Image: Taxes © Thinkstock-SuperStock

The Internal Revenue Service relies on technology more than ever to sniff out tax cheats using robo-audits and data mining—but so far it has caught lot of minnows, and big fish are still eluding detection.

Even as millions of people's accounts are screened online and matched against their digital files elsewhere, the IRS's data-detection tools come nowhere close to collecting the $400 billion in tax dodges estimated to take place each year. The area in which its robo-audits have had the most impact is on tax returns for low-income taxpayers who try to claim the Earned Income Tax Credit.
 
In total, fraudulent claims totaled $2 billion, just 0.01 percent of the total of individual taxes. The EITC was the biggest single compliance problem cited.
 
That amount is expected to rise in the tax year ahead as the IRS extends the use of data mining to include the personal data of millions more taxpayers. Its sophisticated data-matching and pattern-recognition technology, largely developed by IBM over the past decade, will reach up the income ladder to include more middle-income and small-business filers who itemize deductions, although it is unlikely to have any impact on the complicated filings of high-net-worth taxpayers in the top 5 percent of income earnings, say tax experts who have studied the IRS plans.

"Real time" audits of electronic tax returns. The IRS's next phase in high-tech tax collection will be to create a "real-time" check of tax returns to "match them to third party information," said U.S. Treasury Inspector General for Tax Administration J. Russell George in testimony before Congress. Starting this year, the IRS tools will be able to track all credit card transactions, for starters.

The agency has also instructed agents on using online sources such as social media and e-commerce sites including eBay, as well as the rich data generated by mobile devices. In one controversial disclosure in April, the ACLU showed documents in which the IRS general counsel said the agency could look at emails without warrants, but the IRS has said it will not use this power.

While the agency has declined to give details about what third-party personal data it will use in robo-audits and data mining, it has told government and industry groups that its computers are capable of scanning multiple networks at the same time to collect "matching" comprehensive profiles for every taxpayer in America. Such profiles will likely include shopping records, travel, social interactions and information not available to the public, such as health records and files from other government investigators, according to IRS documents.

The IRS did not respond to written requests for information on its program. But George gave Congress an outline last April of the ambitious aims of its $1 billion "modernization" that gave it access to dozens of databases it has not previously used.

"This capability is designed to feed into a single, consolidated taxpayer-account database that will support the deployment of the next generation of taxpayer service and enforcement functions," George said in testimony.

But the many problems encountered already with the relatively simple screens used for Earned Income Tax Credit filers suggest there may be greater problems ahead for taxpayers. Already, independent advisers to the IRS have publicly warned that the agency is not prepared to take the next big step.

"Little has been accomplished to remove the confusion and uncertainty related to these rules," said a report from an IRS Tax Advisory Committee group that studies emerging issues such as use of third-party networks and data mining. Other IRS advisers have cautioned that the lack of transparency and secretiveness of the IRS could undermine the credibility of the tax-collection system.