University of Southern California Dissertations and Theses

Detecting the effects social and business pressures on small California trucking firm tax compliance

(USC Thesis Other)

Detecting the effects social and business pressures on small California trucking firm tax compliance

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content NOTE TO USERS This reproduction is the best copy available. UMI Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. DETECTING THE EFFECTS OF SOCIAL AND BUSINESS PRESSURES ON SMALL CALIFORNIA TRUCKING FIRM TAX COMPLIANCE by Paul G. Haywood A Dissertation Presented to the FACULTY OF THE SCHOOL OF POLICY, PLANNING, AND DEVELOPMENT UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PUBLIC ADMINISTRATION December 2003 Copyright 2003 Paul G. Haywood Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UM I Number: 3180773 INFORMATION TO USERS The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleed-through, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion. UMI UMI Microform 3180773 Copyright 2005 by ProQuest Information and Learning Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. ProQuest Information and Learning Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, Ml 48106-1346 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UNIVERSITY OF SOUTHERN CALIFORNIA SCHOOL OF POLICY, PLANNING, AND DEVELOPMENT UNIVERSITY PARK LOS ANGELES, CALIFORNIA 90089 This dissertation, written by P a u l G. Haywood under the direction o f /z..i.s.. Dissertation Committee, and approved by all its members, has been presented to and accepted by the Faculty o f the School o f Policy, Planning, and Development, in partial fulfillment o f requirements fo r the degree o f Dean Date DISSERTATION COMMITTEE j O . a ...... airpers Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. TABLE OF CONTENTS LIST OF TABLES............................................................................................ v LIST OF FIGURES.......................................................................................... vii ABSTRACT .............................................................................................. viii Chapter I. INTRODUCTION............................................................................... 1 Statement of the Problem............................................................ I Importance of the Analysis.......................................................... 5 Purpose of the Analysis .............................................................. 6 Scope of the Analysis.................................................................. 9 Data Limitations.......................................................................... 12 IPaCkda.............................................................................. 12 Data from Other Non-IRS Sources...................................... 14 Demographic Sensitivity...................................................... 15 Arrangement of the Dissertation ................................................ 18 II. REVIEW OF THE LITERATURE..................................................... 21 Economic Perspective of Tax Compliance ................................ 21 The Audit Rate .................................................................... 24 Sociological Perspective of Tax Compliance.............................. 26 Tax Administration: Taxpayer Categories for Study.................. 29 Analysis Tools Designed to Detect Noncompliance .................. 32 The Taxpayer Compliance Measurement Program (TCMP) ............................................................ 32 Agent-Based Modeling................................................................ 37 Neural Networking...................................................................... 38 Theoretical Framework of Compliance Behavior ...................... 40 An Individual’s Standing in the World ...................................... 46 Environmental Influence on Behavior........................................ 49 1 1 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Human Creativity, New Ideas, and Social Networking.............. 52 The Implications ofTrust in Social Networks............................ 55 Diffusion of Innovation Theory; Evasion as an Innovation 58 in. RESEARCH HYPOTHESES AND METHODS.............................. 63 Research Hypotheses.................................................................. 64 Hypothesis 1 ........................................................................ 64 Hypothesis 2 ........................................................................ 65 Research Design.......................................................................... 66 Dependent Variable .................................................................... 67 Independent Variables ..................................... 69 Research Procedures.................................................................... 74 Data Analysis.............................................................................. 76 IV. RESEARCH FINDINGS.................................................................... 79 Descriptive Analysis.................................................................... 79 Issues with Nonparametric Statistical Techniques...................... 83 Truck Ownership Lifecycle Costs and Regulatory Requirements........................................................................ 91 Other Fixed Costs........................................................................ 93 Legal and Professional Costs .............................................. 94 Geographic Analysis .................................................................. 98 Geographic-Based Dataset .................................................. 101 Testing of the Individual Compliance Clusters .......................... I l l Cluster Position and Descriptive City (External) Data................ 125 Temporal Considerations............................................................ 126 V. CONCLUSIONS, DISCUSSION, AND SUGGESTIONS FOR FUTURE RESEARCH...................................................................... 128 Conclusions ................................................................................ 128 Hypothesis 1 ................................................................................ 129 Hypothesis 2 ................................................................................ 133 Discussion of the Results and Sample Challenge ...................... 135 Suggestions for Future Research ................................................ 137 SELECTED BIBLIOGRAPHY........................................................................ 143 1 1 1 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. APPENDICES 1. Examples of Transaction Code Definitions................................. 157 2. Final Cluster Notes....................................................................... 159 3. Detailed Cluster N otes................................................................. 169 4. Map of California (Route 99 Corridor Highlighted) ................... 215 5. TY 1999 Compton (Example) Dendrogram Using Average Linkage (Between Groups).................................................. 216 IV Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. LIST OF TABLES 1. Business Social Pressure Variables.......................................................... 72 2. Environmental Factors.............................................................................. 74 3. Reported Income on Form 1040 .............................................................. 80 4. Total Income of Owner/Operators—Modified Sample............................ 85 5. Car and Truck Expenses—Modified Sample .......................................... 89 6. Annual Costs of Repair and Maintenance............................................... 92 7. Depreciation—Modified Sample.............................................................. 94 8. Legal and Professional Expenses............................................................. 95 9. Tax Year 1999 External and Internal Data for Test Cities ..................... 105 10. Tax Year 2000 External and Internal Data for Test Cities ..................... 107 11. Tax Year 2001 External and Internal Data for Test Cities ..................... 109 12. Primary Clusters ..................................................................................... 112 13. Clusters of Interest Based on Regression Analysis.................................. 116 14. Correlations for Fresno Cluster 2 TY 2000 .............................................. 119 15. Residuals for Fresno Cluster 2 TY 2000 .................................................. 121 16. Correlations for Compton Cluster 7 TY 2000 .......................................... 122 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 17. Residuals for Compton Cluster 7 TY 2000 ............................................ 124 18. The Triad of Clusters.............................................................................. 132 VI Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. LIST OF FIGURES 1. TheoryofReasoned Action, Fishbone Diagram..................................... 51 2. Independent Variables and Their Impact on Compliance........................ 70 3. TY 1999 Total Income—Modified Sample............................................. 86 4. TY 2000 Total Income—Modified Sample............................................. 87 5. TY 2001 Total Income—Modified Sample............................................. 88 6. TY 1999 Car and Truck Expenses........................................................... 90 7. TY 1999 Legal Expenses......................................................................... 96 8. Generic Representation of a Social Network........................................... 140 V ll Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ABSTRACT In the past, large-scale tax compliance studies characterized taxpayers utilizing internalized categories that are defined by federal tax law. Taxpayers were lumped together based on income brackets, filer type (e.g., small business, partnership, etc.), or tax scheme. Unfortunately, this approach neither takes into account how taxpayers view themselves as human beings nor considers the role of socialization in tax compliance behavior. This research used existing self-reported data from taxpayers via the Internal Revenue Service Form 1040, including Schedules A and C, in an effort to demonstrate whether a detectable relationship exists between known compliance indicators and social and business pressures as reported on the forms. Moreover, it explored the role of social pressure as competitive behavior that could possibly play a key role in deciding not to comply with one’s tax obligations. Another important concept was introduced in this work, the characterization of noncompliance as an innovation. Diffusion of innovation theory was used to offer one possible explanation of how noncompliant behavior, treated as a new idea, may cascade through communities and other social settings. Taxpayers risk being viu Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. noncompliant based on the perceived viability and success of the approach or innovation of noncompliance. A sample of California-based long and short haul truckers was used for the analysis. This group was chosen because the occupation, in many instances, removes the taxpayer from traditional community-based social settings and presents the greatest challenge to the fundamental hypothesis of this work. While only one sample group of taxpayers displayed substantial correlation between the social/business pressure variables and compliance, the research demonstrated that such modeling is viable. More importantly, it opens the door to perfecting this type of research to further explore tax compliance from a perspective of behavioral factors. IX Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER I INTRODUCTION Statement of the Problem Do citizens identify and define themselves as taxpayers or as doctors, lawyers, bankers, and fishermen? There is little substantive work in the taxation literature on identifying compliance vulnerabilities as shaped by social and business pressures unique to a specific occupation or industry. Davis, Hecht, and Perkins (2000) acknowledged that social norms and personal acquaintances influence tax compliance behavior. Conversely, they lament that there is no theoretical foundation that can be used as a basis for understanding the effects of these influences on compliance. According to Carroll (1992) “Taxpaying is also a social process: information, experiences, and strategies are passed around occupational and friendship networks” (p. 45). These networks may be difficult to model, but there could be consistency based on a lifetime of shared experiences (Kirchler, Maciejovsky, & Schneider, 2001). In addition, Carroll (1989, p. 15) proposes that the act of taxpaying from the citizen’s standpoint may depend on his/her experience, profession and knowledge. Research must consider taxpayers as active and dynamic individuals interacting within an environment that is defined by their own human emotions and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. self-identification, not by a generic characterization invented by economists to aggregate behaviors at the highest level. A fundamental realization of sorts has taken place; data at the highest aggregate level does little to assist researchers in predicting either future behavior or motivation. Woodward’s (2000) recent book about the Federal Reserve’s Alan Greenspan describes his (Greenspan) own epiphany on data: Over the past few years he had asked Research and Statistics to work on special studies that “disaggregated” data and broke it into finer pieces. Government statistics tended to come in overall total numbers that lumped together quite different segments of the economy—industrial and non industrial production, for example—eradicating some important distinctions in the process.. . . Greenspan wanted the numbers so he could get his hands around different kinds of businesses. (Washingtonpost.com, 2000) Data aggregated by bureaucratic or legal definition of taxpayer types does not serve to provide the basis for understanding the depth of issues that taxpayers face as people. Even previous attempts to segregate taxpayers by industries such as casinos, real estate, food and beverage, etc. does not fully take into account the localized norms of society within which people live (General Accounting Office [GAO], GAO/GGD-96-72,1996, p. 26). The trucking industry is an example in which there is little study or understanding of the taxpayer’s motivation to comply with their tax obligations. According to the University of Michigan’s Transportation study group, there is almost a mystique regarding truckers’ attitudes toward tax compliance (Belman, Monaco, & Brooks, 1998). The nature of the small trucking business environment is Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. an important consideration and an obvious draw to the lost aura of independence which seems to characterize the mystique of trucking. Truckers are unique in that, even when they work for a major motor carrier, in many cases they are considered independent contractors (Hardman, 1994). The truckers’ reality is best described by Hardman (1994): “The tractor is their place of business; the kitchen table is their home office. They have one tractor, which they normally drive themselves” (p. 5). The single owner/operator does not fit the traditional brick and mortar image of the small business owner. Stability, a necessary function of social normalization, may or may not be present among the trucker population given the dynamics of their work or the sometimes “old west” stylized outlaw mentality of their profession. That is, resistance to formal laws is strong and unlike most occupations is an integral part of their culture. The study of single owner/operator trucker represents an opportunity to understand a unique perspective on both the social norms within an occupation and the potential affects on tax compliance. Another aspect of tax compliance studies which is consistently overlooked is the person’s desire to compete with a friend’s, or neighbor’s perceived level of financial success. Are there shared pressures on taxpayers based on where they live as well as the nature of their occupation? If so, social pressure becomes an essential consideration in the study of compliance since people may seek to increase their incomes in any way possible to compete with their neighbors, including adoption of tax evasion tactics (Putnam, 1993). Consequently, how can researchers approach an Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. occupation where the open road/highway may very well be the primary social network medium in which these small business owners interact? This analysis explores the possibility that very basic human emotions, in combination with business pressures, not mechanistic decision making chains, influence a citizen to evade his or her tax obligations under the law. As with any act of defiance, human emotions such as pride, jealousy, and competitiveness may be the primary drivers in an individual’s decision to evade his or her tax obligation. How these emotions are forged, through time, may depend on a number of factors including community, socialization, and shared interests that are integral aspects of an individual’s occupation. In addition, this analysis attempts to identify spatial, geographically-based patterns of tax evasion in the form of clusters using techniques similar to those used by sociologists in law enforcement to model patterns of criminal activity within a community. Evasive behavior disperses from a central core of expertise and may be promoted and adopted as an innovation or a fad. This approach to modeling tax evasion differs from past studies since it does not focus on the negative connotations of evasion or the personal risks associated with getting caught. This analysis explores impacts of external and social variables on compliance via a specific group of taxpayers: California-based truckers. This is achieved by meeting three data-driven goals: Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1. Discuss, in depth, the sociological implications of tax evasion including how it may manifest itself starting from small groups to identifiable patterns of noncompliance; 2. Use easily obtainable owner/operator self-reported data that are associated with social and business pressures to identify potential with compliance; 3. Use this data, in concert with external demographics such as poverty rate, to demonstrate the existence of geographic clusters of noncompliance. Importance of the Analysis This research could serve to provoke a more advanced debate regarding how the 1RS collects, studies, and interprets data about taxpayers. From a tax administration standpoint, the study looks for linkages between a historically unconventional grouping of taxpayers and their relationship to the environment. Unlike conventional auditing, it is based on more passive surveillance using pre existing voluntary taxpayer-supplied data via individual tax returns. If a linkage is found between the person’s self-reported data and compliance, the analysis could provide tax administrators with a fundamental understanding of the environment in which evasion could flourish. The recent modernization of the 1RS as a result of Restructuring and Reform Act of 1998 (RRA98) institutionalized organizations within taxpayer focused operating divisions to perform “outreach” and provide pre-filing Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. services. Equipped with the results of this analysis and future follow-up work, tax administrators could potentially develop treatments that would preempt the spread of evasive behavior by limiting the growth of compliance clusters among groups. The ability to target compliance fads take on additional importance when one understands the relationship between the 1RS and state and local tax authorities. Through the IRS’s Data Exchange program, taxpayer data flows primarily, with some exceptions, in one direction, from the 1RS to the states. This dependency implies that what is good for the 1RS is good for all players throughout the intergovernmental chain. As with other crimes, the preventative costs of evasion have never been identified. Tax gap measures and enforcement statistics do not reflect the true costs considering sociological impacts that may not be quantifiable. This study provides a theoretical means to identify issues that impact the taxpayer’s decision to evade. If properly identified, these factors could prevent the proliferation of evasion across social networks. Finally, if a correlation is determined, this methodology, using modified independent variables, could be applied to other industry groupings of taxpayers to advance the understanding of how tax evasion grows. Purpose of the Analysis The purpose of this analysis is to learn whether external variables, including income and expenses in the context of geographic data, can be used in tax administration for identifying clusters of noncompliance by occupation. For tax 6 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. administration purposes, this offers the possibility of using pre-existing data without invasive auditing to identify clusters of noncompliance. The analysis does not represent a case study of the 1RS but explores how taxpayer noncompliance is detected within the context of tax administration. The states and some localities are codependent upon data and research provided by the 1RS to locate tax evaders. The analysis attempts to transcend 1RS and other tax authority traditionalism in taxpayer research by identifying compliance vulnerabilities as shaped by distinct normative pressures unique to specific occupations or professions. The analysis explores the relationship of taxpayers to their environments and how they may be influenced to behave towards the state sponsored requirement of taxation. At the same time, it does not try to underestimate the importance of self- interest in making the decision to evade, nor does it disregard the importance of the classic economic model which will be discussed later. More importantly, the analysis attempts to build an argument for considering a construct where social pressure to perform a negative behavior, such as tax evasion, is mellowed by social values that actually cherish mild outlaw group imagery. This is one of the reasons the trucking occupation was chosen for the analysis; the truckers’ unconventional view towards tax compliance. In the truckers’ world, civic virtue may embrace evasion as a virtue, something opposite to what may be considered the norm: Self-interest and civic virtue are not in direct competition with each other; rather, self-interest (that is personal self-interest) is likely to motivate behavior when people see themselves as individuals (in contrast with other individuals), while civic Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. virtue (what is good for the group collectively) is likely to motivate behavior when people see themselves as being members of (positively valued) social categories, in contrast to other (negatively valued) social categories. An appeal to civic virtue changes the psychological situation by situating the recipient in a wider, more inclusive category in a different social context. (Taylor, 2001, p. 5) In the case of truckers, the decision to evade may be reinforced as a positive behavior as opposed to what would be considered the norm. Civic virtue, a concept which is discussed later in this analysis, is usually considered as a positive aspect of societal relations. However, smaller groups, maintaining different societal views of right and wrong may celebrate an evasion tactic as a positive service, something to be shared and cherished among the group, akin to cleaning a section of the highway. Information regarding the possibility of getting caught may be feeding those attempting to evade. While it is doubtful that the average trucker has the time to peruse the vast collection of U.S. General Accounting Office (GAO) bluebook reports, one item of particular interest is the geographic difference in auditing performed by the 1RS. GAO reports that audit rates vary by geographic location which is attributed to differences in compliance, with higher noncompliance in the western and southwestern regions and lower noncompliance in the central and eastern regions (GAO, GAOGGD-99-19,1998). As a result of this report, California was chosen as the sample base for this analysis. From a tax administration standpoint, external data must be easily obtainable to facilitate duplication of the study for other occupations. Data must be available Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. from the taxpayer group’s geographic region without invasive auditing or requesting the affected state for a data extract (special programming may be required) for licensing, sales or other state collected data. This is especially true during times when states’ budgets are in deficit and governors and legislatures are struggling to finance basic services even though such efforts might provide increased revenue in the long run. Scope of the Analysis Trucking companies are generally categorized into two types: the less-than- truckload (LTL) and the full truckload (TL) (Belman, Monaco, & Brooks, 1997, Industry Affiliation, Equipment, and Load section). The less-than-truckload carriers tend to move freight that is smaller than the capacity of the tractor-frailer. In order for LTL companies to be profitable, they maintain terminals or collection points so that trucks can gather freight into a full truckload for shipment to another terminal. Some of the larger firms maintain nationwide networks (Belman et al., 1997). The most famous of the less-than-truckload are the U.S. Postal Service and UPS. Prior to 1980, many firms provided both LTL and TL services. After deregulation the trucking firms became divided rigidly along the two categories (Belman et al., 1997). Entry in LTL is difficult since the costs of startups for building the necessary infrastructures are prohibitive (Belman et al., 1997). Because most of these companies exceed $10 million gross adjusted income, they are classified as mid to large-size business by the 1RS. Moreover, they may be Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. participants in the 1RS’ coordinated examination program (CEP). For large corporations under CEP, senior 1RS revenue agents work on-site within the corporate accounting structure. Full truckload companies tend to be the smaller, individual, self-employed trucker. The trucker owns or leases his rig and either works independently or under contract for a larger firm. In some instances, they perform both services by contracting on a limited basis to the larger firm. The small owner- operator trucker will serve as the basis for this analysis. An important issue that should be brought forward is that owner-operators’ gross income is substantially higher than that of corporate employees, but it must be adjusted for operating expenses (Belman et al., 1997, Annual Income section). As the 1997 survey from the University of Michigan Trucking Industry Program implies, after deductions, the owner-operator’s pay slips are below those of union-employee counterparts: Some owner-operators approximated their net income before taxes with their after expense income reported to the 1RS. Trucking lore suggests that owner-operators have a liberal attitude toward the deductibility of personal expenses. This may result in the under-reporting of some income. (Annual Income section) Consequently, expenses represent a key set of variables for this analysis that is scrutinized from a business standpoint. What can be considered reasonable expenses for the owner/operator are compared to actual industry data for the annual operating costs of the large tractor-frailer. 10 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. One extraneous variable originally considered for this analysis was the impact of union membership on tax compliance. However, since the industry’s deregulation in 1980, the nature of union membership has changed. Union membership has declined sharply in the truckload sector versus LTL sectors. Belman et al. (1997) found in their survey of the industry that only 3% to 4% of owner-operators were union members. In addition, there has been a gradual decline in LTLs (Verhoogen, Burks, & Carpenter, 2002). The legendary power of the Teamsters Union and its impact on Union membership would certainly be an important variable if: (1) it pertained to a majority of the truckers in the sample and (2) the truckers could be identified as union members without invasive research. Based on these fimdamental limitations, it was decided to exclude the variable of union membership from the analysis. The scope of the analysis focuses on compliance reporting issues around small, privately owned TLs who are self-employed or may fimction as part-time contractors to a larger firm but still maintain ownership of their rigs. The primary source for 1RS data is data fields from the Form 1040 series including Schedules A and C, using the North American Industry Classification System (NAICS) Codes 484110 (short-haul) and 484120 (long-haul) Transportation. The NAIC has replaced the U.S. Standard Industrial Classification (SIC) system. The sample is limited to trucker’s reporting positive income (net before major deductions) ranging fi-om $0 to approximately $140,000. 11 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Data Limitations 1RS Data There are numerous limitations on the availability and quality of taxpayer data as well as the statistical techniques used for analysis. Data captured by 1RS tax forms is flawed in several areas. For example, not all taxpayers include their NAIC codes which assist in identifying their profession. There is no practical way to resolve this issue for the purposes of this analysis since it would require massive legal and tax administration changes. In addition, during the three tax years used for the analysis, accounts are continuously added and dropped as new businesses file while others declare bankruptcy, or simply close their doors. Therefore, the sample is not static; rather it reflects a mixture of active and inactive accounts. Because of the massive size of the 1RS Individual Master File (IMF) and Business Master File (BMF), at approximately 13 terabytes, where the primary taxpayer data is stored, there is only enough storage room for three years worth of tax data. Unless the taxpayer account is moved into the compliance stream (e.g., account is queued into 1RS collection or examination functions), taxpayer data from the first tax year of a three-year cycle drops off as the new tax year’s data is rolled over. Different 1RS data systems are used to track taxpayer activities usually based on functional activity such as collection or examination. These systems are primarily designed for inventory tracking and do not translate well into an effective and efficient basis for studying long-term tax compliance issues. Consequently, the only means to validate this analysis would be to 12 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. perform an audit on the identified accounts within the sample. It is easy to predict that, if an audit were performed by NAIC code, a public backlash would occur since it would be focusing on a specific occupation. This type of enforcement activity could violate moral tenets that ensure equal treatment under the law. Moreover, an audit, while serving as one form of validation for tracking compliance behavior by group, would violate the intent of this work since it would be both politically unviable and reactive to noncompliance that has already occurred. In addition, truckers who do not file a tax return or nonfilers are not captured in the sample and, for the purposes of this study, are not pursued since it would require a difficult process including: (1) Matching state tax records, sales receipts, and other localized data (truck driver licenses); (2) Securing current addresses; and (3) Determining if they indeed have tax obligations since they may not be receiving enough income to justify such scrutiny. The type of effort required to identify nonfilers has been undertaken by a cooperative data exchange project between the State of California’s Franchise Tax Board (FTB) and the 1RS. However, the FTB project was attempting to identify nonfilers and does not locate specific professions prevalent within the group (FTB, 2000, Non-filer Program: Identification, Notification, Compliance presentation). A fundamental weakness in any tax compliance study is the use of self-reported data. Self-reported expenses can only be validated through the audit process which presents the issue of burden. In the 13 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Federalist 35, Alexander Hamilton’s (n.d.) essays written to develop an effective system of public finance warned about taxpayer burden: There is no part of the administration of government that requires extensive information and a thorough knowledge of the principles of political economy, so much as the business of taxation. The man who understands these principles best will be least likely to resort to oppressive expedients, or to sacrifice any particular class of citizens to the procurement of revenue. It might be demonstrated that the most productive system of finance will always be the least burdensome, (para. 11) Hamilton was concerned about the social impact of burden, both economic and social, that a tax system places on society. The methodology expressed in this analysis is an attempt to identify the least burdensome and most economical means to understand taxpayer behavior and possibly provide a building block approach to measure compliance. The introduction of a burdensome audit follow-up process for data validation would be self-defeating. Based on this fundamental restriction, the only way to draw concrete conclusions from this type of data is to attempt to utilize contextual data from external sources. Data from Other Non-IRS Sources Another issue is that social indicators such as unemployment, average income and other Census type data are usually presented in the aggregate and, many times, are only available for medium to large cities. A limitation within the analysis is the identification of social impact variables that are localized. Hence, much Bureau of Labor Statistics (BLS) data is only available to distinct population sets (e.g., 14 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. population greater than 25,000). This presents a severe restriction on the ability to cluster populations using easily obtainable data. For example, data on fuel costs was considered and later dismissed since truckers may shop for the best price on fuel in a wildly fluctuating petroleum market. At the same time, the cost of #2 diesel fuel is only available for regions (north, central, and southern) for California including some select cities. There are other data source options regarding specific taxpayers (e.g., real estate data from Westlaw), but it is impractical to go through the various commercial and state databases to identify taxpayer groupings given account data is presented individually. Finally, while California may represent an abundance of data collection opportunities, data on the sub-state level for less developed states may be less available and reliable (Howell & Wells, 1998). Howell and Wells (1998) provide an example of fundamental issues with crime reporting statistics: The FBI indicates that as many as 30 percent of the potential “reporting agencies” in Mississippi do not file crime statistics data summaries through the Uniform Crime Reporting system each year, largely because this is a voluntary program. Consequently, the UCR Program has developed methods for interpolating the data temporally so that some of the data for these non-reporting counties are estimated, (p. 8) Given the limitations of the availability of good data, maintaining the simplicity of the variables and the dependability of their source data is critical. Demographic Sensitivity The relationship between the taxpayer and the tax authority is to some degree adversarial. As Prescott (1998) found: 15 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. For the purpose of defining the role of the lawyer as taxpayer representative, the profession takes the position that the tax compliance process involves a dispute between two adversaries—the taxpayer and the 1RS—and should be treated as an adversarial proceeding warranting partisan advocacy on the part of the taxpayer’s lawyer, (p. 24) The adversarial relationship is based both in the inherent requirements of reporting income and its impact on individual privacy. The 1RS is unique in that it is a regulatory and an administrative body, making it both tribunal and opponent (Prescott, 1998, p. 24). The 1RS must be cognizant of the public’s tolerance for invasive activities, especially data collection. The audit rate, which is normally around 1%, has been reduced based on the public’s limited patience with auditing, making its usefulness as a data collection tool questionable. Classic data collection exercises such as the IRS’s Taxpayer Compliance Measurement Program (TCMP) is a good example of the Service’s efforts to gather data for the purposes of forecasting and audit selection was essentially shut down by political pressure based on taxpayer burden. There are several concerns when approaching tax compliance detection in this manner. For example, by studying specific professional groupings of taxpayers, a risk is that the group could protest the activity as being discriminatory. This has occurred in the past when the 1RS has tried to study groups by specific occupation. The 1RS is acutely sensitive about studies that categorize taxpayers by something other than legal definition or issue. Another issue is that the study could actually cause groups within the sample population to behave in manners inconsistent within their normal behaviors 16 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (something akin to the Hawthorne effect) if they realized that their activities were being scrutinized. In the case of truckers, it is feasible that such studies could cause truckers to react in a defiant manner resulting in greater non-compliance. Based on this experience, 1RS studies of taxpayer behavior are extremely sensitive to the practice of using distinct demographic variables, especially those related to gender, ethnicity, and even occupation. For tax administration purposes, approaching taxpayers fi-om this standpoint has only been used in cases where the specific demographic variable has essentially been the issue. For example, there have been instances where an evasion tactic has been promoted as a legitimate tax credit or break. The 1RS recently had to deal with an onslaught of several thousand cases where taxpayers have claimed a credit for slavery reparations, although there is no authorizing legislation to support such claims. This practice prompted the 1RS to use a communication strategy that focused on working with predominantly Afiican- American leaders to spread the message that, indeed, there was no slavery reparation credit and filing for one would have negative consequences. Compliance studies have tended to maintain this sensitivity to demographic variables since their usefulness would be extremely limited for tax administration. There is little tolerance within Congress and the general public for the level of invasiveness that such studies could obtain. 17 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Arrangement of the Dissertation The dissertation is presented in five chapters including the preceding introductory piece that presents the issues related to non-invasive detection of noncompliance while remaining sensitive to the general public’s perception of the 1RS and the potential burden of tax administration. A discussion is built around the purpose, importance, and scope of the study which is to explore noncompliance from a behavioral standpoint using taxpayer self-reported data as the basis for a model construct. 