2-1 provides a summary of some of these variables categorized into one of three groups -- psychological (or attitudinal), economic, and demographic. TABLE 2-1 TAX COMPLIANCE FACTORS T Economic a Level of income a Type of income a Tax rates a Cost of compliance a Sanctions (penalties) Table 2-2 y Demoqraphic Age a Occupation a Gender n Education a Race a Marital status a Geographic location m Contact from authority b Psvcholoqical Compliant peers Ethics Fairness Complexity a Sanctions (guilt) m Personal control a Decision framing m m a m (Jackson and Mill iron [1986]) references 14 of these variables, and notes miscellaneous others, in a chronological listing of 42 compliance studies. research: The compliance studies are categorized by type of survey, experimental, or analytical. The table indicates whether increasing the magnitude of the compliance variable is positively associated with compliance (+), whether increasing the magnitude is negatively associated with compliance indeterminate (0). (-), or whether the results are 30 TABLE 2-2 RELATIONSHIP OF KEY VARIABLES TO TAX COPLIANCE SURVEY STUDIES Author/ Oate Age Gender Educat­ Income ion (Fern) level Spicer, 1974 0 Vogel, 1974 ♦ 4 Mason & Calvin, 1974 ♦ 4 Song & Yarbrough, 1978 0 0 4 CSR (Aitken & Bonneville), 1980 + 4 - 4 4 + 0 0 0 Wth'hld Occupa­ Com­ Income tion/ pl iant Peers Source Status 0 - 0 4 4 - 4 0 4 0 4 0 - - Westat, 1980(a) Westat, 1980(b) Hotaling & Arnold, 1981 0 Richards & Tittle, 1981 - - 4 - 4 0 4 - Probab. of Sanc­ tions Detect. 0 0 0 0 0 4 4 4 - 4 4 4 4 4 - 4 0 4 4 4 0 0 4 0 ----- ----0 4 4 Scott & Grasmick, 1982 0 - 4 Tax Rates 0 - Scott & Grasmick, 1981 Uorneryd & Walerud, 1982 4 4 Grasmick A Green, 1980 Tittle, 1980 Ethics Fair­ Com­ Agency ness plexity Contact 4 4 4 4 4 4 4 4 4 4 4 4 - Furnham, 1983 4 0 4 4 Groenland S vanVeldhoven, 1983 4 Grasmick, Finley and Glaser, 1984 4 0 - - 4 4 0 4 4 -----Mason & Calvin, 1984 4 4 Thurman, St. John, & Riggs, 1984 4 0 Yankelovich, Skelly & White, 1984 0 0 Wallschutzky, 1984 - + 4 4 4 4 0 4 0 0 4 4 4 0 0 0 0 0 4 ♦ Positively associated with compliance - Negatively associated with compliance 0 link to compliance is indeterminate Multiple symbols indicate that the association with compliance differs for different segments of the taxpayer population. Source: Jackson and Kilt iron (1986] 0 - 31 TABLE 2-2 (CONTINUED) EXPERIMENTAL STUDIES Author/ Date Age Gender (Fem) Schwartz & Orleans, 1967 Educat­ Income ion level Wth'hld Occupa­ Com­ tion/ pliant Income Peers Source Status Ethics + + + Friedland, Maital, and Rutenburg, 1978 ¥ - - Spicer & Becker, 1980 0 ■f 0 0 Fair­ Com­ Agency ness plexity Contact Probab. Sanc­ of tions Detect. + 0 0 0 0 + - - Friedland, 1982 Spicer & Thomas, 1982 0 Chong, 1984 - Jackson & Jones, 1985 0 Mi 11 iron, 1985 0 0 0 0 0 0 + 0 - 0 0 0 0 + + 0 0 + 0 + + .0 0 ♦ Spicer & Hero, 1985 Kaplan & Reckers, 1985 Tax Rates + 0 ANALYTICAL STtBIES Author/ Date Age Gender (Fem) Educat­ Income Level ion A U Ingham & Sandro, 1972 Wth'htd Occupa­ Com­ Income tion/ pliant Source Status Peers Ethics Fair­ Com­ Agency plexity Contact ness - Probnb. of Sanc­ tions Detect. + + + + + #- + Groves, 1958 Srlnivasan, 1973 Yitzhaki, 1974 0 Hork, 1975 Clotfelter, 1983 *■f - - ♦ - + - - Cox, 1984 0 Cowell, 1985 Witte & Woodbury, 1985 Tax Rates + ♦ + Madeo, Schepanski, & Uecker, 1985 +# - + + ■f + +t- + ,- + + + + + Positively associated with compliance - Negatively associated with compliance 0 Link to complionce is indeterminate Multiple syniwls indicate that the association with compliance differs for different segments of the taxpayer population. Source; Jackson and Mill iron [1986] +f- 0 32 2.1.3 Discussion of Economic and Demographic Factors Although all throe categories of factors are an important part of the noncompliance decision, only certain variables will be analyzed in this study. economic and demographic A review of previous research related to the variables of interest follows. Gender. Most studies to date have determined that men are generally less compliant that women. Mason and Calvin [1984] through a personal interview survey of 800 Oregon adults found men significantly more 1 ikely to evade. Of those who admitted any one form of evasion (overstatement of deductions, underreporting of income, and failure to file) 57 percent were men. They also concluded that the proportion of women who did admit an act of evasion was substantially higher than the proportion of women who are charged with other forms of crime. Spicer and Hero [1985] in a lab experiment using 36 University of Colorado undergraduate students found that men tended to evade more taxes than women. Tittle difference [1980] has proposed that part is attributable to women's of the roles conservative 1ifestyle, and better moral s. in a 1984 study, classified for this as conformers, a more More recent research has attempted to focus on why a difference exists. Glaser, reason women Grasmick, Finley and as traditional or nontraditional, and then, control 1ing for age compared these groups to males in terms of self-reported involvement in illegal behavior. Their results among conclude nontraditional category cheating). -- that the level of anticipated females approaches the level economic offenses (including criminality among males in only one illegal gambling and tax 33 Several studies have found no difference between men and women. Song and Yarbrough [1978] in a survey of about 300 residents in a North Carolina city found no difference between the tax ethics of males and females. (see, Other survey research has yielded similar inconclusiveness for example, Westat [1980(b)]; Yankelovich, Skelly, and White [1984]). Since the more recent studies have generally found no significant difference between men and women, it may be that the differences noted in the earlier studies may be vanishing. females to be less compliant. One study, in fact, has found However, Friedland, Maital & Rutenburg [1978] evaluated undergraduate Israeli students, and as a result, this finding is not likely to apply to women in the United States. Income Level. Income level (usually defined as adjusted gross income or total positive income) has been used to explain tax evasion because of the belief that it affects the opportunity to underreport income. This certainly is the case for income from nonwithholding sources (e.g., self-employment, certain interest, rentals, etc.), and these types of income are likely to be concentrated among the affluent. However, much of the research results to date conflict -- not only as to whether level of income is a significant compliance variable, but also which type of taxpayers arc more likely to evade -- high income or low income. Groves [1958] estimated that Wisconsin landlords and farmers underreported their income by as much as 50 percent and 25 percent, respectively. However, many low-paying occupations also furnish sources of nonwithholding income (e.g., certain service workers -- maid, child care, etc.). 34 Clotfelter [1983], in a study of TCMP data, found that income level had a significant effect on income underreporting, with noncompliance increasing with income. Allingham and Sandmo's initial tax evasion model [1972] reached the same conclusion. Friedland, Maital, and Rutenberg [1978] came to the same result in a simulation study of income tax evasion. Other research has found no significant effect between level of income and noncompliance. In many of these cases, the researchers have concluded that there is an effect, but that it is curvilinear, with middle income taxpayers taxpayers noncompliant very compliant, and low and upper income (see, for example, Witte and Woodbury [1985], Mason and Lowry [1981], Grasmick and Scott [1982], Spicer [1974], and Spicer and Becker [1980]). If, in fact, there is a curvil inear rel ationship between income level and noncompliance, the data must be transformed prior to the use of 1 inear model s. Without such a transformation, results may be indeterminate or only find a significant relationship with one segment or the other. Most researchers have strongly suggested an examination of the data prior to analysis to see if such a transformation is warranted. Other research has concluded that income level alone is not a significant variable, but may be significant when coupled with occupation and income source (i.e., differing income levels are associated with different types of income earned). Frank and Dekeyser-Meulders [1977] found that high income professionals in Belgium had the opportunity to evade because of the type of income they earned (cash sources; selfemployment.). They found that low income taxpayers derived the majority of their income from sources where there was withholding of taxes, and 35 had less of an opportunity to evade. This concept of "opportunity" was carried over to the United States in survey research by Yankelovich, Skelly, and White [1984]. The researchers built an opportunity variable with occupation (self-employed, level ($30,000 business, professional, or sales), income or greater), and access to cash characteristics. Their research found income a significant sources and as positive relationship between opportunity and noncompliance. Occupation. the compliance The occupation of a taxpayer has been associated with decision longer than any other variable.. Originally, noncompliance was associated with only professionals and taxpayers with high levels of income (i.e., only those occupations with levels were associated with noncompliance). high income However, in today's society, because of the many different occupations, sources of income, and income levels where tax is applied, earlier associations can no longer be relied upon. Research to establish a link between occupations and compliance has not yet reached a consensus. Westat [1980(b)], in a survey of 500 taxpayers, concluded that the occupational categories of professional, sales, management/executive, and certain associated with low levels of compliance, service occupations were and blue-collar employees (generally, trade unions) were associated with higher compliance levels. Witte and Woodbury [1985] found taxpayers employed in manufacturing positions to be more compliant. On the other hand, Song and Yarborough [1978] found no difference between compliance and employment status. And, of the three types of noncompliance they evaluated (overstatement of deductions, underreporting 36 of income, and failure to file) Mason and Calvin [1978] found "occupational prestige" only associated with failure to file. Enforcement Agency Contact. the form of letters requesting Contact by the enforcement agency, in additional information, notices of correction, telephone inquiries, or audits, has received a great deal of interest over the past few years. Some research has led to the conclusion that as contact increases, compliance decreases. Spicer and Lundstedt [1976] concluded that taxpayers who had been audited were more likely to evade than other taxpayers. The reason for this relationship was not entirely clear, but they felt that suffering the penalties or embarrassment of an audit might lead to increased resistance to the tax laws and more evasion. [1969] Strumpel in an evaluation of European countries, noted that stringent assessment (a form of punishment) may lower compliance and willingness to cooperate. On the other hand, Witte and Woodbury [1985] determined that audits result in increased compliance, and noted that the reduction of IRS personnel devoted to audit and other taxpayer contact activities has led to increased noncompliance. audit, Spicer and Hero [1985] found that after an taxpayers are more compliant because of concerns over being audited again, feeling that they will be watched more closely because of their past noncompliance. They concluded that random audits would lead to significantly higher levels of compliance among those audited. Westat [1980(b)] had mixed results in its analysis of TCMP data. Although the analysis showed a significant increase in reported income and related tax payments in taxpayers recently audited, the compliant reporting behavior returned to pre-audit levels within three years. In 37 addition, in a survey of taxpayers who had contact with the IRS, Westat found that about 70 percent of taxpayers were more conscientious after the audit, while the other 30 percent stated that they became more aggressive. Tax Rates. It seems logical to conclude that as the rate of tax on income increases, the compliance level will decrease. Tax legislation over the past few years (particularly the Tax Reform Act of 1986) has adopted this logic to a certain degree. In an attempt to increase compliance, the TRA of 1986 dramatically reduced marginal tax rates and eliminated various deductions. Because of Congressional attempts to increase compliance, and discussions about instituting a flat tax system, academic researchers have included this variable in numerous studies. Allingham and Sandmo [1972] in their model of the compliance process that, for a risk:neutral individual, maximization of expected utility implies that evasion will increase. tend to increase as marginal tax rates Srinivasan [1973] generally supports this proposition. Clotfelter [1983], in a study of TCMP data, concluded that there was an increasing level of noncompliance as tax rates increased. The elasticities he reported indicated that a 10 percent cut in the tax rate would lead to a 5 percent to 8 percent increase in reported income. Cox [1984] however, argues that Clotfelter's finding may be a reflection of level of income rather than tax rates. Since a taxpayer's marginal rate is determined by the level of income, the results could be interpreted as showing that compliance decreases as income increases. Influence of Tax Return Preparers. According to the Department of Treasury, over 40 percent of all individual taxpayers use professional preparers [Department of the Treasury (1985)]. A logical question is 38 what impact, if any, these individuals have on compliance. research to date has been inconclusive. Empirical Westat [1980(b)], however, reported a significant and positive impact on compliance among taxpayers who used professional return preparers. In a recent study, Long and Caudill [1987] reported that upperincome, elderly, and self-employed taxpayers are more likely to use a tax return preparer than other taxpayers. that professional In addition, their study showed tax return assistance is directly related to the complexity of the tax return and the marginal tax rate. Finally, the income tax liability of taxpayers who used tax return preparers was significantly lower than self-prepared returns with the same income, filing status, and other characteristics. Geographic Location. Geographic location of the taxpayer has also been posited as part of the noncompliance decision. Clotfelter [1983] analyzed six major regions of the country using TCMP data and found that for non-farm returns, compliance was best in the northeast and central region, and worst in the west, southwest and southeast regions. returns, the midwest and the southeast had the lowest For farm 1evels of compliance. Witte and Woodbury [1985] (also using TCMP data) concluded that higher levels of noncompliance were associated with areas of high unemployment and poverty, in better educated areas, and in areas with large that student populations. In addition, they concluded noncompliance was lower in middle class areas generally inhabited by whites. 39 2.1.4 Conclusion As is evident from the previous discussion, and the information in Table 2-2, there is little agreement in the literature regarding the relative salience of identified compliance factors and the manner in which these variables are related to tax compliance. Additional research is needed at both the theoretic and empirical levels. Empirical testing of the factors that affect the noncompliance decision is needed to determine which are a part of the decision making process. 2.2 Methodological Considerations - Extensions of Analytical Research Given the present state of analytic tax compliance research, it appears that efforts in shaping a proper tax compliance model should be directed towards two areas: 1. Refining methods of building models, and 2. Testing factors against empirical data. Refining Methods of Building Models. Allingham and Sandmo [1972] were the first to build a basic model of tax compliance, built around a risk-neutral, utility maximizing individual. They conclude their research by acknowledging the simplicity and inadequacy of their tax behavior model. They then encourage other academic economists to expand the model in a number of areas, including labor supply, or taking the model into the area of optimal taxation theory. basic Allingham uncertainty, and Sandmo model and In 1985, Cowell took the analyzed labor supply under establishing a framework for discussing various evasion opportunities and related enforcement actions. 40 Testing Hypothesized Noncompliance Factors Against Empirical Data. Progress in this area started with Mork [1975], who took the economic models of Allingham and Sandmo [1972] and Srinivasan [1973], empirically tested them using income tax data from Norway. analysis, he concluded that relative risk aversion and From this is a decreasing function of income. In the last few years, analytical models have been tested against TCMP data (see, for example, Clotfelter [1983], [1985], and Madeo, et a l . [1985]). Witte and Woodbury The findings of the Clotfelter and Witte and Woodbury research have been previously discussed in this chapter. The Madeo, et a l . [1985] research involved the construction of a judgment, model of taxpayer behavior based on a survey of certified public accountants in attendance at a continuing education seminar. Based on these responses, the model included variables for income level, income source, tax penalties, and tax rate structure. Subsequent to construction, the model was tested against TCMP data and yielded a .92 correlation. Their conclusion was that CPA tax professionals have a great deal of insight when it comes to properly modeling the compliance process. The Next Generation of Analytical Research. The next generation of analytical research can best be summarized in the following researcher's comments: "Progress in theoretical research may depend on stratifying taxpayer groups before performing the economic analyses. Such stratification depends on defining relevant segments of the taxpaying population, identifying their risk behavior, and measuring the gains and penalties they face when contemplating noncompliance." Witte and Woodbury [1985, p. 10] 41 "Further advances in the economic modeling of noncompliance decisions appear to be predicted on increased behavioral realism of the factors involved in the noncompliance decision. This added insight, necessary to develop more meaningful models, is likely to be obtained either through analysis of taxpayer data bases (such as the TCMP file research of Clotfelter [1983] and Witte and Woodbury [1985]), or behavioral lab studies (such as the research of Chang [1984] and Jackson and Jones [1985])." Jackson and Mill iron [1986, p. 30] This study is designed to provide added insight into the factors that are a part of the noncompliance decision. It is hoped this insight will be beneficial in advancing the economic modeling of this intricate decision making process. CHAPTER 3 DATA COLLECTION. CREATION OF DATA BASE. AND LIMITATIONS OF DATA 3.1 Data Collection - In General During the amnesty program, almost 75,000 taxpayers called "amnesty taxpayers") filed tax returns. (henceforth As part of the processing procedures, the submitted returns were reviewed by Department of Treasury staff for computational accuracy. However, no other procedures were performed during this process (i.e., the returns were not audited or reviewed with the intention of challenging the disclosed information). There was, however, a desire by the Department of Treasury to investigate the participants further. As a result, 10 percent of all amnesty taxpayers were randomly selected in order to collect information available in the materials (primarily tax returns) submitted by these taxpayers. This information would be compiled in a data base, and made available to the Department of Treasury and academic researchers for further analysis. The data collection Amnesty Data Base (MADB) efforts, are and construction of the topics discussed in this theMichigan chapter. Although there is a great deal of information in the MADB, the amnesty materials submitted bv taxpayers are the only source of information in the data base at the present time. returns were checked only provides this information. As mentioned above, the submitted tax for computational errors, and the data base It is possible, however, that amnesty taxpayers disclosed only a portion of their unreported income. There is no way of assessing the level of this nondisclosure. 42 43 Under the amnesty program, taxpayers with accounts receivable outstanding were allowed to pay these amounts without additional filings. Although these taxpayers were included in the 10 percent random sample, there was very little information to be collected from the amnesty materials. As a result, these taxpayers were set aside during the data collection efforts. Sample Selection. Two amnesty return samples were selected during the amnesty program: 1. Random Sample: A random sample of every tenth taxpayer who filed for amnesty was selected. Taxpayers who filed for amnesty were required to submit an amnesty application form and related returns. Each package was processed, batched, and dated by the Department of Treasury during the amnesty program. The 10 percent sample was drawn from these processed amnesty filings, and the data base constructed from these returns is known as the Michigan Amnesty Data Base (MADB). 2. Large Tax Paid Sample: In addition to the random sample, the following returns also have been selected and separately coded: a. b. c. All taxpayers with individual income tax payments of $10,000 or more, All taxpayers with single business tax payments of $20,000 or more, and All taxpayers with sales/use tax payments of $20,000 or more. Although the large tax paid sample may be of interest for future research, it has been set aside for the present. This study uses the individual income tax portion of the MADB as the basis for constructing an appropriate individual research data base. 1imited to individual income tax returns, the Since this study is discussion of data collection and entry procedures used in constructing the Michigan Amnesty Data Base is 1imited to those returns. A summary of all the returns filed under amnesty, and the 10 percent MADB sample (excluding accounts receivable participants) is presented separately as Table 3-1. As can be seen from the table, there are 2,985 44 individual income tax returns in the MADB sample. Table 3-2 presents a summary of amnesty collections by type of tax for the entire amnesty program, and the 10 percent MADB sample receivable participants). (again, excluding accounts The returns in the sample were segregated by the Department of Treasury during completion of the MADB. Subsequent to data collection and coding, the Department of Treasury filed all returns filed during amnesty with non-amnesty returns. were destroyed. The amnesty applications 45 TABLE 3-1 AMNESTY PARTICIPATION BY TAX, EXCLUDING ACCOUNTS RECEIVABLE Entire Amnesty Program Michigan Amnesty Data Base Tax % of Tax % of Tax Tvoe A d d !icants Returns Returns Applicants Returns Returns 20,496 69.13% 71.02% Individual 32,614 1,948 2,985 606 Withholding 1,045 2.22 37 65 1.55 12.16 Intangibles 2,311 5,361 11.36 230 511 Sales/Use 5.10 5.28 1,395 2,409 114 222 Single Business 2,487 5,130 10.87 207 420 9.99 1.30 Miscellaneous 616 Total 47.175 100.00% 4.203 TABLE 3-2 AMNESTY COLLECTIONS BY TAX, EXCLUDING ACCOUNTS RECEIVABLE Tax Tvpe Individual Withholding Intangibles Sales/Use Single Business Miscellaneous Total Tax Tvpe Individual Withholding Intangibles Sales/Use Single Business Miscellaneous Total Entire Amnesty Program Tax Interest Total Amount Amount Amount (Audited) (Audited) (Audited) $11,566.4* $ 2,041.6* $13,608.0* 134.7 1,084.0 1,218.7 5,009.5 865.8 5,875.3 9,227.9 2,068.8 11,296.7 10,395.5 2,013.4 12,408.9 174.8 25.0 199.8 $37,458.2 $ 7.149.3 $44,607.1 Michigan Amnesty Data Base Tax Interest Total Amount Amount Amount (Audited) (Audited) (Audited) $ 1,292.1* $ 213.9* $ 1,506.0* 14.0 87.8 73.8 351.0 80.0 431.0 150.7 989.4 838.7 1,199.8 315.0 1,514.8 $ 3.755.4 $ * Amounts in Thousands of Dollars 773.6 $ 4.529.0 100.00% 46 Returns not a part of the MADB or the Large Tax Paid samples also were separated from the amnesty application and filed with non-amnesty returns. The amnesty applications were shredded. Some general information regarding the nonselected returns remains in the Department of Treasury computer system [e.g., taxpayer name and address, identifier, type of tax, return year(s), tax paid, and interest paid]. No other information is readily available for these taxpayers, although the actual tax returns filed during amnesty could be requested if needed. As mentioned in Chapter 1 (Table 1-9), 20,496 taxpayers filed for individual income tax amnesty. Of these taxpayers, 2,115 were included in the original 10 percent random sample. of these taxpayers. Data was collected from 1,948 Information from the remaining 167 taxpayers was not collected for a variety of reasons. Some of these taxpayers filed for amnesty and owed no tax 1iability, and any information submitted with the amnesty application was often destroyed. Other taxpayers were denied amnesty and, therefore, excluded from the data collection process. Based on comparisons between data in the final individual income tax data base (IITDB) and the population of amnesty participants (as detailed in Table 3-3), the IITDB appears to be representative of the population of all individual amnesty taxpayers. Table 3-3 provides two sets of comparisons. The first part of the table compares the table compares the distribution of amnesty payments by size, detailing the percentages of the amnesty population and amnesty taxpayers in the IITDB making payments in the specified dollar ranges. Based on a review of this information, it appears that the IITDB and population of all individual amnesty taxpayers are similar. The only significant difference between the population and the IITDB is for 47 amnesty payments of $100 or less. This difference could be due to the fact that taxpayers who owed no amnesty tax were excluded from the IITDB, although part of the original 10 percent sample. The second part of the table examines the number of years filed by amnesty taxpayers in the population and IITDB. This information also appears to be similar, with the only significant difference being that the IITDB does not include as many taxpayers who filed for 10 or more years. TABLE 3-3 INDIVIDUAL AMNESTY INCOME TAX PARTICIPANTS: COMPARISON OF DATA BASE TO TOTAL POPULATION Percentage Distribution of Amnesty Payments by Payment Size Population IITDB $100 or or Less 38.7 34.0 $100.01 to $500 37.2 38.4 Ranoe (In Dollars) $500.01 $1,000.01 to $1,000 to $10,000 10.6 12.9 12.1 14.6 More Than $10,000 0.68 0.88 Percentage Distribution of Number of Years for Which Amnesty Was Filed 1 Population IITDB Source: Summary. 71.0 72.5 Number of Years 2 to 4 5 to 9 25.0 23.7 3.7 3.7 10 or More 0.34 0.15 Goddeeris. Martin, and Youna f19881. The IITDB appears to be representative of the population from which it was drawn. Although the data base provides a great deal of information about certain amnesty participants (i.e., those participants without outstanding accounts receivable balances prior to amnesty), the 48 IITDB only includes information on these amnesty participants. It does not include information on tax evaders who chose not to participate in the amnesty program. These evaders may have different characteristics than those amnesty taxpayers in the IITDB. Therefore, the make-up of the IITDB may not be reflective of all tax evaders in Michigan. Construction of the Michigan Amnesty Data Base. A separate data base has been constructed from the 10 percent random sample for each of the tax types identified in Table 3-1. These separate data bases (which in total make up the Michigan Amnesty Data Base) can be 1 inked (using a special identifier developed by the Department of Treasury) across tax types. The data bases consist of 2 sets of variables: 1. Common variables - variables consistent across all types of tax, and 2. Unique variables - variables unique to a type of tax. Composition of the data bases was determined based on previous academic research and in discussions with Treasury Department officials over a six week period during July and August 1987. occurred Data collection at various times during the months of August, October, and November 1987. September, A more detailed discussion of the variables selected for the individual income tax data base and the data collection procedures related to that portion of the Michigan Amnesty Data Base follows. Purpose of Data Base. The Michigan Amnesty Data Base is intended to benefit both the Department of Treasury and academic researchers. The Department of Treasury will be using the data base to provide descriptive information about amnesty filers, refine their audit selection process, and target their enforcement efforts. 49 Academic research (including this dissertation) will investigate various factors that are a part of the noncompliance decision process. A natural extension of this research will be an attempt to model the noncompliance decision process. During this variables were not sought for the taxpayers Michigan Amnesty Data Base. research, that In future research, additional are a part of the other sources of information may be brought into the research process (e.g., TCMP data, additional Department of Treasury information, other states which have run amnesty programs). The Department of Treasury has committed itself and its resources to understanding the noncompliant taxpayer. To that end, the Department is in the process of developing a policy regarding approval of academic research projects using the Michigan Amnesty Data Base or other data. As part of this process, the Department has developed a new identifier for each taxpayer in the data base. This will protect the identity of taxpayers that are a part of the data base and allow future researchers to request additional Department of Treasury information not in the data base regarding these taxpayers. Such a policy would allow, for example, time series analysis of these taxpayers. 3.2 Individual Tax Returns Variable Selection and Data Collection 3.2.1 Determination of Variables Based upon a review of previous academic research in the fields of accounting and economics, a 1 ist of variables to be collected for the IITDB portion of the MADB was submitted to the Department of Treasury for their review. This preliminary list was discussed with the Commissioner of Revenue and circulated to the Individual Income Tax Division for their 50 review and analysis. Based upon these discussions and review, a final list of variables was developed. A process similar to that discussed above intangibles, single business, sales, and use taxes. was In addition, A list of the common variables and unique individual income tax variables is Figure 3-1. for For each tax type, a list of variables unique to the tax was developed. variables common across tax types were identified. followed presented in 51 FIGURE 3-1 MICHIGAN AMNESTY DATA BASE: INDIVIDUAL INCOME TAX RETURN VARIABLES COMMON VARIABLES FOR ALL TAX RETURNS 1. 2. 3. 4. 5. 6. 7. 8. Taxpayer ID Number(s) Zip Code Type of Tax * Tax Return Year Amnesty Tax Paid, As Audited Interest Paid, As Audited Prior to Amnesty Period, No Acceptance Letter Sent Does return contain a letter concerning amnesty? UNIQUE VARIABLES - INDIVIDUAL RETURNS VARIABLES COLLECTED FROM ALL RETURNS; 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. Occupation - Taxpayer Is taxpayer self-employed? Is SE income primary or secondary? Occupation - Spouse Is the spouse self-employed? Is SE income primary or secondary? Residency Filing Status Gender (Single/MFS returns only) Exemptions Adjusted Gross Income (AGI) Additions Subtractions Taxable Income Tax Tax Withheld Estimated Tax Payments Balance Due Is a W-2 present? Treasury Department letter re: no state tax return filed present? Preparation of return 1986 individual return filed? N o t e : Items 31 and 32 collected from returns that were NOT "amended". 31. Additions to Income: ■ Non-MI municipal interest ■ Capital gains ■ Losses from other states 32. Subtractions from Income: ■ US government interest ■ Military benefits ■ Retirement benefits u Income from other states PROPERTY TAX CREDIT INFORMATION: The following information was collected from MI-1040CR, if available. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. Rent or own home Salaries, wages, tips, etc. Interest and dividends Rent, royalty, business income Annuity and pension benefits Net farm income Capital gains less capital losses Alimony, other taxable income Child support Work er’s compensation Household income Property taxes paid Rent paid 52 FIGURE 3-1 (Continued) VARIABLES RETURNS2 ; COLLECTED FROM AMENDED N o t e ; The following amounts are as ORIGINALLY REPORTED. Note; The following amounts are detailed. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. Residency Filing Status Exemptions Adjusted Gross Income Additions Subtractions Taxable Income Tax Liability Property Tax Credit Tax Withheld Estimated Tax Payments Amount paid with original return Refund shown on original return Reason for change in number of exemptions Note; 60. 61. 62. 66. The following information was inferred from reading Part VII of MI-1040X. AGI Errors (maximum of 2) 3 Addition Errors (maximum of 2) ^ Subtraction Errors (maximum of 2) 5 Reason for amended return was IRS audit FOOTNOTES 1 Tax T y p e ; 01=Accounts Receivable; 0 2 = Individual; 03=Sing1e Business Tax; 04=Int.angibles; 05=Sales; 06=Withholding; 07=Inheritance; 08=Miscellaneous; 09=Use 2 Amended returns are defined as those with some nonzero numbers in column A of the MI-1040X. 3 AGI Error C ode s; l=Wages; 2=Interest or dividend income; 3=Rent or net business income; 4=Capital gain or loss; 5=0ther ^ Addition Error Codes; l=Non-MI municipal 3=Losses from other states; 4=0ther interest; 2=Capital gains; ^ Subtraction Error C ode s; l=Income from US government obligations; 2=Mil1tary benefits; 3=Income from other states; 4=Retirement benefits; 5=0ther 53 3.2.2 Data Collection Procedures and Verification of Data Data Collection Planning. After the final list of variables had been identified, data collection procedures were analyzed. A two-step data collection process was proposed and accepted by the Department of Treasury. First, a team of individuals selected by the Department would transfer the data from the returns to a document. Second, the data collected would be entered from the data collection documents into a data base by data entry personnel. Although it would have been possible to collapse this two-step process into a single step (i.e., having the individuals collecting the data enter it directly into the data base, thereby eliminating the process of transferring the data from the tax returns to a collection form), it was felt that the two-step process provided added benefits. First, upon completion of data collection, a separate copy of the data remains. Second, the two-step process made verification of the data collected easier. A four page data collection document was drafted and printed. In order to ensure that data was drawn in a consistent manner from the returns selected, a data collection procedures memorandum was drafted that explained the data collection process. In addition to a 1 ine by 1ine discussion of the data collection document, the memorandum included Michigan income tax returns from 1978 to 1984 coded to match the data collection document. This provided a visual means of assisting the data collection efforts, allowing the personnel to match the return selected to a coded return form. By laying the actual tax return next to the coded return, information to be collected could be easily identified and matched to the data collection document. 54 Data Collection. The Department of Treasury selected ten individuals to collect the data under the supervision of the author, John Goddeeris (Associate Professor, Department of Economics), and Stan Borawski (Michigan Department of Treasury). The staff consisted of five individual income tax auditors (average experience of ten years), and five student interns. A conference room in the Treasury Building was used for data collection, and the individuals were seated around two large conference tables. Student interns were placed in between the auditors and encouraged to ask the auditors for assistance, if needed. In addition, all were encouraged to ask questions. A two hour training session took place on the first morning of data collection. In addition to reviewing the data collection memorandum and related documents, information from several practice returns were coded onto data collection documents by the individual s. training session, emphasized. the qua!itv of the data As part of the collection efforts was In addition, as time was not a factor, the personnel were encouraged to be careful and consistent in their data collection efforts. As a fol 1 ow up to the training, question and answer sessions were held each morning during data collection. Three different forms of quality control took place during data collection. First, at the close of each morning and each afternoon, the data collection personnel took 20 minutes to review another person's collection efforts. Returns were randomly selected and pertinent information was traced from the return to the data collection document. Errors found were corrected, and if a systematic error occurred, the error was pointed out. After each of these reviews, personnel were asked if they had any questions or had noted anything that would be relevant 55 for the group. Second, on several evenings, the supervisors randomly selected returns for comparison to the related data collection document. Few errors were noted during this process, and those discovered were corrected. Finally, quality review, the five student interns performed an overall spending four hours randomly selecting returns and comparing tax return information to the related data collection document. In all, over 850 hours were spent collecting and reviewing the information during eight working days. Data Entry. Upon completion of data collection, four data entry specialists were chosen by the Treasury Department to load the collected data into the data base. memorandum was written. In preparation for this process, a second The memorandum briefly explained the Michigan Amnesty Data Base and the type of information that is part of it. It also contained a detailed explanation of the data entry process, verbally explaining the process and visually correlating the data collection document to the computer screens within the data base. A two-hour training session was held with the data entry personnel on the first day of data entry. The session included a verbal explanation of the process, a review of the memorandum, and an hour of practice loading dummy data into a data base that replicated the actual data base. Data entry took place over a five day period. A quality review was performed daily, examining data entered on the previous day for errors, duplications, and/or omissions. After completion of the data entry, every twentieth record in the data base (5 percent) was printed out and compared to the related data collection document for errors. Errors noted minor. during both of these quality reviews were generally 56 However, based on the review of the 5 percent sample, several errors were noted in the amended return data of one of the data entry clerks. As a result, all amended return information entered by this individual was printed out and compared to the corresponding data collection documents. The systematic error that had been noted in the initial 5 percent sample did not replicate itself in the other records. Summary. The training sessions with staff, participative management of the project, and the many quality control reviews established have served to create a data base which is reflective of the entire population of amnesty filers in Michigan. As previously mentioned, the individual portion of the Michigan Amnesty Data Base was used as the basis for constructing an appropriate research data base. During the construction of the research data base, a detailed review of the selected data was performed prior to analysis. descriptive statistics, This review, accomplished by the use of a general review of the data, and tracing questionable items back to the Department of Treasury files, yielded 133 errors in 44,598 data points. analysis. The noted errors were corrected prior to Chapter 4 contains a more detailed discussion of this review. A complete copy of the various documents and memoranda used in the data collection and coding process can be found in Appendix A. 3.3 The Advantages and Limitations of the Michigan Amnesty Data Base The most significant advantage of Michigan Amnesty Data Base over Taxpayer Compliance Measurement Program (TCMP) data is the inclusion of information from taxpayers omitted from the TCMP (i.e., taxpayers who had not filed returns previously and taxpayers who had not made a full disclosure of income). It appears that the MADB is unequaled as a source 57 of data on amnesty participation and effects. Most analyses of state tax amnesty programs published over the past few years rely on aggregated data provided by state revenue agencies [1986]; Parle and Hirlinger [1986]). (see, for example, Mikesell These analyses provide very little detail about the various types of noncompliance uncovered during amnesty or their relative importance. Michigan resources appears necessary participants. to be the only state to have committed the to The data compile detailed contained information in the MADB on provides amnesty a unique opportunity to analyze the various factors that are a part of the noncompliance decision for amnesty participants. Although these are significant improvements, several comments regarding the limitations of the data base also are warranted. These 1 imitations may ultimately lead to an expansion of the data contained in the Michigan Amnesty Date Base. In addition, these 1 imitations provide some guidance for future research. First, the data base includes only taxpayers from Michigan. Generalizability of research findings using this data to the national population is not warranted without further investigation and analysis. Although this 1 imitation applies to this study, it is interesting to note that most of the noncompliance behavioral research done to date has used geographic-specific data. Second, as previously mentioned, the taxpayers in the data base are a subset of all tax evaders and delinquent taxpayers (i.e., those evaders who chose to disclose themselves during the amnesty program). In addition, it does not necessarily include all the taxable income of amnesty taxpayers (i.e., these amnesty participants may have chosen to disclose only a portion of this income). It may be that the MADB is not representative of all tax evaders in Michigan. Finally, the data base does not include any psychological or attitudinal variables related to the noncompliance decision. CHAPTER 4 CREATION OF THE RESEARCH DATA BASE AND A DESCRIPTIVE ANALYSIS OF THE DATA 4.1 An Overview of the Individual Tax Participant Table 4-1 provides descriptive statistics on Michigan individual income tax amnesty participants. These statistics are derived from the 1,948 taxpayers and 2,985 returns included in the individual income tax data base (IITDB). A complete descriptive analysis of the IITDB on a taxpayer basis has been performed and may be consulted if additional information is needed (Goddeeris, Martin, and Young [1988]). Table 4-1 is a summary of that analysis. For each of several variables, the mean and standard deviation are presented. In addition, this information is compared to data on the population of 1984 individual income taxpayers provided by the Michigan Department of Treasury. The table also includes information on the number of years filed for amnesty, the number of exemptions claimed, a percentage distribution of the filing status indicated, and for single taxpayers, the percentage of males. The average total amnesty tax payment (combining all years for which the taxpayer filed) was $664.82. $108.46. The average interest payment was The average adjusted gross income (AGI) figure reported on returns was $56,047.41. The AGI amount contrasts with an average AGI reported on all 1984 Michigan income tax returns of $23,384. Without further investigation, this difference would appear to indicate that the typical amnesty participant had a relatively high income. However, by examining the standard deviation for AGI (as well as amnesty tax and interest amounts), it is evident that considerable variability exists. 59 60 TABLE 4-1 SUMMARY CHARACTERISTICS OF MICHIGAN INDIVIDUAL INCOME TAX AMNESTY PARTICIPANTS (BY TAXPAYER) Standard Deviation Mean Characteristic Amnesty Tax Paid $ 664.82 Interest Paid $ Number of Tax Years 2,253.46 - 108.46 502.73 - 1.53 1.16 - Tax Per Year $ AGI Per Year $ 56,047.41 Median AGI $ 20,600.00 $ Comparison Mean Values From 1984 Taxpaying Population 432.36 Percent Not Full-time Resident 4.90% Number Of Exemptions 2.46 NA 628,397.05 NA NA 1.51 $ 870 $ 23,129 $ 18,400 NA 2.44 Filina Status: Married Filing Jointly Married Filing Separately Single Of Single, Percent Male 49.50% 1.95% 48.55% NA NA NA 64.80% NA Source: Goddeeris, Martin, and Young (1988). 