recipients to deter fraud -- a concern in the densely populated Northeast -- remains a pipedream.

Another nagging concern about these systems has been the amount of savings they actually generate. Biometric identification systems dont catch recipients in the act of double dipping, but deter fraud by scaring away potential abusers. Officials compare welfare application rolls before and after the system starts operating to measure whether cheaters have left the system and are no longer claiming benefits.

In 1999, Texas conducted a thorough examination of its finger imaging system and its impact on cost savings. The state, which has more than 975,000 clients, estimated that it saves between $6 million and $12 million annually by denying fraudulent claims for duplicate benefits. The savings were based on figures taken from several regional claims offices showing a less than one-half percent reduction in claims when applicants had to submit to finger imaging.

But actually proving these systems prevent fraud and save governments money has been difficult. In 1997, New York state conducted a study of its welfare finger imaging system and came to the conclusion that it was "statistically impossible to calculate how many people stay away [from welfare] because of finger imaging rather than other reasons. Caseloads were already declining sharply, including a rebounding economy, tough work requirements, anti-fraud home inspections and shortened eligibility."

Unclear numbers also killed what was seen as a promising application for biometrics in health care. Floridas Medicaid program proposed that pharmacies finger scan all Medicaid recipients who need a prescription filled. The goal was to improve identification and reduce the number of unauthorized prescriptions issued by pharmacies either through fraud or error, thereby saving the states Medicaid system millions of dollars.

But the Agency for Health Care Administration, which runs Medicaid in Florida, dropped the plan after it realized there was no real way to measure how the system would save money, according to Connie Ruggles, a senior management analyst. "We found there were other, more direct changes we could make that produced measurable savings."

Knowing how to accurately measure fraud prevention through finger imaging has been a shortcoming that needs to be addressed, according to Mintie. "People arent so sure about the savings," he said. "We need some standards in this area. Right now, states count prevention different ways."

Digital DMVs

When administrators of motor vehicle agencies took a look at biometrics for identifying license holders, they tackled the standards issue head on. AAMVAnet Inc., an affiliate of the American Association of Motor Vehicle Administrators, published a national standard for storing fingerprint minutia on license cards. According to Nathan Root, AAMVAnets standards program director, the standard is in widespread use and has made print matching fast, an important issue when searching a database that may contain millions of other prints.

Whats not so widespread is the use of biometrics to identify drivers. "The big hurdle is the privacy concern," said Root. "There are a lot of public fears about Big Brother."

Those fears stifled the use of biometrics just as it was getting started. A few years ago, six states mandated the use of fingerprints to identify drivers to reduce fraud and identity theft. Today, the number of states collecting drivers fingerprints hasnt changed.

Some states, including West Virginia and Illinois, have tried to reduce the invasiveness of fingerprinting by using facial scanning similar to the system used at the Super Bowl instead. In 1999, Illinois deployed the worlds largest drivers license facial recognition system, built by Viisage Technology. When complete, the database will contain more than 20 million images. West Virginia uses a facial recognition system built by Visionics and Polaroid to verify the identification of drivers whose licenses came up for reissue. West Virginias system has a database of more than 2 million facial images.

Tod Newcombe  |  Features Editor