IRS and Agriculture efforts strike pay dirt

Using analysis tools is necessary because 'part of our problem is we're almost drowning in data,' Agriculture's Garland Westmoreland says.

Susan Whitney-Wilkerson

The IRS and Agriculture Department are mining data to ferret out frauds who cheat American taxpayers.

The IRS' Offshore Credit Card Program is designed to identify tax evaders who take their money to banks in countries with lax and secretive banking laws and access it through credit cards.

Data-mining applications help the Agriculture Department's Risk Management Agency identify suspicious crop insurance claims filed by farmers.

Data-mining and link analysis applications flag anomalies among masses of data, making it easier for investigators to spot people trying to take advantage of limitations in federal enforcement.

The cost

'There's nothing illegal about having an offshore bank account or credit card, but if you're not reporting it and you're using it as a vehicle to hide your income, then that's a problem,' said John McDougal, a lawyer with the IRS' Small Business/Self-Employed unit and the Offshore Credit Card Program. The IRS loses at least $70 billion a year in tax revenue from such tricks, he said.

'We've been trying to deal with this for many years but never had an adequate way to get a handle on it,' he said. Data-mining applications help turn up information about offshore cardholders, giving the IRS leads for investigations.

The IRS studies credit card transactions and spending patterns of those using offshore cards.

The card-issuing banks might be in the Bahamas or Cayman Islands, but the credit card transactions generally are processed on computers in the United States. The IRS has gained court-approved blanket summonses to secure offshore transaction data from MasterCard International Inc. and Visa USA for 31 countries that are known tax havens.

'The credit card companies don't know who the cardholders are. Only the issuing bank knows that, and the issuing banks are in the secrecy jurisdictions,' McDougal said. But the transactions on the cards offer clues with merchant names and geographic locations.

The IRS collects massive amounts of transactional data from the companies. The challenge is to piece together data from the transactions and figure out the cardholder's name, location and Social Security number, so the IRS can pull tax records and determine if there's a problem that merits an audit.

The tax agency uses commercial data-mining and analysis technology to analyze the transactions. The key is to find the appropriate tools to use in the right combination. 'We're adding technology incrementally,' McDougal said, but he would not detail the tools the IRS has now or is planning to deploy because the agency doesn't want to help tax cheats find ways to further avoid discovery.

The IRS has identified thousands of suspicious cardholders. To do that, the agency amassed all the data collected from Visa, MasterCard and major merchants with whom the cardholders conducted business, such as airlines, online service providers and car rental agencies.

By analyzing the number and location of transactions, IRS officials use the data-mining software to identify cardholders whose status they want to investigate further.

Once the IRS selects a case for examination, agents follow standard examination procedures, contacting the taxpayer, explaining the investigation, determining if there's a problem and giving tax evaders a chance to comply and pay back taxes. Those who don't comply face further examination and civil fraud prosecution.

Last year, the IRS turned up thousands of people who avoided paying taxes by using offshore accounts, McDougal said.

Maximizing resources

At Agriculture's Risk Management Agency, data mining helps the agency get the most out of its resources, both by helping target funds where they're needed most and by weeding out fraudulent claims, said Garland Westmoreland, director of strategic data acquisition and analysis.

The agency helps farmers manage their risk by providing crop insurance. RMA, which reinsures companies that provide crop insurance, also monitors compliance with program rules by insurers and policyholders.

The agency created a data warehouse of several years' claims , to which analysts applied algorithms to detect anomalies.

'We did an analysis and looked for folks who were having losses that were difficult to understand because other people with the same crop and the same practices in the same area weren't having losses or perhaps not as big or as frequent,' Westmoreland said.

Agriculture then notified policyholders with questionable losses that an agent might visit their farm to conduct an inspection. The letters prompted by the data mining resulted in farmers dropping more than $250 million in claims over the last three years.

The Center for Agribusiness Excellence, a partnership between Planning Systems Inc. of Reston, Va., and Tarleton State University in Stephenville, Texas, provides the data-mining and data warehouse applications for Risk Management.

The data warehouse is built on Teradata hardware and software from NCR Corp. The system uses an Oracle9i database as well as geographic and mapping software from ESRI of Redlands, Calif.

Each year Risk Management has expanded its technology to analyze broader issues and drill deeper into the data. For example, applications evaluate the operation of a pilot program in one part of the country and compare it with a similar project in another region.

'Part of our problem is we're almost drowning in data,' Westmoreland said.

The agency now uses a dashboard to make sense of the data. The dashboard summarizes the information graphically so investigators can view records for any state, county or crop and determine how a program is faring.

The results might indicate a need for a review, for instance, if a program is doing well in one county but experiencing losses in another.

Reader Comments

Fri, Jul 10, 2009

data mining? why not just get UBS to turn over their records of 50000 0ffshore accounts

Please post your comments here. Comments are moderated, so they may not appear immediately after submitting. We will not post comments that we consider abusive or off-topic.

Please type the letters/numbers you see above