The New York Times and Forbes Magazine are reporting on a study by the Transactional Records Access Clearinghouse, a research group affiliated with Syracuse University, which claims to show that the number of audits of corporations reporting assets of $250 million or more has dropped between 2005 and 2009, while audits of smaller companies have increased during that same time.
Believing that the IRS has misallocated its resources, TRAC blames this development on a “perverse quota system” within the IRS. The IRS, however, disputes the study’s findings. From the NY Times article:
I.R.S. officials, who have for years disputed the methodology used by TRAC, were quick to rebut the study’s findings. Steven T. Miller, the I.R.S. director of enforcement, said the study was skewed because it failed to take into account a surge in hours that I.R.S. agents spent working with businesses before they filed their returns to prevent errors or underpayments.
He asserted that the data actually showed that the agency had become more efficient in recovering unpaid taxes from the largest corporations because the average amount of money the auditors recovered per hour had risen to $9,704 in 2009 from $6,928 in 2005.
“We believe we’re looking at the right cases, we’re looking at the largest cases, we’re adding folks to these programs and have worked to really focus on the largest corporations,” Mr. Miller said. “This issue is about more than just the number of feet on the beat.”
Having not had the opportunity to review the TRAC study in detail, it would be difficult to draw any strong conclusions. One question immediately jumped out: to what extent (if any) has a reduction in the time spent auditing large corporations been due to the use of alternative programs like pre-filing agreements and advance pricing agreements, which are specifically intended to reduce the IRS’s examination burden (while still ensuring compliance)? If some portion of the reduction in auditing is due to an increase in alternative compliance-promoting techniques, it would seem that the reduction itself is not a meaningful data point without reference to the larger context.