IRS Criminal Enforcement Data Systems
The information provided the public by the Internal Revenue Service about the agency's criminal enforcement activities are substantially inaccurate. Taken as a whole, the data provide a highly misleading picture of the enforcement efforts of one of the IRS's most important components, the Office of Criminal Investigation.
Because these data cannot be relied upon to describe IRS's criminal enforcement efforts, we have not used them for our Web sites.
The counts on IRS criminal enforcement activities published for many years by the agency in its annual publication -- the Internal Revenue Service Data Book -- are significantly different than those produced by two independent agencies which track the same events: the United States Justice Department and the United States Courts. In addition, the IRS data are internally inconsistent. Coverage shifts as cases move through the system, resulting in an erroneous picture of the flow of workload from IRS investigators to federal prosecutors and then to the federal courts and final sentencing.
These problems are of a long standing nature. For example, similar problem were found and documented in a 1978 study by the National Academy of Public Administration (Report by a Panel of the National Academy of Public Administration (NAPA), Planning and Analysis Concerning the Intelligence Function of the Internal Revenue Service). A second 1980 study commissioned by the government and authored by one of TRAC's co-directors also turned up serious problems of data quality and reliability (see "Criminal Enforcement Actions: Information Systems and Detailed Statistics," pp. 149-179 (Chapter 6) in the monograph: The Internal Revenue Service: Measuring Tax Offenses and Enforcement Response, by Susan B. Long, U.S. Government Printing Office: September 1980 (National Institute of Justice, U.S. Department of Justice)).
According to those reports and TRAC's recent study, serious problems in IRS criminal record keeping systems are of at least three types:
As a result of these failings, IRS criminal enforcement data cannot be relied upon and therefore do not meet minimal standards for acceptable government data quality ((see Judging the Quality of Government Data).
- Cases Disappear. Cases simply vanish from the system without explanation. Cases, recorded by the IRS at one point in the process from investigation to prosecution and sentencing are not consistently recorded and tracked.
- Counts Don't Match Other Independent Data Systems. A second problem is that the counts of the Office of Criminal Investigation don't match other independent data systems -- both those within the agency and those outside the agency. In general other data systems (for example, IRS Chief Counsel's Office and the Justice Department's U.S. Attorney databases) record receiving many more cases from IRS Criminal Investigation than the IRS reports it referred to them.
- Information Recorded on Cases Doesn't Agree with Other Data Systems. Third, even when the same case is recorded in IRS Criminal Investigation databases and other independent data systems, the information recorded about these cases differed. For example,in the NAPA study, the size of the tax deficiency recorded by IRS Criminal Investigation was three times larger than that recorded by the IRS Examination Division on the cases selected for study. Declination rates (the proportion of matters the prosecutors refuse to handle) recorded in the IRS system were only half that recorded by federal prosecutors on IRS criminal referrals in the government's 1980 published study. The IRS's criminal data claimed that a much higher percent of individuals were sentenced to prison than did the U.S. Office of the Courts.
TRAC's current investigation -- extending for more than a year -- shows these identical problems have continued to plague IRS's Criminal Investigation tracking system. If anything, they appear to have grown worse over time.
TRAC has made repeated efforts to get IRS's response to these discrepancies. We have also asked IRS to provide TRAC with more detailed information about how their data system is maintained, about any studies they have conducted on data discrepancies across systems, and for the actual referral-by-referral data. Up until now, these efforts have not met with any success. (For a description of these contacts, see attachments.)
Last year we furnished the Internal Revenue Service with findings from our study (see December 31, 1996 letter). In contrast with TRAC's findings with respect to IRS's data systems, it found that the record systems of the Executive Office of United States Attorneys and those of the Administrative Office of the United States Courts produce closely parallel counts. (See, for example, Federal Tax Prosecutions, Federal Drug Prosecutions, ATF Weapons Investigations and Prosecutions Peak). However, court and federal prosecutor data differ substantially from the information IRS publishes on federal criminal enforcement. Significant differences were found not only in the counts, but in the general patterns of prosecution, conviction and prison sentencing rates, as well as in recorded trends overtime.
TRAC's study covered the period of 1981 through 1995. Similar to the findings in the 1980 study, it found that in every year the IRS recorded sending to federal prosecutors fewer criminal referrals than the prosecutors say they received. Once the matter reached court, however, especially in recent years, the IRS has claimed credit for more work than the prosecutors said they achieved. In 1995, for example, the IRS data claims almost 50% more prosecutions than the Justice Department, and over twice the numbers of individuals sentenced to prison. IRS also recorded the completion of two times more tax prosecutions than the U.S. courts said the courts had handled from all federal investigative agencies. (The IRS is not the only agency that refers tax cases.) The discrepancies between the data from the IRS and that data from the prosecutors and the courts have grown steadily more serious during the last decade.
Prison sentences appear to be even more inflated in IRS tracking system than prosecutions. In 1995 IRS says that its Criminal Investigation Program sent 2,229 individuals to prison. During the same period federal prosecutors reported that IRS referrals resulted in 947 individuals being sentenced to prison terms. Focusing only on criminal prosecutions where tax fraud was the lead charge, IRS systems recorded 733 of these had a tax offense (Title 26) as the lead charge. The federal courts and federal prosecutors each recorded less than half that number (337 and 318, respectively) as sentenced to prison from all sources for tax fraud.
In addition, IRS has withheld the information that TRAC requested which might allow TRAC to carry out a systematic referral-by-referral examination of these discrepancies. Finally, IRS has taken a very insular stance stating that it was satisfied with its data and saw no need to compare it with other sources to check its reliability -- this despite studies spanning a twenty year period finding this IRS data system seriously flawed.