TREASURY INSPECTOR GENERAL FOR TAX ADMINISTRATION

 

 

An Estimated $1.6 Billion in Fraudulent Refunds Was Issued During the 2006 and 2007 Filing Seasons

 

 

 

September 22, 2008

 

Reference Number:  2008-10-172

 

 

This report has cleared the Treasury Inspector General for Tax Administration disclosure review process and information determined to be restricted from public release has been redacted from this document.

Redaction Legend:

1 = Tax Return/Return Information
2(c) = Law Enforcement Tolerance(s)
3(d) = Identifying Information - Other Identifying Information of an Individual or Individuals

Phone Number   |  202-622-6500

Email Address   |  inquiries@tigta.treas.gov

Web Site           |  http://www.tigta.gov

 

September 22, 2008

 

 

MEMORANDUM FOR DEPUTY COMMISSIONER FOR SERVICES AND ENFORCEMENT

 

FROM:                            Michael R. Phillips /s/ Michael R. Phillips

                                         Deputy Inspector General for Audit

 

SUBJECT:                    Final Audit Report – An Estimated $1.6 Billion in Fraudulent Refunds Was Issued During the 2006 and 2007 Filing Seasons (Audit # 200710024)

 

This report presents the results of our review of the Questionable Refund Program.[1]  The overall objectives of this review were to determine the effectiveness of the Internal Revenue Service (IRS) processes to identify and stop questionable refunds during Processing Year (PY) 2007 and to evaluate the effect the failure of the Electronic Fraud Detection System (the fraud detection system) had on PY 2006.  In addition, we analyzed the demographic profiles of taxpayers who submitted potentially fraudulent returns.  This audit was conducted as part of our response to a request made by the House Committee on Ways and Means.

Impact on the Taxpayer

The Criminal Investigation (CI) Division Questionable Refund Program is a nationwide, multi-functional program designed to detect and stop fraudulent claims for refunds on income tax returns.  We estimated that the number of potentially fraudulent returns that would have been identified without threshold restrictions rose by an alarming 70 percent between PYs 2006 and 2007.  Due to resource limitations, the IRS worked only 48.7 percent of these returns, potentially allowing $742 million in fraudulent refunds to be issued.  If this trend continues over the next few years, the IRS might issue an even greater number of fraudulent refunds, possibly resulting in a significant annual revenue loss to the Federal Government.  As a result, additional burden is placed on honest taxpayers whose tax dollars are being used to support this criminal activity.

Synopsis

During the past several years, the CI Division has relied primarily on the fraud detection system to identify fraudulent refunds.  However, because the system was not operational during PY 2006 due to system development problems, the IRS issued millions of dollars in fraudulent refunds.  The system was operational during PY 2007, which enabled the CI Division to identify more than 240,000 fraudulent tax returns with almost $1.5 billion in refund claims.

We estimated that during PY 2006, when the fraud detection system was not operational, the IRS failed to stop approximately $894 million in fraudulent refunds.  This represents an alarming increase from prior years and is substantially more than the IRS initially reported.  The IRS had previously advised the Senate Appropriations Committee Staff that it did not plan to recover the fraudulent refunds paid during PY 2006 because the IRS believed that the total was only about $200 million.[2]

The increase in the number of fraudulent returns continued in PY 2007.  The Fraud Detection Centers (the Centers) initially stopped more than $1.2 billion in refunds during PY 2007.  This is a significant increase over the $412 million in refunds stopped during PY 2005, the last year in which the fraud detection system was operational.  During this same time, the CI Division became more efficient in stopping fraudulent refund claims, mainly because dollar thresholds in place limited the selection of returns to only those with the highest potential for fraud.  While the dollar limits allowed the Centers to verify a higher percentage of fraudulent returns, they excluded thousands of returns from the screening process. 

The exponential growth in fraud in PY 2007 presented a challenge for the IRS, which did not have the resources to handle this volume.  The limited Examination function resources available for processing fraudulent returns directly influenced the number of fraudulent returns that the CI Division was able to identify and stop.  The CI Division and the Examination functions[3] agreed to limit the number of fraudulent return referrals to ensure that examination resources were available to address other areas that are also critical to compliance enforcement.  Therefore, the CI Division focused on identifying those returns with higher dollar values and higher data-mining scores, which precluded more than 500,000 potentially fraudulent returns from entering the Centers’ screening process.  Had these returns been included, we estimated that the Centers would have identified an additional potential $742 million in fraudulent refunds. 

We estimated that the number of potentially fraudulent returns that would have been identified without the dollar value and data-mining score restrictions rose by an alarming 70 percent between PYs 2006 and 2007.  If this trend continues over the next few years, the IRS might issue an even greater number of fraudulent refunds, possibly resulting in a significant annual revenue loss to the Federal Government.  We understand that the IRS has numerous compliance priorities in addition to the Questionable Refund Program, and that it must balance these issues against its limited resources needed to address a number of noncompliance issues.  However, this problem is becoming unmanageable, and the IRS cannot afford to continue handling it in the same manner as in the past.  We believe that the IRS needs to address the use of limited compliance resources corporately–as a top compliance priority–and to proactively determine what it needs to do to counter the growth in questionable refund returns and stop the dollar loss.  This determination should be made based on an understanding of the growth in total refund fraud while considering a balance of available resources and other competing compliance workload.

We also identified some areas in which the CI Division can improve its processing of fraudulent returns.  Although we determined that returns were verified and transferred to the Accounts Management Organization and the Examination functions in a timely manner, a number of returns were incorrectly identified by the CI Division as being not referable.  This put about $15 million in refunds at risk of potentially being issued.  In addition, the CI Division was prevented from sending returns to the Examination functions during ongoing examinations of the taxpayers, which placed even more revenue at risk.  We also determined that while the CI Division made strides in ensuring that it notified taxpayers that their claims for refunds were being held for review, some taxpayers were not issued timely notification that their returns were determined to be fraudulent.  Lastly, the process instituted in PY 2007 for freezing refunds resulted in the inadvertent issuance of approximately $2 million in false refunds.

Recommendations

We recommended that the Deputy Commissioner for Services and Enforcement:

·         Work with the Assistant Secretary for Tax Policy to develop a more urgent approach to achieving the legislative change that will exempt the IRS from having to issue statutory notices of deficiency on fraudulent returns. 

·         Develop a long-term, strategic approach to balancing available resources with the growth in refund fraud and other compliance priorities. 

We recommended that the Chief, CI, 1) coordinate with the Examination functions to review potentially fraudulent ****3(d)**** identified during PY 2006 and pursue recovery or offset through future nonfraudulent refunds, as appropriate, 2) identify all fraudulent PY 2006 tax returns for which PY 2007 returns were verified as being false, to assist in the investigation of refund schemes, 3) implement procedures to ensure that suspicious tax returns filed ****3(d)**** are identified by the fraud detection system during future processing years and coordinate with the Commissioner, Small Business/Self-Employed Division, to ensure that resources are available to verify the accuracy of these returns, 4) establish procedures requiring each Center to use the automated quality review tool to ensure that fraudulent returns are properly classified by the Centers prior to referral, 5) coordinate with the Examination functions to establish a process ensuring that fraudulent returns identified by the CI Division can be referred during open examinations, 6) revise the language contained in the first notice sent to taxpayers to reflect a more realistic time period for the IRS to contact them and ensure that all taxpayers whose returns are identified as fraudulent are notified of the status of their returns, and 7) revise the process for controlling and freezing refund claims identified as potentially fraudulent. 

Response

IRS management agreed with five of our nine recommendations, neither agreed nor disagreed with three recommendations, and disagreed with one recommendation.  Management acknowledged that refund fraud is an increasing concern and believes that employing an enterprise approach will effectively address this compliance threat, while maintaining its commitment to improved taxpayer protection.  In October 2006, the IRS established the Pre-Refund Program Office, which coordinates the activities of all business units that are part of the Questionable Refund Program.  The IRS agreed that a strategic approach to managing the Questionable Refund Program was necessary.  It developed a 5-year plan in February 2008 and formed an Executive Steering Committee comprised of IRS executives providing strategic guidance.  IRS management established procedures requiring each Center 1) to use the automated quality review tool to ensure that returns are properly classified, and 2) to coordinate with the Examination functions to establish a process ensuring that fraudulent returns can be referred during open examinations.  In addition, management agreed to conduct an analysis to determine if the established 3-week time period stated within the first notice needs to be revised.  Finally, management agreed that the process for controlling and freezing refund claims needed improvement and implemented an automated, streamlined process for PY 2008.  

IRS management neither agreed nor disagreed with the following three recommendations:

·         Develop a more urgent approach to achieving a legislative change that will exempt the IRS from having to issue statutory notices of deficiency on fraudulent returns (Recommendation 1). 

·         Identify all fraudulent PY 2006 tax returns filed by taxpayers whose PY 2007 returns were verified as fraudulent (Recommendation 3).

·         Implement procedures to ensure that suspicious returns ****3(d)**** are identified by the fraud detection system in future years (Recommendation 5). 

In each of these instances, IRS management indicated that ongoing efforts and/or existing procedures would address some of the concerns.

IRS management did not agree with Recommendation 2 to coordinate with the Examination functions to review potentially fraudulent ****3(d)**** filed during PY 2006 and pursue recovery or offset through future nonfraudulent refunds.  The IRS stated that attempting to recover these payments would not be a reasonable use of limited examination resources.  Management’s complete response to the draft report is included as Appendix VII.