1RS and other public data sources are presented and critiqued for usefidness because of the difficulties in storing and maintaining such large amounts of data. Chapter 2, the literature review, opens with the traditional economic and sociological perspectives from which tax compliance studies have been formulated in the past. The classic fear of getting caught versus adherence to social norms arguments are drawn out and brought into question. As a follow-up to these traditional arguments, the 1RS perspective on classifying taxpayers is presented. The analysis also explores past large-scale efforts to detect non-compliance via the massive Taxpayer Compliance Measurement Program (TCMP) which served as the basis for macro-economic measures of compliance such as the vaunted tax gap. There is also a discussion regarding newer techniques for modeling and ultimately detecting noncompliant behavior, including agent-based modeling and neural networking that are presently being contracted by the 1RS. The literature review 18 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. moves toward examining modem psychological theory, including social networking and group behavior that forms the basis for a paradigm shift. This different way of thinking treats tax evasion not as a crime but as an innovation that diffuses within market-like strata. Chapter 3 walks through the steps for data testing in an effort to identify a viable methodology for data reduction and recognizing clusters of similar taxpayers. The variables identified in the sample are not normally distributed. Therefore, the numbers of cases are reduced to ensure that only owner/operators remain within the sample. This is achieved by ensuring that each case contains values fi'om a set of mandatory variables including car and truck expenses as home mortgage. This iterative and tedious approach was applied because normal statistical techniques did not offer the necessary reduction capabilities to identify similar cases. Two groupings of communities were chosen that are connected by major highways which potentially could serve as a diffusion medium. Hierarchical cluster analysis was applied to the remaining cases in the sample. Finally, linear regression analysis was applied to the clusters searching for correlation between independent variables and the dependent variable, known compliance. Chapter 4 presents the research findings from the initial descriptive analysis of the larger dataset to the observations from the cluster analysis and linear regression. The data is integrated into several tables that are presented throughout the chapter, providing perspective in relation to how communities and finally, clusters, compare to 19 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. each other. The chapter also provides a detailed description of the techniques used to normalize the data. Finally, Chapter 5 presents conclusions, discussion, and suggestions for future research. This chapter builds upon the data presented in Chapter 4, identifying the fundamental flaws in the analysis and how this particular approach could be perfected. This includes the idea of repeating the analysis on a career group that is traditionally more community based. 20 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER II REVIEW OF THE LITERATURE Economic Perspective of Tax Compliance The understanding of taxpayer behavior and compliance is promulgated primarily from the study of economics and sociology. There are two recognized motivations for complying with one’s tax obligations: the fear of getting caught cheating and the duty to obey (Scholtz & Finney, 1995). From a public policy standpoint, almost every treatise on tax compliance begins with a discussion of Allingham and Sandmo’s (1972) work (Yitzhaki, 1994) which forms the basis of the economic perspective: the taxpayer as a rational actor. In effect, taxpayers make choices based on imperfect information (driven by an intrinsic motivation to maximize income) regarding how much income to report on their tax returns. The taxpayer essentially works through a standard expected utility maximization problem to determine if conditions might require the option of evasion and if this option is tenable (French, 1986). A taxpayer’s propensity to pay his or her taxes or to evade is based on a decision process that involves the individual analyzing the risk consequences of his/her behavior (Carroll, 1992, p. 45). The amount of reported income is a function of the taxpayer’s individual income, the marginal tax rate, the audit rate, and the 21 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. penalty rate which is contingent upon the individual’s risk tolerance. The taxpayer’s decision can be based on a multitude of considerations such as record keeping, tax planning, opportunities to optimize itemization for small businesses, and other exogenous variables (Kirchler et al., 2001, p. 3; Carroll, 1992, p. 44). Within this framework, the taxpayer must decide the trade-off of underreporting his true tax obligation versus the risks of audit and ensuing penalties. Individuals have the opportunity to cheat because the economy creates varied levels of incentives for differently situated taxpayers. The problem with this assumption is that individual taxpayers, small businesses, and even some large businesses may not behave rationally. For example, consider the recent accounting scandals at Enron and WorldCom; can this be considered rational behavior? More probable is that they will act rationally within the context of their individual environments. Andreoni and Feinstein (1998) cite the temporal nature of the tax evasion decision, recognizing the fact that the penalties and interest for getting caught evading is a future event compared to the immediate nature of tax evasion. Because of uncertainty, taxpayer’s income fluctuates along with the shallow price of income. This leads Andreoni and Feinstein (1998) to surmise that people may “borrow” from the 1RS. Because of relatively high interest and penalties that are attached to underreporting of income, in Slemrod’s (1989) words, the 1RS becomes a “loan shark” (p. 14). While the goal of a taxpayer might be the repayment of the loan during better times, the reality is that, in many instances, the revenue is lost forever. 22 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In considering the effectiveness of deterrence (e.g., penalties, interest) on tax compliance behavior, Autunes and Hunt (1973) identify two types of deterrence: (1) Special—or the specific deterrence of a given individual and (2) General—the general reduction of crime [evasion] because of the inhibitory effect of sanctions on the general population. Unfortunately, these two types of deterrence are quite different. For example, it has been shown that deterrence actions have little impact on repeat offenders who have experienced criminal punishment. On the other hand, they argue that this example only casts doubt on special deterrence but says little about the effects of general deterrence. Conclusions regarding the effectiveness of penalty sanctions on compliance are mixed. Games and Englebrecht (1995) concluded that there is limited evidence that higher penalties produce higher compliance as opposed to less severe penalties. Aim (1992), however, concluded that increased penalties did indeed positively impact compliance. Games and Englebrecht noted a major issue with these studies that the proposed penalty was as high as 1500% of the norm. Moreover, while the penalty used in this analysis was politically untenable, it did not take into account the possibility that certain groupings of taxpayers could react differently. Adding to the absurdity of untenable penalties, consider Kopczuk Wojciech and Joel Slemrod’s (2001) “Dying to Save Taxes: Evidence from Estate Tax Retums on the Death Elasticity.” In this study, Wojciech and Slemrod attempted to analyze U.S. Federal tax retums to explore whether the timing of one’s death is correlated to its tax consequences. This effort resulted in the authors winning the 23 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2001 Ig Nobel Prize in Economics, a less than stellar honor. One obvious lesson that can be taken from these papers is that a great divide exists between academic exercises and realistic tax administration policy. The Audit Rate A major aspect of the rational actor concept is consideration of the decision making process. The decision to pay or evade one’s tax obligations is based on several stimuli including possible exposure to penalties, but also incorporates the individual perception of what he or she has learned over time. This decision making process becomes standardized as the taxpayer acclimates to his or her situation and environment. Only when the taxpayer must react to new information such as the introduction of a new tax law does a change to the decision routine occur. Based on this learning routine, audits have played a central role in the tax administration toolkit. This form of sanction, serves to educate the taxpayer on past behaviors as well as setting the course for future behavior. Unfortunately, the result of the audit foci is that inordinate amounts of attention and resources are paid to the use of audits and from a macro-economic standpoint, the audit rate by tax administrators. While it is generally believed that higher audit rates will result in higher compliance some have argued the opposite could occur. Carroll (1992) points out, that if there were an effort to increase the general knowledge of the possibility of audit; increased noncompliance could occur. The basis for this reasoning is that taxpayers may seek to rebel against the intrusiveness of the tax authority. At the same time, the economic 24 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. school continuously produces studies that suggest a higher audit rate will result in a higher level of compliance which ultimately drives tax administration practice (Aim, Jackson, & McKee, 1992). Concerns regarding the audit rate, which has slipped below 1% for small business and self employed taxpayers, have forced the 1RS to request additional resources for post-filing activities (enforcement). For example, the relationship between the threat of being detected for noncompliance via the audit and the taxpayers’ intention to evade taxes in the future was found to be statistically significant by most studies (Grasmick & Scott, 1982). Conversely, Kirchler et al. (2001) present that little tax knowledge is associated with low tax compliance. Kirchler et al. believe the explanation for this phenomenon is that tax knowledge is positively correlated with attitudes toward legal tax avoidance and at the same time negatively correlated with attitudes toward illegal tax evasion. Egashira and Hashimoto (2000) maintain that contemporary economists believe social loss is caused by imperfect information related to the irrational behavior of economic agents. This disagreement among economists suggests that there is no foundation from which to derive an optimal audit rate. hi general, formal economic models of tax behavior lack conclusive empirical evidence that deterrent effects (such as the frequency of audits) are the basis to fully describe and explain individual taxpayer behavior (Kirchler et al., 2001, p. 5). Brehm (1996) is more straightforward in his observation of the effectiveness of auditing: Although the presumption is strong that expected sanctions should be a significant deterrent to tax fraud, the evidence at 25 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. best points to a more complicated relationship and at worst highlights the inefficiency of coercive power of the state as a means for obtaining taxes, (p. 12) Formal economic models fail to account for significant reasons for compliance by assuming divergence of preferences or single-minded individuals seeking leisure. Empirical analysis of compliance suffers from an excess of competing motives, not all of which are easily observable to the analyst (p. 19). Aim suggests that standard economic theory fails to account for the impact of other taxpayers’ behaviors and social norms in compliance (Aim, 1988). These fundamental weaknesses with traditional economic theory require a thorough discussion of another approach to behavioral taxation: the sociological school. Sociological Perspective of Tax Compliance Sociologists and psychologists believe that the decision to comply with tax obligations may involve more than simply considering the consequences of their behavior. The sociological view includes higher order considerations such as moral and civic duty (Etzioni, 1988). Guilt and the possible social stigma of being labeled a tax cheat have also been strongly associated with deterrence. Traditional economic models do not include intrinsic motives such as fairness and the consideration of social norms (Kirchler et al., 2001, p. 3). According to Scholtz and Pirmey (1995), people’s behaviors may not be focused on self-interest alone. Sociological-based theories tend to differ from their economic counterparts hased on the concept of social capital. Basically, the ability to obtain social capital is not tangible material that 26 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. can be obtained by a single individual like money or education. Therefore, the actions of individuals cannot necessarily be isolated from their environment. Reckers, Sanders, and Roark (1994, p. 833) found that individual moral beliefs are highly significant in making the decision to be compliant. They found that when tax evasion was seen as a moral issue by taxpayers they were less likely to evade regardless of their financial circumstances. An important aspect of the sociological standpoint is the taxpayer’s perception of fairness. The literature portrays individuals as complying with their tax obligations if they believe others are also complying. Conversely, if an individual believes that others are cheating, he/she has a tendency not to comply (Carroll, 1992). In addition, the overall tax system must be perceived as being fair, otherwise essentially the same effect is observed. However, fairness and attitudes toward taxpaying constantly change over time (Taylor, 2001). Some survey evidence has reported a positive correlation between perceptions of fiscal inequity and tax evasion (Torgler, n.d., Internet, p. 5). Accordingly, the lack of equity within an exchange relationship creates a sense of stress for the victim. Torgler argues that this tax evasion is a reaction to inequity stress, the taxpayer’s attempt to put things back on equal ground (p. 5). In this scenario, a taxpayer who believes he/she is not receiving a fair trade in government provided goods and services will withhold income to achieve equity. Sociologists use taxpayer surveys to solicit feedback regarding the perceived systemic fairness of the tax system. Since the 1980s, the 1RS h ^ been using taxpayer 27 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. surveys developed by Roper or Gallup to monitor taxpayer perceptions of fairness. There are several difficult issues with collecting normative data from citizens and developing profiles of taxpayer behavior. It is probable that willful tax evaders normally do not complete in-depth attitudinal surveys, even if they are administered by a third-party who pledges to safeguard their identity. Researchers are dependent upon attitudinal data from those who are likely to comply. Brehm (1996) proposes that one approach to gain exogenous information is to use external data: The state may enjoy higher compliance rates with individuals than survey organizations on such matters as provision of information about the value of property, or enumeration of household characteristics. To the extent that this information is a matter of public record, it provides better information about the potential characteristics of evasive people than those obtained solely by observation or direct inquiry. But it also raises significant ethical concerns about potential breaches of privacy, and unwarranted intrusion, (p. 17) From a practical standpoint, basic external data is available via state provided databases, commercial credit reports, and other commercially available sources (e.g., WestLaw, Choicepoint). However, because of the multiple sources of data, it is difficult to design a viable methodology to profile taxpayer behavior (e.g., there can be three different commercial providers in one state alone). More importantly, the collection of external data alone does not provide insight into the intrinsic motivation of the taxpayer. One example where external data is currently used to track is the U.S. Treasury’s Financial Crimes Enforcement Network (FinCEN). However, this 28 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. data is not used to profile taxpayers. This small agency, consisting of approximately 200 employees, is responsible for tracking money laundering crimes. FinCEN uses a numerous public information databases (e.g., CBI-EQUIFAX, Dallas Computer Services, and TRW Consumer Credit) to research specific individual and group related activities. In addition, they use artificial intelligence to model relationships among individual account owners to identify criminal networks. Unfortunately, this model is not easily transferable to the tax compliance sector since it involves a targeted approach and is resource intensive. Of greater concern is the level of invasiveness that this methodology deploys against individuals that would be impossible to justify except in cases of high dollar tax fraud where there are obvious connections to other crimes such as racketeering or even terrorism. Tax Administration: Taxpayer Categories for Study Substantial theoretical work has been turned into tax policy, such as the institutionalization of auditing and survey administration as data collection practices. The primary vehicle for identification in the case of economic studies of taxpayers is the tax form itself. Tax forms are based on legal definitions of taxpayer types that categorize by the type of income they receive. Besides providing the basis for grouping taxpayers, these categories, such as the wage and investment, provide the 1RS with the ability to appear unbiased in the collection of taxpayer data. In addition, the 1RS tax administration systems for detecting and measuring taxpayer noncompliance are designed at the aggregate level focusing on legal definitions, which 29 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. have been designed for the purpose of management and reporting. For example, the Form 1040 filer who uses the Schedule C for reporting small business or sole proprietor income is one of the most common classifications of taxpayers. Other taxpayers are categorized by a specific legal issue, such as an Offer-in-Compromise (QIC), or by claiming a certain credit (e.g.. Earned Income Tax Credit [EITC]). Issue oriented studies of taxpayer behavior have become the norm for tax administrators since they tend to receive the highest political scrutiny. It is estimated that EITC-based false claims amount to over $30 billion a year in lost revenue. On average, 1RS estimates a 97% compliance rate for wage earning and salaried employees while the percentage for all other income is approximately 80% (1RS, 2002, Small Business & Self Employed Web site). The disparity in reporting is directly related to the visibility of the income or the opportunity to evade. Wage earners, salaried employees, and small business owners are monitored via Form W-2 (as well as Form 1099 for other types of income) which is submitted by the employer. The 1RS, through the Information Return Processing (IRP) program (as well as state governments using 1RS data), maintains an extensive program which matches these documents against what is reported by the taxpayer on the Form 1040. This program has become more robust over the years that the definition of audit has been revised to exclude contacts relating to the 1RS information reporting program or CP 2000. A full appreciation of the IRS's ability to examine the accuracy of retums requires going beyond the revised definition and considering the matching system. 30 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. According to Kagan (Carroll, 1989), compliance is nearly perfect for taxpayers (primarily wage earners) where income is matched as opposed to small business owners and self-proprietors. The 1RS information reporting and matching system represents the electronic verification of more than 90% of all income and more than 40% of all deductions for individual income tax retums (1RS, 2002, Small Business & Self Employed Web site). This translates into the complete electronic verification of all income and deductions for more than 45% of all individual retums. The ability to match data has also led to the ad-hoc identification of non-filers. Eleven states including California, Florida and New York are leading the way in identifying non filers through localized matching of 1RS data. The 1RS classifies individual retums into ten categories based on total positive income (TPI) for non-business retums and total gross receipts (TGR) for business retums. The audit rate for each category is computed by multiplying 100 times the number of audits performed during the year and dividing by the number of retums filed during the year. The number of audits performed is primarily for retums from one of the three prior years while the number of retums filed is for the current year. In general, audit rates for high-income, non-business retums and business retums are about twice the 1.67% overall rate. Approximately 33 million self-employed and supplemental income eamers differ from traditional wage and investment taxpayers simply because they file a Schedule C (1RS, 2002, Small Business & Self Employed Web site). Because of the 31 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. nature of the reporting requirements on the Schedule C, as well as the multitude of business types, monitoring compliance is difficult. The burden on small business filers is substantial, since they file twice the number of forms and schedules. Tax compliance issues often stem from a lack of understanding of tax law requirements, inadequate accounting practices, and resources and cash flow problems. From a tax administration standpoint, based on how they report taxes via unmatchable deductions on the Schedule C, the small business and self-employed taxpayer is the most problematic type or category of taxpayer for the 1RS and for state and local tax authorities. Analysis Tools Designed to Detect NoncompUance The Taxpayer Compliance Measurement Program (TCMP) Formerly, a prominent method for collecting taxpayer data and detecting noncompliance was performed through the 1RS TCMP, an intensive auditing program. The TCMP data consisted of detailed information reported on tax retums that have been subjected to intensive line-by-line review by a trained cadre of 1RS revenue agents. The 1RS established the TCMP in 1964, using tax year (TY) 1963 retums as the first base population for examination. The audit was conducted generally in three-year cycles for individual taxpayers. The last full TCMP was performed in 1988, primarily on wage and investment taxpayers and consisted of approximately 54,000 records (1RS, 1993). 32 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The retums selected for examination are not chosen randomly. 1RS policy is to identify and audit those retums with the most potential for noncompliance. This policy was designed to better utilize limited funds and helps prevent burdening compliant taxpayers with unnecessary audits. The 1RS reports that programs to identify and audit those retums with the most potential for noncompliance has helped to decrease its "no change" findings fi-om more than 40% to approximately 15% of all individual audits (1RS, 1993). Nonrandom selection of retums means that the popular belief that audit rates represent the statistical chances of being audited is not accurate. The 1RS uses several nonrandom methods to select retums for auditing purposes. For example, unallowable items, which are the most prominent method, accounted for 42% of all audited retums for 1996. The second most important source (18%) of audited retums was discriminant function analysis (DIF), a sophisticated scoring system that relies on statistical data collected from the most recent TCMP. Other countries use a similar DIF constmct (Torgler, 1993). The third most important audit source (11%) was the non-filer program, which is based on matching, while the fourth (4%) was the state information program. Together, these methods accounted for 75% of all audited retums for 1996. Besides providing data for examination audit selection, TCMP data was used to develop aggregate measures of filing compliance. This included a macro series of “tax gap” measures which serve as the primary indicators of noncompliance (e.g., the gross tax gap, which is the amount of income tax owed but not voluntary paid for the 33 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. entire U.S. population). In 1998, the 1RS estimated the gross tax gap at $280 billion. The net tax gap is the gross tax gap less tax paid as a result of 1RS enforcement actions. The tax gap included unpaid individual and business income taxes. TCMP was used as a tool to allocate resources within the 1RS since it provided clues to the source of noncompliance activities (e.g., the shadow economy). Moreover, TCMP data was incorporated into methodologies for determining the required resources for tax enforcement. The enforcement population is a substantial part of the approximately 100,000 employees of the 1RS overall. The 1RS 2003 budget justification identifies 46,872 full-time equivalents (FTE) out of 101,080 FTE or 46.4% requested for tax law enforcement alone (1RS, 2002, Budget Justification). The FTE used for law enforcement translates to $3.9 billion in budget costs for fiscal year 2003. TCMP Data was also generated for other 1RS research efforts and supported state revenue collection efforts. The TCMP presented several issues that led to its cancellation in the early 1990s. The line-by-line audit process was considered burdensome which made it easy prey for political manipulation. Political opposition mounted against what was termed by many "the audit from hell." Cementing this opposition was IRS's move to increase the sample size of the TCMP to 135,000 taxpayers in 1994 to improve reliability. Moreover, the amount of 1RS resources deployed to perform the TCMP type audits was substantial. Since there was not a full accounting of the resource 34 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. drain from regular audits, a thorough cost-benefit analysis of the overall TCMP process was difficult to accomplish. Besides practical issues associated with administrating the TCMP, there are issues associated with auditing as a practice to detect noncompliance. Clotfelder (1983) used TCMP data in his investigation into how noncompliance responded to changes in the environment. He developed a model in which noncompliance was a fiinction of the combined federal and state marginal tax rate, after-tax auditor adjusted income, and a set of demographic variables that were available on tax retums. Clotfelder concluded that noncompliance is strongly related to the marginal tax rate, a controversial position. Sheffrin and Triest (1992), in their study of TCMP data also found conflicting results: In the TCMP audits, a significant proportion of taxpayers made mistakes that overstate their taxable income. Thus, there are presumably also mistakes in the other direction. Auditors would view this income as noncompliance, while taxpayers would not. In addition, audits are poor vehicles for uncovering unreported income, (p. 196) The uncertainty of the TCMP data, which is based on a random sample, suggests a flawed methodology in efforts to apply the study’s assumptions to the general population. Concerns with auditing as an effective compliance tool have been aggravated by charges that the 1RS was not using the data for root-cause analysis. For example, the GAO (Tax Notes, 1994) reported that the 1RS was not using TCMP data as well as it could to determine the cause of some non-compliance. 35 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The GAO said the Service was not using TCMP data as well as it could to determine the cause of some noncompliance. It (the GAO) found out, for example, that many truckers’ noncompliance stemmed from inadequate record keeping. The GAO recommended that the Service (1RS) develop a system for monitoring all sole proprietor compliance projects, and link them to nationwide plans. It also recommended meeting with trucking industry groups and insurance companies to improve record keeping and information reporting requirements. Although the original TCMP was cancelled, the culture that developed and performed this type of study within the 1RS research community and the Department of the Treasury is still prevalent, resulting in the revival the limited audit-based study process, the National Research Program. To address the issue of burden, the 1RS is collecting more data on the subject in an attempt to minimize the possibility of an audit. The 1RS refers to this activity as “case building” which involves data from several different sources. Although there has not been renewed debate on privacy or invasiveness, the 1RS plans to use: (1) income tax retums from current and past years; (2) information documents such as W-2s and Form 1099; (3) state tax refunds; (4) mortgage interest; (5) Department of Treasury Currency Transaction Reports (CTR) and other banking data; (6) dependent database; and (7) retums from related entities (Brovra & Mazur, 2002). The retums and associated data will be classified by a “well trained group of experts, prior to any examination.” Additional training will be provided to the actual examiners or auditors (numbering over 1000) as the process begins on 2001 tax retums. 36 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. As a result of the 1RS Restructuring and Reform Act of 1998 (RRA 98), the 1RS was restructured into operating divisions designed to service specific types of taxpayer groups (e.g., Small Business/Self Employed, Wage and Investment, Large and Midsize Business, etc.). This is a departure from traditional 1RS reorganizations that tended to maintain the functional focus of the organization (e.g., collection, examination). Each new operating division includes an education and communications subdivision with the primary purpose of deterring taxpayer noncompliance by proactively providing training and educating taxpayers regarding complexity issues. At the same time, the 1RS still dedicates a large amount of its employees to enforcement based activities. While the National Research Program is the most visible of 1RS compliance efforts, there are other efforts that should be discussed. For the Internal Revenue Service, tax research has taken new directions including the use of agent-based modeling and neural networking, both rooted in biology, to model taxpayer behavior. Agent-Based Modeling Agent-based modeling recognizes people as individuals who maintain their own, unique internal states and endowments. This type of behavioral modeling departs fi’ om traditional economic models which portray the agents acting in a relatively pure rational state. More importantly, agent-based modeling avoids the singular focus of a representative individual. The agent-based approach begins with an institutionalized set of rules from which basic behaviors are carefully articulated 37 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (Davis, Hecht, & Perkins, 2000). Agents interact within the predefined framework without the requirement of being completely rational. Aggregate-level behavior becomes a dependent variable as the agents learn from other agents within the environment. Some aspects of these internal states and knowledge are fixed, while others change as a result of interactions with other agents. What emerges is described by Davis et al. . . . a society of taxpayers who are initially endowed with limited knowledge about the enforcement regime and heterogeneous reporting decision rules that are a function of perceived enforcement severity, social norms, and the reporting behavior of acquaintances . . . taxpayers develop periodically updated beliefs regarding social norms and enforcement severity. Individual taxpayer reporting behavior is then permitted to evolve over time, in response to the behavior of other agents. (Introduction section, para. 6) Agent-based modeling allows for the groA vth o f social networks and attempts to present a dynamic stage for predicting future behaviorisms. Recognizing the potential value of this technique, the 1RS is using this Agent-based modeling in an effort to gauge the potential levels of acceptance of electronic filing. This application is of special concern since the 1RS is committed to receiving 80% of all tax returns via electronic filing by 2007. However, the 1RS recognizes that an aging population base is less likely to embrace electronic filing. Neural Networking Efforts are also underway to use neural networking to detect noncompliance of individual high-income filers. This technology is being deployed, with some degree 38 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. of success, in the health insurance field for fi*aud detection, and in the automobile insurance field for warranty abuse detection. Essentially, neural networking models the functions the human brain. A network can vary in size, consisting of a small number to a few billion neurons that can be connected in an array of different methods. Neural networks attempt to model these biological structures both in architecture and operation. Each neuron has a unique threshold value, and if the net is greater than the threshold, similar to a binary switch, the neuron fires (or outputs equal to “1”; the switch is on), otherwise it remains static (outputs is equal to “0”; the switch is off). The output is then fed into all interconnected neurons. A number of issues with neural networks may confound those using this methodology to detect noncompliance. Since biologists do not understand fully how the human brain works, this presents a dilemma in deciding what represents an ideal model of brain activity. There are no set rules to help construct a network and many factors must be taken into consideration: the learning algorithm, architecture, number of neurons per layer, number of layers, and data representation. In addition, computer hardware is designed to operate in serial, not parallel, which is the ideal state for processing in a neural net. 1RS is attempting to use neural networks to detect anomalies among high-income filers. While these new methods show some promise for tax administrators, a fundamental problem remains regarding the predictive capabilities and treatments for deterring tax evasion. Taxpayer profiling requires a dynamic monitoring process that 39 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. senses changes in the propensity to comply. Audits are post-mortem (they capture, to some degree, the behavior after the fact) while providing some insight into why the taxpayer (a misnomer in this case) evaded, and they are still a snapshot of the past. To understand compliance behavior requires a thorough discussion of its theoretical basis in modem contextual psychology. Theoretical Framework of Compliance Behavior The study of human compliance includes many interesting experiments such as Milgram’s work on obedience and Kelman’s experiments on the importance of the perceived power of the requestor (Brehm, 1996). Milgram observed the importance of social context during his power experiments of the 1960s when he pronounced “Under certain circumstances, it is not so much the kind of person a man is as it is the kind of situation in which he is placed that determines his actions” (Haney, 2002, p. 5). Admittedly, Milgram’s statement indicates a fimdamental weakness of laboratory studies, i.e., while a subject’s behavior can be manipulated in a controlled environment, the environment itself is a function of the imagination of the experimenter. The environment, under some circumstances, might reflect such a complex construct of different social pressures that it becomes almost impossible to imitate in the laboratory. Fortunately, classic empirically based psychology, the foundation of compliance studies, has recognized that laboratory experiments do not necessarily provide the best possible framework for describing human behavior. 