Identifying and Quantifying Rates of State Motor Fuel Tax Evasion (2008)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

63 5.1 Introduction Chapters 1 through 4 and Appendix D, the Annotated Bib- liography, describe fuel tax evasion techniques and issues as well as strategies and methods used in the past to estimate fuel tax evasion. Chapter 5 puts this information together and highlights approaches that could be used to estimate evasion for each state. Evasion estimation approaches are, in turn, evaluated based on their ability to estimate tax losses associ- ated with specific evasion techniques, which are assessed later in this chapter and in Chapter 2. This chapter includes five sections, the first being this in- troduction. The second section presents approaches for estimating tax losses associated with EOE. Section 5.3 pres- ents a decision tree to assist states in conducting analysis of EOE. Section 5.4 examines several evasion methods, presents EOE estimation approaches, and models and re- views the data needed to support the proposed estimation approaches. Section 5.5 provides detailed data collection recommendations. It is important to note that in some cases this chapter repeats information about evasion techniques, including figures, con- tained in Chapter 2. The decision to repeat information was made because this chapter was designed to be a stand-alone EOE estimation pamphlet, which can be used on its own to document evasion techniques and present approaches for estimating EOE. 5.2 EOE Estimation Approaches The approach described in this chapter provides a method- ology to estimate the EOE level for each type or groups of types of evasion described in Chapter 2. The methodology provides a strategy that allows the sum of the individual types of EOE to equal the amount of total EOE, as demonstrated in the following equation: where E = Estimated EOE in State i; EM1 . . . EMn = Estimated EOE for technique 1 through n; 1 . . . n = Evasion techniques (use of dyed fuel on- road, tampering with fuel dye equipment, illegal removal of dye from exempt fuel, abuse of the IFTA return process, false re- funds or credits, import-export schemes across state lines, illegal importation of fuel from foreign refineries, abuses due to the presence of Native American reserva- tions, false product labeling, cocktailing, failure to remit tax payments, and daisy chains) The strategy implies that no one approach can be used to accurately estimate overall motor fuel tax EOE in a state. The level and quality of compliance and enforcement differs by state and the approach to calculating EOE will differ. For states with significant enforcement and compliance efforts and good databases, the approach can provide a much more accurate answer to EOE, while states with less enforcement and com- pliance activities will have less accurate answers to EOE for their state. One issue that will arise for external analysis of audit information for taxpayers is privacy concerns under the law. For example, other than IFTA data, most states will not consider releasing audit information for analysis, even if the identities of the taxpayers were removed. In this case only an internal analysis can be undertaken unless confidentiality rules can be extended to the outside evaluator. In addition, if data limitations prevent estimates of evasion by type, re- gression analysis or statistical sampling could provide an overall estimate of the amount of EOE occurring rather than by evasion technique. E = nEM EM EM EM1 2 3+ + + +. . . C H A P T E R 5 Methodology for Identifying and Quantifying State-Level Fuel Tax Evasion and Required Data

5.2.1 Tiered Approach to EOE Estimation The approach focuses on measuring the tax dollars lost to EOE, or the amount under-reported. In all cases, estimates represent the amount of under-reporting that occurs, whether intentional or unintentional. The estimate will contain the amount of tax dollars intentionally or fraudulently evaded as well as errors and omissions. Previous chapters have highlighted state-by-state variation in the data quality and quantity available, considering some of the following factors: • Varied motor fuel tracking systems, • Differing data on aspects of audits and inspections (i.e., some states have considerable data and some do not), • Differing characteristics that lead to evasion (i.e., some states have Native American reservations and/or on-road diesel programs while others do not), • Level of fuel tax compliance and enforcement in a state, and • Varied requirements regarding access to existing (but re- stricted) data. For these reasons we eliminated as possibilities the more rigid Simultaneous Equation Approach (SEA). The SEA ap- proach requires that a consistent set of state explanatory vari- ables be developed (e.g., proximity to high–low tax states, in- ternational borders, amount of enforcement applied, state accounting rules, and what level the state collects fuel taxes). In addition to state explanatory variables, the SEA approach relies on estimates of demand from VMT data supplied to the FHWA by the states. Based on discussions with the FHWA, data did not appear to be estimated accurately enough to be used to estimate EOE for individual states (Oregon DOT, 2000). It appears that the reliability of the VMT data is at best plus or minus 5 percent nationally and could be much worse for individual states. Furthermore, the SEA approach depends on independent, but reliable, estimates of average fuel effi- ciency for vehicles. Current estimates of fuel efficiency pub- lished by FHWA are largely based on VMT and taxes collected, which negates the ability to estimate motor fuel evasion. The first approach proposed in this chapter uses audit and inspections data to develop estimates of the amount of EOE occurring. There are various approaches outlined in this re- port that could be used to examine audit and inspections data, including statistical sampling, and regression tech- niques such as OLS, Tobit, and logit analysis. The second ap- proach uses a tracking system to follow fuel from terminals to taxpayer and calculate the difference of fuel supplied to taxes paid. The third, and final, approach is recommended for es- timating evasion losses due to the presence of Native Ameri- can retail outlets. The approach recommended compares the amount of gallons in question and calculates a percent of the total fuel consumption for that state associated with the non- payment of taxes. If more variables and data are available, more sophisticated techniques can be used to calculate amount of taxes forgone. 5.2.2 Approach #1: Audit and Inspections Analysis The audit and inspections review method produces esti- mates of evasion by examining audits and inspection records of motor fuel excise taxpayers. These evasion estimates are ag- gregated assessments based on the percentage and degree of fraudulent activity found through random audits. In two stud- ies reviewed (see Chapter 3), audit and inspections reviews were combined with other methods to gauge fuel tax evasion and only were used to supplement other forms of evasion analysis. Information collected from the audit reviews and in- spections could be applied to improve the understanding of motor fuel excise tax evasion. At the very least, audit analysis could be used to gain a grasp of the potential monetary range that individual schemes cost in terms of EOE. Tax audits result in an estimate of tax liability. This estimate is typically equal to or greater than the tax liability reported by the taxpayer. The taxpayer’s reported liability may be coun- tered by an assessment from the auditor. The assessment is typically based on incomplete information, and the audited party can provide information to reduce the assessment. The ultimate goal of the audit is to determine true tax liability. While perfection is not possible in reality, for the purposes of this study, the amount arrived at as the true tax liability will be treated as the best estimate of this amount. An audit reveals the difference between reported tax liability and true tax lia- bility or the amount of EOE. Actual tax collections may be less than the additional amount identified at the end of the audit, but it is the tax liability rather than the tax collected that we are using with underpayments to determine EOE. Audit programs that include random and targeted elements are considered best in targeting evasion and maximizing as- sessments per auditing dollar spent. Random audits could be used to establish an unbiased sample to support EOE estima- tion and to uncover forms of evasion not well understood by state auditors. Targeted audits, on the other hand, maximize returns on investment by using a screening process to identify high-risk taxpayers whose tax returns under audit are most likely to yield assessments. 5.2.2.1 Approach #1-A: Simple Average Approach The most simplistic approach to using audit and inspec- tions data is to extrapolate the average or weighted average EOE rate uncovered through audits and inspections to the 64

overall taxpayer population. Under this approach, if 1 per- cent of the heavy trucks inspected on-road were found to be misusing dyed diesel, it would be assumed that 1 percent of all heavy trucks were burning dyed diesel on-road. This approach, however, is fraught with complications and potential problems. For example, audit data often include a small sub-set of the overall taxpayer population. Expanding the sample uncritically to the universe of taxpayers, whose characteristics may not match the sample, could lead to sig- nificant under- or over-estimation. Second, there is often sig- nificant bias in the sample because auditors often use screen- ing methods or triggers (e.g., tips, questionable tax returns) to target the taxpayers whose returns are most likely to gen- erate an assessment. Finally, targeted inspections and audits not part of a more comprehensive enforcement program could fail to capture evasion occurring outside of the legiti- mate fuel supply system through techniques such as daisy chains and import/export fraud. Problems associated with the simple average approach limit its usefulness. It is recommended to use this approach only when there appears to be no bias in the process for se- lecting companies or vehicles for audit or inspection and the other approaches outlined in this section are not feasible due to data limitations. 5.2.2.2 Approach #1-B: Statistical Sampling Taxpayer audits may be treated as a random sample of all taxpayers, although this is not likely to be true. If the sample were truly random, the difference between true tax liability and reported tax liability found in the sample could be ap- plied to the population as a whole to get an unbiased estimate of the total amount of EOE. Evasion estimates could be cre- ated by taking the percentage of all audits/inspections in which illegitimate activities occurred generalizing to an EOE assessment over the reported tax liability for the sample pop- ulation and applying this factor to reported tax liability for the entire population of taxpayers. As noted, audits are sel- dom done on a truly random basis. Hence, some analysis would be required of the basis for selection, and the resulting information used to correct for the bias associated with ap- plying information from a non-random sample. When using this method, analysts must control for bias associated with se- lecting the sample of businesses for audits. For example, the basic equation would be as follows: where E = $ fuel evasion or EOE; X = Violations (in gallons of fuel); E = X n × ×N t n = Sample size—total inspections or audits (in gallons of fuel); N = Population size—target population susceptible to spec- ified evasion scheme (in gallons of fuel); and t = Fuel tax rate in state. The characteristics of a very small sample of taxpayers could not be applied universally to all taxpayers; efforts must be taken to understand the sample characteristics and how well the sample represented the overall population. Second, audits and inspections are not always randomly selected; sometimes they are the product of a tip, suspicious returns, or previous issues with high-risk taxpayers. Sample results cannot be ag- gregated with reasonable confidence unless they are either random or steps are taken to control for bias. For example, if audits are conducted on 5 percent of the taxpayers, but these taxpayers are selected for audit because the auditors have found that a particular group within that population has twice the EOE liability of the population, then adjustments for that bias must be made. Applying the ratio from this sample to the entire population would overstate the actual amount of EOE expected. In this simple case, the correction is obvious; but in more realistic cases, it is likely to be complex. Although statistical sampling may not be the best approach for estimating total evasion in a state, it can be useful for esti- mating EOE for specific evasion techniques where the targeted population can be clearly defined or where more explicit data that could be used in statistical regression analysis are unavail- able. Due to the complexity of business operations, however, audits and inspections may fail to uncover all fraudulent activ- ity occurring within the investigated entities. Audit and in- spections data might fail to capture evasion occurring outside legitimate fuel supply systems through techniques such as daisy chains and fuel smuggling; however, through the examination of distributor field audit data combined with analysis of re- tail-level data, this can be an appropriate estimation approach for these types of schemes. 5.2.2.3 Approach #1-C: OLS Analysis Economists often deal with data that does not represent a random sample. The typical method to correct for the non- random nature of the data is to control for various character- istics using regression analysis. In regression analysis, mea- sures of important characteristics likely to affect the outcome variable are included in the analysis. The regression is said to “control for” the influence of these characteristics to deter- mine the true effect of the items of interest. For example, if businesses that had started within the past year are generally found to have more EOE than established businesses, auditors are more likely to audit such businesses. By including the amount of time in business in explaining EOE, the analyst can 65