Office of Audit Comment

While we are encouraged with the IRS’ efforts to identify new ways to meet the growing challenge of refund fraud, we believe that 1) several of the corrective actions to our recommendations will not fully address the concerns we reported and, 2) the recommendations need to be reconsidered as part of the Pre-Refund Program Office’s long-term strategy.  For example, in response to Recommendation 1, the IRS stated that it will continue to evaluate possible legislative proposals but did not indicate whether it will make this a priority.  Further, the example included in the response relates to the IRS disclosing tax violations to Federal Government ****3(d)****, not to our recommended proposal.  We had previously recommended pursuit of a legislative change to the Chief, CI.  According to the Department of the Treasury system used for tracking IRS management’s corrective actions, the CI Division indicated a January 31, 2009, completion date to allow for the time-intensive clearing process for legislative proposals.  However, the IRS response to this report makes no mention of any continuing activity related to this legislative proposal.  As a result, we are uncertain of the IRS’ plans to pursue the legislative change exempting it from having to issue statutory notices of deficiency on fraudulent returns. 

In response to Recommendation 3, the IRS stated that existing procedures already require prior year returns to be evaluated to determine if the returns are also fraudulent and–if so–to include those returns in any enforcement action.  However, we identified 14,000 potentially fraudulent returns filed in PY 2006 that were not included for enforcement action.  At CI Division management’s request, we provided a list of these 14,000 potentially fraudulent returns and asked the Division to advise us if these returns were verified as fraudulent based on its existing procedures.  The CI Division’s response indicated that its initial analysis confirmed our findings.  The PY 2006 returns associated with false PY 2007 returns were not captured in the Division’s Scheme Tracking and Referral System.  The Division indicated that it would conduct some additional research and provide us with the results.  However, we never received any additional evidence to suggest that any of these returns were included in or considered for current enforcement action. 

Regarding Recommendation 5, we remain concerned about the IRS’ reliance on the Dependent Database to identify suspicious returns ****3(d)****.  Our concern is supported by the CI Division’s conclusion from its prior analysis that there is a gap between the returns identified by the CI Division and those identified by the Dependent Database.  If the IRS continues with its current practice of relying only on the Dependent Database, we believe that there is an increased risk that the IRS will not identify all fraudulent refund returns ****3(d)****  Although the IRS plans to conduct additional research in this area in the coming years, we believe that these types of returns should be identified by the fraud detection system. 

Regarding Recommendation 2, we are disappointed in the IRS’ decision to forego taking any action against fraudulent ****3(d)**** claims detected during PY 2006.  IRS management restated its position that fraudulent ****3(d)**** claims continue to be a priority for IRS enforcement and acknowledged that the failure of the fraud detection system allowed ****3(d)**** claims to be inappropriately paid.  Based on available data, we estimated that those claims totaled about $32 million.  When determining the cost and benefits of pursuing recovery or offset in these cases, the IRS should consider the valuable deterrent effect of ****3(d)****realizing that the IRS will catch attempts to file false returns while ****3(d)****  The decision to forego taking any action on these cases only encourages this type of behavior.

Finally, although the IRS indicated that steps have been taken for PY 2008 to ensure that fraudulent refunds are properly held (Recommendation 9), we remain concerned that refunds are still susceptible to erroneous issuance.  During our current audit on the IRS’ efforts to stop economic stimulus payments on false returns identified during PY 2008,[4] we identified several instances in which refunds and stimulus payments were allowed to be issued when one of the freezes being used expired, and we notified both the CI Division and the Wage and Investment Division.  In its response to this report, the IRS stated that the CI Division has taken proactive measures to ensure that procedures will hold the refund until all issues are resolved.  However, based upon our preliminary results of PY 2008, we are concerned that the freezes are not always working as intended.

Copies of this report are also being sent to the IRS managers affected by the report recommendations.  Please contact me at (202) 622-6510 if you have questions or Nancy A. Nakamura, Assistant Inspector General for Audit (Headquarters Operations and Exempt Organizations Programs), at (202) 622-8500.

 

 Table of Contents

 

Background

Results of Review

Approximately $894 Million in Fraudulent Refunds Was Issued During Processing Year 2006

Recommendations 1 and 2:

Recommendation 3:

Internal Revenue Service Decisions Led to About $740 Million in Fraudulent Refunds Being Issued During Processing Year 2007

Recommendations 4 and 5:

The Processing of Returns Identified As Potentially Fraudulent Can Be Improved

Recommendations 6 through 8:

Recommendation 9:

Appendices

Appendix I – Detailed Objectives, Scope, and Methodology

Appendix II – Major Contributors to This Report

Appendix III – Report Distribution List

Appendix IV – Outcome Measures

Appendix V – Demographic Analysis of Filers of Potentially Fraudulent Returns

Appendix VI – Glossary of Terms

Appendix VII – Management’s Response to the Draft Report

  

Abbreviations

 

CI

Criminal Investigation

****3(d)****

****3(d)****

IRS

Internal Revenue Service

PY

Processing Year

QRP

Questionable Refund Program

 

 Background

 

The mission of the Internal Revenue Service (IRS) Criminal Investigation (CI) Division is to serve the American public by investigating potential criminal violations of the Internal Revenue Code and related financial crimes in a manner that fosters confidence in the tax system and compliance with the law.  The CI Division Questionable Refund Program[5] (QRP) is a nationwide, multi-functional program designed to detect and stop fraudulent claims for refunds on income tax returns.  Responsibility for coordinating the QRP resides with the CI Division’s 10 Fraud Detection Centers (the Centers), which are located at each of the 10 IRS campuses where individual tax returns are filed and processed. 

During the past several years, the CI Division has relied primarily on the Electronic Fraud Detection System (the fraud detection system) to identify fraudulent refunds.  However, because the system was not operational during Processing Year (PY) 2006 due to system development problems, the IRS issued millions of dollars in fraudulent refunds.  The system was operational during PY 2007, which enabled the CI Division to identify more than 240,000 fraudulent tax returns with almost $1.5 billion in refund claims.  This represents a significant increase in both the number of fraudulent returns and refund amounts from previous years.  Figure 1 shows the number of fraudulent returns identified by the QRP during PY 2007 and the previous 4 processing years, as well as the refund amounts that were claimed and stopped. 

Figure 1:  Fraudulent Returns and Refunds Identified and Stopped
 by the CI Division (PYs 2003 – 2007)

Processing Year

Number of Fraudulent Refund Returns Identified

Number of Fraudulent Refund Returns Stopped

Amount of Fraudulent Refunds Identified

Amount of Fraudulent Refunds Stopped

2003

96,953

73,400

$349,515,144

$266,423,786

2004

118,075

82,099

$440,773,403

$309,961,554[6]

2005

132,945

103,537

$515,548,186

$412,184,202

2006

52,255

35,523

$271,180,566

$188,715,519

2007

240,406

189,915

$1,467,762,110

$1,203,795,853

Source:  CI Division fraudulent return statistics for PYs 2003 - 2007.

Through the fraud detection system, each individual income tax refund return receives a data-mining score based on the characteristics of the return and other data.  A private vendor developed the data-mining score, which is validated and updated before each processing year based on experience and new information.  The higher the data-mining score, the greater the probability that the return might be fraudulent.  For purposes of this report, the term “potentially fraudulent return” refers to a return identified by the fraud detection system as having the characteristics of fraud.

The Centers screen returns identified by the fraud detection system and determine each return’s fraud potential.  If a return is selected by the Center and included in its workload, the refund is held until employers or third parties are contacted to verify wage information on the return.  If the Centers do not complete the verification process within a certain time period, the held refund is automatically released. 

In general, if the Centers conclude that a return contains false information (e.g., false or inflated wages), the return is referred to either the Accounts Management Organization or the Examination functions[7] for resolution.  Returns with refundable credits, such as the ****3(d)**** and returns for which the refunds were issued must be sent to the Wage and Investment Division Examination function because the law requires issuance of a statutory notice of deficiency whenever an assessment is made to increase the tax on an account.  During PY 2007, the CI Division reported that 53,622 fraudulent tax returns were sent to the Examination functions and 94,654 were sent to the Accounts Management Organization.  Some returns, including those with identity theft issues, are classified as not referable and are resolved by the CI Division.  For purposes of this report, the term “fraudulent return” refers to a return verified by the Centers as containing false information. 

As part of their mission to provide support to the CI Division field offices, the Centers also analyze fraudulent returns to identify schemes.  These are returns with common characteristics and patterns.  Schemes can include only one return but generally include numerous returns.  The Centers refer schemes to the CI Division field offices for further evaluation if the total amount of the refund(s) claimed exceeds a certain dollar amount.  The field offices then review these schemes for potential criminal prosecution and refer the results of their investigations to the United States Attorney’s Offices for prosecution.  However, some United States Attorney’s Offices prefer a high actual dollar loss before they will accept a case for prosecution.  Consequently, as we have previously reported, the majority of refund schemes are not referred for criminal prosecution.[8]

In PY 2006, the Office of Refund Crimes and an IRS Executive Steering Committee[9] worked together to address problems identified in the National Taxpayer Advocate’s 2005 Annual Report to Congress regarding the QRP.  This initiative resulted in changes that included new procedures and controls for freezing tax returns, notifying taxpayers that their refunds were being delayed, releasing some existing freezes, and adjusting taxpayer accounts. 