40 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Haney has worked to describe this shift in psychological studies based on work by Manicas et al.: Explaining the behavior of particular individuals is now understood to require not only psychological theory but also situational, biographical, and historical information . . . what has been demonstrated through a host of celebrated laboratory and field studies is that manipulations of the immediate social situation can overwhelm in importance the type of individual differences in personal traits or dispositions that people normally think of being as determinative of social behavior. G».S) Using contextual psychology as a starting point, it is apparent that studies which generalize human behavior fall short in their relative ability to predict such attributes as compliance. Haney (2002) points out research on a broad range of human behaviors, research exploring the effects of living in certain neighborhoods on individuals, families, peer groups, and other social networks has mushroomed in the last several years. An important conclusion fi-om this research is that neighborhoods are an important consideration which has broad implications on localized compliance behavior (p. 7). Studies of compliance, like other behavior-based studies, are in a constant state of flux, reflecting the context or environmental change. Therefore, unlike the static variable tests of past compliance efforts, contextual environments should be reanalyzed relentlessly to understand the forces upon individual behavior. Compliance studies, similar to Hanley’s (2002, p. 9) characterization of the law, are constantly out of date. This implies that, without constant monitoring, tax administration systems designed to detect noncompliance and their supporting 41 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. studies, such as the original TCMP, may not offer viable means to monitor or model compliance behavior. Traditional compliance models often simply attempt to attribute behavior to internal disposition rather than to complex situational or contextual factors that indeed may influence behavior. Of greater concern is that individuals seek out a single causal explanation (motive) rather than considering complex multiple causes for explaining behavior. What is missing from these models is the role of decision making theory in how individuals perceive and process environmental information. An individual’s (in this case a trucker’s) decision to comply with his/her tax obligation is based in complex decision-making theory. Ruback and Wroblewski (2001, p. 10) reveal three critical aspects of complex decision making: (1) it is often based on errors, biases and heuristics; (2) the decision maker has little insight into his/her own decision making process; and (3) actuarial models are more accurate, more consistent, and thus fairer than nonstatistical decision making. The use of heuristics within the realm of complex decision making is an example of where individuals attempt to limit information processing within the bounds of their own cognitive abilities (p. 11). These heuristics provide a comfort zone based on experience using limited and easily available information to make a decision regarding one’s tax obligations. Kahneman and Tversky developed a choice model based on conditions of uncertainty known as Prospect Theory (Reckers, Sanders, & Roark, 1994). This theory departs from traditional expected utility models by considering the context or 42 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. frame of a decision as a factor that impacts the choice of the decision maker. They define the fi’ ame of a decision point as the presentation, reference points, alternatives, outcomes, and their probability of occurrence. Unlike the pure rational actor, prospect theory recognizes that the decision maker must identify the inputs to the decision process. More importantly, the theory recognizes that individuals have limited cognitive abilities. The theory includes an editing step where the decision maker chooses to disregard some information focusing on other perceived key points which he translates into a meaningful form such as income gain or loss. Reckers et al. describe the use of prospect theory in tax compliance studies as violating the expected utility tenet of description invariance known as the reflection effect (gain/loss framing effect): The same outcome can be edited either as a gain or loss depending on the reference point presented in the decision frame, and a different decision is likely depending on the frame adopted. The decision maker focuses on changes in wealth or welfare from the current reference point (current wealth position) rather than focusing on the final wealth state. Income then is not the only argument of the cardinal utility fimction, and marginal utility is not always positive and strictly decreasing. Rather, prospect theory proposes that individuals will display a value (utility) function that is concave for gains and convex for losses, with the latter being steeper than the former, (p. 828) The so-called framing effect is not applicable to all decision making schemas. It captures an effect that occurs for some people under certain conditions and does not include the effect of social norms, ethics, and personal characteristics. Reckers et al. 43 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. recognize that without incorporating these factors much of the work in tax compliance using prospect theory has provided limited positive result. In 1991 the 1RS attempted to perform a study using the concept of self-audit for taxpayers claiming higher than normal deductions in Massachusetts. This treatment was not advertised in an effort to deprive the state’s taxpayers of the necessary foreknowledge to plan a reaction. Therefore, strategies to avoid reporting income based on this knowledge could not be implemented during the initial tax-filing period, in a sense depriving taxpayers of their compliance heuristic. Actual taxpayer responses were not required and waivers of penalties were not offered. Unfortunately, 70% of the taxpayers did not respond to the request. Hemmer, Stinson and Vaysman (1994) found that taxpayers will seldom respond positively to self audits when they are voluntary. This response may simply reflect a phenomenon of heuristic shock; i.e., the taxpayer simply did not know how to react. The Minnesota Department of Revenue performed a study that approaches the compliance issue differently than past efforts on the Federal level (Blumenthal, Christian, & Slemrod, 1998). The study used the threat of audit through the notice process as a treatment to measure the possible impact on compliance. The study was unique since it was limited to a sample of Minnesota taxpayers and while this group was still quite heterogeneous, it provided a snapshot specific to a limited region of the country. The results of the study did note that certain types of filers reacted more positively to the treatment than others. Specifically, the financial and construction 44 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. industries were singled out as more sensitive to the enforcement based treatment. This work could not be repeated at the Federal level since it involves the threat of audit to taxpayers which would have obvious repercussions in Congress. However, there is relative value in the observation that certain groups, identified by North American Industry Classification (NAIC) system, behaved differently than their counterparts within the sample. The Minnesota study found that for low and middle- income taxpayers the threat of audit resulted in slightly increased reported income, but the opposite was true for high-income taxpayers. The authors of the Minnesota study believed that phenomenon can be explained as the high income taxpayer viewing the audit as a process of negotiation with reported income as a first bid. High-income filers may believe that, in fact, they will not be penalized or pay interest on audit-based additional income because past experiences have taught them to adopt this heuristic. Clotfelder found that underreporting varies with respect to employment group (Kirchler et al., 2001, p. 6). Kirchler et al. found that individual tax compliance is lower for business entrepreneurs than for lawyers and fiscal officers who are better educated on the myriad of possible penalties or simply more sophisticated in their tax preparation. Contrary to the Minnesota study findings, Blumenthal, Christian, and Slemrod (1998, p. 28) found no evidence that the written threats of audit had any impact on aggregate compliance. Again, competing evidence makes it difficult to draw specific conclusions on how to approach studying compliance. At the same time, regardless of the stimuli (e.g., audit or threat of audit), these studies seem to 45 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. indicate that specific groups of taxpayers, in this case high versus low-income filers and more importantly occupations, may display some shared attributes that characterize their behavior overall. An Individual’s Standing in the World Popular perceptions of justice, fairness and equality may not be objective standards but are more likely formulated by the individual (Taylor, 2001). Taylor (2001) uses Smith and Tyler’s (1996) argument for considering subgroup identity as being a significant aspect of how people perceive fairness: Distinguishing between levels of inclusiveness suggests that when a superordinate category is more important to people, inequities between different groups represent an intragroup situation with implications for collective cooperation and harmony. In contrast, if a particular group is more important to people, inequities between different groups represent an intergroup situation with different groups competing for resources and power. (Social Identity, Justice, and Compliance section) Festinger (1957) also argues that, given a range of possible persons for comparison, someone close to one’s own ability or opinion will be chosen for comparisons. A discussion on the exercise of power over another individual or informal group is essential consideration for understanding behavior changes such as the adoption of a tax evasion tactic. While several types of power are identified in the literature, such as reward, coercive, expert, and legitimate, these imply a hierarchical or vertical distribution (French & Raven, 1989, p. 440). Within an informal group or personal fi-iendships, social influence may be impressed upon one individual by another through 46 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. a strong feeling of association. Referent power is based on the level of identification between two individuals or at least the aspiration of one individual to feel oneness with the other. The follower may not be cognizant of his /her referent state with the dominant individual. Moreover, the dominant individual has the ability to influence the follower even though the latter may not be aware of this aspect of the relationship. The theory of relative deprivation suggests that people often specifically compare their relative individual fortunes with those of others in similar or close circumstances. Davis (1959) suggested that comparisons with similar versus dissimilar individuals leads to different kinds of emotional reactions. The concept of what is fair is influenced by these comparisons. Taylor (2001) proposed that taxpayers see themselves as interchangeable with other taxpayers and interchangeable within a subgroup of taxpayers in contrast to another group. Perceptions of group- based injustice become stronger when social rather than personal identity is considered. What is perceived as fair is ultimately based on references to what other groups of people of similar characteristics have. This is similar to horizontal inequity, except that inequity is based on comparisons to groups as opposed to individuals (Taylor, 2001). Duclos, Jalbert, and Araar (2001) suggest that it is possible that individuals make comparisons of their income and other internalized values across their reference group. 47 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Verhoogen et al. (2002) identify a robust association between the rate of unemployment in the local labor market and fair wage perceptions of employees. Accordingly, Fair Wage theory is simply a causal proposition: the fairer an employee considers his wage, the harder he will work. Little is known about exogenous influences of workers’ fairness judgments in the workplace. Experimental literature in economics and sociology has established that people tend to form fairness judgments in reference to an alternative transaction or aspect of behavior that they view as fair. For the subject, this is his/her reference point of fairness. Important consideration is that an individual’s reference point is sensitive to how an experiment is framed (Verhoogen et al., 2002). Experimental results may not be reliable indicators of whether employees in real workplaces are being influenced by external factors. For fairness perceptions, these reference points may include salaries of fellow employees in the workplace, or some other point of comparison. There is a causal relationship between local unemployment rates and employees’ perceptions of a fair wage. For example, if a surrounding area’s unemployment rate is high, employee perceptions of wage fairness increase. Verhoogen et al. (2002) observed the causal effect of wage-faimess perceptions on employee performance in a geographically dispersed freight handling business. This was a unionized organization, so wages were set at the national level through collective bargaining. Their findings concluded that economic shocks in the local area surrounding the trucking firms’ operational units generated exogenous 48 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. variation in the attractiveness of the wage paid by the firm relative to the employee’s option in the outside labor market. They related this geographic variation to the employee’s perception of fairness for each operational unit and ultimately linked this perception of fairness to employee performance. According to Strauss (1997, p. 25), the taxpayer’s location is not an important factor for most Federal tax calculations. However, for the states, the taxpayer’s location of an individual or business is a crucial aspect of the revenue gathering process. Geographic location is a key element when considering how to study compliance issues. Without context, or consideration of the issues peculiar to a distinct geographic location, human behavior is being analyzed in a vacuum. A key element of compliance studies is the identification of extraneous variables that are peculiar to a taxpayer’s environment. Environmental Influence on Behavior For an individual to adopt an evasive tactic, a tacit behavior must be observed or developed as a reaction to a stimulus. Attribution theory is based on the concept that an individual’s behavior is caused by certain aspects of the environment (Luthans, 1995, p. 162). Since these environmental factors are unobservable, individuals are dependent upon cognitions, primarily perception. Accordingly, attribution theory portrays individuals as rational and motivated to identify and understand the casual structure of their relevant environment. Luthans describes attribution theory’s initiator Fritz Holder’s belief that both internal forces and external factors combine additively to determine behavior. In addition, Heider stressed that it is perception, 49 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. not necessarily actual determinants» that influence behavior. Attribution theory has been used extensively in attempts to explain work motivation in organizational theory. The locus of control in the work environment is comparable to how individuals perceive the variability of control they exercise in social settings. If individuals perceive the ability to personally influence their work environment, then their relative satisfaction with their work environment is higher (Luthans, 1995, p. 169). If individuals perceive they are externally controlled and outcomes are beyond their control, then their satisfaction with their endeavors is less. Kruglanski, Shah, Pierro, and Mannetti (2002) also developed a theory that acknowledged externally driven events had a negative effect on individual satisfaction because of the feeling that one is compelled to do some undesired activity as a means of ultimately attaining gratification. Ajzen and Fishbein (1980) developed the theory of reasoned action, which attempts to explain why individuals engage in certain types of behavior. Figure 1 provides a fishbone diagram of how the theory forms individual behavior. The theory describes three different types of external variables which could affect the decision making process: demographic variables, attitudes toward target individuals, and personality traits. The three variables work together on two other variable groups that lead to intention: attitudes and subjective norms. 50 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CD ■ D O Q . C g Q . ■ D CD C/) C/) 8 ( O ' 3. 3 " CD CD ■ D O Q . C a o 3 " O o CD Q . ■ D CD C/) C/) Ajzen and Fishbein 1980 Theory of Reasoned Action Behavior Intention Attitude toward the behavior Relative Importance of attltudlnai and normative considerations Relative importance of attltudlnai and normative considerations The person's beliefs that the behavior leads to certain outcomes and his evaluations of these outcomes The person's beliefs that specific Individuals or groups think he should or should not perform the behavior and his motivation to comply with the specific referents Figure 1. Theory of Reasoned Action, Fishbone Diagram. Source: Available at www.indstate.edu/nurs/mary/fish.htm. V I Attitudes and specific norms ultimately determine whether a specific behavior is attempted. Therefore, a person’s decision to evade or comply is the result of personal and social pressures individually evaluated by each decision-maker in accordance with the perceived importance to the individual. The trucker may use other truckers as a referent, a yardstick to measure one’s wealth, including equipment investment (i.e., the “rig”). Alvarez and Brehm (1998) cite Hawthorne and Jackson’s work (1987) demonstrating that material self-interest affects opinions on tax policy, one of the very few areas where scholars have been successful in demonstrating direct effects of material self-interest on opinions. Measuring one’s self against another, using such criteria as material wealth, provides a possible stimulus for evasion but does not explain how evasive tactics are developed or dispersed beyond the individual. Noncompliance cannot spread beyond the individual business owner without a medium of dispersal and a fundamental belief among future evaders that the tactic is viable for their own use. Moreover, an evasion tactic may actually be tested in a localized group by the initial developer. This scenario requires that those consulted actively trust both the creativity of the developer and those within the potential circle of adopters. Human Creativity, New Ideas, and Social Networking Woodman, Sawyer, and Griffin (1993) model human creativity as the complex product of a person’s behavior in a given situation. They define individual creativity as: 52 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. . . . a function of antecedent conditions (e.g., past reinforcement history, biographical variables), cognitive style and ability (e.g., divergent thinking, ideational fluency), personality factors (e.g., self-esteem, locus of control), relevant knowledge, motivation, social influences (e.g., physical environment, task and time constraints), (pp. 295- 296) These individual characteristics are influenced by social and contextual factors. Within a group setting, social influences impact individual creativity. Consequently, the individual’s ability to be creative contributes to the group’s preeminence. According to Woodman et al. (1993), the group’s ability be creative is not simply the sum value of the aggregate contributions of its membership. Factors that must be considered include the group’s composition (e.g., diversity), group characteristics (e.g., cohesiveness, group size), group processes (e.g., problem-solving strategies, social information processes), and contextual influences stemming from the organization (p. 304). While the focus of their work has mainly been on formal organizations. Woodman et al. suggest that the same characteristics can be applied to informal groups. Within these informal groups are those who champion new ideas and approaches within the broader societal setting. This proposes that evasion, like any new idea (without the necessary moral context) may be developed and distributed in relatively small informal groups. Friendship networks may offer one medium in which the idea or technique to evade is introduced. External factors such as perceived enforcement activity are considered as a constraint on the perpetuation of the idea within and outside the 53 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. formal group setting. Group norms may offer another constraint on the formulation and perpetuation of an idea. The reality is that behavior is naturally constrained by many of the same factors that empower new ideas. Behavioral experiments are built around the effective introduction and manipulation of constraints to produce the desired result (Woodman et al., 1993, p. 312). Woodman et al. (p. 312) hypothesized that individual creative performance will be decreased by group norms that create high conformity expectations and visa versa when group norms support open sharing of information. Unfortunately, both types of conformance expectations can occur for different issues within an informal group. To consider the existence of informal groups within the trucking industry requires an understanding of the contextual existence or how information is dispersed in a larger environment. Fundamental weaknesses are related to the development of informal groups: These solely interpersonal arrangements are undependable and open to manipulation for anti-social and competitive as well as mutually protective ends. A number of isolates is left over. The total face group is incapable, except defensively, of acting as a socially responsible whole, since not even private allegiances are owed outside the small informal groups. (Trist & Bamforth, 1989; Oft, 1989, p. 386) If informal groups are “too close,” as Trist and Bamforth suggest, how can these structures disperse evasion tactics beyond their clique? Is it possible that groups behave like loose organizations due to business pressures and the general scarcity of resources? Many organizations must constantly struggle to survive. A key to survival is acquiring resources from the environment, which is composed of other 54 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. organizations. The problem is that the environment is unstable and undependable (Pfeffer & Salancik, 1978). Past organizational research has mostly focused on the problems associated with resource utility or the efficiency of processes within a single organization. But resources must be acquired before they can be used, and this involves exchange vsdth the environment. Organizational effectiveness is an organization’s “ability to create acceptable outcomes and actions” (Pfeffer & Salancik, 1978). Traditionally, this means being acceptable to outside constituents; effectiveness is an external standard of performance. In the trucking culture, groups may very share information with other groups as a venerable extension of their prowess or from a business standpoint, effectiveness. The ability to spread new ideas (in this case, evasion) requires a medium of dispersion that bridges the single group setting. However, for an idea to disperse effectively, its source and the people that adopt it must be viewed as a trusted source. The Implications of Trust in Social Networks Robert Putnam’s (1993) definition of social capital is built around civic involvement (e.g., the success of regional governments in Northern Italy compared to Southern Italy). His basic argument is that communities with highly developed social networks of communications and a strong normative structure based in reciprocity and trust are more likely to result in high performing governments. The key to functionality in Putnam’s work is trust. In a functional group setting, trust is often shared among friends where the agents tend to have similar interests; this is referred 55 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. to as the Homophily Principle (Blau, 1977). Trust must be built upon a foundation which has been described as “thick trust” which is based on a deep familiarity or intimacy with another person (Putnam, 1993, p. 171). Putnam describes the necessary dependencies for trust: You do not trust a person (or an agency) to do something merely because he says he will do it. You trust him only because, knowing what you know of his disposition, his available options and their consequences, his ability and so forth you expect that he will choose to do it. (p. 171) The question then becomes: how does interpersonal trust translate to larger social groups? Five central aspects characterize social capital (Pretty & Ward, 2001): relations of trust; reciprocity and exchanges; common rules, norms and sanctions; and connectedness of networks and groups: Connectedness, networks, and groups and the nature of relationships are a vital aspect of social capital. There may be many different types of connection between groups (trading of goods, exchange of information, mutual help, and provision of loans, common celebrations, such as prayer, marriages, and funerals). They may be one-way or two-way, and may be long-established (and so not responsive to current conditions), or subject to regular update. The key to social networking is that individuals of different social stature interact to allow mobility. Social capital theory relates to the relations, networks and obligations that exist and finally the products of these interactions. Slemrod (1989) highlights Knack’s and Keefers’ work that found social capital variables exhibit a strong and significant positive relationship to economic growth. Trust could be a result of growth of income and/or perceived prosperity. Knack’s and Keefers (Slemrod, 1989) 56 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. observed that trust is more correlated with per capita income in later years than income with earlier years, suggesting that the causation runs from trust to growth more than visa versa. Slemrod describes the process as follows: The idea is that a high level of social capital (or trust, or rule obedience, or civic duty) contributes to affluence and also facilitates a higher level of government activity than otherwise. Thus, cross-country variations in social capital will generate a positive association between affluence and government activity, one that is not indicative of causation in either direction. It is not because affluence causes higher taxes, or because high taxes cause affluence, but rather that high civic duty societies tend to exhibit both affluence md higher taxes. (p. 6) Brehm and Rahn (1997) identify the mechanism that connects interpersonal trust and sustains the belief that cooperation is based in the Prisoner’s Dilemma (PD): In iterated PD games, successful strategies are “nice” ones where the player is never first to defect (Axelrod, 1984), an instantiation of some initial level of trust. After the first play, successful strategies simply echo the behavior of the other behavior, reciprocating cooperation for cooperation or defection for defection. If cooperators expect other people to cooperate—and experimental research suggests they do (Orbell and Dawes, 1991) they are more likely to engage in cooperative endeavors, setting in motion a “virtuous circle” in which trust promotes cooperation and cooperation promotes trust (Putnam, 1993). (Brehm & Rahn, 1997, Structure Model of Social Capital section, para. 3) While small business truckers may represent the antithesis of civic engagement, civic engagement may follow the exact lines of Putnam’s examples of Parent-Teacher Associations or Italian soccer clubs. For this research sample civic engagements may include fraternal organizations such as the Elks or Lions Club. One important factor 57 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. that could impact this study is the age of the taxpayer in regard to trust. There is substantial evidence (Brehm & Rahn, 1997; Putnam, 1995) regarding mistrust among youth. The generations defined by large-scale collective experiences including the Great Depression and World War II, for tax compliance purposes have moved into retirement. Over the past twenty years, true economic decline has not been as serious as A vith other generations. Younger generations have not truly experienced collective economic misfortune or large-scale, invasive warfare similar to World War II. While it can be argued that recent terrorist attacks of 2001 and thereafter have brought together the country to some degree, they have not forcefully impacted the lives of all Americans (as did, for example, the institution of a draft, fuel and food rationing, etc.). Diffusion of Innovation Theory: Evasion as an Innovation Research on the dispersion of innovations has demonstrated that social stmctures are important variables in the overall explanatory equation. How a structure contributes to a dispersion process varies with different points in time and with every person involved (Coleman, 1964). One of the oldest and most fundamental theoretical concepts developed in diffusion research is the concept of exposure. The exposure model postulates that an individual engages in a collective behavior based upon the proportion of people in his/her personal environment that are already active. An individual’s tendency to adopt a specific behavior is assumed to be a function of the behavior of others in his/her immediate social environment 58 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (Granovetter, 1978). Social ties used to be categorized as either strong or weak (Granovetter, 1973). Tie strength is normally defined by four dimensions: interaction frequency, friendship duration, intimate talks and activities, and mutual services. Emotional support in intimate behaviors is the most important indicator of strong ties. In general, trust is thought to produce strong ties, especially those involved with intimate behaviors showing friendship, rather than traditional business relationships; i.e., weak ties (Uzzi, 1996). Krackhardt and Hanson (1993) drew an overall organizational network picture by diagramming three types of social networks: trust network, advice network and communication network. He specified the functions of trust network as sharing political information and backing one another in a crisis. In another paper, Krackhardt (1992) presented the "philos" network, (i.e., network of strong friendship ties), and he identified it as a trust network, since the central person of a friendship network is generally an underground leader whom other employees trust most. Davis et al. (2000, p. 2) model the dispersion of a tax evasion tactic similar to the spread of a virus. Unlike traditional diffusion of innovation theory, the taxpayer has no choice in moving (as one has little active choice in catching a disease) from a compliant state to being noncompliant since their behavior is an internal response to another taxpayers’ behavior (Davis et al., 2000). Future acts of noncompliance after the crossover are characterized as an intentional act. Davis et al. present Lemer’s argument that many individuals have a need to believe the world is just. This reaction 59 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. may serve as a coping mechanism for dealing with perceptions of fairness of evasion versus compliant behavior. As a result, the more often noncompliant taxpayers are encountered in the environment, the further along the compliant taxpayer slips towards noncompliant behavior. This is a dynamic process of association built on the pre-existing rules of behavior that promote the emotional justification differently for each individual. Modeling noncompliance behaviors as innovations suggests that, like new technology, a new technique to evade may be diffused across potential adopters. Bandwagon Theory suggests the existence of a positive feedback loop which causes pressure on potential adoptees to grow as the number of adopters grows (Abrahamson & Rosenkopf, 1997). Many diffusion of innovation theories examine how an innovation is adopted across organizations not individuals. Like economic rational actor theories, there are “efficient choice theories” that assume organizations make rational decisions to adopt innovations based on information regarding the profitability of the technical efficiency of the innovation (p. 3.). Under the condition of ambiguity, fad theories assume that an innovation is adopted not on the perceived technical prowess or profitability of the innovation but rather on the reputation and numbers of previous adopters. The result of the bandwagon process is that, instead of adopting innovations based on efficiency and profitability of the data, pressure grows among non-adopters to “fit-in” because of social pressures, even if the innovation offers no viable advantage; 60 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. When confronted with empirically ambiguous questions: that is, questions that cannot easily be answered by pointing to concrete facts such as innovation or profitability, organizational decision makers tend to base their decisions on social cues (such as, how many other organizations have adopted this innovation and what is their reputation), (p. 7.) Treating geographical clusters of small business truckers as loose organizations, the technique or technical approach to evasion can be modeled as innovation for resource restricted truck drivers. Key factors in the adoption of such an approach may include the reputation of the leaders or respected members that are internal to the group. It is possible that evasion transcends the inter-group network as known members of other locations intermingle at socialization points. In the case of truckers, these socialization points may include fiiendship networks, truck stops, or contracting colleagues working for larger companies. Abrahamson and Rosenkopf (p. 293) describe a process in which each potential adopter finds out about how an innovation is dependent upon the structure of his/her social network that disseminates the information on the innovation. The network structure influences the strength of the bandwagon pressure on each potential adoptee. Schermerhom (1975) describes collaboration among groups as positively enhanced by the physical proximity of the stakeholders (Gray, 1985, p. 930). Gray believes that collaboration among groups provides the basis for problem-solving. For the trucking industry, these problems may include the scarcity of resources when considering how to comply with government tax obligations. Gray demonstrates that physical proximity facilitates the fi-equency of contact necessary to build 61 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. interdependencies among groups. Collaborative activities include sharing of information and client resources. The theories discussed above provide the framework for the development of tax compliance hypotheses using the sample of California truckers. The single theme for these hypotheses is simple: Can uncontrolled externalities such as level of income, reported expenses, unemployment levels, and education impact trucker tax compliance? Moreover, will these taxpayers maximize or overstate deductions in response to environmental pressures? 62 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER III RESEARCH HYPOTHESES AND METHODS To compete effectively in a business environment, an entrepreneur may develop an evasive tactic which he or she might test among trusted friends or acquaintances who disperse the tactic. Business pressure may embolden some hesitant informal group members to risk adopting the tactic whether they are cognizant of risk factors or not. This grouping or cluster may view themselves as a competitive grouping or stand-alone network of friends or acquaintances within the given environment. Below are presented two research hypotheses that, while they seem to be simply restatements of common sense observations of phenomena, lack fundamental evidence to ensure their viability as factual statements. This is especially true when applying these hypotheses to a sample of California-based truckers. The challenge here is essentially threefold: (1) using self-reported expenses from the Form 1040 Schedule C, designating some expenses as indicators of business and others as social pressure variables, take a relatively dirty sample from the 1RS Master File and iteratively reduce the population size to a representative set of truckers who relay identifiable operating and living costs that can be associated to the occupation; 63 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (2) discover if there are clusters of noncompliance in a these groupings in the context of a select set of California communities; and (3) attempt to identify if these groupings exist as solitary units within the community or are postured along other groups that could represent a basic form of competition (both from business and social standpoints). The research design identifies the single dependent and two types of independent variables that are used in the analysis. The latter grouping of independent variables is divided into business and social contexts. Finally, the research procedures are described that are designed to address the challenges described above. First, the initial sample is analyzed using traditional statistical techniques. Second, a hierarchical cluster analysis is performed searching for groupings of noncompliance in various California communities without consideration for compliance. Finally, linear regression is used to test for correlation between the independent variables (both sets) and the dependent variable, identifiable noncompliance. Research Hypotheses Hypothesis 1 Independent/sole proprietor clusters of California-based truckers decrease their level of voluntary compliance where business pressures are higher than their competition in other, more affordable parts of the community. 64 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. In an effort to maintain levels of income and business viability, will the owner/operator behave more aggressively when reporting expenses and deductions? The objective of this hypothesis is to determine the relationship between the self- reported items as outputs or indicators of business performance, assuming they are manipulated to some unknown degree to reduce taxable income and relative compliance levels within a community setting. This hypothesis attempts to understand the role that a competitive environment plays in compliance behavior and whether geographical clusters of truck drivers use of an evasion tactic as a tool, acting as an informal group or organization, to maintain competitiveness. Self-reported items such as car and truck expenses, repairs and maintenance, meals and entertainment, and other variables described earlier are used in the final phase of the effort to search for correlation between these elements and compliance. For the initial cluster analysis, both business and social pressures variables are used for the clustering algorithm. The data is analyzed over the course of three tax years (TY) 1999,2000, and 2001. These variables are compared with readily available descriptive data, and generalized environmental data including median income levels for truckers as described by BLS data as well unemployment levels. Hypothesis 2 Independent/sole proprietor California-based truckers may display group behavioral characteristics and decrease their levels of voluntary compliance if they operate or reside in communities 65 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. where their social ranking is below the median social ranking within their occupation. Previous research has not examined explicitly the relationship between an individual’s profession and the effects of social competition within a community or localized setting. The objective of this hypothesis, similar to hypothesis 1, is to use a Form 1040 reported item, in this case, indication of a permanent residence through home mortgage costs as well as total reported income, as possible determinants of compliance. The hypothesis attempts to reveal contextual behavior or, in this case, the truck driver trying to “keep up with the Joneses” at the expense of his or her tax compliance. Given the nature of small trucking business, this is probably one of the least likely populations to display localized social interaction since truckers tend to spend an inordinate amount of time on the road. If networks are detected using the social context variables, it should be considered an impetus for attempting this form of analysis on other occupation groups. At the same time, another possibility should be considered. In communities where tax compliance is lower than the norm, the trucker may simply be a leader or a follower within a smaller cluster that is part of a larger super communal cluster of noncompliance which is actually the accepted norm of that community. Research Design In 1978 the 1RS identified 64 potential compliance factors, including occupation and geographic location. Witte and Woodbury found that taxpayer 66 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. location appeared to be a compliance variable but it has been difficult to prove correlation which may have been caused by the researcher’s inability to stratify the sample to a reasonable grouping of taxpayers for analysis (Young, 1994). Westat (Young, 1994) found that manufacturing and trade organizations were associated with higher compliance while professional, managerial and sales occupations displayed a tendency to be less compliant. Again, these findings provide additional, though limited evidence that both location and occupation are important factors in modeling compliance behavior. The following section describes the dependent variable, identifiable compliance. Dependent Variable The dependent variable for this analysis is the individual’s compliance with his/her tax obligation as indicated by transaction codes, found on the taxpayer’s account establishing a tax module within the 1RS masterfile. By itself, a transaction code is not an indicator of willful noncompliance nor can such codes be used as determinants of a person’s aggressive interpretation of tax laws. This is a fundamental weakness when using this coding as it does not shed light on the possible human motivations of noncompliance. A potential future study could compare the types of noncompliant indicators in one occupation and compare them to another. As stated previously, certain shared characteristics that may be common to an occupation, such as a fiscal officer or lawyer used in another example, may correspond back to these code listings. 67 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. IRS transaction codes (TC) consist of three digits and are used primarily by field functions and the Service Center Campuses nationwide to maintain a history of the taxpayer’s account. The transaction codes allow the 1RS to maintain accounting controls of debits and credits and is used as the basis for pulling account oriented data for other issue related systems. These codes number from 001 to 999 but are not a sequential count totaling 999 different possible actions. In some cases, these transaction codes indicate the type of noncompliance. For example, a delinquency penalty is a computer generated assessment of returns posted after the due date without reasonable cause and for returns containing penalty interest codes. Followed by other code sequences, the indicator of noncompliance can essentially be canceled out requiring any inquiry to examine the entire taxpayer account. The sample taken for this dissertation eontains 172 distinct transaction codes. Appendix 1 provides examples of non-compliance transaction code definitions that are prevalent within the sample excluding normal accounting tracking codes used by the 1RS. For the approximately 34,000 owner-operators within the original sample draw over the three tax-year periods (1999,2000, and 2001), there are approximately 280,000 transaetion code entries within the IMF or 8.2 entries per account per tax year. It is important to note that transaction codes are simple account indicators which provide a history of transactions within the taxpayer’s account, including both positive and negative events (e.g., refimd paid, or refund held). 