JUNE 29, 2005




Chairman Ramstad, Representative Lewis, and Members of the Subcommittee, thank you for the opportunity to discuss with you today a particularly troublesome tax administration issue:  tax refund fraud committed by Federal and State prisoners. 


In response to a request by this subcommittee, the Treasury Inspector General for Tax Administration is conducting an audit of the extent of prisoner refund fraud and IRS efforts to combat it.  While our work is ongoing, we are able to draw some conclusions at this point and recommend solutions to this growing problem.


Prisoners, like all other taxpayers, have a legal obligation to pay their taxes and have the legal entitlement to a refund of overpaid taxes.  This civic duty and legal right only partially explain why the IRS received approximately 455,000 tax returns from prisoners last year.  Another explanation for some of these tax returns is that prisoners have found ways to exploit weaknesses in IRS operations in order to receive refunds to which they are not entitled.  The IRS must close gaps in its policies and procedures to prevent this affront to the American public from continuing to expand.   


Findings on Prisoner Tax Fraud


The number of tax fraud schemes perpetrated by prisoners is on the rise.  According to the IRS, prisoners filed approximately 4,300 fraudulent returns in processing year 2002.  Two years later, that number quadrupled to over 18,000.  It is worth noting that these figures only account for those prisoner returns that the IRS identified as fraudulent during tax return processing.  During the past three years, prisoners have filed over 1.3 million returns, so the risk that fraudulent returns are slipping through the system undetected is great.


During the course of our review, we obtained data from the IRS Criminal Investigation Division (CI), the IRS entity responsible for detecting fraudulent returns. The chart on the following page provides statistics on the total number of returns filed and the refund amounts claimed by all individuals and by prisoners as of April 1, 2005.



Processing Year (PY) 2004

Number of Returns








Total Returns Filed


Not Available[1]


Refund Returns Filed




Returns Reviewed for Potential Fraud








False Refund Returns[2]




False Refunds Stopped




False Refunds Issued








Amount of Refunds








Refunds Requested

 $ 227,573,835,000

 $ 758,951,862






False Refunds

 $ 440,773,403

 $ 68,179,070


False Refunds Stopped

 $ 309,961,554

 $     53,456,963


False Refunds Issued

 $ 130,811,849

 $     14,722,107






Average False Refund

 $                   3,733

 $              3,755


Average False Refund Stopped

 $                   3,784

 $              3,809


Average False Refund Issued

 $                   3,618

 $              3,568



This chart shows that a disproportionately higher percentage of fraudulent returns are filed by incarcerated individuals.  Although prisoner returns account for only .43 percent of all refund returns, they account for over 15 percent of the fraudulent returns identified by the IRS. It is difficult to be surprised that those already imprisoned for committing a crime are more prone than the general public to commit another crime.


Of particular concern is the fact that the IRS frequently pays refunds on returns it has identified as fraudulent.  In 2004, the IRS paid 36,000 refunds on returns that it determined to be fraudulent; 4,100 of these refunds were issued to prisoners.  Stated another way, if a taxpayer submitted a fraudulent return in 2004 – and the IRS determined that return to be fraudulent – that taxpayer still had a 31 percent chance of receiving a refund.  Taking into account that the average false refund amount last year was about $3,600 per return, the IRS paid $131 million in refunds on returns that it identified as fraudulent.  Of this amount, $14.7 million was erroneously paid to prisoners.


Prisoners who cheat the tax system use a variety of tactics.  For example, two Louisiana inmates were sentenced this past March for using the names and Social Security Numbers (SSNs) of other inmates to file fraudulent tax returns.  This scheme eventually resulted in their conviction for conspiring to receive $266,000. In Missouri, two former inmates, while in prison, charged fellow prisoners $100 each to prepare a false tax return.  These returns were then filed by co-conspirators outside of the prison to obtain illegitimate refunds.  In April 2005, one was sentenced to one year and nine months in prison; the other was sentenced to two years and six months.