50.00% .90% 49.10% NA 61 Specifically, it would appear that the means may be heavily influenced by small numbers of very large values. To further explore this possibility, median values were also computed for each of these variables. The median figures computed were less than half the corresponding means. The median total amnesty payment (combining both amnesty tax and interest) is $196, well below the mean total of $773.28 ($664.82 in amnesty tax and $108.46 in interest). The median AGI per year in the amnesty sample is $20,600, or less than half the mean. This is still about $2,000 higher than the median AGI reported on all 1984 Michigan income tax returns ($18,400), and indicates that amnesty participants did have a marginally higher level of income than the 1984 taxpaying population. The individual. amnesty participant in the IITDB filed for about 1.5 years. This is consistent with the information reported in Table 3-3, indicating that about 71 percent of amnesty participants only filed for one year. As would be expected, most participants were full-time Michigan residents in the tax years for which they filed (only 4.9 percent were part-year residents or non-residents). Of those taxpayers filing as single individuals, about 64.8 percent were male. This percentage is higher than the percentage of males in the labor force in 1984 (about 56 percent according to the U.S. Bureau of Labor Statistics). Other data presented in the table (percentage of filing status for married couples filing a joint return and single individuals; exemptions claimed) are very similar population in Michigan. to the values from the 1984 taxpaying 62 4.2 Creation of the Research Data Base 4.2.1 General Discussion A total of 2,985 data lines (one for eachtax return) exist in the individual portion of the Michigan Amnesty Data Base. Of these data lines, two classes of returns were segregated out of the data base in creating the research data base (RDB) used in this study. No tax return data were available for some amnesty participants when the data were collected, coded, and keyed into the data base. these taxpayers filed an amnesty application and paid the Either tax and interest due but did not file a state tax return, or the state tax return was in use for some other Treasury procedure (audit or review) at the time of data collection. These returns were identified as profile only returns, in that only certain profile information was available (name and address, social security number, tax years covered by amnesty application, and the amount of tax and interest paid during amnesty); other economic characteristics were not available. In addition, certain amnesty returns in the sample showed a zero amnesty tax payment, or refund due the taxpayer as a result of filing. Other 1 ines of data were unusable due to a lack of other data. These 1ines were also removed during creation of the research data base. In total, these data 1ines comprised 515 returns and $283,058 in amnesty tax. It should be noted that the average amnesty payment of these returns is $549.63 -- more than 25 percent higher than the average amnesty tax paid for the entire sample ($432.36). These returns, not analyzed in this study, may contain important information if analyzed. As a result, a later study may request that the Department of Treasury search for returns related to these data lines. If the returns can be 63 located, and data collected, this information could be incorporated into the present RDB, and the present study expanded. When these two groups of returns are removed, 2,470 lines of data are left, forming the primary research data base used in this study. 4.2.2 Verification of Data and Error Correction Procedures Once the 2,470 returns had been secured in a separate data base, a series of range and accuracy tests were performed on the data. These tests were primarily performed to assess the accuracy of the data. As mentioned in Chapter 3, a great deal of time was spent on the data collection and data entry processes, supervision, and accuracy checks. including significant training, However, no specific judgement on error rates in the data were made. It is important to note that the range and accuracy test performed did not encompass the entire RDB, but only focused on variables that were of specific interest in this study. The various tests performed on the data include the following: 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. Amnesty tax paid greater,than tax computed on return; Tax divided by taxable income greater than 6.35 percent; Occupation (taxpayer or spouse) coding greater than maximum possible (33); Tax year prior to 1970; Comparison of gender and filing status; Zip code in Michigan, but 1isted asnon-resident; Zip code outside Michigan, but1isted as resident; Taxable income greater than AGI; Data 1ine missing AGI; and Data 1ine missing total tax. Upon completion of these tests, a list of questionable items was compiled, by return. As previously mentioned, after the completion of data collection and data entry, the Department of Treasury, the original returns were filed and the amnesty applications were destroyed. The 64 questionable data noted during the range and accuracy tests were traced bacy. to the original documents in Department of Treasury files. During this process, about 44,598 data points were subjected to examination, excluding missing values. The review process disclosed 133 errors in the data. The errors were corrected in the research data base used in this study. However, the Michigan Amnesty Data Base has not been changed. Instead, researchers requesting the data will be provided with a list of the errors in the data. 4.3 Summary Characteristics of the Research Data Base 4.3.1 An Overview of the Research Data Base Table 4-2 provides some descriptive information about the entire research data base. The information is similar to that previously presented on a participant basis (note, again that the research data base is on a return basis -- one taxpayer could have several lines of entry in the research data base). The average amnesty tax payment is $408.51, while the average adjusted gross income (AGI) figure reported on returns was $57,629.49. As in the by participant information, the amnesty tax and AGI amounts have a standard deviation several times larger than the mean. The median values computed for each of these variables resulted in lower figures, less than half the corresponding means. amnesty payment is $138.06, well below the mean. The median tax The median AGI per year is $20,158.25, or less than half the mean. Most returns in the research data base were from full-time Michigan residents (96.42 percent). Only 1.48 percent were part-time residents and only 2.10 percent were non-residents. TABLE 4-2 SUMMARY CHARACTERISTICS OF THE RESEARCH DATA BASE Taxpayer Group Number Research Data Base Prior Contact (Group 2) Average Average Amnesty Tax AGI 2,470 100.00 $ 408.51 $ 57,629.49 571 23.12 463.86 153,433.61 1,899 76.88 100.00 391.87 28,822.66 1,272 51.50 66.98 317.35 35,042.55 627 25.38 33.02 543.05 16,204.33 Amended Returns (Group 1) Non-Amended Returns Percent No Prior Contact (Group 3) Item Number Mean Standard Deviation Median Amnesty Tax AGI Exemptions 2470 2470 2470 408.51 57,629.49 2.42 1,571.76 697,362.20 1.49 138.06 20,158.25 2.00 Characteristic Minimum .34 25.00 0.00 Maximum 64,498 24,115,974 11 Number % Filina Status: Single Married Filing Joint Married Filing Separate 2433 2433 2433 50.10 47.47 2.43 Residency: Resident Part-Year Resident Non-Resident 2428 2428 2428 96.42 1.48 2.10 Gender: Male Female 1228 1228 66.69 33.31 Other Characteristics: Non-filers in 1986 Opportunity to Evade Likely Self-Employed Urban (living in SMSA) Contact Indicated 2470 2010 2470 2470 2457 26.96 2.79 10.00 82.79 74.48 66 The mean number of exemptions is 2.42, similar to both the sample data and the 1984 Michigan taxpaying population. The percentages of returns filed by single individuals is 50.10 percent, while married couples filing joint returns account for 47.47 percent of the research data base, and married couples filing separately account for 2.43 percent of returns in the research data base. filed jointly, Among the amnesty returns not 66.69 percent were filed by males. Several other interesting characteristics of these returns should be mentioned. Non-Filers in 1986. During the data collection process in August 1987, it was suggested to the Department of Treasury that all taxpayers filing a tax return under amnesty (generally for 1984 and earlier tax years) be monitored to ensure compliance in 1986 and subsequent tax years. As a result, the Department cross-matched all individual amnesty participants against the 1986 tax return master file. This match was performed on three occasions, once prior to August 15, 1987 (the extended due date of 1986 returns), once in September 1987, and a third time in late October 1987 (subsequent to the second extended due date for 1986 individual returns -- October 15, 1987). In addition, amnesty filers identification numbers also were matched against a file which contained all extension requests. Letters were sent to all amnesty participants who hadn’t been matched during this process, requesting information as to why a state return hadn’t been filed. One could think of these taxpayers as chronic non-filers (which may be different than chronic evaders). However, this action (of not filing a return) does provide some measure of an attitude regarding compliance. This information was included in the Michigan Amnesty Data Base, and is 67 further analyzed as a part of this study. Of taxpayers in the research data base. 26.96 percent failed to file a return in 1986. Opportunity to Evade. Yankelovich, Skelly, and White [1984] built a variable called "opportunity" from a composite of occupation (self- employed, business, professional, or sales), income level of $30,000 or more, and access to cash income sources. from data in the research data base. A similar variable was created Of taxpayers in the research data base, 2.79 percent are classified as having the opportunity to evade taxes (although they accounted for 10.08 percent of the amnesty tax paid within the research data base). This low percentage would seem to indicate that the amnesty program did not attract many taxpayers with these characteristics. However, there has been no measure of this variable within the individual taxpaying population in Michigan. As a result, no benchmark exists for comparison purposes. should attempt to identify the Future research percentage of taxpayers with these characteristics not only in Michigan, but also in the United States. Likely Self-Employed. In addition to those taxpayers who identified themselves as self-employed in the occupation information, the data was examined to alter this classification. In addition to those taxpayers disclosing self-employment, additional taxpayers were added to this class if: 1. No taxes were withheld, 2. No W-2 was present with the amnesty return information, and 3. The occupation disclosed was not retired, student, deceased, or other. Given this definition, 10 percent of the returns in the research data base are related to taxpayers who are likely self-employed. 68 Standard Metropolitan Statistical Areas (SMSAl/Urban. As part of the process of preparing the research data base for analysis, taxpayers were assigned to Standard Metropolitan Statistical Areas (SMSA) Michigan using the zip code data in the Michigan Amnesty Data Base. in This allows for a reasonable use of the zip code data, creates a smaller number of regions, and also differences in the data base. SMSA Number provides information on urban/rural The following SMSAs exist in Michigan: _ _ _ _ _ _ _ City_ _ _ _ _ _ _ 0440 0780 0870 2160 Ann Arbor Battle Creek Benton Harbor Detroit 2640 3000 3520 3720 4040 5320 6960 FI int Grand Rapids Jackson Kalamazoo Lansing Muskegon Saginaw/Bay City/Midland Counties_ _ _ _ _ _ _ _ Washtenaw Calhoun Berrien Wayne, Oakland, Macomb, Monroe, Livingston, Lapeer, St. Clair Genesee Kent, Ottawa Jackson Kalamazoo Ingham, Eaton, Cl inton Muskegon Bay, Midland, Saginaw Using a zip code directory for Michigan (which provides not only a 1isting of all zip codes, but also the county in which each is located), zip codes for each SMSA were assembled. data then was made, with new independent variables being created (one for each SMSA). A transformation of the zip code In addition, an urban variable (if in a SMSA, the taxpayer is considered urban) and rural variable (if the taxpayer is not in an SMSA) was also created. Of taxpayers in the research data base, 82.79 percent were 1iving in a SMSA. Groupings Within the Research Data Base. For purposes of providing some descriptive information about the RDB, all taxpayers in the research data base (2,470 returns or 82.75 percent of the IITDB) can be assigned to one of three groups, detailed in Figure 4-1 [amended (Group 1), non- 69 FIGURE 4-1 A SUMMARY OF THE MICHIGAN AMNESTY DATA BASE TOTAL INDIVIDUAL INCOME TAX DATA BASE a Returns 2,985 m Amnesty Tax Paid $ 1,292,087 a Average Amnesty Tax Paid $ 432.36 _ _ _ _ _ _ _ RESEARCH DATA BASE_ _ _ _ _ _ _ a Returns 2,470 a Amnesty Tax Paid $ 1,009,029 a Average Amnesty Tax Paid $ 408.51 AMENDED RETURNS (GROUP 1) ■ Returns a Amnesty Tax Paid $ a Average Amnesty Tax Paid $ a a a 571 264,866 463.86 NON-AMENDED RETURNS (GROUPS 2 AND 3) Returns 1,899 Amnesty Tax Paid $ 744,163 Average Amnesty Tax Paid $ 391.87 a a a RETURNS WITH PRIOR TREASURY CONTACT (GROUP 2)1 Returns 1,272 Amnesty Tax Paid $ 403,672 Average Amnesty Tax Paid $ 317.35 RETURNS WITHOUT PRIOR TREASURY CONTACT (GROUP 3) Returns 627 Amnesty Tax Paid $ 340,491 Average Amnesty Tax Paid $ 543.05 a a a a a a TAX RETURN DATA NOT AVAILABLE 2 Returns 515 Amnesty Tax Paid $ 283,058 Average Amnesty Tax Paid $ 549.63 1 Returns With Prior Treasury Contact. Taxpayers were categorized as having prior Treasury contact if state taxes were withheld, or estimated taxes were paid, or a W-2 was submitted with the tax return, or if the taxpayer had been identified by the Department of Treasury as having filed a federal return but not a state return. 2 Tax Return Data Not Available. These taxpayers filed an amnesty return and paid any tax and interest owed, but did not file a state tax return. The Michigan Department of Treasury requested that appropriate returns be filed, but any returns filed as a result of this request were received after data collection occurred. 70 amended with prior contact (Group 2), non-amended with no prior contact (Group 3)]. These groupings are discussed below. A fourth group identified in Figure 4-1 is made up of the 515 returns (17.25 percent of the IITDB) for which no tax return information was available, or for which no amnesty payment was due (these returns are not included in the RDB). Initially, amnesty taxpayers can be divided into one of two groups- those taxpayers filing amended returns (Group 1) and those filing new (non-amended) returns (Groups 2 and 3). During data collection, an amended return was defined as one disclosing original return information, revised information and reasons for the revision, and recomputed tax liability. If an amended return was filed but contained no original return information, the return was classified as a non-amended return (i.e., it was possible for a new filer to use an amended return form to file under amnesty, even though it was not truly an amended return). Based on the definition of amended returns used in data collection, taxpayers classified in this group were clearly known to the Department of Treasury prior to the amnesty program. Participants filing non-amended tax returns as part of amnesty can be divided into two groups (Groups 2 and 3). From the information collected, it appears that the Department of Treasury had access to data that should have identified certain taxpayers as nonfilers. The categorization of amnesty taxpayers into Group 2 or 3 was based on this knowledge of the taxpayer prior to amnesty. Taxpayers were categorized into Group 2 (new returns with prior contact) if one or more of the following were true: 1. The return indicated that taxes had been withheld from the amnesty taxpayer; 71 2. The return indicated that estimated tax had been paid by the amnesty taxpayer; 3. A W-2 form was attached to the return; or 4. The amnesty materials contained a copy of a letter from the Michigan Department of Treasury indicating that the IRS had notified the Department of Treasury that a federal filing had been made from a Michigan address. The letter also indicated that a cross-match by the Department of Treasury disclosed that no Michigan return had been filed. The other group of taxpayers filing non-amended returns (Group 3) are those who meet none of the above 1isted conditions. These taxpayers were probably unknown to the Department of Treasury prior to amnesty. Although the Department of Treasury had tax information available which could have identified individuals in Group 2 as nonfilers, it is possible that some taxpayers in Group 2 had never filed a tax return. In addition, it is possible that, the Department of Treasury had knowledge about some of the taxpayers in Group 3 (e.g., although the information collected during amnesty would indicate that the Department of Treasury did not have knowledge concerning them, these taxpayers may have filed a return in the past). The results of categorizing participants among the various groups indicates that 571 participants filed amended returns accounting for only 23.12 percent of the research data base. new (non-amended) returns (Group 1), Filers of (Groups 2 and 3) represent 76.88 percent of the data base (1,899 returns). Among the participants for whom non-amended returns are present, Group 2 comprises 1,272 returns (51.50 percent of the research data base; 66.98 percent of non-amended returns). As noted, some of these individuals filed federal tax returns but not Michigan tax returns, but 72 even more common are individuals who paid a portion of their tax liability (through withholding or estimated payments), but did not file a state return. In many instances, these individuals owed only a small amount of additional tax. The reason for their failure to file is not known -- some probably were just lax in their efforts; others probably thought they had paid in a sufficient amount via withholding or estimated payments, and therefore did not file a return. Taxpayers who filed new income tax returns in amnesty and had not been in the tax system (Group 3) account for 25.38 percent of the research data base (627 returns), and 33.02 percent of the non-amended returns in the research data base. Table 4-3 provides some descriptive information about each of these groupings within the research data base (amended, all non-amended, nonamended with contact, non-amended without contact). For amended returns (Group 1), the mean values for amnesty tax ($463.80), AGI ($153,433.61), and exemptions claimed (2.76) were higher than the means for non-amended returns (Groups 2 and 3 combined: $391.87, $28,822.66, and 2.32, respectively). However, the mean-amnesty-tax for those taxpayers without prior Treasury Department contact (Group 3: $543.05) is the largest among the groups, even though the mean AGI for this group is the lowest ($16,204.33). The majority of amended returns were jointly filed by married taxpayers (71.70 percent); only 40.08 percent of the combined non-amended returns were so filed. Among Group 3 returns contact), 67.10 percent are single taxpayers. (taxpayers without Of the returns not filed jointly, the most significant gender difference was among non-amended returns without contact where males filed 71.39 percent of these returns. 73 TABLE 4-3 SUMMARY CHARACTERISTICS OF RETURNS AMENDED RETURNS WITHIN RESEARCH DATA BASE IGROUP II Item Number Amnesty Tax AGI Exemptions 571 571 564 Mean Standard Deviation Median 463.86 2,824.30 119.08 153,433.61 1,380,818.10 31,715.48 2.76 1.27 2.00 Characteristic Minimum Maximum .71 64,498 3,047.48 24,115,974 1.00 8 Number % Filina Status: Sinale Married Filing Joint Married Filing Separate 569 569 569 27.42 71.70 0.88 Gender: Male Female 156 156 51.92 48.08 Other Characteristics: Non-filers in 1986 Opportunity to Evade Likely Self-Employed Urban (living in SMSA) 571 115 571 571 9.98 6.96 4.20 78.98 TABLE 4-3 (CONTINUED) NON-AMENDED RETURNS WITHIN RESEARCH DATA BASE (GROUPS 2 AND 31 Item Number Mean Standard Deviation Median Amnesty Tax AGI Exemptions 1899 1899 1899 391.87 28,822.66 2.32 903.83 237,542.30 1.54 143.07 17,353.00 2.00 Characteristic Minimum Maximum .34 16,188.00 25.00 10,309,092.00 0.0 11.0 Number % Filino Status: Sinale Married Filing Joint Married Filing Separate 1864 1864 1864 57.02 40.08 2.90 Gender: Male Female 1072 1072 68.84 31.16 Other Characteristics: Non-filers in 1986 Opportunity to Evade Likely Self-Employed Urban (living in SMSA) 1899 1895 1899 1899 32.07 2.43 14.27 83.94 TABLE 4-3 (CONTINUED) RETURNS WITH CONTACT IN RESEARCH DATA BASE (GROUP 2) Item Number Mean Standard Deviation Median Amnesty Tax AGI Exemptions 1272 1272 1272 317.35 35,042.55 2.47 843.35 289,710.10 1.59 120.34 21,321.00 2.00 Characteristic Filina Status: Single Married Filing Joint Married Filing Separate Gender: Male Female Other Characteristics: Non-filers in 1986 Opportunity to Evade Likely Self-Employed Urban (living in SMSA) Minimum Maximum .34 15,947.08 150.00 10,309,092.00 0.0 9.0 Number % 1253 1253 1253 52.12 45.33 2.55 656 656 67.23 32.77 1272 1272 1272 1272 31.37 2.44 5.50 86.08 TABLE 4-3 (CONTINUED) RETURNS WITHOUT CONTACT IN RESEARCH DATA BASE (GROUP 31 Item Number Mean Standard Deviation Median Minimum Maximum Amnesty Tax AGI Exemptions 627 627 627 543.05 16,204.33 2.03 999.28 20,793.38 1.38 224.00 9,810.60 1.00 1.46 25.00 1.00 16,188.00 257,923.00 Characteristic 11.00 Number % Filina Status: Single Married Filing Joint Married Filing Separate 611 611 611 67.10 29.30 3.60 Gender: Male Female 416 416 71.39 28.61 Other Characteristics: Non-filers in 1986 Opportunity to Evade Likely Self-Employed Urban (living in SMSA) 627 623 627 627 33.49 2.57 24.40 79.59 77 This can be contrasted with 68.84 percent for all non-amended returns, 67.23 percent for non-amended returns with contact, and only 51.92 percent for amended returns. Only 9.98 percent of the amended return filers failed to file a return in 1986,while 33.49 percent without contact failed to of the non-amended return filers file. Amended filers had the highest percentage of taxpayers with the opportunity to evade (6.96 percent), while amnesty participants filing non-amended returns without contact had the highest percentage of likely self-employed taxpayers (24.40 percent), and apparently also had a greater opportunity to evade (2.57 percent) than the non-amended returns with contact (2.44 percent). In summary, the non-amended no contact taxpayers appear to be a group of particular research interest. The average tax payment for this group is 1 arger than for the entire research data base (as well as the non-amended filers with contact) while the mean AGI is smaller, indicating that the majority of their tax 1iability was paid during amnesty. Occupations. As part of the data collection efforts, occupational information was sought for inclusion in the Michigan Amnesty Data Base. The only available source of this information, however, was a 1 ine on the Michigan tax return labeled "your occupation," requesting a written response from the taxpayer. These written responses were classified into 33 categories, adapted from the two-digit level occupational categories defined by the U.S. Department of Labor (Dictionary of Occupational Titles. 4th Edition, 1977). A few caveats regarding the occupational data are warranted. First, recall that for 515 data lines (about 17.3 percent of the IITDB) no tax 78 returns were available when the data collection occurred. Second, among the returns available during the data collection process, more than 15 percent had no entry on the occupation line. Third, classification among For example, occupations was sometimes a problem. although a taxpayer disclosed his occupation as "self- employed," there was an effort made to classify this taxpayer into another occupational group based on other information in the return. However, if no other information was available, the taxpayer was put in the self-employed category. an On the other hand, if a taxpayer disclosed occupation of "doctor," thetaxpayer was categorized using that information, even though it is possible that the taxpayer was also selfemployed. Another frequent response was "management" or "executive." Taxpayers disclosing one of these occupations were placed in a single category called "management/executive" if this was the only information available. Table 4-4 provides the percentage breakdown full by occupation of the IITDB, in order of frequency, along with brief descriptions of occupations included in various categories. The table does not provide any information beyond the occupational categorizations. is not proper to conclude that indicative of tax evasion. As a result, it occupations with high percentages are In fact, these disclosures not only reflect the propensities of their members to participate in amnesty, but also the taxpaying population in Michigan. 79 TABLE 4-4 PERCENTAGE DISTRIBUTION OF OCCUPATIONS REPORTED BY MICHIGAN INCOME TAX AMNESTY PARTICIPANTS Category Code Occupation 30 9 13 28 27 15 Retired Management/Executi ve Sales Student Self-Employed Building Trades (includes carpenters, pi umbers, electricians, etc.) 22 Machine Trades (includes metal workers, tool and die, mechanics) 23 Fabrication of Products (includes assembly workers) 24 Transportation (drivers, pilots, fl ight attendants) Clerical 11 2 Architecture, Engineering, Surveying Education (teachers at all levels, 1ibrarians) 4 3 Medicine and Health (doctors, dentists, pharmacists, nurses, etc.) 32 Unskilled Laborer 16 Protective Services (police, fire fighters, security) 6 Law (lawyers, judges) 7 Creative Arts (writing, journal ism, art, acting) 14 Food, Beverage, Lodging 18 Personal Services (includes barbers, tailors, dry cleaners, etc.) Building Services (janitors, maintenance) 17 Computer Related 12 19 Agriculture, Fishery, Forestry 8 Accounting/CPAs 5 Clergy 1 Sciences (mathematics, physical/life/social sciences) 26 Amusement, Recreation, Radio/Television, Motion Picture 25 Packaging and Materials Handling Professional Support (legal assistants, 10 dental assistant, etc.) Printing and Paperworking 21 Natural Resources Processing and Extraction 20 29 Housewife 31/33 Other (including Deceased) Source; Goddeeris, Martin, and Young (1988) Percent of Sample 13.30% 9.03 8.35 6.91 6.37 5.75 5.61 5.54 3.83 3.76 3.42 3.35 2.60 2.40 1.51 1.44 1.44 1.37 1.30 1.30 1.23 1.23 1.03 0.89 0.82 0.68 0.68 0.55 0.48 0.34 0.34 3.15 80 For purposes of this study, taxpayers in these 33 occupational groups have been consolidated into 7 major categories -- primarily for ease of analysis. Occupational data was only available for 2,085 of the 2,470 data lines in the research data base. The following summarizes the consolidation process: Michigan Amnesty Data Base Occupations (bv Cateqorv) 1 to 9 10 to 12 13 15, 18, 19, 21, 22, 24 14, 16, 17, 20, 23, 25, 26, 32 27 28 to 31, 33 Research Data Base Cateqorv Title Number Professional 507 Professional Support 116 Sales 201 Skilled Labor 410 Unskilled Labor 264 Self-Employed 135 Other (students, retired) 452 Total 2.085 5I 24 32 5 56 9 64 19 66 12 66 6 47 21 68 100.00% The next set of tables (Tables 4-5 through 4-11), provide the same summary information as Table 4-2, for each of the above occupational categories. A number of differences across category are evident. Average amnesty payments are highest among the self-employed, sales, and professional categories. They are lowest among the professional support, unskilled labor, and other (retired and students) categories. Average AGI values follow a similar pattern, with the occupational categories with high amnesty payments generally reporting high incomes. the self-employed and sales categories, However, both of which have amnesty payments well above the overall average do not have AGI levels as high as the professional category. Means of other variables differ across patterns. employed categories in expected The category with the highest opportunity to evade is self(24.00 percent) professional (4.58 percent). followed by sales (6.75 percent) and The lowest levels of contact occurred in the self-employed (41.79 percent), retired/students (59.59 percent) and sales (66.17 percent). The professional support category is heavily represented by females while several others are predominantly male. The highest level of amended filings occurred in the professional category (32.54 percent). Non-filers in 1986 were highest in the self-employed (38.52 percent) and sales (29.35 percent) categories, while the category most likely to include self-employed taxpayers (other than the selfemployed category) was sales (11.44 percent). Joint returns were filed more often by taxpayers in the professional (62.60 percent) and sales (59.80 percent) categories, while single returns dominated in the professional support (68.81 percent) and retired/students (66.00 percent) categories. 82 TABLE 4-5 SUMMARY CHARACTERISTICS OF RETURNS BY OCCUPATION CATEGORY; Item Number Amnesty Tax AG I Exemptions 507 507 507 Mean 694.95 188,741.73 2.96 PROFESSIONAL Standard Deviation Median Minimum Maximum 3,055.45 1,538,716.69 1.65 176.00 37,284.00 3.00 .71 25.00 1.00 64,498 24,115,974 9.00 Characteristic Number % Filina Status: Single Married Filing Joint Married Filing Separate 500 500 500 34.80 62.60 2.60 Gender: Male Female 188 188 58.51 41.49 Other Characteristics: Non-filers in 1986 Opportunity to Evade Likely Self-Employed Urban (living in SMSA) Amended Return Contact Indicated 507 306 507 507 507 507 28.80 4.58 3.95 87.77 32.54 87.97 83 TABLE 4-6 SUMMARY CHARACTERISTICS OF RETURNS BY OCCUPATION CATEGORY; PROFESSIONAL SUPPORT Item Number Mean Standard Deviation Median Minimum Maximum Amnesty Tax AG I Exemptions 116 116 116 155.73 22,147.27 2.09 191.13 17,705.43 1.25 91.92 18,495.15 2.00 1.94 1,872.00 1.00 1,358.72 117,467.60 6.00 Characteristic Filina Status: Single Married Filing Joint Married Filing Separate Gender: Male Female Other Characteristics: Non-filers in 1986 Opportunity to Evade Likely Self-Employed Urban (living in SMSA) Amended Return Contact Indicated Number % 109 109 109 68.81 27.52 3.67 75 75 32.00 68.00 116 92 116 116 116 116 29.31 0.00 2.59 80.17 20.69 89.66 84 TABLE 4-7 SUMMARY CHARACTERISTICS OF RETURNS BY OCCUPATION CATEGORY: SALES Item Number Mean Standard Deviation Median Amnesty Tax AG I Exemptions 201 201 201 837.97 36,733.67 2.88 1,514.08 36,775.13 1.65 378.00 29,632.00 3.00 Characteristic Filina Status: Single Married Filing Joint Married Filing Separate Gender: Male Female Other Characteristics: Non-filers in 1986 Opportunity to Evade Likely Self-Employed Urban (living in SMSA) Amended Return Contact Indicated Minimum 3.00 1,550.00 1.00 Number Maximum 13,908.00 287,555.00 11.00 % 199 199 199 38.19 59.80 2.01 80 80 85.00 15.00 201 163 201 201 201 201 29.35 6.75 11.44 79.10 18.91 66.17 85 TABLE 4-8 SUMMARY CHARACTERISTICS OF RETURNS BY OCCUPATION CATEGORY: SKILLED LABOR Item Number Mean Standard Deviation Median Amnesty Tax AG I Exemptions 410 410 410 319.41 25,180.62 2.55 574.83 31,021.64 1.61 137.00 21,230.50 2.00 Characteristic Minimum .76 150.00 1.00 Maximum 4,709.00 558,249.00 9.00 Number % Filino Status: Single Married Filing Joint Married Filing Separate 405 405 405 44.69 53.09 2.22 Gender: Male Female 185 185 81.62 18.38 Other Characteristics: Non-filers in 1986 Opportunity to Evade Likely Self-Employed Urban (living in SMSA) Amended Return Contact Indicated 410 339 410 410 410 409 29.76 0.00 9.76 82.93 17.32 78.48 86 TABLE 4-9 SUMMARY CHARACTERISTICS OF RETURNS BY OCCUPATION CATEGORY: UNSKILLED LABOR Item Number Mean Standard Deviation Median Amnesty Tax AG I Exemptions 264 264 264 212.91 21,725.59 2.41 334.15 16,008.90 1.49 96.04 19,811.00 2.00 Characteristic Minimum .34 1,575.98 1.00 Maximum 2,112.00 177,508.00 8.00 Number % Filina Status: Single Married Filing Joint Married Filing Separate 261 261 261 52.11 44.83 3.06 Gender: Male Female 132 132 80.30 19.70 Other Characteristics: Non-filers in 1986 Opportunity to Evade Likely Self-Employed Urban (living in SMSA) Amended Return Contact Indicated 264 222 264 264 264 263 24.62 0.00 9.47 81.06 15.91 84.41 87 TABLE 4-10 SUMMARY CHARACTERISTICS OF RETURNS BY OCCUPATION CATEGORY: SELF-EMPLOYED Item Number Mean Standard Deviation Median Amnesty Tax AG I Exemptions 135 135 135 826.74 43,012.25 2.17 1,976.21 212,126.00 1.35 394.00 14,433.82 2.00 Characteristic Filina Status: Single Married Filing Joint Married Filing Separate Gender: Male Female Other Characteristics: Non-filers in 1986 Opportunity to Evade Likely Self-Employed Urban (living in SMSA) Amended Return Contact Indicated Minimum Maximum 2.00 1,526.00 1.00 16,188 2,452,679 7 Number % 135 135 135 48.89 47.41 3.70 69 69 86.96 13.04 135 125 135 135 135 134 38.52 24.00 100.00 82.96 14.07 41.79 88 TABLE 4-11 SUMMARY CHARACTERISTICS OF RETURNS BY OCCUPATION CATEGORY; Item Number Mean Amnesty Tax AG I Exemptions 452 452 452 174.67 13,925.82 1.95 OTHER (RETIRED. STUDENT1 Standard Deviation 216.79 13,994.92 1.03 Median Minimum Maximum 98.95 9,897.50 2.00 1.46 836.00 0.00 1,767.58 134,406.00 6.00 Characteristic Number % Filina Status: Single Married Filing Joint Married Filing Separate 447 447 447 66.00 33.33 0.67 Gender: Male Female 275 275 50.55 49.45 Other Characteristics: Non-filers in 1986 Opportunity to Evade Likely Self-Employed Urban (living in SMSA) Amended Return Contact Indicated 452 452 452 452 452 451 22.35 0.00 0.22 80.09 24.34 59.59 89 Other Characteristics. Tables 4-12 to 4-15 provide summary information on gender, filing status, non-filers in 1986, and taxpayers with opportunity to evade taxes. In addition to being larger in number, males had a 1arger mean amnesty tax ($302.05) than females ($193.55), although males had a lower mean AG I (Table 4-12). Both the mean amnesty tax payments were below that of the research data base. Males were more 1ikely to be self-employed (11.97 to 6.11 percent), while females were more 1ikely not to have filed a return in 1986 (32.76 to 25.89 percent). Although more taxpayers in the research data base filed as singles, the mean amnesty tax paid by these taxpayers ($261.77) was less than half the mean amnesty tax paid by married taxpayers filing joint returns ($571.15) income (Table 4-13). (AGI) However, the proportion of tax to level of is almost two and a half times as great with single taxpayers than with married couples filing a joint return. Married taxpayers filing joint had a larger percentage of taxpayers with the opportunity to evade taxes (5.02 to 0.82 percent), and amended returns (35.41 to 12.96 percent). Taxpayers who failed to file a return in 1986 (Table 4-14) had larger mean amnesty tax payments ($594.98) and AGI ($110,121.01) than the entire research data base. In addition, they were more 1 ikely to be self-employed (15.02 percent) and had a greater opportunity to evade (3.96 percent) than the entire research data base. failed to file in 1986 than females, Although more males (61.27 to 38.73 percent), this result is primarily due to the fact that twice as many males fi 1ed for amnesty as females. As noted previously, females had a higher level of non-filing in 1986 than males. TABLE 4-12 SUMMARY CHARACTERISTICS OF RETURNS BY GENDER MALE Item Number Mean Standard Deviation Amnesty Tax AGI Exemptions 819 819 819 302.05 18,241.15 1.33 537.13 25,462.05 .81 Characteristic Non-filers in 1986 Opportunity to Evade Likely Self-Employed Urban (1iving in SMSA) Amended Return Contact Indicated Median Minimum Maximum 1.00 150.00 0.00 5,564.00 334,940.00 2.00 122.25 12,752.00 1.00 Number % 819 794 819 819 819 817 25.89 0.88 11.97 86.32 10.01 63.53 FEMALE Item Number Mean Amnesty Tax AGI Exemptions 409 409 409 193.55 22,501.20 1.57 Standard Deviation 265.40 123,078.34 .75 Characteristic Non-filers in 1986 Opportunity to Evade Likely Self-Employed Urban (living in SMSA) Amended Return Contact Indicated Median Minimum 101.00 .76 11,055.00 1,575.98 1.00 1.00 Number % 409 310 409 409 409 406 32.76 0,65 6.11 86.80 19.07 70.69 Maximum 2,202 2,452,679 5.00 91 TABLE 4-13 SUMMARY CHARACTERISTICS OF RETURNS BY FILING STATUS SINGLE Item Number Mean Standard Deviation Median Minimum Maximum Amnesty Tax AGI Exemptions 1219 1219 1219 261.77 19,485.38 1.39 442.61 74,512.27 .78 115.79 12,169.00 1.00 1.00 150.00 1.00 4,674.81 2,452,679.00 2.00 Characteristic Number % Gender: Male Female 1161 1161 66.24 33.76 Other Characteristics: Non-filers in 1986 Opportunity to Evade Likely Self-Employed Urban (1iving in SMSA) Amended Return Contact Indicated 1219 1097 1219 1219 1219 1210 28.30 0.82 9.93 86.22 12.96 66.12 MARRIED. FILING JOINT Item Number Amnesty Tax AGI Exemptions 1155 1155 1155 Mean 571.15 101,026.35 3.54 Standard Deviation Median Minimum Maximum 2,238.49 1,015,404.76 1.27 177.16 31, 889.00 4.00 .34 25.00 2.00 64,498 24,115,974 9 Characteristic Non-filers in 1986 Opportunity to Evade Likely Self-Employed Urban (1iving in SMSA) Amended Return Contact Indicated Number % 1155 897 1155 1155 1155 1151 24.94 5.02 9.96 78.61 35.41 84.45 92 TABLE 4-14 SUMMARY CHARACTERISTICS OF RETURNS NON-FILERS IN 1986 Item Number Amnesty Tax AGI Exemptions 666 666 666 Mean 594.98 110,121.01 2.41 Standard Deviation Median 2,788.68 1,304,892.21 1.52 182.94 20,092.50 2.00 Characteristic Minimum Maximum .76 64,498 25.00 24,115,974 1.00 11 Number % Filino Status: Single Married Filing Joint Married Filing Separate 652 652 652 52.91 44.17 2.92 Gender: Male Female 346 346 61.27 38.73 Other Characteristics: Opportunity to Evade Likely Self-Employed Urban (living in SMSA) Amended Return Contact Indicated 581 666 666 666 662 3.96 15.02 83.18 8.56 68.28 93 Taxpayers interesting with opportunity information to evade (Table 4-15). taxes also provide some They carry the highest mean amnesty tax payment of any of the groups investigated ($1,815.86), have a large mean AGI ($108,801.25), have the highest likely self-employed proportion (58.93 percent), and were more likely to be non-filers in 1986 (42.86 percent). In addition, 79.63 percent of the taxpayers were married filing a joint return. 4.3.2 Distributions of Tax Payments and Reported Incomes The differentiation among the various groups of amnesty participants is reinforced by the data related to the distribution of amnesty tax payments as shown in Table 4-16. The numbers across each row (excluding the means in the two right hand columns) add up to 100 percent, and indicate percentages of taxpayers with amnesty tax payments falling into the specified dollar ranges. Although the mean amnesty tax payment was $408.51 for the research data base, 41.21 percent paid $100.00 or less, and 81.21 percent paid $500.00 or less. There are information. some In the interesting patterns occupational in categories, the distributional payments by the professional support, skilled labor, unskilled labor, and retired/student categories are more concentrated in the smaller payment ranges than the entire research data base. Conversely, the professional, sales, and self-employment categories have relatively high percentages in the more than $500 ranges. Married couples filing joint returns have higher percentages than the entire research data base in the more than $500 ranges, as do taxpayers who failed to file a return in 1986, those with opportunity to evade and taxpayers who likely are self-employed. Taxpayers with the 94 TABLE 4-15 SUMMARY CHARACTERISTICS OF RETURNS OPPORTUNITY Item Number Amnesty Tax AGI Exemptions 56 56 56 Mean 1,816.86 108,801.25 3.30 Standard Deviation Median 2,926.31 323,918.50 1.41 1,178.81 47,368.00 3.00 Characteristic Filina Status: Single Married Filing Joint Married Filing Separate Gender: Male Female Other Characteristics: Non-filers in 1986 Likely Self-Employed Urban (living in SMSA) Amended Return Contact Indicated Minimum Maximum 68.00 16,188 30,092.00 2,452,679 1.00 7 Number % 54 54 54 16.67 79.63 3.70 9 9 66.67 33.33 56 56 56 56 56 42.86 58.93 82.14 16.07 70.91 95 TABLE 4-16 PERCENTAGE DISTRIBUTIONS AND MEANS OF AMNESTY TAX PAID BY VARIOUS CATEGORIES $100 or Less $101 to $500 $501 to $1,000 $1,001 to $5,000 More Than $5,000 Amnesty Tax Mean AGI Mean 41.21 40.00 10.20 8.02 0.57 408.51 57,629.49 Professional 38.02 36.81 11.58 11.93 1.66 694.95 188,741.73 Prof. Support 52.57 40.47 5.62 1.34 0.00 155.73 22,147.27 Sales 24.78 35.98 18.05 20.35 0.84 837.97 36,733.67 Skilled Labor 41.51 42.01 10.31 6.17 0.00 319.41 25,180.62 Unskilled Labor 51.49 38.98 4.77 4.76 0.00 212.91 21,725.59 Self-Employed 19.85 43.90 20.02 14.98 1.25 826.74 43,012.25 Retired/Student 49.30 42.86 6.48 1.36 0.00 174.67 13,925.82 Single 46.02 40.36 8.70 4.92 0.00 261.77 19,485.38 Married, Joint 36.10 39.48 12.03 11.18 1.21 571.15 101,026.35 Married, Separate 38.98 44.08 5.08 11.86 0.00 332.04 17,897.44 Male 44.08 39.69 9.89 6.34 0.00 302.05 18,241.15 Female 49.88 42.05 5.87 2.20 0.00 193.55 22,501.20 Non-filer in 1986 35.89 40.39 10.96 12.01 0.75 594.98 110,121.01 Opportunity 11.41 18.58 17.84 47.63 4.54 1,816.86 108,801.25 Likely Self-Employed 17.00 40.08 23.08 19.03 0.81 773.63 33,039.92 Urban (in SMSA) 41.32 40.59 9.78 7.63 0.68 419.60 64,021.84 Amended Return 45.83 38.37 9.38 5.73 0.69 463.86 153,433.61 Contact Indicated 44.63 39.76 9.17 5.79 0.65 317.35 35,042.55 Total Occupation: Filina Status: Gender: Other: 96 opportunity to evade and/or likely significant differences in these ranges. self-employed have the most Contact seems to be negatively related to higher tax payments. Table 4-17 provides information on the distribution of AGI per year for the same categories included in Table 4-16. Again, the numbers across each row (excluding the means in the two right hand columns) add up to 100 percent, and indicate percentages of taxpayers with AGI falling into the specified dollar ranges. The means categories. of AGI per year vary a great deal across various However, the median AGI value is less than $25,000 in all but five categories (professional, sales, married couples filing a joint return, taxpayers with opportunity to evade, and taxpayers filing amended returns). As mentioned before, the high means can be attributed to a small number of very large values. 97 TABLE 4-17 PERCENTAGE DISTRIBUTIONS AND MEANS OF ADJUSTED GROSS INCOME BY VARIOUS CATEGORIES Category $7,500 or Less $7,501 to $15,000 $15,001 $25,001 to to $25,000 $50,000 More Than $50,000 AGI Mean Amnesty Tax Mean 18.79 18.91 21.98 28.14 12.18 57,629.49 408.51 3.55 7.89 18.15 38.86 31.55 188,741.73 694.95 12.93 23.28 32.76 25.00 6.03 22,147.27 155.73 8.96 9.95 24.87 36.32 19.90 36,733.67 837.97 Skilled Labor 14.39 15.37 30.98 31.70 7.56 25,180.62 319.41 Unskilled Labor 15.15 21.97 26.14 33.33 3.41 21,725.59 212.91 Self-Employed 22.22 30.37 20.74 15.56 11.11 43,012.25 826.74 Retired/Student 41.59 28.10 16.15 11.73 2.43 13,925.82 174.67 33.06 24.61 23.46 16.08 2.79 19,485.38 3.03 12.38 20.52 41.39 22.68 101,026.35 571.15 27.12 23.73 28.81 13.56 6.78 17,897.44 332.04 Male 29.67 26.37 22.83 17.83 3.30 18,241.15 302.05 Female 37.16 22.74 24.69 12.71 2.70 22,501.20 193.55 17.11 19.22 25.07 27.93 10.67 110,121.01 594.98 0.00 0.00 0.00 51.79 48.21 108,801.25 1,816.86 Likely Self-Employed 24.69 25.91 21.05 18.22 10.13 33,039.92 773.63 Urban (in SMSA) 18.63 18.63 21.76 28.27 12.71 64,021.84 419.60 Amended Return 3.47 13.19 20.66 37.33 25.35 153,433.61 463.86 11.26 16.28 24.21 33.61 14.64 35,042.55 317.35 Total Occupation: Professional Prof. Support Sales Filina Status: Single Harried, Joint Married, Separate 261.77 Gender: Other: Non-filer in 1986 Opportunity Contact Indicated CHAPTER 5 AN OVERVIEW OF THE RESEARCH STUDY AND STATISTICAL METHODOLOGY 5.1 Introduction The proper modeling of the noncompliance decision appears to be predicated on identification of factors that are a part of the decision making process. This study provides an opportunity to identify factors via an analytical examination of a portion of the Michigan Amnesty Data Base. Previous taxpayer data base research primarily has been performed on IRS Taxpayer Compliance Measurement Program (TCMP) data. The TCMP survey itself consists of extensive audits of a stratified random sample of the taxpaying population. The information compiled on each individual includes the reported amounts on the original return and amounts the auditor deemed "correct" following the audit. Despite the obvious advantages of TCMP surveys over alternative data on tax evasion, it is important to identify two weaknesses regarding the usefulness of the TCMP difficult for wage, salary, data in measuring tax evasion. First, it is auditors to detect many forms of unreported income. and some interest, dividend, independently reported to the IRS. and rental Only income are While auditors are able to identify some unreported income, other types of income (e.g., moonlighting cash-only to businesses) noncompliance are very difficult in these areas is understated, this data will be somewhat erroneous. identify. and Since empirical research using Second, the TCMP is unable to reflect information on taxpayers who did not file returns. Nonfilers accounted for approximately 41 percent of unreported income in 1981 98 99 [Table 1-5]. It has been estimated that the TCMP data measures only 37 percent to 47 percent of all unreported income (i.e., filers who omit income and overstate deductions, and nonfilers) (Clotfelter [1983]). Finally, the TCMP data omit information on many demographic and all psychological variables that have been posited as part of the noncompliance decision process. Previous research has called for the study of the characteristics of non-filers (Clotfelter [1983]). In addition, this research has encouraged a more complete examination of the characteristics of filers who have underreported income (Clotfelter [1983], Spicer [1986], Jackson and Mill iron [1986]). The problem to this point has been a lack of new empirical data to examine. Since the Michigan Amnesty Data Base includes both types of taxpayers omitted by the TCMP, it may provide more useful information on these factors. 