68 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Another major aspect of this study will be to identify social networks or a possible competitive environment among owner/operators. The aspect of this study involving individual pride and competition to “keep up with the Joneses” will be explored in correlation with compliance behavior among groupings. The individual’s tendency to form social clusters based on this competitive drive could be demonstrated either as clusters of compliance or noncompliance within a limited geographic area. The contextual arena in which the taxpayer exists will be driven by individual indicators of personal wealth and success and the very nature of the population (e.g., poverty stricken vs. affluent) that surrounds them. Independent Variables Figure 2 is a flow diagram of independent variables and their potential impacts on compliance behavior. Included within the flowchart are demographic variables that are not incorporated in this analysis. However, the impact that demographics have on compliance behavior and their affect upon the independent variables within this analysis cannot be ignored. Figure 2 could easily be expanded to incorporate white collar crime, the perceived level of honesty of local officials, police presence, and innumerable other environmental factors that are not easily quantified or captured within conventional databases. Simply attempting to use demographic variables is a difficult political challenge for the tax administrator given the difficulty that analysis of distinct occupations presents. The use of demographic variables, as a compliance tool, would be disastrous. 69 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CD ■ D O Q . C g Q . ■ D CD C/) C/) 8 ( O ' 3. 3 " CD CD T3 O Q . C a o 3 T3 O CD Q . Social Pressure Variables; income Housing Costs Medical/ Dental Other taxation Compliance Outcome Spectrum of Evasion Environmental Factors: Median Income ■ Community) Unemployment Crime Uroanization Business Pressure J Demographics: Variables: 1 i Car and Truck 1 Legal and Professional Education Other Expenses 1 Meals and E xpenses Language Repairs \ ....................... .............. ►Personal Debt Depreciation Criminal Record ■ D CD C/) C/) Figure 2. Independent Variables and their Impact on Compliance. Social ranking independent variables are listed in Table 1 by their relative position on the Form 1040 and Schedule C. Urbanization (e.g., level of housing density: urban, suburban, small city, town, rural) was considered but it was decided not to incorporate urbanization since many of the towns and cities captured within the sample are part of a greater megalopolis. This position is reinforced by Putnam (1993) who demonstrated that urban centers both in the northern and southern regions of Italy did not necessarily play a role in the level of civic participation. The variables are intended to portray the self-reported business pressure experienced by the owner/operator. To provide context, the truck driver’s physical location is identified using the standard five-digit zip code. A basic constraint on this study is the use of possible taxpayer identification information requiring strict measures to ensure privacy protection required under Internal Revenue Code. Therefore, for this purpose, zip codes are limited to the traditional five-digit code, although the four digit suffix is of course, maintained on the 1RS Entity data base (taxpayer identification data). 71 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 1 Business Social Pressure Variables (Trucker-Claimed Income and Expenses) INDEPENDENT VARIABLE* Location of Form 1040, and Schedule C Verifiable by match or audit? Variable Name Business Pressure (Expenses Reported on the Schedule C) Car and Truck Expenses Line 10-From Schedule C Audit only CARTRUCK_C Legal and Professional Services Line 17-From Schedule C Audit only LEGAL_C Other Expense Line 18-From Schedule C Audit only OTHER_C Meals and Expenses Line 24b - From Schedule C Audit only M&E_C Repairs Line 21 - From Schedule C Audit only REPAIR, Depreciation Line 13 - From Schedule C Audit only DEPREC_ Travel Line 24a - Form Schedule C Audit only TRAVBL_C Social Pressure Variables Total Income Line 22 Calculation - Audit only TOTAL_INCOME Medical and Dental Expenses Not Paid by Insurance Line 01 - From Schedule A Audit only MED_DENTAL State/Local Income Taxes Line 05 - From Schedule A Match Form W- 2,1099-G, 1099- R, 1099-MICSC, 1040ES STATE LOG INC _TAX Home Mortgage Line 10-From Schedule A Match Form 1098 HOME_MORT *Reflects the structure of Form 1040 for Tax Year (TY) 2001 72 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Average fuel cost was considered a primary independent variable for inclusion in the analysis. Logically, this is one of the trucker’s greatest expenses, and California maintains some of the highest at-the-pump fuel costs in the United States. At the same time, given the distance a trucker must travel, it is likely the trucker will “shop around” for fuel where possible, making planned fuel stops at the lowest-cost stations along a route. Therefore, while fuel costs are a significant burden on the small firm trucker, localized costs cannot be used as a possible determinant for noncompliant behavior. In addition, the stability and availability of localized data for average fuel costs is not easily obtainable. To realize the environmental context from which the truckers work and live, four additional data points were included in the analysis. Table 2 identifies the second set of extemal-to-IRS independent factors that may drive competitive behavior. These factors are generally fixed for an entire community and cannot necessarily be described as “variables,” especially when they are being applied to different clusters within the same city. This is a limitation because these datasets are not disaggregated to a degree necessary to make them usable at the local level, especially where the analysis deals with small to medium cities with populations above 50,000. 73 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 2 Environmental Factors INDEPENDENT VARIABLE* DATA SOURCE TYPE OF DATA Average Income Bureau of Labor Statistics (BLS) for calendar years 1999, 2000, and 2001 Available for certain cities/regions within California for entire counties or population centers above 25,000. Unemployment Levels BLS Available by cities with populations greater than 25,000 Crime Levels Department of Justice, Federal Bureau of Investigation: “Sourcebook of Criminal Justice Statistics” Available by cities with populations greater than 25,000 Population Census Data Level of population concentration *Reflects the structure of Form 1040 for Tax Year (TY) 2001 Research Procedures The research procedures consist of three primary components: (1) a descriptive phase; (2) a reduction phase using hierarchical cluster analysis and, (3) linear regression phase (compliance as a localized phenomenon). First, the descriptive phase is performed on each primary independent variable, searching for consistency and reliability factors. A table is constructed to identify the arithmetic means for social ranking and business pressure variables. 74 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Second, to identify possible social networks or groupings of people within the sample, hierarchical cluster analysis is used as the overall statistical technique. Cluster analysis is generally used in marketing studies and in many cases is dependent upon census data similar to this study. In addition, studies in law enforcement have focused on hot spot analysis using cluster analysis to identify high crime areas for targeted enforcement activities. Recent 1RS contracted studies using neural networks have used cluster analysis within the methodology to identify groupings of high-income taxpayers. This technique was used to discover systems of organizing observations usually based on human social factors which can be interpreted as groupings. Moreover, it is simpler to predict a person’s behavior if the properties or attributes of their group memberships are known (Stockburger, 1998). The advantage of cluster analysis over discriminant function analysis (as well as other techniques) is that it classifies unknown groupings while the latter is reliant upon the researcher pre identifying group memberships. There are four steps in performing a normal cluster analysis: (1) data collection and selection of variables for analysis; (2) generation of a similarity matrix; (3) decision about number of clusters and inteipretation; and (4) validation of cluster analysis. Consequently, there have been a number of concerns over the use of cluster analysis for market based segments. Carmone, Kara, and Maxwell (1999) identify specific issues with the methodology including: (1) the choice of an appropriate clustering algorithm; (2) the selection of variables; (3) the number of clusters; (4) cross-validation; (5) variable standardization; and (6) external 75 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. validation. The greatest challenge has been determining the proper set of variables for clustering where multiple options exist (Carmone et. al., 1999). Another issue is the question of how to pre-elect variables for clustering (or weighing them) from a larger set of similar variables. Carmone et al. identify the criticality of identifying true subsets of variables: With a given data set, it may happen that one variable is relevant for one grouping structure but not for others. It is also possible that some variables constitute only noise, which will mask the clear structure portrayed by other variables and yield less homogenous clusters. Therefore, there is general agreement that selecting the appropriate variables is crucial for the success of the cluster analysis, (p. 502) Given the nature of the data (primarily interval), linear regression is used in the third phase to search whether the identified clusters could possibly be explained by the variables identified within the Schedule C. If indeed a cluster demonstrates a high R squared value as well as the appropriate statistic of significance (less than .05 probability), then the cluster is descriptively compared to the BLS and census data. To accept or reject both hypotheses will require that a minimum of 90% of the clusters chosen from clusters display correlation between social or business pressure variables and the dependent variable compliance. Data Analysis The raw tax data equivalent to approximately 34,000 tax accounts or individual taxpayer records for TY 1999, 2000, and 2001 was refreshed in November 2002 from an original 1RS masterfile extract conducted by an 1RS authorized 76 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. mainframe programmer. The original sample included TY 1998,1999, and 2000 data which were performed in January 2001. Masterfile cases are renewed every tax year with new data “aging off’ the earliest year in three year cycles. The data was downloaded via a secure dial-in connection in ASCII text format to CD-ROM medium. There are two primary downloads performed for the analysis. The first download contained the Entity or case information that is transcribed from the taxpayer’s forms. The second download, which was also refreshed in January 2002 to include TY 2001 data, contains the specific account history by transaction code. It is important to understand that, for account purposes, the tax year is an artificial segmentation. Accounts that were determined to be noncompliant during one tax year may not be resolved for a number of years or as limited by the statute. Moreover, account histories are always active offering the possibility that subsequent payment activities may have occurred by the taxpayer to take corrective action for any mistake discovered by the system. The test files were downloaded via secure personal computer to a Microsoft Access database that was divided into three primary tables corresponding with the three tax years. The data was scrubbed for outliers, including cases going back to TY 1990. A query was developed combining the transaction data with each corresponding tax year. As described previously, a special table was developed assigning a compliance indicator for each of the 172 transaction codes found in the sample and incorporated into the query. Several iterations of queries were performed 77 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. to ensure that each case was designated as either being in compliance or noncompliance. The completed queries were moved into SPSS version 11.5 for Windows for the descriptive, grouping, and correlation phases of the analysis. 78 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER IV RESEARCH FINDINGS Descriptive Analysis Initial analysis was performed on the overall raw sample for each tax year. The reported mean income as shown on line 22 of Form 1040 for the three tax years is displayed in Table 3. The data shows wide ranges between minimum and maximum reported income. It should be noted that in cases where extreme amounts of income were reported, it was observed that a majority of the income loss or profit was attributable to investments as reported on the Form 1040, Schedule D. The U.S. Department of Labor, Bureau of Labor Statistics, for calendar years 1998,1999, and 2000 (equivalent to TY 1999,2000, and 2001) identify covered employment and wages for all California industries statewide as $35,348, $37,577, and $41,186, respectively. Considering the wide variance in income, the skewness and Kurtosis statistic for all three years demonstrates that the data, as it stands, is not normal. While this is typical for many 1RS datasets or large population sets overall, additional data reduction must occur to hone the sample. This reduction requires a closer examination of the nature of the self-reported items. 79 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 3 Reported Income on Form 1040 Tax Year 1999 2000 2001 Sample Size (N) 33,971 34,093 34,120 Mean $27,339 $29,366 $28,691 Standard Error of Mean 263.37 275.58 206.55 Standard Deviation 48524.1 50884.3 38153.1 Range $5,700,000 $5,600,000 $3,300,000 Minimum Data Unavailable Maximum Data Unavailable Skewness -9.599 33.657 -3432 Skewness Standard .013 .013 .013 Kurtosis 1428.054 2406.889 503.235 Kurtosis Standard Error .027 .027 .027 The competitiveness business variable family includes costs related directly to truck maintenance and business expenses, as well as common household expenses (e.g., medical and dental). These are not considered to be pure business expenses since each expense could incorporate other costs that only an audit would uncover. At the same time, fiscal pressures may cause the small owner/operator to maximize deductions or crossover to noncompliant behavior to maintain viability. Wage income, in many cases, is secondary income serving to complement Schedule C income for most owner/operators. However, in many instances, owner/operators may lease their services to larger firms. In these cases, owner/operators report both Schedule C and wage income generated from an 1RS Form 1099. Actual truck driver income varies by location throughout California. The 2002 Directory of California Local Area Wages provides a range of hourly 80 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. income from approximately $8.00 an hour for entry level drivers to a high of $25.00 for experienced drivers with three years or more experience. Taxpayer reported wages may actually reflect a population that includes small non tractor-trailer or large, straight-box drives, for which income can be considerably smaller. Another consideration is that the tax return may reflect other sources of income, including from a spouse’s income, a second job, or another home-based business. To determine the type of taxpayer, repair and maintenance costs were used as a determinant of ownership. Drivers identified within the sample that did not report repair and maintenance costs are classified as company employed truck drivers and were dropped from the sample. Price pressure of business ownership does not directly impact these taxpayers, since they do not own their own equipment. While corporate employed drivers might be an important aspect of the social dynamic among truckers, they do not necessarily compete in the same environment. Another consideration is that trucks, like automobiles, can be purchased with extensive warranty coverage. Several large truck engine suppliers such as Cummins Diesel (Cummins Bulletin 3385896, 1992)) offer extended warranty packages that can extend the engine warranty well over 250,000 miles. While it is possible that those reporting few or no repair costs were purchasers of new vehicles, it is highly unlikely. For example, in 1990, Mack Truck approved 25,000 mile crankcase drain intervals (90,000 Mile Diesel Engine Oil Changes, Internet, 2003) up from 15,000 and 20,000 mile intervals formerly recommended under on-highway conditions. Many truckers 81 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. are more prudent, staying with traditional oil change intervals, believing that this will incur fewer long-term repair costs. A typical over-the-road diesel truck using fourteen gallons of oil per change, ranging from $15 for regular 15W40 to $36 a gallon for synthetic oils: Synthetic oil—14 gallons X $36.00/gallon X 3 changes in 75,000 miles = $1512.00 Regular 15W-40—14 gallons X $15.00/gallon X 5 changes in 75,000 miles = $1050.00. (“90,000 Mile Diesel Engine Oil Changes,” Internet, 2003) This does not include labor and disposal costs (e.g., storage, contract pick-up costs, hazardous material disposal cost, and sludge disposal costs). Considering that truckers average approximately 100,000 miles a year, and oil changes are only one example of costs associated with doing business, it would be reasonable to exclude those not reporting repair and maintenance costs from the sample. This exclusion reduces the original sample size from approximately 34,000 to 17,000 owner/operators for each tax year. Taking normality into consideration, the refined sample for each tax year was scrutinized for skewness and kurtosis statistics. Based on the skewness statistic(s) it was considered prudent to develop a histogram of the total income data for each year to provide a graphic understanding of the nature of the distribution. The data was skewed to the right and represents an abnormal distribution of total income for the sample. In addition to the descriptive statistics for the total income variable, further 82 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. histograms and dichotomized variable correlations were performed to determine the viability of all the data within the masterfile dependent variable sample. Unfortunately, the simple examination of skewness and the histograms indicates that the distributions of the data remain non-normal. Consequently, conventional statistical approaches that depend on data normality have been brought into question, making it extremely difficult to draw any reliable conclusions based on the data as it is stands (Kerlinger, 1986, p. 267). Issues with Nonparametric Statistical Techniques Several paths can be followed when data distributions are not normally distributed: (1) non-parametric statistics; (2) sub group averaging; (3) segmentation of data; (4) transformation of data; or (5) use of different distributions. Indeed, there is a precedent for simply ignoring population normalcy. Micceri (Yu, 2003), in his analysis of over 400 large data sets, found that a great majority of this work in behavioral sciences did not follow normal distributions. Nonparametric techniques tend to present several weaknesses which played a role in the researcher’s decision to attempt to use transformation techniques. While Chi-square may represent a traditional statistical technique, there are several issues that can threaten the integrity of the analysis. Yu (2003) presents five distinct criticisms of nonparametric techniques: 1. Estimations of the population: Nonparametric tests do not make strong assumptions regarding the population. Ultimately, the researcher can not 83 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. make inference that the sample statistic is a viable estimate of the population parameter. 2. Degradation of precision: If other more precise measurements are available it could be detrimental simply to transform measures into ranked data. 3. Low Power: Only in a few instances do nonparametric tests provide the same statistical power of conventional parametric tests. 4. False sense of security: The false belief that nonparametric techniques are immune to the negative impact of outliers. Yu is adamant in his criticism of this weakness: “Zimmerman found that the significance levels of the WMV test and the KW test are substantially biased by unequal variances, even sample sizes in both groups are equal.” Further, non parametric tests tend to produce biased results when multiple assumptions, such as non-normality and unequal variances are violated. The presence of outliers is also detrimental to non-parametric tests. Zimmerman outliers modify Type II error rate and power of both parametric and non-parametric tests in a similar way. 5. Software issues: This weakness is beginning to be addressed within the major software packages. 6. Testing Distributions: Yu (2000, p. 3) provides the example of whether procedure tests for two distributions are different in some manner but do 84 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. not display how they are different in regard to means, variance, and shapes which form the basis for his preference towards transforming data. Based on Yu’s observations, the sample was truncated to eliminate high positive and negative income outliers. Table 4 represents descriptive statistics for the sample which still includes a small number of high negative income cases. Figures 3,4 and 5 are the histograms for the three tax year of total income data. Table 4 Total Income of Owner/Operators— Modified Sample Tax Year 1999 2000 2001 Sample Size (N) 16,327 16,652 17,357 Mean $27,835.72 $28,665.14 $28,785.58 Standard Error of Mean 186.18 193.11 181.37 Standard Deviation 23789.907 24918.810 23894.864 Range $296,000 $299,000 $290,000 Minimum Data Unavailable Maximum Data Unavailable Skewness .904 .770 .935 Skewness Standard Error .019 .019 .019 Kurtosis 4.283 4.827 4.342 Kurtosis Standard Error .038 .038 .037 While not completely normalized from the reduction of outliers, the data is closer to a normal distribution than the original sample. TY 2000 and 2001 data approximates this distribution. 85 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UJ g I — 0 3 C /5 1 i a i I H I g : I 86 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 Z UJ O h- iü î I o i a i I H I & I 87 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. s 5 (û ^ ^ 8 Il 00 N > CM ^ g " g I I § I § o .9 V y i i I t» I o i I a I H e I 88 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 5 illustrates car and truck cost descriptive statistics which have been artificially ranged from $0 to approximately $250,000. The original sample contained approximately 15,000 cases that reported $0 on their tax returns on line 21 of the Schedule C. This range was chosen since $500 approximated a quarter-year of first-year ownership repair and maintenance expenses based on data from First Fleet (2003). Table 5 Car and Truck Expenses— Modified Sample Tax Year 1999 2000 2001 Sample Size (N) 16,327 16,652 17,357 Mean $12,899.75 $14,312.03 $14,435.17 Standard Error of Mean 146.48 155.39 151.74 Standard Deviation 18716.939 20052.084 19990.976 Range $244,000 $242,500 $246,000 Minimum Data Unavailable Maximum Data Unavailable Skewness 3.653 3.085 3.151 Skewness Standard Error .019 .019 .019 Kurtosis 22.821 15.957 16.877 Kurtosis Standard Error .038 .038 .037 The histogram (Figure 6) provides a visual representation of the data for the TY 1999 repairs and maintenance variable and is representative of the other tax years within the sample. As with total income, the distribution is skewed to the right. However, the non-normal characteristics of the distribution are not as pronounced as with the pre-stratified example. 89 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. s < 6 0 0 Il < M & M 8 " S I I Z % I I « T > § « u I v-4 & \d I 90 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. The car and truck expenses sample, line 10 of the Schedule C, is used for taking the standard mileage rate or reporting actual expenses related to gasoline, oil, repairs, insurance, tires, license plates, etc. There are two basic scenarios for using the standard mileage rate: (1) the taxpayer must own the vehicle and use the standard mileage rate for the first year the vehicle is placed in service, or (2) the taxpayer leases the vehicle and is using the standard mileage rate for the entire lease period of the vehicle. If the taxpayer is deducting actual expenses, then he or she can deduct the expenses described above. Truck Ownership Lifecycle Costs and Regulatory Requirements As previously discussed using data from one major engine manufacturer, the owner/operator faces price pressure related to the maintenance and repair of his or her primaiy capital investment. The major large-truck leasing corporation. First Fleet (2003), estimates that repair and maintenance costs naturally increase over a five-year lifecycle. Table 6 provides First Fleet's estimate for a five year lifecycle. Another estimate of actual mileage that owner/operators drive is approximately 100,000, a relatively small difference in comparison to FirstFleet's estimate. To minimize or offset costs, owner/operators may try to optimize the use of their trucks. Berwick and Dooley (1997) argue that owner/operators attempt to maximize economies of utilization: 91 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 6 Annual Costs of Repairs and Maintenance Years in Service Average Miles of Usage Annual Cents /Mile Annual Owner's Cost 1 110,000 Z31 $2,541 2 220,000 4.11 $9,042 3 330,000 5.63 $18,579 4 440,000 8j3 $36,652 5 550,000 9.41 $51,755 Source: FirstFleet Inc. (2003). Finding the sweet spot. In Patricia McCullough Smith (Ed.), Heavy Duty Trucking. Retrieved March 11,2003, from http://www.firstfleet.co/heavydt.htmI. Cost ininimization for the owner/operator encottrages high usage of equipment. The concept of economies of utilization is allocation of fixed costs over increased output and is realized by increasing the use of those fixed assets. Fixed costs are short-run costs that cannot be avoided and do not vary with output. Variable costs change with output and may easily be changed. The distinguishing characteristic between fixed and variable cost is time. In the long run, all costs are variable, or can be changed, (pp. 13-14) There are practical limitations to the number of miles a driver can accrue during the course of a year from both a practical and legal standpoint (driver hours are limited by federal regulations). Federal statute states that no driver shall work more than 70 hours in eight days or drive more than 10 hours in one day (Berwick & Dooley, 1997). In addition, there is additional work time that is not included within the 10- hour limitation. For example, fiieling, loading and unloading are classified as work time. Overall, drivers are limited to a total of 15 hours of duty but can only drive 10 92 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. hours of the 15. Moreover, drivers must take at least 8 hours of off duty time between shifts. The only practical way for the owner/operator to increase the utility of the investment is by adding a redundant driver (sometimes a spouse). Variable costs include repair and maintenance, fuel, labor, and tires. Fixed costs include equipment costs, license fees and taxes, insurance, and general overhead. Utilization is the most important factor impacting owner/operators (Berwick & Dooley, 1997). Other Fixed Costs An important fixed-cost price pressure on the owner is whether he or she is maintaining a lease on the vehicle. The class eight (tractor-trailer) is considered a 3- year property for depreciation purposes. 1RS Section 179 allows the taxpayer to claim a high level of depreciation during the first year of use. If the owner/operator leases, actual payments are deducted. To determine whether this price pressure exists within the California sample. Schedule C line 13 expenses were reviewed. Table 7 provides descriptive statistics for depreciation derived from line 13. Unfortunately, line 20a for rented or leased vehicles, machinery or equipment is not transcribed from the Schedule C and would require a full review (audit) of a sample of paper returns, which is not practical for the objectives of this analysis. In addition, business insurance (another fixed cost) was also considered a viable expense to the small owner/operator. However, this expense is not transcribed from the Schedule C, again requiring a manual review of returns. 93 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 7 Depreciation— Modified Sample Tax Year 1999 2000 2001 Sample Size (N) 16,327 16,652 17,357 Mean $6,539.57 $7,085.01 $6,851.12 Standard Error of Mean 135.05 143.87 241.39 Standard Deviation 17255.915 18565.047 31802.812 Range $1,100,000 $1,300,000 $3,800,000 Minimum Data Unavailable Maximum Data Unavailable Skevyness 25.005 28.456 98.379 Skewness Standard Error .019 .019 .019 Kurtosis 1238.148 1602.422 11667.370 Kurtosis Standard Error .038 .038 .037 Legal and Professional Costs Legal and professional expenses captured on line 17 of the Schedule C are characterized as management and overhead costs by Berwick and Dooley (1997). These costs tend to be short-run fixed costs that are designed to capture tax preparation related expenses. Consequently, they may represent an indicator of compliance, providing some insight into the owner/operator’ s viability, and they are characterized as a dependent variable of compliance to be used with the earlier identified compliance variable based on 1RS transaction coding. Table 8 represents reported legal and professional expenses as reported within the sample. The histogram in Figure 7 reflects the legal and professional expenses for TY 1999. The unusual shape of the legal and professional histogram degrades the dataset’s 94 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. usefulness since it still contains a number of outliers as well as S O values. The $0 is a legitimate entry, since many taxpayers self-prepare and therefore should not be excluded from the sample. In an attempt to remedy this issue, the sample will have to be exposed to multiple rounds of reduction to eliminate cases that are either outliers or other anomalies. Table 8 Legal and Professional Expenses Tax Year 1999 2000 2001 Sample Size (N) 16,327 16,652 17,357 Mean $451.52 $409.60 $398.04 Standard Error of Mean 19.46 10.94 9.75 Standard Deviation 2486.212 1411.433 1284.540 Range $180,000 $81,000 $42,000 Minimum Data Unavailable Maximum Data Unavailable Skewness 50.565 20.297 14.052 Skewness Standard Error .019 .019 .019 Kurtosis 3452.539 797.322 315.641 Kurtosis Standard Error .038 .038 .037 95 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C M C O C O " 4" to o C M o I I N È I I c s M 1 I I z I I !■ ! o\ tH K I 96 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Additional issues related to reporting legal and professional services should be considered. This particular expense should be explored for possible misreporting issues especially among self-preparers. For example, on Schedule C, line 17 the language on the tax form is not necessarily descriptive since it reads Legal and professional services. The 2001 Instructions for Schedule C, Profit or Loss from Business are simple: “Include on this line fees for tax advice related to your business and for preparation of the tax forms related to your business.” If the taxpayer fails to read the instructions and simply interprets legal and professional services at face value, costs reported on this line could include actual legal expenses related to the profession or even where there is an impact on their capital expenditures (e.g., speeding fines. Department of Transportation fines, divorce representation, etc.). A line-by-line analysis of legal expenses for TY 1999 reveals some interesting observations. The maximum expenditures, in two instances, are approximately $180,000 and while they are not visibly related (different SSN and address), both accounts displayed the same car and truck, other C, repairs, and total deductions. These outliers (probable noncompliance) aside, the data reveals that a full 36.2% do not report any legal costs at all and a cumulative 80% of the sample report $500 or less (99% of the cases report $5,500 or less). The question that must be answered: is what are reasonable and typical “Legal and Professional” costs for the average owner/operator? Naturally, this question can be applied to any of the variables. As with repairs, tax preparation and in some instances, accountant and legal expenses (if 97 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. they are related to taxes only) are presumably proportional to the size of the small business operation. A cursory review of total income and legal expenses provided no reasonable indicator of correlation. (Pearson Correlation is .029, which is of minor significance at the 0.01 level [2 tailed].) A review of commercially available data on long-distance freight and specialized freight revealed that legal and professional services account for 0.4% of total revenue (Biznotes.com, 2002). An attempt to stratify the data in relation to gross income on line 7 of the Schedule C (1999) was found to be untenable, since this data is available by individual audit only. An additional effort was made to correlate legal expenses against line 31 (Net profit or loss, 1999) which is also audit only and line 48 (Total other expenses, 1999). Both efforts yielded unsatisfactory results. In a final effort to eliminate additional potential outliers, all cases that had negative reported income were eliminated. This step was taken in preparation for the geographical analysis that follows. Geographic Analysis The overall sample size for all tax years was reduced again based on geographical location of the taxpayers clustered by zip code. Using Arcview Geographical Information Software (GIS), the sample for TYs 1999,2000, and 2001 were reduced to 4,549; 4,718; and 4,191 taxpayers, respectively. The zip code clusters retained within the sample representing concentrations of small business truckers were then broken down by three groupings: (1) small to medium cities with 98 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. multiple zip codes equating to more than 1% of the total number in the sample; (2) groupings of 3 or 2-3 zip codes that equate to 0.5% of the total sample; and (3) single zip codes that equate to 0.4% or more of the total sample. The base year for cluster identification is TY 1999, which is retained for TY 2000 and 2001, since massive taxpayer migration was not noted. Because of the size of the immediate Los Angeles area and its underlying social complexity, only outlying communities were taken for the sample. These included minor cities and bedroom communities outside of Los Angeles were used for the analysis. A hierarchical cluster analysis was then performed in SPSS version 11.5 for using the variables described for social rank (e.g., general business expenses and family income pressure). It is possible that, within the larger clusters, two or more groups of geographically co-located noncompliant mini clusters may exist. Concurrently, several taxpayers inside the cluster may share reporting characteristics that are coincidental and are not physically co-located within a geographical cluster, or are located far from the center of the cluster. If the analysis was simply focusing on taxpayer reporting, the sample size could stand as is. However, to determine the element of social/economic competition among co-located operator/owners, it is essential that the sample focus only on those taxpayers who are geographically co located within the cluster. To find these mini-clusters a manual stratification of each cluster for all tax years was performed. 99 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. To limit the sample to possible social clusters of owner/operators additional criteria were implemented to screen out implausible groupings of taxpayers. First, clusters of three or fewer owner/operators were dropped from the sample unless all three were geographically co-located within the same city or town (not necessarily the same zip code). Second, clusters where fewer than 50% were designated to be non compliant (using of 1RS transaction coding) were eliminated unless there was two or more geographically collocated mini-clusters within the larger cluster. Again, it should be noted that transaction coding is a result of the use of 1RS discriminant frmction (DIF) formulas, which as a system, is not designed to identify clusters of taxpayers or derive intent. Furthermore, the cluster analysis was intended to identify groupings of taxpayers that might very well represent successfid innovations of evasion that are not easily identifiable by DIF. The third criteria involved the screening of three self-reported items: (a) “Car and Truck” expenses reported above $30 thousand; (b) the presence of a home mortgage; or (c) “Other” Schedule C expenses above $30 thousand. Clusters were retained in this round of stratification only if a large majority, greater than 75%, reported these expenses based on the dollar criteria described where applicable. Based on the above criteria, the sample was finally reduced to 4,132; 3,639; and 4,025 for TY 1999, 2000, and 2001, respectively. As a result of the initial cluster analysis and review/removal of some cases from the sample, another cluster analysis was performed on the remaining cases using the same specifications. The clusters throughout all sample tax years were grouped 100 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. together as an assortment of taxpayers from different cities and communities based on similar variable patterns. While this would be useful if the intent of this work were to market a product to a specific group, it does little to support hypothesis testing. Moreover, it does not provide the necessary background to identify and analyze social/business competition or the possible diffusion of evasion behavior. Geographic-Based Dataset To establish the necessary dataset, the taxpayer’s geographic dispersion was again analyzed relative to (1) the number of clusters and their population within a community or city; (2) the proximity of the community to a major highway; and (3) median income for the community versus the median income of the trucker owner/operators. An important concept that has not been discussed previously is how truckers might communicate information with others in their occupation or disperse new ideas to maintain competitiveness. Given that their profession is built around mobility on roads and highways this medium might serve as the basis for dispersion of information among truckers. Hence, the proximity of the community to a major highway was used as the basis for developing the final dataset. Seven cities were chosen along California Highway 99 as well as five communities along Interstate 710 in the Los Angeles area. The seven cities along the California Highway 99 corridor, from north to south, are Yuba City, Sacramento, Stockton, Modesto, Turlock, Fresno, Tulare, and Bakersfield. For the Interstate 710 corridor, specific 101 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. data was collected on the communities of Maywood, Bell, Bell Gardens, Compton, and Cudahy. The hierarchical analysis relies on the same parametric assumptions as other conventional statistical techniques (e.g., normal frequency distributions). Again, all of the variables identified to represent business and social pressures have relatively high variances, with a large number of zero values that result in non-normal frequency distributions. As presented without transformation, there is little correlation among the variables, even when nonparametric techniques (e.g.. Chi-square) are applied to the data. Efforts to transform data using the Log-fimction: did not result in satisfactory X-Log(x/mean(x) + k) where k = < 1 frequency distributions for the raw data, even stratified by community. This Log function is traditionally used where there are 0 values for variables. It should be noted that hierarchical cluster analysis is dependent upon Gaussian type of frequency distribution. SPSS provides two important options that assist in addressing the issues discussed above: (1) variable transformation (e.g., range 0 to 1) and (2) the use of Pearson Correlation for interval data instead of traditional Euclidean distance. The cluster analysis was performed by case, not variable, and the compliance nominal variable was excluded since it provided misleading results based on the initial cluster tests. Moreover, it was not desirable to have exclusive clusters 102 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. of completely non-compliant and compliant taxpayers. Again, the issue of IRS’s ability to detect and identify non-compliance is a point of contention. It is postulated that, if indeed there are clusters of taxpayers who have adopted non-compliance as a strategy to cope with either social or business pressures, then there are other taxpayers “waiting in the wings” who have not joined the bandwagon or have joined but have not yet been detected by the 1RS. Therefore, a cluster that is purely non-compliant might not offer the same level of diversity that is necessary to identify the attributes that drive non-compliance. Other SPSS options included the development of agglomeration schedule and proximity matrix for each city for each tax year. Cluster membership was limited to a single solution of 15 clusters (an arbitrary choice). A between-group linkage was used as the cluster method. Finally, vertical dendrograms were developed for each city-based cluster. Appendix 5 contains a sample dendrogram for TY 1999 Compton. Tables 9,10, and 11 represent the data for each city and tax year providing basic data as well as socio-economic external data and represent the final sample size for these communities. The final sample was selected based on the prevalence of economic factors, including 90% or more reporting car and truck repairs and 90% home mortgages. Secondary criteria includes 50% or more of the taxpayers reporting other C expenses, depreciation, legal expenses, medical dental expenses, meals and entertainment, repairs, state and local taxes and travel. In the iterative method of data reduction, the final step was developed to ensure the sample was established to 103 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. capture social pressure variables. Unfortunately, the method eliminates many owner/operators who do not own their own homes but are potentially noncompliant. Future analysis should pay close attention to business pressure variables only, since the expenses or limitations on mobility of owning a residence in California may simply be too much for owner/operators, whether they are honest or evasive. If the clusters found within the final sample base met were comprised of 15 or more cases, linear regression was applied to the individual cluster using the social rank and family income data variables (1RS data) within the model. As a result of this final specification, several smaller communities did not have a sufficient population within a single cluster and were excluded from further analysis. However, Appendix 2, Final Cluster Notes, provides observations on all clusters with unusual characteristics. This may include clusters with only two cases yet contain unusually high OTHER C, REPAIRS C or other expenses combined with identified noncompliance. While these cases may not be candidates for statistical analysis as stand-alone members of geographical clusters, they potentially could represent fragments of a larger super cluster whose origins are not from the base community. Further, they may represent the remains of an unsuccessful noncompliance innovation that has begun to dissipate. More importantly, they may represent detectable portions of a cluster that was part of a larger successful evasive cluster of taxpayers eliminated by the data reduction methodology. 