This audit was a continuation of our Fiscal Year 2007 audit[10] of the QRP.  In that audit, we reported that the growth in refund fraud led to the QRP’s management and inventory problems, but we recognized the quick efforts by the IRS to address the concerns raised by the National Taxpayer Advocate.  However, we raised concerns about some of the changes that had been implemented in PY 2006.  For example, the IRS decided to limit the number of fraudulent returns referred to the Examination function, limit the length of time potentially fraudulent refunds were frozen, and reduce the number of subsequent year freezes placed on accounts.  All of these actions allowed several millions of dollars of potentially fraudulent refunds to be issued.

For the current audit, work was performed at the CI Division Office of Refund Crimes in Washington, D.C., and in the Austin, Texas; Cincinnati, Ohio; and Memphis, Tennessee, Centers during the period January through November 2007.  In addition, we performed work at the Wage and Investment Division Accounts Management Organization in Memphis, Tennessee.  We also contacted personnel within the Wage and Investment Division Pre-Refund Program Office and the Wage and Investment Division and Small Business/Self-Employed Division Examination functions.  We conducted this performance audit in accordance with generally accepted government auditing standards.  Those standards require that we plan and perform the audit to obtain sufficient, appropriate evidence to provide a reasonable basis for our findings and conclusions based on our audit objectives.  We believe that the evidence obtained provides a reasonable basis for our findings and conclusions based on our audit objectives.  Detailed information on our audit objectives, scope, and methodology is presented in Appendix I.  Major contributors to the report are listed in Appendix II.

  

Results of Review

 

We estimated that during PY 2006, when the fraud detection system was not operational, the IRS failed to stop approximately $894 million in fraudulent refunds.  This represents an alarming increase from prior years and is substantially more than the IRS initially reported.  The IRS had previously advised the Senate Appropriations Committee Staff that it did not plan a recovery program for fraudulent refunds paid during PY 2006 because the IRS believed that about $200 million[11] in fraudulent refunds was issued.

The growth in fraudulent returns continued in PY 2007, and the Centers stopped more than $1.2 billion.  This is a significant increase over the $412 million stopped during PY 2005, the last year in which the fraud detection system was operational.  During this same time, the CI Division became more efficient in stopping fraudulent refund claims, mainly because dollar thresholds in place limited the selection of returns to only those with the highest potential for fraud.  While the dollar limits allowed the Centers to verify a higher percentage of fraudulent returns, they excluded thousands of returns from the screening process. 

The exponential growth in fraud in PY 2007 presented a challenge for the IRS, which did not have the resources to handle this volume.  The limited Examination function resources available for processing fraudulent returns directly influenced the number of fraudulent returns that the CI Division was able to identify and stop.  The CI Division and the Examination functions agreed to limit the number of fraudulent return referrals to ensure that examination resources were available to address other areas that are also critical to compliance enforcement.  Therefore, the CI Division focused on identifying those returns with higher dollar values and higher data-mining scores, which precluded more than 500,000 potentially fraudulent returns from entering the Centers’ screening process.  Had these returns been included, we estimated that the Centers would have identified an additional $742 million in fraud.

We estimated that the number of potentially fraudulent returns that would have been identified without the dollar value and data-mining score restrictions rose by an alarming 70 percent between PYs 2006 and 2007.  If this trend continues over the next few years, the IRS might issue an even greater number of fraudulent refunds, possibly resulting in a significant annual revenue loss to the Federal Government.  This problem is becoming unmanageable, and the IRS cannot afford to continue handling it in the same manner as in the past.  We believe that the IRS needs to address the use of limited compliance resources corporately–as a top compliance priority–and to proactively determine what it needs to do to counter the growth in questionable refund returns and stop the dollar loss.

We also identified some areas in which the CI Division can improve its processing of fraudulent returns.  Although we determined that returns were verified and transferred to the Accounts Management Organization and the Examination function in a timely manner, a number of returns were incorrectly identified by the CI Division as being not referable.  This put about $15 million in revenue at risk of being lost.  In addition, the CI Division was prevented from sending returns to the Examination function during ongoing examinations of the taxpayers, which placed even more revenue at risk.  We also determined that while the CI Division has made strides in ensuring that it notified taxpayers that their claims for refunds were being held for review, some taxpayers were not issued timely notification that their returns were determined to be fraudulent.  Lastly, the process instituted in PY 2007 for freezing refunds resulted in the inadvertent issuance of approximately $2 million in fraudulent refunds. 

As part of this audit, we also analyzed the characteristics of the returns that would have been identified as potentially fraudulent during PY 2006.  Appendix V presents those results. 

Approximately $894 Million in Fraudulent Refunds Was Issued During Processing Year 2006

The IRS failed to stop approximately $894 million[12] in fraudulent refunds during PY 2006 when the fraud detection system was not operational.  This amount is significantly higher than the IRS’ July 2006 estimate of about $200 million to $300 million and our August 2006 estimate of $318.3 million (both of these estimates were based upon data available at the time).[13]  The IRS had previously advised the Senate Appropriations Committee Staff that it did not plan to recover the fraudulent refunds paid because the IRS believed at the time that the total was only $200 million.  While we recognize the magnitude of the lost revenue, we believe that IRS efforts to recover the money at this point, using current procedures, would require a resource-intensive commitment.  An attempt to recover this money might become more likely if the tax law is changed to exempt the IRS from having to issue statutory notices of deficiency for fraudulent returns.  To do this, the IRS will need to take a more aggressive approach in pursuing our May 2007 report recommendation that it initiate a legislative proposal to exempt it from having to issue statutory notices of deficiency on returns determined to be fraudulent.[14]

In November 2006, the Office of Refund Crimes processed the PY 2006 individual refund returns through the fraud detection system and identified 593,820 potentially fraudulent returns.  Using this information along with workload estimates and historical fraud detection rates, we estimated that the fraudulent refunds associated with these returns totaled approximately $1.1 billion.  The CI Division stopped approximately $188 million in fraudulent refunds.  Therefore, approximately $894 million in fraudulent refund claims were paid.  Figure 2 shows the estimated amount of fraudulent refunds paid during PY 2006.

Figure 2:  Estimated Fraudulent Refunds Paid During PY 2006

PY 2006 Actual and Our Estimates

Refund Amounts

Total Estimated Amount of Fraudulent Refund Claims

$1,082,991,893

Total Fraudulent Refund Claims Stopped by the CI Division

$188,715,519

Total Estimated Amount of Fraudulent Refund Claims Paid

$894,276,374

Source:  Our analysis of fraud detection system data provided by the CI Division.

The IRS decided not to pursue recovery of the fraudulent refunds paid during PY 2006.  In a December 2006 USA Today[15] article, the former IRS Commissioner stated that there would be little chance the IRS could collect the bulk of the erroneously issued checks because it is extremely difficult to retrieve money once it has been issued.  In addition, Office of Refund Crimes personnel stated that they focused their efforts on moving forward and getting the fraud detection system operational for PY 2007.

Under current procedures, to proceed with a recovery program would require a resource-intensive commitment.  For each of the 593,820 returns identified by the fraud detection system as potentially fraudulent, the Centers would first need to verify the authenticity of the wage and withholding information on the returns with employers or third parties.  Then each return determined to be fraudulent would have to be sent to the Examination function so an assessment could be made to increase the tax on the account.  Once assessments were made, the IRS could recover funds on delinquent accounts through the notice process and by offsetting future refunds.  Our prior audit found that most fraudulent tax returns were not sent to the Examination functions when the refunds had already been issued.  We projected that the IRS could have collected $27.5 million through refund offsets alone on the estimated $81.5 million in fraudulent refunds issued during PY 2004 for which the IRS took no recovery action.[16]

The IRS has indicated that resources within the Examination functions are limited because of the need to adequately address other areas of noncompliance.  This has had a direct impact on its decision not to recover fraudulent refunds paid during PY 2006 and on the number of fraudulent returns identified during PY 2007.  As previously discussed, to make an attempt at recovery possible, a law change is needed to exempt the IRS from having to issue statutory notices of deficiency for fraudulent returns.  Doing so would allow other functions within the IRS to resolve the fraudulent returns more efficiently and with fewer resources than those used by the Examination functions.  In response to our prior recommendation, the CI Division indicated that it planned to work with the IRS Office of Legislative Affairs and the Office of Chief Counsel to discuss initiating a legislative proposal.  However, little progress has been made on this initiative. 

Given our current estimate that about $894 million in fraudulent refunds was paid during PY 2006, we believe that the IRS needs to vigorously pursue this change to the statutory notice of deficiency requirement.  Rescinding the notice requirement in these cases would provide the opportunity for the IRS to address more fraudulent refund returns with its available resources. 

While the IRS is pursuing this change, there are two areas in which it should take immediate action.  Efforts should be made to recover fraudulent refunds paid during PY 2006 to ****3(d)****  Although the Centers identified some ****3(d)**** in PY 2006, for 8,975 returns claiming about $32 million in refunds, we could not determine if the IRS had verified whether the returns were fraudulent and pursued recovery.  Congress has previously voiced concern over the number of ****3(d)**** filing fraudulent tax returns and receiving refunds.  Congress indicated that it was unacceptable for the American public to be victimized by ****3(d)**** 

Similarly, the IRS should verify all potentially fraudulent PY 2006 returns for taxpayers whose PY 2007 returns were verified as being fraudulent.  While the Centers identified some of the PY 2006 returns as fraudulent, for an additional 14,000 returns claiming about $64 million, we could not determine if the IRS had verified whether the returns were fraudulent.  If a taxpayer filed one fraudulent return, it is possible that his or her prior and subsequent years’ returns will be fraudulent as well.  This is especially important because Office of Refund Crimes management had previously advised us of their plans to migrate to refund scheme development, looking for groups of returns with similar patterns and characteristics.  Because the refunds on these 14,000 returns have already been issued, actions to verify whether these returns were fraudulent could help the CI Division build schemes that show an actual dollar loss, making them more likely to be accepted for prosecution by the United States Attorney’s Offices.

