During this 2012 political season, perhaps no other specific government program has received more time in the spotlight than the USDA’s SNAP program, formerly known as food stamps. An article published this week in the Huffington Post by Lisa Baertlein about food stamp fraud is a clear example of how the program is receiving intense scrutiny.
The SNAP program currently serves over 46 million Americans – that’s 1 in 7 Americans!
One reason for the scrutiny is that fraud is a major issue for the SNAP program. According to an article this week in the Business Insider, approximately $750 million of your hard earned tax dollars are wasted each year due to fraud in the program. As a prime example, according to Feb 2012 article in the McAlister News-Capital, a woman in Pittsburgh is accused of fraudulently pocketing more than $21,000 from the program by hiding income that would have otherwise disqualified her from benefits.
While many are placing blame on the SNAP program for this $750 million in fraud, my industry is looking elsewhere – the private EBT vendors through whom all the benefits are paid. EBT stands for “Electronic Benefits Transfer.” Long gone are the days of paper stamps. Since 2004, all SNAP benefits are provided electronically through a debit-like card (more correctly known as an EBT Benefits Card). Now, participants go to the store and checkout by swiping their SNAP EBT benefits card just like any other shopper using debit or credit.
While the financial industry is hard at work using data mining to profile and rapidly detect fraud in both credit and debit cards, there is currently little to no effort on behalf of EBT vendors servicing the SNAP program to do any form of automated fraud detection. This is absurd! A quick review of the web sites for every EBT vendor (found at http://www.fns.usda.gov/snap/ebt/pdfs/state-lines.pdf) reveals not a single one mentioning the use of even basic data mining to catch fraud.
Even the most basic, primitive use of data mining and mathematics can have a colossal impact on fraud detection. For example, a friend of mine, Dr. Richard Cleary at Bentley University, is a leading researcher on Benford’s Law. Put simply, Benford’s Law states that the first digit in a list of numbers, such as transaction values, follow a unique, non-uniform distribution. In other words, the first value in dollar amounts for financial transactions follow an interesting and predictable pattern. Deviations from this pattern are a red flag for fraud. We commonly use Benford’s Law in the financial industry.
Guess what – SNAP EBT transactions would be a prime candidate for Benford’s Law. If the same EBT participant has numerous transactions that break Benford’s Law, in all likelihood, there is probably fraud on the account. Furthermore, we can compute the probability that the case is fraudulent.
Something like Benford’s Law and the fraud detection described above doesn’t take years and millions of dollars to incorporate. This is something measured in hours! Within a few hours of starting to apply Benford’s Law, you can literally have a set of names, addresses, and phone numbers of potential fraud candidates. Start pounding doors to stop that $750 million loss the same day! It wouldn’t surprise me if this technique alone shaves millions… and it’s only the beginning.
By using simple techniques like Benford’s Law combined with more advanced strategies that automatically detect suspicious patterns and behavior, a comprehensive approach to proactive fraud detection can be developed with limited investment, but substantial impact.
This begs the question, given all the SNAP fraud funnels through electronic EBT payments overseen by private vendors, why aren’t these private vendors dedicating teams of experts to data mining and fraud detection like any other credit or debit card in the world? It stumps us.
Boston Decision is Boston’s leading predictive modeling firm (#1 on Google and #1 on Yahoo). If you have a question about EBT fraud detection or would like help implementing a data mining fraud detection capability, contact us at www.bostondecision.com.