104 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C D " O O Q . C g Q . " O CD C/) W o " 3 O 3 CD 8 3. 3 " CD CD " O O Q . C a O 3 " O O CD Q . " O CD C / ) C / ) Table 9 (Part 1) Tax Year 1999 External and Internal Data for Test Cities (1) - Bureau of the Census results from the 2000 Census (2) = FBI Crime Report (2002) City Name Yuba City Sacramento Stockton Modesto Turlock Fresno Tulane External Data Population (I) 36,758 407,018 243,771 188,856 55,810 427,652 43,994 Median Income One ($)(!) 32,858 37,049 35,453 40,394 39,050 32,236 33,637 Percentage of Population at or 14.5 15.3 18.9 12.2 12.4 20.5 16.9 Crime Statistics (2) 2102 27,540 16,681 10,240 2971 29,577 2797 Compliance 0.44 0.47 0.56 0.55 0.53 0.58 0.57 Total Number of Taxpayers 75 222 150 128 66 180 30 Variables (Approximate Mean Value) Income 2 25,200 23,800 27,800 27,000 27,000 26,000 29,000 Home Mortgage 2600 2300 2500 3100 3700 2200 3300 Car and Truck 8200 13,000 15,000 10,000 6500 7000 8300 Meals and Entertainment Data Unavailable Other C Travel Medical/Dental Repairs Depreciation Legal Schedule C Wages State & Local Income Tax C D " O O Q . C g Q . " O CD C/) o " 3 O 8 ci' 3 3 " CD CD " O O Q . C a O 3 " O O CD Q . " O CD C /) C /) Table 9 (Part 2) T oa: Fear External and Internai Data for Test Cities (1) = Bureau of the Census results from the 2000 Census (2) = FBI Crime Report (2002) City Name Bakersfield Maywood Bell Bell Gardens Compton Cudahy Long Beach External Data Population (1) 247,057 28,083 36,664 44,054 93,493 24,208 461,522 Median Income One ($) (I) 39,982 30,480 29,946 30,597 31,819 29,040 37270 Percentage of Population at or 14.6 23.1 21.2 25.3 25.5 26.4 Crime Statistics (2) 11,320 588 939 1399 4903 535 18,372 Compliance 0.56 0.63 0.63 0.65 0.58 0.78 0.55 Total Number of Taxpayers 188 48 46 71 137 36 284 Variables (Approximate Mean Value) Income 2 23,000 18,600 18,400 18,600 18,800 18,600 20,600 Home Mortgage 2400 3100 900 1600 3100 300 2400 Car and Truck 8800 11,000 13,500 11,800 12,500 10,500 10,000 Meals and Entertainment Data Unavailable Other C Travel Medical/Dental Repairs Depreciation Legal Schedule C Wages State & Local Income Tax g C D " O O Q . C g Q . " O CD C/) o " 3 O 8 CD 3. 3 " CD CD " O O Q . C a O 3 " O O CD Q . " O CD C /) C /) Table 10 (Part 1) Tax Year 2000 External and Internal Data for Test Cities City Name Yuba City Sacramento Stockton Modesto Turlock Fresno Tulane External Data Population (1) 36,758 407,018 243,771 188,856 55,810 427,652 43,994 Median Income One ($)(!) 32.858 37,049 35,453 40,394 39,050 32.236 33,637 Percentage of Population at or 14.5 15.3 18.9 12.2 12.4 20.5 16.9 Crime Statistics (2) 2056 27,705 16,991 10,892 3200 33.332 2483 Compliance 0.70 0.71 0.74 0.73 0.68 0.70 0.58 Total Number of Taxpayers 92 241 148 120 71 179 26 Variables (Approximate Mean Value) Income 2 26,600 27,000 28,000 27,000 25,500 25,700 27,500 Home Mortgage 2700 2500 2600 3600 4500 2800 3700 Car and Truck 5000 10,000 12,000 8500 7000 9600 13,500 Meals and Entertainment Data Unavailable Other C Travel Medical/Dental Repairs Depreciation Legal Schedule C Wages State & Local Income Tax (1) = Bureau of the Census results from the 2000 Census (2) = FBI Crime Report (2002) C D " O O Q . C 8 Q . " O CD C/) o " 3 O 8 ci' 3 3 " CD CD " O O Q . C a O 3 " O O CD Q . " O CD C / ) C / ) Table 10 (Part 2) Tax Year 2000 External and Internal Data for Test Cities (1) = Bureau of the Census results from the 2000 Census (2) = FBI Crime Report (2002) City Name Bakersfield Maywood Bell Bell Gardens Compton Cudahy Long Beach External Data Population (1) 247,057 28,083 36,664 44,054 93,493 24,208 461,522 Median Income One ($)(!) 39,982 30,480 29,946 30,597 31,819 29,040 37,270 Percentage of Population at or 14.6 23.1 21.2 25.3 25.5 26.4 19.3 Crime Statistics (2) 10,414 633 894 1224 4973 609 17,929 Compliance 0.68 0.72 0.75 0.66 0.61 0.70 0.69 Total Number of Taxpayers 161 29 48 47 119 37 264 Variables (Approximate Mean Value) Income 2 26,300 19,700 21,300 20,400 20,500 23,000 24,000 Home Mortgage 2800 3000 1500 800 3100 600 3000 Car and Truck 9000 11,300 15,100 14,300 13,900 10,500 12,100 Meals and Entertainment Data Unavailable Other C Travel Medical/Dental Repairs Depreciation Legal Schedule C Wages State & Local Income Tax o 00 C D " O O Q . C g Q . " O CD C/) o " 3 O 8 ci' 3 3 " CD CD " O O Q . C a O 3 " O O CD Q . " O CD C /) C /) Table 11 (Part 1) Tax Year 2001 External and Internal Data for Test Cities (1) = Bureau of the Census results from the 2000 Census (2) = FBI Crime Report (2002) City Name Yuba City Sacramento Stockton Modesto Turlock Fresno Tulane External Data Population (1) 36,758 407,018 243,771 188,856 55,810 417,652 43,994 Median Income One ($) (1) 32,858 37,049 35,453 40,394 39,050 32,236 33,637 Percentage of Population at or Below the Poverty Level (1) 14.5 15.3 18.9 12.2 12.4 20.5 16.9 Crime Statistics (2) 2002 31,131 19,843 12,096 3,439 35,229 3049 Compliance 0.42 0.40 0.48 0.40 0.50 0.56 0.34 Variables (Approximate Mean Value) Income 2 24,500 25,000 29,600 27,500 28,300 28,800 32,800 Home Mortgage 2000 2600 3400 4300 3800 2600 4500 Car and Truck 4200 13,100 12,200 11,200 8000 8700 5400 Meals and Entertainment Data Unavailable Other C Travel Medical/Dental Repairs Depreciation Legal Schedule C Wages State & Local Income Tax § C D " O O Q . C g Q . " O CD C/) o " 3 O 8 ci' 3 3 " CD CD " O O Q . C a O 3 " O O CD Q . " O CD C /) C /) Table 11 (Part 2) Tkr Tear 2007 External and Internal Data for Test Cities City Name Bakersfield Maywood Bell Bell Gardens Compton Cudahy Long Beach External Data Population (1) 247,057 28,083 36,664 44,054 93,493 24,208 461,522 Median Income One ($) (1) 39,982 30,480 29,946 30,597 31,819 29,040 37,270 Percentage of Population at or Below the Poverty Level (1) 14.6 23.1 21.2 25.3 25.5 26.4 19.3 Crime Statistics (2) 10,502 643 819 1149 5412 561 18,734 Compliance 0.46 0.53 0.51 0.46 0.54 0.51 0.48 Variables (Approximate Mean Value) Income 2 25,900 18,200 20,800 18,000 20,400 22,400 22,200 Home Mortgage 2000 2700 1100 1500 3200 600 2500 Car and Truck 9800 11,600 12,400 12,200 13,000 9200 10,400 Meals and Entertainment Data Unavailable Other C Travel Medical/Dental Repairs Depreciation Legal Schedule C Wages State & Local Income Tax (1) - Bureau of the Census results from the 2000 Census (2) = FBI Crime Report (2002) Testing of the Individual Compliance Clusters In an effort to identify the explanatory power of the independent variables on known compliance, a linear regression model was used on each cluster meeting the prescribed specifications outlined above. Traditionally, the R or regression line calculation is used to demonstrate how closely data points (independent variables) bunch along the line. More specifically, the sum of the squares of the placement attributed to the regression line is divided by the total sum of the squares. High R value (the highest value is R=l) is translated to represent a more concentrated cluster of data points. However, unless R is at least 0.7, the reliability or consistency to predict the impact of the independent variables on compliance comes into question. R-squared, a stricter indicator of predictability and also more widely prescribed by statisticians is used in this analysis as a final determinant in hypothesis testing. As a result, the R value must exceed 0.7, since using the R-squared (0.49) translates only to approximately half the number of data points being explained by the regression line. In addition, the statistic of significance should not exceed 0.05, since a greater number could be attributable to normal random distribution. A summary table (Table 12) was developed incorporating the primary regression statistics as well as the descriptive statistics from the Bureau of Labor Statistics (ELS), Census, and the FBI. Clusters that meet or exceed the R-squared and statistic of significance are displayed in Table 12 as shaded areas. I l l Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C D " O O Q . C g Q . " O CD C/) o " 3 O 8 ci' 3 3 " CD CD " O O Q . C a O 3 " O O CD Q . " O CD C /) C /) Table 12 (Part 1) Primary Clusters City Name Yuba City Sacramento Sacramento Sacramento Sacramento Sacramento Tax Year 2001 1999 2000 2000 2001 2001 Cluster Number of 15 3 1 7 11 1 5 Sample Size N 21 37 36 28 33 40 R .748 .626 .581 .706 .772 .604 R Square .560 .392 .338 .499 .596 .365 Adjusted R Square .022 .159 .035 .204 .384 .144 Standard Errors of the Estimate .458 .461 .446 .348 .389 .391 Durbin Watson 2.039 2.617 1.667 2.404 2.383 1.588 Significance .484 .140 .393 .164 .020 .137 Population (1) 36,758 407,018 407,018 407,018 407,018 407,018 Median Income ($) (1) 32,858 37,049 37,049 37,049 37,049 37,049 Percentage of Population at or Below the Poverty Level (1) 14.5 15.3 15.3 15.3 15.3 15.3 Crime Statistics (2) 2002 27,540 27,705 27,705 31,131 31,131 Compliance (General Population) 0.42 0.47 0.71 0.71 0.40 0.40 (1) = Bureau of the Census results from the 2000 Census (2) = FBI Crime Report (2002) C D ■ D O Q . C g Q . ■ D CD C/) o' 3 CD 8 CD 3. 3 " CD CD ■ D O Q . C a o 3 ■ D O CD Q . ■ D CD C /) C /) Table 12 (Part 2) Primary Clusters City Name Stockton Stockton Stockton Stockton Modesto Modesto Modesto Tax Year 1999 1999 2000 2001 2000 2001 2001 Cluster Number of 15 1 7 2 3 4 1 9 Sample Size N 20 15 22 36 38 31 17 R .666 .675 .846 .630 .712 .661 836 R Square .444 A55 .715 397 .508 .437 698 Adjusted R Square -.057 -.271 .402 120 .299 .112 ^35 Standard Errors of the Estimate .525 .467 .272 469 329 .477 .430 Durbin Watson 2.187 1.781 1.921 1.607 2.169 1824 2436 Significance j6 8 J36 .102 321 .030 376 .513 Population (1) 243,771 243,771 243,771 243,771 188,856 188,856 188,856 Median Income ($) (1) 35,453 35/K3 35463 35,453 40,394 40,394 40,394 Percentage of Population at or Below the Poverty Level (1) 18.9 18.9 18.9 189 12.2 12.2 123 Crime Statistics (2) 16,681 16,681 16,991 1RJW 3 10,892 12,096 12,096 Compliance (General Population) 0.56 0^6 0.74 0.48 0.73 0.40 0.40 (1) = Bureau of the Census results from the 2000 Census (2) = FBI Crime Report (2002) u > C D ■ D O Q . C g Q . ■ D CD C/) o' 3 O 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 ■ D O CD Q . ■ D CD C /) C /) Table 12 (Part 3) Primary Clusters City Name Turlock Fresno Fresno Fresno Bakersfield Bakersfield Bakersfield Tax Year 2001 1999 2000 2001 1999 2001 2001 Cluster Number of 15 1 1 2 4 3 3 12 Sample Size N 16 31 31 16 48 31 18 R 956 503 j? 3 618 .434 253 .794 R Square .914 253 281 .189 205 261 Adjusted R Square .679 -.179 624 -1220 -.059 -.097 -.066 Standard Errors of the Estimate 290 j5 2 .273 .787 .520 .518 .501 Durbin Watson 2.124 2.271 1.625 2.615 2441 1293 2.221 Significance .101 .818 .001 978 .674 .673 278 Population (1) 55,810 427,652 427,652 427,652 247,057 247,057 247,057 Median Income ($) (1) 3%050 32236 32236 32236 39,982 39,982 39,982 Percentage of Population at or Below the Poverty Level (I) 12.4 2fr5 20.5 202 14.6 14.6 14.6 Crime Statistics (2) 3,439 2%577 33232 35229 11,320 10,502 1% 5(% Compliance (General Population) 0.50 0.58 0.70 0.56 0.56 0.46 046 (1) = Bureau of the Census results from the 2000 Census (2) = FBI Crime Report (2002) C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 ■ D O CD Q . ■ D CD C /) C /) Table 12 (Part 4) Primary Clusters City Name Compton Compton Compton Long Beach Long Beach Long Beach Tax Year 1999 2000 2001 1999 2000 2001 Cluster Number of 15 2 / 7 10 2 10 1 Sample Size N 37 18 25 36 38 48 R .531 .895 .744 .531 .673 623 R Square .282 .801 .554 .282 .454 288 Adjusted R Square .108 577 235 .069 251 .244 Standard Errors of the Estimate .477 .333 .446 .489 286 291 Durbin Watson 1903 2.621 1.429 1.691 1.624 1893 Significance .168 .043 .167 .277 .047 .016 Population (1) 93,493 93,493 93,493 461,522 461,522 461,522 Median Income ($) (1) 31,819 31,819 31,819 37,270 37270 37270 Percentage of Population at or Below the Poverty Level (1) 25.5 2 5 j 19.3 19.3 192 Crime Statistics (2) 4903 4973 5412 18272 17,929 18,734 Compliance (General Population) 0.58 0.61 0.54 0.55 0^9 0.48 (1) = Bureau of the Census results from the 2000 Census (2) = FBI Crime Report (2002) Only four communities, TY 2001 Sacramento, TY 2000 Modesto, TY 2000 Fresno and TY 2000 Compton, contained R values greater than or equal to 0.7 with a statistic of significance lower than 0.05. The City of Long Beach for TY 2000 and 2001 contained clusters for each year that had an R value between approximately 0.62 and 0.68 that did not meet specifications for a sufficient R value. This phenomenon is discussed later in this paper under temporal considerations. Table 13 provides a synopsis of the 4 communities: Table 13 Clusters of Interested Based on Regression Analysis City Name Sacramento Modesto Fresno Compton Tax Year 2001 2000 2000 2000 Cluster Number of 15 1 4 2 7 Sample Size N 33 38 31 18 R .772 .712 .873 R Square .596 .508 .762 .801 Adjusted R Square .384 .299 .624 .577 Standard Errors of the Estimate .389 .329 .273 333 Durbin Watson 2.383 2.169 1.625 2.621 Significance .020 .030 .001 .043 Population* 407,018 188,856 427,652 93,493 Median Income ($)* 37,049 40,394 32,236 31,819 Percentage of Population at or Below the Poverty Level 15.3 12.2 20.5 25.5 Crime Statistics (Occurrences) 31,131 10,892 33,332 4973 Compliance (General Population) 0.40 0.73 0.70 0.61 *As identified by the 2000 Census R-values for each city in Table 9 exceed 0.7 with R-square values ranging from 0.596 to 0.895. It appears that, within these four clusters, the socio-economic variables are 116 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. at least a 50% determinant of compliance. Normally, an analysis of residuals and correlations would be in order for additional evidence that the variables impact levels of compliance. Examples of the correlation table and residuals for both Fresno and Compton are provided below from Appendix 3. These charts (Tables 14,15,16, and 17) were chosen because of the vitality of both R-Squared scores. It should be noted that the variable names were adjusted based on their log transformation. Referencing correlation Tables 14 and 16, there is the expected marginal correlation between the independent Income2 variable and Comp2 (compliance) dependent variable, 0.725 and 0.619 for Fresno and Compton respectively. lncome2 and State2 variables (formerly STATE LOC INC—state and local taxes) for Fresno are correlated with a value of 0.723, which should not be considered unusual. Two anomalies appear on Table 11 regarding the relationship between Repair2 and Other2 variables as well as State2 and Other2. The correlation for Repair2 and Other2 (repairs and other expenses as reported on the Schedule C is 0.707 is not repeated in any other set of clusters. Theoretically, several interpretations could be made of this phenomenon, including disregarding it simply as an anomaly. Another, more intriguing interpretation is that, under some circumstances, repair costs (which is ill-defined by the Schedule C instructions) and other Schedule C expenses (which can include a plethora of items) may be related, both being moved upward in an attempt to raise deductions if repair costs become untenable by the taxpayer. One notable observation is the relationship between state/local taxes and 117 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. other expenses, recorded as a negative correlation, -0.766. Is there actually any meaningful correlation between these two variables in this single cluster? Although they might be a strong indicator of correlation, this does not imply causality between the variables and may simply reflect a mathematical error. A review of all of correlation tables reveals little evidence of significant and consistent variable interrelationships for any of the clusters identified in the sample. Reviewing the residual tables for Fresno and Compton shows that the residuals for both cities (Fresno min -0.39 and max 0.42, Compton min -0.44 and max 0.48) are unremarkable. The Cook’s Distance mean for Fresno and Compton, 0.054 and 0.201, respectively, provide some assurance that the data points are near the regression line for these two clusters. However, this phenomenon does not occur consistently throughout the other two clusters identified in the sample. 118 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 ■ D O CD Q . ■ D CD C /) C /) Table 14 (Part 1) Correlations for Fresno Cluster 2 TY 2000 COMP2 CARTRU 2 DEPREC2 HOME2 INC0ME2 LEGAL2 Pearson Correlation COMP2 1.000 -.379 .283 -.117 .725 -.096 CARTRU 2 -.379 1.000 -.209 .336 -.149 -.109 DEPREC2 .283 -.209 1.000 -.142 .416 .498 H0ME2 -.117 .336 -.142 1.000 .236 -.029 INC0ME2 .725 -.149 .416 .236 1.000 .021 LEGAL2 -.096 -.109 .498 -.029 .021 1.000 M E C2 .243 -.214 .137 -.312 -.019 .056 MEDEN2 .293 -.437 -.069 -.484 .042 -.045 OTHER2 .353 -.379 .516 -.391 .236 .265 REPAIR2 .093 .045 .066 .149 .153 .114 STATE2 .496 -.226 .402 .254 .723 -.034 TRAVEL2 .314 -.161 .477 -.274 .129 .333 Sig. (1-tailed) C0MP2 .018 .061 .265 .000 .304 CARTRU 2 .018 .129 .032 .212 .280 DEPREC2 .061 .129 .223 .010 .002 HOME2 .265 .032 .223 .101 .438 INC0ME2 .000 .212 .010 .101 .456 LEGAL2 .304 .280 .002 .438 .456 M E C2 .094 .124 .231 .044 .460 J82 MEDEN2 .055 .007 .357 .003 .412 .404 0THER2 .026 .018 .001 .015 .101 .075 REPAIR2 .310 .404 .362 .211 .206 .270 STATE2 .002 .111 .012 .084 .000 .428 TRAVEL2 .043 .194 .003 .068 .245 .034 o C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 ■ D O CD Q . ■ D CD C /) C /) Table 14 (Part 2) Correlations for Fresno Cluster 2 TY2000 M E 02 MEDEN2 OTHER2 REPAIR2 STATE2 TRAVEL2 Pearson Correlation C0MP2 .243 .293 .353 .093 .496 .314 CARTRU 2 -.214 -.437 -.379 .045 -.226 -.161 DEPREC2 .137 -.069 .516 .066 .402 .477 H0ME2 -.312 -.484 -.391 .149 .254 -.274 INC0ME2 -.019 .042 JZ 36 .153 .723 .129 LEGAL2 .056 -.045 .265 .114 -.034 .333 M E C2 1.000 .450 .442 .282 -.283 .011 MEDEN2 .450 1.000 .334 -.049 -.312 .018 OTHER2 .442 .334 1.000 .405 -.049 .404 REPAIR2 .282 -.049 .405 1.000 -.026 -.265 STATE2 -.283 -.312 -.049 -.026 1.000 .070 TRAVEL2 .011 .018 .404 ^265 .070 1.000 Sig. (1-tailed) C0MP2 .094 .055 .026 .310 .002 .043 CARTRU_2 .124 .007 .018 .404 .111 .194 DEPREC2 .231 .357 .001 .362 .012 .003 H0ME2 .044 .003 .015 .211 .084 .068 INC0ME2 .460 .412 .101 .206 .000 .245 LEGAL2 .382 .404 .075 .270 .428 .034 M E C2 .006 .006 .062 .062 .477 MEDEN2 .006 .033 .396 .044 .461 0THER2 .006 .033 .012 397 .012 REPAIR2 .062 .396 .012 .444 .075 STATE2 .062 .044 .397 .444 .354 TRAVEL2 .477 .461 .012 .075 .354 C D ■ D O Q . C g Q . ■ D CD C/) C/) CD 8 Table 15 Residuals for Fresno Cluster 2 TY2000 CD 3. 3 " CD CD ■ D O Q . C a O 3 " O O CD Q . ■ D CD C /) C /) Minimum Maximum Mean Std. Deviation N Predicted Value .89 2J9 1.74 .388 31 Standard. Predicted Value -2.186 1.671 .000 1.000 31 Standard Error of Predicted Value .111 .210 .168 .025 31 Adjusted Predicted Value .83 2.73 1.76 .407 31 Residual -.39 .42 .00 .217 31 Standard. Residual -1.433 1.553 .000 .796 31 Deleted Residual -.73 .64 -.02 .355 31 Mahalanobis. Distance 3.975 16.896 10.645 3.382 31 Cook's Distance .000 .283 .054 .072 31 Centered Leverage Value .132 .563 .355 .113 31 C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 ■ D O CD Q . ■ D CD C /) C /) Table 16 (Part 1) Correlations for Compton Cluster 7 TY2000 C0MP2 CARTRU 2 DEFREC2 HOME2 INCOME2 Pearson Correlation C0MP2 1.000 .099 .105 .205 .619 CARTRU_2 .099 1.000 .261 .191 .082 DEPREC2 .105 .261 1.000 .040 .229 H0ME2 .205 .191 .040 1.000 .177 INC0ME2 .619 .082 329 .177 1.000 LEGAL2 .089 .005 .085 -.055 -.154 M E C2 -.144 .611 .193 .296 ^303 0THER2 329 -.152 .252 .215 -333 REPAIR2 .221 -.396 .147 .030 .003 STATE2 .020 .503 .082 -.113 .491 Sig. (1-tailed) C0MP2 .348 339 .207 .003 CARTRU 2 .348 .148 .223 .373 DEPREC2 339 .148 .437 .181 H0ME2 .207 323 .437 341 INC0ME2 .003 .373 .181 .241 LEGAL2 .362 .492 369 .415 .271 M_E_C2 .285 .004 .222 .116 .110 OTHER2 .091 .273 .157 .196 376 REPAIR2 .190 .052 .280 .453 .496 STATE2 .469 .017 .374 .328 .019 C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 ■ D O CD Q . ■ D CD C /) C /) Table 16 (Part 2) Correlations for Compton Cluster 7 TY2000 LEGAL2 M E C2 OTHER2 REPAIR2 STATE2 Pearson Correlation C0MP2 .089 -.144 .329 .221 .020 CARTRU 2 .005 .611 -.152 -.396 .503 DEPREC2 .085 .193 .252 .147 .082 H0ME2 -.055 .296 .215 .030 -.113 INCOME2 -.154 -.303 -.233 .003 .491 LEGAL2 1.000 -.185 .462 .588 -.242 M E C2 -.185 1.000 .201 -.310 -.046 0THER2 .462 .201 1.000 .704 -.766 REPAIR2 .588 -.310 .704 1.000 -.526 STATE2 -.242 -.046 -.766 -.526 1.000 Sig. (1-tailed) C0MP2 362 .285 .091 .190 .469 CARTRU 2 .492 .004 .273 .052 .017 DEPREC2 369 .222 .157 .280 .374 H0ME2 .415 .116 .196 .453 328 INCOME2 .271 .110 .176 .496 .019 LEGAL2 .231 .027 .005 .166 M E C2 .231 .212 .105 .428 0THER2 .027 .212 .001 .000 REPAIR2 .005 .105 .001 .012 STATE2 .166 .428 .000 .012 to C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 8 ci' Table 17 Residuals for Compton Cluster 7 TY2000 3 3 " CD CD ■ D O Q . C a O 3 ■ D O CD Q . ■ D CD C /) C /) Minimum Maximum Mean Std. Deviation N Predicted Value .81 2.21 1.56 .458 18 Standard. Predicted Value -1.632 1.436 .000 1.000 18 Standard Error of Predicted Value A85 .308 .245 .038 18 Adjusted Predicted Value -.07 2.34 1.50 .619 18 Residual -.44 .48 .00 .228 18 Standard. Residual -1.333 1.430 .000 .686 18 Deleted Residual -1.18 1.07 .05 .592 18 Mahalanobis Distance 4.292 13.616 8.500 2.863 18 Cook's Distance .002 .912 .201 .318 18 Centered Leverage Value .252 .801 .500 A68 18 Cluster Position and Descriptive City (External) Data Since the external data represent constants for their respective communities, a simple comparison of the median income for each cluster versus the median income as identified by the Bureau of Labor Statistics may provide some clues to where these truckers are located in the social strata of their respective communities. Referring to Tables 10, 11 and 12, it should be noted that the truckers’ reported income is substantially lower than the general populations’ median income for all four cities. The Sacramento cluster median income was 33% lower than the general population’s median income for this city. The Modesto cluster was 32% lower than the general population’s median income. Members of the Fresno cluster were only 11% lower than the rest of the population in their city. The Compton cluster was the worst example, earning a median income 36% lower than the general population. Moreover, Compton had by far the worst poverty rate with 25.5% of the population below the poverty line. While the data does not fully support a conclusion that social pressures were in play in the Compton cluster, intuitively this type of environment could only promote noncompliance and the introduction of innovative tax avoidance techniques to maintain a minimal standard of living. The population size for the four clusters varies and the nature of these communities must be taken into account. Sacramento and Fresno, by themselves, are mid- size cities with central city populations of approximately 400,000. Sacramento is part of a greater megalopolis consisting of approximately 1.4 million people. 125 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Modesto is a smaller city which is along the Route 99 corridor along with Sacramento and Fresno. However, Compton is part of the greater Los Angeles megalopolis and it might have been prudent to seek additional clusters around this community, not simply along the interstate. Temporal Considerations As demonstrated by Table 12, Sacramento, Stockton, Modesto, Fresno, Bakersfield, Compton, and Long Beach display clusters for least two or more tax years. Some are relatively similar in cluster size. For example, Sacramento cluster 1 for TY 1999 and cluster 7 for TY 2000 have membership of 36 and 37, respectively. In addition, these clusters demonstrate similar R-Squared values of 0.392 for TY 99 and 0.581 for TY 00. Sacramento cluster 1 for TY 01, one of the chosen four clusters, has a reduced membership of 33 with an R-squared value of 0.596. Is it possible that these clusters are related and are an indicator of a matured grouping over the three years? As demonstrated by Table 12, another example with a similar membership size is Stockton, with clusters 1 for TY 99 with 20 and cluster 2 for TY 00 with 22members. The R-squared values for these two clusters are 0.666 and 0.846, respectively. Other cities where this phenomenon occurs is Modesto (TY 00 clusters 4 and TY 01 cluster 1), Fresno (TY 99 cluster 1 and TY 00 cluster 2), and Long Beach (TY 99 cluster 2 and TY 00 cluster 10). (See Table 12). The question arises whether the three years’ data are sufficient to capture ongoing clusters of noncompliance. At the same time, smaller unsuccessful clusters 126 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. of noncompliance innovation may expire without the necessary historical footprint from which to learn. Time impacts all phenomena, and only through additional analysis and comparisons of these findings with other clusters based on similar variables for different professions is it feasible that time can be incorporated into a viable statistical model. 127 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. CHAPTER V CONCLUSIONS, DISCUSSION, AND SUGGESTIONS FOR FUTURE RESEARCH Conclusions The hypotheses below represent phenomena that likely occur under the correct conditions and may be described as intuitively correct. There is substantial risk of committing a Type I or Type II error under the conditions in which this analysis was performed. (A Type I error occurs when a true null hypothesis is rejected; a Type II error takes place when a false null hypothesis is not rejected.) Human behavior is driven, to some degree, by environmental demands as well as by personality. Translating the conditions under which this phenomenon occurs can be extremely difficult since possible environmental pressures may be traced down to the household level. In this study, an effort was made using clustering techniques to understand social and economic pressures on an occupational group as opposed to traditional studies which mainly focus on filer types (e.g., high-income). The nature of the sole proprietor or small business trucker work activity does not allow for traditional networking. The home base is simply that, a hub for the operations of the business that may be rarely used. Networking in this occupation occurs via whatever 128 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. means the open road provides. Therefore, it is perceived that this occupation was among the most difficult to prove the viability of the 1RS data as possible indicators (and predictors) of noncompliance since competition and networking may not be a localized occurrence. This effort could be considered the mine test, reminiscent of the U.S. Navy’s testing protocol for combat ships. Listed below are the hypotheses followed by a brief discussion of the body of evidence to conclude whether these theories are applicable. Hypothesis 1 Independent/sole proprietor clusters of California-based truckers decrease their level of voluntary compliance where business pressures are higher than their competition in other, more affordable parts of the community. Several linkages had to be established to build the context for this environmentally-based hypothesis to describe the mechanism from which business and social interaction would drive compliance behavior. To reach a viable conclusion on whether to accept or reject this hypothesis, several methodological steps had to be taken. First, the original sample of NAIC 48110 and 48120 for the State of California was drawn for three tax years. The sample was fundamentally flawed since it contained a plethora of business types and income levels displaying non-Gaussian characteristics. To address this flaw, an examination of the data, both logically and statistically, reduced the overall sample size and also produced a more analogous 129 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. grouping of taxpayers. The sample was further limited once it was reduced by only including communities along two major highway corridors. This action was taken in an effort to simulate how an evasion tactic could disperse beyond its localized routes for this particular occupation. Second, hierarchical cluster analysis was performed on the test cities and communities without regard for the level of known noncompliance. Cities with clusters with 15 members or more displaying the same reporting characteristics were incorporated into the analysis. Third, linear regression was performed on each city’s cluster(s) using independent variables based on business and social pressure and gauging the impact on known compliance within the cluster. Finally, clusters displaying high relative R-squared values were to be compared with local communities in their corresponding highway corridors if this step proved tenable or there were enough clusters to make a meaningful comparison. Table 1 provided a list of variables including car and truck, legal, meals and expenses, travel, and other expenses used as business pressure variables. A review of correlation tables (Tables 14,15,16, and 17) for the four primary clusters revealed no significant correlation between these variables and the dependent variable, compliance. Additional review of the data for all primary clusters using regression analysis again revealed no significant correlation. At the same time, it should be noted that each of the possible evasion clusters existed within communities that displayed relatively high poverty rates and whose average income within the cluster was significantly lower than median income for the community at large. BLS data 130 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. from the 1999-2001 Metropolitan Area Occupational Employment and Wage Estimates were to be used for income comparisons to the general trucker population but was not included because of the lack of numerous identifiable and robust cluster formations. Conversely, three of the identified cities, Sacramento, Modesto, and Fresno are in relative proximity along the Highway 99 corridor. Referencing Appendix 3, the map of California, Sacramento and Fresno are the two furthest endpoints of the triad, approximately 170 miles apart with Modesto near the midpoint. Table 18 displays the three cities (a derivative of Table 13 excluding the Compton cluster which is discussed later). There are several similarities of note. First, both Sacramento and Fresno are relatively close in population size, approximately 400,000 residents, although both are central in larger metropolitan regions. The levels of crime in both are relatively the same at above 30,000+ occurrences. However, the level of known compliance between the 2001 Sacramento cluster and the 2000 Fresno is significantly different. The difference, 0.40 for Sacramento for TY 2001 versus 0.70 for Fresno in TY 2000 is difficult to explain given the relative crime rates for both. A breakdown by white collar and non-violent crime might provide a better basis for understanding these differences and should be applied in future analysis. A cursory review of compliance rates for the specific clusters in comparison to the general sample population reveal a relatively high rate of known compliance among the clusters except for Sacramento. For the both Fresno and the State of California, the cluster 131 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. taxpayers’ known compliance rate was actually higher than the general sample population. These competing results only provide additional background for speculation, not the basis for identifying the phenomenon that was the focal point of this hypothesis. At the same time, general compliance rates for tax year 2000 are higher than for TY 1999 and TY 2001 indicating a possible anomaly in the data. Initial investigation of the original sample and steps taken to reduce the sample revealed nothing extraordinary that points to either an error in the analysis or issues with the sample itself. Table 18 The Triad of Clusters City Name Sacramento Modesto Fresno Tax Year 2001 2000 2000 Cluster Number of 15 1 4 2 Sample Size N 33 38 31 R .772 .712 .873 R Square .596 .508 .762 Adjusted R Square .384 .299 .624 Standard Errors of the Estimate .389 .329 .273 Durbin Watson 2.383 2.169 1.625 Significance .020 .030 .001 Population* 407,018 188,856 427,652 Median Income ($)* 37,049 40,394 32,236 Percentage of Population at or Below the Poverty Level 15.3 12.2 20.5 Crime Statistics (Occurrences) 31,131 10,892 33,332 Compliance (General Population) 0.40 0.73 0.70 S pecific C lu ste r C om pliance 0.40 0.82 0.74 *As identified by the 2000 Census 132 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Interpretative results aside, there is little evidence to conclude that the clusters of truckers share anything more than similar reporting characteristics. The Sacramento cluster display the same low level of known compliance as the general population at 0.40 for TY 2001, but this level was not observed in TY 2000. At the same time, the TY 1999 Sacramento cluster was rejected for further analysis because it did not display strong correlation between the business pressure independent variables and compliance. Another, larger cluster exists for Sacramento (n=40) for TY 2001 but was rejected for additional analysis since it showed little correlation between the business pressure independent variables and compliance. Since there is no défendable evidence that a clustering effect has taken place based on noncompliant behavior, and no separable activity that can be compared to the general community, it is prudent to reject this hypothesis. Hypothesis 2 Independent/sole proprietor California-based truckers may display group behavioral characteristics and decrease their level of voluntary compliance if they operate or reside in communities where their social ranking is below the median social ranking within their occupation. The most critical social pressure variable for this analysis was the home mortgage expense that is confirmed via matching with the Form 1098. Unfortunately, in efforts to hone the sample, those without mortgage payments were eliminated. 133 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. This group probably contains a mixture of both renters and those that may have paid- off a home. Unfortunately, city specific data for home mortgage costs was not consistently available. The U.S. Census and Department of Housing and Urban Development did perform its American Housing Survey for Sacramento in 1996, but this data was considered too old to be useful for comparison purposes. Moreover, the survey did not break down the population into specific occupations. The social and business pressure variables were reviewed together for correlation regarding their impacts on compliance. Only income and, in some instances, state and local taxes displayed significant correlation to the compliance dependent variable. Unfortunately, these correlations were not consistent enough to effectively separate the cluster members’ behavior from the general population nor to justify a comparison. At the same time, revisiting the income variances observed in the four primary clusters (ranging from 10 to 38% below the median income for all wage earners) is a stark reminder that the regression model might still provide clues to the potential power of income and taxation on compliance. Compounding this is the poverty level for the test cities. In Compton, the poverty level was 25.5% of the population, a full 5% higher (e.g., Fresno = 20%) than any other city-based clusters within Table 12. While the identified social variables offer little evidence of impacting tax compliance as stand-alone factors, broader issues which are directly related to income such as the overall poverty rate offer some insight on a larger phenomenon, the overall environmental effect on compliance. 