At a minimum, we believe that the IRS should verify these highly suspect returns and take proactive steps to recover those refunds deemed to be fraudulent.  The IRS can accomplish this recovery by offsetting any future nonfraudulent refunds claimed by these taxpayers or by attempting to recover the funds directly from the taxpayers through the notice process. 

Recommendations

Recommendation 1:  The Deputy Commissioner for Services and Enforcement should work with the Assistant Secretary for Tax Policy to develop a more urgent approach to achieving the legislative change exempting the IRS from having to issue statutory notices of deficiency on fraudulent returns.  This should include developing milestones to show the progress.

Management’s Response:  The IRS did not agree or disagree with this recommendation.  It stated that working with the Office of the Assistant Secretary for Tax Policy, the IRS plans to continue to evaluate possible legislative proposals that will strike the right balance between leveraging resources and protecting taxpayer rights.  For example, the IRS had previously proposed a legislative provision that would authorize it to disclose certain return information about tax violations by ****3(d)**** so that specified Federal Government ****3(d)**** could punish and deter such conduct through administrative sanctions.  The safeguard provisions, as well as criminal and civil sanctions, would apply.

Office of Audit Comment:  We are concerned that IRS management might not be pursuing the recommended legislative change as a priority.  The response simply states that the IRS will continue to evaluate possible legislative proposals, but it does not indicate whether the IRS will make this a priority.  Further, the example included in the response relates to the IRS disclosing tax violations to Federal Government ****3(d)****not to our recommended proposal.  We had previously recommended pursuit of a legislative change to the Chief, CI.  According to the Department of the Treasury system used for tracking IRS management’s corrective actions, the CI Division indicated a January 31, 2009, completion date to allow for the time-intensive clearance process for legislative proposals.  However, the IRS response to this report makes no mention of any continuing activity related to this legislative proposal.  As a result, we are uncertain of the IRS’ plans to pursue the legislative change exempting it from having to issue statutory notices of deficiency on fraudulent returns. 

The Chief, CI, should:

Recommendation 2:  Coordinate with the Wage and Investment Division and the Small Business/Self-Employed Division Examination functions to review potentially fraudulent ****3(d)**** identified during PY 2006 and pursue recovery or offset through future nonfraudulent refunds, as appropriate.

Management’s Response:  The IRS disagreed with this recommendation, stating that fraudulent ****3(d)**** claims are a priority for IRS enforcement.  The IRS stops all identified fraudulent ****3(d)**** claims, regardless of the amount of the claim.  In addition, returns filed by those ****3(d)**** in subsequent years are automatically identified and evaluated for fraud.  However, to the extent that the failure of the fraud detection system in PY 2006 allowed some ****3(d)**** to be inappropriately paid, the IRS did not agree that it would be a reasonable use of limited examination resources to attempt to recover those payments.

Office of Audit Comment:  We are disappointed in the IRS’ decision to forego taking any action against fraudulent ****3(d)****claims detected during PY 2006.  IRS management restated its position that fraudulent ****3(d)**** claims continue to be a priority for IRS enforcement and acknowledged that the failure of the fraud detection system allowed ****3(d)**** claims to be inappropriately paid.  Based on available data, we estimated that those claims totaled about $32 million.  When determining the cost and benefits of pursuing recovery or offset in these cases, the IRS should consider the valuable deterrent effect of ****3(d)**** realizing that the IRS will catch attempts to file false returns while ****3(d)****  The decision to forego taking any action on these cases only encourages this type of behavior. 

Recommendation 3:  Identify all fraudulent PY 2006 tax returns for which the PY 2007 returns were verified as fraudulent, to assist in the investigation of refund schemes.

Management’s Response:  The IRS did not agree or disagree with this recommendation, stating that existing IRS procedures already require prior year returns to be evaluated to determine if the returns are also fraudulent.  If so, those returns are included in any enforcement action.

Office of Audit Comment:  We remain concerned with the IRS’ indication that its existing procedures would identify prior year returns to determine if the returns were fraudulent.  At CI Division management’s request on May 19, 2008, we provided a list of the 14,000 potentially fraudulent returns filed in PY 2006 and asked the Division to advise us if these returns were verified as fraudulent based on its existing procedures.  The Division responded on May 29, 2008, indicating that its initial analysis confirmed our findings.  The PY 2006 returns associated with false PY 2007 returns were not captured in the Division’s Scheme Tracking and Referral System.  The Division indicated that it would conduct some additional research and provide us with the results.  However, we never received any additional evidence to suggest that any of these returns were included in or considered for current enforcement action. 

Internal Revenue Service Decisions Led to About $740 Million in Fraudulent Refunds Being Issued During Processing Year 2007

The CI Division was more successful in stopping fraudulent refund claims during PY 2007 than in prior years.  It reported stopping more than $1.2 billion during PY 2007.  This is a 192 percent increase from PY 2005, the last year in which the fraud detection system was operational.  The increase could be largely attributable to the higher data-mining scores and dollar thresholds in place that limited the selection of returns to only those with the highest potential for fraud. 

However, use of higher data-mining scores and dollar thresholds–and other decisions made by the IRS prior to and during PY 2007–excluded a great number of returns from the screening process.  We determined that additional returns were scored as potentially fraudulent but were not referred to the Centers for evaluation.  This decision resulted in issuance of an estimated $742 million in fraudulent refunds.[17]  These returns were excluded primarily because of the limited Examination function resources available, which caused the IRS to implement refund dollar thresholds designed to limit the number of fraudulent returns identified during the year.  The CI Division also excluded returns ****3(d)**** from its screening process when it mistakenly relied on another system to identify these returns, and that system did not identify the returns. 

We understand that the IRS has numerous compliance priorities in addition to the QRP, and that it must balance these issues against its limited resources needed to address a number of noncompliance issues.  However, as discussed previously, the amount of refund fraud committed each year is growing exponentially, and if changes are not made, an even greater amount of fraudulent refunds will be issued in the coming years.  This problem is becoming unmanageable.  Unlike the past, the IRS needs to address the use of limited resources corporately–as a top IRS compliance priority–and to proactively determine what it needs to do to counter the growth in questionable refund returns and protect the revenue.  This determination should be made based on an understanding of the growth in total refund fraud while considering a balance of available resources and other competing compliance workload.

Nearly one-half of the fraudulent returns are not being addressed

During PY 2007, the CI Division excluded from its screening process an estimated 500,000 returns that would have resulted in the identification of an estimated $742 million in fraudulent refunds.[18]  These returns were excluded because of limited Examination function resources and the CI Division’s decision to rely on another system to identify returns ****3(d)****  These decisions were short-term fixes to respond to resource needs rather than part of a long-term strategy to curb the growth of refund fraud and its effect on tax administration.

To illustrate, Figures 3 and 4 show what the CI Division identified as potentially fraudulent using the fraud detection system and what the Division did not address because of decisions to exclude returns from screening.  While the fraud detection system identified approximately $3 billion in potential fraud, another $2.2 billion in refunds was excluded from processing because of decisions to raise the data-mining scores and refund dollar threshold. 

Figure 3:  PY 2007 Fraudulent Refunds

Total Fraud Identified by the Fraud Detection System

(No Tolerances)

 

 

Worked by the CI Division

(Actual)

Not Worked by the CI Division

(Estimated)

Totals

(Estimated)

Fraud Detection System’s Potential Fraud

$3,003,536,195

$2,152,607,185

$5,156,143,380

Actual and Estimated Fraud

$1,467,762,110

$742,063,173

$2,209,825,283

Initially Identified and Stopped

$1,203,795,853

$0

$1,203,795,853

Estimated Amount Released

$263,966,257[19]

$742,063,173

$1,006,029,430

Amount Stopped and Subsequently Released

$22,610,962

$0

$22,610,962

Total Released

$286,577,219

$742,063,173

$1,028,640,392

Source:  The CI Division and our analysis of data provided by SRA International, Inc.

Although the CI Division initially stopped $1.2 billion dollars, it subsequently released more than $20 million in refund claims it had already identified as being fraudulent because the Examination functions did not have the resources to process them.  Additionally, as discussed later in this report, more than $2 million in fraudulent refund claims were erroneously released due to processing errors.  As a result, we estimated that a total of $1.03 billion ($264 million plus $742 million plus $23 million) in fraudulent refunds was released in PY 2007. 

Figure 4 shows the number of returns excluded, which meant that the CI Division worked only 48.7 percent of the volume of potentially fraudulent returns.[20] 

Figure 4:  PY 2007 Fraudulent Returns

 

Total Returns Identified by the Fraud Detection System (No Tolerances)

 

 

Worked by the CI Division

(Actual)

Not Worked by the CI Division

(Estimated)

Totals

(Estimated)

Fraud Detection System’s Potential Fraud

490,921

517,914

1,008,835

Actual and Estimated Fraud

240,406

196,802

437,208

Initially Identified and Stopped

189,915

0

189,915

Estimated Released

50,491

196,802

247,293

Source:  The CI Division and our analysis of data provided by SRA International, Inc.