control for the effect of time in business on the amount of EOE detected. Applying this information to the population at large would control for the difference in time in business be- tween the audit sample and the rest of the population. If this were the only difference between the audit sample and the rest of the population, the resulting estimate of EOE for the pop- ulation as a whole would be an unbiased estimate. An unbi- ased estimate is one that would be equal to the actual EOE if the experiment were conducted many times. If the analyst did not control for time in business, then applying the EOE from the audit sample to the entire population would overstate the amount of EOE since the rest of the population has more time in the business and, hence, is less likely to have more EOE than the audit sample. Many important characteristics (e.g., trigger for the audit, experience of the auditor) can be entered in this way to improve the ability to generalize the estimates from the sample group to the entire population. 5.2.2.4 Approach #1-D: Tobit Analysis The simplest type of regression analysis (ordinary least squares or OLS) may not work well on some audit data. If many of the observations are cut off (censored) at a particular level (e.g., no evasion), the standard regression model does not work well and a different modeling approach must be taken. Audit data are often clustered at a zero evasion level; this is known as censored data since no negative results are reported. If audit results are censored, then a model approach called the Tobit procedure can be used to determine a more appropriate estimate of how various characteristics affect EOE. The Tobit model allows adjustment of the estimated slope when data are censored. In Tobit analysis, a single maximum likelihood es- timate of the slope coefficients is generated that corrects for the bias associated with use of censored data. Feinstein (2003) suggests extensions to the simple Tobit model described above. In the extension, a second step is undertaken if data are available on what auditor performed the assessment. Further analysis may capture how much more eva- sion could be detected based upon what the best auditor could find. New variables of probability are calculated for complete detection, fractional detection, and no detection. Feinstein fur- ther suggested a third step. In the third step, the probabilities to evade and not evade are combined with the probability of detection (if they exist) based on the characteristics of the population. Using assumptions about the distribution of the populations, an estimate of the propensity to evade can be developed for the type(s) of fuel evasion being analyzed. An additional caveat with the Tobit procedure needs to be addressed. The Tobit procedure assumes that the same vari- ables affect the level of evasion and whether evasion is found or not. Sometimes there is information that this assumption is not correct. When there are variables that affect whether evasion occurs or not and they are different than the variables that affect the amount of evasion, an alternative procedure may be used. One approach to correct for the shortcoming in the Tobit procedure is the Sample Selection procedure developed by Heckman (Crown, 1998). First a probit model is estimated based on whether evasion is detected or not. These estimated probabilities are used to generate what is known as an inverse mills ratio. The resulting values are then included as a vari- able in an OLS regression to determine the amount of evasion detected. For example, when vehicles are inspected for dyed fuel, audit information on violators is often much more com- plete than for nonviolators. Nonviolator information only may include whether it is an in-state vehicle or out-of-state vehicle, type of vehicle license, and region where the violation occurred; while information on the violators could include primary industry, how many gallons, the type of vehicle, etc. The sample selection procedure allows for estimates of the affect of these latter variables on EOE that can then be applied to all vehicles. 5.2.2.5 Approach #1-E: Logit Analysis When appropriate data are not available on individual cases or information is lacking on the amount of fuel being misused or taxes are not paid, a simplified approach called logit analy- sis may be used to determine the percentage of fuel usage not in compliance. Logit analysis determines the probability for EOE based upon the characteristics of those complying and those not complying. More importantly, logit analysis can be based upon the use of grouped data. Grouped data means entities with a similar set of characteristics can be grouped together in the analysis. For example, when vehicles are inspected for dyed fuel, informa- tion on violators is often much more complete than for non- violators. Nonviolator information only may include whether it is an in-state vehicle or out-of-state vehicle, type of vehicle license, and region where the violation occurred, while infor- mation on the violators could include primary industry, how many gallons, the type of vehicle, etc. Thus, logit analysis can use reduced information to calculate probabilities of violations based on the limited characteristics available in the data. 5.2.3 Approach #2: Fuel Tracking Approach In states where fuel tracking systems exist, the tracking sys- tem can be used for examining fuel evasion for some evasion techniques. Such a system then could be used to correlate state collections to the fuel usage values for a state. A first step is to examine existing state fuel distribution systems and as- sess their viability in terms of estimating evasion. Note, how- ever, that the small number of fuel tracking systems available 66

at the state level and the lack of uniformity, with respect to such models, constrains their application as a viable method- ology for many states. That is, the overall methodology and model for each state must be designed to allow states to ig- nore or take advantage of fuel tracking system data, depend- ing on the state-specific circumstances. 5.2.4 Approach #3: Statistical Analysis of Sales The primary approach recommended for estimating eva- sion losses due to the presence of Native American retail out- lets is a hybrid approach that compares consumption with taxed gallons and motor fuel tracking approaches. This ap- proach can be used when consumption data and tracking sys- tems are in place. The alternative approach, however, in- volves the development of a statistical model for estimating motor fuel sales on Native American reservations. In the model, the number of gallons of motor fuel sold by Native American establishments serves as the dependent variable, and a number of explanatory variables are used to estimate the sales volumes. Independent variables could include: • The number of Native American reservations in a state; • The number of retail motor fuel outlets on reservations; • The number of pumps located at each Native American retail outlet; • State motor fuel excise tax rates; • State populations; and • Proximity to high-tax states. Data provided by states with agreements with Native Americans could serve as the base data required to test the predictive capacity of the model, or data tied to other retail establishments in the state could serve as the underlying data for the model. Estimates of motor fuel sales, in turn, would be compared to estimated gallons consumed by Native Amer- ican populations based on the number of enrolled members and estimated per-capita consumption of motor fuel. The difference between the modeled estimates of motor fuel sales and estimated Native American motor fuel consumption would represent EOE. 5.2.5 Remaining Issues and Caveats The biggest issue facing anyone estimating the level of eva- sion will be the quality of the data the state possesses. Those states with good data from audits, inspections, and tracking systems may provide good bases for estimates of evasion. Conversely, because of their good data (presumably used for enforcement and compliance) those states also may experi- ence less evasion. The methods chosen are believed to be the most suitable approaches considering the likely data quality and availability. Data availability and quality will have an im- pact on the method that can be used. • The lower the quality of information retained associated with audits and inspections, the greater the likelihood that the approach would not be able to remove bias from estimates. • For each state, careful attention is needed to ensure that the individual estimation approach to evasion does not overlap and that all attempts have been made to remove double counting. The quality of data that a state possesses may require a method to capture and understand the levels of double counting. • Limitations of each estimation method, including the audit review approach, were identified in Chapter 3; however, with appropriate use of statistical techniques to remove the bias, the approach is believed the most viable. The following sections provide a decision tree to help states work through estimating evasion and then provide methods for estimating EOE for each general category of evasion. In gen- eral, EOE categories are grouped together by similar charac- teristics of the evasion technique. The techniques analyzed for methodological approaches include false refunds or credits, use of dyed fuel on-road, tampering with fuel dye equipment, illegal removal of dye from exempt fuel, abuse of the IFTA re- turn process, import–export schemes, illegal importation of fuel from foreign refineries, abuses due to the presence of na- tive American reservations, false product labeling, cocktailing, failure to remit tax payments, and daisy chains. 5.3 Decision Tree for Individual State Analysis This section illustrates the decision tree recommended to follow to get an estimate of the amount of fuel tax EOE for a state. Key questions and their answers will lead to a total estimate for each state. 5.3.1 Key Questions 1. Does state have an electronic fuel tracking system that tracks fuel at each stage in the distribution process from the terminal rack to the retail level? a) Is the difference in taxes paid versus taxes owed tracked by fuel type, e.g., gasoline, diesel, etc.? If the answer to these questions is yes to both, then calcu- late the amount of fuel tax owed based on initial tracking of taxes paid versus taxes owed by fuel type. In other words, cal- culate the percent EOE based on the amount of tax found owed on the initial assessment, not after the correction has 67

been made by the taxpayer. The tracking system approach, however, may not capture some forms of import/export fraud. Thus, retail audit data may be required to augment the fuel tracking approach. If the answer to question 1a is no, then calculate the percentage by all fuel types and apply to total fuel taxed. If the state has no fuel tracking system or the fuel tracking system does not achieve total fuel accountabil- ity, move on to Question 2. 2. Does state have a method to track audits by type of evasion found? a) What types of evasion do the audits detect? b) Is the difference tracked by fuel type (e.g., gasoline, diesel)? If the answers to Questions 2 and 2a are both yes, then use Sections 5.4.1 to 5.4.9 to determine the appropriate approach to calculate the percent EOE for each evasion technique. Go to Question 3 to determine which calculation method for EOE is appropriate. If the answer to Question 2b is yes (the audit reveals the EOE found by fuel type) then calculate the amount owed by each fuel type. If the answer to Question 2 is no, then use one of the techniques described in Section 5.2 to calculate EOE for all types of evasion found through the audit process and apply it to all taxed fuel by determining the answers to Questions 3 through 9. 3. How is the audit investigation started? a) Is the investigation started randomly (e.g., no processes are used to determine what entities should be investi- gated)? b) Was some tip or other method used to develop the list of taxpayers audited? c) Is the selection method unknown? If the answer to Question 3a is that the entities to be audited were selected randomly from the population, then the weighted percent EOE from the sample population can be attributed to the whole population using the statistical sampling method. If the answer to Question 3b is that some tip was used to determine who is audited, then determine what proportion of the population the sample population represents. The percent of the total population represented by the audits can be used to calculate the EOE based on those audited. It should not be extrapolated to the whole population. If there is a mix of tips and random selection and the au- dits that are randomly selected can be determined, or the answer to 3c is that the method is unknown, then those au- dits that are randomly selected or for which the selection method is unknown could be analyzed based on the answers to Question 4. 4. How robust are the characteristics of the taxpayer in the sample data? a) Does the audit data contain enough information to characterize the entity audited? If the answer to 4a is yes, then choose the statistical approach that reflects the best approach in 5.2 Approaches 1-C through 1-E by answering Question 5. If the answer to question 4a is no, there is not enough information and the best answer that can be stated is the percent EOE for that type of evasion. However, it should be noted that the results aren’t statistically valid but the EOE estimate is the best available answer. 5. Does the sample reflect the overall population of motor fuel taxpayers? If the answer to Question 5 is yes then the weighted aver- age percent EOE of the population can be applied to amount of fuel reflected by the population. If the answer to Question 5 is no, then go to question 6. 6. Does the sample data appear to be censored? If the answer to Question 6 is that data are censored, then the approach chosen should be a Tobit analysis following Section 5.1 Approach 1-D. Reread Section 5.1 Approach 1-D if unsure whether the data are censored. If the answer is no, then go to Question 7. 7. Does data have a differential in the amount of information available for violators as opposed to nonviolators? Typi- cally dyed fuel investigations have little information col- lected about nonviolators whereas violators have signifi- cantly more information about them collected. If the answer is yes, then a Logit analysis is appropriate. Follow the approach in Section 5.2 Approach 1-E. The ap- proach is usually appropriate with dyed fuel investigations if the data are collected. Calculate the probability of evasion based upon the available characteristics of the sample data and apply to the population. If the answer is no, then use OLS analysis following Section 5.2 Approach 1-C to obtain a set of coefficients that can be applied to the population. 8. Have all types of evasion been accounted for by the data? If the answer to this question is yes, then go to Question 10. If the answer is no and for some states this should be true, es- pecially those with Native American issues, follow the approach discussed in Section 5.4.8. 68