134 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. As with Hypothesis 1, neither the number/size of the clusters nor the correlation factors justify accepting the hypothesis. Because there was no substantive indication within the clusters (other than possibly Compton) that the cluster membership behaved uniformly, the remainder of the hypothesis falls by the wayside. However, a possible improvement in the analysis may have entailed reviewing all the cluster data for a single city such as Compton. Discussion of the Results and the Sample Challenge The approach used in this analysis for data reduction in an attempt to identify clusters of noncompliance is not new. The focal point of using occupation however, is unique and should not be abandoned based on the results of this analysis. Again, the trucker population was used to challenge these hypotheses. It may have been easier simply to use an occupation which is more community dependent, such as medicine or contracting. While there is only scant evidence (the four primary clusters) that the 1RS data had some predictive power over compliance, the fact is there is some evidence that an effect exists. This should be used as the foundation for additional research, not as the basis for dismissing the approach. Cluster analysis and other data mining techniques offer powerfid tools for detecting growing cells of noncompliance. If compliance is treated as an innovation, the possibility of detecting it may be enhanced. Innovation usually begins with the most sophisticated individuals attempting to design and introduce a new idea. Based on the perceived success of the noncompliance innovation, it may collapse or expand 135 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. exponentially. Therefore, a methodology must be discovered to detect noncompliance innovations in their infancy as opposed to the traditional 1RS methods where the noncompliance practice is allowed to spread nationally (as with slavery reparations). A system designed to monitor and detect trends within and among taxpayer clusters may offer such an alternative. Moreover, it was earlier postulated that owner/operators may use the highway system as a means to inadvertently spread noncompliance behavior. This is one example of a possible medium for spreading noncompliance. The media, professional organizations, churches, and other settings may offer additional paths for spreading the innovation of noncompliance. While the four primary clusters did not provide any evidence that this effect was occurring. Table 12 is primarily populated by cities along the Route 99 corridor. Many of the cities along the Interstate 710 corridor, including Bell and Bell Gardens did not support a large enough population for both clustering and the use of regression analysis. (Long Beach is the exception.) Consequently, smaller, developing clusters of taxpayers may have been missed. Another consideration is that several occupations specific algorithms could be used side-by-side to identify interrelationships among the occupations. In addition, this method could be used to identify clusters that cross occupational boundaries indicating the existence of social institutions as pathways for the expansion of a noncompliance behavior. 136 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Suggestions for Future Research An important division can be drawn between an evasion tactic and a tax scheme. Tax schemes are often issue driven. For example: “Resisters who are interested in learning about the advantages and risks associated with self employment should read NWTRCC's excellent pamphlet, “Self Employment: An Effective Path for War Tax Refusal: Practical War Tax Resistance #4" (Goldberger & Feldman, 2001, p. 1). Tax authorities cannot identify a person’s political affiliation and, in the case of war protestors, their incomes may be so modest as not to warrant attention. These flash-in-the-pan schemes tend to exhibit a limited life span, are fairly visible, and in many cases have known promoters. Another example, cited previously, was the slavery reparations scheme. An evasion tactic tends to be more subtle, like a technique or tool which is developed over the course of time. Schemes use evasion tactics which are the manifestation of reporting and monitoring vulnerabilities (e.g., self-reported expenses, off-shore income, etc.). While the Federal Government tends to focus on the large-scale issues associated with evasion it has performed in a less than satisfactory manner in understanding human vulnerabilities that cause taxpayers to embrace an evasion tactic. The method used in this paper should not be rejected based on this analysis. More viable population sets that are directly connected to their communities should be tested. The methodology advanced within this analysis requires perfection to build more sensitivity into the model. If compliance indicators can be captured, regardless 137 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. of their ability to indicate malice, aggressive attempts to design models should be pursued. More importantly, external data sources which provide context for the noncompliance activity should be better defined and explored as primary factors in searching for noncompliance clusters. The benefit for tax administrators is developing the capability, early-on, to detect evasion tactics that are in their dispersal infancy. This analysis attempted to capture clusters of possible evaders at an early stage and to apply this knowledge to two possible compliance scenarios that are encapsulated in the hypotheses. The key to finding where an evasive tactic is bom is locating the small group of allies that test and eventually may serve as the decision point to disperse the tactic to more people within their social network. It has been argued that entrepreneurs are more likely to take risks than ordinary people (Knight, 1971). Entrepreneurs can exist in all aspects of society, including occupation. Is it possible that this form of leadership exists within informal groupings of taxpayers? Of greater interest may be that these leaders will use trusted lieutenants and allies as test beds for noncompliant behavior. In addition, the testers may look to the entrepreneur to lead them out of business difficulties or to provide momentum to move up the social ladder. The IBM Institute for Knowledge-Based Organizations uses social network analysis as an integral tool for understanding how information, and knowledge in general, is disseminated through an organization. These same principles can be applied to communities: In deciding whether or not to seek out an individual for information or advice, a person must have some perception of 138 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. the relevance of the other person’s knowledge, skills and abilities in relation to the current problem. Although this perception might be wrong or biased by a variety of factors, it is still the basis for deciding to whom to turn for information or advice on a given problem. Thus, understanding how well members of a group know each others’ knowledge, skills and abilities is a first step to understanding how effective they are in terms of knowledge sharing and creation. (Cross, Parker, & Borgatti, 2002, p. 7) The principal “glue” for these social networks, as described earlier in this analysis, is trust among the entrepreneur and a small circle of confidents that expands over time. Unfortunately, this analysis cannot rule out friendship or family networks based on trust. The only reasonable approach to perform an analysis on a large population is to use data mining techniques such as cluster analysis that are dependent upon the limited number of social and business pressure variables that are freely available. Figure 8 is a generic representation of social networks providing some insight into how different nodes connect. These linkages cannot be easily modeled in the context of real-world research, since it would require invasive data gathering on the scale of the FINCEN model. What is required, at a very basic level, is a thorough understanding of the business and social pressures that may be unique to a certain occupation. For example, physicians incur costs related to the purchase of new medical equipment and malpractice insurance. Small defense contractors incur costs related to travel, meals, and entertainment. Sales occupations experience the same types of cost pressure. Understanding these cost pressures as well as the market in which these businesses compete could greatly enhance tax authorities’ understanding 139 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. I % I o iZJ e s 1 I 140 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. of the potential environments where evasive tactics are bom and expand. However, to be effective, this understanding cannot be at the aggregate level. Tax administrators must gain a localized understanding of these environments because this is where the taxpayer lives and works. Moreover, this is where evasion tactics are bom. From a public administration perspective, noncompliance cluster detection is hampered by the basic quality of the data within 1RS systems. To legitimately collect the proper tax, 1RS information systems must be enhanced beyond simply applying sample-based algorithms and projecting expected levels of taxpayer compliance on a macro-economic scale. 1RS is making a genuine effort to modernize its systems, but it is unclear whether these future systems will provide the necessary data structure to more efficiently model taxpayer activity and behavior. It is hoped that efforts such as this study will develop behavioral research platforms that are essential to identify evasion tactics in their infancy. While much of the attention of this analysis has been focused on the IRS’s ability to analyze taxpayer behavior, it is state and local governments that may ultimately benefit fi"om this type of approach. Given that many state and local governments are cash-strapped and have few resources to launch new initiatives, the early identification of new evasive tactics and schemes represents an opportunity to avoid revenue loss and consume valuable enforcement resources. Moreover, they are 141 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. doser to their populations than the Federal Government, and some way have with access to crucial local data that the 1RS must search out and mine. One potential approach is to establish a local experimental environment testing different versions of the methodology (different variables). This could be done as a partnership between Federal, state, and local tax authorities focusing on several different occupations. This analysis might be expanded to search for linkages between professional classes and constructing the basis for an event history. For the construct of the experiment to be successfiil, the environment in which the taxpayers work and live must be carefully recorded. Therefore, changes to the environment and possible corresponding changes to known levels of compliance can be observed. 142 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. SELECTED BIBLIOGRAPHY 2002 Directory of California Local Area Wages (2002). Retrieved from the Internet on June 5,2002, at www.calmis.ca.gov/file/occup$/dclaw$.htm. 90,000 Mile Diesel Engine Oil Changes. Retrieved from the Internet on February 8, 2003, at http://leeric.lsu.edU/bgbb/7/ecep/diesel/a/a.htm. Abrahamson, E., & Rosenkopf, L. (n.d.). Toward a synthesis of fad and efficient- choice theories of bandwagon innovation diffusion. (E. Abrahamson, personal communication, February 5,1999), 1-36. Abrahamson, E., & Rosenkopf, L. (1993). Institutional and competitive bandwagons: Using mathematical modeling as a tool to explore innovative diffusion. Academy of Management Best Paper Proceedings, 487-517. Abrahamson, E., & Rosenkopf, L. (1997). Social network effects on the extent of innovation diffusion: A computer simulation. Organization Science, 8 ,289- 309. Ajzen, 1., & Fishbein, M. (1980). Understanding attitudes and predicting social behavior. Englewood Cliffs, NJ: Prentice Hall. Allingham, M. G., & Sandmo, A. (1972, November). Income tax evasion: A theoretical analysis. Journal o f Public Economic, 1, 323-338. Aim, J. R. (1988, March). Uncertain tax policies, individual behavior and welfare. The American Economic Review, 78 (1), 237-245. Aim, J., Bahl, R., & Murray, M. M. (1990, November). Tax structure and tax compliance. The Review o f Economics and Statistics, 72, 603-613. Aim, J., Jackson, B. R., & McKee, M. (1992). Estimating the determinants of taxpayer compliance vrith experimental data. National Tax Journal, XLV, 1. 143 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Aim, J., McClelland, G. H., & Schulze, W. D. (1992, June). Why do people pay taxes? Journal o f Public Economic, 48,21-3%. Alvarez, M. R., & Brehm, J. (1998, April). Speaking in two voices: American equivocation about the Internal Revenue Service. American Journal of Political Science, 42,418-452 An Analysis of1RS Compliance Measurement Techniques (1998). Produced for the U. S. Internal Revenue by the Sag Corporation, 1-14. Andreoni, J., Erard, B., & Feinstein, J. (1998, June). Tax compliance. Journal of Economic Literature, 36, 818-860. Autunes, G., & Hunt, A. L. (1973). The impact of uncertainty and severity of punishment on levels of crime in American states: An extended analysis. Journal of Criminal Law and Criminology, 64. 486-493. Belman, D. A., Monaco, K. A., & Brooks, T. J. (1997). Let it be palletized: A portrait o f truck drivers ’ work and lives from the 1997 survey o f truck drivers, University o f Michigan Trucking Industry Program. Milwaukee, WI: Department of Economics, University of Wisconsin Milwaukee. Bemasconi, M. (1998). Tax evasion and orders of risk aversion. Journal o f Public Economics, 6 7 ,123-134. Berwick, M., & Dooley, F. (1997). Truck costs for owner/operators. Fargo, ND: Department of Transportation, University (North Dakota) Transportation Centers Program. Bishop, Y. M., Feinberg, S. E., & Holland, P. W. (1995). Discrete multivariate analysis: Theory and practice. Cambridge, MA: The MIT Press. Blau, P. M. (1977). Inequality and heterogeneity. New York: Free Press. Blumenthal, M., Christian, C., & Slemrod, J. (1998, May). The determinants of income tax compliance: evidence from a controlled experiment in Minnesota [Working paper 6575]. Cambridge, MA: National Bureau of Economic Research. 144 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Bowles, S., & Gintis, H. (2000, December). Social capital and community governance. Economic Journal Presented to a Symposium for the Department of Economics, University of Massachusetts, Amherst, Massachusetts. Brand, P. (n.d.). Compliance: A 2U * century approach. National Tax Journal, 44, 413-419. Brehm, J. (1996). Steps towards a political science of compliance: Common insights and recurring problems. Prepared for presentation at the 5 4 * * * Annual Meeting of the Midwest Political Science Association, Chicago, II. Brehm, J., & Hamilton, J. T. (1996, May). Noncompliance in environmental reporting: Are violators, ignorant, or evasive, of the law. American Journal of Political Science, 40,444-477. Brehm, J., & Rahm, W. (1997, July). Individual-level evidence of the causes and consequences of social capital. American Journal of Political Science, 41, 999-1023. Brooks, J. J., & Fall, S. (1996, Fall). Taxation and human capital. The American Journal of Tax Policy, 1-18. Brown, R. E., & Mazur, M. (2002, June 12). The National Research Program: IRS’ s new approach to measuring taxpayer compliance. U.S. Internal Revenue Service White Paper. Washington, DC: U.S. Government Printing Office. Brusco, M. J., Credit, D., & Stephanie, S. (2002, Winter). A simulated annealing heuristic for a bicriterion partitioning problem in market segmentation. Journal of Marketing Research. Burtless, G. (1999). Effects ofgrowing wage disparities and changing family composition on the U.S. income distribution (Working Paper Number 4). Washington, DC: Brooking Institution, Center on Social and Economic Dynamics. Campbell, D., Parisi, M., & Balkovic, B. (2000, Fall). Individual income tax returns 1998. Internal Revenue Service Statistics o f Income Bulletin, 8-46. 145 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Carmone, F. J., Kara, A., & Maxwell, A. (1999). HINoV: A new model to improve market segmentation definition by identifying noisy variables. Journal of Marketing Research, 36, 501-509. Carnes, G. A., & Englebrecht, T. D. (1995, Spring). An investigation of the effect of detection risk perceptions, penalty sanctions, and income visibility on tax compliance. The Journal of the American Taxation Association, 17 26-41. Carroll, J. S. (1989). A cognitive-process analysis of taxpayer compliance. In J. Roth, J. T. Scholz & A. Dryden Witte (Eds.), Taxpayer compliance. Vol. 2: Social science perspectives, Philadelphia, PA: University of Pennsylvania Press. ________ (1992). How Taxpayers Think About Their Taxes: Frames and Values. In J. Slemrod (Ed.), Why people pay taxes: Tax compliance and enforcement. Ann Arbor, MI: University of Michigan Press. Casavant, K. (1993). Basic theory o f calculating costs: applications to trucking. Upper Great Plains Transportation Institute No. 118. Fargo, ND: North Dakota State University. Chang, O. H., & Schultz, J. J. (1990, Fall). The income tax withholding phenomenon: Evidence firom TCMP data. The Journal o f the American Taxation Association, 12, 88-93. Christian, C. W. (1994). Voluntary Compliance with the Individual Income Tax: Evidence from TCMP Study. The 1RS Research Bulletin (1RS Publication 1500, Rev. 9-94), 35-42. Citrin, J. (1979). Do people want something for nothing: Public opinion on taxes and government spending. National Tax Journal, 3 2 ,113-129. Cleary, R. (1971). Political education in the American democracy. Scranton, PA: International Textbook Company. Clotfelder, C. T. (1983, August). Tax evasion and tax rates: An analysis of individual returns. The Review o f Economics and Statistics, 65, 363-373. Cohen, S. H., & Ramaswamy, V. (1998, Spring). Latent segmentation models: New tools assist researchers in market segmentation. Marketing Research: A Magazine o f Management Applications, 10. 146 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Coleman, James S. (1964). Introduction to mathematical sociology. New York: Free Press. Cox, D. (1984). Raising revenue in the underground economy. National Tax Journal, 57,283-288. Cross, R., Parker A., & Borgatti, S. P. (2002). A bird’ s-eye view: Using social network analysis to improve knowledge creation and sharing. Somers, NY : IBM Institute for Business Value. Cummins Engine Co., Inc. (1992). Cummins Bulletin 3385896. Columbus, IN: Cummins Engine Co., Inc. Dagal, J. F. (1988). Diesel engine andfuel system repair (2nd ed.). New York, NY: John Wiley & Sons, Inc. Davis, J. A. (1959). A formal interpretation of the theory of relative deprivation. Sociometry, 22 ,280-296. Davis, J. S., Hecht, G., & Perkins, J. (2000, August). Social behaviors, enforcement, and compliance dynamics (Draft 2.0). Champaign, IE: Department of Accountancy, University of Illinois at Urbana-Champaign. Do new engines mean new PM programs? (1992, March). Fleet Equipment Pages, 24-28. Dooley F., Bertram, L., & Wilson, W. (1988). Operating costs and characteristics of North Dakota grain trucking firms. Upper Great Plains Transportation Institute No. 67. Fargo, ND: North Dakota State University. Dooley, F. (1996). Interview for trucking characteristics. Wallwork Truck Sales, Fargo, ND. Duclos, J., Jalbert, V., & Araar, A. (2001, January). Classical horizontal inequity and reranking: An integrated approach. Quebec, Canada: Universate Laval Quebec. Dusenbury, R. (1994, Spring). The effect of prepayment position on individual taxpayers’ preferences for risky tax-filing options. The Journal o f the American Taxation Association, 1 6 ,1-16. 147 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Economie action and social structure: The problem of embeddedness (1985, November). American Journal o f Sociology, 97,481-510. Egasbira, S., & Hasbimoto, T. (n.d.). The formation of common norms on the assumptions of “fundamentally” imperfect information. Paper on to appear in Rosaria Conte & Chris Dellarocas (Eds.), Social order in multiagent systems. Dordrecht, The Netherlands: Kluwer. Hide, E. (2001, November). Rank dependent expected utility models o f tax evasion. Oslo, Norway: University of Oslo, Department of Private Law. Etzioni, A. (1988). The moral dimension. New York: Free Press. Evan, W. N. (Ed.) (1980). The sociology o f law: A social-structured perspective. New York: The Free Press. Fallan, L. (1999, January). Gender, exposure to tax knowledge, and attitudes towards taxation: An experimental approach. Journal Business Ethics, 18, 173-184. Feld, L. P., & Frey, B. S. (2001, December). The tax authority and the taxpayer: An exploratory analysis. Paper presented to the 2 "'* International Conference on Taxation, Canberra, Australia. Festinger, L. (1957). A theory o f cognitive dissonance. Evanston, IE: Row- Peterson. FirstFleet Inc. (2003). Finding the sweet spot. In Patricia McCullough Smith (Ed.), Heavy Duty Trucking. Retrieved March 11, 2003, from http://www.firstfleet.co/heavydt.html Franzoni, L. A. (1999). Tax evasion and tax compliance. Bologna, Italy: University of Bologna. French, S. (1986). Decision theory: An introduction to mathematics of rationality. New York: John Wiley. French, J. R. P., Jr., & Raven, B. (1989). The basis of social power. In O. J. Steven (Ed.) Classic readings in organizational behavior. Belmont, CA: Brooks/Cole Publishing Company, 440-455. 148 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Ghosh, D., & Crain, T. T. (1995, May). Ethical standards, attitudes toward risk, and intentional noncompliance: An experimental investigation. Journal o f Business Ethics, 14, 353-365. Goldberger, P., & Feldman, J. H., Jr. (2001, April). Beware of schemes that promise easy ways to reduce your taxable income to zero. More than a paycheck, news from the war tax resistance. Retrieved from the Internet on July 5, 2000, http://www.nwtrcc.org/mtap01/mtap0401 .htm. Gouveia, M., & Strauss, R. P. (1999, November). Effective tax functions for the US individual income tax: 1966-1989. Proceedings of the 1999 National Tax Association, 92“* Annual Conference, Atlanta, GA. Granovetter, M. (1973). The Strength of Weak Tie. American Journal o f Sociology, 7 & , 1360-80 ________(1978). Threshold models of collective behavior. American Journal of Sociology, 83, 1420-1443. Grasmick, H. G., & Scott, W. J. (1982). Tax evasion and mechanisms of social control: A comparison of grand and petty theft. Journal o f Economic Psychology, 2 ,213-230. Gray, B. (1985). Conditions facilitating interorganizational collaboration. Human Relations, 38 ,911-936. Hamilton, A. (n.d.). Federalist. Retrieved from Internet on September 4,2003, http://memory.loc.gov/const/fed/fed_35.html. Haney, C. (2002, March). Making Law Modem: Toward a Contextual Model of Justice. Psychology, Public Policy, and Law. Hardman, J. C. (1994, November). Unemployment compensation and independent contractors: The motor carrier industry. Transportation Law Journal, 22. Hashimoto, T., & Egashira, S. (2000). Formation o f social norms in transferring information and revising the cognitive frameworks. Proceedings of KSS2000 Internal Symposium on Knowledge and System Sciences: Challenges to complexity. Hawthorne, M. R., & Jackson, J. E. (1987). The individual political economy of federal tax policy. American Political Science Review, 81. 757-774. 149 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Hemmer, T., Stinson, C. H., & Vaysman, I. (1994, Fall). Self-audits, penalties, and taxpayer compliance. The Journal o f the American Taxation Association, 16, 63-81. Howell, F. M., & Wells, J. G. (1998, February). Social Indicator Analysis Study (SIAS) o f the demandfor substance abuse treatment in Mississippi. Jackson, MS: Social Science Research Center, Mississippi State University. Jones, M. (2001, September). Ability to evade taxes and equilibrium tax reporting and auditing (Draft), Binghamton, NY : State University of New York. Joulfaian, D., & Rider, M. (1996, December). Tax evasion in the presence of negative income tax rates. National Tax Journal, 44, 553-570. Joulfaian, D., & Rider, M. (1998, December). Differential taxation and tax evasion by small business. National Tax Journal, 51, 676-687. Kerlinger, F. N. (1986). Foundations of behavioral research ( 3 * ^ ^ ed.). Philadelphia, PA: Holt, Rinehart, and Winston, Inc. Kirchler, E., Maciejovsky, B., & Schneider, F. (2001, April). Everyday representations of tax avoidance, tax evasion, and tax flight: Do legal differences matter! (Working Paper). Vienna, Austria: University of Vienna, Department of Psychology, Unit of Economic Psychology. Knight, F. H. (1971). Risk, uncertainty and profit. Chicago: The University of Chicago Press. Krackhardt, D. (1992). The strength of strong ties: The importance of philos in organizations. In Nitin Nohria and Robert G. Eccles (Eds.), Networks and organizations. Boston: Harvard Business School Press. Krackhardt, D., & Hanson, J. R. (1993, July-August). Informal networks: The company behind the chart. Harvard Business Review,\DA-\ 11. Kruglanski, A.W., Shah, J. Y., Pierro, A., & Mannetti, L. (2002). When similarity breeds content: Need for closure and the allure of homogeneous and self- resembling groups. Journal of Personality and Social Psychology, 83, 648- 662. 150 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Lazarsfeld, P. F., & Merton, R. K. (1954, 1964). Friendship as social process: A substantive and methodological analysis. In M. Berger, T. Abel & C. H. Page (Eds.), Freedom and control in modem society (pp. 18-66). New York: Octagon Books. Lévi-Strauss, C. (1963). Structural anthropology. New York: HarperCollins Publishers. ________ (1966). The savage mind. Chicago: The University of Chicago Press. Loucks, A. D., & Zambie, E. (n.d.). Predictors of criminal behavior and prison misconduct in serious female offenders. Ontario, Canada: Correctional Service Canada.. Luthans, F. (Ed.) (1995). Organizational Behavior (7* ed.). New York: McGraw- Hill, Inc. Martin, J. (1981). Relative deprivation: A theory of distributive injustice for an era of shrinking resources. Research in Organizational Behavior, 3 ,53-107. May, P. J., & Soren, W. (1999). Regulatory enforcement and compliance: Examining Danish Agro-Environmental Policy. Journal o f Policy Analysis and Management, 18, 625-651. Milliment, D. L., & Racine, J. (2002). A nonparametric analysis o f the determinants o f child schooling in Indonesia. Dallas, XX: Department of Economics, Southern Methodist University. Mills, L. F. (1996, September). Corporate Tax Compliance and Financial Reporting. National Tax Journal, 44,421-435. Nantrell, S. C. (1999, Spring). Symposium: Federal Tax Policy in the new millennium: A cultural perspective on American Tax Policy. 1999 Chapman Law Review, 33, Chapter 2. New PM Programs (1992, May). Fleet Equipment Pages, Part 2,16-21. Nigrini, M. J. (1999, May). I’ve got your number: How a mathematical phenomenon can help CPAs uncover fraud and other regularities. Journal o f Accountancy, 79-83. 151 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Northrup, H. R. (1999, Spring). The failure of the Teamsters’ Union to win railroad- type labor protection for mergers or deregulation. 1995 Transportation Law Journal, 22. Oil and the engine make (1993, January). Fleet Equipment Pages, 22-25. Pfeffer, J., & Salancik, G. R. (1978). The external control of organizations. New York: Harper and Row. Prescott, L. D., Jr. (n.d.). Challenging the adversarial approach to taxpayer representation. Los Angeles Law Review, 50,693-773. Pretty, J., & Ward, H. (2001). Social Capital and the Environment World Development, 29,209-227. Putnam, R. D. (1993). Making democracy work: Civic traditions in modern Italy. Princeton, NJ: Princeton University Press. ________(1995). Bowling alone: America’s declining social capital. Journal of Democracy, 6, 65-78 Rakove, J. N. (1996). Original meanings: Politics and ideas in the making o f the Constitution. New York: Alfred A. Knopf. Reckers, P. M. J., Sanders, D.L., & Roark, S.L. (1994, December). The influence of ethical attitudes on taxpayer compliance. National Tax Journal, 47, 825-836. Ritsema, C. M., Thomas, D. W., & Perrier, G. D. (2003). Economic and behavioral determinants o f tax compliance: Evidence from the 1997Arkansas Tax Penalty Amnesty Program. Paper presented to the 2003 1RS Research Conference in Albuquerque, New Mexico. Roberts, M. L. (1994, Spring). An experimental approach to changing taxpayers’ attitudes towards fairness and compliance via television. The Journal of the American Taxation Association, 16, 67-86. Rodgers, E. (1983). Diffusion o f Innovations. New York: The Free Press. Ruback, R. B., & Wroblewski, J. (2001). The Federal sentencing guidelines: Psychological and policy reasons for simplification. Psychology, Public Policy, and Law, 7, 739-775. 152 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Ruebeck, C. J. (1999). Imitation dynamics in the Repeated Prisoner’s Dilemma: An exploratory example. Journal of Economic Behavior & Organization, 40, %\- 104. Schermerhom, J. R. (1975). Determinants of interorganizational cooperation. Academy o f Management Journal, 18, 846-856. Scholz, J. T., & Lubell, M. (1998). Adaptive political attitudes: Duty, trust and fear as monitors of tax policy. American Journal of Political Science, 42, 903- 920. Scholtz, J. T., & Finney, N. (1995, May). Duty, fear, and tax compliance: The heuristic basis of citizenship behavior. American Journal o f Political Science, 39490-512. Scholz, J. T., & Wood, D. B. (1998). Controlling the 1RS: Principles, principles, and public administration. American Journal of Political Science, 4 2 ,141-162. Shef&in, S. M., & Triest, R. K. (1992). Can brute deterrence backfire? Perceptions and attitudes in taxpayer compliance. In Joel Slemrod (Ed.), Why people pay taxes: Tax compliance and enforcement (pp. \92-2lS). East Lansing, MI: University of Michigan Press. Slemrod, J. (1989, September). On voluntary compliance, voluntary taxes, and social capital; symposium: What do we mean by “taxpayer relief’? National Tax Journal, 51. Smith, H. J., & Tyler, T. R. (1996). Justice and power: When will justice concerns encourage the advantaged to support policies which redistribute economic resources and the disadvantaged to willingly obey the law? European Journal o f Social Psychology, 26,171-200. Spencer, J. (n.d.). Concentrated poverty in Los Angeles County and the geographic distribution o f economic change. Los Angeles: Institute for Labor and Employment, University of Southern California. The State of California Franchise Tax Board (2000, May). Non-filer program. Presentation provided at a meeting between the Franchise Tax Board and the 1RS, Sacramento, CA. 153 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Stockburger, D. W. (1998). Introductory Statistics: Concepts, Models, and Applications. Retrieved from the Internet on December 3,2002, at http://psychstat.smsu.edu/introbook/sbkOO.htm. Strauss, R. P. (1997, Fall). Administrative and revenue implications of alternative federal consumption taxes for the state and local sector. The American Journal of Tax Policy, 14, 361. Sutherland, H., Taylor, T., & Gomulka, J. (2001). Combining household income and expenditure data in policy simulations. (The Microsimulation Unit, Number MUOlOl, DAE Working Papers, Amalgamated Series No. 0110.) Cambridge, England: Department of Applied Economics, University of Cambridge. Tax Notes 14440 (1994, September 12). 1RS could improve pursuit of sole proprietor deadbeats, GAO says. Taylor, N. (2001, July). Understanding taxpayer attitudes through understanding taxpayer identities (Working Paper No. 14). Canberra, Australia: The Australian National University, Australian Taxation Office, Centre for Tax System Integrity. Teitelbaum, D. (2000). Analysis of idea mind virus propagation: The Bios Group solution. Arlington, VA: The Bios Group. Torgler, B. (1993). Vertical and exchange in a tax morale experiment. University of Basel, Wirtschaftswissenschaftliches Zentrum, Abteilung Wirtschafrspolitik, Basel, Switzerland. Retrieved on November 8,2002, at http://polis.unipmn.it/epcs/papers/torgler.pdf. Trist, E. L., & Bamforth, K. W. (1989). The basis of social power. In J. S. Ott (Ed.), Classic readings in organizational behavior (pp. 373-391). Belmont, CA: Brooks/Cole Publishing Company. U.S. General Accounting Office. (1994, September). 1RS can better pursue noncompliant sole proprietors. (GAO/GGD-94-175), Washington, DC. U.S. General Accounting Office (1996, May). Internal Revenue Service: Results of Non-filer Strategy and opportunities to improve future efforts. (GAO/GGD- 96-72). Washington, DC: General Accounting Office. 154 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. (1998, December). Tax administration: 1RS ’ audit and criminal enforcement rates for individual taxpayers across the country. (GAO/GGD- 99-19). Washington, DC: General Accounting Office. (2000, March). Tax administration: 1RS’ use o f non-audit contacts. (GAO/GGD-00-7). Washington, DC: General Accounting Office. U. S. Internal Revenue Service (1978). A Dictionary o f Compliance Factors. Source unknown, 1-33. ________ (1983). Income tax compliance research: Estimates for 1973-1981. Washington, DC: U.S. Government Printing Office, 1983-381 -544-1056. (1993). Federal tax compliance research: Gross and net employment tea gap estimates for 1984-1997. Department of the Treasury, Publication 1415- E (Rev. 10-93), Catalog Number 20478E. Washington, DC: U.S. Government Printing Office. (1994). The 1RS Research Bulletin: 1993/1994. Department of the Treasury, Publication 1500 (Rev. 9-94), Catalog Number 11546J. Washington, DC: U.S. Government Printing Office. (2001). ADP and IDRS information. Department of the Treasury, Document 6209 (Rev. 3-2001), Catalog Number 614620. Washington, DC: U.S. Government Printing Office. (2001). Study of taxpayer attitudes and noncompliance (Draft Report). Department of the Treasury, Roper Starch Worldwide Inc.,1983-381-544- 1056. (2002, February). FY 2003 budget justification. Unpublished document. as used. Uzzi (1996). The sources and consequence of embeddedness for the economic performance of organizations: The network effect. American Sociological Review, 61,674. Verhoogen, E., Burks, S. V., & Carpenter, J. P. (2002, September). Fairness and freight-handlers: A test offair-wage theory in a trucking firm (Working Paper No. 96). Berkeley, CA: Center for Labor Economics, University of California. 155 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Vihanto, M. (2000). Tax evasion in a transition from socialism to capitalism: The psychology o f the social contract (Discussion Paper No. 6). Bank of Finland: Institutes for Economies in Transition. Walker, I., & Mann, L. (1987). Unemployment, relative deprivation and social protest. Personality and Social Psychology Bulletin, 75,275-283. Washington State Department of Revenue (1998). Retailers ’ cost o f collecting and remitting sales tax. Presentation provided at a meeting in May, 2000, Research Division, Olympia, WA. Weinberg, S. L., & Abramowitz, S. K. (2002). Data analysis for the behavioral sciences using SPSS. New York: Cambridge University Press. Westat Inc. (1980). Individual Income Tax Compliance Factors Study Qualitative Research. Prepared for the Internal Revenue Service. Wojciech, K., & Slemrod, J. (2001, March). National Bureau of Economic Research (NBER) Working Paper No. w8158. Woodman, R. W., Sawyer, J. E., & Griffin, R. W. (1993). Toward a theory of organizational creativity. Academy of Management Review, 75,293-321. Woodward, B. (2000). Maestro: Greenspan’ s Fed and the American Boom. New York: Simon and Schuster. Retrieved from Washingtonpost.com on November 15,2000, excerpt from Greenspan Locates Link in High-Growth Economy. Yitzhaki, S. (1994). On the progressivity of commodity taxation. In Wolfgang Eichhom (Ed.), Models and Measurement of Welfare and Inequality (pp. 448-465). Heidelberg: Springer-Verlag. Young, J. C. (1994, Fall). Factors associated with noncompliance: Evidence from the Michigan Tax Amnesty Program. The Journal of the American Taxation Association, 16, 82-105. Yu, Chung-Ho (2003). Parametric tests. Retrieved from the Internet on March 15, 2003, at http://seamonkey.ed.asu.edu/~alex/teachingWBI/ parametric_test.html. 156 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. APPENDIX 1 Examples of Transaction Code Defînitions Examples (Non-inciusive)of 1RS Individual Master File Non-compliance Transaction Code Definitions Prevalent in Three Tax Year Sample Definition Count within original Sample IRP (Information Returns Processing, e.g., Form 1099) Delinquency Inquiry: Establishes tax account delinquency inquiry within the affected taxpayer account. 1579 Return Filed and Tax Liability Assessed: Establishes a tax module 69,721 Delinquency Penalty: Computer generated assessment of delinquency penalty on returns posted after the due date without reasonable cause and containing penalty interest codes. 4451 Estimated Tax (ES) Penalty: Computer generated or manually assessed penalty for failure to make adequate ES payments. Applicable to form 1040 and others. 15769 Interest Assessed: Computer generated interest that is due. 34,611 Failure to Pay Tax Penalty: Computer generated penalty assessed if return liability and/or examination adjustment is not paid on or before date prescribed for payment. 28,401 Additional Tax Assessment: Additional tax as a result of an adjustment to a tax account. 21,143 Additional Tax Deficiency: Assesses additional tax as a result of an Examination or Collection adjustment to a tax module which contains a specialized transaction code. 973 Interest Assessment on Additional Tax or Deficiency: Assesses computer-generated interest on additional tax or deficiency assessed upon posting of an examination adjustment and issuance of the adjustment notice. 582 Examination Indicator: Computer generated at Service Center Campuses when opening record is posted. Indicates that return has been referred to the Examination or Appeals Division. 4202 Currently not Collectible Account: A balance due account is considered currently not collectible. 1207 157 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Lien Indicator: Federal Tax Lien has been filed for this tax period. 582 Subsequent Payment Check Dishonored: Records a dishonored check issued as a subsequent payment. 332 IRP Underreporter: IRP Underreporter status transaction. 2753 Notes: 1. BMP Business Master File 2. IMF Individual Master File 3. Sole proprietors may simply file a form 1040 Schedule C that would be captured on the IMF. 4. CP computer paragraph is the IRS= terminology for a notice sent to the taxpayer for a number of different reasons (e.g., incorrect calculation, improper with holding, etc.). 158 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C D ■ D O Q . C g Q . ■ D CD C/) C/) APPENDIX 2 8 ci' Final Cluster Notes 3 3 " CD CD ■ D O Q . C a O 3 " O O CD Q . ■ D CD C /) C /) Tax Year City Cluster Number* N Regression? (Yes/No)** Observations/Notes 1999 Bakersfield 3 48 Yes Home mortgage and car and truck expenses fully populated 1999 Bell 10 2 No Same mortgage expenses (real estate checked), car and truck expenses 1999 Bell Gardens 3 11 No No home mortgage expenses 1999 Compton 2 37 Yes No reported travel and medical costs, home mortgage and car and tmck expenses fully populated 1999 Compton 9 4 No High repair costs 1999 Cudahy No Not significant pattern in clusters 1999 Fresno 1 31 Yes Home mortgage and car and truck expenses fully populated 1999 Fresno 2 5 No Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis) 1999 Fresno 11 4 N Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis) 1999 Fresno 13 3 N Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis) 1999 Long Beach 1 6 N Significant medical costs. Home mortgage and car and tmck expenses fully populated. (Cluster too small for analysis) 1999 Long Beach 2 36 Yes Home mortgage and car and tmck expenses fully populated, however medical/dental, travel, and meals and entertainment were not included in analysis due too the low number making deductions < 25 percent. v o C D ■ D O Q . C g Q . ■ D CD C/) C/) 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 " O O CD Q . ■ D CD C /) C /) Tax Year City Cluster Number* N Regression? (Yes/No)** Observations/Notes 1999 Long Beach 8 8 No Significant medical costs. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 1999 Long Beach 1 3 3 No Significant medical costs. 2 out of 2 cases noncompliant. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 1999 Maywood 8 5 No 3 out of 5 noncompliant. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 1999 Maywood 14 2 No Medical-and dental populated, both cases noncompliant. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 1999 Modesto 3 8 N Medical and dental. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 1999 Modesto 8 4 N Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 1999 Modesto 9 14 No Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 1999 Modesto 10 5 N Other expenses category (Schedule C) fully populated > $25k 1999 Modesto 11 14 N Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 1999 Modesto 12 3 N Significant legal costs. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 1999 Sacramento 1 37 Yes Home mortgage and car and truck expenses fully populated. Medical/dental and travel expenses sparsely populated and therefore not analyzed in the model. 1999 Sacramento 8 16 Yes Home mortgage and car and truck expenses fully populated. O C D ■ D O Q . C g Û . ■ D CD C/) C/) 8 ci' a 3 " CD CD ■ D O Q . C a O 3 " O O CD Q . ■ D CD C /) C /) Tax Year City Cluster Number* N Regression? (Yes/No)** Observations/Notes 1999 Sacramento 11 4 No High medical/dental expenses. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 1999 Sacramento 13 6 No Note car and truck expenses which where > $30k. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 1999 Stockton 1 20 Yes Home mortgage and car and truck expenses fully populated. Travel sparsely populated and therefore not included in analysis. 1999 Stockton 2 X No Home mortgage expenses not populated. Significant car and truck expenses range. Possible that truck is essentially their home. 1999 Stockton 3 4 No Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). All cases include expenses for medical/dental. 1999 Stockton 7 15 Yes Home mortgage and car and truck expenses fully populated. Medical/dental and travel costs sparsely populated and therefore, not included in the analysis. 1999 Tulare 5 2 No Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 1999 Tulare 7 7 No High legal fees, home mortgage, and Other expenses. Car and truck expenses fully populated. High total deductions. (Cluster too small for analysis). 1999 Turlock 6 8 No Significant number of noncompliant cases. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 1999 Yuba City 4 8 No Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). O n C D ■ D O Q . C g Q . ■ D CD C/) C/) 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 " O O CD Q . ■ D CD C /) C /) Tax Year City Cluster Number* N Regression? (Yes/No)** Observations/Notes 1999 Yuba City 4 3 No Significant medical/dental expenses for all cases. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2000 Bakersfield 1 58 No There are no home mortgage expenses reported for these cases - not analyzed. All cases have significant meals and entertainment costs. 2000 Bakersfield 7 4 No 100% noncompliant. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2000 Bakersfield 8 13 No High level of noncompliance. High dollar amount for medical expenses. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2000 Bakersfield 9 6 N High level of noncompliance with significant medical/dental costs. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2000 Bakersfield 10 3 No All cases are noncompliant. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2000 Bakersfield 13 10 No High medical expenses. Legal costs in a number of cases are similar. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2000 Bell 2 17 No High numbers of cases are noncompliant. Cases do not have home mortgage expenses (except for two that have the cost) and were not analyzed. 2000 Bell Gardens 4 2 No Both cases noncompliant, similar repair costs. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). S C D ■ D O Q . C s Q . ■ D CD C /) C /) 8 ( O ' 3. 3 " CD CD ■ D O Q . C a o 3 " O o CD Q . ■ D CD C /) C /) Tax Year City Cluster Number* N Regression? (Yes/No)** Observations/Notes 2000 Bell Gardens 6 11 No Cases do not have home mortgage expenses and were not analyzed. 2000 Compton 2 17 No Only 4 cases had home mortgage expenses and so this cluster was not analyzed. 2000 Compton 7 18 Yes Home mortgage and car and truck expenses fully populated. Medical/dental, travel and legal costs sparsely populated and were not included in the analysis. 2000 Cudahy 1 8 No 5 out of 8 cases noncompliant, 7 out of cases reported legal expenses. 2000 Cudahy 2 27 No 26 out of 27 cases noncompliant, only 2 cases reported home mortgages so this sample was not analyzed. 2000 Fresno 1 23 No Only 5 cases have home mortgages, sample was not analyzed. 2000 Fresno 2 31 No There was no home mortgage expenses claimed in this sample therefore, it was not analyzed. 2000 Fresno 3 31 No 22 out of 31 cases noncompliant. 27 cases reported legal expenses. There was no home mortgage expenses claimed in this sample therefore, it was not analyzed. 2000 Fresno 4 16 Yes Home mortgage and car and truck expenses fully populated. 2000 Fresno 8 6 No 5 out of the 6 cases noncompliant. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2000 Long Beach 2 60 No 44 out of the 60 cases noncompliant. 47 claimed legal expenses and no home mortgage expenses claimed, therefore sample was not analyzed. 2000 Long Beach 4 8 No Cases do not have home mortgage expenses (except for two that have the cost) and were not analyzed C D ■ D O Q . C g Q . ■ D CD C/) C/) 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 " O O CD Q . ■ D CD C /) C /) Tax Year City Cluster Number* N Regression? (Yes/No)** Observations/Notes 2000 Bell Gardens 6 1 1 No Cases do not have home mortgage expenses and were not analyzed. 2000 Long Beach 10 38 Yes Home mortgage and car and truck expenses fully populated. There was no travel reported for these cases and it was excluded from the analysis.. 2000 Maywood 8-15 X No Clusters 8-15 had only one case each. There was no meaningful pattern within the data for Maywood. 2000 Modesto 4 38 Yes Home mortgage and car and truck expenses fully populated. 2000 Sacramento 2 40 No Only 7 cases have home mortgages, sample was not analyzed. 2000 Sacramento 7 36 Yes Home mortgage and car and truck expenses fully populated. 2000 Sacramento 10 6 No 4 out of the 6 cases were noncompliant. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2000 Sacramento 1 1 28 Yes Home mortgage and car and truck expenses fully populated. Travel was sparsely populated and was not included in the analysis. 2000 Stockton 1 6 No Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2000 Stockton 2 22 Yes Home mortgage and car and truck expenses fully populated. 2000 Stockton 3 1 1 No 9 out of 11 noncompliant, 3 cases and 2 other cases had the same legal costs. There were no home mortgage expenses. 2000 Stockton 4 17 No 14 out of 17 noncompliant however, there were no home mortgage expenses. 2000 Stockton 6 6 No 4 out of 6 noncompliant, only 1 home mortgage, high legal expenses. 2000 Stockton 7 2 No High medical/dental, both cases noncompliant. 2000 Stockton 9 5 No 4 out of 5 noncompliant, high medical/dental and all claimed legal expenses. C D ■ D O Q . C g Q . ■ D CD C/) C/) 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 " O O CD Q . ■ D CD C /) C /) o\ L A Tax Year City Cluster Number* N Regression? (Yes/No)** Observations/Notes 2000 Bell Gardens 6 11 No Cases do not have home mortgage expenses and were not analyzed. 2000 Tulare 8 4 No Similar home mortgage costs (Cluster too small for analysis). 2000 Turlock 5 4 No Medical/dental costs fully populated range from. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2000 Turlock 6 3 No Similar car and truck expenses. 2000 Turlock 9 5 No 4 out of 5 noncompliant, high car and truck expenses. 2000 Turlock 13 3 No Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2000 Yuba City 1 1 1 No Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2000 Yuba City 2 28 No 19 out of 21 noncompliant, 6 cases had the same legal expenses, no home mortgage expenses reported. 2000 Yuba City 4 9 No Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2000 Yuba City 7 5 No Similar car and truck expenses . Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2001 Bakersfield 5 7 No High meal and entertainment expenses. 2001 Bakersfield 9 9 No 7 out of 9 cases noncompliant, high medical/dental expenses. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2001 Bakersfield 12 18 Yes Home mortgage and car and truck expenses fully populated. 2001 Bell X X No No pattern observed in the data. 2001 Bell Gardens 1 10 No 8 out of 10 cases noncompliant, no home mortgage expenses reported. 2001 Bell Gardens 5 5 No 4 out of 5 cases noncompliant. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 ■ D O CD Q . ■ D CD C /) C /) g: Tax Year City Cluster Number* N Regression? (Yes/No)** Observations/Notes 2000 Bell Gardens 6 11 No Cases do not have home mortgage expenses and were not analyzed. 2001 Compton 1 25 No 16 out of 25 noncompliant, no home mortgage expenses reported. 2001 Compton 2 5 No Significant repair costs. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2001 Compton 7 4 No 100% noncompliant However, cluster too small for additional statistical analysis. 2001 Compton 9 4 No 100% noncompliant. However, cluster too small for additional statistical analysis. 2001 Compton 10 25 Yes Medical and travel expenses sparsely populated and not included in analysis. Home mortgage and car and truck expenses fully populated. 2001 Compton 12 4 No 2 cases have the same legal expenses, high medical and dental expenses. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2001 Cudahy 2 13 No 9 out of 13 cases were noncompliant, no home mortgage expenses. 2001 Fresno 1 32 No No home mortgage expenses reported. Therefore, cluster was not analyzed. 2001 Fresno 3 11 No High Other (Schedule C) expenses. Cluster too small for analysis. 2001 Fresno 4 16 Yes Home mortgage and car and truck expenses fully populated. 2001 Long Beach 1 48 Yes Travel and medical/dental not populated, home mortgage and car and truck expenses fully populated. 2001 Long Beach 3 49 No 34 out of 49 cases noncompliant, no home mortgage expenses reported so, cluster was not analyzed. 2001 Long Beach 12 25 No Car and truck expenses, no home mortgage expenses reported. C D ■ D O Q . C g Q . ■ D CD C/) C/) 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 " O O CD Q . ■ D CD C /) C /) Tax Year City Cluster Number* N Regression? (Yes/No)** Observations/Notes 2000 Bell Gardens 6 11 No Cases do not have home mortgage expenses and were not analyzed. 2001 Maywood 9 3 No 100% noncompliance, car and truck. 2001 Modesto 1 31 Yes 17 out of 31 noncompliant, home mortgage and car and truck expenses fully populated 2001 Modesto 9 17 Yes Home mortgage and car and truck expenses fully populated 2001 Sacramento 1 33 Yes Home mortgage and car and truck expenses fully populated 2001 Sacramento 2 49 No 23 out of the 49 cases noncompliant. Three sets of cases had the same reported legal expenses. There was no home mortgage expenses reported. 2001 Sacramento 3 128 No 55 out of the 128 cases noncompliant. Multiple duplicate legal costs. There was no home mortgage expenses reported. 2001 Sacramento 5 40 Yes Medical and dental costs sparsely populated and were excluded from the analysis. Home mortgage and car and truck expenses fully populated 2001 Sacramento 9 12 No 6 out of the 12 cases are noncompliant. Medical and dental is fully populated. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2001 Sacramento 12 2 No High repair costs. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2001 Sacramento 15 3 No High Other expenses, home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2001 Stockton 3 36 Yes Home mortgage and car and truck expenses fully populated 2001 Stockton 4 9 No All cases have reported medical and dental expenses and high Other expenses.. 2001 Tulare X X No No recognizable patterns 2001 Turlock 1 16 Yes Home mortgage and car and truck expenses fully populated C D ■ D O Q . C g Q . ■ D CD C/) C/) 8 ci' Tax Year City Cluster Number* N Regression? (Yes/No)** Observations/Notes 2000 Bell Gardens 6 1 1 No Cases do not have home mortgage expenses and were not analyzed. 2001 Turlock 2 2 No Both cases are noncompliant, high Other expenses. Home mortgage and car and truck expenses fully populated. (Cluster too small for analysis). 2001 Yuba City 3 21 Yes Home mortgage and car and truck expenses fully populated 3 3 " CD CD ■ D O Q . C a O 3 " O O CD Q . ■ D CD C /) C /) ON 00 C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O APPENDIX 3 8 ci' Detailed Cluster Notes 2000 Compton Cluster 7 of 15 Descriptive Statistics (Part 1) 3 3 " CD CD ■ D O Q . C a o 3 ■ D O CD Q . ■ D CD C /) C /) N Range Minimum Maximum M ean Statistic Statistic Statistic Statistic Statistic Std. Error CARTRUCK_C 18 7406.389 2096.565 COMPLIANCE 18 0.555556 0.120517 DEPREC_C 18 1842.833 728.2539 HOME_MORT 18 9367.056 676.8087 LEGAL_C 18 98.27778 39.43588 M&E_C 18 455.1667 185.375 MED_DENTAL 18 Data Unavailable 263.8889 175.267 OTHER_C 18 8021.333 2371.457 REPAIRS_C 18 1382.889 488.2288 STATE_LOC_INC_TAX 18 161.7778 51.20844 TOTALJNCOME 18 28655.72 3472.046 TRAVEL_C 18 51.94444 36.11243 Valid N (listwise) 18 C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O Descriptive Statistics (Part 2) 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 ■ D O Std. Deviation Variance Skewness Kurtosis Statistic Statistic Statistic Std. E rror Statistic Std. E rror CARTRUCK C 8894.974 79120561 1.762854 0.536278 2.385541 1.037795 COMPLIANCE 0.5II31 0.261438 -0.24447 0.536278 -2.19938 1.037795 DEPREC C 3089.72 9546367 2.111884 0.536278 4.923301 1.037795 HOME M ORT 2871.456 8245259 2.64079 0.536278 8.792194 1.037795 LEGAL_C 167.3122 27993.39 1.560353 0.536278 1.185916 1.037795 M&E C 786.4794 618549.8 1.852592 0.536278 2.359577 1.037795 MED DENTAL 743.5948 552933.3 2.828479 0.536278 7.130689 1.037795 OTHER C 10061.24 l.OlE+08 0.817864 0.536278 -1.05375 1.037795 REPAIRS C 2071.379 4290612 1.510018 0.536278 1.320676 1.037795 STATE LOC INC TAX 217.259 47201.48 1.286083 0.536278 0.538 1.037795 TOTAL_INCOME 14730.65 2.17E+08 0.766159 0.536278 -0.13923 1.037795 TRAVEL C 153.2121 23473.94 2.840723 0.536278 7.143357 1.037795 Valid N (listwise) CD Q . ■ D CD C /) C /) C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 2000 Compton Cluster 7 of 15 Correlations (Part 1) 8 ( O ' 3. 3 " CD CD ■ D O Q . C a o 3 ■ D O CD Q . ■ D CD C /) C /) COMP2 CARTRU 2 DEPREC2 HOME2 INCOME2 Pearson Correlation C0MP2 1.000 .099 .105 .205 .619 CARTRU 2 .099 1.000 .261 .191 .082 DEPREC2 .105 .261 1.000 .040 .229 H0ME2 .205 .191 .040 1.000 .177 INC0ME2 .619 .082 .229 .177 1.000 LEGAL2 .089 .005 .085 -.055 -.154 M E 02 -.144 .611 .193 .296 -.303 0THER2 .329 -.152 .252 .215 -.233 REPAIR2 .221 -J96 .147 .030 .003 STATE2 .020 .503 .082 -.113 .491 Sig. (1-tailed) C0MP2 .348 J39 .207 .003 CARTRU_2 .348 .148 .223 .373 DEPREC2 J39 .148 .437 .181 H0ME2 .207 .223 .437 .241 INC0ME2 .003 .373 .181 .241 LEGAL2 .362 .492 .369 .415 .271 M_E_C2 .285 .004 .222 .116 .110 0THER2 .091 .273 .157 .196 .176 REPAIR2 .190 .052 .280 .453 .496 STATE2 .469 .017 .374 .328 .019 M I I I ! o ü 8 8 ü u I u M , a 172 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. I i I ? m I $ I o s o § I $ 1 I R S % § u M I u I § R 8 ( 8 I 8 i I 8 I I § I § ■ s ( £ 173 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. : : I 0 M I 174 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C D ■ D O Q . C g Q . ■ D CD C/) C/) 2000 Compton Cluster 7 of 15 Model Summary (b) 8 ( O ' R R Square Adjusted R Square Std. Error of the Estimate Change Statistics Durbin- Watson Model R Square Change F Change dn dt2 Sig. F Change 1 .895(a) .801 .577 333 .801 3.577 9 8 .043 2.621 3. 3 " CD a Predictors: (Constant), STATE2, M_E„C2, DEPREC2, H0ME2, LEGAL2, INC0ME2, REPAIR2, CARTRU_2, 0THER2 b Dependent Variable: COMP2 CD ■ D O Q . C a o 3 " O o CD Q . ■ D CD C /) C /) ANC VA (b) Model Sum of Squares df Mean Square F Sig. 1 Regression 3.560 9 .396 3.577 .043(a) Residual .885 8 .111 Total 4.444 17 a Predictors: (Constant), STATE2, M_E_C2, DEPREC2, H0ME2, LEGAL2, INCOME2, REPAIR2, CARTRU_2, 0THER2 b Dependent Variable: COMP2 < 1 V I C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 2000 Compton Cluster 7 of 15 Coefficients (a) 8 ( O ' 3. 3 " CD CD ■ D O Q . C a o 3 ■ D O CD Q . ■ D CD C /) C /) Model Unstandardized Coefficients Standardized Coefficients t % 95% Confidence Interval for B Correlations Collinearity Statistics B Std. E rror Beta Lower Bound Upper Bound Zero- order Partial Part Tolerance VIE 1 (Constant) -4.785 3.359 -1.424 .192 -12.530 2.961 CARTRU_2 .288 .272 J35 1.058 .321 -.339 .915 .099 .350 .167 .248 4.032 DEPREC2 -.406 .245 -.311 -1.658 .136 -.970 .159 .105 -.506 -.262 .706 1.416 H0ME2 -.391 .874 -.083 -.448 .666 -2.408 1.625 .205 -.156 -.071 .725 1.379 INC0ME2 1.543 .525 .720 2.936 .019 .331 2.755 .619 .720 .463 .413 2.419 LEGAL2 -.171 .298 -.136 -.574 .582 -.859 .517 .089 -.199 -.090 .441 2.268 M E C2 -.601 .406 -.453 -1.479 .177 -1.538 .336 -.144 -.463 -.233 .265 3.768 0THER2 1.760 .623 1.255 2.826 .022 .324 3.197 .329 .707 .446 .126 7.925 REPAIR2 -.612 .436 -.445 -1.405 .198 -1.616 .392 .221 -.445 -.222 .248 4.039 STATE2 .269 .608 .187 .443 .670 -1.134 1.672 .020 .155 .070 .140 7.155 a Dependent Variable: C0MP2 as C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 2000 Compton Cluster 7 of 15 Coefficient Correlations (a) (Part 1) 8 ( O ' 3. 3 " CD CD ■ D O Q . C a o 3 ■ D O CD Q . ■ D CD Model STATE2 M E C2 DEPREC2 HOME2 LEGAL2 1 Correlations STATE2 1.000 -.013 -.180 .192 -.134 M E C2 -.013 1.000 -.072 -.201 .404 DEPREC2 -.180 -.072 1.000 .160 .079 H0ME2 .192 -.201 .160 1.000 .018 LEGAL2 -.134 .404 .079 .018 1.000 INC0ME2 -.506 .394 -.170 -.383 .338 REPAIR2 -.143 .262 -.024 -.017 -.329 CARTRU_2 -.527 -.549 -.068 -.128 -.371 0THER2 .715 -.363 -.279 .002 -.243 Covariances STATE2 .370 -.003 -.027 .102 -.024 M E C2 -.003 .165 -.007 -.071 .049 DEPREC2 -.027 -.007 .060 .034 .006 H0ME2 .102 -.071 .034 .764 .005 LEGAL2 -.024 .049 .006 .005 .089 INCOME2 -.162 .084 -.022 -.176 .053 REPAIR2 -.038 .046 -.003 -.007 -.043 CARTRU 2 -.087 -.061 -.005 -.031 -.030 0THER2 .271 -.092 -.043 .001 -.045 a Dependent Variable: C0MP2 C /) C /) C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O Coefficient Correlations (a) (Part 2) 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 ■ D O CD Q . Model 1NCOME2 REPAIR2 CARTRU 2 OTHER2 1 Correlations STATE2 -.506 -.143 -.527 .715 M E C2 .394 .262 -.549 -.363 DEPREC2 -.170 -.024 -.068 -.279 H0ME2 -.393 -.017 -.128 .002 LEGAL2 .338 -.329 -.371 -.243 INCOME2 1.000 -.079 .018 -.284 REPAIR2 -.079 1.000 .234 -.542 CARTRU 2 .018 .234 1.000 -.204 0THER2 -.284 -.542 -.204 1.000 Covariances STATE2 -.162 -.038 -.087 .271 M E C2 .084 .046 -.061 -.092 DEPREC2 -.022 -.003 -.005 -.043 H0ME2 -.176 -.007 -.031 .001 LEGAL2 .053 -.043 -.030 -.045 INC0ME2 .276 -.018 .003 -.093 REPAIR2 -.018 .190 .028 -.147 CARTRU 2 .003 .028 .074 -.034 OTHER2 -.093 -.147 -.034 .388 ■ D CD a Dependent Variable: C0MP2 c /) c /) < 1 0 0 C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 2000 Compton Cluster 7 of 15 Residuals Statistics (a) 8 ( O ' 3. 3 " CD CD ■ D O Q . C a o 3 ■ D O Minimum Maximum Mean Std. Deviation N Predicted Value .81 2.21 1.56 .458 18 Std. Predicted Value -1.632 1.436 .000 1.000 18 Standard Error of Predicted Value .185 .308 .245 .038 18 Adjusted Predicted Value -.07 2.34 1.50 .619 18 Residual -.44 .48 .00 .228 18 Std. Residual -1.333 1.430 .000 .686 18 Stud. Residual -1.858 1.719 .037 1.036 18 Deleted Residual -1.18 1.07 .05 .592 18 Stud. Deleted Residual -2.305 2.025 .017 1.161 18 Mahal. Distance 4.292 13.616 8.500 2.863 18 Cook's Distance .002 .912 .201 .318 18 Centered Leverage Value .252 .801 .500 .168 18 a Dependent Variable: COMP2 CD Q . ■ D CD C /) C /) lo ' T " 0 K a 1 I I 3 a: I ZdlAlOO 180 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. « o 0 Is. 1 C I I o C L C o c ô 2 .2 •e m CL a ! S 8 .2 k« > I E 0) CL Q o o o O o LU Z cinoo 181 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 2000 Fresno Cluster 2 of 15 Descriptive Statistics (Part 1) 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 ■ D O CD Q . N Range Minimum Maximum Mean Statistic Statistic Statistic Statistic Statistic Std. Error CARTRUCK_C 31 5835.839 1422.753 COMPLIANCE 31 0.741935 0.079889 DEPREC_C 31 5255 892.2923 HOME_MORT 31 8684.645 509.9564 LEGAL_C 31 252.9677 69.95826 M&E_C 31 1000 211.7094 MED_DENTAL 31 Data Unavailable 1046.355 260.625 OTHER_C 31 25046.39 3718.72 REPAIRS_C 31 5513.677 888.5761 STATE_LOC_INC_TAX 31 704.2581 175.2375 TOTAL_INCOME 31 41053.03 3501.836 TRAVEL_C 31 396.2581 131.1437 Valid N (listwise) 31 ■ D CD C /) C /) 00 t o C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O Descriptive Statistics (Part 2) 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 ■ D O Std. Deviation Variance Skewness Kurtosis Statistic Statistic Statistic Std. Error Statistic Std. Error CARTRUCK C 7921.554 62751017 2.086678 0.420536 3.215501 0.820803 COMPLIANCE 0.444803 0.197849 -1.16286 0.420536 -0.69715 0.820803 DEPREC C 4968.073 24681751 0.851054 0.420536 0.236442 0.820803 HOME MORT 2839.317 8061720 0.798288 0.420536 0.41367 0.820803 LEGAL C 389.5111 151718.9 2.699598 0.420536 7.536674 0.820803 M&E C 1178.748 1389448 1.055643 0.420536 0.309482 0.820803 MED DENTAL 1451.099 2105688 1.157069 0.420536 -0.1153 0.820803 OTHER C 20704.96 4.29E+08 0.944452 0.420536 1.343421 0.820803 REPAIRS C 4947.382 24476593 0.630808 0.420536 -0.73042 0.820803 STATE LOC INC TAX 975.6813 951954 1.844308 0.420536 2.968354 0.820803 TOTAL INCOME 19497.4 3.8E+08 0.776 0.420536 -0.02243 0.820803 TRAVEL C 730.1774 533159.1 2.722549 0.420536 9.056111 0.820803 Valid N (listwise) CD Q . ■ D CD C /) C /) 00 w C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 2000 Fresno Cluster 2 of 15 Correlations (Part 1) 8 ( O ' 3. 3 " CD CD ■ D O Q . C a o 3 ■ D O CD Q . ■ D CD C /) C /) COMP2 CARTRU 2 DEPREC2 HOME2 INC0ME2 LEGAL2 Pearson Correlation C0MP2 1.000 -.379 .283 -.117 .725 -.096 CARTRU 2 -.379 1.000 -.209 .336 -.149 -.109 DEPREC2 .283 -.209 1.000 -.142 .416 .498 H0ME2 -.117 .336 -.142 1.000 .236 -.029 INCOME2 .725 -.149 .416 .236 1.000 .021 LEGAL2 -.096 -.109 .498 -.029 .021 1.000 M E C2 .243 -.214 .137 -.312 -.019 .056 MEDEN2 .293 -.437 -.069 -.484 .042 -.045 0THER2 .353 -.379 .516 -.391 .236 .265 REPAIR2 .093 .045 .066 .149 .153 .114 STATE2 .496 -.226 .402 .254 .723 -.034 TRAVEL2 .314 -.161 .477 -.274 .129 .333 Sig. (1-tailed) C0MP2 .018 .061 .265 .000 .304 CARTRU 2 .018 .129 .032 .212 .280 DEPREC2 .061 .129 .223 .010 .002 HOME2 .265 .032 .223 .101 .438 INCOME2 .000 .212 .010 .101 .456 LEGAL2 .304 .280 .002 .438 .456 M E C2 .094 .124 .231 .044 .460 .382 MEDEN2 .055 .007 .357 .003 .412 .404 0THER2 .026 .018 .001 .015 .101 .075 REPAIR2 .310 .404 .362 .211 .206 .270 STATE2 .002 .111 .012 .084 .000 .428 TRAVEL2 .043 .194 .003 .068 .245 .034 2 i ! I o I I I u I w . I 2 i 185 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. R I I o £ I 2 i i i i $ I I I I I r- m o s I I s i ( S o I i I § i m 2 « m 0 M • r t - f N ( N m i I i I I I I I 8 w , I 2 i 5 I § § ■i 186 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 ■ D O M E C2 MEDEN2 OTHER2 REPAIR2 STATE2 TRAVEL2 N C0MP2 31 31 31 31 31 31 CARTRU 2 31 31 31 31 31 31 DEPREC2 31 31 31 31 31 31 H0ME2 31 31 31 31 31 31 INC0ME2 31 31 31 31 31 31 LEGAL2 31 31 31 31 31 31 M E C2 31 31 31 31 31 31 MEDEN2 31 31 31 31 31 31 0THER2 31 31 31 31 31 31 REPAIR2 31 31 31 31 31 31 STATE2 31 31 31 31 31 31 TRAVEL2 31 31 31 31 31 31 Model Summary (b) CD Q . ■ D CD R R Square Adjusted R Square Std. Error of the Estimate Change Statistics Durbin- Watson Model R Square Change F Change dn df2 Sig. F Change 1 ■873(a) .762 .624 .273 .762 5.528 1 1 19 .001 1.625 C /) C /) a Predictors: (Constant), TRAVEL2, M _E__C 2, ÎNC0ME2, CARTRU_2, LEGAL2, REPAIR2, HOME2, MEDEN2, DEPREC2, OTHER2, STATE2 b Dependent Variable: C0MP2 0 0 < 1 C D ■ D O Q . C g Q . ■ D CD C/) o ' 3 O ANOVA(b) 8 ( O ' Model Sum of Squares df Mean Square F Sis. 1 Regression 4.522 1 1 .411 5.528 .001(a) Residual 1.413 19 .074 Total 5.935 30 a Predictors: (Constant), TRAVEL2, M_E_C2, INC0ME2, CARTRU_2, LEGAL2, REPAIR2, H0ME2, MEDEN2, DEPREC2, 0THER2, STATE2 b Dependent Variable: C0MP2 3. 3 " CD CD ■ D O Q . C a o 3 ■ D O CD Q . ■ D CD ( / ) ( / ) 00 00 C D ■ D O Q . C g Q . ■ D CD C/) o' 3 O 2000 Fresno Cluster 2 of 15 Coefficients (a) 8 ( O ' 3. 3 " CD CD ■ D O Q . C a o 3 ■ D O CD Q . " O CD C /) C /) Model Unstandardized Coefficients Standardized Coefficients t Sig. 95% Confidence Interval for B Correlations Collinearity Statistics B Std. E rror Beta Lower Bound Upper Bound Zero- order Partial Part Tolerance VIP 1 (Constant) -2.451 2.152 -1.139 369 -6.956 2.054 CARTRU_2 -.141 .152 -.152 -.927 .366 -.460 .178 -.379 -.208 -.104 .466 2.148 DEPREC2 -.274 J96 -.180 -.926 366 -.894 .346 .283 -.208 -.104 333 3.005 H0ME2 -.420 .497 -.134 -.845 .408 -1.460 .620 -.117 -.190 -.095 .498 2.009 INC0ME2 1.391 .485 .657 2.871 .010 377 2.405 .725 .550 .321 .239 4.181 LEGAL2 -.216 .204 -.153 -1.059 303 -.643 .211 -.096 -336 -.119 .598 1.671 M E C2 .279 .186 .215 1.500 .150 -.110 .668 .243 .325 168 .610 1.640 MEDEN2 .120 .229 .100 .523 .607 -.360 .600 .293 .119 .059 .340 2.942 0THER2 -.181 .382 -.103 -.475 .640 -.980 .617 .353 -.108 -.053 .264 3385 REPAIR2 .217 .263 .136 .824 .420 -.334 .769 .093 .186 .092 .458 2.184 STATE2 .206 353 .153 .584 .566 -.533 .946 .496 .133 .065 .184 5.446 TRAVEL2 .422 .181 368 2.324 .031 .042 .801 .314 .470 .260 .500 1.999 a Dependent Variable: C0MP2 0 0 \ o C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O Residuals Statistics (a) 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 ■ D O Minimum Maximum Mean Std. Deviation N Predicted Value .89 2J9 1.74 .388 31 Std. Predicted Value -2.186 1.671 .000 1.000 31 Standard E rror of Predicted Value .111 .210 .168 .025 31 Adjusted Predicted Value .83 2.73 1.76 .407 31 Residual -39 .42 .00 .217 31 Std. Residual -1.433 1.553 .000 .796 31 Stud. Residual -1.965 1.887 -.023 1.013 31 Deleted Residual -.73 .64 -.02 J55 31 Stud. Deleted Residual -2.143 2.037 -.024 1.060 31 Mahal. Distance 3.975 16.896 10.645 3.382 31 Cook's Distance .000 J83 .054 .072 31 Centered Leverage Value .132 j63 .355 .113 31 a Dependent Variable: C0MP2 CD Q . ■ D CD C /) C /) lo Y ~ 0 CM 1 I 0 1 I 0 C L c ■« 1 .g ■e (0 C L □ a % O g I ZdlAlOO 191 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. > o r- O CM § I I 0 Q. c w Ï 1 t f a . □ DO L U Zd I /M O O 192 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C D " O O Q . C S Q . ■ D CD C/) (fi 2000 Modesto Cluster 4 of 15 Descriptive Statistics (Part 1) 3 3 " CD CD ■ D O Q . C a O 3 " O O CD Q . N Range Minimum Maximum Mean Statistic Statistic Statistic Statistic Statistic Std. Error CARTRUCK_C 38 7129.053 1513.979 COMPLIANCE 38 0.815789 0.06373 DEPREC_C 38 2994.737 523.0067 HOME MORT 38 7712.421 351.9816 LEGAL_C 38 213.2105 72.21106 M&E_C 38 1045.368 237.4868 MED_DENTAL 38 Data Unavailable 1140.632 229.2193 OTHER_C 38 23857.34 3390.275 REPAIRS_C 38 4294.684 699.3896 SCH_C_WAGE 38 26.71053 26.71053 STATE_LOC_INC_TAX 38 168.3158 41.98681 TOTALJNCOME 38 31159.84 1706.966 TRAVEL_C 38 133.8421 48.25872 Valid N (listwise) 38 ■ D CD C /) C /) C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 ■ D O Descriptive Statistics (Part 2) Std. Deviation Variance Skewness Kurtosis Statistic Statistic Statistic Std. Error Statistic Std. Error CARTRUCK C 9332.794 87101046 1.927745 0.382818 3.171677 0.7497 COMPLIANCE 0.392859 0.154339 -1.69696 0.382818 0.925609 0.7497 DEPREC C 3224.03 10394369 1.593053 0.382818 2.39291 0.7497 HOME MORT 2169.76 4707860 1.054834 0.382818 0.829461 0.7497 LEGAL C 445.1389 198148.6 4.259775 0.382818 21.00551 0.7497 M&E C 1463.967 2143199 2.271691 0.382818 6.012908 0.7497 MED DENTAL 1413.002 1996576 0.993438 0.382818 0.047977 0.7497 OTHER C 20899.06 4.37E+08 1.529779 0.382818 3.70522 0.7497 REPAIRS C 4311.327 18587542 1.05863 0.382818 1.007819 0.7497 SCH C WAGE 164.6547 27111.18 6.164414 0.382818 38 0.7497 STATE LOC INC TAX 258.8241 66989.9 2.318352 0.382818 5.29916 0.7497 TOTAL INCOME 10522.45 l.llE+08 0.343703 0.382818 -0.52099 0.7497 TRAVEL C 297.4867 88498.35 3.226462 0.382818 11.22254 0.7497 Valid N (listwise) CD Q . ■ D CD C /) C /) C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 2000 Modesto Cluster 4 of 15 Correlations (Part 1) 8 ( O ' c 3. 3 " CD CD ■ D O Q . C a o 3 ■ D O CD Q . ■ D CD C /) C /) COMP2 CARTRU 2 DEPREC2 H0ME2 INCOME2 LEGAL2 M E 02 Pearson Correlation C0MP2 1.000 -.027 -.351 .457 .131 .091 .037 CARTRU 2 -XB7 1.000 -.004 .218 -.013 -.271 -.347 DEPREC2 -.351 -.004 1.000 -.151 -.013 -.040 .080 H0ME2 .457 .218 -.151 1.000 .187 .024 .011 1NC0ME2 .131 -.013 -.013 .187 1.000 .116 .003 LEGAL2 .091 -.271 -.040 .024 .116 1.000 -.036 M E C2 .037 -.347 .080 .011 .003 -.036 1.000 MEDEN2 .255 -.143 .156 .179 .236 .351 .082 0THER2 -.136 -.684 -.057 -.041 .067 .400 .338 REPAIR2 .065 -.519 .008 .128 .121 .030 .351 STATE2 .131 -.042 .077 .027 .551 .033 -.144 TRAVEL2 -.150 -.086 -.136 -.126 .072 .367 .269 Sig. (1-tailed) C0MP2 .435 .015 .002 .216 .294 .412 CARTRU 2 .435 .490 .094 .468 .050 .016 DEPREC2 .015 .490 .183 .469 .405 .317 H0ME2 .002 .094 .183 .131 .443 .473 [NCOME2 .216 .468 .469 .131 .244 .493 LEGAL2 .294 .050 .405 .443 .244 .414 M E C2 .412 .016 .317 .473 .493 .414 MEDEN2 .061 .196 .175 .141 .077 .015 .312 0THER2 .208 .000 .366 .403 .345 .006 .019 REPAIR2 .350 .000 .482 .222 .235 .429 .015 STATE2 .216 .400 .324 .436 .000 .422 .195 TRAVEL2 .184 .304 .209 .225 .333 .012 .051 L A u M S 0 0 ^ m " o S " m W i o o m 00 m 0 0 m 00 m 00 00 00 < T , I 00 m 00 m 00 m i 00 m 00 m i oo f C oo m ! 00 <n 00 m 00 m oo M . u I I w, i I 196 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. i i I I S i oo N i m O i i o « n i i oo § o o i % m g % o m I i n « n i i i 2 i g m « r i « os i i $ s 2 n I i Ê o U â u i 3 tL I I i •a 197 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. oo m : 00 m I 00 r o 00 m oo m 00 : oo ro 00 m oo 00 < T 5 I 00 00 m 00 e n I u ï I i 198 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O Model Summary (b) 8 ci' Model R R Sauare Adjusted R Sauare Std. Error of the Estimate Change Statistics R Square Change F Change dfl df2 Durbin- Watson 1 .712(a) .508 .299 .329 jOB 2.437 11 26 .030 2.169 3 3 " CD a Predictors: (Constant), TRAVEL2, STATE2, MEDEN2, CARTRU_2, DEPREC2, H0ME2, M_E_C2, LEGAL2, INC0ME2, REPAIR2, OTHER2 b Dependent Variable; COMP2 CD ■ D O Q . C a O 3 ■ D O CD Q . ANC VA (b) Model Sum of Squares df Mean Sauare F Sig. 1 Regression 2.899 1 1 .264 2.437 .030(a) Residual 2.812 26 .108 Total 5.711 37 ■ D CD C /) C /) a Predictors: (Constant), TRAVEL2, STATE2, MEDEN2, CARTRU_2, DEPREC2, H0ME2, M_E_C2, LEGAL2, INC0ME2, REPAIR2, OTHER2 b Dependent Variable: C0MP2 C D ■ D O Q . C S Q . ■ D CD C/) o " 3 O 2000 Modesto Cluster 4 of 15 Coefficients (a) 8 ci' 3 CD 3. 3 " CD CD ■ D O Q . C a O 3 ■ D O CD Q . ■ D CD C /) C /) Model Unstandardized Coefficients Standardized Coefficients t % 95% Confidence Interval for B Correlations Collinearity Statistics B Std. E rror Beta Lower Bound Upper Bound Zero- order Partial Part Tolerance VIE 1 (Constant) -2.710 2659 -1.019 .318 -8.176 2.756 CARTRU 2 -.473 .299 -.353 -1.582 .126 -1.087 .142 -.027 -.296 -.218 .381 2j% DEPREC2 -.555 .204 -.402 -2.724 .011 -.973 -.136 -.351 -.471 -.375 .869 1.150 H0ME2 1.344 .536 .393 2.508 .019 .242 2.446 .457 .441 .345 .773 1.294 INCOME2 -.078 .446 -.031 -.174 .863 -.995 .839 .131 -.034 -.024 .594 1.683 LEGAL2 .117 .201 .103 .582 .566 -296 .529 .091 .113 .080 .599 1.668 M E C2 .190 .205 .159 .931 .360 -.230 .611 .037 .180 .128 .649 1.542 MEDEN2 .323 .173 .309 1.863 .074 -.033 .679 .255 .343 .256 .689 1.451 0THER2 -.266 .150 -.444 -1.773 .088 -.574 .042 -.136 -.328 -.244 .302 3.307 REPAIR2 -.133 .271 -.101 -.491 .628 -.691 .425 .065 -.096 -.068 .446 2.241 STATE2 .099 .220 .085 .453 .655 -.352 .551 .131 .088 .062 .539 1.856 TRAVEL2 -.153 .181 -.151 -.846 .405 -.526 .219 -.150 -.164 -.116 .590 1.694 a Dependent Variable: COMP2 § C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O Residuals Statistics (a) 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 ■ D O Minimum Maximum Mean Std. Deviation N Predicted Value 1.26 2.44 1.82 .280 38 Std. Predicted Value -1.997 2.240 .000 1.000 38 Standard E rror of Predicted Value .111 .237 .183 .026 38 Adjusted Predicted Value 1.21 2.68 1.81 .323 38 Residual -.56 .59 .00 .276 38 Std. Residual -1.701 1.784 .000 .838 38 Stud. Residual -2.001 2.065 .003 1.011 38 Deleted Residual -.77 .79 .00 .403 38 Stud. Deleted Residual -2.134 2.214 -.003 1.038 38 Mahal. Distance 3J39 18.315 10.711 3.273 38 Cook's Distance .000 .128 .039 .041 38 Centered Leverage Value .088 .495 .289 .088 38 a Dependent Variable: C0MP2 CD Q . ■ D CD C /) C /) O « o t - 0 1 I I I o Û _ c 0 1 S ’ (T t ( î C N C L S O 0 m .5 5 ‘ i— > 1 0 CL Û o g o o ° o o o % % ° o o o o o ° o ° r, O O O O Û H 6 z d m o 202 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. »o V- 0 1 I I C M 0_ o O o C L Q ) c JO o .2 O T & _ (0 C O s > > a > < D c D C C D T S 1 c C D " t C Ol ( Ü C D Q _ Q o o L U 3d l/M O O 203 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C D ■ D O Q . C g Q . ■ D CD C/) o' 3 O 2001 Sacramento Cluster 1 of 15 Descriptive Statistics (Part 1) 8 C Q ' a 3 * CD CD ■ D O Q . C a o 3 ■ D O CD Û . N Range Minimum Maximum Mean Statistic Statistic Statistic Statistic Statistic Std. Error CARTRUCK_C 33 3392.212 585.8613 COMPLIANCE 33 0.393939 0.086377 DEPRECJ: 33 4269.727 1064.732 HOME_MORT 33 8754.545 472.317 LEGAL_C 33 80.78788 23.49072 M&E_C 33 1481.061 329.7056 MED_DENTAL 33 Data Unavailable 1056.667 271.213 OTHER_C 33 35088.88 2268.95 REPAIRS_C 33 8568.152 877.508 STATE_LOC_INC_TAX 33 179.3636 55.40824 TOTALJNCOME 33 31104.52 2125.688 TRAVEL_C 33 405.1515 130.7717 Valid N (listwise) 33 ■ D CD ( / ) ( / ) § C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O Descriptive Statistics (Part 2) 8 ci' 3 3 " CD CD ■ D O Q . C a O 3 ■ D O Std. Deviation Variance Skewness Kurtosis Statistic Statistic Statistic Std. Error Statistic Std. Error CARTRUCK_C 3365.517 11326705 2.35164 0.408635 5.761264 0.798414 COMPLIANCE 0.496198 0.246212 0.455074 0.408635 -1.91285 0.798414 DEPREC_C 6116.422 37410618 2.727173 0.408635 8.426641 0.798414 HOME MORT 2713.255 7361751 1.00368 0.408635 0.115326 0.798414 LEGAL_C 134.9439 18209.86 1.723814 0.408635 2.418719 0.798414 M&E_C 1894.015 3587291 1.239536 0.408635 0.666871 0.798414 MED_DENTAL 1558 2427365 2.081772 0.408635 5.054225 0.798414 OTHER_C 13034.13 1.7E+08 0.195105 0.408635 -0.17203 0.798414 REPAIRS_C 5040.9 25410668 -0.34349 0.408635 -0.68159 0.798414 STATE_LOC_INC„TAX 318.2961 101312.4 2.479717 0.408635 5.991757 0.798414 TOTAL INCOM E 12211.15 1.49E+08 -0.05483 0.408635 0.082516 0.798414 TRAVEL_C 751.2264 564341.1 1.927555 0.408635 3.07966 0.798414 Valid N (listwise) CD Q . ■ D CD C /) C /) S C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 2001 Sacramento Cluster 1 of 15 Correlations (Part 1) 8 C Q ' 3 3 " CD CD ■ D O Q . C a o 3 ■ D O CD Q . ■ D CD C /) C /) COMP2 CARTRU 2 DEPREC2 HOME2 INC0ME2 LEGAL2 Pearson Correlation C0MP2 1.000 .082 .054 .226 -.155 .051 CARTRU 2 .082 1.000 -.186 .010 -.107 -.057 DEFREC2 .054 -.186 1.000 .231 .146 .069 H0ME2 .226 .010 .231 1.000 .271 .051 INCOME2 -.155 -.107 .146 .271 1.000 .108 LEGAL2 .051 -.057 .069 .051 .108 1.000 M E C2 -.215 -.089 -.087 .123 .172 -.199 MEDEN2 .268 .052 -.187 -.192 .243 -.036 0THER2 .071 -.312 .253 -.165 .120 -.122 REPAIR2 .275 -.033 .036 .127 -.001 -.095 STATE2 -.467 -.135 ^039 -.005 .488 -.066 TRAVEL2 -.192 -.210 -.101 -.010 .207 .051 Sig. (1-tailed) COMP2 .325 .383 .103 .195 .389 CARTRU_2 .325 .151 .479 .276 .376 DEPREC2 .383 .151 .098 .209 .351 H0ME2 .103 .479 .098 .063 .389 INC0ME2 .195 .276 .209 .063 .274 LEGAL2 .389 .376 .351 .389 .274 M_E_C2 .115 .311 .315 .247 .169 .133 MEDEN2 .066 .386 .148 .143 .087 .422 0THER2 .348 .038 .078 .180 .253 .250 REPAIR2 .061 .428 .421 .240 .497 .299 STATE2 .003 .227 .414 .489 .002 .358 TRAVEL2 .143 .120 .288 .477 .124 .389 i U m « i m m f O m en m en O i en en en en U I S en en en en en en en en en en I i w ,i i i H 207 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. o ( S o o N R I M S S 0 \ o s r- R I I 8 I I m O ( S i I g o s s m s 0 0 s O N m R I I i i § I U w s o R R $ i 00 R (S I I U U g U w . k ; i i 3 § I g e u i 208 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 8 C Q ' 3 3 " CD CD ■ D O Q . C a O 3 ■ D O CD Q . ■ D CD C /) C /) M E C2 MEDEN2 OTHER2 REPAIR2 STATE2 TRAVEL N C0MP2 33 33 33 33 33 33 CARTRU 2 33 33 33 33 33 33 DEPREC2 33 33 33 33 33 33 H0ME2 33 33 33 33 33 33 INC0ME2 33 33 33 33 33 33 LEGAL2 33 33 33 33 33 33 M E C2 33 33 33 33 33 33 MEDEN2 33 33 33 33 33 33 0THER2 33 33 33 33 33 33 REPAIR2 33 33 33 33 33 33 STATE2 33 33 33 33 33 33 TRAVEL2 33 33 33 33 33 33 Model Summary (b) Model R R Sauare Adjusted R Sauare Std. Error of the Estimate Change Statistics R Square Change F Change dfl df2 Sig, F Change Durbin- Watson 1 ■772(a) .596 .384 .389 .596 2.813 1 1 21 .020 2.383 a Predictors; (Constant), TRAVEL2, H0ME2, LEGAL2, MEDEN2, REPAIR2, CARTRU_2, M_E_C2, STATE2, DEPREC2, 0THER2, INC0ME2 b Dependent Variable: C0MP2 I C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O ANOVA(b) CD Sum of 8 Model Squares df Mean Square F Sie. 1 Regression 4.694 1 1 .427 2.813 .020(a) C Q 3 " Residual 3.185 21 .152 O Total 7.879 32 a Predictors: (Constant), TRAVEL2, H0ME2, LEGAL2, MEDEN2, REPAIR2, CARTRU_2, M_E_C2, STATE2, DEPREC2, OTHER2, INC0ME2 b Dependent Variable: C0MP2 3. 3 " CD CD ■ D O Q . C a O 3 ■ D O CD Q . ■ D CD C /) C /) C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O 2001 Sacramento Cluster 1 of 15 Coefficients (a) 8 C Q ' 3 3 " CD CD ■ D O Q . C a O 3 ■ D O CD Q . ■ D CD C /) C /) Model Unstandardized Coefficients Standardized Coefficients t S iR . 95% Confidence Interval for B Correlations Collinearity Statistics B Std. Error Beta Lower Bound Upper Bound Zero- order Partial Part Tolerance VIE 1 (Constant) -9.503 3U M 0 -2.394 .026 -17.759 -1.248 CARTRU 2 .079 .196 .061 .404 .690 -.328 .486 .082 .088 .056 .842 1.188 DEPREC2 -.234 .277 -.142 -.846 .407 -.810 .341 .054 -.182 -.117 .685 1.459 H0ME2 1.612 .655 .407 2.460 .023 .249 2.974 .226 .473 .341 .703 1.423 INC0ME2 -.200 J76 -.104 ^533 .600 -.981 .581 -.155 -.115 -.074 .503 1.989 LEGAL2 .079 .185 .064 .428 .673 -.306 .464 .051 .093 .059 .864 1.158 M E C2 -.527 .253 -.369 -2.080 .050 -1.054 .000 -.215 -.413 -.289 .610 1.638 MEDEN2 .423 .223 .310 1.898 .072 -.041 .887 .268 .383 .263 .723 1.384 0THER2 1.129 .503 .425 2.245 .036 .083 2.176 .071 .440 .312 .538 1.858 REPAIR2 .846 334 .370 2.534 .019 .152 1.540 .275 .484 .352 .901 1.109 STATE2 -.487 .248 -.352 -1.964 .063 -1.002 .029 -.467 -.394 -.272 .598 1.671 TRAVEL2 -.182 .194 -.149 -.937 .359 -.586 .222 -.192 -.200 -.130 .761 1.314 a Dependent Variable: C0MP2 t s J C D ■ D O Q . C g Q . ■ D CD C/) o " 3 O Residuals Statistics (a) 8 C Q ' 3 3 " CD CD ■ D O Q . C a O Minimum Maximum Mean Std. Deviation N Predicted 'V alue .65 2.16 1.39 383 33 Std. Predicted Value -1.943 2.002 .000 1.000 33 Standard E rror of Predicted Value .166 .338 .231 .041 33 Adjusted Predicted Value -.12 2.28 1.33 .510 33 Residual -.63 .64 .00 .315 33 Std. Residual -1.613 1.639 .000 .810 33 Stud. Residual -1.909 2.257 .050 1.050 33 Deleted Residual -.88 1.77 .06 .573 33 Stud. Deleted Residual -2.049 2.531 .059 1.097 33 Mahal. Distance 4.837 23.151 10.667 4.282 33 Cook's Distance .000 1.300 .088 .235 33 Centered Leverage Value .151 .723 .333 .134 33 ■ D O a Dependent Variable: C0MP2 CD Q . ■ D CD C /) C /) to î £ S I I O C L c 0 1 I ■ C a i C L O 0 g "C :> 1 ( D CL Û O # □ o DO O O o a D a £ L I ZdlAlOO 213 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. *o a I I I I 0 DL C ■ § 1 g t l (0 Q _ U J 2d I /M O O 214 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. APPENDIX 4 Map of California (Route 99 Corridor Highlighted) 1 t - 215 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. APPENDIX 5 TY1999 Compton (Example) Dendrogram Using Average Linkage (B^ween Groups) CASE Label M u r a C ase 56 56 C ase 70 70 C ase 54 54 C ase 86 86 C ase 5 5 5 5 C ase 8 7 8 7 C ase 2 2 C ase 13 13 C ase 2 1 21 C ase 1 4 1 4 C ase 26 26 C ase 41 4 1 C ase 6 2 6 2 C ase 6 3 63 C ase 58 58 C ase 46 46 C ase 4 7 4 7 C ase 34 34 C ase 35 35 C ase 36 36 C ase 22 22 C ase 49 49 C ase 61 61 C ase 4 5 4 5 C ase 51 5 1 C ase 57 5 7 C ase 60 60 C ase 50 5 0 C ase 5 9 5 9 Case 68 68 C ase 6 7 67 C ase 4 8 4 8 C ase 101 101 C ase 136 136 C ase 122 122 C ase 128 128 C ase 1 1 C ase 130 130 0 + - 10 1 5 20 25 — + J 1 T T 216 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C ase 7 1 7 1 C ase 73 73 C ase 76 76 C ase 112 112 C ase 2 4 2 4 C ase 42 4 2 C ase 43 43 C ase 77 77 C ase 1 8 1 8 C ase 1 9 1 9 C ase 38 38 C ase 79 79 C ase 9 5 9 5 C ase 121 121 C ase 37 37 C ase 39 39 C ase 5 3 53 C ase 66 66 C ase 6 4 64 C ase 78 78 C ase 6 5 6 5 C ase 72 72 C ase 74 74 C ase 4 4 4 4 C ase 111 111 C ase 3 3 C ase 113 113 C ase 117 117 C ase 7 5 7 5 C ase 116 116 C ase 93 93 C ase 9 4 9 4 C ase 115 115 C ase 120 120 C ase 88 88 C ase 8 8 C ase 23 23 C ase 10 10 C ase 4 0 40 C ase 7 7 C ase 9 9 C ase 11 11 C ase 119 119 C ase 6 6 C ase 28 28 J ZJ 31 3 31 217 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C ase 4 Case 127 C ase 82 C ase 99 C ase 108 Case 110 C ase 29 C ase 30 C ase 3 1 Case 16 Case 15 C ase 32 C ase 3 3 C ase 104 C ase 20 C ase 8 0 C ase 9 1 Case 102 C ase 83 C ase 135 C ase 109 C ase 90 C ase 97 C ase 9 8 C ase 100 C ase 103 C ase 96 C ase 85 C ase 25 C ase 126 C ase 5 C ase 2 7 C ase 124 C ase 123 C ase 125 C ase 8 9 C ase 105 C ase 114 C ase 12 C ase 1 7 C ase 8 4 C ase 92 C ase 8 1 C ase 106 C ase 5 2 4 127 8 2 99 108 110 29 3 0 3 1 16 1 5 32 3 3 104 20 8 0 9 1 102 8 3 1 35 1 09 90 9 7 9 8 100 103 96 85 25 126 5 2 7 124 123 125 8 9 105 114 12 1 7 8 4 92 81 106 5 2 T \ ZJ J J J 218 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C ase 6 9 C ase 118 C ase 129 C ase 131 C ase 132 C ase 133 Case 134 case 107 C ase 137 6 9 118 129 131 132 133 134 107 137 219 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Internal venturing in public agencies