To address workload concerns, the IRS used a dollar threshold (for refund amount) during PY 2006 to limit the number of returns sent to the Examination functions for resolution.  This pattern continued in PY 2007 with the use of a dollar threshold and higher data-mining scores to select returns.  Furthermore, the IRS had to raise the dollar threshold even higher because of the large volumes of fraudulent refund returns the CI Division was identifying in PY 2007.  The impact of raising the dollar threshold in the middle of the filing season was significant.  In addition to excluding returns from the initial screening process, the CI Division had to release more than $20 million[21] in refund claims it had already identified and stopped as fraudulent primarily because the Examination functions did not have the resources to process them through to completion.

The IRS has indicated that resources available for processing fraudulent returns within the Examination functions are limited because of the need to adequately address other areas of noncompliance.  The use of thresholds reduces not only the number of returns sent to the Examination functions but also the number sent to the Accounts Management Organization. 

Additionally, IRS management decided not to have the fraud detection system provide the Centers with suspicious tax returns ****3(d)**** during PY 2007 because the Centers cannot verify income on these types of returns.[22]  Instead, the IRS relied on the Dependent Database, even after we had raised concerns about this in our last report[23] and it had preliminary data to show that this process was not effective.  We estimated that this decision cost the IRS about $27 million[24] in fraudulent refunds because the Dependent Database would not have identified these returns.

We expect that the number of fraudulent refund returns filed will continue to increase in the coming years.  However, if the IRS follows the same approach of excluding returns from identification to fit its current resource availability, countless more fraudulent refunds will be issued.  Therefore, we believe that the CI Division should consider the potential total population of fraudulent returns to assist the IRS in making future decisions about the QRP and not exclude returns due to dollar tolerances and resource constraints.  Figure 5 shows 1) the growth in the amount of fraudulent refunds identified during PYs 2003 through 2005, and 2) the estimated amount of refunds that would have been identified during PYs 2006 and 2007 had the IRS not implemented refund dollar thresholds or other criteria to generate workload based upon available resources. 

Figure 5:  Growth in Fraud[25] During PYs 2003 2007

Figure 5 was removed due to its size.  To see Figure 5, please go to the Adobe PDF version of the report on the TIGTA Public Web Page.

 

If the trend of filing fraudulent refund returns continues, the IRS will increasingly fall behind in its responsibility to combat this criminal activity.  If the IRS is unable to deploy additional resources or does not succeed with a legislative remedy to address the statutory notice of deficiency process, an even greater amount of fraudulent refunds will be issued.  Given the dollar amount of refunds released during PY 2007 and the risk for the future, we cannot overemphasize the need for the IRS to develop a long-term strategy that will allow it to address the problem from a proactive, rather than reactive, stance.  If more effective actions are not taken, this criminal behavior will persist.

Recommendations

Recommendation 4:  The Deputy Commissioner for Services and Enforcement should develop a long-term, strategic approach to balance available resources with the growth in refund fraud and other competing compliance priorities.  This should include gathering data to estimate the total number and dollar amount of potentially fraudulent refund returns, including returns not being reviewed due to dollar thresholds and other criteria.

Management’s Response:  The IRS agreed with this recommendation.  It established a Pre-Refund Program Office with a lead executive in October 2006.  The Office is responsible for developing an enterprise vision and strategy for all IRS pre-refund activities, identifying and implementing business process improvements to achieve the vision, and ensuring full participation of all affected IRS business areas and stakeholders.  The first step in meeting this goal was the development of a Pre-Refund Concept of Operations document (a 5-year plan) in February 2008.

The Pre-Refund Program Office ensures the participation of internal and external stakeholders in decisions about pre-refund programs and activities.  Decisions incorporate the input, engagement, and buy-in of affected stakeholders.  This includes a thorough analysis of pre-refund programs and activities, including workload and resource availability.  The Pre-Refund Program Executive Steering Committee, comprised of executives throughout the IRS, provides strategic guidance and decision making on cross-functional issues, including those relating to the QRP.  The Executive Steering Committee meets monthly to provide executive leadership and oversight of cross-functional tactical operational issues.

Recommendation 5:  The Chief, CI, should implement procedures to ensure that suspicious tax returns filed ****3(d)**** are identified by the fraud detection system during future processing years and coordinate with the Commissioner, Small Business/Self-Employed Division, to ensure that resources are available to verify the accuracy of these returns.

Management’s Response:  The IRS did not agree or disagree with this recommendation.  The IRS indicated that it has an extensive research project–the National Research Program–which is currently evaluating compliance in individual income tax returns (including those ****3(d)**** for Tax Years 2006 and 2007, and it will continue this research for subsequent years.

Office of Audit Comment:  We remain concerned about the IRS’ reliance on the Dependent Database to identify these types of returns.  The Dependent Database is a program designed to identify noncompliant ****3(d)**** and dependent issues and to provide the ability to freeze refunds.  The criteria are based on a limited number of factors and are applied on individual returns.  However, in our opinion, the criteria do not address the many variables and patterns of multiple fraudulent returns found in refund schemes.  Our concern is supported by the CI Division’s conclusion from its prior analysis that there is a gap between the returns identified by the CI Division and those identified by the Dependent Database.  If the IRS continues with its current practice of relying only on the Dependent Database, we believe that there is an increased risk that the IRS will not identify all fraudulent refund returns ****3(d)****  Although the IRS plans to conduct additional research in this area in the coming years, we believe that these types of returns should be identified by the fraud detection system. 

The Processing of Returns Identified As Potentially Fraudulent Can Be Improved

Our analysis of a sample of cases processed in PY 2007 determined that tax returns were being verified as fraudulent and transferred in a timely manner to the Accounts Management Organization and Examination functions for resolution.  However, several processing errors allowed millions of dollars of fraudulent refunds to be issued and increased taxpayer burden. 

Some tax returns were incorrectly identified as not referable for civil resolution

In some instances, the Centers incorrectly identified returns as not referable when they should have been sent to the Accounts Management Organization or the Examination functions for resolution.[26]  This occurred because the manual quality review process, when used, was ineffective[27] and prevented the tax returns from being properly resolved by the civil functions.  During this audit, we requested that Office of Refund Crimes personnel use an automated quality review tool, developed during PY 2007, to identify additional instances of tax returns that were incorrectly identified.  This analysis identified ****2(c)**** with almost $15 million[28] in refund claims as being incorrectly labeled as not referable for civil resolution.  The IRS can now properly resolve these returns. 

Fraudulent returns filed by taxpayers under current examination could not be referred, and refunds were eventually issued

Although the Centers determined that wages reported on some returns were false, the procedures in place at the start of PY 2007 prohibited them from referring this information to the Examination functions during ongoing examinations on the same tax returns.  The CI Division indicated that this affected about 6,000 returns that it had determined were fraudulent in PY 2007.  During the year, the CI Division worked with the Wage and Investment Division Examination function to identify a solution for incorporating the amounts associated with the fraud determinations into the ongoing examinations.  However, in some instances, the CI Division decided that the amounts it had determined were fraudulent should be released if they were below a certain dollar amount.  When asked about the outcome of these cases, the CI Division could not provide us with what, if any, actions had been taken by the Wage and Investment Division Examination function to incorporate this information into its examinations.  However, the CI Division indicated that the dollar amount of fraudulent refund claims being released could reach $217,000.

The solution to incorporate the amounts associated with fraud determinations into ongoing examinations was designed to resolve this type of situation during PY 2007 only.  At the time of our audit, a procedure had not yet been agreed upon for PY 2008 and subsequent processing years.  We are concerned that in future years the portion of the refunds identified by the Centers as being false will not be included in ongoing examinations and will need to be released. 

Some taxpayers are still not receiving timely notification that their returns were determined to be fraudulent

The IRS implemented new procedures during PYs 2006 and 2007 to ensure that taxpayers were notified that their returns 1) had been stopped for review (first notice)[29] and 2) had been subsequently identified as being false (second notice).[30]  The first notice indicates that if the taxpayer is not contacted within 3 weeks, he or she may contact the IRS to inquire about the refund claim.  We reviewed a sample of 100 tax returns identified by the Centers as being false and determined that taxpayers were sent the first notice that their returns were being held for review. 

However, 48 taxpayers in our sample whose returns were sent to the Examination functions for resolution were not sent the second notice indicating that their returns had been identified as being false.  According to CI Division procedures, these taxpayers would not receive the second notice.  Instead, the CI Division was relying on the Examination functions to notify the taxpayers through their initial contact letters.  However, we determined that these taxpayers were often not notified by the Examination functions in a timely manner.  For these 48 taxpayers, the next notification they received from the IRS involved the following:

·         41 taxpayers were sent initial contact letters by the Examination functions an average of more than 3 months after the initial notices were sent (range was more than 1 month to nearly 6 months.)

·         4 taxpayers had yet to be contacted by the Examination functions at the time we were concluding our analysis an average of more than 7 months after issuance of the first notices.

·         3 taxpayers had to inquire about the status of their refunds and were sent second notices an average of 3 weeks after issuance of the first notices. 

In addition, second notices sent by the IRS were rarely sent within the 3-week time period noted on the first notice.  These situations create unnecessary taxpayer burden for honest taxpayers, who either remain without knowledge of their refunds’ status or need to inquire about them. 