9. Has EOE been estimated for all the evasion techniques? If the answer is yes, then go to Question 10. If there are types of evasion still to be estimated for which there is no data, then data could be collected based on availability of funding for such an investigation. Such samples should be stratified and randomly selected to reflect the population. Once the in- vestigation is complete, then the resulting estimate of EOE could be applied to the population from which the sample was selected. 10. Calculate total EOE. Sum together the EOE for each type of evasion technique or group of techniques found in answering Questions 1 through 9. Make sure that each is applied appropriately to the fuel type that the evasion technique covered. For example, IFTA evasion probably mostly reflects diesel fuel. Dyed fuel investigations reflect improper use of dyed fuel as transporta- tion fuel and would be counted as diesel. It is possible if a state is a low-tax state to have a negative EOE. 5.4 Evasion Techniques This section presents approaches for estimating EOE asso- ciated with nine evasion methods: • False refund or credit, • Evasion of untaxed dyed fuel, • Abuse of IFTA return process, • Evasion associated with exporting/importing fuel across state lines, • Illegal importation across international borders, • Failure to remit, • Cocktailing and false labeling, • Abuses due to the presence of Native American reservations, and • Daisy chains. This section presents an overview of, and prescribes ap- proaches for, estimating EOE associated with each evasion technique. It also directs the reader to the data required to support each EOE estimation approach as specified in Sec- tion 5.5 and identifies data limitations. 5.4.1 False Refund or Credit 5.4.1.1 Overview of Evasion Technique The extent to which fuel tax evasion through refund and credit schemes is a significant compliance issue depends largely on the point of taxation and how elaborate the exemptions are within a jurisdiction. For example, tax systems with a point of taxation high in the distribution chain (e.g., the terminal rack) tend to generate higher rates of refund and credit filings as tax- payers recover payments made on taxed fuel used for nontax- able purposes. Also, the more exemptions a jurisdiction allows, the more refund claimants it is likely to have. It also is possible under certain circumstances for a whole- saler to apply for a refund or credit as well. Under a tax-at-the- rack system, a wholesaler can claim to export or sell for export previously taxed fuel. The fuel, in turn, could be sold within the jurisdiction with the wholesaler keeping the refund as profit. Further, motor fuel taxpayers may over-report refunds or credits associated with nontaxable uses (e.g., agriculture, construction, commercial off-road, marine vehicles, home or business heating, etc). It is difficult, if not impossible, to directly measure true motor fuel consumption for tax-exempt uses because travel is not monitored off-road and states often lack the resources necessary to properly enforce motor fuel excise tax programs. Refund schemes have been addressed historically through au- ditor analysis of refund claims, dyed fuel requirements, and stiffer penalties for refund fraud. 5.4.1.2 Model Approach There are two possible approaches: the first approach esti- mates EOE by developing estimates of off-road fuel use based on information collected on tax return forms about off-road vehicles and average fuel use and compares the refund with the amount of estimated off-road use to determine potential evasion. The second approach uses audit data to estimate the percent of total gallons refunded obtained through EOE and apply that to overall refunds using statistical analysis. When states maintain complete records of reported refunds and credits, evasion could be computed by estimating nonhigh- way consumption of motor fuels and comparing to reported refunds and credits. Many states presently analyze company operations to compare refund claims with expected motor fuel consumption, based on some rational measures such as the number of vehicles registered to the company using motor fuel for tax-exempt purposes or the number of acres managed by farming operations. The approach for estimating EOE by comparing actual consumption with refunds requires construction of a spread- sheet-based model at the state level where data are collected on the number of exempted vehicles and the average fuel con- sumption per vehicle for each type of operation receiving an exemption. The approach entails analyzing data relating to the specific operation in question to determine a range of ac- ceptable values in terms of total consumption per vehicle and using this value, combined with registration information, to determine an acceptable range of refunds for the industry 69

under examination. Estimated aggregate consumption could then be compared to data collected from motor fuel tax refund claims to estimate under-collection due to refund fraud. This is an accounting approach that simply compares estimated and reported consumption. The alternative approach is to use a statistical analysis on the refund audits undertaken by the state. As a last approach, the amount of EOE found in the audits could be used to di- rectly calculate how much EOE was found. 5.4.1.3 Data Needs and Limitations The states collect data on refund claims and credits from various sources, including motor distributor schedules (which usually report tax-exempt fuel sales in a separate special work- sheet or column) and the nonhighway fuel consumers’ tax forms. States also may house industry-level data on the num- ber of vehicles registered and operated within the state, miles reported, and motor fuel economy. Data required to support both modeling approaches detailed in the previous section are outlined in Section 5.5. Data requirements are specified for each approach in the refund and tax-exempt fuel data element subsection. Due to the significant data requirements required to im- plement the proposed model approach, tax administrators and modelers must be careful when cross-referencing data and must examine discrepancies and data issues related to de- ferments of tax payments or refunds between periods or re- porting lags, misrepresentation of nontaxable consumption, and the absence of available data related to taxable versus non- taxable consumption. 5.4.2 Evasion of Dyed Fuel Laws 5.4.2.1 Overview of Evasion Technique Thirty-eight states require fuels for particular tax-exempt uses to be dyed at the rack. There are several forms of fuel tax evasion that can occur involving diesel that is untaxed and either dyed or intended to be dyed, but is instead used for tax- able purposes. The three techniques used to evade dyed fuel laws include the following methods: misuse of dyed fuel, dye masking, and/or dye removal. The next section presents an overview of the three techniques. 5.4.2.2 Evasion Techniques 5.4.2.2.1 Misuse of Dyed Fuel. Business operations using off-road vehicles (e.g., farming, logging, construction) pur- chase tax-free dyed diesel. Dyed diesel is often delivered to these businesses and transferred into private storage tanks for use in their off-road vehicles. To evade fuel taxes, on- road vehicles owned by the business or individuals associ- ated with the business use dyed diesel fuel from these tanks. Another common way perpetrators fuel their on-road vehi- cles with dyed diesel is by using card-lock systems at retail stations which allow registered customers to access tax-free fuel by swiping a card. These stations are generally unmanned and a person with an access card can fill their on-road vehi- cle’s fuel tank or fill a container for later use in their highway vehicle. 5.4.2.2.2 Dye Masking (Blending and/or Falsely Label- ing Dyed Fuel). Dyed diesel can be acquired and blended with darker oils (such as waste oils) or green dye to mask the apparent color. This fuel is then sold at the retail level, possi- bly to unsuspecting consumers, under the false presumption that it is taxed fuel. This can be achieved by blending fuels in a warehouse or simply by adding dyes in the fuel tanker. 5.4.2.2.3 Dye Removal. Many terminals with dye injec- tion equipment have card systems in place so that registered drivers can load fuel without assistance, and these terminals are sometimes left unattended at certain times of the week. This situation enables a tax evasion perpetrator to pull up to a loading rack, order a load of dyed diesel, tamper with the fuel dye injection equipment, and then leave with undyed and untaxed fuel that eventually can be sold as taxed fuel at the re- tail level. In some cases, terminals do not have dye injection equipment, so the dye is poured directly into the tanker (splash blending). It is possible, in such cases, for a tax evasion per- petrator to purchase the fuel tax-free, but fail to splash dye it. It is also possible for dyed fuel to be purchased and then have the dye removed by bleaching it, adding sulfuric acid to it, or running it through a filtration and/or refining system. In all cases this fuel officially leaves the terminal rack as dyed un- taxed fuel, but most likely ends up sold on the retail market as undyed taxed fuel, and the perpetrator then pockets the amount of the tax. This form of fuel tax evasion would most likely impact states that are taxing fuel at the terminal rack level or distributor level. Each one of these evasion techniques can be measured separately if data are available (methodology described below). The following balance equation can be used to meas- ure the overall disappearance of dyed fuel from the legitimate market: where Fdyed = Reported dyed fuel that leaves the terminal rack for use in a given state; Fi = Untaxed dyed fuel that reportedly is used by sectors with access to dyed fuel; F Fi tdyed EM= ( )+ ( )∑ ∑ 70

EMt = Total fuel evaded by technique t; and t = Dyed fuel misuse, dyed fuel removal, blending/false labeling. The next two sections examine these evasion techniques in more detail. 5.4.2.3 Dyed Fuel Misuse 5.4.2.3.1 Overview of Evasion Technique. This scheme is thought to be common and generally occurs on a small scale by many separate individuals, particularly individuals who own or work for businesses operating off-road vehicle equipment. This type of evasion only could occur in states that have dyed fuel programs or at the federal level (see Chapter 2). In Fiscal Year 2002, IRS Fuel Compliance Officers assessed more than 900 penalties, totaling more than $1.8 million for misuse of dyed diesel fuels. More than 70 percent of the penalties involved the misuse of fuel by taxpayers in the construction and agriculture industries. Both of these indus- tries are subject to broad-based tax exemptions for non- highway use of motor fuels thereby presenting opportunities for abuse (see Chapter 2). The following addresses a key question that must be an- swered to model this type of evasion about who is deciding to evade and where (at what point in the distribution) the evasion occurs. Figure 5-1 shows where evasion leakage occurs (note “Bulk Plant” is actually part of the nonbulk distribution). Sectors with Access to Tax-Exempt Dyed Fuel: Case stud- ies appear to indicate that the vast majority of the evasion de- cisions about misuse of dyed fuel occur at the end-use/retail level. Businesses that currently use equipment for off-road or tax-exempt purposes also choose to use the fuel illegally for on-road purposes. One northwest television news group cre- ated a team of seven investigators to follow the misuse of dyed fuel in Washington State. This team uncovered an extensive enclave of truckers, loggers, construction crews, and fruit growers using dyed fuel on-road. This would suggest that closer scrutiny of the economic activity of these industries, as well as laws and enforcement methods related to dyed fuels used in these industries, is necessary to estimate the amount of taxes not paid using this evasion method. Motor Carrier Use: It is also possible that fuel taxes are evaded by illegally acquiring dyed fuel permits and/or ob- taining access to dyed fuel that is controlled through a card- lock system. This evaded fuel is sometimes channeled outside the tax-exempt sectors and is used to fuel large trucks. This form of evasion is detected through highway fuel inspections and also can be caught during IFTA audit analysis. 5.4.2.3.2 Model Approach. Three methodological ap- proaches are suggested to estimate the EOE of this evasion technique, depending on data availability: 1. OLS or Tobit; 2. Logit; and 3. Statistical Sampling. 71 $ $ $ $ $ $ Tax-Exempt Uses $ Card-lock Figure 5-1. Nonbulk distribution system.