PDF

Evaluation of professional services consultants in rural government

PDF

A financial audit model for entrepreneurial governments

PDF

A configuration study of multiagency partnerships as practiced in Taipei City government

PDF

Assessing United States information assurance *policy response to computer -based threats to national security

PDF

Educational reform for private state -approved schools in California

PDF

Charter schools' influence on public school administrators' innovative behaviors

PDF

Determining acceptable seismic risk: A community participation-based approach

PDF

Conversion of health care organizations from non -profit to for -profit status

PDF

Drug treatment providers' organizational responses to implementation of California's Proposition 36

PDF

Care management for the uninsured: A force field analysis of the business case

PDF

Determining risk propensity of government program managers for high risk /high payoff projects

PDF

Historical perspectives and future horizons of local government managers and the International City /County Management Association

PDF

Adapting and applying a mission-focused strategic framework for emergency management

PDF

Do uses of human resource information technology (HRInT) tools in federal organizations improve their human resource management productivity?

PDF

Executive spending power: Flexibility in obligation and outlay timing as a measure of federal budgetary and policy control

PDF

Development of a family risk-factor measure that predicts imminent risk of placement and appropriateness for family-based, wrap -around services

PDF

Banking on the commons: An institutional analysis of groundwater banking programs in California's Central Valley

PDF

Bridled boldness: Reengineering the Federal Bureau of Prisons. Lessons learned

PDF

Getting 'how' and 'why' straight: A critical discourse analysis of the National Partnership for Reinventing Government's ideological discourse on information and communication technologies

Asset Metadata

Creator Haywood, Paul G. (author)

Core Title Detecting the effects social and business pressures on small California trucking firm tax compliance

School School of Policy, Planning and Development

Degree Doctor of Public Administration

Degree Program Public Administration

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag business administration, general,OAI-PMH Harvest,political science, public administration

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Starr, Kevin O. (committee chair), [illegible] (committee member), Clayton, Ross (committee member), Newland, Chester (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c16-390158

Unique identifier UC11340980

Identifier 3180773.pdf (filename),usctheses-c16-390158 (legacy record id)

Legacy Identifier 3180773.pdf

Dmrecord 390158

Document Type Dissertation

Rights Haywood, Paul G.

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA