Menu
Tax Notes logo

IRS Data Analytics Could Start Race on Voluntary Disclosures

Posted on June 25, 2019

The IRS’s sophisticated mining of its large data sets for investigative leads should instill taxpayers considering voluntary disclosure of past noncompliance with a sense of urgency.

The IRS Criminal Investigation division’s data analytics tools and its possession of large data sets allow the agency to find issues to investigate more quickly and easily, according to Robert J. Kovacev of Norton Rose Fulbright US LLP.

“So if there is something that you think requires voluntary disclosure, every day you wait increases the chance that you are going to miss out on the ability to make a voluntary disclosure,” Kovacev told Tax Notes.

Sandra R. Brown of Hochman Salkin Toscher Perez PC said the new data sets and analytics tools are allowing the IRS to make connections on its own in situations when it used to rely on informants and use those connections to open civil and criminal investigations. “Therefore, if you have reason to believe you potentially could be part of something the IRS may have . . . there is a potential now that it’s increased the likelihood that the IRS could eventually get to you,” she said.

CI Chief Don Fort announced in March that the IRS’s data analytics approach is here to stay. CI’s use of analytics tools has been exemplified by the nationally coordinated investigations unit (NCIU), and he said he hopes to double that unit’s case referral output.

Last summer Fort also mentioned that CI, at that time, had over 1 billion records to feed into its data analysis.

The last year has also seen the end of the offshore voluntary disclosure program and a new memo revising the general application voluntary disclosure program.

Taxpayer Timeliness

One key element of a voluntary disclosure is its timeliness. Internal Revenue Manual section 9.5.11.9 describes four conditions for a timely disclosure.

A timely disclosure must happen before the IRS begins a civil or criminal investigation of the taxpayer, before the IRS receives information from a third party about the specific taxpayer’s noncompliance, before the IRS has started a civil or criminal investigation of someone else that is “directly related to the specific liability of the taxpayer,” and before a criminal enforcement action has supplied the IRS with information directly related to the taxpayer’s specific liability.

Some of the data sets that CI has announced it is mining, like Swiss bank data and voluntary disclosures, might fit one or more of those timeliness factors.

Brown, the former acting U.S. attorney for the Central District of California and chief of its tax division before that, said one of the IRM examples seems particularly apt in the big data and analytics realm. It involves an IRS civil compliance project investigating a widely promoted scheme and the IRS obtaining “information which might lead to an examination of the taxpayer.” The example concludes that if the taxpayer voluntarily discloses before any examination or investigation begins, that disclosure will be timely.

Brown pointed out that the large data sets at issue are the IRS’s internal data — something Fort has been highlighting in rebuttal to news stories describing “Big Brother” data gathering. CI has been describing attempts to better analyze internal data, she said, adding that that sounds a lot like a project case.

In other words, a disclosure might be timely as long as it comes before an NCIU referral, Brown said.

Kovacev said the IRS should need something more specific than possession of raw data before a disclosure would be untimely. “They actually need to do something with it,” he said, pointing to the IRM investigation initiation tests.

The voluntary disclosure memo says the IRM will be updated to reflect the new framework by November 2020. So far, the IRS website version of IRM 9.5.11.9 doesn’t reflect any changes.

However, elsewhere on its website, the IRS posted a revised portion of that provision dated April 9. The revision tweaks the language in a couple of places. One of those changes is to the IRM compliance project example, but the change doesn’t affect the conclusion.

New Connections

Kovacev said the IRS’s new analytical capabilities should spell the end of the so-called audit lottery. “You can’t assume that the IRS is not going to make connections just because it’s not from an obvious source,” he said.

Anecdotes about real-time investigation and observation in tax cases will become more common, Kovacev said.

Brown said the increased chance of discovery through the IRS’s data analytics tools should be part of taxpayers’ voluntary disclosure deliberations.

Taxpayers and practitioners should be on the lookout for IRS announcements of both receipt of new data sets and of new areas of interest, Brown said. Those announcements send signals to taxpayers who may have something to clean up related to data the IRS is about to feed into its analysis, she said.

In the Dark

One issue that should concern taxpayers with noncompliance issues is the gap in public information about CI’s data sets and analytics tools, according to Kovacev. “We don’t know exactly what tools they are using, and we don’t know exactly what the data sources are,” he said.

Taxpayers are now more uncertain about what information and tools the IRS already has, so “you should always assume that the IRS has the information and that they have the tools that will sooner or later connect the dots,” Kovacev said.

CI is famously — and appropriately — secretive,” Kovacev said. Special agents may be working a taxpayer’s case long before that taxpayer discovers the investigation, he pointed out.

CI has been good about documenting the opening of an examination or investigation, but that may happen without the taxpayer learning about it, Brown said.

The open question for taxpayers now is not whether the IRS can find someone in its data, but when it will do so, Brown said.

Copy RID