As this last example illustrates, prisoners often use individuals outside of prison to perpetrate fraud.  Although the current IRS management information system cannot report specific information on how may prisoners use the names of other individuals to commit fraud, our analysis of 18,343 false refunds involving prisoners for the 2004 processing year:[3]


·         Identified 1,193 schemes claiming $68.7 million in false refunds; and,

·         Included 113,797 returns, not identified as prisoner returns, claiming an additional $380.8 million in refunds.[4]


These figures demonstrate that prisoners often work collaboratively with persons outside of prison in sophisticated and complex refund fraud schemes, as evidenced by the large number of false returns relating to prisoner schemes but not identified as prisoner returns.


At this point during our review, we can confidently state that prisoner tax fraud is rising, prisoners file a disproportionately higher percentage of fraudulent returns than the general public, the IRS frequently pays refunds on tax returns despite identifying these returns as fraudulent, and the IRS lacks adequate data to ascertain the extent of the prisoner fraud problem.


IRS Process for Detecting Fraudulent Refund Returns


As these findings demonstrate, prisoner tax fraud is a serious problem.  To fix this problem, the IRS must improve its ability to identify fraudulent prisoner returns. 


Currently, the IRS sends all refund returns through its Electronic Fraud Detection System (EFDS) before issuing a refund.  If the filer is a prisoner according to information provided by Federal and State prison officials, the return is flagged with a prisoner indicator.  The prisoner indicator is one of several elements used by the EFDS to assign a data mining score to tax returns.  The higher the score a return receives, the greater the likelihood that the return is fraudulent.


In general, returns that receive a high data mining score must also request a refund that exceeds a certain threshold established by CI to be selected for further screening.  The number of returns selected for screening is then based on the amount of resources available to screen the returns.  Thus, detection of false refunds is a function of the data mining score, the threshold established for screening, and the amount of CI resources.  These three criteria determine how many false refunds are identified and how many false refunds slip through the processing system undetected. 


The IRS’ Fraud Detection Centers are the frontline for detecting fraudulent refund schemes.  In processing year 2004, these Centers physically screened only 36,126 of the 455,097 prisoner returns filed.  As we continue our analysis of this data, we will evaluate how – with improved prisoner data and an invigorated commitment by the IRS to aggressively pursue prisoner refund fraud – the IRS can identify more high-risk prisoner returns. 


Reasons False Refunds Were Erroneously Paid


As previously noted, the IRS often issues refunds on tax returns it eventually identifies as fraudulent. The IRS attributed many of its erroneous payments of fraudulent tax refunds to the short time constraints it sets for itself.  Part of its customer service philosophy is to pay tax refunds promptly since the vast majority of tax refunds are legitimate.  This attempt to pay refunds promptly generally gives the IRS about three weeks to prevent a fraudulent refund from being paid.  During that time, CI evaluates the return for indicators of fraud, verifies any wage and withholding information with employers, and scans other potentially fraudulent returns for similar return characteristics, such as a common employer, street address, or bank account.  The IRS must work quickly within this time frame to ensure legitimate refunds are paid promptly.


Paper returns in particular present the IRS with unique challenges.  The CI function must physically review the return to collect return information that is not entered electronically.  CI has even less time on paper returns that request an electronic direct deposit of the refund into a bank account.  Prisoners may have already discovered this weakness in the system, since over three-fourths of all fraudulent refund returns filed by prisoners in processing year 2004 were filed by paper instead of electronically.


We obtained a computer extract from CI that described the reason why a refund was not stopped despite being identified as fraudulent.  As of May 4, 2005, the record contained 4,261 records of false refunds issued to individuals identified in the 2004 prisoner file.  The following table shows the five most common reasons the IRS listed for issuing refunds on false returns.[5]


Reason Refund Not Stopped

Total Refunds Not Stopped

Identified after Cycle Cutoff[6]




No Selection[8]


Tax Examiner Error[9]


Prior Year Return[10]


Total Top Five Reasons




The chart above shows that CI chose the category of “Other” or “No Selection” for why refunds were erroneously issued to prisoners in over 29 percent of its cases. While some use of these reasons is legitimate, TIGTA believes they should be used sparingly to enable CI to identify the actual reason why an erroneous refund was issued and enable CI to correct deficiencies in its system. 