OLS or Tobit. Provided that sufficient data are available for individual filers through the audit and inspections process, the OLS or Tobit approach could be adopted. Explanatory variables would be developed for each sector evaluated. For example, explanatory variables could include: • The difference between the state diesel tax rate and na- tional average; • Gallons of fuel associated with audit or inspection; • Primary industry area; • County (dummy variable); • Examiner or inspector; • Location of fuel source; • Number of inspections/audits by sector; and • Detection rate of inspector/auditor. Logit. Logit analysis can be undertaken if the amount of gallons violated is not available for both the violators and nonviolators but the information on their characteristics is available. The characteristics would be the same for OLS or Tobit analysis. Statistical Sampling. If data that obscures the taxpayer, but leaves remaining information intact cannot be obtained due to a state law or confidentiality issues, the statistical sampling ap- proach can be used. This approach uses a probability sampling method from inspections and audit data. These data are po- tentially available from some states and the IRS. Data must be evaluated to remove bias (e.g., inspections based on tips) so that the sample is random. If the bias cannot be removed, methods to estimate the amount of bias need to be explored. The number of violations (in terms of total fuel evaded) as a percentage of total fuel covered through inspections and audits will be the basic statistic that is used to measure evasion. In each case this percentage will be expanded to the targeted popula- tion by multiplying the proportion of violations to inspections times the total amount of fuel use in the selected population. For example, the estimate would be as follows: where Ei = $ Evasion in Sector i; Xi = Violations attributed to Sector i (in gallons of fuel); Ni = Total inspections of Sector i (in gallons of fuel); Fi = Total amount of fuel used for Sector i purposes; Si = Proportion of Fi used by on-road vehicles; t = diesel tax rate in state; and i = agriculture, construction, logging, manufacturing, min- ing, motor carriers, retailers, etc. Ei will be measured for all sectors with access to tax-exempt dyed fuel and for motor carriers and retailers. E X N F S ti i i i i= × × × If the data cannot be broken into industry-specific cate- gories, then the percentage of violations is applied to the sum of the amount used for all the industries included. The answer would be much better, however, if the data would allow the breakdown by industry type, especially since retailer violation percentages would be applied to retail diesel use. An alternative to applying the direct percentage of those evading is to calculate the number of gallons evaded and divide into the total gallons of undyed fuel to get an EOE percent- age. This can only be done if information is available on ve- hicle tank sizes. This approach underestimates the amount of evasion because it is not applied to the whole population. 5.4.2.3.3 Data Needs and Limitations. Twenty-two states and the IRS have dyed fuel inspections data. The primary problem found in the analysis of two states data was that information on nonviolators is not kept. The best estimate that can be performed is the percentage of violators applied to the amount of fuel assuming equal usage. In one case we were able to estimate directly the amount of gallons evaded and that estimate was applied to the total number of gallons evaded. For states that regularly conduct on-road dyed fuel inspections, estimates of EOE for misusing dyed fuel could be based on inspection data, as specified in the dyed fuel data elements outlined in Section 5.5. By collecting detailed inspec- tion data for both violators and nonviolators, the statistical sampling approach can be refined to generate more reliable estimates of EOE. 5.4.2.4 Dye Removal and Dye Masking 5.4.2.4.1 Overview of Evasion Technique. Schemes that involve tampering with fuel dye injection equipment and the failure to splash dye appear to occur when terminals are un- attended by terminal employees. Some equipment has been altered in the past by simply using a wrench to close the dye injection valve. There have been a number of different cases where dye has been removed illegally from tax-exempt fuel. In some cases household chorine bleach or sulfuric acid has been added to dyed nontaxed fuel to eliminate the visible red color. Green dyes and dark oils (such as waste oil) can also be added to the red-dyed fuel to conceal the appearance of red. Dyed fuel also can be transported to a warehouse where the fuel then is run through a charcoal filtration system until no apparent red color is present. There also have been reported cases where dyed fuel is bought and transported to a leased or owned refinery and then refined to remove the dye (there are a number of refineries that legally carry out the process of extracting red dye from transmix fuel, which is the interface fuel between dyed and undyed diesel in a pipeline). 72

The following addresses a key question that must be an- swered to model this type of evasion about who is deciding to evade and where (at what point in the distribution) the evasion occurs. Figure 5-2 shows where evasion leakage occurs (note “Bulk Plant” is actually part of the nonbulk distribution). • Tampering and Failure to Splash Dye: Tampering with fuel dye equipment and failure to splash dye occur at the ter- minal rack and the perpetrators are motor fuel carriers. • Illegal Removal or Masking of Dye from Tax-Exempt Fuel: Dye can be removed or masked from tax-exempt dyed fuel at any point in the nonbulk distribution process. It is pos- sible for dyed fuel to be masked with other dyes or darker oils within the tanker truck just as it leaves the terminal rack. Other schemes may occur at refineries or warehouses further down the distribution process. This would not be an effective evasion scheme for states that tax at the retail level (e.g., Oregon, North Dakota, New Jersey), as the tax would have to paid even when dye is ille- gally removed. 5.4.2.4.2 Model Approach. There are three methodolog- ical approaches that can be used to estimate the EOE of this evasion technique, depending on data availability: 1. OLS or Tobit; 2. Statistical Sampling; and 3. Motor Fuel Tracking. OLS or Tobit. Provided that sufficient and appropriate data are available for individual filers through the audit and/or inspections process, the OLS or Tobit approach could be adopted. Explanatory variables could include the following: • Size of business (# of employees); • Income/revenue of business; • Type of transport; • Industry affiliation; • Number of motor carriers in fleet; • Number of fuel carrier drivers; • Estimated number of miles; • Estimated fuel use (number of gallons); • Longevity of business (e.g., the year the transport/hauling license was obtained); • Location of terminal from which fuel was obtained; • Examiner or inspector; • Number of inspections/audits; and • Detection rate of inspector/auditor. Statistical Sampling. If sufficient inspections and audit data exist regarding these types of evasion schemes, the methodological approach to estimating the amount of tax evasion that occurs by either withholding or removing dye from tax-exempt fuel uses a probability sampling method from inspections and audit data. Data need to be evaluated to remove bias (e.g., inspections based on tips), such that the sample is random. If the bias cannot be removed, methods to estimate the amount of bias need to be undertaken. The num- ber of violations (in terms of total fuel evaded) as a percent- age of total fuel covered through inspections/audits will be the basic statistic used to measure evasion. This percentage, expanded to the target population, equals the ratio of viola- tions to inspections multiplied the total amount of fuel use in the selected population as shown in the equation below. 73 $ $ Card-lock $ Masking or Dye Removal Retail Figure 5-2. Depiction of where evasion through masking dye occurs.

For example, the estimate would be as follows: where E = $ evasion-altered dyed fuel; X = violations (in gallons of fuel); N = total inspections/audits (in gallons of fuel); F = total amount of taxed motor diesel fuel sold in state; and t = diesel tax rate in state. If key characteristics are discovered about the perpetrators during the data gathering process, these characteristics may be used to adjust the total population. For example, if it is found that this type of evasion only occurs with terminals that have specific types of equipment, the results of the sampling would be expanded only for the total gallons associated with the terminals that possess that specific equipment. Tracking. The tracking approach for estimating motor fuel sales is based on the examination of distributor reports to track motor fuel sales. The amount listed on distributor reports would be compared with motor fuel sales to detect evasion. 5.4.2.4.3 Data Needs and Limitations. The problem found in some states is that they don’t track the type of EOE found in their audits. However, state audit data regarding dyed fuel violations that is kept could be expanded in the fu- ture to include the type of EOE that the audit was finding. Data required to estimate EOE through targeted dyed fuel inspections is outlined in the dyed fuel data element sub- sections in Section 5.5. The distributor audit data outlined in the third general audit data subsection of Section 5.5 also could be used to estimate EOE. 5.4.3 Abuse of IFTA Return Process 5.4.3.1 Overview of Evasion Technique IFTA represents a contract between jurisdictions that sim- plifies fuel tax remittance for multi-jurisdictional motor car- riers by allowing them to file with one base jurisdiction. The base jurisdiction collects and disperses fuel taxes to other ju- risdictions. IFTA is a not-for-profit organization that receives dues from each jurisdiction and serves as support staff to aid communication and organization between these jurisdic- tions. The presence of differentials between state tax rates generates incentives to evade motor fuel taxes by defrauding the IFTA system. Fraud occurs when motor carriers under- report miles in high-tax states or fail to file IFTA reports. IFTA audit analysis also captures failure to remit taxes by motor carriers who use untaxed fuel either by using dyed fuel or untaxed clear fuel. E X N F t= × × IFTA members include all states in the United States and all the provinces in Canada with the exception of Alaska, Hawaii, and the District of Columbia in the United States and the Northwest Territories (Nunavut and Yukon) in Canada. The intent of IFTA is to tax interstate motor carriers on the basis of the quantity of fuel used within the state rather than on the basis of fuel quantities sold within the state. States are required to audit 3 percent of the number of base-state IFTA accounts; however, several states audit less than 3 percent in any given year. Figure 5-3 shows that Alabama, California, Georgia, Kansas, Michigan, Nebraska, Oklahoma, and Rhode Island conducted less than 3 percent of the IFTA audits on average during the last 5 years (2000–2004). Alaska and Hawaii are not members of IFTA (IFTA, 2006). 5.4.3.2 Model Approach The approach used here follows the OLS or Tobit approach outlined above. If data are not censored then the process needs to follow the OLS approach. If data is found to be censored, e.g., evasion versus errors and omissions can not be differen- tiated then Tobit is the approach, otherwise OLS is required. For example, using the case of IFTA abuse, explanatory variables used in the regression model could be extracted di- rectly from the audit records. Examples of the explanatory variables that could be used to conduct the regression model procedure include: 1. Estimated annual total amount of gallons of fuel reported on the IFTA form; 2. Estimated annual total miles in the base state reported on the IFTA form; 3. Total estimate of miles in the base state as percentage of the total miles including travel in other jurisdictions; 4. Number of IFTA decals received; 5. Number of IFTA decals returned (not used by the end of the year); 6. Number of the motor fuel carrier fleet trucks; 7. Number of motor fuel carrier drivers; and 8. Number of states within which the motor carrier operates. 5.4.3.3 Data Needs and Limitations The estimate of total taxes due is a difficult value to obtain. The value is needed to calculate the percent of EOE. The vari- able isn’t the amount owed or the amount assessed, but the amount of tax paid in all jurisdictions. This value must be cal- culated by summing the amount paid in all jurisdictions plus any debits or credits found during the audit. The value is not one normally kept in the process of storing IFTA audit records but it should be if the amount of evasion is to be calculated. The analyst can compute this total if the audit data include gal- lons reported by each jurisdiction. Matching reported gallons 74