The process for freezing taxpayer accounts was complex and confusing, which resulted in the erroneous issuance of fraudulent refunds

In the past, the CI Division froze a taxpayer’s fraudulent return and the subsequent years’ returns until they could be resolved.  Because of criticism from the National Taxpayer Advocate, the IRS revised its process for freezing refunds during PY 2006.  This involved a series of steps to place and remove several freezes on a taxpayer’s account that required coordination by participating functions.  These steps increased the risk that refunds could be released erroneously.  Through our analysis of Master File account information, we identified $2.1 million in erroneously issued refunds.[31]  The erroneous issuance of refunds fell into two main categories:  those issued when one of the freezes being used expired without the knowledge of IRS personnel and those issued after CI Division personnel manually released the freezes.  In the latter, it appears that IRS personnel took actions intended to release only one of the freezes applied to the account.  However, those actions released all freezes and caused issuance of the fraudulent refunds.  The IRS needs to revise its current process for controlling and freezing refunds to one that is not susceptible to erroneous issuance of fraudulent refunds. 

Recommendations

The Chief, CI, should:

Recommendation 6:  Establish procedures requiring each Fraud Detection Center to use the automated quality review tool to ensure that fraudulent returns are properly classified by the Centers prior to referral.

Management’s Response:  The IRS agreed with this recommendation, stating that it has already established these procedures for PY 2008.  The automated quality review tool was updated for the 2008 Filing Season and is being successfully used by all of the Centers.  Specialized training was conducted on February 19, 2008, and the tool was rolled out on February 26, 2008.  Both the criminal investigation and civil processes contain certain validations and data checks to ensure that returns are properly classified.  

Recommendation 7:  Coordinate with the Examination functions to establish a process ensuring that fraudulent returns identified by the CI Division can be referred during open examinations to properly address all issues.

Management’s Response:  The IRS agreed with this recommendation, stating that it has already established and implemented these processes for PY 2008.

Recommendation 8:  Revise the language contained in the first notice sent to taxpayers to reflect a more realistic time period for contacting taxpayers and ensure that all taxpayers whose returns are identified as fraudulent are notified of the status of their returns by issuance of the second notice or a new notice.

Management’s Response:  The IRS agreed with this recommendation.  Automation of the referral process for PY 2008 streamlined the generation and transfer of referral information to the Examination functions and the Accounts Management Organization.  As a result of the PY 2008 automation and procedural changes, the CI Division plans to conduct an analysis to determine if the established 3-week time period stated within the first notice needs to be revised.  The CI Division plans to present the results and proposed recommendations to the Pre-Refund Program Executive Steering Committee for any necessary action.

Recommendation 9:  Revise the process for controlling and freezing refund claims identified as fraudulent to ensure that those refunds are not susceptible to erroneous issuance.

Management’s Response:  The IRS agreed with this recommendation, stating that it has already implemented this process for PY 2008.  The prior, manually intensive process was replaced for PY 2008 with an automated, streamlined process that negates the need for manual intervention in placing controls on a return once it has been verified as fraudulent.  Other freeze conditions are used to hold the refund while the automated referral process is completed.  The CI Division has taken proactive measures to ensure that procedures are in place to hold the refund until all issues are resolved.

Office of Audit Comment:  Although the IRS indicated that steps have been taken for PY 2008 to ensure that fraudulent refunds are properly held, we remain concerned that refunds are still susceptible to erroneous issuance.  During our current audit of the IRS’ efforts to stop economic stimulus payments on false returns identified during PY 2008,[32] we identified several instances in which refunds and stimulus payments were allowed to be issued when one of the freezes being used expired, and we notified both the CI Division and the Wage and Investment Division.  In its response to this report, the IRS stated that the CI Division has taken proactive measures to ensure that procedures will hold the refund until all issues are resolved.  However, based upon our preliminary results for PY 2008, we are concerned that the freezes are not always working as intended.  

 

Appendix I

 

Detailed Objectives, Scope, and Methodology

 

The overall objectives of this review were to determine the effectiveness of the IRS processes to identify and stop questionable refunds during PY 2007[33] and to evaluate the effect the failure of the Electronic Fraud Detection System (the fraud detection system) had on PY 2006.  In addition, we analyzed the demographic profiles of taxpayers who submitted potentially fraudulent returns.[34]  Through assessment of the electronic data sources used in this audit, we concluded that the data were of undetermined reliability.  However, answering the audit’s objectives would not be feasible if the data were not used, and it was our opinion that using the data would not weaken the analysis or lead to an incorrect or unintentional message.  Additional steps to determine data reliability prior to testing were not feasible.  For samples selected, the electronic data were validated to the Master File.  To accomplish our objectives, we:

I.                   Developed an understanding of the data-mining methodology, procedures, and processes and how prior year results affected future development by interviewing appropriate IRS personnel and reviewing applicable documentation.  Additionally, we obtained assistance from an independent contractor that provided an assessment of the validity of the data-mining scoring methodology as it pertained to ****3(d)**** returns.

II.                Determined the dollar amount of refund claims that would have been identified as fraudulent during PY 2006 when the fraud detection system was not operational. 

A.    Interviewed CI Division and contractor personnel to determine the specific data-mining and refund tolerance criteria in effect during PY 2006. 

B.     Obtained from the fraud detection system a data extract containing the universe of 593,820 tax returns that would have been identified as potentially fraudulent had the system been operational during PY 2006.  We determined the dollar amount of the returns identified by each data-mining submodel.  Using tolerance and workload criteria, hit-to-scan ratios, and detection rates, we estimated the dollar amount of refund claims that would have been identified as fraudulent during PY 2006.

C.     Determined the potential effect of the IRS’ decision to not pursue recovery of the fraudulent refunds issued during PY 2006.

1.         Obtained from the Scheme Tracking and Referral System a data extract of about 52,000 returns determined to be fraudulent by the CI Division during PY 2006.  We compared this file to our estimate of refund claims that would have been identified as fraudulent during PY 2006 to determine the amount of fraudulent refunds not identified by the CI Division.

2.         From the fraud detection system’s extract of 593,820 tax returns, identified about 15,000 returns filed by ****3(d)****  We compared these 15,000 returns to the file of 52,000 returns identified as fraudulent during PY 2006 to determine the amount of refund claims filed by ****3(d)**** that were not identified as fraudulent. 

3.         Obtained from the Scheme Tracking and Referral System a data extract containing about 72,000 returns determined to be fraudulent by the CI Division during the first quarter of PY 2007.  We compared this file to the 593,820 returns that would have been identified as potentially fraudulent during PY 2006 and identified more than 17,300 taxpayers who filed a potentially fraudulent return during PY 2006.  We subsequently compared these 17,300 returns to the data extract of about 52,000 returns determined to be fraudulent by the CI Division during PY 2006 to identify the number verified as fraudulent.

D.    Analyzed the fraud detection system’s computer file containing the universe of 593,820 tax returns that would have been identified as potentially fraudulent during PY 2006 to determine the demographics and other characteristics of the taxpayers who filed these returns. 

III.                        Evaluated the effectiveness of the case processing actions made by the Centers during PY 2007.

A.    Identified and reviewed the tolerance criteria and pertinent policies and procedures used by the CI Division in processing potentially fraudulent returns.

B.     Obtained from the data-mining contractor an analysis of the number of potentially fraudulent returns screened based on the Centers’ actual tolerances, the actual number of returns verified as fraudulent, the refunds claimed, and related hit-to-scan ratios and detection rates.  We also obtained the number of potentially fraudulent returns that would have been screened if lower tolerances had been used.  We applied the actual hit-to-scan ratios and detection rates to the potential number of fraudulent returns to estimate the potential fraud that went undetected due to the lack of resources and the higher dollar thresholds. 

C.     Obtained from the CI Division data files containing 16,730 tax returns that had been identified as fraudulent and transferred to IRS civil functions between January 17, 2007, and March 16, 2007.  We selected a judgmental sample of 100 of the 11,023 returns processed by the Atlanta, Georgia; Cincinnati, Ohio; Kansas City, Missouri; and Memphis, Tennessee, Centers.  We used judgmental sampling because our sample was selected early in PY 2007 when the number of fraudulent returns was limited and continuously changing.  We reviewed Master File transcript information for these returns to determine whether:

1.      The Centers’ case actions were taken in a timely manner.

2.      Controls over the referral process, including appropriate input and release of freeze codes, were effective.

D.    Identified a total of 4,270 tax returns classified as not referable to the civil functions by the Atlanta, Cincinnati, Kansas City, and Memphis Centers from the Scheme Tracking and Referral System computer file of about 72,000 returns determined to be fraudulent as of the first quarter of PY 2007.  We selected a judgmental sample of 50 to verify whether the returns were correctly classified.  We used judgmental sampling because our sample was selected early in PY 2007 when the number of fraudulent returns was limited and continuously changing.

E.     Obtained from the CI Division the results of its study of tax returns filed ****3(d)**** during PY 2007 that were eliminated from the fraud detection system screening process.

F.      Obtained electronic Master File information for the approximately 72,000 tax returns identified as fraudulent as of the first quarter of PY 2007 from the Treasury Inspector General for Tax Administration Data Center Warehouse.  This information was used to identify additional refund claims that were inadvertently issued by the CI Division.