CI conducted an additional analysis of the reasons tax refunds were not stopped despite being identified as fraudulent, and identified the causes in the chart on the next page.[11] 

Reason Refund Not Stopped

Total Refunds Not Stopped

Return Preparer Case[12]


Prior Year Return[13]


Did Not Meet Data Mining Tolerances[14]


Untimely Receipt of Paper Return[15]


Direct Deposit Refund of Paper Return[16]


Volume of ELF Scanning (resource issue)[17]


Human Detection, Data, or Input Errors[18]


  Total Refunds Issued


IRS Efforts to Combat Tax Refund Fraud


With the dramatic increase in fraudulent refund returns filed by prisoners, one would expect a strong, coordinated response from the IRS to combat these schemes.  However, until recently, the IRS did not have an overall, comprehensive approach to working with Federal and State prisons to address prisoner tax fraud.  In some locations, prisoner tax refund fraud was not considered to be a prevalent issue.  Further, according to CI management, some U.S. Attorney’s offices are reluctant to pursue these investigations, believing it is not a prudent use of resources, particularly if the person is already incarcerated and another conviction would not likely yield additional punitive sanctions.


Instead of an overall, comprehensive approach, our review has discovered that the IRS let its 10 Fraud Detection Centers located around the country establish their own policies and procedures for working with prisons in their region.  These Centers coordinated to varying degrees with the prisons in their area.  For example, we discovered that 4 of the 10 Fraud Detection Centers have not established any procedures to ensure that prison mail addressed to the IRS is sent directly to these Centers for screening.  As a result, this mail is received by the IRS just like any other tax return, and the IRS must rely on its incomplete and inaccurate prisoner database to identify the filer as a prisoner.  Furthermore, 8 of the 10 Centers do not share information with State tax revenue authorities that a prisoner has been caught filing a fraudulent Federal tax return.  Sharing such information would help ensure that prisoners caught cheating the IRS would not be allowed to cheat the State revenue authorities with impunity.


Despite the IRS’ inconsistent and incomplete approach to address prisoner refund fraud, the CI function has conducted criminal investigations on certain prisoners and prisoner refund schemes.  Our comparison of the prisoner database to the March 2005 Criminal Investigation Management Information System (CIMIS)[19] shows that the CI function had initiated 312 primary investigations on schemes involving 3,069 prisoner returns.  From 75 of these primary investigations, CI has initiated 128 subject investigations.  We will continue to analyze this data and provide the results of our analysis in a report to the IRS.


Proposed Actions to Enhance Tax Refund Fraud Detection


The Criminal Investigation Division has taken some steps to address the concerns expressed by this subcommittee and in response to our review.  Management of CI has proposed several legislative and procedural remedies that will improve coordination with Federal and State prisons.  CI has also requested computer programming changes to provide more time to verify tax return information and to pinpoint common characteristics of prisoner refund schemes.  If properly implemented, I believe these actions will strengthen the ability of the IRS to detect and deter prisoner refund fraud.


Need for Accurate and Complete Prisoner Information


I want to emphasize that the IRS is not solely responsible for confronting this problem.  Federal and State prison officials should be required to transmit complete and accurate information on the prisoner population to the IRS.  The IRS uses the data submitted by Federal and State prison officials in its Electronic Fraud Detection System.  As with any computer-based analytical tool, the ability to quickly and accurately identify potential tax refund fraud is only as good as the data used in such analyses, and we have found that much of this data is inaccurate and incomplete.


During the 2004 processing year, we estimate that approximately 550,000, or almost 20 percent, of the 2.8 million prisoner records the IRS received contained incomplete or inaccurate information.  These records were inadequate for the following reasons:


  • About 255,000 records did not have a valid SSN.  For example, over 252,000 of these records listed 000-00-0000 as the SSN.
  • About 118,000 records had duplicate SSNs.
  • Almost 43,000 records were invalid because they exceed the highest SSN issued by the Social Security Administration.
  • About 134,000 records were not in the 2004 prisoner file because the IRS required all prisoner information to be submitted by August 2003, in order to have time to perfect the data and enter it into IRS computer systems.  This early deadline excluded individuals incarcerated between September 1, 2003, and December 31, 2003.