5.2 Statistical Anal vs is Values of the measure of compliance used in this study are not expected to be concentrated at the 1 imits, and thus, the need to use a technique designed for the analysis of censored dependent variables was eliminated. Since the focus of this research is on the explanation of, or prediction of, a dependent variable (the compliance measure), the regression model is appropriate. Multiple regression is an appropriate method when some of the independent variables are continuous and some are categorical, when cell frequencies in a factorial design are unequal and disproportionate, and when one is studying trends in the data. The classic 1 inear regression (CLR) model using the ordinary least squares estimator is probably the 100 most popular tool used by econometric researchers doing empirical work (Kmenta [1986] and Kennedy [1985, p.10]). The regression model used to represent the relationship between various factors and taxpayer noncompliance can be expressed as follows: Yi = ^0 + ^lXli + /32X2i + • • • + ^kXki + ei where Y. = a measure of taxpayer noncompliance for the ith taxpayer, i = 1 , . . ., n. Estimated for each taxpayer by computing the natural log of unreported income (computed by dividing the amnesty tax paid by the appropriate tax rate). X. .= the kth factor, k = 1, 2, . . .,K, for the ith taxpayer. The factors to be examined are described above. p^ = a parameter to be estimated for the various factors to be examined. pk reflects the impact of on Y, holding all other variables constant. = an error term. The CLR model consists of five basic assumptions about the way in which the observations are generated. 1. The first assumption of the CLR model is that the dependent variable can be calculated as a 1 inear function of an independent variable (or set of independent variables), plus an error term. The unknown coefficients of this 1 inear function form the vector p and are assumed to be constants. Several violations of this assumption, called specification errors, are: Non!inearitv: When the relationship between the dependent and independent variables is not 1 inear; Wrong Regressors: The omission of relevant independent variables or the inclusion of irrelevant independent variables; and Changing Parameters: When the parameters {p) do not remain constant during the period in which data was collected. 2. The second assumption of the CLR model is that the expected value of the error term is zero (i.e., the mean of the distribution from which the error term is drawn is zero). 101 3. 4. 5. The third assumption of the CLR model is that the disturbance terms all have the same variance and are not correlated with one another. Two violations of this assumption which are significant in econometric research are: a Heteroscedasticitv: same variance; and When the errors do not all have the a Autocorrelated Errors: with one another. When the errors are correlated The fourth assumption of the CLR model is that the observations on the independent variable can be considered fixed in repeated samples (i.e., it is possible to repeat the sample with the same independent variables). Econometric research may make several violations of this assumption as well: a Errors in Variables: Errors in measuring the independent variables; a Autoregression: Using a lagged value of the dependent variable as an independent variable; and a Simultaneous Equation Estimation: Situations in which the dependent variables are determined by the simultaneous interaction of several relationships. The fifth assumption of the CLR model is that the number of observations is greater than the number of independent variables and that there are no 1 inear relationships between the independent variables. The problem of multicollinearitv (two or more independent variables being approximately 1 inearly related in the sample data) is associated with this assumption. In the context of the CLR model, the OLS estimator has a large number of desirable properties, making it the overwhelming choice for the "optimal" estimator in much econometric research. The desirable properties of the OLS estimator include: 1. Least Squares: Because the OLS estimator is designed to minimize the sum of squared residuals, it is automatically "optimal" on this criterion. 2. Highest R2 : Because the OLS estimator is optimal on the least squares criterion, it will automatically be optimal on the highest R2 criterion. 3. Unbiasedness: The assumptions of the CLR model can be used to show that the OLS estimator B0,-s is an unbiased estimator of 6. 102 4. Asymptotic Criteria: Because the OLS estimator in the CLR model is unbiased, it is also unbiased in samples of infinite size and thus is asymptotically unbiased. In econometric research, when the estimating situation can be characterized by the CLR model, the OLS estimator meets almost all the criteria that econometricians consider relevant. If, however, violations are made to the assumptions of the CLR model, different estimators may be appropriate. Several violations of the CLR model which may be relevant to this research are discussed below. Non!inearitv. In practice, most econometric models are not 1 inear. Running an OLS regression when this is not true is unsatisfactory, since parameter estimates not only are biased but also are without meaning except in so far as the 1 inear functional form can be interpreted as an approximation to a nonlinear functional form. As the level of income variable in this study has been hypothesized to have a curvilinear relationship with the dependent variable, this data was plotted. relationship of the data was found to be 1 The inear, and as a result no transformations were necessary. Heteroscedasticitv. terms all In the CLR model, we assume have the same variance. variance is known as homoscedasticity. This condition of nonconstant It may be the case, however, that all of the error terms do not have the same variance. known as heteroscedasticity. heteroscedasticity exists that the error This condition is The easiest way to determine whether or not is to visually inspect the residuals. The examination of residuals is discussed as part of the analyses in Chapter 6. Multicollinearitv. The fifth assumption of the CLR model specifies that there are no exact linear relationships between the independent 103 variables. variables Although anexact is unusual, it linear relationship among independent is very possible to have an approximate linear relationship among some variables. Such relationships are common in economics, and this study is not excepted. Although the estimation procedure does not break down when the independent variables are highly correlated (i.e., approximately linearly related), severe estimation problems arise. The OLS estimator unbiased, statistic in the presence of multicollinearity remains and is still the best linear unbiased estimator. is unaffected. The major undesirable The R^ consequence of multicoll inearity is that the variances of the OLS estimates of the parameters of the coll inear variables are quite large. variances arise because in the These high presence of multicollinearity the OLS estimating procedure is not given enough independent variation in a variable to calculate with confidence the effect it has on the dependent variable. Such a situation may result in a large coefficient of determination (R^) accompanied by statistically insignificant estimates of the coefficients of the independent variables. At least one of the independent variables appear to influence systematically the dependent variable, determined. but which variable is causing the influence cannot be In this study, the sample size is large, which minimizes the impact of multicollinearity. Dummy Variables. As certain items of data are categorical variables, dummy coding of the categorical variables was performed to allow analysis. Regression Selection Technique Utilized. the variables will A stepwise selection of be used within the regression model to predict 104 noncompliance. Stepwise selection of the predictor variables is really a combination of the backward and forward selection procedures. The first variable considered for entry into the equation is the one with the largest positive or negative correlation with the dependent variable. An F test for the hypothesis that the coefficient of the entered variable is zero is then calculated. succeeding variable) To determine whether this variable (and each is entered, the probability statistic is compared to a .05 probability level. associated with F A variable enters the equation only if the probability associated with the F test is less than or equal to .05. After inclusion of the first variable (if one is significant), the second variable is selected based on the highest partial correlation. If it passes the entry criterion (probability of .05), it also enters the equation. At this point, the first variable is examined to see whether it should be removed from the regression equation. removal criterion is a probability of .10. of the second variable, For this study, the As a result, after inclusion if the probability associated with the F statistic is greater than .1 0 , the variable is removed from the equation. In the next step, variables not in the equation are examined for entry. After each step, variables already in the equation are evaluated for removal. This process continues until no more variables meet entry and removal criteria. The micro-computer statistical package used to conduct this analysis was SPSS-PC+. 5.3 Description of Variables As previously noted, there is little agreement in the literature regarding the relative salience of identified compliance factors and the 105 manner in which these variables are related to tax compliance. This analysis will provide relevant information regarding certain variables. The variables to be analyzed are predicated on the information obtained during the Michigan Amnesty Program, and are generally limited to certain economic and demographic variables. A list of the variables to be analyzed is contained in Figure 5-1. The independent variables are grouped into two categories -- factors previously identified and new variables. In addition, the expected sign of the regression coefficients is provided. 5.3.1 Dependent Variable This research, like Clotfelter [1983], Madeo, et. al. [1985], and Witte and Woodbury [1985], will use a direct measure of compliance derived from the actual reporting behavior of amnesty participants as the dependent variable. Previous studies have used a variety of measures to measure noncompliance. the independent However, because of the nature of the data and variables available for analysis, the noncompliance measure will be the natural log of unreported income. Many TCMP studies have used the ratio of amnesty tax paid to total correct tax noncompliance. 1 iability (after adjustments) as the measure of This measure cannot be used in this study because about 25 percent of amnesty participants paid in all of their tax 1iability during amnesty. would be 1 For these taxpayers the noncompliance ratio measure , prohibiting the use of regression analysis. Many economic studies use the natural log of a monetary amount as a dependent variable (e.g., many labor economic studies). In addition, the Clotfelter study (mentioned above) used the natural log of unreported income as the dependent variable. 106 FIGURE 5-1 RESEARCH VARIABLES DEPENDENT VARIABLE ■ Log of unreported income (amnesty tax paid/appropriate tax rate) INDEPENDENT VARIABLES Expected Sign of Regression Coefficient ■ Factors Previously Identified: Not Yet Tested Emoiricallv: 1 . Gender (male) + . Tax return preparation 3. Opportunity (occupation, income level, income source) + 2 - Empirically Tested Yielding Ambiouous Results: 1 . Filing (marital) status (single) . Geographic location (utilizing SMSA regions in Michigan) 3. Occupation 4. Self-employment 5. Enforcement agency contact 6 . Income level ilog of adjusted gross income) + 2 +/■ +/■ + - + New Factors To Be Examined: 1. 2. Exemptions Amnesty participants failing to file a return in 1986 + 107 Another benefit of using the log of unreported income as the dependent variable relates to interpretation of regression coefficients. When a regression linear in logarithms includes dummy variables (as this study does), the coefficient of a dummy variable may be interpreted as the percentage impact on the dependent variable (here, unreported income) of the qualitative variable it represents. 5.3.2 Independent Variables Factors Previously Identified. The analysis of previously identified factors is conducted along two lines, analytically examining factors that have previously been identified as potentially a part of the noncompliance decision process. 1. Have not been limitations, or 2. Have been tested results. The study examines variables that: tested with against taxpayer data due taxpayer data yielding to data ambiguous The first line of analysis includes an examination of gender and influence of tax return preparers. opportunity (income level, In addition, the composite variable income source, and occupation) previously posited as a part of the noncompliance decision process, but never before empirically tested, is examined. The second line of analysis includes an examination of filing status, geographic location (using the SMSA variables discussed in Chapter 4), occupation, contact by enforcement agency, and income level. In addition, the data provides an opportunity to look at noncompliance across several years when the Michigan tax rate varied. Additional Factors To Be Analyzed. other factors using the data collected. The research will examine two These factors include exemptions 108 (a potential surrogate for family size), whether or not the taxpayer filed a 1986 income tax return, and a second self-employment variable. As mentioned in Chapter 4, the 1986 nonfiler variable was created by cross-checking participants who filed during amnesty with the various 1986 master files maintained by the Treasury Department. Although no direct attitudinal variables are part of the Michigan Amnesty Data Base, a participant’s action of not filing a return does provide some measure of an attitude regarding compliance. Therefore, this information is analyzed as a part of formal research in this study. Of taxpayers in the research data base, 26.96 percent failed to file a return in 1986. In addition to those taxpayers who identified themselves as self- employed in the occupation information, the data was examined in an attempt to classify taxpayers as self-employed without using the occupational self-employment disclosure data in the Michigan Amnesty Data Base. A separate self-employment variable was then created for analysis. Specifically, taxpayers were assigned to this class if (1) no taxes were withheld, (2 ) no W - 2 was present with the amnesty return information, and (3) an occupation other than retired, student, deceased, or other was disclosed. This classification is different than the self-employment occupational variable since it includes taxpayers who did not disclose self-employment as an occupation, and excludes certain taxpayers who disclosed self-employment as an occupation (if either of the first two conditions for selection were true). 109 5.4 A Summary of the Data Bases To Be Analyzed 5.4.1 Primary Research Data Base A total of 2,985 data lines (one for each tax return) exist in the individual portion of the Michigan Amnesty Data Base. Of these data lines, two classes of returns were segregated out of the data base in creating the research data base used in this study. As mentioned in Chapter 4, no tax return data were available for some amnesty participants when the data were collected, coded, and keyed into the data base. Either these taxpayers filed an amnesty application and paid the tax and interest due but did not file a state tax return, or the state tax return was in use for some other Treasury procedure (audit or review) at the time of data collection. In addition, certain amnesty returns in the sample showed a zero amnesty tax payment, or refund due the taxpayer as a result of filing. These 1ines were also removed during creation of the research data base. In total, these data 1ines comprised 515 returns and $283,058 in amnesty tax. When these two groups of returns are removed, 2,470 1ines of data are left, forming the primary research data base used in this study. 5.4.2 Data Bases Constructed to Evaluate Taxpayers Without Treasury Department Contact As discussed in Chapter 4, participants filing non-amended tax returns as part of amnesty can be divided into two groups. Based on information collected from the amnesty applications, it appears that some amnesty participants were a part of the Michigan and federal tax systems, while other participants probably were outside the systems. The second 110 group of taxpayers is of interest since these taxpayers are more likely to include advertent noncompliers. Taxpayers were categorized into the second group if all of the following were false: 1. The return indicated that the tax liability had partially been paid through withholding; 2. The return indicated that the tax liability had partially been paid through payments of estimated tax; 3. A W-2 form was attached to the return; or 4. The amnesty materials contained a copy of a letter from the Michigan Department of Treasury sent to the taxpayer prior to amnesty. The taxpayers receiving such a letter had been identified from a match of federal taxpayers with Michigan addresses, and indicated that the Department of Treasury had knowledge of a federal return being filed with a Michigan address. Using data 1ines selected, two data bases were created to explore factors that are part of the noncompliance decision for these amnesty participants. The no contact taxpayers (roughly 630 lines of data out of 2470 in the entire research data base) paid in about tax 1 86 percent of their iability during amnesty. Primary No Contact Data Base. and no-contact amnesty participants, To discriminate between the contact approximately the same number of data 1ines were selected from taxpayers known to Treasury who made only small payments during amnesty (these taxpayers paid in approximately 23 percent of their tax 1iability during amnesty). These two sets of data were merged together into a data base (titled CONNCON.SYS) for analysis. As mentioned above, the primary reason for creating this data base was to explore differences between amnesty participants with some prior contact and those who had no prior contact. The determination of prior contact was made with the data in the research data base. The taxpayers Ill with contact in this data base generally paid small amounts of tax via amnesty. As a result, these taxpayers may be thought of as delinquent filers rather than tax evaders. However, this data base (and its subsequent analysis) is not likely to provide insight into differences between compliant and noncompliant taxpayers, because it may not be reflective of the taxpaying population in Michigan. It merely provides some insight into differences between amnesty participants with and without contact. Secondary No Contact Data Base. Of the taxpayers in the primary no contact data base, the retired/student occupational group typically had low incomes and made small amnesty payments. Although they may be nonfilers, perhaps they should not be classified as chronic tax evaders. For this reason, the primary no contact data base was modified, and a second no contact data base is created (a subset of the primary no contact data base; titled C0NNC0N2.SYS) by removing those taxpayers whose occupation was coded as 0CT7 (retired, students, and other). 5.4.3 Stratified AGI Data Bases Previous researchers have indicated that stratification of data bases in different ways (e.g., by level of income) may provide additional useful information (see, for example, Witte and Woodbury [1985]). With this in mind, this study stratified the research data base by level of adjusted gross income. analyzed in this study: The following summarizes the five AGI stratas 112 AG I N Less than $ 7,500 $ 7,500 - 14,999 15,000 - 24,999 25,000 - 49,999 50,000 or more 464 467 543 695 301 Amnestv Tax % 18.8 $ 18.9 28.1 1 2 .2 4.9 9.5 15.1 27.6 42.9 2.470 100.0 $1,009,028.66 100.0 TOTALS 5.4.4 Average AG I 106.28 205.01 280.40 400.58 1,439.57 $ 4,568.04 11,105.55 19,753.05 35,296.83 331,500.89 $ 408.51 $ 57.629.49 % 49,316.15 95,740.59 152,257.47 278,403.86 433.310.59 2 2 .0 Average Amnesty Tax $ Missing Information As would be expected, the data in the research data base has complete income, information in certain amnesty tax paid, circumstances zip code data, (e.g., adjusted gross 1986 return filing, and exemptions) and missing values related to other variables (e.g., filing status, occupation, gender, tax return preparation, and contact with the Treasury Department). Where missing values existed, a mean substitution of the variable in question was deemed to be appropriate. For example, of the 2,470 data lines, only 2,085 had occupational data. For the data lines with missing occupational information, the mean substitution was made by giving each such data seven occupations. 1 ine a decimal equivalent for each of the The decimal equivalent was based on the occupational distribution among the 2,085 valid codings. Other missing value mean substitutions were made in a similar manner. 5.5 Summary This chapter describes the analyses to be performed (including the dependent and independent variables), the data bases used analyses, and the statistical technique used in the analyses. presents the results of the statistical analyses. in the Chapter 6 CHAPTER 6 RESULTS AND ANALYSIS 6.1 Introduction Chapter 5 provided a summary of the research and the methodology to be used in making conclusions regarding the data. This chapter reports on the various regression analyses performed on the data. Before proceeding with a discussion of the results, a short review of the various statistics is warranted. Goodness of Fit. An important part of any statistical procedure that builds models from data is establishing how well the model actually fits. A commonly used measure of the goodness of fit of a 1 inear model is R2 , sometimes called the coefficient of determination. thought of in a variety of ways. correlation It can be Besides being the squared multiple (indicating the proportion of variance of the dependent variable accounted for by the optimally weighted independent variables), it is also the squared product-moment correlation of the observed dependent variables and the predicted values of the dependent variables, which are of course a 1 inear combination of the independent variables. If all the variance in the dependent variable is accounted for, R2 is 1. If there is no 1 inear relationship between the dependent and independent variables, R 2 is 0. It is important to note that R2 is a measure of the goodness of fit of a particular model. mean that there is no association indicates that there is no 1 An R 2 of 0 does not necessarily between fits it only inear relationship. The sample R 2 tends to be an optimistic model thevariables; the population. Normally, 113 the estimate of how well the model does not fit the 114 population as well as it fits the sample from which it is derived. The statistic "adjusted R2" attempts to correct R 2 to more closely reflect the goodness of fit of the model in the population (e.g., see Cohen [1975], p. 105). Regression Coefficients. In the regression equation, the least- squares coefficients (B) are used to estimate the unknown population parameters (6 ). equation Each regression coefficient in a multiple regression indicates the expected change in the dependent variable associated with a one unit change in the variable under consideration while controlling for, or holding constant, the effects of the other independent variables. As a result, B ’s in multiple regression are referred to as partial regression coefficients. Tests of Significance - In General. It is necessary to remember that all that is meant by a statistically significant finding is that the probability of its occurrence is small, assuming that the null hypothesis is true. A It is the substantive meaning of the finding that is paramount. statistically significant substantively meaningful. finding is of no value if it is not It has been said that "we should not feel proud when we see a researcher smile and say ’the correlation is significant at the .01 level.’ Perhaps that is the most he can say, but he has no reason to smile" (Nunally [I960]). Test of the Goodness of Fit. Usually, the first test of interest is to determine the probability that some 1 inear relationship exists between the dependent variable and the set of independent variables. An overall F test is used to test the null hypothesis that there is no 1 inear relationship (i.e., Bj = B 2 = B3 = B 4 = . . . = B^ = 0). The F statistic is the ratio of the mean square regression (R2 divided by the appropriate 115 degrees of freedom) to the mean square residual [(1 - R^) divided by the appropriate degrees of freedom]. Test of Significance of Regression Coefficients. Like other statistics, the regression coefficient has a standard error associated with it. standard The standard error of the regression coefficient is the deviation of the sampling distribution of the regression coefficient, and therefore, it can be used to test the significance of the regression coefficient. Dividing a regression coefficient by its standard error yields a tstatistic (or t ratio). When a given regression coefficient is tested for significance, the question being addressed is whether it differs from zero while variables. 1 control 1 ing for the effects of the other independent Multicollinearity among the independent variables leads to arger standard errors of the regression coefficients (i.e., the larger the intercorrelation among the independent variables, the larger the standard errors). As a result, when independent variables are highly intercorrelated, the chance of finding statistical significance in the regression coefficients diminishes. This test is different than the test of R^ in a multiple regression equation, which is tantamount to testing all the regression coefficients simultaneously. With the use of the t-statistic, confidence intervals can be set around the regression coefficients. The use of confidence intervals in preference to tests of significance has been strongly advocated by many statisticians. Probably the most important argument is that a confidence interval provides more information than that provided by a statement about the rejection of a null hypothesis. In addition, a confidence interval enables a researcher to test simultaneously all possible null 116 hypotheses. The narrower the confidence interval, the smaller the range of null hypotheses, and therefore, more confidence can be placed in the findings. Testing Increments in Proportion of Variance Accounted For. Another way of assessing the relative importance of independent variables is to consider the increase in when a variable is entered into the equation. The test for an increment in the proportion of the variance accounted for is also an F test. Two points should be made about this test. First, testing the increment in proportion of variance accounted for by a single variable is equivalent to associated with the variable. the test of the regression coefficient Second, the increment in the proportion of variance accounted for by a given variable (or set of variables) may be considerably different from the proportion of variance it accounts for by itself, the difference being directly a function of the correlations of the variable with the other variables in the equation. Interpretation of Regression Coefficients of Dummy Variables When Dependent Variable Is Expressed in Log Units. in logarithms includes dummy When a regression 1 inear variables (as this study does), the coefficient of a dummy variable may be interpreted as the percentage impact on the dependent variable (here, unreported income) of the qualitative variable it represents. 6.2 An Overview of the Multiple Regression Analyses 6.2.1 In General Table 6-1 presents a brief summary of the of this study. The summary 8 regressions run as part includes the significant regression coefficients (in selection order) and the t-statistics for each. In * * * * Dependent Variable.. NONCOMP MULTI PL E TABLE 6-1 REGRESSION SUMMARY * * * * LOG OF UNREPORTED INCOME Variables in the Equation RESALL2.SYS (ENTIRE RESEACH DATA BASE) Variable B AGILN CONTACT OPPORT 0CT3 0CT6 N0FILE86 PDPREFSMSA2 RATE635 OCT 5 SELFEMPL GENDER EXEMPT (Constant) .58388 -.93359 .84392 .58912 .53310 .17434 .16548 .33982 -.13441 -.26126 .29934 .19001 .04797 2.27921 Multiple R R Square Adj R Square Std Error .51300 .26317 .25927 1.30457 T 18.679 -13.208 4.393 5.899 4.146 2.923 3.001 2.544 -2.452 -2.958 2.466 2.830 2.140 7.866 CONNCON.SYS (CONTACT Sig T Variable .0000 .0000 .0000 .0000 .0000 .0035 .0027 .0110 .0143 .0031 .0137 .0047 .0324 .0000 SELFEMPL OCT3 OPPORT 0CT6 N0FILE86 GENDER AGILN 0CT7 SMSA10 OCT5 SMSA1 RATE635 (Constant) CONNCON2.SYS (CONNCON.SYS WITHOUT RET'D/STUDENTS) & NO CONTACT TAXPAYERS) B T Sig T Variable 1.50225 1.15694 1.84707 .87073 .38644 .50863 .29101 .56012 -.58539 -.36186 -.21078 -.18244 3.77791 10.557 7.055 4.836 4.604 4.169 5.617 6.018 4.528 -3.426 -2.507 -2.410 -2.082 7.621 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0006 .0123 .0161 .0375 .0000 Multiple R .50019 R Square .25019 Adj R Square .24286 Std Error 1.45664 B T Sig T SELFEMPL OCT3 OPPORT OCT6 AGILN GENDER N0FILE86 SMSA10 OCT 5 EXEMPT SMSA1 (Constant) 1.48814 1.10215 1.80217 .87154 .28616 .73394 .48096 -.66824 -.38250 .09232 -.22488 3.48266 10.095 6.367 4.491 4.388 4.944 6.034 4.462 -3.304 -2.517 2.202 -2.179 6.155 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0010 .0120 .0279 .0296 .0000 Multiple R R Square Adj R Square Std Error .52867 .27949 .27130 1.52482 INDEPENDENT VARIABLES NAME AGILN SELFEMPL GENDER CONTACT OPPORT N0FILE86 PDPREP OCT3 OCT 5 0CT6 OCT7 DESCRIPTION NAME LOG OF AGI SELF-EMPLOYED (PROFESSIONAL, SALES, OR SELF-EMPLOYED; NO WITHHOLDING; ZERO OR POSITIVE ESTIMATED TAXES) SINGLE, MALE TAXPAYERS PRIOR TREASURY CONTACT OPPORTUNITY (PROFESSIONAL, SALES, OR SELF-EMPLOYED; AGI $30000+; ACCESS TO CASH INCOME SOURCES) NONFILER IN 1986 PAID PREPARER SALES UNSKILLED LABOR SELF-EMPLOYED OTHER (STUDENT, RETIRED, OTHER) SMSA1 SMSA2 SMSA3 SMSA9 SMSA10 URBAN EXEMPT RATE635 DESCRIPTION METRO DETROIT ANN ARBOR JACKSON BENTON HARBOR GRAND RAPIDS/HOLLAND TAXPAYER IN SMSA EXEMPTIONS TAX YEAR 1983 (TAX RATE=6.35%) TABLE 6-1 (CONTINUED) Dependent Variable.. NONCOMP LOG OF UNREPORTED INCOME ---------------------------------------------------- Variables in the Equation AGI1.SYS (AGI LT $7500) Variable AGI2.SYS (AGI S 7501-$14999) B T Sig T Variable AGILN EXEMPT CONTACT OCT3 0CT7 GENDER RATE533 (Constant) 1.13508 -.27490 -.32869 .70379 .32602 .27088 -2.48008 -2.15303 13.099 -4.940 -3.898 3.172 3.375 2.995 -2.790 -3.000 .0000 .0000 .0001 .0016 .0008 .0029 .0055 .0028 CONTACT OCT6 AGILN OCT4 EXEMPT SELFEMPL (Constant) -.83076 .84747 .87965 .39623 -.11332 .48258 -.02895 Multiple R .56322 .31722 .30674 .88353 Multiple R .42677 .18213 .17146 1.20504 R Square Adj R Square Std Error R Square Adj R Square Std Error AGI4.SYS (AGI $25000-550000) Variable CONTACT OPPORT AGILN OCT3 PDPREP NOFILE86 SMSA9 (Constant) Multiple R R Square Adj R Square Std Error T Sig T Variable -6.628 4.068 3.025 2.340 -2.366 2.101 -.011 .0000 .0001 .0026 .0197 .0184 .0362 .9914 CONTACT OCT6 OCT3 URBAN SMSA3 SELFEMPL (Constant) Mult'ple R R Square Adj R Square Std Error AGI5.SYS (AGI $50000+) B T Sig T Variable -1.46137 .92541 1.04657 .55846 .36589 .27120 1.28879 -2.10402 -9.159 3.717 4.122 3.439 3.643 2.457 2.434 -.792 .0000 .0002 .0000 .0006 .0003 .0142 .0152 .4289 CONTACT AGILN EXEMPT N0FILE86 OCT1 SMSA2 RATE635 URBAN PDPREP SMSA1 SMSA9 (Constant) .49381 .24384 .23614 1.26407 B AGI3.SYS (AGI $15000-$24999) Multiple R R Square Adj R Square Std Error B T Sig T -2.35573 .73288 .27781 .86327 -.74837 .84350 -.53893 1.24164 .49987 -.53720 -2.06258 .69604 -8.052 6.893 4.370 4.015 -3.802 1.887 -2.841 3.954 2.448 -2.380 -2.272 .568 .0000 .0000 .0000 .0001 .0002 .0602 .0048 .0001 .0149 .0179 .0238 .5703 .64297 .41341 .39108 1.52415 B T Sig T -1.20054 1.04280 .70344 -.49282 1.00124 .58814 8.89459 -6.780 3.788 3.420 -3.236 2.306 2.060 43.865 .0000 .0002 .0007 .0013 .0215 .0399 .0000 .47454 .22519 .21651 1.35489 119 addition, multiple R, and R square for each regression are presented. A detailed discussion of each of the regressions along with a comparison of the results with previous research in the area of noncompliance occurs later in this Chapter. However, a review of the results in Table 6-1 will indicate that the various regression equations confirm that certain variables (previously identified in survey, experimental, and analytical research) are part of the noncompliance decision. This study is unique in that the data being analyzed are related to amnesty participants. of all tax evaders. As stated previously, these taxpayers are a subset Previous analytical work in the noncompliance area has typically used Taxpayer Compliance Measurement Program (TCMP) data for analysis. This data (discussed more fully in Chapter 2) is collected from detailed audits of taxpayers known to the Internal Revenue Service. The results of this study, therefore, may not be directly comparable to prior research since a different type of taxpayer is involved. 6.2.2 A General Discussion About the Assumptions of the Regression Model as it Applies to this Study In Chapter 5, potential violations of the B ^ S regression in this study were briefly discussed. Conclusions regarding these potential violations are discussed here, and in the discussion of each of the regression analyses as needed. Specification Errors. The first assumption of the classic 1 inear regression model states that the dependent variable can be written as a linear function of a specific set of independent variables, plus a disturbance term. Specification errors can occur where relevant independent variables are omitted, or irrelevant independent variables 120 are included. In addition, a nonlinear functional form will lead to specification error. The relationship between the independent variables selected for analysis and the measure of noncompliance used as the dependent variable can be defended based on prior theory (e.g., gender, self-employment, contact, opportunity, et. al.), or the specific nature of the data set (e.g., SMSA’s). Because of the large numbers of small payments in the data set, it was decided to logarithmically transform the dependent variable and adjusted gross income. The added benefit of being able to discuss the regression coefficients of the dummy variables in terms of percentage impact on the dependent variable was previously mentioned. As a result, there appears to be a sufficient basis for concluding that specification error has been minimized in the regression results. Measurement Errors. Many researchers feel that measurement error (or errors in variables) present the greatest drawback to econometric research. Here, we are concerned with the of there was using incorrectly measured variables. discussion of the percent sample data and its relationship to the 10 In Chapter 3, implication a brief population from which it was drawn. In Chapter 4, there was a discussion of several items that might lead to a measurement error problem -- the coding of occupations and the lack of this data on about 10 percent of the returns, and the various range and accuracy tests performed on the research data base. relying on the which 10 However, given the corrections to the data made, and percent sample as representative of the population from it was drawn, there is support for a conclusion of minimal measurement errors in the data set. 121 Heteroscedasticitv. The regression model is relatively robust for violations of the assumption of heteroscedasticity (Pedhazur [1982]). The standardized residuals in each of the regressions were plotted against standardized predicted dependent variables, and against a normal distribution. violated. The piots generally indicate that this assumption is not Further discussion of the residual analysis occurs in the discussion of each of the regressions. Multicollinearitv. Although, technically, multicollinearity unless there is an exact 1 there is no inear relationship between independent variables, there is a possibility that several independent variables are highly correlated (approximately 1 inearly related). in this case, Even the problem can lead to imprecise estimation of the regression coefficients, and adverse effects on their standard errors (leading to a negative impact on their statistical significance and increasing their confidence intervals). The ordinary least squares estimator in the presence multicollinearity remains unbiased, however, and is still the best unbiased estimator. The statistic is unaffected. 1 of inear In this study, because of the size of the data set, and the regression methodology employed (stepwise selection) multicollinearity is minimized. Statistical Package Used for the Regression Analyses. The multiple regression procedure of the SPSS-PC+ package was utilized with stepwise selection (with a significance level of .05) to predict the variables that are part of the noncompliance decision. A more detailed discussion of the stepwise procedure utilized can be found in Chapter 5. Discussion of Results. The discussion partitioned into three sections, as follows: of results will be 122 6.3 Regression Analysis of the Research Data Base 6.4 Regression Analyses of the Non-Contact Data Bases 6.5 Regression Analyses of the Stratified AGI Data Bases As mentioned in Chapter 5, there are two non-contact data bases, one which includes all returns where no contact was indicated, and a second that excludes taxpayers who identified themselves as retired, students, or several other minor stratified AGI data bases. occupational categories. There are five 123 6.3 Regression Analysis of the Research Data Base 6.3.1 In General Table 6-2 presents the regression results for the entire research data base. The analysis of the regression occurs along several lines, including a discussion of the goodness of fit (R2), an analysis of the regression coefficients and their confidence intervals, and a comparison of the results with prior research. 6.3.2 Goodness of Fit The model has a R 2 of .2632, and an adjusted R 2 of .2593 with 13 variables in the equation. As a result, the independent variables in the model account for about 26 percent of the variance variable (unreported income). 1 imitations psychological of the and research has reported set demographic R2 thedependent This result seems reasonable, given the (i.e., this analysis [or attitudinal] variables, and various economic percent. data in only includes no aportion of the variables). Similar noncompliance statistics of from about 20 percent to 45 In addition, most behavioral researchers would consider a R 2 of .2632 to be both meaningful and of large magnitude (Cohen [1977]). The power of this analysis (i.e., the probability of rejecting the null hypothesis that the population multiple correlation is equal to zero) can be computed using a technique detailed in Cohen & Cohen [1975] which makes use of the sample size, the R 2 statistic, and the number of independent variables in the equation. Given 2,470 cases, an R 2 of .2632, and 13 independent variables, the power of the analysis greatly exceeds .99 (a sample size of 135 would have resulted in a power computation of .99; as all analyses performed in this study have at least 124 TABLE 6-2 REGRESSION TO PREDICT NONCOMPLIANCE USING RESEARCH DATA BASE Variable Name Coefficient (Standard Error) t (Significance) R2 Adjusted AGILN .58888 (.03153) 18.679 (.0 0 0 0 ) .1137 .1134 CONTACT -.93359 (.07068) -13.208 (.0 0 0 0 ) .2135 .2129 OPPORT .84392 (.19211) 4.393 (.0 0 0 0 ) .2272 .2263 0CT3 .58912 (.09987) 5.899 (.0 0 0 0 ) .2389 .2376 0CT6 .53310 (.12859) 4.146 (.0 0 0 0 ) .2472 .2456 NOFILE 8 6 .17434 (.05964) 2.923 (.0035) .2501 .2483 PDPREP .16548 (.05514) 3.001 (.0027) .2529 .2508 SMSA2 .33982 (.13357) 2.544 (.0 1 1 0 ) .2550 .2526 RATE635 -.13441 (.05482) -2.452 (.0143) .2569 .2542 0CT5 -.26126 (.08831) -2.958 (.0031) .2588 .2558 SELFEMPL .29934 (.12137) 2.466 (.0137) .2605 .2572 GENDER .19001 (.06713) 2.830 (.0047) .2618 .2582 EXEMPT .04797 (.02242) 2.140 (.0324) .2632 .2593 Constant 2.27921 (.28975) 7.866 (.0 0 0 0 ) 7 125 135 cases, no further mention of the power of these analyses will be made). The order of the variables into the equation, and their impact on R2 is as follows (all variables except AGILN and EXEMPT are dummy variables). of .1137. The log of AGI (AGILN) enters the equation first with a R 2 Prior contact with the Treasury Department (CONTACT) enters next and increases R 2 by .0998. increases R 2 by .0137 to .2272. Opportunity to evade (OPPORT) then Taxpayers who disclosed an occupation of sales (0CT3) enters the equation next and increases R 2 by .0117 to .2389. Taxpayers who disclosed an occupation of self-employed (0CT6) then enters the equation and increases R2 by .0083 to .2472. Among amnesty participants were certain taxpayers who continued their nonfiling in 1986. The dummy variable created to signify these (N0FILE86) taxpayers enters the equation next, increasing R 2 by .0030 to .2501. The variable created to signify amnesty participants who used paid preparers to help with their return process (PDPREP) enters next, with a .0027 contribution to R2 . The next variable to enter is SMSA2, which represents taxpayers in the Ann Arbor SMSA. It increases R 2 by .0021 to .2550. The next two variables to enter the model (RATE635; variable that represents returns filed when unskilled the tax labor) rate was 6.35 percent) each increasing it to .2588. contributes an and 0CT5 additional .0019 (representing to the R2 , The variable that was constructed to indicate self-employment via an examination of other data in the research data base other than occupation coding (SELFEMPL) enters next, with a contribution of .0017 to R2 . The variable constructed to indicate single males (GENDER)enters the regression equation next, increasing R 2 by .0013 to .2618. The number of exemptions (EXEMPT) claimed on the return 126 (a surrogate for family size) enters the model last, and increases R 2 by .0014 to .2632. As can be seen from Table 6-2, five of the thirteen variables (AGILN, CONTACT, OPPORT, 0CT3, and 0CT6) account for 24.72 percent of the variance in the dependent variable, and about 94 percent of the variance accounted for by all thirteen variables (26.32 percent). Although statistically significant, the last eight variables make relatively smal 1 contributions to R2 . Although not significant at the .05 level, four other variables are significant at the .10 level. They are 0CT7 (occupation coded as retired and students), FSTATUS (a dummy variable coded for single taxpayers), SMSA3 (representing Jackson), and SMSA9 (representing Benton Harbor). All four variables are positively associated with the dependent variable. 6.3.3 Analysis of Regression Coefficients and Confidence Intervals Table 6-3 presents the regression coefficients for each of the independent variables in the regression equation, along with their standard error and computed 95 percent confidence interval s. dependent variable in this study is expressed in log As the units, the coefficient of the dummy variables in the regression equation may be interpreted as the percentage impact on the dependent variable (here, unreported income) of the qualitative variable it represents. Log of AGI (AGILN). The estimated adjusted gross income is .58888. regression coefficient of The regression coefficients for a continuous variable expressed in logarithms would normally be interpreted as the percentage change in the dependent variable given a doubling of the independent variable. However, as the regression coefficient 127 TABLE 6-3 CONFIDENCE INTERVALS OF REGRESSION COEFFICIENTS FOR RESEARCH DATA BASE Variable Name Coefficient Standard Error .58888 .03153 .52705 .65070 CONTACT -.93359 .07068 -1.07219 -.79498 OPPORT .84392 .19211 .46721 1.22063 OCT3 .58912 .09987 .39328 .78496 0CT6 .53310 .12859 .28095 .78525 N0FILE86 .17434 .05964 .05739 .29130 PDPREP .16548 .05514 .05736 .27360 SMSA2 .33982 .13357 .07791 .60174 RATE635 -.13441 .05482 -.24190 -.02691 OCT5 -.26126 .08831 -.43444 -.08808 SELFEMPL .29934 .12137 .06134 .53735 GENDER .19001 .06713 .05837 .32166 EXEMPT .04797 .02242 4.017955E-03 .09193 2.27921 .28975 1.71103 2.84740 AGILN (Constant) 95 Percent Confidence Interval 128 increases in size, this interpretation begins to break down. Specifically, a doubling of the independent variable will result in a 2^ increase in the dependent variable (where b is the regression coefficient). Given this fact, amnesty participants with twice as much adjusted gross income relative to other amnesty participants had about 50.41 percent (2-58888 _ This finding more unreported income in AGI, ceteris paribus. suggests that noncompliance increases as income increases, and is consistent with the findings of Mork [1975], Clotfelter [1983], and Witte and Woodbury [1985]. In addition, the result is consistent with the hypothesis of increasing absolute risk aversion, commonly accepted as a reasonable assumption in models of individual choice under uncertainty (see, for example, Allingham and Sandmo [1972]). The regression coefficient is within the range of the Clotfelter study, which estimated regression coefficients of .292 for non-business returns, .620 for non-farm business returns, and .656 for farm returns. However, this variable is dominant (as is the next variable to enter the equation), with Clotfelter’s. A a t-statistic potential almost explanation three times of the as large strength of as its significance is that during the amnesty program, those taxpayers who had the most to lose were those with higher levels of income. it, was these taxpayers who came forward during amnesty. As a result, As a result, a clear finding that noncompliance increases as income increases cannot be made. As previously was mentioned, depending on the data base used, dominance by either low or high income level taxpayer noncompliance may mask noncompliance at the opposite end of the spectrum. This may be the 129 case with this data set. Based on a review of descriptive statistics, the amnesty program was dominated by participants making low amnesty payments. In addition, the highest level of participation by occupation occurred among retired taxpayers, highest group. income. and students comprised the fourth Both of these groups reported relatively low levels of As noted briefly above, the occupation variable for retired and students (0CT7) was significant at the .1 0 positively correlated to noncompliance and level. Witte and Woodbury [1985] suggest that the relationship between income level and compliance is really curvilinear, with low-income and high-income taxpayers being noncompliant. It would appear that this data set might lend some support to that conclusion. A question that must be addressed is why low income taxpayers are noncompliant. Part of it surely is a belief that those with the most to lose are the most at risk (i.e., a to get caught). 