 

Appendix II

 

Major Contributors to This Report

 

Nancy A. Nakamura, Assistant Inspector General for Audit (Headquarters Operations and Exempt Organizations Programs)

Carl Aley, Director

Jeffrey M. Jones, Director

John Wright, Director

Diana Tengesdal, Audit Manager

Timothy A. Chriest, Acting Audit Manager

Jeff K. Jones, Lead Auditor

Thomas M. Dori, Senior Auditor

Michael J. Hillenbrand, Senior Auditor

Mark A. Judson, Senior Auditor

Janice A. Murphy, Senior Auditor

Jeffrey E. Williams, Assistant Director, Data Warehousing

 

Appendix III

 

Report Distribution List

 

Commissioner  C

Office of the Commissioner – Attn:  Chief of Staff  C

Commissioner, Small Business/Self-Employed Division  SE:S

Commissioner, Wage and Investment Division  SE:W

Chief Counsel  CC

National Taxpayer Advocate  TA

Chief, Criminal Investigation  SE:CI

Director, Taxpayer Account Operations, Taxpayer Advocate Service  TA:TAO

Director, Campus Compliance Services, Small Business/Self-Employed Division  SE:S:CCS

Director, Compliance, Wage and Investment Division  SE:W:CP

Director, Operations Policy and Support, Criminal Investigation  SE:CI:OPS

Director, Refund Crimes, Criminal Investigation  SE:CI:RC

Director, Office of Legislative Affairs  CL:LA

Director, Office of Program Evaluation and Risk Analysis  RAS:O

Office of Internal Control  OS:CFO:CPIC:IC

Audit Liaisons:

            Commissioner, Small Business/Self-Employed Division  SE:COM

            Commissioner, Wage and Investment Division  SE:W

            National Taxpayer Advocate  TA

            Chief, Criminal Investigation  SE:CI:S:PS

 

Appendix IV

 

Outcome Measures

 

This appendix presents detailed information on the measurable impact that our recommended corrective actions will have on tax administration.  These benefits will be incorporated into our Semiannual Report to Congress.

Type and Value of Outcome Measure:

·         Revenue Protection – Potential; $575,976,374 in additional fraudulent refunds lost during PY 2006 because the Electronic Fraud Detection System (the fraud detection system)[35] was not operational (see page 5).

Methodology Used to Measure the Reported Benefit:

In November 2006, the Office of Refund Crimes processed the PY 2006 refund returns through the fraud detection system and identified 593,820 potentially fraudulent returns that met the criteria that would have been used during PY 2006 had the system been operational.  Using this information along with workload estimates and historical fraud detection rates, we estimated that the fraudulent refunds associated with these returns totaled $1,082,991,893.  The CI Division stopped $188,715,519 in fraudulent refund claims.  Therefore, $894,276,374 in refunds was paid.

We had previously reported that approximately $318.3 million in fraudulent refunds might have been issued.[36]  Based on our current analysis, we estimated that the IRS lost an additional $575,976,374 in fraudulent refunds because the fraud detection system was not operational during PY 2006. 

Type and Value of Outcome Measure:

·         Revenue Protection – Potential; $742,063,173 in fraudulent refunds lost during PY 2007 due to decisions made by the IRS to exclude certain tax returns from the fraud detection system identification process (see page 9).

Methodology Used to Measure the Reported Benefit:

In response to the ever-growing output from the QRP, the IRS decided to exclude certain tax returns from the fraud detection system’s identification process.  It decided 1) to select tax returns using a dollar threshold and high data-mining scores because of limited resources within the Examination functions, and 2) to program the fraud detection system to omit suspicious tax returns ****3(d)**** during PY 2007.  Had these decisions not been implemented, the CI Division would have identified an additional 196,802 returns as fraudulent during the year.  With these additional tax returns, we estimated that $742,063,173 in refund claims would have been identified as fraudulent.

Type and Value of Outcome Measure:

·         Revenue Protection – Actual; $20,469,941 in fraudulent refunds released when the IRS decided to increase the refund dollar threshold during PY 2007 (see page 9).

Methodology Used to Measure the Reported Benefit:

During PY 2007, the IRS raised the dollar threshold based on the unexpectedly high volume of fraudulent refund returns the CI Division had identified.  The impact of raising the dollar threshold in the middle of the 2008 Filing Season was significant.  In addition to excluding returns from the initial screening process, the CI Division’s reports showed it released $20,469,941 in refund claims for 6,085 tax returns it had already identified as fraudulent because the Examination functions did not have the resources to process them. 

Type and Value of Outcome Measure:

·         Revenue Protection – Actual; $14,915,744 in fraudulent refunds misclassified by the CI Division as not referable (see page 15).

Methodology Used to Measure the Reported Benefit:

The Fraud Detection Centers incorrectly identified as not referable some returns that should have been sent to the civil functions for resolution, which prevented these functions from properly resolving the tax returns.  Subsequently, we requested that Office of Refund Crimes personnel use a recently developed automated quality review tool to identify additional tax returns that were incorrectly identified.  Based on this analysis, ****2(c)**** with $14,915,744 in refund claims were identified as being incorrectly labeled as not referable for civil resolution.  These can now be properly resolved by the IRS.

Type and Value of Outcome Measure:

·         Revenue Protection – Actual; $2,141,021 in fraudulent refunds lost during PY 2007 due to the CI Division’s process for freezing taxpayer accounts (see page 15).

Methodology Used to Measure the Reported Benefit:

The CI Division’s process for freezing tax returns led to issuance of fraudulent refund claims that had previously been stopped.  In some instances, the CI Division used a freeze that expired without the knowledge of IRS personnel.  Use of this freeze led to erroneous issuance of refunds that had been previously identified as fraudulent.  Our review of Master File information identified 10,979 fraudulent returns on which this temporary freeze was input.  Erroneous refunds for 397 of these returns were issued when the freezes were released after a programmed number of days.  This caused the issuance of $2,074,086 in fraudulent refund claims.

We also identified accounts on which freezes were apparently inadvertently released by IRS personnel, causing issuance of refunds.  Our review of Master File information identified issuance of 12 erroneous refunds totaling $66,935.  We believe that IRS personnel took actions intended to manually release only one of the freezes applied to each of these accounts.  However, these actions resulted in the release of all freezes placed on the accounts and issuance of the refunds. 

 

Appendix V

 

Demographic Analysis of Filers of Potentially Fraudulent Returns

 

The National Taxpayer Advocate’s 2005 Annual Report to Congress expressed concerns that the IRS was targeting lower income taxpayers with its QRP[37] case selection process.  The Report described the characteristics of those taxpayers who came to the Taxpayer Advocate Service for assistance because their refunds had been frozen by the CI Division.  The National Taxpayer Advocate concluded that most of the taxpayers reviewed by the Taxpayer Advocate Service were in the lower income strata, with approximately 75 percent of the taxpayers ****3(d)**** Congress reiterated the National Taxpayer Advocate’s concerns about targeting taxpayers ****3(d)**** and recommended that the QRP be suspended until an assessment could be made. 

We provided information on the Electronic Fraud Detection System (the fraud detection system) data-mining models to an independent statistician, asking whether the models unfairly targeted lower income taxpayers.  The statistician indicated that if the data-mining models were unfairly targeting returns ****3(d)**** by setting a less stringent review criterion than that set for returns ****3(d)**** then the rate of fraud detection would be much lower for the returns ****3(d)****  However, the overall fraud detection rates for ****3(d)**** taxpayers were comparable, suggesting that the data-mining models are selecting those returns with a higher likelihood for fraud and do not appear to target returns ****3(d)****

We also analyzed return characteristics for the 593,820 returns filed during PY 2006 that the fraud detection system identified as potentially fraudulent.  Our review revealed the following:

  • Nearly ****3(d)**** with an average refund claim of ****3(d)****  The average adjusted gross income reported for these returns was ****3(d)****
  • Approximately ****3(d)**** returns were filed by ****3(d)**** These returns claimed almost ****3(d)**** in refunds, with an average of ****2(c); 3(d)**** per return.  Of these ****3(d)****returns, more than ****3(d)****, ****1, 3(d)**** Therefore, these approximately ****3(d)**** returns are most likely fraudulent. 
  • More than ****3(d)**** tax returns were filed ****3(d)****attached.  These returns claimed more than ****3(d)**** in refunds, ****2(c); 3(d)****.  Of these ****3(d)**** returns, more than ****3(d)****did not report wages.  Because these approximately ****3(d)**** returns would not have wages to verify, they would need to be sent to the Small Business/Self-Employed Division Examination function for a civil audit.

Table 1 compares the characteristics of the potentially fraudulent returns ****3(d)**** with those ****3(d)****

Table 1:  ****3(d)**** Compared to ****3(d)****

Characteristics

****3(d)****

****3(d)****

Number of Returns

****2(c); 3(d)****

****2(c); 3(d)****

Percentage of Returns

****2(c); 3(d)****

****2(c); 3(d)****

Total Refunds Claimed

****2(c); 3(d)****

****2(c); 3(d)****

Average Refund Claimed

****2(c); 3(d)****

****2(c); 3(d)****

Average Reported Adjusted Gross Income

****2(c); 3(d)****

****2(c); 3(d)****

Refund Percentage of Adjusted Gross Income

****2(c); 3(d)****

****2(c); 3(d)****

Filing Status – ****3(d)****

****2(c); 3(d)****

****2(c); 3(d)****

Source:  Our analysis of information provided by the CI Division for returns filed during PY 2006
that were identified as potentially fraudulent by the fraud detection system.