The inaccuracies in the prisoner file prevent the IRS from detecting all false refund returns filed by prisoners.  The IRS must have a current and accurate prisoner file to alert CI to prisoner-filed returns.  An improved prisoner file would also enable the IRS Submission Processing function to stop false prisoner refunds on those returns that do not meet CI criminal investigative criteria before they are issued.  If Congress were to require Federal and State prisons to provide accurate Social Security Numbers and other prisoner information in a consistent format, the IRS could obtain the information much later in the year and include it in the prisoner database for the upcoming tax season.  This improvement to information quality would help prevent prisoners from bilking the Government out of millions of dollars.




Mr. Chairman and members of the subcommittee, I appreciate the opportunity to discuss this important tax administration issue today.  As I stated initially, our review of the extent of prisoner fraud and the effectiveness of IRS efforts to prevent it is still ongoing, but we will continue to work with the IRS and this subcommittee as we near completion.  I will be happy to answer any questions you have at the appropriate time.





[1] The IRS cannot identify the total number of prisoner returns filed because the prisoner data file is only matched against refund returns.

[2] This could include false returns identified during IRS returns processing by the Electronic Fraud Detection System (EFDS), which is an automated system that is used by personnel in Fraud Detection Centers (FDCs) to review potentially fraudulent tax returns.  False returns can also be identified with the assistance of prison officials, informants, or other sources. 

[3] In the chart on page two, 18,159 false refunds were reported by CI as of April 1, 2005.  Since that date, this number has risen to 18,343 false refunds because false returns are continuously being identified.

[4] Returns not identified as prisoner returns could actually be submitted by prisoners but not identified as prisoners due to the incompleteness of the prisoner file.  There could also be situations of prisoners using the Social Security Numbers of non-prisoners.  Similarly, there could be only a few returns identified as prisoner returns in a scheme with many other non-prisoner returns.

[5] These five reasons were given as explanations for 91% of the refunds issued to prisoners who submitted fraudulent returns.

[6] “Identified after Cycle Cutoff” means that the tax refund, after going through various computer routines at a Submission Processing Center to perfect the data for processing, was issued before certain additional analyses could be completed that would have identified the return as fraudulent.

[7] “Other” is a generic catch-all category that TIGTA believes should be used sparingly, as it may prevent CI from identifying the actual reason a false refund was issued and impair its ability to improve its system.  The CI function advised that time constraints during processing may have led to the high use of this category.

[8] “No Selection” is also a generic catch-all category that TIGTA believes should be used sparingly, as it may prevent CI from identifying the actual reason a false refund was issued and impair its ability to improve its system. The CI function advised that time constraints during processing may have led to the high use of this category.

[9] “Tax Examiner Error” means that an IRS employee may have entered or interpreted data incorrectly, or otherwise possessed information to prevent a refund from being issued, but did not stop the refund.

[10] “Prior Year Return” includes returns that were processed in the previous year but were detected during the current processing year. Some of these returns are identified at a later date due to an informant or the return being associated with a current year scheme.

[11] This analysis is based on 4,000 refunds; therefore, it cannot be precisely compared with the data we obtained on the 4,261 refunds identified in the 2004 prisoner file.  Although CI’s additional analysis provides a better indication of the reason refunds were issued, CI based some of its assumptions on data analysis, not actual reviews of cases.

[12] CI management explained that refunds are usually not stopped in return preparer cases because the focus of the investigation is the return preparer, not the individual taxpayers who, knowingly or unknowingly, are party to fraud.

[13] Prior year returns were processed in the prior year but detected during the current processing year.

[14] This category indicates that the return did not meet data mining tolerances when it was processed.  However, analysts later determined the return to be false upon reviewing a subsequent return with similar characteristics.

[15] In order to identify the return as fraudulent, analysts must review the paper return; however, analysts did not receive the paper return in time to review it and stop the refund.

[16] Similar to the previous category, analysts must obtain the paper return to determine its legitimacy.  In these cases, the refunds were already deposited electronically into bank accounts before the refund could be prevented.

[17] In citing this reason, the CI function assumed that the volume of electronic filing scanning was so large that analysts could not review them in time to determine whether they were false.

[18] This category includes IRS employees who entered or interpreted data incorrectly, or otherwise possessed information to prevent a refund from being issued, but did not stop the refund.

[19] CIMIS is a database that tracks the status and progress of IRS criminal investigations and the time expended by special agents.  It is also used by IRS management as a basis for national and local resource and inventory decisions.