1 ittle fish in a big sea is not 1 ikely In addition, however, it could be that these taxpayers may need additional help from the enforcement agencies, or that they legitimately believe that they do not owe any taxes because they are earning too little. The 95 percent confidence interval for this coefficient ranges from .52705 to .65070, indicating that the population parameter would fall in this range 95 percent of the time. adjusting for the size The range indicates that (after of the coefficients) an individual amnesty participant with twice as much in adjusted gross income as another individual participant could have unreported income, ceteris paribus. from 44.1. to 57.0 percent more 130 Contact (CONTACT). The estimated regression coefficient of the contact variable is -.93359, with a 95 percent confidence interval ranging from -.79498 to -1.07249. Taxpayers having some form of Treasury Department contact prior to amnesty (withheld taxes, estimated taxes, a W-2, or a letter from Treasury requesting information regarding a state tax return) had on average 93.36 percent less unreported income than those amnesty participants who had no contact prior to amnesty, ceteris paribus. The 95 percent confidence interval for the population parameter ranges from 79.5 percent to 107.2 percent less unreported income than amnesty participants with no prior contact. This result confirms the previous analytical research of Witte and Woodbury [1985] who found that receiving notices from the IRS was associated with an increased level of compliance (as was an increased probability of audit). Spicer and Hero [1985] found similar results in a lab experiment. These results, however, conflict with the survey research of Spicer and Lundstedt [1976], who found that prior contact from the IRS (in that study the contact was normally via an audit) led to increased tax resistance and decreased compliance. Opportunity to Evade (OPPORT). This variable was first posited as positively related to noncompliance in survey research by Yankelovich, Skelly and White [1984]. created via a composite As discussed in Chapter 4, this variable was of occupation (self-employed, business, professional, or sales), an income level of $30,000 or greater, and access to cash income sources. This variable had not been tested empirically prior to this study. The estimated regression coefficient of the variable is .84392, with a 95 percent confidence interval of .46721 to 1.22063. The confidence 131 interval is large due to the size of the standard error of the regression coefficient, indicating a great deal of variability in the sampling distribution of the regression coefficient. The regression coefficient indicates that amnesty participants with the opportunity to evade taxes on average underreported 84.392 percent more income than amnesty participants characteristics, ceteris paribus. without the opportunity With 95 percent confidence, these results indicate that the population parameter shows that taxpayers with opportunity underreport income by at least 46.7 percent and possibly as much as 12 2 .1 percent when compared to those amnesty participants without the opportunity characteristics. This result is related to this not surprising given the descriptive statistics group of taxpayers. Although small in number, their mean amnesty tax payment was more than four times as large as the mean of the research data base. Recall that only 2.79 percent of amnesty participants had these characteristics (although they accounted for 10.08 percent of the amnesty tax paid within the research data base). This study indicates that the characteristics associated with opportunity are a strong indicator of potential noncompliance. Occupations (0CT3, 0CT6. 0CT5). Three different occupation variables are part of the regression equation -- sales (0CT3), selfemployed (0CT6), and unskilled labor (0CT5). Occupational information was also used in constructing a composite opportunity variable for this study, which was also found to be significant in the regression equation. The impact of each of these variables as part of the regression equation is briefly discussed below. 132 Sales. For taxpayers who listed sales as their occupation, the estimated regression coefficient in this study is .58912, indicating that when controlling for other independent variables, these taxpayers disclosed about 58.912 percent more unreported income than taxpayers who listed their occupation as one not included in the regression equation (the omitted occupation variables include professional, professional support, skilled labor, and retired/student). The 95 percent confidence interval for the coefficient in measuring the population parameter runs from .39328 to .78496. population This indicates that (with 95 percent confidence) the parameter would disclose thatthese taxpayers reported from 39.3 to 78.5 percent more unreported income than taxpayers who are part of the omitted occupational categories in this regression. Self-Employed. The estimated regression coefficient for self-employed taxpayers is .53310, with a 95 percent confidence interval ranging from .28095 to .78525. These results can be interpreted as indicating that taxpayers who listed their occupation as self-employed disclosed 53.31 percent more unreported income than the omitted occupational categories, ceteris paribus. The 95 percent confidence interval indicates that the population parameter for taxpayers disclosing their occupation as selfemployed likely falls in a range from 28.1 unreported income than those taxpayers to 78.5 percent more in the omitted occupational categories. Unskilled Labor. For taxpayers who listed their occupation as one grouped in the unskilled labor category for this study, the estimated regression coefficient in this study is -.26126, indicating that when 133 controlling for other independent variables, these taxpayers disclosed about 26.126 percent less unreported income than the taxpayers who were grouped occupationally confidence interval in the omitted categories. The 95 percent for the coefficient in measuring the population parameter runs from -.43444 to -.08808. This indicates that (with 95 percent confidence) the population parameter would disclose that these taxpayers reported from 43.4 to 8 .8 percent less unreported income than taxpayers who are part of the omitted occupational categories. As previously mentioned, prior research can be considered generally indeterminate as to the impact of occupation on noncompliance. This study, however, concludes that certain occupations are more apt to be part of a noncompliance decision. activities were more laborers were less 1 1 Taxpayers in sales or self-employment ikely to evade than the base group, while unskilled ikely to evade. These findings generally support the survey research of Westat [1980b] (which found that employment in manufacturing or trade organizations was associated with higher compliance and the occupational categories and service employees were associated with lower levels of compliance) and the analytical of professional/managerial, study of Witte and clerical/sales, Woodbury [1985] (who found higher compl iance in jobs where withholding of taxes occurred [manufacturing]). The professional occupational regression equation. However, category was not significant in the these taxpayers are included in the opportunity variable if the other requirements of this composite variable are met (AGI greater than $30,000, and access to cash income sources). The results seem to conflict with the survey results of Mason and Calvin 134 [1978] who found occupational prestige significantly related only to failure to file; in their study occupation was not associated to the underreporting of income or the overstatement of deductions. Non-Filers in 1986 (N0FILE86). As mentioned in Chapter 5, this variable was included in the study as a surrogate for an attitudinal variable related to the noncompliance decision (i.e., do noncompliant taxpayers have an attitude regarding reporting that will continue even after an amnesty program?). The regression coefficient for this variable is .17434. We can conclude that these taxpayers disclosed 17.434 percent more unreported income than those amnesty participants who filed a tax return in 1986. The 95 percent confidence interval for this coefficient ranges from .05739 to .29130. This group of taxpayers apparently decided to disclose themselves during amnesty, and subsequent to amnesty. then returned to their noncompliant behavior In addition, this study concludes that they had more unreported income than amnesty participants who filed a return in 1986. The real question is why they made this choice. It would appear that fear of sanctions was not a concern, while ethics, or the moral implications of tax evasion may help to explain this behavior. Economic literature related to tax evasion stresses the fear of sanctions as a deterrent to tax evasion, and empirical evidence (survey, experimental, and analytic research) generally supports this contention (see, for example, Mason and Calvin [1978 and 1984], Friedland [1982], and Witte and Woodbury [1985]). Apparently, this group of taxpayers had no fear of sanctions as a result of their continued noncompliance. The Department of Treasury promised to ask no questions or put people on a 135 "hit list" if they filed during the amnesty program. Perhaps the 1986 nonfilers relied on this promise as a basis for not filing a return. Several studies have focused on tax compliers, and determined that these individuals are more likely to view tax evasion as seriously immoral or wrong (see, for example, Tittle [1980], and Scott and Grasmick [1981]). In general, taxpayers who are noncompliant do not accept those beliefs. This conclusion is supported by the survey research of Song and Yarbrough [1978] who found that when tax evasion was compared to other crimes (both violent and property-related), evasion was not viewed as a serious crime. It may be that the amnesty participants in this group do not view evasion as a serious crime. Tax ethics is a difficult concept to define. difficult to determine when a taxpayer meets threshold as it relates to tax evasion. It is even more the "moral failure" The results of this study would support the conclusions of previous research that indicates a need to explore specific ethical measures and their relationship to tax evasion. Descriptive statistics related to this group of amnesty participants are discussed in Chapter 4. Paid Preparers Used to Prepare Returns (PDPREP). coefficient for this variable is .16548, indicating The regression that amnesty participants who used paid tax-return preparers disclosed about 16.5 percent more unreported income than amnesty participants who their own return. The 95 percent confidence interval prepared for this coefficient ranges from .05736 to .27360. This study can make no conclusion regarding the implications of return preparers assisting in the noncompliance decision; the only conclusion that can be made is that these returns reported more income. 136 Ann Arbor and Washtenaw County (SMSA2). Based on the regression coefficient, amnesty participants living in the city of Ann Arbor or Washtenaw county disclosed about 33.982 percent more unreported income than other amnesty participants. The 95 percent confidence interval for the population parameter ranges from about 7.8 percent to 60.2 percent. Although unable to generalize these findings, support the contention that geographic location noncompliance decision. this study does is a part of the The area indicated (Ann Arbor) might initially seem to conflict with the previous finding of Witte and Woodbury [1985] that compliance is generally higher in established areas populated by middle class whites and lower in poverty and high unemployment areas. However, Witte and Woodbury also noted, "disturbingly," that better educated areas with large student populations generally have low levels of compliance. Ann Arbor would certainly fit this description. Two other Standard Metropolitan Statistical Area (SMSA) variables (SMSA3 - Jackson, and SMSA9 - Benton Harbor) not significant at the .05 level, are significant at the .10 level, and one of these regions (Benton Harbor) does have an unemployment percentage that exceeds the statewide average. However, the current study did not specifically explore areas within the SMSA regions where poverty or high unemployment might exist. Therefore, no conclusions can be made since the SMSA variables in this study do not describe where the amnesty participants SMSA regions. 1 ived within the Future research should attempt to focus on segmenting taxpayers among known poverty and high unemployment regions and comparing these taxpayers with those outside these regions. Tax Year 1983 (RATE635). Although the state income tax in Michigan is a flat tax on a tax base derived from federal AGI, during several 137 years in the early 1980’s the flat rate varied from 4.6 percent (the "normal rate" during most years) to 5.1 percent in 1982, 6.35 percent in 1983, 5.85 percent in 1984, and 5.33 percent in 1985. The regression coefficient indicates that amnesty participants who filed a 1983 tax return revealed about 13.441 percent less unreported income than other amnesty participants, ceteris paribus. The 95 percent confidence interval indicates a population parameter in the range of -.24190 to.02691. At face value, this result might appear to indicate that higher tax rates are associated with higher compliance -- a view mildly supported by analytical modeling (see, for example, Allingham and Sandmo [1972], and Yitzhaki [1974]). In fact, although not in the regression equation, two other tax rate years (1985 [RATE533] and 1984 [RATE585]) would also have negative regression coefficients if included in the regression equation. However, several alternative explanations exist. First, this finding could be a reflection of the fact that amnesty participants were disclosing more significant amounts of previously unreported income in years prior to 1983 (i.e., that the more significant noncompliant taxpayers filed for multiple years including one or more prior to 1983). Second, this finding could al so be a reflection of the general economic conditions at the time. It could be that because the economy was good during the years in question, taxpayers were more apt to report and pay taxes on all their income (and, therefore, less income was disclosed during amnesty). Third, this finding could indicate that these taxpayers, although filing for amnesty, did not disclose al_l of their unreported income during these higher rate years. 138 Self-Employed Taxpayers (SELFEMPL). This variable was created as an alternative to the occupational coding, by examining other information available in the data base (the lack of withheld taxes and a Form W-2 in the filings during amnesty). The regression indicates that these amnesty participants revealed about 29.934 percent more unreported income than participants who were not determined to be self-employed under the parameters indicated, ceteris paribus. Although significant, and of further support to the contention that self-employed taxpayers have a greater propensity to evade taxes, this variable is not as significant as the occupational variable. remember that the occupational However, variable was significant based on a comparison to taxpayers in the professional support category. comparison group is taxpayers who are not self-employed. Here, the Another reason may be the restrictiveness with which the variable was created -- any amount of withheld taxes would have categorized a taxpayer as not selfemployed. Gender. The majority of studies (survey and experimental research) testing the compliance level of males versus females have found males less compliant. analytically. males are coefficient However, this variable had not previously been tested This analytical study supports the position that single less compliant than other taxpayers. The regression indicates that males disclosed about 19.00 percent more unreported income than other taxpayers, ceteris paribus. This finding cannot be 1 inked to income level because females and married taxpayers filing a joint return in the data base had a higher level of average income (AGI) then males (i.e., the finding cannot be refuted using the argument that the reason males have a higher level of 139 noncompliance is because they earn more). It could, however, be linked to self-employment (almost twice as many males in the research data base are likely self-employed). The 95 percent confidence interval for the regression coefficient ranges from .05837 to .32165. As a result, the population parameter would indicate single males could disclose from 5.8 to 32.2 percent more unreported income than single females. Exemptions (EXEMPT). This study used exemptions as a surrogate for family size. The regression coefficient for this variable is .04797, which signifies that amnesty participants with twice as many exemptions relative to other amnesty participants disclosed about 4.797 percent more unreported income, ceteris paribus. The 95 percent confidence interval for the population parameter runs from .004179 to .09193. Other Variables. Several other items should be mentioned regarding the results of this regression analysis. First, some studies have indicated that filing status is a significant factor in the noncompliance decision. For example, Clotfelter [1983] who found married couples filing a joint return to be significantly less compliant than single individuals in non-business returns and no significant difference with business or farm returns. In this regression, no significant difference exists between single individuals and married couples filing a joint return (at the .05 level). However, there is a significant difference between the groups at the .1 0 level, with single taxpayers being less compliant than married couples filing a joint return. Second, although several studies have indicated that taxpayers in professional occupations are less compliant than other taxpayers (see, for example Westat [1980b]), this study does not support that conclusion. 140 However, taxpayers listing their occupation as one of those grouped in the professional category in this study are included in the opportunity variable if the other requirements of this composite variable were met (AGI greater than $30,000, and access to cash income sources). Finally, it was previously mentioned that five of the thirteen variables in the regression equation (AGILN, CONTACT, 0PP0RT, 0CT3, and 0CT6) accounted for about 94 percent of the variance accounted for in the dependent variable by all thirteen predictor variables. Although statistically significant, the meaningfulness of these variables as part of the noncompliance decision might be challenged. 6.3.4 Residual Analysis Figures 6-1 through 6-3 are a part of the output from the regression related to an analysis of residuals (the difference between the observed dependent variable and the value predicted by the model). Reviewing the residual information provides information regarding whether some of the assumptions of regression analysis (e.g., are tenable. 1 inearity, homoscedasticity) Probably the simplest and most useful is one in which standardized residuals are plotted against the standardized predicted dependent variables (Figure 6-1). If the points are randomly and evenly distributed in no apparent pattern, indicated. Figure In addition, 6-1 1 inearity and homoscedasticity are presents such a pattern. residuals are assumed to be normally distributed. Figure 6-2 presents a histogram of standardized residuals superimposed over a normal curve, given the mean and variance of the residuals in the regression. Although not perfect, the residuals do appear to have a near normal distribution. As can be seen from the histogram, the model is overpredicting certain cases (i.e., there are outliers at the bottom of 141 the histogram). The data related to each outlier case were examined for potential data coding errors. this process. No data errors were found as a result of However, all outlier cases made very small tax payments (less than $10.00). The conclusion, therefore, is that the model does not fit well for amnesty participants making small tax payments (i.e., for these participants, the model predicts a level of unreported income in excess of actual unreported income). Another way to compare the observed distribution of residuals to that expected under the assumption of normality is to plot the two cumulative distributions against each other for a series of points. If the two distributions are identical, a straight 1 ine results. By observing how points scatter about the expected straight possible to compare the two distributions. probability plot of the residuals. 1 ine, it is Figure 6-3 is a cumulative Again, a near normal distribution is indicated. 6.3.5 Summary Although thirteen variables are included as significant factors in the regression analysis of the entire research data base, five of these variables account for the vast majority of the variance explained by the variables, and of these five, two variables (income level and contact with the Treasury Department) dominate. Because of the domination of the income level variable, and to further explore the data, the research data base was stratified into five AGI levels. bases occurs later in this Chapter. An analysis of these data 142 FIGURE 6-1 RESERACH DATA BASE: SCATTERPLOT OF STANDARDIZED RESIDUALS AGAINST STAIDARDIZED PREDICTED SCORES Across - *ZPRED 3 2 Down - *ZRESID H--------- 1 --------- 1 — Out - 1 • k k. •kk kkk • •k ■ ■ •••' ; A . .2•• « kmkkmkkmkm. k mk »kk •• •• &•■ B kkkkmmkk k ■ k mk m • &4r •itm m kk k. mkkkk kk m kk • kkm * k m km m k Out -1 Symbols: Max N : * 2.0 4.0 10.0 2 Out 143 FIGURE 6-2 RESERACH DATA BASE: NExp N 0 0 0 0 0 0 0 5 6 3 * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * 6 6 6 5 1 * 2.71 1.38 1.98 2.82 3.94 5.43 7.36 9.83 12.9 16.7 21.3 26.7 32.9 40.0 47.9 56.4 65.4 74.7 84.0 92.9 101 109 115 119 122 123 122 119 115 109 101 92.9 84.0 74.7 65.4 56.4 47.9 40.0 32.9 26.7 21.3 16.7 12.9 9.83 7.36 5.43 3.94 2.82 1.98 1.38 2.71 HISTOGRAM OF STANDARDIZED RESIDUALS . : = Normal Curve) (:* = 2 Cases, Out 3.00 2.88 2.75 2.63 2.50 2.38 2.25 *** 2.13 *** 2.00 ** 1.88 ***** 1.75 ******* 1.63 ***************' 1.50 ******************&■**** 1.38 ***********************. *** 1.25 * * * * * * * * * * * * * * * * * * * * * * * * * * * • * * * 1.13 **************************************** 1.00 A * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * . * * * * * * .88 *****************************************■**************** .75 *********************************************■*************** .63 **************************************************.************* .50 ********************************************************************* .38 ********************************************************.********** .25 ***********************************************************•****** .13 0.0 -.13 -.25 -.38 -.50 -.63 -.75 -.88 -1.00 -1.13 -1.25 -1.38 -1.50 -1.63 -1.75 -1.88 -2.00 -2.13 -2.25 -2.38 -2.50 -2.63 -2.75 -2.88 -3.00 Out * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * .* * * * * * * * * * * * * *************************************************************.***** *********************************************************** ***************************************************** ******************************************** ' ********************************************** ************************************************* ********************************* ************************** **************************** ************************ *********************** ********************** ******************* ************* ************ # **********■ *******.* *****• **** * * * * >** * * * .* * * * * . * .* .* * .* * •* * * * * * * * * * 144 FIGURE 6-3 RESERACH DATA BASE: NORMAL PROBABILITY PLOT OF STANDARDIZED RESIDUALS Standardized Residual 1.0 ** < T (t H 5 7 0 75 -■ ** a(5 ** 25 -- ** *** E x p e c t e d 145 6.4 Regression Analyses of the Non-Contact Data Bases 6.4.1 In General As was discussed in Chapter 5, two data bases were created to explore noncompliance decision factors for those amnesty participants who could be classified as not having had any contact with the Treasury Department (as defined in this study) prior to amnesty. The no contact taxpayers (roughly 630 1 ines of data out of 2470 in the entire research data base) amnesty. paid in about 86 percent of their tax 1 iability during To discriminate between the contact and no-contact amnesty participants, approximately the same number of data 1 ines were selected from taxpayers known to Treasury who made only small payments during amnesty (these taxpayers paid in approximately 23 percent of their tax 1iability during amnesty). These two sets of data were merged together into a data base (titled CONNCON.SYS) for analysis. A second data base (titled CONNCON2.SYS) was also created by removing those taxpayers whose occupation was coded as 0CT7 (retired, students, and other) from the initial data base. 6 This section of Chapter discusses the regression analyses of these data bases, and compares the results with that of the entire research data base. 6.4.2 Analysis of Contact/No Contact Data Base (CONNCON.SYS) Table 6-4 presents the regression results for the contact/no contact data base (C0NNC0N). 1 The analysis of the regression occurs along several ines, including a discussion of the goodness of fit (R^), an analysis of the regression coefficients and their confidence intervals, comparison of the results with the entire research data base. and a 146 TABLE 6-4 REGRESSION TO PREDICT NONCOMPLIANCE USING CONNCON DATA BASE Variable Name Coefficient (Standard Error) t (Significance) R2 Adjusted SELFEMPL 1.50225 (.14230) 10.557 (.0 0 0 0 ) .0979 .0971 0CT3 1.15694 (.16398) 7.055 (.0 0 0 0 ) .1361 .1347 OPPORT 1.84707 (.38191) 4.836 (.0 0 0 0 ) .1702 .1682 0CT6 .87073 (.18912) 4.604 (.0 0 0 0 ) .1842 .1816 N0FILE86 .38644 (.09270) 4.169 (.0 0 0 0 ) .1948 .1915 GENDER .50863 (.09055) 5.617 (.0 0 0 0 ) .2029 .1990 AGILN .29101 (.04835) 6.018 (.0 0 0 0 ) .2166 0CT7 .56012 (.12371) 4.528 (.0 0 0 0 ) .2357 .2307 SMSAIO -.58539 (.17085) -3.426 (.0006) .2406 .2351 0CT5 -.36186 (.14436) -2.507 (.0123) .2440 .2379 SMSA1 -.21078 (.08746) -2.410 (.0161) .2475 .2408 RATE635 -.18244 (.08762) -2.082 (.0375) .2502 .2429 (Constant) 3.77791 (.49575) 7.621 (.0 0 0 0 ) .2 1 2 1 7 147 6.4.2.1 Goodness of Fit The model has a R 2 of .2502, and an adjusted R 2 of .2429 with 12 variables in the equation. As a result, the independent variables in the model account for about 25 percent of the variance in the dependent variable (unreported income). This result is similar to (although about 1.3 percent smaller than) the regression analysis of the entire research data base. The order of the variables into the equation, and their impact on R 2 is as follows (all variables except AGILN are dummy variables). The self-employment variable created via an examination of the data other than disclosed occupation (SELFEMPL) enters the equation first with a R 2 of .0979. Taxpayers who disclosed an occupation of sales (0CT3) enters the equation next and increases R 2 by .0383 to .1361. Opportunity to evade (OPPORT) then increases R 2 by .0341 to .1682. Taxpayers who disclosed an occupation of self-employed (0CT6) then enters the equation and increases R 2 by .0140 to .1842. signify amnesty participants who The dummy variable created to failed to file a return in 1986 (N0FILE86) enters the equation next, increasing R 2 by .0105 to .1948. The variable constructed to indicate single males (GENDER) enters the regression equation next, increasing R 2 by .0081 to .2029. The log of AGI (AGILN) enters the equation next and adds .0137 to the R2 , increasing it to .2166. The next variable to enter is 0CT7, which generally represents retired and student taxpayers. .2357. It increases R 2 by .0191 to The next variable to enter is SMSA10, which represents taxpayers in the Grand Rapids/Holland SMSA. It increases R 2 by .0050 to .2406. The next variable to enter the model is 0CT5 (representing unskilled labor), contributing an additional .0034 to the R2 , increasing it to 148 .2440. The variable representing taxpayers in the Metropolitan Detroit SMSA (SMSA1), enters the equation next, and increases .2475. by .0035 to The last variable to enter the model is RATE635 (representing returns filed when the tax rate was 6.35 percent). additional It contributes an .0026 to the R^, increasing it to .2502. There are no additional variables that would be added to the regression equation at a significance level of .1 0 . As can be seen from Table 6-4 and the foregoing discussion, several differences are immediately visible when these results are compared to those of the entire research data base. First, although the self- employment variable is strong, it does not dominate the analysis as the level of income variable dominated the analysis of the entire research data base. Second, regression equation, the level but of income variable it is the is part of the seventh variable to enter the equation. Third, although several variables are strong (as would be expected in stepwise regression), eight of the twelve variables make substantial contributions to the explanation of the variance in the dependent variable; and, the last four variables to enter the regression equation (contributing .0050, .0034, .0035, and .0026 to R^), together increase R^ about as much as the last eight variables to enter the regression equation for the entire research data base (the contributions of these last eight variables ranges from .0014 to .0030). analysis provides stronger evidence for several As a result, this of the variables. Fourth, the standard errors of the regression coefficients are generally greater than those in the research data base. Although this is to be expected as a result of the data set (the independent variables have 149 larger variances associated with them), it will lead to decreased precision in estimating the population parameters since the 95 percent confidence intervals will be larger. Finally, several differences exist regression equation. among the variables in the The Ann Arbor SMSA variable (SMSA2), the variable indicating the number of exemptions claimed on the return (EXEMPT), and the variable related to paid preparation of the return (PDPREP) are not a part of this regression equation. Three new variables are a part of this regression equation -- the occupational variable indicating (generally) retired or student (0CT7), and two SMSA variables (SMSA10 - Grand Rapids/Holland, and SMSA1 - Metropolitan Detroit). 6 .4.2.2 Analysis of Regression Coefficients and Confidence Intervals Table 6-5 presents the regression coefficients for each of the independent variables standard mentioned error and previously, in the computed because regression equation, 95 of percent the confidence larger along with their intervals. standard errors, As the confidence intervals are larger and precision regarding estimation of the population parameters is weakened in certain cases. Self-Employed Taxpayers (5ELFEMPL). The regression indicates that these amnesty participants revealed about 150.225 percent more unreported income (vs. about 29.9 percent for the entire research data base) than participants who were not determined to be self-employed under the parameters indicated, ceteris paribus, variable in the regression equation. and is the most significant In this analysis, this variable is of greater significance than the self-employment occupational variable. This result further corroborates the fact that self-employed taxpayers, 150 TABLE 6-5 CONFIDENCE INTERVALS OF REGRESSION COEFFICIENTS FOR CONNCON DATA BASE Variable Name Coefficient Standard Error SELFEMPL 1.50225 .14230 1.22308 1.78142 0CT3 1.15694 .16398 .83523 1.47866 OPPORT 1.84707 .38191 1.09779 2.59635 0CT6 .87073 .18912 .49969 1.24177 N0FILE86 .38644 .09270 .20457 .56830 GENDER .50863 .09055 .33097 .68629 AGILN .29101 .04835 .19615 .38587 OCT7 .56012 .12371 .31742 .80282 SMSA10 -.58539 .17085 -.92058 -.25019 OCT5 -.36186 .14436 -.64508 -.07864 SMSA1 -.21078 .08746 -.38236 -.03920 RATE635 -.18244 .08762 -.35434 -.01054 (Constant) 3.77791 .49575 2.80531 4.75052 95 Percent Confidence Interval 151 because of their access to cash income sources and control over net income reported, are more likely to evade taxes. The standard error of this regression coefficient (.14230) is comparable to its parallel variable in the regression of the entire research data base (.12137). As a result, precision regarding the population parameter is about the same as it was in the regression of the entire research data base. Here, the 95 percent confidence interval indicates that the population parameter would fall in the range of 1 2 2 .3 percent to 178.1 percent more unreported income than taxpayers who are self-employed. Opportunity to Evade (OPPORT). Not surprisingly, opportunity to evade taxes is a significant variable in this analysis. The estimated regression coefficient of the variable is 1.84707, with a 95 percent confidence interval of 1.09779 to 2.59635. The standard error of the coefficient (and, therefore, the 95 percent confidence interval) is about twice as large as the standard error of this variable in the analysis of the entire research data base. It would appear that some taxpayers with opportunity paid small amounts of amnesty tax. However, although the precision is not good, the significance of the variable is impressive. The regression coefficient indicates that the amnesty participants in this data base with the opportunity to evade taxes on average underreported 184.707 percent more income than amnesty participants without the opportunity characteristics, ceteris paribus. The 95 percent confidence interval indicates that the population parameter could be as low as 109.8 percent or as high as 259.6 percent (i.e., those with opportunity revealed at least twice as much previously unreported income as those without opportunity). 152 Occupations (0CT3. 0CT6, 0CT7. OCT51. Four different occupation variables are part of this regression equation. Three -- sales (0CT3), self-employed (0CT6), and unskilled labor (0CT5) are the same as in the regression of the entire research data base. relates generally to retirees and students. The new variable (0CT7) All are but one (0CT5) are positively related to the dependent variable. The impact of each of these variables as part of the regression equation is briefly discussed below. Sales. For taxpayers who listed sales as their occupation, the estimated regression coefficient in this study is 1.15694, indicating that when controlling for other independent variables, these taxpayers disclosed about 115.694 percent more unreported income than taxpayers in the occupations omitted from the regression equation. This compares to 58.912 percent in the regression of the entire research data base. 95 percent confidence interval The for the coefficient in measuring the population parameter runs from .853523 to 1.47866. The standard error of the regression coefficient is close to twice as large as that of the entire research data base. Again, the precision of the parameter estimation is only fair. Self-Employed. The estimated regression coefficient for self-employed taxpayers in this regression is .87073, with a 95 percent confidence interval ranging from .49969 to 1.24177. Taxpayers who 1isted their occupation as self-employed disclosed about 87.07 percent more unreported income than taxpayers in the occupations omitted from the regression equation, ceteris paribus (vs. 53.3 percent in the entire research data 153 base regression, indicating a higher likelihood of these taxpayers in the no contact group). The standard error of the regression coefficient is about 50 percent larger than that of the coefficient in the previous regression, again negatively impacting precision. Unskilled Labor. These taxpayers again are found to be more compliant. The estimated regression coefficient here is -.35186, indicating that when controlling for other independent variables, these taxpayers disclosed about 35.186 percent less unreported income than taxpayers in the occupations omitted from the regression equation. The coefficient in the study of the research data base was -.26126. This difference indicates that blue collar workers disclosed less unreported income, and are less 1 ikely to be among the no contact group. confidence interval The 95 percent for the coefficient in measuring the population parameter runs from -.64508 to -.07864. This indicates that (with 95 percent confidence) the population parameter would disclose that these taxpayers reported from 64.5 to 7.9 percent less unreported income than taxpayers in the occupations omitted from the regression equation. Retired/Student. Although not significant in the regression analysis of the entire research data base, this occupational category is significant in this comparison of taxpayers with and without contact. The regression coefficient is .56012, which indicates that these taxpayers disclosed about 56.012 percent more occupations omitted unreported income than taxpayers from the regression equation. The in the 95 percent confidence interval for the population parameter runs from .31742 to .81282. 154 When the inclusion of this variable in the regression equation is coupled with the reduced effect of the level of income variable (the regression coefficient decreases by about 50 percent), there appears to be some support to Witte and Woodbury’s contention that there is noncompliance among both high and low income groups. Again, it is important to stress that this finding does not imply that low income taxpayers are more than middle income taxpayers. 1 ikely to be advertent tax evaders Some, in fact, may be willful tax evaders. In addition, there may be a difference between students and retirees. It would seem more plausible that many of these taxpayers may not be filing because of 1 ack of knowledge, confusion over the tax laws, or a real belief that they owe no taxes. Non-Filers in 1986 (NOFILE 8 6 ). This category again is a significant variable in the noncompliance decision as it relates to taxpayers without prior contact. Here, the regression coefficient is .38644 (more than twice the size of the coefficient in the entire research data base [.17434]). We can conclude that these taxpayers disclosed 38.644 percent more unreported income than those amnesty participants who filed a tax return in 1986, and that taxpayers who did not file a return in 1986 were more 1 ikely to be part of the no contact group. The 95 percent confidence interval for this coefficient ranges from .20457 to .56830 (vs. .05739 to .29130 in the previous regression). Obviously, the people who failed to file may have done so for a variety of reasons. This finding also provides support to the contention that advertent noncompliant taxpayers (who 1 ikely make up a portion of 155 this group) have different ethics than those of compliant taxpayers. Further research is needed to determine reasons for this behavior. Gender. Gender was a more significant variable in this analysis than in the previous regression. Here, the regression coefficient was .50863 (as compared to .19001 in the analysis of the entire research data base), strengthening the contention that single males are more evade than other taxpayers. significant segment 1 ikely to In addition, males appear to be a more of taxpayers without contact. The regression coefficient indicates that single males disclosed about 50.863 percent more unreported income than other taxpayers, ceteris paribus. The 95 percent confidence interval for the regression coefficient ranges from .33097 to .68629 (vs. .05837 to .32166 in the previous regression). As a result, the population parameter would indicate single males could disclose from 33.1 to 6 8 .6 percent more unreported income than other taxpayers. Log of AGI (AGILN). The estimated regression adjusted gross income is .29101. coefficient of Thus, amnesty participants in this data base with twice as much adjusted gross income relative to other amnesty participants had (after adjustment) about 22.4 percent more unreported income in AGI, ceteris paribus. As mentioned previously, this regression coefficient is about 50 percent less than in the analysis of the entire research data base (.58888). So, although there is still a relationship between level of income and level of noncompl iance, it is not as significant here as in the entire research data base. Part of the reduced significance could be due to the fact that the retired/student occupational group (and their lower levels of income) are a larger part of the no contact group. This 156 would mitigate the effect of this variable. The inclusion of contact taxpayers with low levels of amnesty payments in the data base (as surrogates for compliant taxpayers) also obviously mitigated the impact of this variable. The 95 percent confidence interval for this coefficient ranges from .19615 to .38587, indicating that the population parameter would fall in this range 95 percent of the time. The range indicates that an amnesty participants in this data base with twice as' much in adjusted gross income as another individual participant could have (after adjustment) from 14.6 to 30.7 percent more unreported income, ceteris paribus. Geographic Regions (SMSA10 and SMSA11. Two Standard Metropolitan Statistical Area (SMSA) variables are a part of this regression equation, and both are negatively related to noncompliance. Grand Rapids/Holland (SMSA101. SMSA10 includes the counties of Kent and Ottawa, and the cities of Grand Rapids and Holland. Based on the regression coefficient, amnesty participants in this data base who live in Kent or Ottawa county disclosed about 58.539 percent less unreported income than taxpayers equation. The 1 iving 95 percent in areas confidence omitted from the regression interval for the population parameter ranges from about 92.1 percent to 25.0 percent less unreported income. It is al so possible to conclude that taxpayers 1iving in this area are less likely to be part of the no contact group. Metropolitan Detroit fSMSAll. SMSA1 all includes a large southeastern Michigan, including of metropolitan regression coefficient indicates that for this data base, part Detroit. of The taxpayers 157 living in this area disclosed 21.078 percent less unreported income than taxpayers living in areas omitted from the regression equation. At face value, it might appear that this information might refute the finding of Witte and Woodbury regarding low compliance in unemployment and poverty areas -- since this region includes many areas of unemployment and poverty. Although interesting, however, the Witte and Woodbury finding cannot be refuted because the SMSA variables do not describe where taxpayers 1 ive within these areas. However, it does appear that this region has fewer members of the no contact group. Tax Year 1983 (RATE635). The regression equation includes this variable (as in the regression equation for the entire research data base) with a regression coefficient of -18.244, indicating that amnesty participants part of this data base who filed a 1983 tax return revealed about 18.244 percent less unreported income than amnesty participant who filed a return during a year when the tax rate was 4.6 percent (the base dummy variable in this study), ceteris paribus. The 95 percent confidence interval indicates a population parameter in the range of.24190 to -.02691. Other Variables. A few other comments are warranted regarding this analysis. Filing status was not a significant variable in this analysis at either the .05 level or at the .10 level. This would indicate that there is no statistical difference between single and married taxpayers in the no contact group. Second, professional occupations were found again to be no more likely to evade than other occupational groups. A caveat is in order for the opportunity classification, as was mentioned in the discussion of the entire research data base. 158 6 .4.2.3 Residual Analysis Figures 6-4 through analysis of residuals. 6 -6 present several figures used as part of the Figure 6-4 plots standardized residuals against the standardized predicted dependent variables. As the points are randomly and evenly distributed in no apparent pattern, linearity and homoscedasticity are indicated. Figure 6-5 presents a histogram of standardized residuals superimposed over a normal curve, given the mean and variance of the residuals in the regression. of the residuals. Figure 6-6 is a cumulative probability plot Although not perfect, the residuals do appear to have a near normal distribution. 