with published tax rate data can be done to construct total tax liability. Another variable that would be useful is the amount owed to other jurisdictions. If the amount owed other juris- dictions were available, the amount of EOE occurring in the home state could be calculated. If it isn’t known, then the value calculated is the amount of EOE incurred by the home state motor carriers against the IFTA system. Section 5.5 includes two sub-sections outlining the data re- quired to estimate EOE associated with IFTA abuse. The first sub-section lists the general IFTA data items needed to estab- lish total tax liability by motor carriers subject to IFTA in the examined state. The second subsection identifies data need- ing to be extracted from individual audit records to estimate EOE for the sample of carriers audited under IFTA. 5.4.4 Evasion Associated with Exporting/ Importing Fuel Across State Lines 5.4.4.1 Overview of Evasion Technique Import/export fraud occurs when the taxpayer chooses not to pay taxes in the high-tax state. For example, this can occur if the taxpayer claims export of fuel across a state line and does not pay the tax to the importing state or if the taxpayer claims the export of fuel from his state and then does not ex- port the fuel. Export evasion occurs when fuel is purchased in a state with a low tax rate and exported to a state with a higher tax rate without paying the higher taxes. The tax evader yields extra profit equal to the difference between the tax rates for each gallon illegally imported. Figure 5-4 demonstrates one form of the import-export evasion technique. In this example, a motor fuel distributor claims to sell 100 gallons of fuel in State A but actually exports it to State B. In turn, the distributor fails to claim imports and remit proper payment to State B. The distributor profits from the differential in tax rates be- tween State A and State B. Profit from tax evasion is captured in Equation 5.6. where E = $ evasion due to bootlegging; TRb = Tax rate in State B; and TRa = Tax rate in State A. E b a= ×( )− ×( )100 100TR TR 75 3% or More Less than 3% Source: IFTA, 2006 Figure 5-3. IFTA audits by state (average annual 2000–2004). State A State B Claim Actual 100 gallons sold in State A 100 gallons sold in State B Figure 5-4. Exporting across state lines without paying State B fuel taxes.

False claim of export is another export-import evasion technique and is similar to the previous technique except the flow of fuel runs in opposite directions. In illegally exporting fuel across state borders, the taxpayer claims to sell the fuel in its home state but transports and sells it in a neighboring state. When making false claims of export, the perpetrator re- verses the operation claiming to export the fuel but selling it in the home state. With false claim of export, perpetrators buy fuel within one jurisdiction and file paperwork claiming it as tax exempt be- cause it is targeted for delivery in another jurisdiction where taxes will be applied upon delivery. In turn, the fuel is actu- ally sold within the original jurisdiction and never exported, thus avoiding the tax. This scheme occurs between states and across international borders. Figure 5-5 demonstrates how to apply this technique. In this example, a distributor claims that 100 gallons are des- tined for export to State B. In turn, the distributor sells the fuel in State A and fails to remit tax payments. In this case, evasion is equal to the product of the 100 gallons and the tax rate in State A. Tax evaders go to extreme lengths to mask their crimes and garner profit from evasion. For example, there is evidence to suggest that perpetrators have dumped their fuel within a state and refilled the carrier tank with water so that a weigh station would assume that there is fuel inside the tank (Turner, 2004). The most significant factor motivating evaders to falsify claims on motor fuel shipments between states is the wide range of differences between state motor fuel tax rates. The JFSMFTCP, which was established to reduce motor fuel tax eva- sion between the states by maximizing efforts in motor fuel tax auditing, criminal investigations, and enforcement, has project goals that include the creation of automated data processing tools to monitor motor fuel production, and imports and ex- ports across state and international borders. To achieve these goals, it is required to implement registration and reporting sys- tems for motor fuel producers, distributors, and retailers. 5.4.4.2 Model Approach Import-export evasion across state lines is relatively hard to evaluate without significant efforts to catch the evasion and evaluate whether there is a problem. Cooperating states with tracking systems can partially capture import-export evasion but they can’t capture all. A tracking system can only capture those entities that fail to pay their taxes if they report they are exporting to State A or importing fuel. However, if the entity claims to not be exporting and then does export, it can only be caught by audits. Joint audits of supplier and purchaser at the distributor or retail level within the receiving state are the only way that type of evasion can be detected fully. The approach to measure import or export evasion is two- fold. If tracking systems are available, differences can account for those who don’t report or underreport. In addition, audit information that targets both sides of a transaction can add further information about the amount of evasion occurring. When performing retail audits to detect cross-border eva- sion, it is important to reconcile inventory at the station by comparing inventory reports, meter readings or stick read- ings with BOLs, accounts payable, and invoices. This practice will ensure that each load of motor fuel is accounted for, that the payments made by the retailer include a tax component, and that each tax payment is received. It is recommended that states reconcile at least two to three months’ records. There are a number of EOE indicators to consider when conducting retail audits, including: • The retailer keeps poor records or is missing records; • The retailer regularly purchases fuel at below-market rates; • Meters break repeatedly; • There are discrepancies between inventory records and BOLs, accounts payable or invoices; • The retailer experiences a large change in sales volumes; and • There are discrepancies between distributor reports and the data obtained from the retailer. Using the approach outlined above, the audit could yield an assessment that would serve as the dependent variable. In- dependent variables include the operational, business, audit, and other characteristics targeted in the second general audit data element subsection in Section 6.5. 5.4.4.3 Data Needs and Limitations Data needed include field audits of both purchaser and suppliers in addition to the desk audits. Sources of export and import data could be obtained through one of two sources: (1) IRS—EXTOLE system or (2) state import/export data and motor fuel tracking systems. Section 5.5 includes a sub- section that outlines the import/export data needed to estab- lish total motor fuel volumes imported to, and exported out of, each state. Retail audit data requirements are outlined in the second general audit data sub-section of Section 5.5. 76 State A State B Claim Actual 100 gallons sold in State A 100 gallons sold in State B Figure 5-5. False claim of export.

5.4.5 Illegal Importation Across International Borders 5.4.5.1 Overview of Evasion Technique Evasion sometimes occurs by way of international borders when untaxed fuel is smuggled into the country and sold to retailers at taxed rates. Perpetrators of this scheme take ad- vantage of the fact that state and federal agencies have no ju- risdiction over foreign fuel supply operations. Fuel can be purchased from foreign entities and brought into the United States and distributed under the radar of the IRS and state tax agencies. 5.4.5.2 Background on Evasion Technique Under this scheme, fuel is bought from foreign refineries or bulk dealers and transported to the United States by truck or shipped by ocean vessel. By truck, fuel can be illegally im- ported and delivered to retail stations or perpetrator-owned terminals or bulk plants. The owed taxes are not paid. If fuel is delivered to terminals or bulk plants, required reports are not filed. At border crossings, truckers are required to present, if re- quested, a bill of lading to U.S. Customs. These BOLs can be forged. Further, there are other border crossings not rou- tinely attended by Customs agents that these trucks can travel on; however, security at all border crossings has tightened considerably since 9/11. Problems with international fuel smuggling into the United States appear to primarily be imports from Canada rather than Mexico; the geographic area of concern includes states close to the northern border. There are 130 land-border points on the Canadian-U.S. border and most of the volume of mer- chandise is transported by motor carriers. In a recent study conducted by the Joint USA/Canadian Motor Fuel Compli- ance Initiative (FHWA, 1994a), cross-border inspections in the Northeast resulted in 2 percent dyed diesel fuel violations and numerous other violations related to improperly docu- menting cross-border fuel shipments. Other states that would need careful scrutiny are those states with access to fuel off- loading with connections to transportation routes. Studies in- dicate that a portion of the fuel tax nonpayment is associated with failure to pay on the entire amount of fuel off-loaded. 5.4.5.3 Model Approach The methodological approach to estimating the amount of fuel tax evasion uses the audit and inspections analysis, as well as an additional tracking approach: 1. Statistical sampling, and 2. Tracking. 5.4.5.3.1 Statistical Sampling. If sufficient cross-border inspections and audit data exist regarding cross-border eva- sion schemes, the methodological approach to estimating the amount of fuel taxes evaded will use probability sampling from inspections and audit data. This data is potentially avail- able from some states and the IRS, as well as recently con- ducted inspections by the Joint USA/Canadian Motor Fuel Compliance Initiative (various years). Data need to be evalu- ated to remove bias (e.g., inspections based on tips) such that the sample is random. If the bias cannot be removed, methods to estimate the amount of bias must be explored. The number of violations (in terms of total fuel evaded) as a percentage of total fuel covered through inspections/audits will be the basic statistic used to measure evasion. For each type of fuel, this percentage will be expanded to the targeted population by multiplying the proportion of violations to inspections times the total amount of fuel that reportedly crosses the border. For example, the estimate would be as follows: where Ei = $ Evasion for Fuel Type i; Xi = violations attributed to Fuel type i (in gallons of fuel); Ni = total inspections of Fuel type i (in gallons of fuel); Fi = total amount of Fuel type i reportedly imported from Canada by a particular U.S. State or the amount of fuel imported by a particular state; ti = tax rate in state for fuel type i; and where i = gasoline, diesel. 5.4.5.3.2 Tracking. If adequate inspections and audit data are not available, the amount of evasion occurring through cross-border schemes could be estimated by comparing the total amount of fuel imported from Canada by specific states with the amount of fuel officially imported and sold in specific states. These states would primarily include northern states and states with fuel off-loading facilities and access for fuel off- loading. Other states that could be susceptible to this type of evasion include states with off-loading facilities for fuel from ocean-going barges and tankers. Southern-border states with Mexico do not appear to be at risk for this type of evasion, as fuel is significantly more expensive in Mexico. 5.4.5.4 Data Needs and Limitations The IRS may have data on fuel imports. In addition, state audit data on dyed fuel import violations could provide further information if available. During the summers of 2003 and 2004, the Canadian/USA Tax Compliance Initiative conducted E X N F ti i i i i= × × 77