 

Appendix VI

 

Glossary of Terms

 

Accounts Management Organization – The organization within the Wage and Investment Division responsible for taxpayer relations by answering tax law/account inquiries and adjusting tax accounts.  In addition, it is responsible for providing taxpayers with information on the status of their returns/refunds and for resolving the majority of issues and questions to settle their accounts.

Adjusted Gross Income – Income amount calculated after certain adjustments are made but before standard or itemized deductions and personal exemptions are subtracted. 

Campus – The data processing arm of the IRS.  The campuses process paper and electronic submissions, correct errors, and forward data to the Computing Centers for analysis and posting to taxpayer accounts.

Data Center Warehouse – A centralized storage and administration of files that provides data and data access services of IRS data.

Data Mining – The process of automatically searching large volumes of data for patterns.

Dependent Database – A system that identifies and selects for examination taxpayer returns with possible erroneous ****3(d)****  During initial processing, the Dependent Database Scoring Program analyzes tax returns that have claimed ****3(d)****  Using data from several sources, it analyzes each tax return for criteria indicating that the taxpayer might not be eligible for ****3(d)**** and assigns a numeric value to each criterion.  The Dependent Database then produces an overall score for the return.  Based on resources available to conduct examinations, the IRS selects certain types and quantities of returns for pre-refund examinations to verify the taxpayers’ eligibility for ****3(d)****

Detection Rate – The number of fraudulent returns meeting the CI Division’s designated data-mining scores and refund dollar tolerances divided by all fraudulent returns identified.

Electronic Fraud Detection System – A computer system, developed in 1994 and implemented in 1996, that automates the identification output for potentially fraudulent electronically filed tax returns, increases data available for analysis, and assists in the development of information relating to paper and electronically filed schemes detected by the CI Division.

Filing Season – The period from January through mid-April when most individual income tax returns are filed.

Fraud Detection CenterThe function responsible for identifying and detecting refund fraud, preventing the issuance of false refunds, and providing support for the Division field offices.

Hit-to-Scan Ratio – The estimation of the number of tax returns that must be scanned or reviewed to find one fraudulent return.  This is calculated as the number of fraudulent returns meeting the CI Division’s designated data-mining score and refund dollar tolerances divided by the number of verified fraudulent returns.

Master File – The IRS database that stores various types of taxpayer account information.  The database includes individual, business, and employee plans and exempt organizations data.

Processing Year – The year in which taxpayers file their returns with the IRS.  For example, most Tax Year 2006 returns were filed in PY 2007.

Questionable Refund Program – A nationwide program established to detect and stop fraudulent claims for refunds on income tax returns.

Refund Offset – A computer program that will automatically apply a refund due a taxpayer to another account on which the taxpayer owes money to the IRS.

Return Preparer ProgramA program that pursues unscrupulous return preparers who knowingly claim excessive deductions and exemptions on returns prepared for clients.  The clients might or might not have knowledge of the false claims.

Scheme – Can include only one return but generally includes numerous returns.  In addition, many small fraudulent refunds that do not have common characteristics may be placed in a “dump” scheme.

Scheme Tracking and Referral System – The system of records maintained at each Fraud Detection Center for QRP and Return Preparer Program schemes.  It was designed to store information, for multiple processing years, that is used for tracking and historical purposes.

Submodel – A specific data-mining model designed to focus on known areas of high risk or other general areas of concern.

Taxpayer Advocate Service – An independent organization within the IRS that helps taxpayers resolve problems with the IRS and recommends changes to prevent problems.

 

Appendix VII

 Management’s Response to the Draft Report

 

The response was removed due to its size.  To see the response, please go to the Adobe PDF version of the report on the TIGTA Public Web Page.


[1] See Appendix VI for a glossary of terms.

[2] This estimate was based on data available at that time.

[3] According to the referral procedures, fraudulent refund returns are forwarded to either the Wage and Investment Division Examination function or the Small Business/Self-Employed Division Examination function.  The majority of fraudulent refund returns are forwarded to the Examination function in the Wage and Investment Division.  For purposes of this report, we will use the generic “Examination functions,” unless we need to refer to a specific Division for clarification.

[4] Review of Procedures to Identify and Stop Economic Stimulus Payments on False Returns (Audit Number 200810032).

[5] See Appendix VI for a glossary of terms.

[6] To facilitate comparisons, we eliminated a PY 2004 refund scheme from the amounts shown as Fraudulent Refunds Identified and Fraudulent Refunds Stopped.  ****1****

[7] According to the referral procedures, fraudulent refund returns are forwarded to either the Wage and Investment Division Examination function or the Small Business/Self-Employed Division Examination function.  The majority of fraudulent refund returns are forwarded to the Examination function in the Wage and Investment Division.  For purposes of this report, we will use the generic “Examination functions,” unless we need to refer to a specific Division for clarification.

[8] Actions Have Been Taken to Address Deficiencies in the Questionable Refund Program; However, Many Concerns Remain, With Millions of Dollars at Risk (Reference Number 2007-10-076, dated May 31, 2007).

[9] An Executive Steering Committee, consisting of members from the Taxpayer Advocate Service, the CI Division, the Small Business/Self-Employed Division, the Wage and Investment Division, and the Modernization and Information Technology Services organization, was formed in response to the National Taxpayer Advocate’s 2005 Annual Report to Congress.  Based on the problems identified in the Report, the Executive Steering Committee was to establish a process to notify taxpayers that their refunds were being held and to revise the QRP.

[10] Actions Have Been Taken to Address Deficiencies in the Questionable Refund Program; However, Many Concerns Remain, With Millions of Dollars at Risk (Reference Number 2007-10-076, dated May 31, 2007).

[11] This estimate was based upon data available at that time.

[12] To arrive at the outcome measure associated with this finding and presented in Appendix IV, we used our estimate of approximately $894 million and subtracted $318.3 million, the amount estimated and claimed in our August 2006 report. 

[13] The Electronic Fraud Detection System Redesign Failure Resulted in Fraudulent Returns and Refunds Not Being Identified (Reference Number 2006-20-108, dated August 9, 2006).  The estimate in this report was based on the amount of fraudulent refunds stopped during PY 2005 and the amount stopped during PY 2006 as of the time of the audit. 

[14] Actions Have Been Taken to Address Deficiencies in the Questionable Refund Program; However, Many Concerns Remain, With Millions of Dollars at Risk (Reference Number 2007-10-076, dated May 31, 2007).

[15] McCoy, Kevin.  2006. “How the IRS failed to stop $200M in bogus refunds.”  December 4, 2006, USA Today.

[16] Actions Have Been Taken to Address Deficiencies in the Questionable Refund Program; However, Many Concerns Remain, With Millions of Dollars at Risk (Reference Number 2007-10-076, dated May 31, 2007).

[17] See Appendix IV for details.

[18] Our estimate is based on the potential amount of fraud using a lower threshold and the actual fraud hit-to-scan ratios and detection rates achieved by the CI Division at the existing thresholds.  Although the hit-to-scan ratios and detection rates could be different at the lower threshold, they could not be readily determined.  We believe that the indicators used provide a reasonable estimate of the potential additional fraud that could be identified using the lower thresholds.  We also recognize that these additional cases could not be worked without additional staffing and changes to existing law.

[19] During each processing year, the CI Division is unable to stop a number of fraudulent refund claims that it identifies, primarily because the Centers identify the returns as fraudulent after they have the ability to stop the refunds from being issued. 

[20] This percentage is derived from information provided by the data-mining contractor dated October 27, 2007.  Although this percentage was calculated prior to the end of the processing year, we believe that this is an accurate representation of the percentage of potentially fraudulent returns that were processed.

[21] See Appendix IV for details.

[22] ****3(d)****  These can be verified only by conducting an examination.

[23] Actions Have Been Taken to Address Deficiencies in the Questionable Refund Program; However, Many Concerns Remain, With Millions of Dollars at Risk (Reference Number 2007-10-076, dated May 31, 2007).

[24] This amount is included in our $742 million estimate.

[25] Data for fraud estimation without thresholds were not available for PYs 2003 through 2005.

[26] Some returns cannot be referred to the civil functions for resolution, such as those involving identity theft and those not meeting established dollar thresholds. 

[27] The Office of Refund Crimes suspended the manual quality review process in March 2007 because the time needed to review the returns was causing a delay in the movement of work.

[28] See Appendix IV for details.

[29] CP05 Notice – Informs the taxpayer that his or her refund is being held for further review.  This notice will generate through an automatic process completed through Master File programming.

[30] CP05A Notice – Is sent to the taxpayer when his or her tax return has been verified as being false by the CI Division.  It requests that the taxpayer provide documentation to substantiate claims made on the return.

[31] See Appendix IV for details.

[32] Review of Procedures to Identify and Stop Economic Stimulus Payments on False Returns (Audit Number 200810032).

[33] See Appendix VI for a glossary of terms.

[34] See Appendix V for details.

[35] See Appendix VI for a glossary of terms.

[36] The Electronic Fraud Detection System Redesign Failure Resulted in Fraudulent Returns and Refunds Not Being Identified (Reference Number 2006-20-108, dated August 9, 2006).  Our estimate in this report was based on the amount of fraudulent refunds stopped during PY 2005 and the amount stopped during PY 2006 as of the time of the audit.

[37] See Appendix VI for a glossary of terms.