ST. LOUIS - Faced with looming budget deficits, a growing number of states are mining for tax cheats.
Armed with the equivalent of electronic picks and shovels, they are using a sophisticated new computer techniques to root out delinquent taxpayers.
They are combing through often disparate records about peoples' personal finances - everything from house payments to motor-vehicle registration - to discover irregularities. For example:
In Texas, revenue officials took records from the Federal Aviation Administration on private jet ownership and compared it to state tax rolls. It found numerous examples of people who hangared their planes in the state but didn't pay taxes on them. The result: $5.8 million in additional revenue taken in.
In Iowa, tax authorities noticed that an out-of-state electronics company was filing unemployment claims for workers - within the state. It checked the corporate tax rolls. The company wasn't listed. Result: $1 million in unpaid taxes collected from the firm.
In California in one day, the state cross-checked mortgage-payment records with its own tax rolls and made an unusual discovery: both a European princess and a prostitute living in pricey Los Angeles homes who had never filed a return. They have now.
"This is a trend you should expect to see more and more," says Harley Duncan, executive director of the Federation of Tax Administrators in Washington, D.C. "States are starting to become highly creative in sifting through data to identify nonfilers and do better audit selection."
The reason for the investment in "data mining" is certainly understandable. The gap between taxes owed and tax money collected is generally believed to average 17 percent nationwide - representing billions of dollars.
States, faced with shrinking revenues and criticized in the past for heavy-handed collection tactics, find the new sleuthing techniques often deliver lucrative returns. And few people seem upset about tax authorities going after European royalty.
The computer techniques being used vary widely. They are able to probe disparate sources in ways that clerks sifting through stacks of paper never could. The most simple ones involve the tried-and-true method of states examining their own returns to look for unusual patterns.
More sophisticated tools mine other state databases for comparisons. They might, for instance, cross-check property- and income-tax information with motor-vehicle records to determine if a resident lives in a million-dollar home and drives a Mercedes but only declares $35,000 in yearly earnings.
The most complex systems reach outside the bureaucracy to tap commercial records. By one estimate, only a handful of states are using these "third-generation" data-mining efforts.
"The whole area is at an early stage," says John LeFaver, a tax-collection specialist with American Management Systems (AMS). "Data mining, to do it effectively, requires a level of expertise and knowledge that at this point most states don't have."
For those states that are delving into the exotic world of "neural networking," "pattern recognition," and "data cleansing," the results can be substantial. Mr. LeFaver estimates the seven states that have installed AMS systems, at a cost of $182 million, have realized additional collections of $904 million.
California is viewed as a leader in using mining techniques. Its system analyzes information from 47 different occupational licensing agencies (doctors, lawyers, accountants), and from a host of commercial databases and millions of businesses.
The system is helping the state differentiate between the steady taxpayer who misses a single payment, often for understandable reasons, such as illness, with the serial tax dodger. The result: Scarce enforcement resources can be focused on true scofflaws.
"By leveraging data about past behavior and determining who is likely to comply and who isn't, it makes for a much less confrontational environment," says John Vranna, an enforcement chief with the California Franchise Tax Board.