a series of cross-border inspections, documenting various fuel tax violations in the Northeast. Data required to estimate EOE using this approach is outlined in the illegal importation data element subsection of Section 5.5. The general audit data ele- ment subsections also could be used to estimate EOE associ- ated with illegal importation across international borders. U.S. Customs and some states have records of the amount of fuel imported from Canada, while the Canadian govern- ment tracks the amount of fuel reportedly exported to the United States by state. This approach requires that a state have a useable tracking system where imported fuel is recorded. 5.4.6 Failure to Remit 5.4.6.1 Overview of Evasion Technique Many states allow distributor registrants to purchase fuel untaxed. Fuel tax evasion perpetrators either obtain a regis- tration legally or illegally or forge the registration documen- tation. Tax-free fuel is purchased and then sold as tax paid to other wholesale distributors or retailers. They evade the taxes by simply failing to file or filing false returns with the state. A perpetrator may get away with this for some time before en- forcement agencies can detect them due to long time periods between the filing of reports and remittance of tax. Further, the state agency must check the evading company’s reports against other businesses to detect discrepancies or must find irregularities in the tax filing during the auditing process. One example of a recent case in Ohio involves a perpetra- tor who forged a registration and used an alias to purchase untaxed diesel, which he then sold to truck stops all over Ohio. This scheme can be more quickly discovered by imple- menting a fuel tracking system that matches terminal dis- bursements with distributor reports. This scheme occurs at the point of taxation. The perpetrators typically fail to file re- turns or file false returns for taxes paid (reporting a certain amount but not the total amount of taxable fuel sold). (See Chapter 2 on Evasion Methods for more details.) 5.4.6.2 Model Approach There are three methodological approaches that are sug- gested in order to estimate the EOE of this evasion technique, depending on data availability: 1. Motor Fuel Tracking; 2. OLS or Tobit; and 3. Statistical Sampling. 5.4.6.2.1 Tracking. Provided that a state has a good track- ing system, the best approach to catching EOE in failure to remit is the tracking system. In the tracking system approach, the amount of fuel delivered for sale is compared with the state’s motor fuel tax receipts. Careful analysis of the infor- mation would need to account for any double counting of the failure to remit on importation. 5.4.6.2.2 OLS or Tobit. Provided that sufficient data are available for individual filers through the audit and inspec- tions process, a regression approach could be adopted. Ex- planatory variables would be developed for each sector eval- uated. Explanatory variables might include • Gallons of fuel; • Type of business; • Examiner; • Industry affiliation; • Fuel types; • Terminals used; • State of origination; • Number of audits; and • Detection rate of examiner. 5.4.6.2.3 Statistical Sampling. This approach will prima- rily rely on audit data. If a sufficient amount of audit infor- mation exists regarding this evasion scheme, the methodolog- ical approach to estimating the amount of fuel evaded uses a probability sampling method from audit data produced. This data is potentially available from some states. Data must be evaluated to remove bias (e.g., inspections based on tips), such that the sample is random. If the bias cannot be removed, methods to estimate the amount of bias will be explored. In- stances of EOE (in terms of total fuel taxes not remitted con- verted to gallons) as a percentage of total fuel covered through audits will be the basic statistic used to measure EOE. This per- centage will be expanded to the targeted population by multi- plying the proportion of EOE to inspections times the total amount of fuel used in the selected population. For example, the estimate would be as follows: where Ei = $ Evasion of fuel i; Xi = number of cases where EOE (in gallons of fuel) for fuel i were found; Ni = total inspections/audits (in gallons of fuel) for fuel i; Fi = total amount of taxed motor fuel sold in state for fuel i; and ti = tax rate in state for fuel i. An alternative to this approach would be to examine track- ing data for states to determine the amount of taxed fuel re- ported and the amount of taxes collected. The difference E X N F ti i i i i= × × 78

between the two would be the level of EOE for that state. If audit data and tracking systems were available, estimates from the two could be used to determine the level of bias in the estimates from the audit data sampling approach. 5.4.6.3 Data Needs and Limitations State audit data as outlined in the general audit data elements in Section 5.5 are required to estimate EOE. In Section 5.5, the first general audit data element would be used to establish the characteristics of the taxpayer population. The remaining general audit data elements are differentiated based on the point in the distribution system where the audits are being conducted. 5.4.7 Cocktailing and False Labeling 5.4.7.1 Overview of Evasion Technique Many products are tax-redeemed, not taxable, or not tracked by that state or the IRS, which can be obtained under false pretenses and used in gasoline or diesel engines. Examples in- clude aviation fuel, used motor oil, and mineral spirits. By blending these products with taxable fuel, fuel volumes can be extended. Perpetrators either can blend these products for their own use or they can profit from the tax collected on sales or the number of extra gallons created through blending. Sometimes fuels that are untaxed or have reduced tax rates (e.g., kerosene, jet fuel) can be used as a substitute for taxable fuels without any blending necessary. In all cases, fuel tax evaders would most likely falsely label taxable products as nontaxable products at the point of taxation, but eventually sell it for a taxable purpose. A recent scheme uncovered in Massachusetts involved an oil company that was blending untaxed kerosene and home heating oil with diesel and not reporting tax on the blend. An- other scheme in Florida involved an airport employee who was illegally siphoning jet fuel from the airport system and then labeling and selling it as taxed diesel at the retail level. This scheme occurs below the terminal rack. Many perpe- trators have access and licenses to distribute untaxed or tax re- duced fuel, such as heating oil and kerosene. Figure 5-6 shows where evasion leakage occurs (note “Bulk Plant” is actually part of the nonbulk distribution). 5.4.7.2 Model Approach The methodological approach to estimating the amount of fuel tax evasion uses an approach under the audit and in- spections analysis as well as an additional tracking approach. For blending agents where the intended use is relatively easy to identify, isolate and measure, such as jet fuels, the supply and use approach is suggested. For all other blending agents, the statistical sampling approach is recommended. 1. Statistical sampling and 2. Supply and use. 5.4.7.2.1 Statistical Sampling. The first approach will in- volve statistical sampling, where the blending agent and its var- ious uses will be used as the expansion coefficient for each sam- ple. The number of violations (in terms of total fuel evaded) as a percentage of total fuel covered through inspections/audits will be the basic statistic used to measure EOE. For each blend- ing agent or substitute fuel, this percentage will be expanded to the targeted population by multiplying the proportion of vio- 79 $ $ Card-lock $ Cocktailing Retail Figure 5-6. Where cocktailing occurs in the distribution system.

lations to inspections times the total amount of fuel used in the selected population. For example, the estimate would be: where Ei = $ Evasion for fuel type i; Xi = Violations involving blending or falsely labeling fuel type i (in gallons of fuel); Ni = Total inspections involving fuel type i (in gallons of fuel); Fi = Total amount of fuel type i used in a particular state; ti = Tax rate in state for i type of fuel; and i = Gasoline, kerosene, jet fuel, mineral spirits, and other tax-exempt or tax-reduced fuels. Ei will be measured for all fuel types that are typically sub- stituted or blended with diesel or gas. Considering not every state will have inspections data, states could be grouped into statistically appropriate clusters for each type of evasion measurement. To extrapolate the total popula- tion, it will be important to examine the use of untaxed (or tax- reduced) fuels that are commonly used as substitutes for gas and diesel and/or can be easily blended with gas and diesel. This would involve closer examination of jet fuel use, kerosene, bio- diesel, mineral spirits, etc. It would be much more difficult to examine the total amount of waste oils available for blending, as this is not routinely tracked. 5.4.7.2.2 Supply and Use. The supply and use approach could be applied to blending agents that are relatively easy to identify, isolate, and measure, such as jet fuel. In such a case, the supply and disappearance of jet fuel would be measured, where final discrepancies between supply and use would be attributed to fuel tax evasion. 5.4.7.3 Data Needs and Limitations Inspections data could reveal the degree to which cocktail- ing is occurring. Inspection data requirements are outlined in the cocktailing and false labeling inspection data element sub- section of Section 5.5. In addition, state audit data regarding fuel tax violations could provide further information if avail- able. The data required to estimate EOE are outlined in the general audit data element sub-sections of Section 5.5. 5.4.8 Abuses Due to the Presence of Native American Reservations 5.4.8.1 Overview of Evasion Technique Issues faced by tax agents and compliance officers due to the presence of the Native American exemption are signifi- cant. According to the Bureau of Indian Affairs, there are E X N F ti i i i i= × × 562 federally recognized tribal governments in the United States. These governments are spread out geographically over the United States, from Alaska to Florida and from Maine to California. There are concentrations of Native American tribal governments in New Mexico, Arizona, Colorado, and Nevada. Figure 5-7 presents a map of the Native American Reservations in the continental United States. As noted in FTA’s Survey of Native American Issues, a number of states have entered into agreements for the collec- tion of taxes with Native American Tribes (Arizona, Louisiana, Minnesota, Montana, Nebraska, Nevada, Oklahoma, South Dakota, Utah, Washington, and Wisconsin), are in active negotiations with tribes (Arizona, Connecticut, Montana, Nebraska, Nevada, North Dakota, Oregon, Utah, and Wiscon- sin) and are currently embroiled in litigation with tribes over the issue of motor fuel taxation (Idaho, Kansas, Minnesota, Nevada, and Pennsylvania) (FTA 2002a). While some states do have agreements about administering state fuel taxes in place with tribes, court cases in other states have determined that taxation of fuel in these lands would vio- late the sovereignty of these nations. In many states, Native American retail outlets may purchase tax free fuel or obtain a refund for fuel distributed to reservation residents. One evasion scheme arises from the fact that fuel can be imported from other states and foreign points of origin and delivered directly to Native American reservations without taxes being remitted. Native American tribes have the responsibility to collect state fuel taxes when a non-Native American purchases gaso- line or diesel fuel from a tribal retail outlet. Battelle discussed this issue with the JFSMFTCP contact in Idaho and explained the only determining factor associated with motor fuel tax compliance is the relationship between the state and the tribal government (Walters, 2004). In some states, there is an open dialogue between the tribal governments and the state. In others, the dialogue is not as open and data on motor fuel sales is more difficult to obtain. 5.4.8.2 Model Approach Two approaches could be used to estimate evasion due to Native American sales of motor fuel. Both rely on a compar- ison of estimated motor fuel sales and consumption by regis- tered tribal members, combining elements of the supply and use approach and tracking, when possible. 5.4.8.2.1 Tracking. The tracking approach for estimating motor fuel sales is based on the examination of distributor re- ports to track motor fuel sales to reservations. The amount listed on distributor reports would be compared with motor fuel sales to detect evasion. 5.4.8.2.2 Statistical Analysis of Sales. In the absence of complete distributor reports, an alternative approach would 80

be to construct a model where the number of gallons of motor fuel sold by individual Native American establish- ments would serve as the dependent variable, and a number of independent variables that could be used to estimate gal- lons sold would be identified and tested. Independent vari- ables could include: • The number of Native American Reservations in a state; • The number of retail motor fuel outlets on reservations; • The number of pumps located at each Native American re- tail outlet; • State motor fuel excise tax rates; • State populations; and • Proximity to high-tax states. Data provided by states with agreements would serve as the base data required to test the predictive capacity of the model. The model would be designed to test the correlation between the variables outlined above (and others tested during model development) and actual motor fuel sales for the Native American establishments reporting data to state taxing au- thorities. In turn, the model can be used to predict motor fuel gallons sold by Native American establishments not report- ing to state taxing authorities. Tribal member consumption would be estimated based on data relating to the number of registered members and esti- mates of average per-capita fuel consumption for state resi- dents. Estimated tribal member consumption would be com- pared to modeled estimates of Native American motor fuel sales. The difference between these two estimates would rep- resent EOE. 5.4.8.3 Data Needs and Limitations State distributor fuel tax forms often include fields for reporting sales to Native American Tribes. Further, Native American retail outlets generally cannot obtain fuel from sources other than the distributors that are reporting to the state, with the exception of tribes that operate refineries. To the extent that data are provided in paper form or gaps ap- pear in data collection from distributors, the data could be in- sufficient to support the proposed primary analysis. Data required to support the EOE estimation methods are detailed in the Native American data element subsection of Section 5.5. 81 Source: Bureau of Indian Affairs. Figure 5-7. Native American reservations in the continental United States.