6 .4.2.4 Summary This regression provides stronger support for the inclusion of several variables in the noncompliance decision paradigm. statistically significant, and therefore, Although a part of the regression equation for the entire research data base, gender, self-employment, and continued nonfilers contributed only mildly to the explanation of the variance in the dependent variable. In this analysis of taxpayers with no contact, however, they make significant contributions to the model. In addition, the significant (and powerful) variables in the initial regression reappear in this model (sales, self-employed, and unskilled labor occupations; dominance. opportunity; level of income), but without their Finally, the inclusion of the retired/student occupational group along with the mitigation of the level of income variable provides some mild support for the contention that low income taxpayers are part of the noncompliance puzzle. 159 FIGURE 6-4 COHNCON DATA BASE: SCATTERPLOT OF STANDARDIZED RESIDUALS AGAINST STANDARDIZED PREDICTED SCORES Across - *ZPRED Down - *ZRESID Out Out Out Symbols: Max N 2.0 4.0 9.0 160 FIGURE 6-5 CONNCGN DATA BASE: NExp N 0 1.36 2 .69 1 1.00 1 1.42 5 1.98 * 2.73 0 3.70 3 4.93 5 6.48 5 8.38 * 10.7 * 13.4 * 16.5 * 20.1 * 24.0 * 28.3 * 32.8 * 37.5 * 42.2 * 46.7 * 50.8 * 54.5 * 57.6 * 59.9 * 61.3 * 61.8 * 61.3 * 59.9 * 57.6 * 54.5 * 50.8 * 46.7 * 42.2 * 37.5 * 32.8 * 28.3 * 24.0 * 20.1 9 16.5 6 13.4 * 10.7 * 8.38 5 6.48 9 4.93 2 3.70 1 2.73 5 1.98 1 1.42 4 1.00 0 .69 3 1.36 i [* = 1 Cases, Out 3.00 2.88 2.75 2.63 2.50 2.38 2.25 2.13 2.00 1.88 1.75 1.63 1.50 1.38 1.25 1.13 1.00 .88 .75 .63 .50 .38 .25 .13 0.0 -.13 -.25 -.38 -.50 -.63 -.75 -.88 -1.00 -1.13 -1.25 -1.38 -1.50 -1.63 -1.75 -1.88 -2.00 -2.13 -2.25 -2.38 -2.50 -2.63 -2.75 -2.88 -3.00 Out HISTOGRAM OF STAWARDIZED RESIDUALS . : = Normal Curve) :* ; * ■*** ********** m *** ***** ***** ************ ************• ****************** *******************•* ***********************-** ************************** *********************** ********************************** ************************************ ************************************************** *************************************** ******************************************************•* *********************************************************• ************************************************************ ********************************************************* ******************************************************************************* ************************************************************ ************ ************************************************************ ********************* ********************************************************* ********** ******************************************* ****************************************** **********************************************■******* **************************************************** **************************************** ********************************b ******************** **************** ************ ********* ****** **********a *******-****** *****a ********* ** * ***** : .*** .** 161 FIGURE 6-6 GOMMGON DATA BASE: NORMAL PROBABILITY PLOT OF STANDARDIZED RESIDUALS Standardized Residual 1 Q . f c < - J < D W 0 ’O .75 ** ** ** 25 -** Expected E x p e c t e d 162 6.4.3 Analysis of Contact/No Contact Data Base Excluding the Retired/Student Occupational Group (C0NNC0N2.SYS) Table 6-6 presents the regression results for the contact/no contact data base excluding the retired/student occupational group (C0NNC0N2). The analysis of the regression will be limited to a review of the goodness of fit (R^), and a comparison of the regression coefficients to the Contact/No Contact data base including the retired/student occupational group. 6.4.3.1 Goodness of Fit The model has a R^ of .2795, and an adjusted R^ of .2713 with 11 variables in the equation. Therefore, the independent variables in the model account for almost 28 percent of the variance in the dependent variable (unreported income). This result is about 3 percent greater than the analysis of the CONNCON data base, and about 2 percent greater than the analysis of the entire research data base. Two predictors in the CONNCON regression equation are not a part of this equation -- the retired/student occupational group (obviously) representing tax year 1983 (RATE635) -- and a new and the variable predictor variable appears -- number of exemptions claimed on the return (EXEMPT). Not surprisingly, this regression is similar to that of the CONNCON data base. Of the variables in the equation, the first four variables are identical to the CONNCON regression and in the same order (SELFEMPL, 0CT3, OPPORT, and 0CT6); the next five variables are the same as in the CONNCON regression (excluding 0CT7) but in a siightly different order (AGILN, GENDER, N0FILE86, SMSA10, and 0CT5). The impact on R^ and their order in the equation, is as fol 1ows (all variables except AGILN and EXEMPT are dummy variables). The self- 163 TABLE 6-6 REGRESSION TO PREDICT NONCOMPLIANCE USING CONNCON2 DATA BASE Variable Name Coefficient (Standard Error) t (Significance) R2 Adjusted R^7 SELFEMPL 1.48814 (.14742) 10.095 (.0 0 0 0 ) .1078 .1069 OCT3 1.10215 (.17310) 6.367 (.0 0 0 0 ) .1532 .1515 OPPORT 1.80217 (.40125) 4.491 (.0 0 0 0 ) .1923 .1898 0CT6 .87154 (.19862) 4.388 (.0 0 0 0 ) .2109 .2076 AGILN .28616 (.05789) 4.944 (.0 0 0 0 ) .2269 .2230 GENDER .73394 (.12163) 6.034 (.0 0 0 0 ) .2457 .2410 N0FILE86 .48096 (.10780) 4.462 (.0 0 0 0 ) .2622 .2569 SMSAIO -.66824 (.20224) -3.304 (.0 0 1 0 ) .2683 .2622 OCT5 -.38250 (.15195) -2.517 (.0 1 2 0 ) .2719 .2651 EXEMPT .09232 (.04192) (.0279) .2760 .2685 SMSA1 -.22488 (.10322) -2.179 (.0296) .2795 .2713 (Constant) 3.48266 (.56582) 6.155 (.0 0 0 0 ) 2 .2 0 2 164 employment variable created via an examination of the data other than disclosed occupation (SELFEMPL) enters the equation first with a R 2 of .1078 (vs. .0979 in the CONNCON regression). The occupation of sales (0CT3) enters the equation next and increases R 2 by .0454 (vs. .0383 in the CONNCON regression). Opportunity to evade (OPPORT) then increases R 2 by .0391 (vs. .0341 in the CONNCON regression). An occupation of selfemployed (0CT6) then enters the equation and increases R 2 by .0186 (vs. .0140 in the CONNCON regression). The log of AG I (AGILN) enters the equation next and adds .0161 to the R2 , (vs. .0137 in the CONNCON regression). The variable constructed to indicate single males (GENDER) enters the regression equation next, increasing R 2 by .0188 (vs. .0081 in the CONNCON regression). The dummy variable created to signify amnesty participants who failed to file a return in 1986 (N0FILE86) enters the equation next, increasing R 2 by .0165 (vs. .0105 in the CONNCON regression). The next variable to enter is SMSA10, which represents taxpayers in the Grand Rapids/Holland SMSA. It increases R 2 by .0060 (vs. .0050 in the CONNCON regression). The next variable to enter the model is 0CT5 (representing unskilled labor), contributing an additional .0036 to the R2 , (vs. .0034 in the CONNCON regression). The number of exemptions (EXEMPT) claimed on the return (a surrogate for family size) next enters the model, and increases R 2 by .0041. The variable representing taxpayers in the Metropolitan Detroit SMSA (SMSA1), enters the equation last, and increases R 2 by .0035 (the same as in the CONNCON regression). Dropping the retired/student occupational category strengthens the effect of several variables (GENDER and N0FILE86 most notably), while decreasing the effect of none (SMSA10, 0CT5, and SMSA1 increase R 2 by 165 almost the same amounts as in the CONNCON regression). There are no additional variables that would be added to the regression equation at a significance level of .1 0 . 6.4.3.2 Analysis of Regression Coefficients and Confidence Intervals Table 6-7 presents the regression coefficients for each of the independent variables in the regression equation, along with their standard error and computed 95 percent confidence intervals. Although most regression coefficients are similar to their counterparts in the CONNCON analysis (i.e., the removal of the retired/student occupational group had no large impact on certain variables), several change significantly (GENDER, N0FILE86, and SMSA10), and are discussed briefly below. The regression coefficients have standard errors similar to those of the analysis of the CONNCON data base, as expected. is that the confidence intervals are 1 The result, again, arger and precision regarding estimation of the population parameters is weakened as compared to the research data base. Gender. Gender, once again, has a larger regression coefficient in this analysis than in the previous regression. Here, the regression coefficient is .73394 (as compared to .50863 in the analysis of CONNCON and .19001 in the analysis of the entire research data base). This finding suggests that among unknown taxpayers (exclusive of retirees and students, who in this study had comparatively lower income levels), males are much more noncompliant than other taxpayers. The regression coefficient indicates that single males disclosed about 73.394 percent more unreported income than other taxpayers, ceteris paribus. When the 166 TABLE 6-7 CONFIDENCE INTERVALS OF REGRESSION COEFFICIENTS FOR CONNCON2 DATA BASE Variable Name Coefficient Standard Error SELFEMPL 1.48814 .14742 1.19884 1.77744 OCT3 1.10215 .17310 .76245 1.44184 OPPORT 1.80217 .40125 1.01475 2.58959 OCT 6 .87154 .19862 .48176 1.26132 AGILN .28616 .05789 .17257 .39976 GENDER .73394 .12163 .49525 .97264 N0FILE86 .48096 .10780 .26942 .69251 SMSA10 -.66824 .20224 -1.06512 -.27137 OCT5 -.38250 .15195 -.68069 -.08431 EXEMPT .09232 .04192 .01006 .17458 SMSA1 -.22488 .10322 -.42745 -.02231 (Constant) 3.48266 .56582 2.37228 4.59303 95 Percent Confidence Interval 167 retired/student occupational group is removed from the analysis, the proportion of unreported income among single males increases by about 44 percent. This result was expected based on the descriptive statistics of gender and income levels in the retired/student occupational group. Within this group, there is a higher proportion of females, and generally lower levels of income, than in the entire research data base. The 95 percent confidence interval for the regression coefficient ranges from .49525 to .97264 (vs. .33097 to .68629 in the CONNCON regression). As a result, the population parameter would indicate single males could disclose from 49.5 to 97.3 percent more unreported income than other taxpayers. Non-Filers in 1986 (N0FILE86). The regression coefficient for this category of amnesty participants is .48096 (vs. .38644 in the CONNCON analysis and retired/student .17434 in the entire research data base). occupational group is removed from the When the data base, nonfilers in 1986 disclose about 24 percent more unreported income than when the retired/student occupational group is in the analysis. There are several explanations to this increase. First, based on the descriptive analysis of the retired/student taxpayer group, a smaller proportion failed to file a return in 1986 when compared to the entire research data base. more 1 For some reason, the taxpayers in this group were ikely to file after amnesty -- maybe they were fearful of sanctions subsequent to amnesty; maybe they didn’t realize that they needed to file prior to amnesty and once they understood, they complied; maybe they had been noncompliant prior to amnesty, were fighting a guilty conscience, and when amnesty came along, they took the opportunity to make amends, 168 and return to their better ethics towards compliance after the amnesty program. Second, this result could be expected because of the generally lower income levels among the retired/student group. In other words, those amnesty taxpayers in the retired/student occupational group who did not file in 1986, disclosed lower levels of income during amnesty than other nonfilers in 1986. Finally, it appears that taxpayers in the retired/student group who filed during amnesty were more likely to continue filing post-amnesty than other taxpayers in the no contact group. The 95 percent confidence interval for this coefficient ranges from .26942 to .69251 (vs. .20457 to .56830 in the CONNCON regression and .05739 to .29130 in the entire research data base regression). Grand Rapids/Holland SMSA (SMSA10). As mentioned above, SMSA10 includes the counties of Kent and Ottawa, and the cities of Grand Rapids and Holland. The regression coefficient for this category of amnesty participants is -.66824 (vs. -.58539 in the CONNCON analysis), indicating that amnesty participants part of this data base 1 iving in Kent or Ottawa county disclosed taxpayers 1 about 66.824 percent less unreported income than iving in areas omitted from the regression equation. When the retired/student occupational group is removed from the data base, remaining taxpayers in the Grand Rapids/Holland SMSA disclose even less unreported income (about 14 percent less) retired/student occupational group is in the analysis. than when the It would appear that taxpayers in this SMSA are less likely to be in the no contact group, and more likely to report smaller amounts of unreported income. 169 Exemptions (EXEMPT). Although not a part of the CONNCON regression equation, this variable appears as part of the regression equation when the retired/student occupational group is removed. The regression coefficient for this variable is .09232 (vs. .04797 for the regression on the entire research data base), signifying that amnesty participants in the C0NNC0N2 data base with twice as many exemptions relative to other amnesty participants disclosed about 9.232 percent more unreported income, ceteris paribus. Since the retired/student occupational group claimed fewer exemptions (on average) than all taxpayers without contact, it may be that these taxpayers were mitigating the effect of family size in the CONNCON analysis. Once they are removed, a significant effect is indicated -- with a regression coefficient almost twice as large as that of the entire research data base. Here, the 95 percent confidence interval for the population parameter runs from .01006 to .17458 (vs. .004179 to .09193 for the entire research data base). 6 .4.3.3 Residual Analysis Figures 6-7 through 6-9 present several figures used as part of the analysis of residual s. Figure 6-7 plots standardized residuals against the standardized predicted dependent variables. As the points are randomly and evenly distributed in no apparent pattern, 1 inearity and homoscedasticity are indicated. Figure 6-8 presents a histogram of standardized residuals superimposed over a normal curve, given the mean and variance of the residuals in the regression. of the residuals. Figure 6-9 is a cumulative probability plot 170 These figures indicate a near normal distribution of residuals; and more normal than the CONNCON analysis, 6 .4.3.4 Summary The analysis of this data base further confirmed the existence of certain variables in the noncompliance decision-making process for taxpayers with no contact from the Department of Treasury (positively: self-employed taxpayers, taxpayers indicating sales as an occupation, opportunity to evade, 1986 nonfilers, negatively: labor, unskilled and gender, taxpayers and level of income; living Rapids/Holland and metropolitan Detroit SMSA’s). in the Grand It also increased the magnitude of several regression coefficients (gender, 1986 nonfilers, and taxpayers living in the Grand Rapids/Holland SMSA). 171 FIGURE 6-7 CONNCON2 DATA BASE: SCATTERPLOT OF STAMDAHHZED RESIDUALS AGAINST STANDARDIZED PREDICTED SCORES Across - *ZPRED + Out 3 2 1 Down - *ZRESID • - I J ■I * • it. •kit ■kit m • it • • .it -1 -2 - -■ -3 Out -t -3 Symbols: Max N 2.0 4.0 8.0 Out 172 FIGURE 6-8 00NMC0N2 DATA BASE: NExp N 0 1.07 1 .55 0 .79 1 1.12 1 1.56 6 2.15 6 2.92 5 3.89 3 5.12 3 6.62 * 8.43 * 10.6 * 13.1 * 15.9 * 19.0 * 22.4 * 25.9 * 29.6 * 33.3 * 36.8 * 40.1 * 43.1 * 45.5 * 47.3 * 48.4 * 48.8 * 48.4 * 47.3 * 45.5 * 43.1 * 40.1 * 36.8 * 33.3 * 29.6 * 25.9 • 22.4 * 19.0 * 15.9 * 13.1 6 10.6 8 8.43 5 6.62 * 5.12 2 3.89 3 2.92 3 2.15 3 1.56 2 1.12 2 .79 0 .55 1 1.07 (* = 1 Cases, Out 3.00 2.88 2.75 2.63 2.50 2.38 2.25 2.13 2.00 1.88 1.75 1.63 1.50 1.38 1.25 1.13 1.00 .88 .75 .63 .50 .38 .25 .13 0.0 -.13 -.25 -.38 -.50 -.63 -.75 -.88 -1.00 -1.13 -1.25 -1.38 -1.50 -1.63 -1.75 -1.88 -2.00 -2.13 -2.25 -2.38 -2.50 -2.63 -2.75 -2.88 -3.00 Out HISTOGRAM OF STANDARDIZED RESIDUALS . : = Normal Curve) ; m : *. ****** ****** ***** *** *** *********** **********■* ************.*** ************** ********************* *********************•* ***************** ************************** a ****************** ************************************•*********** ****************************** ******************************************s **************************************** ********************************************* ********************************************** ************************************************•******* ***********************************************-**************************** **********************************************.******** ********************************************a ******************************************** ***************************************•****** ***************************************** ********************************•** ************************** ********************* ******************* ************* ************* ********** ****** ******** ***** ********** ** **. *•* *.* :* •* : 173 FIGURE 6-9 CGNHC0N2 DATA BASE: NORMAL PROBABILITY PLOT OF STAWARDIZED RESIDUALS Standardized Residual 1 75 ■ 25 -- ** ** Expected E x p e c t e d 174 6.5 Regression Analyses of the Stratified AGI Data Bases 6.5.1 In General Previous researchers have indicated that stratification of data bases in different ways (e.g., by level of income) may provide additional useful information (see, for example, Witte and Woodbury [1985]). With this in mind, the research data base was stratified by level of adjusted gross income. However, the recommendation of previous research was only one of the reasons for the stratification. The other, more practical, reason was to attempt to mitigate the effect of level of income variable which dominated the regression analysis of the entire research data base. The following summarizes the five AGI stratas analyzed in this study: _ _ _ _ _ AGI Less than $ 7,500 15.000 25.000 50.000 N_ $ 7,500 464 - 14,999 467 - 24,999 543 - 49,999 695 or more 301 TOTALS 2.470 % 18.8 $ 18.9 22.0 28.1 12.2 Amnesty Tax 49,316.15 95,740.59 152,257.47 278,403.86 433.310.59 % Average Amnesty Tax Average AGI 4.9 $ 106.28 $ 4,568.04 9.5 205.01 11,105.55 15.1 280.40 19,753.05 27.6 400.58 35,296.83 42.9 1,439.57 331,500.89 100.0 $1.009.028.66 100.0 $ 408.51 $ 57.629.49 The analysis of the five AGI stratas proceeds as follows: 6.5.2 Analysis of Taxpayers With AGI Less Than $7,500 6.5.3 Analysis of Taxpayers With AGI From $7,500 to $14,999 6.5.4 Analysis of Taxpayers With AGI From $15,000 to $24,999 6.5.5 Analysis of Taxpayers With AGI From $25,000 to $49,999 6.5.6 Analysis of Taxpayers With AGI More Than $50,000 6.5.7 Comparison of the Strata Results 175 Table 6 -8 presents a summary of the five AGI regression analyses, identifying significant variables across the five stratas and providing the regression coefficients. This summary does not list the variables in the order they entered the various regression equations utilizing the stepwise method; it is only meant to provide an overview of the results for comparison purposes. As expected, the regression results vary among the AGI stratas with several new variables a part of the regressions (e.g., the URBAN variable and the SMSA9 variable [the Benton Harbor SMSA] are significant in two of the regressions, the professional occupational category is significant in the largest AGI strata. each of the stratas follows. A further analysis of TABLE 6-8 A SUWURT OF SIGNIFICANT VARIABLES IN TIE STRATIFIED AGI REGRESSIONS Dependent Variable.. NONCOMP LOG OF UNREPORTED INCOME Variables in the Equation AGI1.SYS (AGI LT S7500) Variable CONTACT AGI2.SYS (AGI $7501-$14999) B -.32869 Variable CONTACT B -.83076 AGI3.SYS (AGI $15000-S249991 Variable CONTACT B -1.20054 AGI5.SYS (AGI $50000*1 AGI4.SYS (AGI £25000-S50000) Variable CONTACT OPPORT -1.46137 .92541 Variable B CONTACT -2.35573 OCT1 OCT3 .70379 OCT4 OCT6 OCT 7 AGILN EXEMPT GENDER RATE533 .39623 .84747 OCT3 .70344 0CT6 1.04280 OCT3 .55846 AGILN 1.04657 .32602 1.13508 -.27490 .27088 SELFEMPL AGILN EXEMPT .48258 .87965 -.11332 SELFEMPL .58814 -2.48008 N0FILE86 .27120 PDPREP .36589 SMSA3 1.00124 -.49282 8.89459 (Constant) .47454 .22519 .21651 1.35489 Multiple R R Square Adj R Square Std Error SMSA9 (Constant) Multiple R R Square Adj R Square Std Error -.74837 -2.15303 .56322 .31722 .30674 .88353 (Constant) -.02895 URBAN (Constant) Multiple R R Square Adj R Square Std Error .42677 .18213 .17146 1.20504 Multiple R R Square Adj R Square Std Error 1.28879 -2.10402 .49381 .24384 .23614 1.26407 AGILN EXEMPT .73288 .27781 N0FILE86 .86327 RATE635 PDPREP SMSA1 SMSA2 SMSA9 URBAN (Constant) Multiple R R Square Adj R Square Std Error -.53893 .49987 -.53720 .84350 -2.06258 1.24164 .69604 .64297 .41341 .39108 1.52415 177 6.5.2 Analysis of Taxpayers With AGI of Less Than $7.500 6.5.2.1 In General Table 5-9 presents the regression results for the data base created for taxpayers with AGI less than $7,500 (AGI1). regression occurs along several lines, The analysis of the including a discussion of the goodness of fit (R2), an analysis of the regression coefficients and their confidence intervals, and a comparison of the results with other analyses. 6 .5.2.2 Goodness of Fit The model has a R 2 of .3172, and an adjusted R 2 of .3067 with seven variables in the equation. As a result, the independent variables in the model account for about 32 percent of the variance in the dependent variable (unreported income). The order of the variables into the equation, and their impact on R 2 is detailed in Table 6-9. As it was in the regression of the entire data base, the log of AGI (AGILN) is again powerful, with an R 2 of .2049 (about 65 percent of the variance explained by all seven independent variables). equation (along with their R 2 Other variables in the regression contributions) include the number of exemptions claimed on the return (.0417), prior contact with the Treasury Department (.0245), retired/student the occupational sales group occupational (.0118), the single males (.0120), and tax year 1985 (.0117). significant at the .1 0 level. group (.0108), variable the signifying No other variables are 178 TABLE 6-9 REGRESSION TO PREDICT NONCOMPLIANCE USING AGI1 DATA BASE Variable Name Coefficient (Standard Error) t (Significance) R2 Adjusted AGILN 1.13508 (.08665) 13.099 (.0 0 0 0 ) .2049 .2032 EXEMPT -.27490 (.05565) -4.940 (.0 0 0 0 ) .4966 .2466 CONTACT -.32869 (.08432) -3.898 (.0 0 0 1 ) .5206 .2711 0CT3 .70379 (.22191) 3.172 (.0016) .5309 .2818 OCT7 .32602 (.09660) 3.375 (.0008) .5419 .2936 GENDER .27088 (.09043) 2.995 (.0029) .5528 .3056 RATE533 -2.48008 (.88889) -2.790 (.0055) .5632 .3172 (Constant) -2.15303 (.71765) -3.000 (.0028) 7 179 6 .5.2.3 Analysis of Regression Coefficients and Confidence Intervals Table 6-10 presents independent variables the regression coefficients for each of the in the regression equation, along with their standard error and computed 95 percent confidence intervals. Log of AGI (AGILN). The estimated regression adjusted gross income is 1.13508. coefficient of Thus, within this strata, amnesty participants with twice as much adjusted gross income relative to other amnesty participants had (after adjustment) about 119.6 percent more unreported income in A G I , ceteris paribus. is about twice as 1 This regression coefficient arge as the coefficient for the entire research data base. Exemptions (EXEMPT). In this analysis, the exemption variable has a negative regression coefficient, probably exemptions in the computation of tax paid. for this variable is indicating the use of The regression coefficient -.27490, which signifies that amnesty participants with twice as many exemptions relative to other amnesty participants disclosed about 27.490 percent less unreported income, ceteris paribus. Of course, the exemption amount (which reduces AGI in computing taxable income) could explain this Contact (CONTACT). contact variable is smaller amount of unreported income. The estimated regression coefficient of the -.32869, with a 95 percent confidence interval ranging from -.16299 to -.49439. This coefficient is about three times smaller than the coefficient for the entire research data base (- .93359), indicating a smaller level of distinction between taxpayers with and without contact (i.e., those taxpayers without contact indicated did not have as great a difference in unreported income as those taxpayers 180 TABLE 6-10 CONFIDENCE INTERVALS OF REGRESSION COEFFICIENTS FOR AGI1 DATA BASE Variable Name Coefficient Standard Error 95 Percent Confidence Interval AGILN 1.13508 .08665 .96479 1.30537 EXEMPT -.27490 .05565 -.38426 -.16554 CONTACT -.32869 .08432 -.49439 -.16299 .70379 .22191 .26770 1.13988 .32602 .09660 .13618 .51585 .27088 .09043 .09316 .44860 RATE533 -2.48008 .88889 -4.22691 -.73325 (Constant) -2.15303 .71765 -3.56334 -.74271 0CT3 0CT7 * GENDER 181 with contact). It is noteworthy that a smaller proportion of taxpayers in this range had contact with the Treasury Department when compared to the entire research data base (i.e., although there were more no contact taxpayers in this group, the difference between the groups is smaller than the entire research data base). Occupations (0CT3, 0CT7). Two different occupation variables are part of the regression equation -- sales (0CT3), and retired/student (0CT5). The impact of each of these variables as part of the regression equation is briefly discussed below. Sales. For taxpayers who listed sales as their occupation, the estimated regression coefficient in this study is .70379 (as compared to .58912 for the entire research data base), other independent variables, indicating that when controlling for these taxpayers disclosed about 70.379 percent more unreported income than taxpayers in the occupations omitted from the regression equation. The 95 percent confidence interval for the population parameter runs from .26770 to 1.13988. Retired/Student. Although not significant in the regression analysis of the entire research data base, this occupational category is significant in this stratified AGI analysis. The inclusion of this variable in the regression equation is not surprising based on the income level being examined. The regression coefficient is .32602, which indicates that these taxpayers disclosed about 32.602 percent more unreported income than taxpayers in the occupations omitted from the regression equation. The 95 percent confidence interval for the population parameter runs from .13618 to .51585. It is important to stress,once again, that this 182 finding does not imply advertent tax evaders. that retired taxpayers and/or students are There may be other reasons why they failed to comply (e.g., lack of knowledge, confusion, etc.). Gender. disclosed The regression coefficient indicates that single males about taxpayers, 27.088 ceteris percent paribus. more This unreported coefficient income is coefficient for the entire research data base (.19001). larger than than other the The increase is significant, because females made up a larger proportion of this AGI strata than in the entire research data base. In addition, there was more information available regarding gender in this AGI strata (about 89 percent of the returns in the strata were filed by single individuals). The 95 percent confidence interval for the regression coefficient ranges from .09316 to .44860. As a result, the population parameter would indicate single males could disclose from 9.3 to 44.9 percent more unreported income than other taxpayers. Tax Year 1985 (RATE533). The regression coefficient indicates that taxpayers who filed a return during 1985 revealed about 248 percent less unreported income than taxpayers who filed a return during a year omitted from the regression equation. Based on a review of the data in the strata, there was only 1 return (out of 464 in the strata) filed in 1985. As a result, it would seem that the inclusion of this variable in the model should be viewed skeptically. In addition, the standard error of the regression coefficient is very large (.88889). 6 .5.2.4 Summary Several variables and their signs that are part of the regression equation are not surprising -- contact with the Treasury Department 183 (negatively related), (positively related). and the retired/student occupational group More surprising was the strength of the level of income variable (which the stratification sought to weaken), indicating a positive relationship between level of income and unreported income. The inclusion of the single male variable was also not surprising, given the descriptive statisties of the retired/student occupational group (lower income levels, and greater proportion of single male taxpayers than the research data base). 6.5.3 Analysis of Taxpayers With AGI From $7.500 to $14.999 6.5.3.1 In General Table 6-11 presents the regression results for the data base created for taxpayers with AGI of $7,500 or more but less than $15,000 (AGI2). The analysis of the regression includes a discussion of the goodness of fit (R2 ), an analysis of the regression coefficients and their confidence intervals, and a comparison of the results with the results from other analyses. 6 .5.3.2 Goodness of Fit The model has a R 2 of .1821, and an adjusted R 2 of .1715 with six variables in the equation. As a result, the independent variables in the model account for only about 18 percent of the variance in the dependent variable (unreported income). The order of the variables into the equation, and their impact on R 2 is detailed in Table 6-11. The first variable to enter the regression equation is contact, with an R 2 of .1076. Other variables in the regression equation (along with their R2 contributions) include the self-employed occupational group (..0345), the level of income (.114), the skilled labor occupational group (. 1 1 0 ), the 184 TABLE 6-11 REGRESSION TO PREDICT NONCONPLIANCE USING AGI2 DATA BASE Variable Name Coefficient (Standard Error) t (Significance) CONTACT -.83076 (.12534 ) -6.628 (.0 0 0 0 ) .1076 .1057 0CT6 .84747 (.20835) 4.068 (.0 0 0 1 ) .1421 .1384 AGILN .87965 (.29079) 3.025 (.0026) .1535 .1480 0CT4 .39523 (.16935) 2.340 (.0197) .1645 .1573 EXEMPT -.11332 (.04790) -2.366 (.0184) .1743 .1653 SELFEMPL .48258 (.22968) (.0362) -.02895 (2.67789) (.9914) (Constant) 2.1 0 1 -.011 r2 Adjusted .1821 .1715 ? 185 number of exemptions claimed on the return (.0098), and the non- occupational self-employment variable (.0078). The sales occupational category (not significant at the .05 level) is significant, and positively related to the dependent variable at the .10 level (.0891). 6 .5.3.3 Analysis of Regression Coefficients and Confidence Intervals Table 6-12 presents the regression coefficients for each of the independent variables in the regression equation, along with their standard error and computed 95 percent confidence intervals. Contact (CONTACT). The estimated regression coefficient of the contact variable is -.83076, with a 95 percent confidence ranging This coefficient is only siightly from -.58949 to -1.107707. interval smaller than the coefficient for the entire research data base .93359), but is almost three times as large as coefficient. Apparently, (- the AGI1 regression as income increases, so does the reporting difference between the contact and no contact groups. Occupations (0CT6, 0CT5). part of Two different occupation variables are the regression equation -- self-employed (0CT6), and labor (OCT4). skilled The impact of each of these variables as part of the regression equation is briefly discussed below. Self-Emploved. The estimated regression coefficient for self-employed taxpayers is .84747, with a 95 percent confidence interval ranging from .43804 to 1.25690. The regression coefficient for the entire research data base was .53310. These results can be interpreted as indicating that taxpayers who listed their occupation as self-employed disclosed 186 TABLE 6-12 CONFIDENCE INTERVALS OF REGRESSION COEFFICIENTS FOR AGI2 DATA BASE Variable Name CONTACT Coefficient Standard Error 95 Percent Confidence Interval -.83076 .12534 -1.07707 -.58445 0CT6 .84747 .20835 .43804 1.25690 AGILN .87965 .29079 .30821 1.45110 0CT4 .39623 .16935 .06344 .72902 -.11332 .04790 -.20745 -.01919 .48258 .22968 .03123 .93394 -.02895 2.67789 -5.29136 5.23346 EXEMPT SELFEMPL (Constant) 187 84.747 percent more unreported income than taxpayers in the occupations omitted from the regression equation, ceteris paribus. The 95 percent confidence interval indicates that the population parameter for taxpayers disclosing their occupation as self-employed likely falls in a range from 43.8 to 125.7 percent more unreported income than those taxpayers in the occupations omitted from the regression equation. Skilled Labor. For taxpayers who listed their occupation as one grouped in the skilled labor category for this study, the estimated regression coefficient in this study is .39523, indicating that when controlling for other independent variables, these taxpayers disclosed about 39.623 percent more unreported income than taxpayers in the occupations omitted from the regression equation. The 95 percent confidence interval for the coefficient in measuring the population parameter runs from .06344 to .72902. This indicates that (with 95 percent confidence) the population parameter would disclose that these taxpayers reported from 6.3 to 72.9 percent more unreported income than taxpayers in the occupations omitted from the regression equation. This variable had not been included in any of the previous regression equations. Log of AGI (AGILN). The estimated regression adjusted gross income is .87965. coefficient of Thus, within this strata, amnesty participants with twice as much adjusted gross income relative to other amnesty participants had (after adjustment) about 84.0 percent more unreported income in AGI, ceteris paribus. This regression coefficient is about 50 percent larger than the coefficient for the entire research data base, but smaller than the coefficient in the AGI1 regression. 188 Exemptions (EXEMPT). As in the analysis of AGI1, the exemption variable has a negative regression coefficient, probably indicating the use of exemptions in the computation of tax paid. The regression coefficient for this variable is -.11332, which signifies that amnesty participants with twice as many exemptions relative to other amnesty participants disclosed ceteris paribus. about 11.332 percent less unreported income, Of course, the exemption amount (which reduces AGI in computing taxable income) could explain this smaller amount of unreported income. Self-Emploved Taxpayers (SELFEMPL1. The self-employed variable created as an alternative to the occupational coding has a regression coefficient of .48258, indicating that these amnesty participants revealed about 48.258 percent more unreported income than participants who were not determined to be self-employed under the parameters indicated, ceteris paribus. This coefficient is about 60 percent larger than the coefficient in the regression of the entire research data base (.29934). Although significant, and of further support to the contention that self-employed taxpayers have a greater propensity to evade taxes, this variable is not as significant as the occupational variable (as was the case in the regression for the entire research data base). However, the significance of the occupational variable is based on a comparison to taxpayers in the professional support category. Here, the comparison group is taxpayers who are not self-employed. 6 .5.3.4 Summary Contact with the Treasury Department became the dominant variable in the analysis of this AGI strata, and the regression coefficient increased 189 dramatically over the previous AGI strata, indicating difference between taxpayers with and without contact. more of a The regression indicated a new variable not a part of previous analyses -- the skilled labor occupational group (positively related to noncompliance). Self- employed taxpayers (both the occupational and non-occupational variable) were positively related to noncompliance in this AGI strata, and the sales occupational group (although not significant at the .05 level) was significant and positively related to noncompliance in this strata at the .10 level (.0891). 6.5.4 Analysis of Taxpayers With AGI From $15.000 to $24.999 6.5.4.1 In General Table 6-13 presents the regression results for the data base created for taxpayers with AGI of $15,000 or more, but less than $24,999 (AGI3). The analysis of the regression occurs along several lines, including a discussion of the goodness of fit (R2), an analysis of the regression coefficients and their confidence intervals, and a comparison of the results with other analyses. 6 .5.4.2 Goodness of Fit The model has a R 2 of .2252, and an adjusted R 2 of .2165 with six variables in the equation. As a result, the independent variables in the model account for about 23 percent of the variance in the dependent variable (unreported income). The order of the variables into the equation, and their impact on R 2 is detailed in Table 6-13. As it was in the regression of the second AGI strata (AGI2), the first variable to enter the regression equation is contact, with an R 2 of .1648. This is a powerful variable in this regression equation, accounting for about 73 190 TABLE 6-13 REGRESSION TO PREDICT NONCOMPLIANCE USING AGI3 DATA BASE Variable Name CONTACT Coefficient (Standard Error) t (Significance) R2 Adjusted R2 -1.20054 (.17706) -6.780 (.0 0 0 0 ) .1648 .1633 0CT6 1.04280 (.27532) 3.788 (.0 0 0 2 ) .1840 .1810 0CT3 .70344 (.20568) 3.420 (.0007) .1982 .1938 URBAN -.49282 (.15228) -3.236 (.0013) .2117 .2058 SMSA3 1.00124 (.43424) 2.306 (.0215) .2191 .2118 SELFEMPL .58814 (.28549) 2.060 (.0399) .2252 .2165 (Constant) 8.89459 (.20277) 43.865 (.0 0 0 0 ) 191 percent of variance in the dependent variable accounted for by all six independent variables. with their Other variables in the regression equation (along contributions) include the self-employed occupational group (.0192), the sales occupational group (.0142), taxpayers living in a SMSA (.0134), taxpayers living in the Jackson SMSA (.0074), and the nonoccupational self-employment variable (.0061). The unskilled labor occupational category (not significant at the .05 level) is significant, and negatively correlated to the dependent variable at the .10 level (.0866), as it was in the analysis of the entire research data base. 6 .5.4.3 Analysis of Regression Coefficients and Confidence Intervals Table 6-14 presents the regression coefficients for each of the independent variables in the regression equation, along with their standard error and computed 95 percent confidence intervals. Contact (CONTACT1. The estimated regression coefficient of the contact variable is -1.20054, with a 95 percent confidence interval ranging from -.85272 to -1.54836. This coefficient is significantly larger than the coefficient for the entire research data base ( -.93359), and continues to increase across AGI stratas. Occupations (0CT6. 0CT31. Two different occupation variables are part of the regression equation (0CT3). -- self-employed (0CT6), and sales The impact of each of these variables as part of the regression equation is briefly discussed below. Self-Employed. The estimated regression coefficient for self-employed taxpayers is 1.04280, with a 95 percent confidence interval ranging from 192 TABLE 6-14 CONFIDENCE INTERVALS OF REGRESSION COEFFICIENTS FOR AGI3 DATA BASE Standard Error Variable Name Coefficient CONTACT -1.20054 .17706 -1.54836 -.85272 0CT6 1.04280 .27532 .50197 1.58363 0CT3 .70344 .20568 .29941 1.10748 URBAN -.49282 .15228 -.79195 -.19368 SMSA3 1.00124 .43424 .14823 1.85426 .58814 .28549 .02733 1.14894 8.89459 .20277 8.49626 9.29292 SELFEMPL (Constant) 95 Percent Confidence Interval 193 .50197 to 1.58353. data base was The regression coefficient for the entire research .53310. Therefore, taxpayers in the self-employed occupational group revealed about 50 percent more unreported income than those in the entire research data base. Within this AGI strata, taxpayers who listed their occupation as self-employed disclosed 104.280 percent more unreported income than taxpayers in the occupations omitted from the regression equation, ceteris paribus. In addition, the regression coefficient in this strata is about 23 percent larger than the coefficient in the AGI2 strata (.84747). As income increases, therefore, the size of underreporting increases within this occupational group. Sales. For taxpayers who listed sales as their occupation, the estimated regression coefficient in this AGI strata is .70344 (as compared to .58912 for the entire research data base, and .70379 in the AGI1 strata), indicating that when controlling for other independent variables, these taxpayers disclosed about 70.344 percent more unreported income than taxpayers in the occupations omitted from the regression equation. The 95 percent confidence interval for the population parameter runs from .29941 to 1.10748. Taxpayers Living in a 5M5A (URBAN). This variable was created as a composite of all the SMSA variables, and is meant to test for a difference between urban and rural taxpayers. Although not significant in any other regression equations, this variable is significant here, with a regression coefficient of -.49282, indicating that taxpayers in this AGI strata living in a SMSA disclosed about 49.282 percent less unreported income than taxpayers living in a rural area. The 95 percent 194 confidence interval for estimating the population parameter ranges from.19368 to -.79195. Jackson and Jackson County (SMSA31. Based on the regression coefficient, amnesty participants in this AGI strata living in the city of Jackson or Jackson county disclosed about 100.124 percent more unreported income than amnesty participants in this strata other Standard Metropolitan Statistical Areas. 1 iving in Because of the large standard error, the 95 percent confidence interval for the population parameter ranges from about 14.8 percent to 185.4 percent. This, too, is a new variable in a regression equation. When viewed in 1 ight of the previous variable, it becomes evident that the rural area had some very significant noncompliers in this AGI strata. «r9' Self-Employed Taxpayers (SELFEMPL). As in the prior AGI strata, the self-employed variable created as an alternative to the occupational coding is a part of this regression equation. regression coefficient of .58814, indicating The variable has a that these amnesty participants revealed about 58.814 percent more unreported income than participants who were not determined to be self-employed under the parameters indicated, ceteris paribus. This coefficient is about 60 percent larger than the coefficient in the regression of the entire research data base (.29934), and 20 percent larger than the coefficient in the regression equation of AGI2 (.48258). 6 .5.4.4 Summary Contact with the Treasury Department again was the dominant variable in the analysis of this AGI strata. The regression coefficient has increased in size over each of the previous AGI stratas, indicating increasing differences between taxpayers with and without contact. In 195 addition, it accounted for the majority of the variance in dependent variable among the independent variables in the regression equation (about 73 percent). Self-employed taxpayers (both the occupational and non-occupational variable) and the sales occupational group were also positively related to noncompliance in this AGI strata. Two geographically related variables are a part of the regression equation, and indicate that taxpayers living outside a SMSA or living inside the Jackson SMSA are more 1ikely to be noncompliant. absence is the level of income variable. This is the only AGI strata where it was not a part of the regression equation. occupational group Conspicuous by its The unskilled labor (although not significant at the .05 level) was significant and negatively related to noncompliance in this strata at the .10 level (.0855). 6.5.5 Analysis of Taxpayers With AGI From $25.000 to $49.999 6.5.5.1 In General Table 6-15 presents the regression results for the data base created for taxpayers with AGI of $25,000 or more but less than $50,000 (AGI4). The analysis of the regression includes a discussion of the goodness of fit (R2 ), an analysis of the regression coefficients and their confidence intervals, and a comparison of the results with the results from other analyses. 6.5.5.2 Goodness of Fit The model has a R 2 of .2438, and an adjusted R 2 of .2361 with seven variables in the equation. As a result, the independent variables in the model account for about 24 percent of the variance in the dependent 196 TABLE 6-15 REGRESSION TO PREDICT NONCONPLIANCE USING AGI4 DATA BASE Variable Name Coefficient (Standard Error) t (Significance) r2 Adjusted CONTACT -1.46137 (.15955) -9.159 (.0 0 0 0 ) .1569 .1557 OPPORT .92541 (.24897) 3.717 (.0 0 0 2 ) .1809 .1785 AGILN 1.04657 (.25388) 4.122 (.0 0 0 0 ) .1992 .1958 OCT3 .55846 (.16241) 3.439 (.0006) .2 1 1 0 PDPREP .36589 (.10044) 3.643 (.0003) .2301 .2245 N0FILE86 .27120 (.11036) 2.457 (.0142) .2373 .2307 SMSA9 1.28879 (.52959) 2.434 (.0152) .2438 .2361 -2.10402 (2.65814) -.792 (.4289) (Constant) .2156 ? 197 variable (unreported income). The order of the variables into the equation, and their impact on R 2 is detailed in Table 6-15. As in the two previous regression equations (related to AGI2 and AGI3), the first variable to enter the regression equation is contact, with an R 2 of .1569. Once again, this is a powerful variable in this regression equation, accounting for about 64 percent of variance in the dependent variable accounted for by all seven independent variables. Other variables in the regression equation (along with their R2 contributions) include opportunity (.0240), level of income (.0184), the sales occupational group (.0164), prepare their returns (.0145), taxpayers using paid preparers to 1986 nonfilers (.0073), and taxpayers 1iving in the Benton Harbor SMSA (.0065). The unskilled labor occupational category (not significant at the .05 level) is significant, variable at the .10 level and negatively related to the dependent (.0673), as it was in the analysis of the entire research data base. 6.5.3.3 Analysis of Regression Coefficients and Confidence Intervals Table 6-16 presents the regression coefficients for each of the independent variables in the regression equation, along with their standard error and computed 95 percent confidence intervals. Contact (CONTACT). The estimated regression coefficient of the contact variable is-1.46137, with a ranging from -1.14810 to -1.77464. 95 percent confidence interval This coefficient is significantly larger than the coefficient for the entire research data base ( -.93359), and even larger than the coefficient related to the AGI3 strata (- 1.20054), continuing its pattern of increasing as income level increases. 198 TABLE 6-16 CONFIDENCE INTERVALS OF REGRESSION COEFFICIENTS FOR AGI4 DATA BASE Standard Error Variable Name Coefficient CONTACT -1.46137 .15955 -1.77464 -1.14810 OPPORT .92541 .24897 .43659 1.41424 AGILN 1.04657 .25388 .54810 1.54504 0CT3 .55846 .16241 .23959 .87733 PDPREP .36589 .10044 .16868 .56309 N0FILE86 .27120 .11036 .05451 .48789 1.28879 .52959 .24897 2.32861 -2.10402 2.65814 -7.32307 3.11503 SMSA9 (Constant) 95 Percent Confidence Interval 199 Opportunity to Evade (OPPORT). Recall that this variable is created via a composite of occupation (self-employed, business, professional, or sales), an income level of $30,000 or greater, and access to cash income sources. This is the first AGI strata where it is a relevant variable because of the AGI requirement. The estimated regression coefficient of the variable is .92541 (compared to .84392 for the entire research data base). The 95 percent confidence interval runs from .43659 to 1.41424. is large due to the size of the standard The confidence interval error of the regression coefficient, indicating a great deal of variability in the coefficient. The regression coefficient indicates that amnesty participants with the opportunity to evade taxes on average underreported 92.541 percent more income than amnesty participants characteristics, ceteris paribus. without the opportunity This result is not surprising given the descriptive statistics related to this group of taxpayers. Although small in number, their mean amnesty tax payment was more than four times as large as the mean of the research data base. With 95 percent confidence, these results indicate that the population parameter shows that taxpayers with opportunity underreport income by at least 43.7 percent and possibly as much as 141.4 percent when compared to those amnesty participants without the opportunity characteristics. Log of AGI (AGILN). The estimated regression adjusted gross income is 1.04657. coefficient of Thus, within this strata, amnesty participants with twice as much adjusted gross income relative to other amnesty participants had (after adjustment) about 106.6 percent more unreported income in AGI, ceteris paribus. This regression coefficient 200 is about twice as large as the coefficient for the entire research data base. Sales Occupational Category (0CT31. The estimated regression coefficient for self-employed taxpayers is .55846, with a 95 percent confidence interval ranging from .23959 to .87773. This coefficient is smaller than the coefficients in AGI1 (.70379) and AGI3 (.70344), but is similar to that for the entire research data base (.58912), although the standard error is much larger. Within this AGI strata, taxpayers in the sales occupational group disclosed about 55.846 percent more unreported income than taxpayers in the occupations omitted from the regression equation. Paid Preparers Used to Prepare Returns (PDPREP). coefficient for this variable is .36589 The regression (as compared to .16548 for the entire research data base) indicating that amnesty participants who used paid tax-return preparers disclosed about 36.6 percent more unreported income than amnesty participants who prepared their own return. The 95 percent confidence interval for this coefficient ranges from .16868 to .56309. Once again, it is not possible to make a conclusion regarding the implications of return preparers assisting in the noncompliance decision; the only conclusion that can be made is that these returns disclosed more previously unreported income. variable has regressions. not been However, it is interesting that this a part of any of the previous AGI strata It therefore appears that this variable may be related to income level (i.e., as income level increases taxpayers seek out tax return preparers). 201 Non-Filers in 1986 (N0FILE86). study as a surrogate for noncompliance decision. an This variable was included in the attitudinal variable related to the The regression coefficient for this variable is .27120 (as compared to .17434 for the entire research data base). We can conclude that these taxpayers disclosed 27.120 percent more unreported income than those amnesty participants who filed a tax return in 1986. The 95 percent confidence interval for this coefficient ranges from .05451 to .48789. Bonton Harbor and Berrien County (SMSA9). Based on the regression coefficient, amnesty participants in this AGI strata 1iving in the city of Benton Harbor or Berrien county disclosed about 128.879 percent more unreported income elsewhere. than Because amnesty of the participants 1arge standard in this error, strata the 95 1 iving percent confidence interval for the population parameter ranges from about 24.9 percent to 232.9 percent. This is another new geographic variable in a regression equation. 6 .5.5.4 Summary Contact with the Treasury Department again is a dominant variable in the analysis of this AGI strata. The regression coefficient increased in size over the previous AGI stratas, indicating increasing differences between taxpayers with and without contact. Although not as powerful as in the previous AGI strata, it still accounted for the majority of the variance in dependent variable among the independent variables in the regression equation (here, about 64 percent). This AGI strata was the first where opportunity to evade taxes could be evaluated, and it was the second variable to enter the regression equation. The level of income variable was significant again in this AGI strata. Taxpayers who failed 202 to file a return in 1986 also was a part of the regression equation for this strata, and positively related to noncompliance. It had not appeared in any previous AGI strata regression equations. The sales occupational group were also positively related to noncompliance in this AGI strata. The Benton Harbor SMSA was found to be positively related to noncompliance, as was the variable related to paid preparation of the tax return. Finally, as in the AGI3 strata, the unskilled labor occupational category (not significant at the .05 level) is significant, and negatively related to the dependent variable at the .10 level (.0673). 6.5.6 Analysis of Taxpayers With AGI of $50.000 or More 6.5.6.1 In General Table 6-17 presents the regression results for the data base created for taxpayers with AGI of $50,000 or more (AGI5). The analysis of the regression includes a discussion of the goodness of fit (R2 ), an analysis of the regression coefficients and their confidence intervals, and a comparison of the results with the results From other analyses. 6 .5.6 .2 Goodness of Fit The model has a R 2 of .4134, and an adjusted R 2 of .3911 with eleven variables in the equation. As a result, the independent variables in the model account for about 41 percent of the variance in the dependent variable (unreported income). This is the greatest amount of variance accounted for in any of the regression equations. The order of the variables into the equation, and their impact on R 2 is detailed in Table 6-17. As in the three previous regression equations (related to AGI2, AGI3, and AGI4), the first variable to enter the regression equation is 203 TABLE 6-17 REGRESSION TO PREDICT NONCOMPLIANCE USING AGI5 DATA BASE Variable Name CONTACT Coefficient (Standard Error) t (Significance) R2 Adjusted Rz9 -2.35573 (.29256) -8.052 (.0 0 0 0 ) .1305 .1276 AGILN .73288 (.10632) 6.893 (.0 0 0 0 ) .2356 .2304 EXEMPT .27781 (.06357) 4.370 (.0 0 0 0 ) .2744 .2671 N0FILE86 .86327 (.21504) 4.015 (.0 0 0 1 ) .3096 .3003 0CT1 -.74837 (.19685) -3.802 (.0 0 0 2 ) .3322 .3208 SMSA2 .84350 (.44704) 1.887 (.0602) .3486 .3353 RATE635 -.53893 (.18970) -2.841 (.0048) .3654 .3502 URBAN 1.24164 (.31401) 3.954 (.0 0 0 1 ) .3807 .3637 PDPREP .49987 (.20418) 2.448 (.0149) .3949 .3762 SMSA1 -.53720 (.22569) -2.380 (.0179) .4029 .3823 SMSA9 -2.06258 (.90786) -2.272 (.0238) .4134 .3911 (Constant) .69604 (1.22495) .568 (.5703) 204 contact, with an overpower the of .1305. other Although powerful, this variable does not variables in the regression equation. variables in the regression equation (along with their Other contributions) include level of income (.1050), number of exemptions claimed (.0389), 1986 nonfilers (.0352), the professional occupational group (,0225), the Ann Arbor SMSA (.0164), tax year 1983 (.0168), taxpayers living in a SMSA (.0153), taxpayers using paid preparers to prepare their returns (.0143), taxpayers living in the metropolitan Detroit SMSA (.0080), and taxpayers 1iving in the Benton Harbor SMSA (.0105). Opportunity to avoid taxes, although not significant at the .05 level is significant and positively correlated to the dependent variable at the .10 level (.0613), occupational group (.0857). as is gender (.0533), and the sales The retired/student occupational group is also significant at the .10 level (.0545), but negatively correlated with the dependent variable. 6 .5.6 .3 Analysis of Regression Coefficients and Confidence Intervals Table 6-18 presents the regression coefficients for each of the independent variables in the regression equation, along with their standard error and computed 95 percent confidence intervals. Contact (CONTACT). The estimated regression coefficient of the contact variable is -2.35573, with a 95 percent confidence interval ranging from -1.77991 to -2.93154. The size of the confidence interval is due to the size of the standard error of the regression coefficient. This coefficient has become 1arger with each AGI strata, and the pattern continues in this AGI strata. In addition, it is significantly larger than the coefficient for the entire research data base ( -.93359). 205 TABLE 6-18 CONFIDENCE INTERVALS OF REGRESSION COEFFICIENTS FOR AGI5 DATA BASE Variable Name Coefficient CONTACT -2.35573 .29256 -2.93154 -1.77991 AGILN .73288 .10632 .52361 .94215 EXEMPT .27781 .06357 .15269 .40293 N0FILE86 .86327 .21504 .44004 1.28651 OCT1 -.74837 .19685 -1.13581 -.36094 SMSA2 .84350 .44704 -.03637 1.72337 RATE635 -.53893 .18970 -.91230 -.16557 URBAN 1.24164 .31401 .62361 1.85968 PDPREP .49987 .20418 .09801 .90173 SMSA1 -.53720 .22569 -.98141 -.09300 SMSA9 -2.06258 .90786 -3.84943 -.27572 .69604 1.22495 -1.71492 3.10699 (Constant) Standard Error 95 Percent Confidence Interval 206 Log of AGI (AGILN). The estimated regression adjusted gross income is .73288. coefficient of Thus, within this strata, amnesty participants with twice as much adjusted gross income relative to other amnesty participants had (after adjustment) about unreported income in AGI, ceteris paribus. This 6 6 .2 percent more regression coefficient is about 25 percent larger than the coefficient for the entire research data base, and is the smallest level of income coefficient among the AGI stratas. Exemptions (EXEMPT). Although the regression coefficient for this variable was negative in the lowest two AGI stratas, it is positive in this AGI strata. The regression coefficient is .27781 (compared to .04797 in the regression of the research data base). This indicates that amnesty participants within this strata with twice as many exemptions relative to other strata members disclosed about 27.781 percent more unreported income, ceteris paribus. The 95 percent confidence interval for the population parameter runs from .15269 to .40293. Non-Filers in 1986 (N0FILE86). study as a surrogate noncompliance decision. for an This variable was included in the attitudinal variable related to the The regression coefficient for this variable is .86327 (a significantly higher coefficient when compared to a .27120 coefficient in the AGI4 strata, and a .17434 coefficient for the entire research data base). It is interesting to note that the level of unreported income increases as income level increases. We can conclude that these taxpayers disclosed 86.327 percent more unreported income than those amnesty participants who filed a tax return in 1986. The 95 percent confidence interval for this coefficient ranges from .44004 to 1.28651. 207 Professional Occupational Category (0CT1). the professional regression occupation equation. Even category more This is the first time variable surprising regression coefficient is negative (-.74837). interval ranges from -.36094 to -1.13581. has is the appeared fact in that a the The 95 percent confidence Previous research has found that taxpayers in this occupational category (see, for example, Westat [1980b] and Witte and Woodbury [1985]) are generally associated with lower levels professional of compliance. occupational In this group disclosed analysis, taxpayers in the about 74.837 percent less unreported income than taxpayers in the occupations omitted from the regression equation. Tax Year 1983 (RATE635). As mentioned before, during several years in the early 1980’s, the Michigan flat tax rate varied from 4.6 percent (the "normal rate" during most years) to 5.1 percent in 1982, 6.35 percent in 1983, 5.85 percent in 1984, and 5.33 percent in 1985. The regression coefficient indicates that amnesty participants who filed a 1983 tax return revealed about 53.893 percent less unreported income than amnesty participant who filed a return during a year when the tax rate was 4.6 percent (the base dummy variable -- all tax years prior to 1982), ceteris paribus. This compares to a regression coefficient of .13441 for the regression of the entire data base. The 95 percent confidence interval indicates a population parameter in the range of -.16557 to.91230. As was mentioned before, this finding could merely be a reflection of the economic conditions in the early 1980’s, rather than an indication that higher tax rates are associated with higher rates of compliance -- it could be that because the economy was good during 1982, 208 1983, and 1984, taxpayers were more apt to report and pay taxes on all their income. Taxpayers Living in a SMSA (URBAN!. This variable was created as a composite of all the SMSA variables, and is meant to test for a difference between urban and rural taxpayers. This variable was also significant in the analysis of the A6I3 strata (although no where else). Here, however, this variable has a regression coefficient of 1.24164 (as compared to -.49282 in the AGI3 analysis), indicating that taxpayers in this AG I strata 1iving in a SMSA disclosed about 124.164 percent more unreported income than taxpayers 1iving in a rural area. The 95 percent confidence interval for estimating the population parameter ranges from .62361 to 1.85968. Paid Preparers Used to Prepare Returns (PDPREP1. The regression coefficient for this variable is .49987 (as compared with .36589 in the analysis of the AGI4 strata and .16548 for the entire research data base) indicating that amnesty participants who used paid tax-return preparers disclosed about 50 percent more unreported participants who prepared their own return. income than amnesty The 95 percent confidence interval for this coefficient ranges from .09801 to .90173. Once again, it is not possible to make a conclusion regarding the implications of return preparers assisting in the noncompliance decision; the only conclusion that can be made is that these returns disclosed more previously unreported income. It is, however, interesting that this variable only appears in the regression equations for the highest two AGI stratas. 209 Geographic Regions (SMSA1, SMSA2. and SMSA9). Three SMSA variables are a part of this regression equation; one is positively related to noncompliance and two are negatively related to noncompliance. Metropolitan Detroit southeastern Michigan, fSMSAl). SMSA1 including all includes a of metropolitan large part Detroit. of The regression coefficient indicates that for this AG I strata, taxpayers 1iving in this area disclosed 53.720 percent less unreported income than taxpayers 1iving in regions omitted from the regression equation. This finding is similar in direction, but much larger than the regression coefficients of the no contact data bases, where the regression coefficients were -.21078 (CONNCON) and -.22488 (C0NNC0N2). Benton Harbor and Berrien County (SMSA9). Based on the regression coefficient, amnesty participants in this AG I strata 1iving in the city of Benton Harbor or Berrien county disclosed about 206.258 percent less unreported income than amnesty participants in this strata regions omitted from the regression equation. 1 iving in The direction of the coefficient is the opposite of what it was in the previous AG I strata (AGI4). In addition, because the coefficient has a very large standard error (.90786), the 95 percent confidence interval for the population parameter ranges from about -27.6 percent to -385.0 percent. Ann Arbor and Washtenaw County (SMSA2). Based on the regression coefficient, amnesty participants 1iving in Ann Arbor or Washtenaw county disclosed about 84.350 percent more unreported income than amnesty participants living in regions omitted from the regression equation. As 210 can be seen from the 95 percent confidence interval (which includes zero), this variable is not significant at the .05 level, although it was when it entered the equation. The 95 percent confidence interval for the population parameter ranges from about -3.64 percent to 172.34 percent. 6 .5.6.4 Summary Contact with the Treasury Department continued to be the most significant variable in the regression equation, and by increasing in magnitude again completed a pattern that was noted beginning with the second AG I strata -- as income level increases, so does the size of the regression coefficient. the regression Level of income was the second variable to enter equation. Surprisingly, the professional occupational category was a significant variable for the first time in any regression equation, and was negatively related to noncompliance (when compared to the base occupational group -- professional services). The variable related to taxpayers who failed to file a return in 1986 also was a part of the regression equation noncompliance. for this strata, positively related to It increased in magnitude over the previous AGI strata. Four geographic variables are part of the regression equation -- three SMSA variables [metropolitan Detroit and Benton Harbor (both negatively related to noncompliance), and Ann Arbor (positively related to noncompliance)] and the variable indicating taxpayers living in a SMSA (positively related to noncompliance). As in the previous AG I strata, the variable indicating the use of professional tax return preparers was part of the regression equation. Opportunity to avoid taxes, although not significant at the .05 level is significant and positively related to the dependent variable at the .10 level (.0613), as is gender (.0533), and the sales occupational 211 group (.0857). The retired/student occupational group is also significant at the .10 level (.0545), but negatively related with the dependent variable. 6.5.7 Several Comparison of the AGI Strata Results interesting comments can be made analyses of the AGI stratas. about the regression In addition, the regressions provide some insight into how the noncompliance decision may change across levels of income. Contact. Contact was the only variable to appear in all five AGI strata regressions -- always with a negative regression coefficient. In addition, the size of the regression coefficient consistently increased across the stratas. It could be concluded that at higher levels of i' come, certain taxpayers without contact had more control over their income, and therefore, a larger difference existed between taxpayers with and without contact. At lower levels of AGI, although significant, there was a much smal1 er difference between taxpayers with or without contact. This result indicates that taxpayers without contact have significant amounts of income not being disclosed, and highlights the differences between taxpayers with and without contact. It would seem that increasing the presence of an enforcement agency (e.g., via audits) at higher income levels would be beneficial. Opportunity. Opportunity to avoid taxes could be evaluated in only two AGI stratas due to the AGI definition that was part of its makeup. The variable, not surprisingly, was significant at the .05 level in the AGI4 strata, and significant at the .10 level in AGI5 strata. 212 Level of Income (AGILN). The level of income variable was significant in four of the five AGI stratas (only in the third strataAGI from $15,000 to $24,999 - was it not significant), indicating that as income level increased, the amount of unreported income stratification, however, did mitigate but the lowest income strata, where increased. The the effect of this variable in all it continued to be the dominant variable in the regression equation. Occupational Categories. the AGI strata Several interesting findings come out of regressions. First, the professional category was significant in only the highest AGI strata. occupational This was not surprising since this category contained the majority of occupational codings in the strata (about 58 percent). sign of the regression coefficient What was surprising was the -- negative -- indicating less noncompliance when compared to all other occupational categories. Second, the sales occupational group was significant and positively related to noncompliance at the .05 level in three stratas (AGI1, AGI3, and AGI4), and significant at the .10 level in the other two (AGI2 and AGI5). It appears from these results that noncompliance occupational group crosses all levels of income. taxpayers (both the occupational in this Third, self-employed and non-occupational variable) were positively related to noncompliance in two AGI stratas (AGI2 and AGI3). Although initially it might appear unusual that these variables were not significant in the higher AGI stratas, remember that these taxpayers have a great deal of control over the level of their income (i.e., they control both the receipt of income and related expenses). As a result, chances are they will attempt to smooth income across years, when 213 possible, and be more likely to claim deductions aggressively. As a result, net income (and AGI) will be lower. Fourth, the skilled labor category, not previously significant in any regression equation, was significant noncompliance in the second AGI strata. and positively related to Finally, as expected, the retired/student occupational group was significant and positively related to noncompl iance in the lowest AGI strata. In addition, this group was al so significant at the .10 level, in the highest AGI strata, but negatively related to noncompliance. Gender. At the .05 level, gender was only significant in the lowest AGI strata, where it was positively related to noncompliance. However, it was also significant at the .10 level in the highest AGI strata, and again positively related to noncompliance. Although not constant across these AGI stratas, this analysis appears to validate previous research-males are less compliant than females. Non-Filers in 1986. This variable which was intended to be a surrogate for an attitudinal variable in this study was significant in only the highest two AGI stratas (AGI4 and AGI5). In the highest AGI strata, the regression coefficient was almost three times the size of the coefficient in the second highest AGI strata. In both cases the coefficient was, as expected, positively related to noncompliance. A couple of interesting observations can be made about the variable given the stratified regressions. First, based on its significance in only the higher AGI stratas, this variable supports the contention that in this study, income level and noncompliance are positively correlated. Second, this variable was not significant in the lower AGI stratas. Therefore, this stratified analysis corroborates the finding in the 214 second no contact data base (C0NNC0N2) where the regression coefficient for this variable increased (relative to the initial no contact data base - CONNCON) when the retired/student occupational group was dropped from the analysis. At that time, it was hypothesized that the taxpayers in the retired/student group were different than higher income taxpayers as it related to continued compliance post-amnesty. Once these taxpayers filed during amnesty, the results indicated a higher chance level of compliance (i.e., keeping them in the system). It was conjectured that the taxpayers in the retired/student group may have had a different set of ethics and/or were fearful of the consequences if they returned to their pre-amnesty practices. For whatever reason, once they filed under amnesty, they have continued to file. In summary, once taxpayers in lower income levels (and particularly within the retired/student occupational group) filed under amnesty, the results indicate that they have continued to file (i.e., it appears that one of the goals of amnesty -- adding taxpayers to the tax rolls -- may have "worked" for these taxpayers). On the other hand, the same cannot be said of taxpayers in higher income stratas. CHAPTER 7 IMPLICATIONS OF FINDINGS AND EXTENSIONS OF THE RESEARCH EFFORT 7.1 Introduction This research effort began contributions in several areas. with the objective of making First, this research was designed to increase our understanding of what factors are part of the noncompliance decision; second, it was designed to investigate characteristics of the amnesty participant in Michigan; and third, the analysis of the results were to be used to make recommendations about future enforcement efforts. The findings of this study as they relate to each of these areas are discussed below. In addition, based on the results of this research effort of this study, several extensions are warranted.Although these extensions may not be feasible at this time, it is appropriate to discuss them briefly. 7.2 Implications of Findings 7.2.1 Academic Noncompliance Research The implications of the findings of this study from an academic research standpoint must be viewed in 1ight of the data examined. data used in this study are from amnesty participants. The As has been mentioned previously, this limits the findings from several perspectives. First, amnesty participants are a subset of all tax evaders and delinquent taxpayers. population; Another group of tax evaders exists those who chose not to participate in amnesty. in the It is possible that the characteristics of these non-amnesty evaders are quite different from the amnesty participants who are part of the Michigan Amnesty Data Base (MADB). 215 216 Second, the information in the MADB does not necessarily include all the taxable income of amnesty taxpayers (i.e., these amnesty participants may have chosen to disclose only a portion of this income). It is possible, therefore, that the MADB may not be comprehensive with respect to data on every subject. from Michigan. Third, the data base includes only taxpayers Generalizability of research findings using this data to the national population is not warranted without further investigation and analysis. Finally, the data does not include any psychological or attitudinal variables related to the noncompliance deci si on. On the other hand, the MADB does offer some significant advantages over the primary analytical data base that has been used in academic research to date, the Taxpayer Compliance Measurement Program (TCMP) data base. Specifically, the MADB includes information from taxpayers omitted from the TCMP (i.e., taxpayers who had not filed returns previously and taxpayers who had not made a full disclosure of income). Although not a perfect data base, the MADB offers characteristics of amnesty participants. an opportunity to explore the Specifically, this study offers insight into differences between amnesty participants who were merely delinquent in their filing (i.e., the majority of their tax liability had been remitted to the State of Michigan), and participants who paid in a substantial portion of their 1 iability during amnesty. The goal of academic research efforts in this area is to completely understand the economic, demographic, and attitudinal characteristics of tax evaders and their noncompliance decision process. Once this has been accomplished, a more complete modeling of the noncompliance decision is possible. Admittedly, this effort falls short of that goal. On the other hand, this study was not designed to achieve that objective. 217 Instead, it was hoped that this research would add to the knowledge currently in existence regarding the noncompliance decision. In Chapter 2, a reference was made to Denzin [1978] who said: I conclude that no single method will ever permit an investigator to develop causal propositions free of rival interpretations. . . multiple methods of observations must be employed. This is termed triangulation, and I now offer as a final methodological rule the principle that multiple methods must be used in every investigation, since no method is ever free of rival causal factors. Given this MADB perspective, and the 1 imitations and advantages of the discussed above, this study has accomplished its academic objectives. Analysis of the Research Data Base. Specifically, this effort has confirmed the existence of certain economic and demographic factors in the noncompliance decision process. Although only briefly discussed here, these findings are discussed in greater detail in Chapter 6 . A comparison of the results of this study's findings to that of other analytical studies can be found in presentation of the analytical studies reflect the results of this study. Table 7-1. Table 7-1 is a portion of Table 2-2, revised to In addition to identifying factors that are part of the noncompliance decision, Table 7-1 discloses that this study has provided the first analytical evidence for several significant variables (gender, opportunity, and family size). Specifically, as to the analysis of the research data base, this study suggests that noncompliance increases as income increases, and is consistent with the findings of Mork [1975], Clotfelter [1983], and Witte and Woodbury [1985]. Taxpayers having some form of Treasury Department contact prior to amnesty (withheld taxes, estimated taxes, a W-2, or a letter from Treasury requesting information regarding a state tax return) TABLE 7-1 COPARISON OF RESULTS TO OTHEB ANALYTICAL NONCOPLIANCE STUDIES AMALYTICAL S U S I E S Author/ Date Age Gender (Fern) Educat­ Income level ion Allingham & Sancino, 1972 Wth'hld Occupa­ ComIneons tion/ pli ant Peers Source Status Com­ Agency plexity Contact - Sanc­ tions Probab. of Detect. Tax Rates + ♦ + + Srinivasan, 1973 Yitzhaki, 1974 0 Mork, 1975 + ,- • + - + - - Cox, 1984 0 + Cowell, 1985 Witte & Woodbury, 1985 + + Madeo, Schepanski, & Uecker, 1985 Young, 1988 Religion + Groves, 1958 Clotfelter, 1983 Other Variables - + + t- + + + - •+ + + + + +, - + - + 0 0 Interactions of Variables Opportunity, Geographic location. Family size, Noncompliant attitude + Positively associated with compliance - Negatively associated with compliance 0 Link to compliance is indeterminate Multiple symbols indicate that the association with compliance differs for different segments of the taxpayer population. 219 had substantially less unreported income than those amnesty participants who had no contact prior to amnesty. amnesty participants with the Just the opposite was found among opportunity to evade taxes. The opportunity variable was created via a composite of occupation (selfemployed, business, professional, or sales), an income level of $30,000 or more, and access to cash income sources; it had not been tested analytically prior to this study. The occupations of amnesty participants were found to be a part of the decision process. Specifically, sales and self-employed taxpayers were more be 1ikely to non-delinquent amnesty participants, while unskilled laborers were more likely to be merely delinquent in fil ing their tax returns (i.e., these individuals were more 1ikely to have paid in most of their tax 1iabi1ity prior to amnesty). In addition, the special the self-employed variable created (ignoring occupational disclosure of the participant) was positively related to noncompliance, again indicating that these amnesty participants were more 1 ikely to be evaders rather than merely delinquent in filing returns. This study determined that those amnesty participants who made a choice not to file a return in 1986 had more unreported income than amnesty participants who filed a return in 1986 (i.e., these taxpayers were more 1ikely evaders than delinquent taxpayers). This would seem to indicate that the decision making process of evading taxes warrants further exploration. Although I am unable to generalize geographic findings, the study results do support the contention that geographic location is a part of the noncompliance decision. The area indicated in the regression analysis (Ann Arbor) corresponds to a finding of Witte and Woodbury that 220 better educated areas with large student populations generally have low levels of compliance. Ann Arbor would certainly fit this description. The majority of studies (survey and experimental research) testing the compliance level of males versus females have found males less compliant. This analytical study supports the position that single males are less compliant then other taxpayers. This study used exemptions as a surrogate for family size, and found that as family size increases, so does the likelihood of evasion (vs. delinquency of filing returns). Analysis of the AGI Stratas. This study segmented the research data base into five AGI strata, and analyzed each. Although this type of stratification had been suggested as an interesting and worthy extension in prior studies, this study appears to be the first to have actually performed this type of analysis. Specifically, the regressions provide some insight into how the noncompliance decision may change across levels of income. Contact was the only variable to appear in all of the five AGI strata regressions -- always with a negative regression coefficient. In addition, the size of the regression coefficient consistently increased across the strata. The relevant conclusion, therefore, is that those with contact were more 1ikely to be delinquent amnesty participants (rather than tax evaders), and as income increases, those taxpayers who had no contact (and were more 1ikely evaders) disclosed significantly more unreported income. Opportunity to avoid taxes could be evaluated in only two AGI strata due to the AGI definition that was part of its makeup. The variable, not 221 surprisingly, was significant at the .05 level in the AGI4 strata, and significant at the .10 level in AGI5 strata. The level of income variable was significant in four of the five AGI stratas (only in the third strata - AGI from $15,000 to $24,999 - was it not significant), indicating that as income level increased, the amount of unreported income increased. The stratification, however, did mitigate the effect of this variable in all but the lowest income strata, where it continued to be the dominant variable in the regression equation. Several interesting findings come out of the AGI strata regressions related to occupations. First, the professional occupational category was significant in only the highest AGI strata. carried a negative regression coefficient, In that analysis, it indicating that this occupation was more likely to have been delinquent in their filing rather than evaders when compared to all other occupations. Second, the sales occupational group appeared positively (at the .10 level) in every AGI strata regression, indicating that the evasion occupational group crosses al 1 levels of income. taxpayers (both the occupational tendencies of this Third, self-employed and non-occupational variable) were positively related to noncompliance in two AGI stratas (AGI2 and AGI3). It would appear that the control these individuals exhibit over receipts and disbursements (which provides an ability to manipulate income to a lower level) might explain their lack of significance in the higher AGI stratas. Fourth, the skilled labor category, not previously significant in any regression equation, was significant and positively related to noncompliance in the second AGI strata. Finally, as expected, the retired/student occupational group was significant and positively related 222 to noncompliance in the lowest AGI strata. This result indicates that these individuals were more likely to be evaders (rather than delinquent in filing returns) for low levels of income. At the .05 level, gender was significant only in the lowest AGI strata, where it was positively related to noncompliance. However, it also was significant at the .10 level in the highest AGI strata, and again positively related to noncompliance. Although not constant across these AGI stratas, this analysis appears to validate previous research which found males to be less compliant than females. The variable constructed to measure the effects of those amnesty participants who failed to file a return in 1986 was significant in only the highest two AGI stratas (AGI4 and AGI5). In the highest AGI strata, the regression coefficient was almost three times the size of the coefficient in the second highest AGI strata. that these individuals were more of income. lower 1 The analysis indicates ikely to be evaders with higher levels Given this finding, it would appear that once taxpayers in income levels (and particularly within the retired/student occupational group) filed under amnesty, they have continued to file (i.e., it appears that one of the goals of amnesty -- adding taxpayers to the tax rolls -- may have "worked" for these taxpayers). On the other hand, the same cannot be said of taxpayers in higher income stratas. 7.2.2 The Amnesty Participant in Michigan It appears that the MADB is unequaled as a source of data on amnesty participation and effects. Most analyses of state tax amnesty programs publ ished over the past few years rely on aggregated data provided by state revenue agencies (see, for example, Mikesell Hirlinger [1986]). [1986]; Parle and These analyses provide very 1 ittle detail about the 223 various types of noncompliance uncovered during amnesty, or their relative importance. Michigan resources appears to be the only state to have committed necessary participants. to The compile detailed data contained in the information MADB on provides the amnesty a unique opportunity to analyze amnesty participants in order to provide insight into some of their characteristics. This information would be beneficial in conceptualizing and administering an amnesty program -- specifically by providing a profile of the typical amnesty filer. In addition, it could be used to target the marketing efforts of the program and design the enforcement efforts subsequent to amnesty. Based on the descriptive analysis of these taxpayers, some basic conclusions can be reached. First, most of the participants had not previously filed returns (as contrasted with amnesty participants who filed amended returns). research data base. small tax payments. Nonfilers comprised almost 77 percent of the Second, most of the participants made relatively Although the average amnesty payment in the research data base (RDB) was $408.51, the median was $138.06, indicating half of the amnesty participants in the RDB paid less than that amount. Third, the majority of amnesty participants (about 71 percent) filed for only one year. This information can serve to indicate the general type of amnesty participants in Michigan. The regression analyses indicate that certain of these participants were more returns. 1 ikely to be evaders than merely delinquent in filing their For example, the regression analysis of the RDB would indicate that the evaders among amnesty participants had one or more of the 224 following characteristics: a higher level of income, no previous Treasury Department contact, opportunity to evade, either self-employed or employed in sales, living in the Ann Arbor area, a male, a higher number of exemptions, and one who disregards the enforcement system. Each of the AGIstratas can be evaluated in a similar fashion. 7.2.3 Designing An Amnesty Program As a precursor to discussing the structure and operation of an amnesty program, it is important to understand what type of taxpayers have come forward during the programs to date. useful since it enables us to visualize the type participate inan amnesty program. With This information is of taxpayer1 ikely this data in hand, to it is possible to properly structure an amnesty/enhanced enforcement program. In addition, it may provide some insight into those taxpayers who are not participating in tax amnesties. Some Demographical Information Regarding the Amnesty Participant. This study has determined that certain factors are part of the makeup of a typical amnesty participant. above. These factors were briefly discussed In addition to this information, the amnesty data provide additional insight into these participants. From the information at hand, reached. several other conclusions can be First, the vast majority of participation is in the individual income tax. This result applies not only in Michigan (where almost 70 percent of the returns were filed by individuals), but in most other states that have run amnesty programs to date. Second, the majority of amnesty participants filed returns never filed before (vs. filing amended returns to correct deficiencies). returns were non-amended. In the RDB, almost 77 percent of the Finally, most of the amnesty participants made 225 relatively small payment was tax payments. $138.06, In the RDB, the median amnesty tax indicating that 50 percent of the amnesty participants disclosed $3,000 or less previously unreported income during amnesty. The mean amnesty tax payment was $408.51, which corresponds to about $8,900 of previously unreported income. The Psychology of the Amnesty Participant. As previously mentioned, amnesty participants are a subset of all tax evaders. A question which must be addressed relates to their decision to disclose themselves during the amnesty program. It would appear that these individuals were motivated in some fashion to come forward during amnesty. For example, given the marketing of amnesty in Michigan ("get to us before we get to you") and the increases in penalties and interest post-amnesty, it would seem that the typical amnesty participant is risk-averse (feeling that there is a relatively high probability of detection post-amnesty), and with some guilt (a conscience that wants to rectify a situation).Any amnesty effort, therefore, should capitalize on these elements. feeling of Somewhat surprisingly, prior research would indicate that these same items are what motivates people to be in compliance. reason, it failed to work for these individuals until For some the amnesty program. The Amnesty Program. considering an amnesty Most states have several goals in mind when program. The most obvious outstanding state revenues at 1imited expense. is collecting These revenues may include current assessments (i.e., accounts receivable) in addition to revenues which are uncollectible due to 1imited enforcement efforts. is obvious from thisstudy, that many amnesty participants It were potentially known to the Department of Treasury prior to amnesty (e.g., 226 withheld taxes, estimated taxes, W-2, etc.) but had not filed returns. These delinquent filers, usually making small significant portion of amnesty participants. tax payments, were a A second goal for the amnesty program is to promote improved compliance with the tax laws subsequent to amnesty. Finally, the amnesty program hopes to add taxpayers to the tax rolls who had managed to stay outside the tax system prior to amnesty. Based on this study, and a tangential involvement in the Michigan amnesty program, several factors in the design of an amnesty seem to be important. These factors include public relations efforts, program publicity, scope of the program, and enforcement agency funding. Each of these is discussed briefly below. Public Relations. "It's the message that's important" is a phrase applied in many situations, and one that applies to an amnesty program as well. How the amnesty program is communicated to the taxpaying population will contribute heavily to the program's success or failure. Failure, here, is not immediately measured in dollars, as any amnesty program is expected to accelerate revenue collections. If, however, taxpayers view amnesty as "unfair," future compliance may be impaired. It is vital enforcement effort that citizens view amnesty as part of an overall designed to find noncompliant taxpayers. The marketing of the program must communicate that amnesty is the final opportunity to "wipe the slate clean." Amnesty should be viewed as the beginning of a new era in enforcement efforts -- efforts that include, for example, increased audit staffing, greater risk of audit selection, an active discovery unit designed to search for noncompliant taxpayers, 227 and a revised penalty and interest structure. In this way, amnesty is a public acknowledgement of a change in the rules. In addition, participants have noncompliance. it is important not benefitted to communicate financially from that amnesty their previous Most amnesty programs require the payment of all taxes and related interest and waive related penalties. It is important that compliant taxpayers understand that the waiver of penalties is not a significant benefit for amnesty participants. In most cases of self­ disclosure, an enforcement agency will waive all penalties in response to the taxpayer coming forward. As a result, there is very 1ittle difference (if any) between an amnesty program and normal enforcement policies in this regard. Subsequent to amnesty, distribution of information regarding the amnesty and its participants is vital, as the enforcement agency applies this knowledge to its enforcement efforts. taxpaying population It is critical that the see evidence of an active effort to identify taxpayers who remained noncompliant. Program Publicitv. If a government intends to influence a large number of taxpayers, the effective marketing of the program is critical to success. Obviously, significant amounts of publicity will require considerable funding. However, most states who have run amnesties and subsequently evaluated the program have stated that additional funding should have been sought for effective media presentation. The most successful amnesties have relied on an extensive media effort, including press releases, advertising. public service spots, and significant paid media Without question, publicity is the largest fixed cost to be covered in the amnesty program. However, given the need to communicate 228 the change in enforcement policies properly, it is obviously a key factor in amnesty design. Scope of the Program. Most states have run general amnesties rather than selective amnesties. general Generally, the reasons in favor of running a amnesty include the political environment, the potential for large revenue collections, and publicity costs being similar regardless of the size of the program. Given this background, several comments can be made based on this study. Prior research has indicated that taxpayer psychology plays a key role in amnesty. Compl iant taxpayers must view amnesty as part of a larger enforcement effort. Noncompliant taxpayers must view amnesty as a last chance to correct their evasion behavior. Since publicity is key to the amnesty program, enforcement agencies should take advantage of media coverage prior to amnesty. Several opportunities establishment exist (or enlargement) for enforcement agencies. The of a discovery unit to identify tax evaders should precede the amnesty program. This unit could be used for a number of projects prior to amnesty, all of which would result in favorable publ icity for enforcement prior to amnesty. The most visible would be an active identification program. Using agency and other government information, along with factors identified from this and other studies, potential amnesty. requiring evaders can be targeted For example, 1 all professions This unit could also identify a key 1ist of largt tax evaders prior to amnesty. to a matching program between prior to icensure in the state and state income tax return files might yield nonfilers. prior and contacted amnesty, In Michigan, a "dirty dozen" 1ist was created and the press informed of its existence. The 229 Department of Treasury indicated that they would be watching for these taxpayers to file for amnesty. If amnesty was not sought, they would prosecute them immediately after amnesty. Several of these individuals did apply for amnesty; those that did not were prosecuted within 60 days of the program's conclusion. Since the descriptive analysis of the RDB indicated a significant number of small payments (over 50 percent less than $110), a reasonable question is how to handle these participants. For the most part, these participants were delinquent filers rather than tax evaders. portion of their tax 1 iability was paid in prior to withholding or estimated payments). A 1 arge amnesty (via During amnesty, these individuals filed the return in question and paid the balance due. It would seem that an enforcement agency could capitalize on this by actively pursuing these individuals prior to amnesty, thus creating another publicity mechanism for the enhanced enforcement effort. Retired taxpayers and students made up a large portion of the RDB. These taxpayers participated in amnesty for a variety of reasons, including disclosing evasion actions. these taxpayers were not more occupational groups. 1 However, this study concludes that ikely to be tax evaders than other Although some taxpayers in these groups were evaders, itappears 1 occupational ikely that many of these taxpayers may not have known there was a need to file. If so, taxpayer service programs must be evaluated to see if adequate assistance rendered. is being An evaluation program performed prior to amnesty with changes in the assistance programs can also be an effective publicity vehicle. Obviously, the filing process must also be evaluated for these taxpayers. It could be that returns and instructions are too complicated, and need 230 revision. A pre-amnesty effort to assist these individuals in filing returns could be beneficial to these taxpayers and viewed positively by compliant taxpayers. Alternatively, it has been argued to restrict amnesty to only small amounts of amnesty tax (e.g., less than $200). Given the level of effort required during an amnesty program, it has been argued that these small payments could be collected with and making 1 ittle effort, thus decreasing the cost/benefit ratio amnesty more palatable to compliant taxpayers. The rationale here is that this type of program could allocate the scarce resources of the enforcement agency to discovery efforts targeted at evaders. Accessibility to the amnesty program also must be evaluated. for amnesty must be as easy as possible. Filing The design of the amnesty application should only require necessary information to properly process the amnesty request (e.g., taxpayer identification, summary of returns filed, total remittance). From an academic research standpoint, some information not otherwise available might be requested as part of the amnesty program. example, disclosures regarding participants might be sought. age and education level of For amnesty In addition, since taxpayer attitude plays a role in the noncompliance decision, follow-up interviews or surveys might be required (or agreed to if needed) application. This requirement must be balanced with the anonymity status of most participants in amnesty programs. made to as part of the amnesty participants to ensure good Assurances would have to be participation. If, however, interviews or surveys could be sought, the acquired information would be immensely valuable in modeling the noncompliance decision. 231 Enforcement Agency Requirements. The amnesty program is only one part of a larger enforcement effort. The message communicated to compliant taxpayers is that subsequent to amnesty increased enforcement efforts will be used to identify tax evaders and delinquent nonfilers. As a result, increased funding for the enforcement agency is mandatory as part of the effort. The creation of a discovery unit within the enforcement agency, increased staffing of the agency, computerization of enforcement efforts, and increased compliance. taxpayer services In addition, all will contribute to increased # the imposition of a revised penalty and interest structure will provide the agency with the tools needed to properly enforce the tax laws. What About a Second Amnesty Program? Given the above discussion, it would appear that such an opportunity is not feasible. Most states view amnesty as a one-time program since it is intended to be the key notification of a change in the compliance system. viewed negatively because taxpayers will compliance system, Second amnesties are perceive a change in the allowing for noncompliance until the next amnesty program. However, potentially a selective beneficial) second amnesty if restricted to might certain be feasible types (and of evaders identified as a result of amnesty program data, and efforts of the discovery unit. For example, a certain class of taxpayers identified as noncompliant during the amnesty program or via discovery could be offered amnesty. In order to distinguish this selective amnesty from the general amnesty, the enforcement agency would need to aggressively support its position, relying on its analysis of the original amnesty data. In this 232 study, the Ann Arbor SMSA was identified as a noncompliant region. Prior to enhanced enforcement efforts, a general amnesty could be offered for taxpayers 1iving in this SMSA. such a 1 The additional information gathered from imited amnesty could be beneficial and might be applied to other regions in the state (e.g., regions with higher incomes, a college or university, and high-tech industries in the case of the Ann Arbor SMSA). 7.2.4 Implications for Enforcement Efforts The regression results can be used to help in the shaping of the enforcement policies of the Michigan Department of Treasury (and could be used to support changes at the federal level). The regression results indicate that contact with the enforcement agency is a key compliance variable. At higher levels of income, there is a great deal of difference between amnesty participants who were merely delinquent in filing their returns and amnesty participants who were evaders. Coupled with the fact that amnesty participants probably chose to disclose themselves because of fear of detection or guilty feelings, it would seem that the increased presence (or perceived presence) of the enforcement agency would do much to reduce noncompliance -- at least among those taxpayers who are 1ikely to be amnesty participants. Some suggested actions would be a greater audit and compliance awareness, increasing fines and penalties, and an effective public relations campaign that makes taxpayers aware of the 1 ikelihood and consequences of detection. Given the fact that the retired/student group disclosed themselves in large numbers, but paid 1 ittle in tax, it would seem that an education program related to return requirements would be beneficial. It seems that these taxpayers may have evaded not because they wanted to, but because they were ignorant of the requirements. 233 Given the fact that three variables in the regression equation are related to cash income sources (opportunity to evade, self-employed, and sales occupations) it would seem that increased reporting requirements in this area would be beneficial (although it may be practically difficult). However, over the past few years, Congress has increased the reporting requirements in areas such as rental payments and cash transactions of $10,000 or more. An expansion of these requirements would be appropriate to evaluate. 7.3 Extensions of the Research As described in Chapter 2, the first step in the process of modeling taxpayer noncompl iance is the proper characteristics of the noncompliant taxpayer. identification of the Several extensions of this study seem warranted. First, a sample of compliant taxpayers in Michigan could be drawn from Department of Treasury files. The same data analyzed in this study could be collected from these taxpayers with an analysis of differences between the compliant and noncompliant taxpayers performed. Second, the taxpayers in the Michigan Amnesty Data Base could be cross-matched to IRS tapes provided to Michigan each year identifying all taxpayers who filed a federal Form 1040 using a Michigan address. Taxpayers identified as in the Michigan Amnesty Data Base, but not appearing on the federal tapes could be separately analyzed as noncompliant federal taxpayers (since the IRS was unaware of their existence). The results of that analysis could also be compared to the findings from the present study. 234 Further extensions could be carried out along several fronts. First, other characteristics outside those examined in this study could be explored. ■ An examination of taxpayers who filed amended returns during amnesty could be performed to determine the reason for the amendment (e.g., underreporting of income, too many exemptions, inappropriate deductions, IRS audit, etc.). ■ Tax evasion behavior by taxpayers across different types of taxes could also be investigated. It would be interesting to know the degree to which individuals or businesses who had a delinquency in one tax (e.g., single business tax) also exhibited a pattern of tax delinquency in another type of tax (e.g., sales and use tax). m Specific characteristics beyond those identified previously could be developed for business taxpayers, including the form of entity (e.g., corporation, partnership, S corporation, etc.) and business endeavor (e.g., manufacturing, sales, etc.). Second, after characteristics of amnesty participants have been identified, it would then be possible to examine hypotheses about why taxpayers choose to participate in an amnesty and to study the implications of that taxpayer behavior for various tax enforcement strategies. Taxpayers who participate in amnesty are, of course, a subset of all delinquent taxpayers. While the Michigan Amnesty Data Base does not include those taxpayers who chose not to participate in the amnesty program, it may be possible to gain some information about those taxpayers indirectly by comparing the characteristics participants to previously identified delinquent taxpayers. of amnesty Two sources of data for such taxpayers are taxpayers with outstanding accounts receivable to the Michigan Department of Treasury prior to amnesty, and the IRS TCMP sample. Third, it may be possible to develop a model of taxpayer choice concerning participation in a tax amnesty related to the economic models of tax evasion. As a general outline of the model, taxpayers will select 235 a pattern of tax behavior satisfaction or utility, enforcement, and penalties. in order to gain the highest possible given their perceptions about tax costs, A model of the amnesty tax decision is different from a general model of the decision to evade taxes because the amnesty is unanticipated and when it is proposed, the taxpayer has already made a decision about compliance. Finally, it might be possible to generalize the findings of the amnesty taxpayer choice model general. However, such to gain insight about tax evasion a generalization will in be conceptually and statistically difficult. This sort of disaggregated analysis will contribute to a better understanding of the likely factors associated with participation in a state tax amnesty, of its longer run revenue effects, implications of a federal amnesty. and of the APPENDIX A APPENDIX A MICHIGAN TAX AMNESTY DATA BASE - INITIAL DATA COLLECTION INDIVIDUAL AND INTANGIBLES TAX RETURNS AUGUST 6. 1987 MICHIGAN TAX OTCSTY DftTft BASE - INITIAL DATA COLLECTION IICIVIDUAL AND INTANGIBLES TAX RETURNS COHCT VARIABLES FDR ALL TAX RETURNS _ _ _ _ _ _ _ _ Item/Variable EColums)_ _ _ _ _ _ _ _ 1. Taxpayer ID Nutters: m Taxpayer ID (9) _ = ■ Other ID (9) ^ = 2. Zip Code (5) 3. Type of Tax (2) 4. Tax Return Year (2) 5. Annesty Tax Paid, As Audited (10) 6. __ __ __ _ Interest Paid, As Audited (10) UNIQUE VARIABLES FOR ItOIVIDUAL 1NCPE TAX RETURNS VARIABLES TO BE COLLECTED FROM ALL RETURNS; ________ Item/Variable (Columsl________ 1. Occupation - Taxpayer (2)_______________ _____________________ 2. Occupation - Spouse (2)_________________ _____________________ 3. Residency (1) 4. Filing Status (1) 5. Sex (Single and ITSreturns only) (1) 6. Exemptions (2) 7- AGI (10) .= = 8. Additions (10) ===== > = ,= 9. Subtractions (10)__________________________________ __ 10. Taxable Income (10) = = > = ===== » . 11. Tax (10) = = = > = = = . = 12. Tax Withheld (10) = = = . = 13. Estimated Tax Payments (10) 14. Balance Due (10) = . = _ _ ___, _ _ - PAGE 1 - . 238 ___________ Item/Variable_(Colurrts)__________ 15. Is a W-2 present? (1) = 16. Does return contain a letter,explanation, etc.concerningannesty? (1) __ 17. Is a Michigan Dept, of Treasury letter re: nostatetax return __ filedattached? (1) 17a. Preparation of return. Note: 18. 19. Collect the information for items 18 and 19 from all returns that are NOT "amended". Additions to Income: Non-MI Mmi Interest (10) __ __ , _ _ _ _ _ _ , _ = _ = _ = .=== ^ Capital Gains (10) = T_ = , _ _ _ _ _ _ , _ _ = _ _ _ ._ _ _ _ Losses fran Other States (10) == ,___ _ _ _ _ , _ _ = _ _ _ ._ _ _ Subtractions from Income: US Government Interest (10) ___ _ = , _ _ = = Military Benefits (10) __ , _ _ _ _ =_ . _ _ __ Retirement Benefits (10) _= __ , _ _ _ _ =_ IncoiK from Other States (10) __ __ , _ _ _ _ _ _ , _ _ _ _ _ _ ._ _ PROPERTY TAX CREDIT INFORMATION: , _ = === _ = ._ _ _ The following information will be collected frcm MI-1M0CR, if available. 20. Rent or Own Home (1) =_ 21. Salaries, wages, tips, other comp. (10) 22. All dividends and interest (10) .= = = = = = = 23. Rent, royalty, and net business income (10) _ _ _ _ , _ _ _ _ _ _ ,_ _ _ _ _ _ ._ _ _ _ 24. Annuity and pension benefits (10) ._ _ _ _ = = __ __ 25. Net farm income (10) >= 27. Alimony and other taxable income (10) 29. Worker’s conpensation, other benefits (10) 30. Household income (10) Rent paid (10) -= = = = = ===== === = = , ^ ,__ = = _= . =_ _ _ _ _ • _ _ === ______________________________________ 28. Child support (10) 32. = = = = = ■ , _ _ _ _ _ _ ,_ _ _ _ _ _ __ __ , __ ■ Source(s) of other taxable income Property taxes paid (10) = > . 26. All capital gains less capital losses (10) _ _ _ _ , 31. = - PAGE 2 - 239 VARIABLES TO BE COLLECTED ON AMFJPED RETURNS: _ _ _ _ _ _ _ _ Item/Variable FColums)_ _ _ _ _ _ _ _ Note; For the following items, detail the amourts ORIGINALLY REPORTED. 1. Residency (1) 2. Filing Status (1) 3. Exenptions (2) 4. AGI (10) = = . = = = 5. Additions (10) __ 6. Subtractions (10) __ __ , __ __ __ , __ __ 7. Taxable Income (10) = = > 8. Tax Liability (10) 9. Property Tax Credit (10) = = === . _ __ __ , __ __ __ 10. Tax Withheld (10) ^ 11. Estimated Tax Pc^yments (10) __ __ , __ __ ^ ^ ^ Note; Detail the following amomts. 12. Amount Paid With Original Return (10) 13. Refund Shown on Original Return (10) 14. Reason for Change in Ntnlber of Exenptions Note; The following information will be inferred from reading Part VII of MI-1040X. 15. AGI Error 1 (1) 16. AGI Error 2 (1) 17. Addition Error 1 (1) 18. Addition Error 2 (1) 19. Subtraction Error 1 (1) 20. Subtraction Error 2 (1) 21. Reason for Amended Return was IRS Audit (1) - PAGE 3 - 240 UNIQUE VARIABLES FOR INTANGIBLES TAX RETURNS VARIABLES TO BE COLLECTED FROM ALL RETURNS: ___________Item/Variable (Colums)___________ 1. Filing Status (1) 2. Stocks and Bonds(10) 3. 4. _ = = >_ _ Accounts and Notes (10) ___ _ _ ,_ s_ Mortgages and Land Contracts (10) __ _ 5. Annuities (10) 6. Statutory Deduction (5) _ _ = = = ,= = _ = __ = = _ _ ._ _ === = _ _== . ,_ _ _ _ _ _ _ __ _ __ __ __ . __ __ _ Tax Due (10) Note; _ _ ,_ _ ,_ _ _ = _ _ , _ _ 7. Cash on Hand (10) 8. _ ;!;===, ._ _ . If the return is an amended return, collect the following information frcm the original return. 9. Filing Status (1) __ 10. Stocks and Bonds(10) _ _ _ _ ,_ _ _ = _ _ , _ _ _ _ _ _ 11. Accounts and Notes (10) _= _ 12. Mortgages and Land Contracts (10) _ _ _ _ ,_ _ _ _ _ _ , _ _ 13. Annuities (10) = = * ,=_ =_ _ = , = = = >_ 14. Statutory Deduction (5) Tax Due (10) . __ __ _ _ _ _ ._ _ _ _ _ _ ■_ _ . 15. Cash on Hand (10) 16. ___ ._ _ = = , = = = > = - °m. 4 - ._ _ _ __ ___ 241 MICHIGAN TAX WNESTY DATA BASE DATA COLLECTION PROCEDtglES MEM0WH3UM M O T H m W O INTANGIBLES TAX RETURNS The following infonnation is to assist in the collection of data from the Michigan Tax Annesty returns selected for analysis. One data collection docunent (DCD) will be filled out for each return selected. As a result, there may be several collection docunents for each taxpayer amnesty profile (computer generated printout). The ultimate goal of this process is to collect data on a consistent basis. proper analysis of the data collected. References within this memorandim are to pages of the DCD. This will allow for a 242 GENERAL PROCESS TO BE FOLLOWED 1. Remember that you will have a Data Collection Document (DCD) for each tax return. An Annesty Taxpayer Profile may include more than one tax return. If so, some information you will be required to collect for the DCD may be repetitive across returns (e.g., Social Security nmfoer, zip code, etc.) but be careful, as even this information may change between returns. 2. Section I of Page 1 of PCD. The information under the category "Cannon Variables For All Tax Returns" will be collected from every return. This information will be drawn from the carputer printout (Amnesty Taxpayer Profile) that surrounds an atmescy form and related returns. 3. Balance of DCD. Which remaining port’on(s) of the DCD you use will be dependent on the type of tax form you are working with: If tax form is: d Form MI-1040 Data Collection Process In general, you will be using pages 1 and 2 of the DCD. If the taxpayer has used a normal MI-1040 to file an amended return, you will use pages 1, 2, and 3 of the [CD (normally, the word "tended'' will be written across the top of Form MI-1040). Use "emended" information (i.e., the "correct" amounts) to fill out pages 1 and 2 of the DCD. Use "as originally filed" information to fill out page 3 of the DCD (if this information is not available, write 'WA" in each entry area on paqe 3 of the DCD). Form MI-1040X Some taxpayers used Form MI-1040X (tended Return) even though they were filing a return for the first time (i.e., they should rat have used a MI-1040X). For our purposes, a "true" amended return will have sone nonzero ntnters in Colmn A of Parts II, III, and IV of the form. a If you a return with some nonzero nuibers in Colum A, the process below applies, a If not, the process for Form MI-1040 (above) applies. "True" Amended Return; In general, you will be using pages 1, 2, and 3 of the DCD. Pages 1 and 2 of the DCD: The information detailed "On This Return" in Part I of the form, or in Colum C ("Correct Amount") in Parts II, III and IV of the form will be used to fill out pages 1 and 2 of the DCD. If a MI-1040 CR is attached, use the amended version to fill out the bottom of page 2. This information is the equivalent of the "final return" filed by the taxpayer. Paqe 3 of the DCD: Page 3 of the DCD is intended to capture information originally reported. The information detailed "Or Original Return" in Part I of the form, or in Colum A ("As Originally Reported") in Parts II, III, and IV of the form will be used to fill out page 3 of the DCD. Intangibles Fill out page 4 of the DCD. If the return is an amended return, use section 1 for the "correct" amounts, and section 2 for the "as originally filed" amounts. 4. Upon completion of a DCD, place the tax return inside the DCD, and move on to the next return. Upon completion of all returns within a taxpayer profile, place DCD’s and related returns inside the profile, and move on to the next profile. 243 INFORMATION RELATED TO PAGE 1; Section 1: Cannon Variables For All Tax Returns Items 1 through 6: These items are to be drawn from the annesty taxpayer Item 3: Code profile (conputer printout). as follows based on type of tax: Type of Tax IIT INF Code 02 04 Items 5 and 6 are the amounts after audit by Treasury personnel. Selection of data for coding should be as follows: First Choice Written amounts on the "Audited" line of the profile under the "Tax" and "Interest" colurns. Second Choice Printed amounts on the "Claimed" Tine of the profile under the "Tax" and "Interest" colurns. Section 2; bhioue Variables for Individual Income Tax Returns In General; This information will generally be drawn from page 1 of Form MI-1040 or Form MI-1040X. Itans 1 throutft 5: This information will generally be found in Part I of the tax form. Items 1 and 2: Print occupation listed in Part I of the tax form for both the taxpayer and spouse. Item 3: For Form 1040X, code based on amended information ("On This Return"). Code as foTlows: Resident Nonresident Part-Year Resident Item 4: For Form 1040X, code based on amended information ("On This Return"). Code as follows: Single Married, FilingJoint Married, FilingSeparately Item 5: 1 2 3 Collect information only frcmSingle or Married. Filing Separately Returns. Code as follows: Male Female Item 6: 1 2 3 MI-1040: 1 2 List total exemptions claimed. MI-1040X: List total exenptions claimed, as mended ("On This Return"). Items 7 throuch 14: This information will generally be found in Part II of the tax form. For Form MI-1040X, collect information reported in Colum C of the form. 244 ItFORfflTION RELATH) TO PAGE 2; Items 15 throudi 17a: Item 15: Code as follows: Yes No 1 2 Line 16: If there is a letter, attachment, sane other docunent concerning annesty, or an explanation regarding annesty on the return itself, code this request "yes." Code as follows: Yes No 1 2 Item 17: Code as follows: Yes No 1 2 Item 17a: Based on an examination of the Declarations section of the tax form (MI-1040, Part X; M1-1040X, Part VIII), code as follows: Paid preparer Taxpayer 1 2 Items 18 and 19: These items are to be collected for all returns except Form Ml-1040X’s with sans nonzero nitrbers in Colum A of Parts II, III, or IV (i.e., these are "true" amended returns, which will be analyzed on page 3 of the DCD). The information requested will appear on page 2 of Form MI-1040. If you are analyzing a MI-1040X with zeros in Colum A, attenpt to determine this information from the Explanation of Changes (Part VII) on page 2 of the form. ■ If there is no entry on theappropriate line of the tax form, leave the DCD blank for that item. ■ Not all items of additionsand subtractions are requested. form that is not requested, ignore it. If an amount appears on a line of the tax a Consult the various years’ tax forms for assistance in selecting the correct data. PROPERTY TAX CREDIT INFORMATION: Items 20 throudi 32: These items are to be collected from Form MI-1040CR (Homestead Property Tax Credit) if attached to the return. If the form is not present, leave these items blank. a Consult the various years’ tax forms for assistance in selecting the correct data. a Specific requests are as follows: Item 20: Code as follows: Rent Hone Own Hone 1 2 Line 27: In addition to the nuneric information requested, list sources and amounts of other taxable incane in the space provided. 245 INFORMATION RELATED TO PAGE 3; In General: Page 3 of the DCD is used for "true" amended returns. For our purposes, a "true"amended return will have some nonzero nurtiers in Colum A of Parts II, III, and IV of the form. This portion of the OCD is intended to capture information originally reported by the taxpayer(s). The information detailed "On Original Return" in Part I of the form, or in Colum A ("As Originally Reported") in Parts II, III, and IV of the form will be used to fill out page 3 of the DCD. If the taxpayer has used a normal MI-1040 to file an amended return (normally,the word"Amended" willbe written across the top of Form MI-1040), use "as originally filed" information to fill outpage 3 of the DCD (if this information is not available, write "M/A" in each entry area on page 3 of the DCD). Items 1 throudi 11: Item 1: Code based on original information ("On Original Return"). Code as follows: Resident Nonresident Part-Year Resident Item 2: Code based on original information("On Original Return"). Code as follows: Single Married, Filing Joint Married, Filing Separately Item 3 : 1 2 3 1 2 3 List total exenptions claimed, "Or Original Return". Items 4 through 11: The information requested will appear in Colum A ("As OriginallyReported") II, III, and IV of the form. in Parts ■ Consult the sarple MI-1040X for assistance in selecting the correct data. Items 12 and 13: Detail these items from the appropriate line of the form. Item 14: Write down any reason given by the taxpayer for a change in the nurber of dependents claimed. information will be contained in Part VI of the form. This Items 15 through 21: The information requested must be inferred from reading Part VII of the form. Two errors may be coded for changes in each of the follcwing categories: Adjusted Gross Income Additions Subtractions If more than two errors exist in one of these areas, code the twr most significant errors on the DCD. other errors. Items 15 and 16: Code AGI errors as follows: Wages, conpensation, etc. 1 Interest and dividends 2 Rental, royalty, or net business income 3 Capital gains and losses 4 Other ■ 5 Ignore any 246 IffORHATICN RELATED TO PAGE 3 (CONTINUED!: Items 17 and 18: Code "Addition" errors as follows: Non-Michigan municipalinterest Capital gains Losses from other states Other Items 19 and 20: 1 2 3 4 Code "Subtraction" errors as follows: Income from US Government obligations Military pay and benefits Income from other states Retirement benefits Other 1 2 3 4 5 Item 21: Code as follows: Yes No 1 2 INFORMATION RELATED TO PAGE 4: In General; Page 4 of the DCD is used to collect information from intangibles returns filed during annesty. Items 1 throutii 8: Items 1 through 8 of the DCD are for all intangibles returns filed during annesty. return is an anertded intangibles return, code the amended information here. Line 1: If your Code as follows: Individual Fiduciary Lines 2through 8: 1 2 Consult the sample returns for assistance in selecting the correct data. Items 9 through 16; Items 9 through 16 of the DCD are for amended intangibles returns filed during annesty. Code information as originally filed here. If this information is not available, write "f(/A" in each entry area. Line 1: Code as follows: Individual Fiduciary Lines 2 1 2 through 8:Consult the sample returns for assistance in selecting the correct data. 247 MICHIGAN TAX AMNESTY DATA BASE OCCUPATIONAL CATEGORIES 01. Sciences (Mathematics, Physical Sciences, Life Sciences, Social Sciences) 02. Architecture, Engineering, Surveying 03. Medicine and Health (Doctors, Dentists, Veterinarians, Nurses) 04. Education (Primary, Secondary, College) 05. Clergy (Priests, Ministers, Rabbis, Nuns) 05. Law (Lawyers, Judges) 07. Creative Arts (Writing, Art, Journalism, Acting) 08. Accounting/CPAs 09. Management/Executive 10. Professional Support Services 11. Clerical 12. Computer Related Fields 13. Sales 14. Food, Beverage, or Lodging 15. Building Trades 16. Protective Services 17. Building Services 18. Personal Services (Barber, Cosmetology, Apparel) 19. Agricultural, Fishery, and Forestry 20. Natural Resources Processing and Extraction (Typing, Filing, Stenography, etc.) 21. Printing and Paperworking 22. Machine Trades (Skilled Labor) 23. Fabrication of Products 24. Transportation (Motor Freight, Transportation Fields) 25. Packaging and Materials Handling 26. Amusement, Recreation, Radio and Television, Motion Picture 27. Self-Employed 28. Student 29. Housewife 30. Retired 31. Deceased 32. Laborer 33. Other 248 MICHIGAN TAX AMNESTY DATft BASE OCCUPATIONAL CATEGORIES Attempt to determine occupation class using the following detail. If needed, consult a Form W-2, or the return itself for assistance 01. Sciences ■Mathematics ■Physical Sciences (Astronomy, Chemistry, Physics, Geology, Meteorology) ■Life Sciences (Biology, Psychology) ■Social Sciences (Economics, Political Science, History, Sociology) 02. Architecture, Engineering, Surveying ■Drafting ■Cartography 03. Medicine and Health ■Doctors ■Dentists ■Veterinarians ■Pharmacists ■Nurses ■Physicians Assistant 04. Education (Primary, Secondary, College) ■Vocational Education ■Librarians ■Archivists aMuseum Curators 05. Clergy (Priests, Ministers, Rabbis, Nuns, Gurus) 06. Law (Lawyers, Judges) 07. Creative Arts (Writing, Art, Journalism, Acting) ■Editor ■Photography ■Commercial Artist ■Fine Artists (Painter, Sculptor) ■Musician ■Dancer ■Model 08. Accounting/CPAs ■Auditor ■Taxation ■Controller ■Treasurer ■Budget/Cost Accounting ■Bookkeeping 09. Management/Executive ■Financial Analyst ■Corporate Officer ■Comment; If "Executive" can be categorized in one of the above 8 categories by examining other information in the return, classify the occupation in that category. Use this category as a last resort. 249 10. Professional Support Services ■Legal Assistant ■Dental Assistant ■Hospital Orderly 11. Clerical (Typing, Filing, Stenography, etc.) ■File Clerk ■Mail Room 12. Computer Related Fields ■Computer Programmer ■Systems Analyst ■Computer Operator 13. Sales ■Manufacturers Rep ■Sales Manager ■Insurance Agent ■Stock Broker 14. Food, Beverage, or Lodging ■Caterer ■Waiter/Waitress ■Maitre’d ■Bellman ■Concierge ■Housekeeper 15. Building Trades ■Carpentry ■Building ■Excavating ■Painting ■Plastering ■Welding ■Electrical ■Heating/Air Conditioning ■Plumber 16. Protective Services ■Police Officer ■Sheriff ■Fire Fighter ■Armed Forces ■Security Guards 17. Building Services ■Janitor ■Maintenance 18. Personal Services ■Barber ■Cosmetology ■Seamstress ■Tailor ■Dry Cleaner ■Launderer ■Masseurs ■Housekeeper 250 19. Agricultural, Fishery, and Forestry 20. Natural Resources Processing and Extraction ■Paper ■Petroleum, Coal, Natural Gas ■Chemicals, Plastics, Rubber, Paint ■Wood Products ■Stone, Clay, Glass 21. Printing and Paperworking ■Photocopying ■Typesetter ■Printing ■Binding ■Paper Cutting 22. Machine Trades {Skilled Labor) ■Metalworking ■Machinists ■Tool and Die ■Textiles ■Mechanics 23. Fabrication of Products ■Assembly Operations ■Metal ■Scientific ■Medical ■Photographic ■Optical ■Electrical Products and Equipment ■Automobiles ■Jeweler 24. Transportation (Motor Freight, Transportation Fields) ■Truck Driver ■Bus Driver ■Railroad ■Water Transportation ■Pilot ■FIight Attendant ■Passenger Transportation 25. Packaging and Materials Handling ■Packaging ■Moving Goods ■Storing Goods ■Baggage Handling 26. Amusement and Recreation ■Athlete ■Sporting Activities ■Theater Projectionist 27. Self-Employed ■In addition to coding as self-employed, occupational category, if possible. also determine a specific 251 S t a t e of M i c h i g a n D e p a r t m e n t of T r e a s u r y Treasury Building, Lansing, Income Tax Div ision D i s c l o s u r e Unit Michigan 48922 ss // I n f o r m a t i o n o b t a i n e d f r o m the I n t e r n a l R e v e n u e S e r vice, u n d e r the a u t h o r i t y of S e c t i o n 6 1 0 3 ( d ) of t h e I n t e r n a l R e v e n u e C o d e , i n d i c a t e s y o u m a i l e d y o u r 1983 Federal i n c ome tax return from a M i c h i g a n address. We hav e no record of r e c e i v i n g y o u r 1 9 8 3 M i c h i g a n i n c o m e t a x r e t u r n u n d e r t h e n a m e or s o c i a l s e c u r i t y n u m b e r as s h o w n a b o v e . M i c h i g a n h a s a t a x a m n e s t y p r o g r a m f r o m M a y 12 - J u n e 30 , 1 9 8 6 . T a x a m n e s t y is a o n e - t i m e o p p o r t u n i t y f o r d e l i n q u e n t t a x p a y e r s to p a y b a c k t a x e s w i t h o u t f e a r of p e n a l t y o r p r o s e c u t i o n . D e l i n q u e n t t a x p a y e r s m a y r e c e i v e a w a i v e r of t h e 257. p e n a l t y by f i l i n g t he ir r e t urn(s ) and an a m n e s t y form w i t h f ul l p a y m e n t o f t a x a n d i n t e r e s t b e f o r e J u n e 30. V i s a or M a s t e r C a r d will be a c c e p t e d for p a y m e n t . E n c l o s e d are M i c h i g a n i n c o m e tax forms, an a m n e s t y form and a q u e s t i o n n a i r e d e s i g n e d to h e l p us e v a l u a t e y o u r f i l i n g s t a t u s . P l e a s e r e a d t h e q u e s t i o n n a i r e c a r e f u l l y to d e t e r m i n e M i c h i g a n f i l i n g r e q u i r e m e n t s . If y o u h a v e i n c o m e s u b j e c t to F e d e r a l i n c o m e tax b ut e x e m p t f r o m M i c h i g a n i n c ome tax y o u a r e s ti ll r e q u i r e d to file a M i c h i g a n r e t u r n and re p o r t s u c h in c o m e as a d e d u c t i o n f r o m a d j u s t e d g r o s s i n c o m e . If y o u t h i n k y o u ov>e b a c k t a x e s , p l e a s e r e p l y b y f i l i n g a 1 9 8 3 M i c h i g a n i n c o m e t a x r e t u r n , i n c l u d i n g r e m i t t a n c e of t a x d u e p l u s p e n a l t y of 257. a n d i n t e r e s t at 3 / A o f 17 p e r m o n t h f r o m t h e d u e d a t e of t h a t r e t u r n . If y o u f i l e a n a m n e s t y f o r m a n d p a y t a x a n d i n t e r e s t d u e , p e n a l t y cf 257. w i l l be w a i v e d . A f t e r a m n e s t y , t h i s r a t e o f p e n a l t y w i l l i n c r e a s e to 5 0 7 a n d i n t e r e s t w i l l b e 17 a b o v e t h e p r i m e r a t e o n J u l y 1, 1 9 8 6. If y o u f i l e d a 1 9 8 3 r e t u r n o r a r e n o t r e q u i r e d t o f i l e , the c o m p l e t e d q u e s t i o n n a i r e and a ny n e c e s s a r y do c u m e n t s our records. You must rep ly by June 30 your Federal Return, will or t h e t a x , p e n a l t y a n d be ass essed . please respond with an d we will correct interest due bas ed P l e a s e c o n t a c t o ur n e a r e s t o f f i c e or c al l: For a m n e s t y 1 - 8 0 0 - 4 6 8 - 2 9 3 7 ( 1 ~ 8 0 0 - I - 0 ~ T A X E S ) . For other inf o r m a t i o n 1-800-292-6424. PLEASE ATTACH E n c 1o s u r e s YOUR REPLY TO THE E N C L O S E D C O P Y OF THIS LETTER. on 252 Office Use Michigan Department of Treasury b. 84 MICHIGAN INDIVIDUAL INCOME TAX RETURN C. MI-1040 1. For 1984, or taxable year beginning _ d. .,1 9 8 4 , ending. 19____ _ This form is issued under the authority of the Income Tax Act of 1967, as amended. Filing is mandatory. See filing requirements and penalty and Interest statement In the MI-1040 Instruction Booklet. ID EN TIFICA TIO N (Please 8ypa or print) Firstname & Initial(ifJointreturn,use firstnanus & initialsofbath) Your socialsocutilynumber Page 1, item g Present homo addross(numbor and strootorruralroute) 3b. YOur occupation Poge City,townorpostoffice,and Slate Schooldistrictcode (see 1nsl.,pg.6) 4b. Schooldistrictname RESIDENCY STATUS H i FILING STATUS 9 m o. I ISingle a. 11 Resident Page 3d. Spouse'soccupation rage i, item 2 II STATE CAMPAIGN FU N D l, item 4 b. D Nonresident b. f IMarried, filingjointly c. □ Port-year resident— Dat03: c. f 1Married, filing separately — Enter 9pou3o's social security number on tine3c andontor spouse's name hero: Page 1, M o m 3 Yr. From ______ |______ 1____ I TO l, item #S|j| Spotiso'ssocialsecuritynumber 8 EXEMPTIONS Do you (or your spouse if joint return) want $2 of your taxes to go to this fund? Your decision wilt not increase your tax or reduce your refund. YES a. You ri b. Your Spouse n a. Your allowable Fodorat exemptions... NO b. Special exemption for homiplegics, paraplegics and quadriplegics..... 11 11 . TOTAL EXEMPTIONS Add tines Q(a)and 8(b) Poge : 1 "IN C O M E AN D A D JU S T M E N T S i. Item 6 Adjusted gross income which should be reported on Federal form 1040, line 32; Item I__ m or 1040A, line 14 or 1040 EZ, line 3. (See page 7 of Instructions).................................................................. 10a. Additions to adjusted gross Income (from line 42 of this form ).................................. .. 10b. Enter amount of Federal employed married couple adjustment from Federal 1040, lino 30 or Federal 1040A, line 1 2 .................................... »............ 10a. Pogc lt 10c. Add lines 10a and 1 0 b ........................................................................................................................................ 11. Total. Add lines 9 and 10c.................................................................................................................................... 12 . Subtractions from adjusted gross Income (from page 2, line 50 of this form)................................................. 13. Income subject to tax. Subtract line 12 from line 1 1 ........................................................................................................ 13. 10b.______ sum 11. ii Page 1, Item 9 14. Exemption allowance. Multiply $1500 by line 8c. Part-year and nonresidents enter line 6 5 . .. 15. Taxable income. Subtract line 14 from line 1 3 ................................................................................................................ 15. Page l, Item 10 16. Tax. Multiply line 15 by 5.05% (.0 5 8 5)............................................................................................................................. 16. Page l. Item 11 Page 1. Item 14 Page 17. 2, . .............................. m 7 H R F D IT S (S E E IN S T R U C T IO N S — P A G E S 7 T H R O U G H 9) " ~ Item 13 ------ 15---------------------------------------------------------1----------------------- Income tax paid to Michigan cities See page 7 of instructions............................................................ 17a 18. Public contributions. (See instr. page 8 .).................................... 19. Income tax paid to another state. (Attach copy of return)............... 20 . 21. firaj Gleaning Credit. Attach form M I-1040CR-8............................................. Total credits. Add lines 17a, 18a, 19aand20 ............................................................................................................... 21. 22. Income tax. Subtract line 21 from line 13. If line 21 is greater than line 16, enter "NONE" |M B M Ifla. 19.. ............................. P R O P E R T Y TAX AN D H O M E H EA TIN G C R E D IT S A N D PA Y M E N T S 23. Property Tax Credit. Attach MM040CR-1,2 ,3 or 4 ................................................. 24. Home Heating Credit. Attach MI-1040CR-7................................ 25. Farmland Preservation Credit. Attach MI-104GCR-5................. 26. Solar Energy Credit. Attach MI-1040CR'6 27. MICHIGAN TAX WITHHELD. ATTACH STATIE COPY OF W-2 . 28. Michigan estimated tax payments . . . H m m ............................ . . . m BD .............................. Pogc 1* Item Pago Item 29. 1983 overpayment credited to 1984............................................. 30. Add lines 23.24, 25,26, 27,28 and 2 9 .......................................................................................................... B M S ifg j 31. m I . . . H I 30. TAX D U E ST A T E O R O V E R P A Y M E N T S A N D C R E D IT S D U E YOU If line 30 is less than line 22, enter BALANCE OF TAX DUE STATE, . . . . . PAY^tm including interest______ and penalty_____ if applicable. See Instructions, page 7 ................................. 32. II line 30 is greater than line 22, enter overpayment...................................................................................................... 32. 33. CONTRIBUTIONS: 34. See Instr. page 10 I + --------- = Subtract tine 33 tram line 32 and enter difference........................................................................................................ 35. Amount ol line 34 to be credited to your 1985 ESTIMATED T A X ..........................................E S I _________ I 36. Subtract line 35 bom line 34. This Is your REFUND...................................................................................................... CHILDREN S TRUST FUND m N I 8sg| NONGAME WILDLIFE FUND V * SSB- - - - - - - - - 1 33. 34. ____ 253 ADDITIONS TO INCOME 37. Gross interest and dividends from obligations issued by states 38. Capital gains (from MI-1040D)............................................................................................. other than Michigan or their political subdivisions......................................................................................................... 37. 38. 39. Other gains (from MI-4797)................................................................................................................................................ 39. 40. Losses attributable to other states (see instructions, page 1 0 ) .................................................................................... 40. 41. Other (see instructions, page 10). Describe:_________________________________________________________ 41. 42. Total additions. Add lines 37 through 41. Enterhere andon page 1, line 1 0 a ............................................................. 42. 2. Item in Page 2. Item 10 Page 2. Item 10 Page ::r I SU B T R A C T IO N S FR O M IN CO M E 43. Income from U.S. government bonds and other U.S. obligations included in line 9 ................................................... 44. Military pay or military retirement benefits from U.S. Armed Forces included in line 9 (attach W - 2 ) ........................ 44. 43. 45. Capital gains (from MI-1040D).................................. 45. 48. Other gains (from MI-4797)..................... 46. 47. Income attributable to another 9tate. Explain type and source____________________________________________ 47. 48. Retirement or pension benefits included in line 9. See page 11 of 2. item 19 page 2. Item 19 Page 2. Item 19 page 2, Item 19 49a 49a. 49b. page Miscellaneous subtractions (see instructions page 11 for listing of eligible deductions.) Desorihe- 49b. 50. 50. mtmimm part-vfap A ^ n NntyRFSinFM? i w c o m f ai ■o c a t i o n TOTAL INCOME COLUMN A (Seepage 12 ofthe instructions) INCOME 51 Wages, salaries, tips, etc................................................................. 52. Interest and dividends 53 54. Business or farm in co m e............................................................... MICHIGAN INCOME COLUMN B INCOME FROM OTHER STATES — COLUMN C , 55. 56 Other (Describe) 57 Gross income. Add lines 51 Ihrough 5 6 ........................................ 58. Enter amount of Federal employed married couple adjustment ADJUSTMENTS TO INCOME 59. Enter the total of all other adjustments on lines 24-30 of your Federal return. Describe: 60. 81. Subtract line 60 from line 57. The amount in col. A should equal line 9. Enter the amount in col, B on line 63. Enter amount in col. C on line 47 or. If a negative amount, enter as a positive amount on line 4 0 ........................................................... PA R T-Y EA R AND N O N R E S ID E N T E X E M PT IO N A L L O W A N C E If you received gross Income not Included In line 9, see page 12 of the instructions before completing this schedule. 62. Multiply number of exemptions on line 8c by $ 1.500.00.............. 62. 63. Michigan source income: enter amount from line 61, Column B 63. 84. Divide line 63 by line 61. Column A. and enter percentage . . . . 64. 65, Multiply line 62 by line 64. Enter here and on line 1 4 ................. 65. | D E C L A R A T IO N S — S i g n b e lo w . If f ilin g jo in tly b o t h h u s b a n d a n d W ife m u s t s i g n . T h is re tu rn is d u e April 1 5 ,1 9 8 5 , o r o n th e 15 th d a y <>f th e fo u rth m o n th a fte r th e c lo s e o f y o u r ta x y e a r. /declareundeI rsp e a, lt o f p e ra ju yc to hm ap tl t h eeI t rn ue cy or r e c t nr d e t .nformationinthisreturn,and 1 do e co lf ar uc nh de al fnp e rl je ud ry t. hatthisreturnisbasedonallinforma­ ti n we hi 1r hp ae vn e at ny yo k o w ge aftech/nenfs, YourSignature Dale Page Spouse’sSignaturo(Iftiling|o[nlly,GOTH must elgneven IIonlyono had income) Mailing Instructions Pfoparor’s Signaturo. Business Namo. Address and IdentificationNumber PAY: Make checks payable to “ State of Michigan” . Record your social security number on the face of your check. Mail check and return to: Michigan Department of Treasury, Lansing, Ml 48929. 2. Item 17n REFUND OR CREDIT: Mail your return to: Michigan Department of Treasury, Lansing, Ml 48956 | 254 Office Use Michigan Department of Treaaury MICHIGAN INDIVIDUAL INCOME TAX R E T U R N MI-1040 , 1983, ending- 1. For 1983, or taxable year beginning- , 19- ID E N f iF iC A tlO N (Please type or print) First nsms & Initial(iffolnt return, use first nemos & Initials of botti) Lest nemo {|^ Your social security number Paste i, item 9 Present homo address (number and street or rural route) if Page City, town or post office, and State School district codo (see Inst.,pq. 7) S9 RESIDENCY STATUS I*! FILING STATUS a. □ Resident a. □ Single Page 3d. Spouse's occupation Page Item M B. EXEMPTIONS STATE C A M P A I G N F U N D Do you (or your spouso Ifjoint return) want $2 of your taxes to go to this fund? (Your docislon will not Incroeso your tax or reduce your refund.) i, Item oo instr., pg 7) I_ _ .i YiES NO □ □ 8 EXEMPTIONS FILING STATUS o. □ Single Pn0 e a. Your allowable Item 4 Federal exemptions ..... _ b. CD Marriod. filing jointly c. CD Married, filing separately — Enter spouse's social security number on lino 3c and enter spouse's name hero: b. Special exemption for ■ homlploglcs. paraplegics and quadriplegics ....... c. TOTAL EXEMPTIONS JL _L b. Your Spouse CD CD Add linos 8(a) and 8(b) Page 1, INCOME AND ADJUSTM ENTS 9. Adjusted gross Income which should be reported on Federal Form 1040, line 31, or 1040A, line 10 (Attach copies of any federal schedules that .® Indicate a I039 or adjustment from gross Income. See page 7 of Instructions) ................... 10. Additions to adjusted gross income (from page 2, line 38 of this form) ............................... Page I. Item 7 Page 1, Item Page li Item 9 I 11. Total. Add lines 9 and 10 ....................................................................................................................... . 11 12. Subtractions from adjusted gross income (from page 2, line 46 of this form) ..................... 14. Exemption allowance. Multiply S1500 by line 8. Part-year and nonresidents enter line 59 © ■® 15. Taxable Income. Subtract lirie 14 from line 13 ............................................................................... . 15 Page 1» Itern 10 16. Tax. Multiply line 15 by 4.6% (.046) .................................................................................................... . 16 Page I. Item 11 Page 1, Item . 13 13. Income subject to tax. Subtract line 12 from line 11 ................................................................... _______ I CREDITS (SEE INSTRUCTIONS — PA GES 8 THROUGH 10)__________ Pine 2. lie ® io AMOUNT PAID 17. Incoms tax paid to Michigan cities .............................. © . — — --------- )___ CREDIT 17a— 18. Contributions to Michigan colleges, universities, public libraries and public broadcasting, stations . . . 19. Income tax paid to another (attach ' (® .------- ----------- :— [------- 18a- state copy of return) •........................................................ 19. ----------- 20. Total credits. Add lines 17a, 18a ahd 19a i ................................................................................................................. 20. 21. Income tax. Subtract line 20 from line 16. If line 20 Is greater than line 16 enter "NONE" .....................® PROPERTY TAX AND HOME HEATING CREDITS AND PAYMENTS •/ TH 22. Property Tax Credit. Attach MI-1040CR-1, 2, 3, or 4 ...................................... 23. Home Heating Credit. Attach MI-1040CR-7 ....................................................... 24. Farmland ■Preservation or Solar Energy Credit. Attach MI-1040CR*5 or 6 25. Michigan tax withheld. Attach State copy of W-2 ........... ................... ........ Page 26. Michigan estimated tax payments ........................................................ ............. 12 in 27. 1980 overpayment credited to 1981 ............................. .................................... 28. Add lines 22. 23. 24. 25. 26 and 27 ..................... l. Item Item .......................................... j TAX DUE STATE OR OVERPAYMENTS AND CREDITS DUE YOU 29. II line 20 Is I0S3 than line 21, enter BALANCE OF TAX DUE STATE, Including Interest._______ and penalty , II applicable (see Instructions p. 6) 30. If line 28 Is greater than line 21, enter BALANCE DUE YOU .......... ■PAY1 ........... 30. 31. Amount of line 30 to be PAID TO YOU ................................................................ % 32. Amount of line 30 to be credited to 1982 ESTIMATED TAX (OVER) ^ Q . _L 1-1 259 PAGE 2, MI-1040 K U ADDITIONS TO INCOME i l 1 33. Qross interest and dividends from obligations issued by states . 33. . . 34. . . 35. . . 36. j 2. Item Page 19 2. Item T o Page 2< Item 10 37. 38. ' L _ ._ B iH ira W Page SUBTRACTIONS FROM INCOME | 39. Income from U.S. government bonds and other U.S. obligations Included in line 9 ......................... !. . 39. 40. Military oav or military retirementbenefits from U.S. Armed Forces included In line 9 (attach W-2) . 40. . 41. Page 2. Item 19 Page 2. item 19 . 42. _ 43 44. Page 2. Item 19 Page 2. Item 19 45.' Miscellaneous subtractions, including deductible portion of state income tax refund, ’ = political contributions (max. $50 single. $100 joint return), Michigan lottery winnings and other 45. 46. Total subtractions. Add lines 39 through 45. Enter here and on page 1, line 12 46. PART-YEAR AND NONRESIDENT INCOME ALLOCATION (See page 12 of the instructions) TOTAL INCOME MICHIGAN INCOME INC6ME FROM OTHER COLUMN A COLUMN fe STATES — COLUMN C INCOME 47. Wages, salaries, tips, etc................................................................. 48. Internet and Dividends •Exclusion • 49. Business or farm Income Rnlanre ............................................................. 50. Capital gains, from Federal schedules ..................................... 51. Income reported on Federal Schedule E ................................ 52. Other. (Describe) 53. Grcss income. Add lines 47 thru 5 2 .......................................... ADJUSTMENTS TO INCOME 54. Enter total adjustments claimed on line 30 of'your Federal income tax return. Describe. 55. Subtract line 54 from line 53. Amount in Column A should agree with line 9. Enter amount in Column B on line 57. Enter amount in Column C on line 43, or if negative amount, enter on line 36 ......................................................................................... PART-YEAR AND NONRESIDENT EXEMPTION ALLOWANCE If you received gross Income not Included In line 9, se e page 13 of the Instructions before completing this echeduH?. 56. Multiply number of exemptions on line 8 by $1,500.00 ...................................... • 56. 1 57. Michiaan source income: enter amount from line 55. Column B ................... 57. 1 58. Divide line 57 by line 55, Column A, and enter percentage • 58. 59. Multiply line 56 bv line 58. Enter here and on line 14.......... • 59. % 1 | DECLARATIONS — S ig n b e lo w . If filing Jointly b o th h u s b a n d a n d w ife m u s t sig n . T his retu rn Is d u e April 15, 1982, o r on th e 15th tlay of th e fourth m onth a fter th e c lo s e of your ta x y ear. I / { attachments, /d eclareundorpei n ye, otco pr er re jc ut ryen th tm hp ele It ne f. ormationInthhreturn, Ind erwh pe pea rn ju al ted th s.returnIsbasedonall fe oc rl ma ar te iou nnd of in ca hlt 1yho af ve yry knt oh w gi e a nd sal tt ru datco Your Signature Dale Spouse s Signature (iffiling Jointly. BOTH must sign even Ifonly one had income) Preparer s Signature. Business Name. Address and Identification Number Page 2. lies 170 ► Mailing Inatructlons: PAY — Make checks payable to "State ol Michigan". Record your social security number on the face of your check. Mall check and return to: Michigan Department of Treasury, Lansing, Ml 4B929. REFUND OR CREDIT — Mall your return to: Michigan Department of Treasury, Lansing, Ml 48956. | 260 Michigan Department of Treasury 1 For 1980, or taxable yoar beginning IDENTIFICATION Office Use 1980 MICHIGAN INDIVIDUAL INCOME TAX RETURN MI-1040 ________ . 1980. ending (Pleasetypeorprint > . . 19 _______ © Your socia l secu rity num ber P reso nt ho m o ad d re s s (n u m b e r a n d atre nt o r ru ra l to u to ) 3b Y o u i occu p a tio n Lily lown «>< |n)Al i*tin* .ind © School tlntncf code («?oo histf ,pg B) 3d Q f*»U n im » b in iita i M | re tu rn U*C tir ^ l na m *4 l> iftiTr o f b n lh j Poge 1, Item 3 Page 4a O RESIDENCY STATUS a 1 ,iResident b I )Nonrfl^KJent c 1 lPart-year resident Dates Page 1. Item 3 _____ I____ 1 lo YFS Your Spouse ! . S in g le M am ed POOO 1. J tfim Federal exemptions filin g join H v b Spocial exemption for Married, lilmg separately -Enter spouses social security number on line 3c and enter spouse s name here NO n n f.I n l. Item 6 EXEMPTIONS a Your allowable 1 F IL IN G S T A T U S Do you (or your spouse d joint return) wanl $2 of your taxos lo go lo '.his fund7 Your decmon will nol increase your la* or roduce your rotund l S p n g te S OCCupJI'On Page * ST A TE C A M P A IG N I U N O 1. Item hemiplegic*, paraplegics and quadriplegics c TOTAL EXEMPTIONS Add tines 0(a) and 8(b) Q . ------------- Pag* 1. Hr I nco m e a n d a d ju stm en ts 9. Adjusted gross Income which should bo reported on Federal Form 1040, lino 31, or 1040A, lino 11 (Attach copies of any Federal schedules that © Page t, Item (D Pago 1. Item n 11 Total Add fines 9 and 10 12 Subtractions from adjustod gross income (from page 2, line 46 of this fo rm ).................... 13 Income subject lo tax Subtract line 1? fromline it 11 0 Pago 1. I Item 9 14 15 16 0 Page 1. Itn IT) 10 Page 1. Item 11 Page 1. Item indicate a loss or adjustment from gross income. See page 0 of instructions) .................. 10. Additions to adjusted gross income (from page 2, lino 30 of this form) ............................. 13 Exemption allowance. Multiply $1500 by lino8Part-year and nonresidents enter line 60 Taxable income Subtract line 14 from line 13 Tax Multiply line 15 by 4 6% ( 046) 15 16 CREDITS (SEE INSTRUCTIONS — PAGES 8 THROUGH 10) Page 2. Item *3 17. Income tax paid to Michigan cities .................... 18. Contributions to Michigan colleges,universities, public libraries and public broadcasting stations 19 Income lax paid to another slate (attach copy of return) CREDIT A M O U N T PAID © 30. .... ►© 14 261 PAGE 2, MI-1040 ADDITIONS TO INCOME 33 Gross interest and dividends from obligations issued by states other than Michigan or their political subdivisions ............................................... 34. Capital gains (from MI-1040D) ................ ................................................... 35 Other gains (from MI-4797) .......................................... ............................. 36. Losses attributable to other stales (see instructions, page I I ) . .................. 37 Other (see instructions, page 11) Describe .......... .............. ... 38 Total additions. Add lines 33 through 37 Enter here and on page 1, lino 10 33. 34. 35. 36. 37. 38. Page 2, item 10 Pnge 2. Item in Page 2. Item SUBTRACTIONS FROM INCOM E 39 Income from U S government bonds and other U S obligations included in line 9 ................ 40 Military pay or military retirement benefits from U S Armed Forces included inline 9 (attach W-2) . 41 Capital gains (from MI-1040D) ................................................. 42 Other gains (from MI-4797)........................................... 43 Income attributable to another state. Explain typo and source 44 Retirement or pension benefits included in line 9 N a m e of Payor: ...... 45. Other (see instructions, page 11). Describe: 46 39. 40 19 Page 2. Item 19 41. 42 2. Item 19 43. Page 2. H e m 19 44 Pngn ................... .... .... 45 -------- ----- ----- Total subtractions. Add lines 39 through 45 Enter hero and on page 1. line 12.................... m a m Pago 2. Item 46. PART-YEAR A N D N O N R E S ID E N T IN C O M E A LLO CATION TOTAL INCOME COLUMN A (See page 12 of the instructions) INCOME --MICHIGAN I N C O M E INCOME F R O M O TH E R COLUMN B S TATES — C O L U M N C 47. Wages, salaries, tips, etc 46. Dividends Balanco __ Exclusion 49. Interest ....... ........ 50. Business or farm income .... 51 Capital gains, from Federal schedules. 52 Income reported on Federal Schedule 53 Othor. (Describe) 54 Gross income Add lines 47 thru 53 . E ADJUSTMENTS TO INCOME your Federal 55 Enter total adjustments claimed on tine 30 income tax return Describe........... 56 .... ... ................ Subtract line 55 from lino 54. Amount in Column A should agree with line 9 Enter amount in Column B on line 58. Enter amount in Column C on lino 43, or if negative amount, enter on line 36 PART YEAR AND NONRESIDENT EXEM PTION ALLOWANCE If you received gross Income not Included In line 9, or Income that can be deducted on lines 39 and 40, see page 13 of the Instructions before completing this schedule. 57. Multiply number of exemptions on lino 0 by $1,500 00 .................................................................................. 57. 58. Michigan source income: enter amount from lino 56. ColumnB .................................................................. 59. Divide line 58 by lino 56, Column A. and enter percentage.................................................................... 50. 59. 60. Multiply line 57 by line 59. (Cannot exceed line 57). Enterhere and on line 14........................................ 60. I B Q ^ I E C L A R A T I O N S — Sign below . If filing Jointly both h u sb a n d a n d wife m u st sign. This return h duo April 15, 1981, or on tho 15th day of th e fourth month after the clo se of your ta s year. rU pu ee n, allyolperjurythattheInformationInthisreturn.and tdeclareunderwh pe yho pea rn ju to hw at hg is atd te ac cl ha mr ee ntu sn ,de is in ca hlt I af ve yry kn let d ereturnlabasedonall / correct and compfoie Your Signature Spouoe'e Signature (Iffiling jointly, OOTH muat sign avan Ifonly one had Income) < Mailing Instructions: information of Preparer a Signature. BuaJnaaa Name, Addreea and Identification Number Page 2. Item 17a PAY - Make checks payable to "State of Michigan". Record your social security number on the face of your check. Mail check and return to: Michigan Department of Treasury. Lansing. Mt 48929. REFUND OR CREDIT - M ail your return to: Michigan Department of Treasury. Lansing. M l 48958. 262 Michigan Department of Treasury Office Use 1979 MICHIGAN INDIVIDUAL INCOME TAX RETURN MI-1040 1. For 1979, or taxable year beginning 1979, ending I d e n t i f i c a t i o n jp/ease type or print) ^ First noma & initial (if joint return, use first namos & initials of both) Pane | Last nams Your social security number i 1 I. Item 9 3b. Your occupotion Address (number and street or rural route) Page ZIP Code 4a. School district code (seo inslr.. pg 7) ^ 1 RESIDENCY STATUS CDResident ID]Nonresident c. EDPart-year resident-Dates: a b. Page To © . STATE CAMPAIGN FUND ...\ P A R T II I © Do you (or your spouse ifjoint return) want $2 of your ,uxos to go to this fund? (Your decision will not Increase your tax or roduco your refund.) YES NO a. You f.~l 1 1 b. Your Spouse ... O CD 1. Item a 1 3d. Spouse's occupation | 4b. School district nnmo FILING STATUS a. 0 Singla 1 Spouse's social security number • Page D 1. Item 1, Item 2 8. EXEMPTIONS Pngo 1. Item 4 a. Your allowable ED Married, filing jointly c. ED Marriod. filing separately — Enter spouse's social security Federal exemptions .... b. b. Spocial oxomptlon for hemiplegics, paraplegics number on line 3c and enter spouse's name hero: and quadriplegics...... c TOTAL EXEMPTIONS Add linos 0(a) and 8(b) © INCOME AND ADJUSTM ENTS Page ______ 1. Item 6 9. Adjusted gross income which should be reported on Federal Form 1040, line 31, or 1040A, line 11, {Attach copies of any federal jrhedules that Page 1. Item 7 Page 1. Item 0 Page 1. Item 9 15. Taxable income. Subtract tine 14 from line 13 ....................................................................................................... 15 Page 1. Item 10 16. Tax. Multiply line 15 by 4.6% (.046) ............................................................................................................................ 16 Page indicate a loss from or adjustment to gross income. See page 7 of instructions) 10. Additions to adjusted gross income (from page 2, line 37 of this form) .................................... © ....................................................... ® 11. Total. Add lines 9 and 10 ............................................................................................................................................... 11 12. Subtractions from adjusted gross income (from page 2, line 45 of this form) ......................................... © 13. Income subject to tax. Subtract line 12 from line 11 ............................................................................................ 13 14. Exemption allowance. Multiply line 8 by $1500.00. Part-year and nonresidentsenter line 59 © CREDITS (SEE INSTRUCTIONS — PA GES 8 THROUGH 10) Page 2, item 13 CREDIT AMOUNT PAID 17. Income tax paid to Michigan c itie s ........................... ©. 17a.. 16. Contributions to Michigan colleges, universities and public libraries......................................................... 19. Income tax paid to another state 180 - (attach copy of return) .................................................. 19. Total credits. Add lines 17a, 18a and 19a ................................................................................................................. 20. Income tax. Subtract line 20 from line 16. If line 20 is greater than line 16 enter "NONE” ..................... © • PROPERTY TAX AND HOME HEATING CREDITS AND PAYMENTS 22. Property Tax and Home Heating Credit(s). AttachMI-1040CR-1. 2, 3.or 4 23. Farmland Preservation or Solar Energy Credit. AttachMM040CR-5 or6 24. Michigan tax withheld. Attach State copy of W-2 .... @. .. .. ’. @ .................................................. 25. Michigan estimated tax payments............................................................................ 26. 1970 overpayment credited to 1979 .................................................................................. 27. Add lines 22, 23, 24. 25 and 26 © . Page 1, Item Page 1. Item ©. ©. I .................................................................................................................................. 27. p B S S S M p TAX DUE STATE OR OVERPAYMENTS AMD CREDITS DUE YOU j 28. If line 27 is less than line 21, enter BALANCE OF TAX DUE STATE ................................................. PAY ► © . 29. II line 27 is greater than line 21. enter BALANCE DUE YOU ........................................................................... 29. 30. Amount of line 29 lo be PAID TO YOU ..................................................................................................................^ 0 31. Amount of line 29 lo be credited to 1980 ESTIMATED TAX ................► © -----------------------------1-------- . (OVER) Page 1. Item 14 263 PAGE 2, MI-1040 ADDITIONS TO INCOME PAR T VI 32. Gross interest and dividends from obligations issued by states other than Michigan or their political subdivisions ........................................................................................................ 32. . Page 2. Item 18 33. Capital gains (from MI-1040D) ...................................................................................................................... 33. . 34. Other gains (from MI-4797) .................................. 34. 35. Losses attributable to other states (see instructions, page 1 0 ).......................................................................... 35. . 36. Other (see instructions, page 10) Describe:_____ .............. ........................................... Page 2, item 10 37. . SUBTRACTIONS FROM INCOME 38. Income from U.S. government bonds and other U.S. obligations included in line9 ................................ 39. Military pay or military retirement benefits from U.S. Armed Forces includedin line 9 (attach W -2)... 38. 39. 40. Capital gains (from MI-1040D) .................................................................................................................................. 40. 41. Other gains (from MI-4797) ........................................................................................................................................ 41. 42. Income attributable to another state. Explain type and s o u rc e .__________ 42. 43. Retirement or pension benefits, included in line 9. Name of Payer:__________________ 44. Other (see instructions, page 11). D e s c r ib e :_________ 43. 44. 45. Total Subtractions. Add lines 38 through 44. Enter here and on page 1, line 45. B .- i- ~ - f 2, item .............................................................................. 36. . 37. Total additions. Add lines 32 through 36. Enter here andon page 1, line 10 P AR T VII. Page 12 ................................... Page 2. Item Pago 2. Item Page Page 19 119 _L19~ Item 2, tiem PART-YEAR AN D NONRESIDENT INCOME ALLOCATION (See page 11 of the instructions) TOTAL INCOME MICHIGAN INCOME INCOME FROM OTHER COLUMN A COLUMN B STATES-COLUMN C INCOME 46 Wages, salaries, tips, etc............. 47 Dividends________Exclusion ____ . Balance 48 Interest ............... ........................... 49. Business or farm income ............................................................ 50. Capital gains, from Federal schedules ...................................... 51. Income reported on Federal Schedule E ................................ 52. Other. (Descrlbo)_______________________________________ 53. Gross income. Add lines 46 thru 5 2 .......................................... ADJUSTMENTS TO INCOME 54. Enter total adjustments claimed on line 30 of your Federal income tax return. D e s c r ib e ._________________ ..__________________________ 55. Subtract line 54 from line 53. Amount in Column A should agree with line 9. Enter amount in Column B on line 57. Enter amount in Column C on line 42 ............................................... ' [ PART-YEAR AND NONRESIDENT EXEMPTION ALLOWANCE If you received gross Income not Included In line 9, or Income that can bo deducted on Ines 38 and 39, see page 12 of the Instructions before completing this schedule. 56. Multiply number of exemptions on line 8 by $1,500.00 ..................................................................................... 56. 57. Michigan source income: enter amount from line 55, Column B .................................................................. 57. 58. Divide line 57 by line 55, Column A, and enter percentage ............................................................................ 58 59. Multiply line 56 by line 50. Enter hero and on line 14................................... 59. K M R l DECLARATIONS — Sign below. If filing jo intly, both husband and w ife m ust sign. T his return Is d u e April 15, 1950, o r on th e 15th day of th e fourth m onth after th e c lo s e of your tax y ear. do er cm la at ri eonuo nf de ph en1 al fny pe j uw rl yedt a. tthisreturnisbasedonall It dt ea cc lh am re enu r a, lt or fep ju t hm ap tlt wr hic ht ay veoa kr n o gh e a tn sd ,ei sp te rn ue cy or ce tra nr dyc o eh te e.informationinthisreturn,and inf / Your Signature Dato Preparer's Signature. Business Name. Address end Identification Number ► Spouso'e Signature (it tiling Jointly. OOTH must sign oven if only one l^Tfe. Va. 'bC.k. p/©vj| l~\ ,»A-- o-Hor* ( A t A - "i.cJb.. 271 1979 M ichigan Department o t Treasury Office Use GENERAL PROPERTY TAX CREDIT AND HOME HEATING CREDIT MI-1040 CR-4 JT'V'yi■ . .r1 J IDENTIFICATION .t - n ■>*$ ,. (PIggso typo or p rin t ) | 0 First n a m o fj in itia l fit jo in t re tu rn , use fir s t n a m e s & in itia ls o l b o th ) j A d d re s s (nu m b er an d s tre e t o r ru ra l rou te) La st na m e 11 1 0 Y o ur so c ia l s e c u rity n u m be r 0 S p o u s e 's so c ia l se c u rity n u m be r ^ Y o u r a llo w a b le M ic h ig a n e x e m p tio n s . .. S c h o o l d is tric t co d e 1 City, to w n o r p o s t o ffic e , a n d S ta te ZIP C ode 1 5a. 5b. S c h o o l d is tr ic t n a m e (s e o In s tr u c tio n s , p a g e 17) . J i SCHEDULE OF HOUSEHOLD INCOME - INCLUDE ALL INCOME OF YOU AND YOUR SPOUSE 6. 7. Pnge 2. I t pm 21 Pnge 2. Ite m 22 Ite m Ite m 23 24 8. P ng e 2. Annuity and pension benefits — Name of payer. 9. P ng e 2. Net form Income (see instructions, poge 1 7 ) ___ 10. P ng e 2. Ite m Atl capital galn9 loss capital losses (see Instructions, page 1 7 ) ............. Other taxable income and adjustments fse© Instr.. page 17). Describe: . 1 t. P o g e 2. Ite m Social security, supplemental security Income (SSI) or railroad retirement benefits 23 26 ............................... Pnge 2. Item 27 16. Other nontaxable income (see instructions, page 18). D es c rib e :____________________________________ _ 16. 17. Workers’ compensation and unemployment insurance benefits................................................................................ 1 7 , P o o r 2. Item 29 Alimony and child support .............................................................................................................................................. PT4T 15. Veterans' disability compensation and pension benefits .......................................................................................... 15. 18. ADC (attach Department of Social Services Annual Statement)......... ...................................................................... Q 19. All other public assistance payments (see instr. page 18). Describe:__________________________________ © . 20. SUBTOTAL — Add lines 6 through 19 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21. Insurance premiums you paid for medical care of yourself and family 20. .............................................................. 21. 22. HOUSEHOLD INCOME — Subtract line 21 from line 20 .......................................................................................... PART III Page 2. Item 30 Page 2. Item 3! I PROPERTY TAX CREDIT 23. Property taxes on your home for 1979, or amount from lino 39. 44 or 46 ......... 24. Rent paid from tine 42 ..................................................................................................... © . rage 2- Ilem 3 2 ____ 25. Multiply line 24 by 17% (.17) .......................................................................................................................................... 25. 26. Total. Add lines 23 and 25 .............................................................................................................................................. 26. 27. Amount not refundable. Multiply line 22 by 3.5% (.035) ....................................................................................... 27. 28. Subtract line 27 from line 26. If line 27 Is greater than line 26, enter NONE................................................. 28. 29. PROPERTY TAX CREDIT. Multiply line 28 by 60% (.60). (Cannot exceed $1,200.00.) Enter here and also on line 33 If you do not claim a Home Heating C re d it.................................................................................. © . |~HOME HEATING CREDIT AND CREDIT SUMMARY | 30. Standard allowance from Table 4, page 31. Multiply household income (line 22) by 3.5% (.035). Part-year residents, seeInstructions,page 1 6 ) ............ 16. Part-year residents, see Instructions,page1 6 ........................... 30. 31. 32. HOME HEATING CREDIT. Subtract line 31 from line 30............................................................................................ © . 33. SUMMARY: TOTAL CREDIT. Add lines 29 and 32. Enter here and also on line 22 of Form MI-1040 If you are required to file a Michigan Income Tax retu rn........................................................................................ @ (OVER) -^cn4\. IjAC*— (4« Ct4 OAAJ=L OL*UOwi>C*flE» 2,*7 <5VS 'kc.'fcs . L’l 'SpOLcJL. O _ 272 M ichigan D e p a rtm e n t o l T rea su ry Otfice^Use_ AMENDED MICHIGAN INDIVIDUAL INCOME TAX RETURN MI-1040 X (Rev. 12/84) Ib ::i (Issued under authority of Act 281, P.A. 1967) r mMmmmmmmmmmm ENTER CALENDAR YEAR OR ENDING DATE OF FISCAL YEAR (MO./DAY/YR.) OF THIS RETURN I PART I ID E N T IF IC A T IO N (P le a s e type o r p rin t) ? rMSI "amp A initial (if inint irliirn use first names A initials of both) Page I. Item ?J Mnmp adrtipss I" Page i post ofticp and Slate I, item 1 Fage 1. Item 2 4 Spouses sorc.il security no IM P O R T A N T : P le a s e a n s w e r all qu estio ns, fill In a pplicab le Item s, and exp la in c h a n g e s on P age 2. S rnlrr n (NOM and address on o rannni change fir On Original Return L.1 Resident th is Return J •hanging from v passed fix fit Pnge Pnge Resident Poge 3. Item 1 I I Non-resident L I Part-year resident from I. Item 3 [J 3. Item 2 On O riginal R e tu rn ....... Non-resident __ Single tiling II On This Return .. . [.. 1 Part-year resident, from Married filing sppa'atHv 8 ’n Exemptions: IN C O M E , A D D IT IO N S , and D E D U C TIO N S (see 11. Total income. Add lines 9 and 1 0 ............................................................................. 11 12. Subtractions from adjusted gross income . .................................................... 12 CO'tprt d orron sc - e specfrr ms! ) 9. Adjusted gross incom e Explain changes on page 2 9 _ r n? ° 10. Additions to adjusted gross in c o m e ....................................................................... 10. p nge 13. Balance. Subtract line 12 from line 1 1 ................................................................... C 3. Item 3 H Nf‘t CMarigi* A Ar. ortgin.111' repotted or as (increase or a d ju s te d !. Item On This Return Pnge P°0G »• Itom 4 PART It Page On Ofioinnl Return on Page ?) 3‘ ,t,?m a. item n Pnge I, Ito rr Pnge 1. M om I) ::::::::zj Page !, Item 3 p° g c 3. item PQ9e 3, Hem o Page 3. Item 9 Page 3. Item 10 Pnge I. Item 12 Page 3, Item 1! Pnge 1. Item 13 Poge 3. Item 12 13 14. Exemption allowance. M ultiply number of exem ptions by applicable rate (see instructions) ............................................................................................... 14 15. Taxable income Subtract line 14 from line 13 .................................................... 15 16. Tax. M ultiply line 15 by tax rate (see in s tru c tio n s ).............................................. 16. PART III l. Item 10 Pnge I, Item II N O N R E F U N D A B L F C RED ITS 17. City income tax credit .............................................................................................. 17 18. College contribution credit ..................................................................................... 18 19. Credit for tax paid to another s ta te ....................................................................... 20. Total non-refundable credits. Add lines 17 through 1 9 ...................................... 20 21. Pago 19 Balance. Subtract line 20 from line 16 (If line 20 is greater than lino 16 enter " N O N E ")..................................................................................... 21. PART IV 22. ! C R E D IT S A N D PA YM E N T S Homestead Property Tax Credit and Home Heating Credit (1st yr. 197 8 )................................................................................................................ 22 Farmland Preservation Tax Credit ...........................................................................23 Solar Energy Credit ................................................................................................... 24 Michigan income tax withheld . . . ,2’. .*t.er?\ 25 Michigan estimated tax p a y m e n ts ............................................................... Overpayments from prior year claimed on o riginal re tu r n ..................... Amount paid w ith original return, plus additional tax paid after filin g . 28.. Total credits and payments. Add lines 22 through 20 of colum n C . . . . 29 . n R E F U N D OR B ALAN C E DUE 30. Refund, if any. shown on original re tu rn ............................................................................................................................................ 30 Pnge 3. Item 13 31. 32. Enter the difference between lines 29 and 30 (see in s tru c tio n s ).................................................................................................. 31 (W ith in te r ,is ! o l b f 1“ c per m o n th 32. If line 21 , colum n C, is greater than line 31 . enter BALANCE DUE. Pay in full from the dam ia* was ongnntiy dm-i Page 1. Item 14 33. If line 21, column C, is less than line 31, enter REFUND to be received !nq ^rta y i^irM date leium iTTecLiS )....... (OVER) 33. 273 f E 34. B l E X E M P T IO N S Show exemptions claimed on your original return. A Yourself □ 3 Spouse [1 35. I [J I R09UlM ( l H E n te r N u m b e r 65 o f boxos vo' k checked ______ Enter first names of dependent children who lived with you. and their social security numbers. 36. Enter first names of dependent children who did not live with you. If pre*1905 agreement, check here $ Q 37. N um ber r N u m l i. . . .. y T E n tu r ^ Enter full names of other dependents and social security numbers, if any. N u iM b r r 30. Total number of exemptions claimed on your original return. 39. Other dependents not claimed on original return. N AM F I n ip t In ju re \ m th e Ia s i to iu n m to itie n q itt to r e a c h nam e b s lp d (it m orp s p a c e is n p pd prt a t i. ir h s c h e d u le ) 40. i ip p o il Explain change in number of dependents. Pnfjo 3. Item m i 41. II 100".. M nno E X P L A N A T IO N O F C H A N G E S Explain changes to Income. Deductions, and Credits. Show computations in detail and attach applicable schedules. If Taxpayer Used MI-1040X To File Initial Return: If you are analyzing a MI-1040X with zeros in Colmn A, attenpt to determine the information for Items 18 and 19 on page 1 of the DCD from information listed here. If This Is A "True" Amended Return; If you are analyzing a MI-1040X with some nonzero nutters in Colmn A, the information requested for Itans 15 through 21 on page 3 of the DCD must bo inferred from reading this part of the form. Two errors may be coded for changes in each of the following categories. Adjusted Gross Income Additions Subtractions If more than two errors exist in one of these areas, code the two most significant errors on the DCD. other errors. j D E C L A R A T IO N S — Sign belo w . If filing jointly, both husband and w ife must sign. 1 d e c la re u n d e r p e n a lly o f p e rju ry th a t th e in fo r m a tio n in this return, a n d a tta c h m e n ts , Y o ur S ig n a ln re is t r u e , correct and co m p le te D ale ► 1 d e c la r e u n d e r p e n a lt y o f p e r ju r y th a t th is in fo rm a tio n of w h ic h 1 have any k n o w le d g e P ip p a rp f s S ig n a tu re Tnge S p o u s e s S ig n a lu rp III filin g jo in tly UOTH m us t s ig n even it o n ly o n e ha d in c o m e ! ► Ignore any B u sin e ss N am e 2. Item 17a return is h a s e d o n A d d re ss a n d Id e n tilir a l'O r' N um oer BIBLIOGRAPHY BIBLIOGRAPHY Abdel-khalik, A. R. and B. B. Ajinkya, Empirical Research in Accounting: A Methodological Viewpoint (Sarasota, Florida: The American Accounting Association, 1979). Allingham, M. and A. Sandmo, "Income Tax Evasion: A Theoretical Analysis," Journal of Public Economics (November 1972), pp. 323-338. American Bar Association Commission on Taxpayer Commpliance, "Report and Recommendations" (Chicago, Illinois: American Bar Foundation, 1987). Benjamini, Y. and S. Maital, "Optimal Tax Evasion and Optimal Tax Evasion Policy: Behavioral Aspects," in the Economics of the Shadow Economy, edited by W. Gaertner and A. Wenig. (Berlin: SpringerVeri ag, 1985), pp. 245-264. Chang, 0., Tax Avoidance: A Prospect Theory Perspective. Ph.D. Dissertation, University of 111inois, 1984. Unpublished Clotfelter, C., "Tax Evasion and Tax Rates: An Analysis of Individual Returns," The Review of Economics and Statistics (August 1983), pp. 363-373. Cowel1, F., "Tax Evasion with Labour Income," Journal of Public Economics (February 1985), pp. 19-34. Cox, D., "The Income Tax and the Underground Economy," National Tax Journal (September 1984), pp. 283-288. Dean, P., T. Keenan and F. Kenney, "Taxpayers' Attitudes to Income Tax Evasion: An Empirical Study," British Tax Review 1 (1980), pp. 28-44. Denzin, N., Sociological Methods: Inc., 1978). A Sourcebook, (New York: McGraw-Hill, Department of the Treasury, Statistics of Income Bulletin 4 (3) (Winter 1984-85). Ekstrand, L., "Factors Affecting Compliance: Focus Group and Survey Results." 1980 proceedings of the 73rd Annual Conference on Taxation, National Tax Association-Tax Institute of America (November, 1980). 274 275 Farrington D. and R. Kidd, "Is Financial Dishonesty a Rational Decision?" British Journal of Social and Clinical Psychology 16 (1977), pp. 139-146. Feige, E. L., "How Big is the Irregular Economy," Challenge. November/December, 1979. Festinger, L., A Theory of Cognitive Dissonance. (Evanston: Peterson, 1957). Row, Frank, M. and D. Dekeyser-Meulders, "A Tax Discrepancy Coefficient Resulting from Tax Evasion or Tax Expenditures," Journal of Public Economics (August 1977), pp. 67-78. Friedland, N., "A Note on Tax Evasion as a Function of the Qual ity of Information About the Magnitude and Credibil ity of Threatened Fines: Some Preliminary Research," Journal of A p p I ied Psychology (February 1982), pp. 54-59. Friedland, N., S. Maital, and A. Rutenberg, "A Simulation Study of Income Tax Evasion," Journal of Public Economics (August 1978), pp. 107-116. Furnham, A . , "The Protestant Work Ethic, Human Values, and Attitudes Towards Taxation," Journal of Economic Psychology 3 (1983), pp. 113-128. Goddeeris, J. H., S. W. Martin, and J. C. Young, "Characteristics of the Individual Michigan Amnesty Participant," Michigan Department of Treasury, Lansing, Michigan (1988). Graetz, M. J. and L. L. Wilde, "The Economics of Tax Compliance: and Fantasy," National Tax Journal 38 (1985), pp. 355-364. Fact Grasmick, H., N. Finley, and D. Glaser, "Labor Force Participation, SexRole Attitudes, and Female Crime: Evidence from a Survey of Adults," Social Science Quarterly 65 (1984), pp. 703-718. Grasmick, H., and W. Scott, "Tax Evasion and Mechanisms of Social Control: A Comparison with Grand and Petty Theft," Journal of Economic Psychology 2 (1982), pp. 213-230. Groenland, E. and G. van Veldhoven, "Tax Evasion Behavior: A Psychological Framework," Journal of Economic Psychology 3 (1983), pp. 129-144. Groves, H., "Empirical Studies of Income-Tax Compliance," National Tax Journal (December 1958), p. 291-301. Gutmann P., "The Subterranean Economy," Financial Analysts Journal (November-December 1977), pp. 26, 27, 34. 276 Hotaling, A. and D. Arnold, "The Underground Economy," Massachusetts CPA Review (May-June 1981), pp. 8-14. Internal Revenue Service, "Estimates of Income Unreported on Individual Income Tax Returns," Washington, D.C., September, 1979. Internal Revenue Service, "Income Tax Compliance Research," Washington, D.C., July, 1983. Internal Revenue Service, A Dictionary of Compliance Factors. Office of the Assistant Commissioner Research Division, undated. Jackson, B. and S. Jones, "Salience of Tax Evasion Penalties Versus Detection Risk," -Journal of the American Taxation Association (Spring 1985), pp. 7-17. Jackson, B. and V. Mill iron, "Tax Compliance Research: Findings, Problems, and Prospects," Journal of Accounting Literature (1986). Kahneman, D., P. Slovic and A. Tversky, Judgement Under Uncertainty: Heuristics and Biases. (Cambridge University Press, 1982). Kahneman, D. and A. Tversky, "Prospect Theory: An Analysis of Decision Under Risk," Econometrica (March 1979) pp. 263-291. Kahneman, D. and A. Tversky, "The Psychology of Preferences," Scientific American (January, 1982), pp. 100-173. Kahneman, D. and A. Tversky, "Choices Values, and Frames," American Psychologist. (April 1984), pp. 345-350. Kaplan, S. E. and P. M. J. Reckers, "A Study of Tax Evasion Judgements," National Tax Journal (March 1985), pp. 97-102. Kelejian, H. K., and W. E. Oates, Introduction to Econometrics (New York: Harper and Row, 1981). Kmenta, J . , Elements of Econometrics (New York: Company, 1986). Macmillan Publishing Kennedy, P., A Guide to Econometrics (Cambridge: The MIT Press, 1985). Kerlinger, F.,Foundations of Behavioral Research 2nd ed.(New York: Holt, Rinehart and Winston, Inc., 1973). Laver, M . , The Politics of Private Desires. (New York: 1981). Penguin Books, Leonard, H. B. and R. J. Zeckhauser, "Amnesty, Enforcement, and Tax Policy," in Tax Policy and the Economy. L. Summers, ed. (Cambridge: The MIT Press, 1986), pp. 55-85. 277 Long, J. E., and S. B. Caudill, "The Usage and Benefits of Paid Tax Return Preparation," National Tax Journal (March 1987), pp. 35-46. Madeo, S., A. Schepanski, and W. Uecker, "Modeling Judgments of Taxpayer Compliance," working paper, University of Iowa, January 1985. Mason, R. and L. Calvin, "A Study of Admitted Income Tax Evasion," Law and Society Review (Fall 1978), pp. 73-89. Mason, R. and L. Calvin, "Public Confidence and Admitted Tax Evasion," National Tax Journal (December 1984), pp. 489-496. Mason, R. and H. Lowry, "An Estimate of Income Tax Evasion in Oregon," Survey Research Center, Oregon State University, Corvallis, Oregon, 1981. Michigan Department of Treasury, R. C. Fisher, and J. H. Goddeeris, "Michigan Tax Amnesty Report," Lansing, Michigan (1987). Mikesell, J. L., "Amnesties for State Tax Evaders: The Nature of and Response to Recent Programs," National Tax Journal (September 1986), pp. 507-525. Mikesell, J. L., "Tax Amnesties as a Tool for Revenue Administration," State Government 57 (1984), pp. 114-117. Mill iron, V. C., "A Behavioral Study of the Meaning and Influence of Tax Complexity," Journal of Accounting Research. (Autumn 1985). Mill iron, V. C. and D. R. Toy, "Tax Compliance: An Investigation of Key Features," Journal of the American Taxation Association (Spring 1988), pp. 84-104. Mork, K., "Income Tax Evasion: 30 (1975), pp. 70-76. Some Empirical Evidence," Public Finance Parle, W. M . , and M. W. Hirlinger, "Evaluating the Use of Tax Amnesty by State Governments," Public Administration Review (May/June 1986), pp. 246-255. Pindyck, R. S. and D. L. Rubinfeld, Econometric Models and Economic Forecasts. (New York: McGraw-Hill, Inc., 1981). Reinganum, J. F. and L. L. Wilde, "Income Tax Compliance in a Principal Agent Framework," Journal of Public Economics (February 1985), pp. 1-18. Schmidt, P. and A. Witte, An Economic Analvsis of Crime and Justice: Theory. Methods, and Applications (Orlando, Florida: Academic Press, Inc., 1984). Schmolders, G., "Survey Research in Public Finance-A Behavioral Approach to Fiscal Theory," Public Finance 25 (1970), pp. 300-306. 278 Schwartz, R. and S. Orleans, "On Legal Sanctions," University of Chicago Law Review (Winter 1957), pp. 274-300. Scott, W. and H. Grasmick, "Deterrence and Income Tax Cheating: Testing Interaction Hypotheses in Utilitarian Theories," The Journal of Applied Behavioral Science (July-September 1981), pp. 395-408. Song, Y. and T. Yarbrough, "Tax Ethics and Taxpayer Attitudes: A Survey," Public Administration Review (Septemer-October 1978), pp. 442-452. Spicer, M. W., "CiviIization At A Discount: The Problem of Tax Evasion." National Tax Journal 39 (1986), pp. 13-20. Spicer, M. arid L. Becker, "Fiscal Inequality and Tax Evasion: An Experimental Approach," National Tax Journal (June 1980), pp. 171-175. Spicer M. and R. Hero, "Tax Evasion and Heuristics: A Research Note," Journal of Public Economics (February 1985), pp. 253-267. Spicer, M. and S. Lundstedt, "Understanding Tax Evasion," Public Finance 31 (1976), pp. 295-305. Srinivasan, T., "Tax Evasion: A Model," Journal of Public Economics (November 1973), pp. 339-346. Sutherland, E., White Collar Crime (New York: Dryden, 1949). Tanzi, Vito, "The Underground Economy in the United States: Estimates and Implications," Banca Nazionale del Lavoro Quarterly Review (December 1980), pp. 427-453. Tittle, C. R., Sanctions and Social Deviance: Deterrence. (New York: Praeger, 1980). The Question of Van den Doel, H., Democracy and Welfare Economics. (Cambridge: Cambridge University Press, 1978). Warneryd, K. E. and B. Walerud, "Taxes and Economic Behavior: Some Interview Data on Tax Evasion in Sweden," Journal of Economic Psychology (1982, Vol. 2), pp. 187-211. Westat, Inc., Individual Income Tax Compliance Factors Study Qualitative Research. Prepared for the Internal Revenue Service, February 4, 1980a. Westat, Inc., Self-Reported Tax Compliance: A Pilot Survey Report. Prepared for the Internal Revenue Service, March 21, 1980b. Witte, A. D. and D. F. Woodbury, "The Effect of Tax Laws and Tax Administration on Tax Compliance: The Case of the U.S. Individual Income Tax," National Tax Journal (September 1985), pp. 1-13. 279 Yankelovich, Skelly and White, Inc., Tax Attitudes Studv: Prepared for the Internal Revenue Service, 1984. Final Report. Yitzhaki, S., "A Note on Income Tax Evasion: A Theoretical Analysis," Journal of Public Economics (Vol. 3, 1974).