5.4.9 Daisy Chains 5.4.9.1 Overview of Evasion Technique In a daisy chain scheme, a ring of artificial companies transact several fallacious purchases of fuel without paying taxes. The fuel is eventually sold at taxed rates to a legal re- tail operation. The daisy chain represents a multiflow fraud scheme that involves the creation of entities that use artifi- cial trusts and accounts to avoid tax obligations. When in- vestigators track the purchases of the fuel in an effort to track tax liability, one of the dummy companies, known as the burn company, dissolves along with any tax liability. This scheme could still be used in some states; however, its significance as an evasion technique has declined due to the movement of the point of taxation for the federal govern- ment and many states up the distribution chain to the ter- minal rack. Hwang et al. (2003b) described the daisy chain as a long, indirect, and complex paper trail of motor fuel tax docu- mentation, which makes it difficult for auditors to track and discover the evasion. This practice can be used for evasion at both federal and state fuel tax levels. Daisy chains are not successful when the points of taxation are at the retail or ter- minal rack level. Taxing at the retail level thwarts daisy chains because tax must be remitted once fuel is sold to the motorist. Thus, any amount of misdirection in the paper- work accompanying fuel shipments fails to hide tax liability because it is not incurred until the fuel is sold to the motor- ing public. At the terminal level, large terminal operators pay the tax when it breaks bulk and is purchased by distributors. The daisy chain works effectively only when the tax is at the distributor level; the distributor can purchase the fuel tax- free at the terminal rack, run it through the daisy chain, and then sell it at taxed rates to unwitting retailers at the other end (Figure 5-8). In a study examining the optimum point of taxation for motor fuel excise taxes, Brand (1996) found that the move- ment of the federal diesel fuel point of taxation in 1994 re- duced noncompliance significantly. This reduced the num- ber of taxpayers from more than 50,000 to around 1,500. The study also concluded there was an immediate and significant increase in revenue due to increased motor fuel excise tax compliance. 5.4.9.2 Model Approach The model must account for the dampening impact on daisy chains when the point of taxation is either at the retail or terminal level or when tax is remitted high in the distrib- utor level. The first step in the estimation of tax evasion due to daisy chains is the assessment of the state’s motor fuel tax program to determine whether or not the program can be exploited by the daisy chain. In the event the assessment determines that motor fuel tax evasion may be occurring through the daisy chain mechanism, the next step in the model approach would employ some elements of the audit and inspections analysis by examining historical auditing records to determine the extent shipments are being received by retailers or transactions are made between distributors with companies not licensed by the state. This approach would broaden the daisy chain method to include transac- tions between reputable licensed companies and unlicensed criminal operations. 5.4.9.3 Data Requirements and Limitations The proposed model depends in large part on sound dis- tributor and/or retailer audit data. Alternatively, states that have implemented state motor fuel tracking systems would have the advantage of a more complete and potentially com- pletely electronic set of distributor sales data. These records could be used to identify fuel sales and purchases by unli- censed, unlawful distributors. Errors in data reported by distributors and paper reporting in many states make this model approach more difficult, more time consuming, and less accurate. Data required to estimate EOE are outlined in the first three general audit data element subsections of Section 5.5. 5.5 Detailed Data Recommendations 5.5.1 Introduction This section presents detailed recommendations govern- ing data collection in support of the EOE estimation ap- proaches outlined in Sections 5.4.1 through 5.4.9. The data outlined, however, are not tied in all cases to a specific esti- mation approach or even an evasion method. Rather, the 82 Terminal Rack Distributor A Distributor B Distributor C “Burn Company” Distributor D Retailer Figure 5-8. Daisy chain.

data elements are categorized based on the specific types of investigations and audits generating the original data. These investigations and audits are those that states already per- form, or could perform, as part of their regular enforcement programs. To understand the link between the data collec- tion recommendations and EOE estimation, refer to the data requirements and limitations subsection of Sections 5.4.1–5.4.9. Modeling approaches also are outlined in those sections. The remainder of this section details a small number of is- sues to consider when collecting data, including the quantity of information to be collected, the temporal element of data collected, how to define EOE for modeling purposes, and the importance of controlling for double counting. 5.5.2 Quantity of Information When constructing a database to support the EOE calcu- lation, it is important to consider the quantity of the infor- mation required. When conducting an analysis using the statistical sampling approach, it could require collection of hundreds of observations to generate estimates with margins of error of less than plus or minus 4 percent. This confidence level likely would be achievable for some audits and inspec- tions, such as IFTA audits or on-road dyed diesel inspections. However, this level of precision likely would not be feasible for most other forms of audit or inspection given time and budget limitations. For other more costly elements that are often performed with less frequency, it is recommended that at least 30 audits or inspections be performed in support of the EOE estimation approach. In most cases, results from 30 observations could yield results that would be considered sta- tistically valid. 5.5.3 Temporal Element of Data Collected The temporal element also must be addressed when col- lecting data. Past models used to estimate evasion have gen- erally been highly unstable, yielding results inconsistent from one year to the next. Estimated EOE under any model will vary from year to year based on data anomalies, inconsisten- cies in data collection techniques, real changes in evasion, en- forcement activities, and changes in tax code. Analysts have the option of collecting data over an extended period of time to determine the impact of various tax code or programmatic changes (e.g., moving the point of taxation up the distribu- tion chain); however, long-term data collection efforts are not required to estimate EOE. Thus, it is recommended that data sets used to estimate EOE cover at least four years when there was general consistency with tax codes, enforcement programs, and data collection techniques. To the extent there are inconsistencies in available data, it will be important to document the factors (e.g., changes in data collection tech- niques, responsibility for data collection being shifted from one public agency to another) that impact consistency in rel- evant data series. 5.5.4 How to Define EOE EOE is a term developed for this study to describe the value to include as the dependent variable in any EOE model. The distinction between this term and evasion is that it does not attempt to attach intent to the act of failing to make a full tax payment. That is, an assessment may result due to an omission on the part of a taxpayer, an inadver- tent error, or willful evasion. The intent of the taxpayer is often impossible to ascertain. For the purposes of this re- port, intent is not considered. It is also important to note that taxpayers generally have a mechanism to appeal the findings of a tax audit. In many instances, this appeal will lead to an adjustment in the assessment amount. To the ex- tent that an appeal or any other audit review process leads to an adjustment to the initial assessment, the final assess- ment is what should be considered in the EOE calculation. Also, penalties and interest should not be included in the EOE calculation. 5.5.5 The Need to Control for Double Counting It is important to control for double counting as necessary. For example, IFTA audits may include much more than sim- ply IFTA-related EOE. IFTA audits may capture illegal blend- ing, failure to remit, and other evasion methods used by motor carriers to evade taxes. To the extent these evasion methods are picked up through other EOE estimation ap- proaches, the modeler should be careful to not double count the EOE. For example, one evasion study relied on a model to estimate cross-border distributor EOE (Balducci et al., 2006). The authors of this study also reviewed but did not use audit data that could have been used to examine cross-border EOE. As noted in the report, the use of both techniques would have resulted in double counting. 5.5.6 Data Recommendations Data outlined in this section of the report should be consid- ered an ideal case rather than a requirement. States need not collect all the data outlined in this section to estimate EOE. As more data are collected, EOE estimates will be more precise and more confidence can be attributed to the results. Many states collect little of the data outlined in this section or collect detailed 83

data on violators while collecting limited data on audits and in- spections that do not yield assessments. For those states with limited existing data, this section could be used to design a data collection program to support future evasion studies. 5.5.6.1 Dyed Fuel Data Element #1 The following general data items are needed on dyed fuel inspections: • Number of dyed fuel violations classified by the type of violation; • Estimate of the total value of dyed fuel violations by viola- tion type (misuse, dye removal; from fuel, mislabeling) not including penalties or interest; • Estimate of the number of gallons associated with dyed fuel violations (misuse, dye removal from fuel, mislabeling); • Triggers for inspections (e.g., tax evasion hint or report by a third party, random sampling, regular on-road inspec- tion, or other . . . specify); • Total number of inspections by type of inspection; • Total number of dyed fuel gallons consumed in state; • Total number of gallons inspected by type of inspection (on-road, site inspection); and • Total taxable gallons of diesel burned on-road in state by all taxpayers. 5.5.6.2 Dyed Fuel Data Element #2 The following data items extracted from individual inspec- tion data are needed for all dyed fuel inspections: • General Inspection Information – Date inspection performed – Location (county, city) where inspection conducted – Highway number – Location of inspection (e.g., road inspection, site visit, weigh station) • Driver Information – D.L. State • Vehicle Information – Registered weight – Vehicle type (e.g., car, pickup, single-unit truck or com- bination) – Fuel tank capacity – Private or for hire – Commodity code – Interstate or Intrastate – Leased or owned • Sample Information – Number of samples taken – Tank location and capacity – Name of fueling location – Terminal code • Business characteristics of inspected companies – Years in operation – North American Industry Classification System (NAICS) code – Annual revenue – Number of employees – Motor fuel types • Trigger for inspection (e.g., tax evasion hint or report by a third party, random sampling, regular on-road inspection, or other . . . specify) • Type of operation inspected (e.g., retail gas stations, farm, construction, motor fuel carriers, logging, motor carrier, individual) • Types of violations found if any (e.g., dyed fuel used on- highway, dyed fuel signs missing from pump station, dis- tributor sold dyed fuel for consumption on the highway, no violation found . . . specify) • If violation found, the following variables are needed – Type of enforcement taken ▪ Civil (reason for considering civil enforcement) ▪ Criminal (reason for considering criminal enforcement) – First, second, or third offense or greater – Number and tax value of gallons in which the assess- ment was based. 5.5.6.3 Refund and Tax-Exempt Fuel Data Element The following refund and tax-exempt fuel data items are needed: • General taxpayer information needed to estimate total motor fuel consumption and true tax liability – Summary of bulk storage data for both gasoline and undyed diesel ▪ Annual total beginning inventory ▪ Annual total fuels received into storage ▪ Annual total ending inventory of fuel ▪ Annual total dispensed into vehicles ▪ Annual total dispensed into equipment ▪ Annual total miles traveled in all jurisdictions reported by refund claimants for each year for both on- and off-road – Annual total miles traveled by claimants for each year on public roads within the state estimating evasion – Annual total miles traveled by claimants for each year off-road in state – Annual total taxed gallons within the state estimating evasion at the pump placed into equipment for gasoline and undyed diesel separately 84

– Annual gallons for which refunds were applied within the state estimating evasion by fuel type – Number of registered off-highway vehicles/equipment and average fuel consumption classified by type of vehi- cle and type of fuel (examples provided below): ▪ Government vehicles and equipment (federal, state, counties, and cities government agencies) ▪ Agricultural equipment, (e.g., tractors and combines) ▪ Commercial equipment ▪ Logging equipment ▪ Construction and mining equipment (e.g., graders, cranes, paving equipments, and earth moving equip- ment) ▪ Industrial equipment (e.g., forklifts, aerial lifts, min- ing equipment and logging equipment) ▪ Recreational vehicles and equipment (e.g., boats) ▪ Residential and commercial lawn and garden equip- ment ▪ Marine vehicles and equipment ▪ Locomotive equipment ▪ Airport Equipment ▪ Aircraft ▪ Pleasure craft ▪ Other exempt uses – Number of special fuel registrations for out-of-state users for recreation or for religious, charitable, educa- tional, or other purposes • Individual refund audit data required – Annual total number of false refund claims classified by violation and fuel type – Total number of inspections by type of inspection – Audit data ▪ Trigger for audit (e.g., tax evasion hint or report by a third party, random sampling, regular on-road in- spection, or other . . . specify) ▪ The IFTA EOE dollar value assessment (minus penalties and interest) ▪ Gallons on which the EOE assessment was made by fuel type – Auditor information ▪ Years in service ▪ Rank ▪ Detection rate for the auditor – Business characteristics of inspected companies ▪ Years in operation ▪ NAICS code ▪ Annual revenue ▪ Number of employees ▪ Motor fuel types sold or used – If violation found the following variables are needed ▪ Number of gallons on which assessment was made by type of violation and fuel type ▪ Amount of the assessment minus penalties and inter- est by type of violation and fuel type ▪ Type of enforcement taken • Civil (reason for considering civil enforcement) • Criminal (reason for considering criminal enforce- ment) • First, second, or third offense or greater 5.5.6.4 Import/Export Fuel Data Element The following import/export fuel data items are needed: • Monthly number of businesses requesting tax exemption due to exporting fuel to the other states • Monthly gallons exported from state by type of fuel and the destination jurisdiction as reported to the state • Monthly gallons imported to the state conducting the eva- sion analysis classified by type of fuel as reported by desti- nation jurisdiction • Monthly gallons exported and imported via modes of trans- portation other than tanker trucks (pipeline, barges, and rail) 5.5.6.5 IFTA Data Element #1 The following general IFTA data items are needed: • Estimated total amount of gallons of fuel reported in IFTA forms by fuel type • Estimated total miles traveled in the base state reported in IFTA forms by fuel type • Total estimated miles traveled in the base state as percent- age of the total miles traveled including travel in other ju- risdictions by fuel type • Number of IFTA decals purchased • Number of IFTA decals returned (not used by the end of the year) • Number of IFTA audits completed • Number of IFTA audits resulting in an assessment • Dollar value of IFTA assessments made by base-state on behalf of other jurisdictions classified by jurisdiction and fuel type • Total dollar value of the EOE assessments made on behalf of the base-state classified by jurisdiction and fuel type • Total dollar value collected on behalf of the base-state clas- sified by jurisdiction and fuel type • Dollar value of total IFTA tax collections • The percentage of IFTA audits completed and assessed rel- ative to the total number of IFTA registered motor carriers • Total number of IFTA accounts • Percentage of total number of motor carriers audited under IFTA • Percentage of total gallons consumed by motor carriers au- dited under IFTA 85

5.5.6.6 IFTA Data Element #2 The following data items extracted from individual audit records are needed for all IFTA audits: • Date IFTA audit performed • Number of IFTA decals the motor carrier used • Annual taxable miles traveled in state by fuel type • Annual taxable gallons used in state by fuel type • Annual total taxable miles traveled in other jurisdictions by fuel type • Annual total taxable gallons used in the other jurisdictions (classified by jurisdiction) by fuel type • Miles per gallon registered by motor carrier • Base state percentage of total miles traveled • The dollar value of the IFTA EOE assessment (if no assess- ment made then assessment amount equals zero) • The dollar value of the IFTA EOE assessment made on behalf other IFTA jurisdictions (classified by jurisdiction) • Other types of motor fuel tax EOE detected during the IFTA audit (specify the technique of evasion e.g., dyed fuel, failure to remit, importation violation . . . etc.) • If available, any other relevant business information of the motor fuel carrier – Years in operation – NAICS code – Annual revenue – Number of employees – Primary commodity type hauled or type of operation (e.g., agriculture products, concrete and aggregate, forest products) – Number of safety violations • Auditor information – Years in service – Rank – Detection rate for the auditor • Trigger for audit (e.g., tax evasion hint or report by a third party, random sampling, red flag that triggered audit, or other . . . specify) 5.5.6.7 General Audit Data Element #1 General audit data needed: • Percent of the total population audited by operation type (distributors, retailers etc.) • Percentage of total gallons consumed by audited companies by fuel and operation type • Trigger of audits (third party tip, random sampling, flagged return or other . . . specify) by fuel and operation type • Percentage of all audits assessed from total completed audits by fuel and operation type • Total number of audits by type (field audit or office audit) 5.5.6.8 General Audit Data Element #2 Individual general audit data for all audits performed by states taxing at the retail level: • Date audit performed • Type of audit (field audit or office audit) • Trigger for the audit (third party tip, random sampling, flagged return or other . . . specify) • Operational characteristics – Type of operation audited (truck stop, card lock fueling station, gas station) – Is the company licensed to sell exempt fuel? – What motor fuel products does the company sell? – Location(s) of station(s) audited – States in which the company is licensed to operate • Type of violations found • If violation found, the following variables are needed – Number of gallons on which assessment was made by type of violation and fuel type – Amount of the assessment minus penalties and interest by type of violation and fuel type – Type of enforcement taken ▪ Civil (reason for considering civil enforcement) ▪ Criminal (reason for considering criminal enforcement) ▪ First, second or third offense or greater • Business characteristics of inspected companies – Years in operation – NAICS code – Annual revenue – Location of company headquarters – Number of employees – Number of taxed gallons – Motor fuel types • Auditor information – Years in service – Rank – Detection rate for the auditor 5.5.6.9 General Audit Data Element #3 Individual general audit data for all audits performed by states taxing at the distributor level: • Date audit performed • Type of audit (field audit or office audit) • Trigger for the audit (third party tip, random sampling, flagged return or other . . . specify) 86

• Operational characteristics – Type of operation audited (distributor, importer, al- ternative fuel producer, bulk purchasers, special fuel dealers) – Is the company licensed to distribute exempt fuel – If feasible, determine the average number of times that loads change ownership prior to delivery – What motor fuel products does the company distribute – Is the distributor an importer and/or exporter – States in which the company is licensed to operate – Terminals from which distributor obtains motor fuel • Types of violations found • If violation(s) found the following variables are needed – Number of gallons on which assessment was made by type of violation and fuel type – Amount of the assessment minus penalties and interest by type of violation and fuel type – Type of enforcement taken ▪ Civil (reason for considering civil enforcement) ▪ Criminal (reason for considering criminal enforcement) ▪ First, second or third offense or greater • Business characteristics of inspected companies – Years in operation – NAICS code – Annual revenue – Location of company headquarters – Number of employees – Number of taxed gallons – Motor fuel types • Auditor information – Years in service – Rank – Detection rate for the auditor 5.5.6.10 General Audit Data Element #4 Individual general audit data for all audits performed by states taxing at the terminal rack level: • Date audit performed • Type of audit (field audit or office audit) • Trigger for the audit (third party tip, random sampling, flagged return or other . . . specify) • Operational characteristics – Type of operation audited (position holder, importer, alternative fuel producer, terminal operator, vessel op- erator, pipeline operator, train operator) – Is the company licensed to sell exempt fuel? – What motor fuel products do they store, transport or sell? – States in which the company is licensed to operate – Terminals where company operates – Location(s) of terminal(s) where violations are discovered • Types of violations found • If violation(s) found the following variables are needed – Number of gallons on which assessment was made by type of violation and fuel type – Amount of the assessment minus penalties and interest by type of violation and fuel type – Type of enforcement taken ▪ Civil (reason for considering civil enforcement) ▪ Criminal (reason for considering criminal enforcement) ▪ First, second or third offense or greater • Business characteristics of inspected companies (does not include on-road inspections) – Years in operation – NAICS code – Annual revenue – Location of company headquarters – Number of employees – Number of taxed gallons – Motor fuel types • Auditor information – Years in service – Rank – Detection rate for the auditor 5.5.6.11 Illegal Importation Data Element The following data items extracted from individual inspec- tion data are needed for all inspections targeting cross-border evasion: • General Inspection Information – Date inspection performed – Location (county, city) where inspection conducted – Highway number – Location of inspection (e.g., road inspection, site visit, weigh station) • Driver Information – D.L. State • Sample Information – Number of samples taken – Tank location and capacity – Name of fueling location – Terminal code • Business characteristics of inspected companies (does not include on-road inspections) – Years in operation – NAICS code – Annual revenue – Number of employees – Motor fuel types – States in which company is licensed to operate – Location of company headquarters 87

• Trigger for inspection (e.g., tax evasion hint or report by a third party, random sampling, regular on-road inspection, or other . . . specify) • Types of violations found if any • If violation found, the following variables are needed – Type of enforcement taken ▪ Civil (reason for considering civil enforcement) ▪ Criminal (reason for considering criminal enforcement) – First, second or third offense or greater – Number and tax value of gallons on which the assess- ment was based 5.5.6.12 Cocktailing and False Labeling Inspection Data Element The following data items extracted from individual inspec- tion data are needed for all inspections capturing cocktailing and false labeling: • General Inspection Information – Date inspection performed – Location (county, city) where inspection conducted – Location of inspection • Sample Information – Number of samples taken – Name of fueling location – Terminal code • Business characteristics of inspected companies (does not include on-road inspections) – Years in operation – NAICS code – Annual revenue – Number of employees – Motor fuel types – States in which company is licensed to operate – Location of company headquarters • Trigger for inspection (e.g., tax evasion hint or report by a third party, random sampling, regular inspection, or other . . . specify) • Types of violations found if any • If violation found the following variables are needed – Type of enforcement taken ▪ Civil (reason for considering civil enforcement) ▪ Criminal (reason for considering criminal enforcement) – First, second, or third offense or greater – Number and tax value of gallons on which the assess- ment was based 5.5.6.13 Native American Data Element The following data items are needed to estimate evasion as- sociated with Abuses due to the presence of Native American reservations: • Number of Native American reservations in state • Number of retail motor fuel outlets on reservations • Number of pumps located at each Native American retail outlet • Number of enrolled members located on each reservation where retail outlets are located • Location of each retail outlet in relation to population centers and high volume roads and highways • Type of retail operations (truck stop, casino/fueling sta- tions, card lock station, gas station/convenience store) • Data on retail sales of motor fuel by operation type, loca- tion, and number of pumps for non-Native American retail outlets 88