Electronic Benefits Transfer and Supplemental Nutrition Assistance Programs are taxpayer-funded benefits programs that ensure sure children and families receive the nutrition they need. However, the Congressional Research Service estimates that $1.1 billion in benefits are trafficked fraudulently each year in the US, undercutting public trust.
In 2017, South Carolina reported about 855,000 residents had received taxpayer-funded SNAP benefits totaling more than $1.2 billion per year. South Carolina Attorney General Alan Wilson announced $610,975.97 in restitution from cases that determined benefits fraud since Jan. 5, 2015. This is just one state that realized the extent to which EBT and SNAP are being misused, alongside sales tax and income tax fraud.
In fiscal year 2021, the federal government spent about $111 billion on SNAP and other related food assistance programs. Addressing benefits trafficking at the state level would save billions of dollars per year in tax-supported food assistance. Benefits trafficking is often associated with other types of tax evasion. A big data analytics approach leveraging sales tax filings, business registrations, and publicly available data resources can be used to detect benefits trafficking and associated sales tax and income tax abuse.
Benefits Trafficking
Trafficking includes the sale of SNAP benefits for cash, which is a major abuse of public resources and criminal activity. In many cases, businesses accepting SNAP are entitled to claim sales tax exemptions of up to 100% of the value of the items purchased with benefits.
SNAP benefits trafficking hurts both federal and state funding for the program. It impacts sales tax revenue due to the noncompliant business operator pocketing cash, as they charge these transactions to the SNAP benefits card while claiming sales tax exemptions on the purchase. The activities of businesses that are trafficking SNAP benefits are detectable and exhibit telltale evidence of noncompliance when advanced analytics methods are applied.
Multiple news reports on SNAP fraud looked at the successful prosecution of business operators engaged in trafficking. These activities leave distinct fingerprints in periodic tax filings and can be identified using a combination of methods such as unsupervised and supervised learning algorithms.
A modern approach—leveraging big data and sophisticated analytics techniques that would assess 100% of businesses—is necessary. This approach would have to provide strong evidence to support enforcement and prosecution of businesses’ fraudulent activity.
EBT fraud takes place in a variety of ways. A person operating a cash register will offer cash to purchase a customer’s benefits card. The employee or proprietor could offer 25 cents on the dollar, buying the card for $25 cash that now can be used to buy tobacco and alcohol. Through the rest of the month, this card can be swiped to cover cash purchases on eligible items, allowing the fraudster to pocket the cash.
Many products available for purchase with SNAP benefits are tax-exempt, meaning the fraudster is committing tax fraud and possibly income tax evasion while committing benefits fraud. There are multiple ways to accomplish this fraud, but this can be more common among smaller grocers and corner markets that accept benefits.
Often, the only way for this scheme to be uncovered is by a whistleblower or random audit selection. However, big data and analytics approaches can compare similar business segments, identifying outliers that exhibit particular patterns. Approaching this problem as a big data and analytics exercise allows analytical and computational expertise to be leveraged in a way that reduces the reliance on random audit selection and whistleblowers.
An Advanced Analytics Approach
A big data and cloud computing approach to the problem of benefits abuse enables 100% of the data to be analyzed more thoroughly than methods being used today. Cloud computing eliminates the huge expense to scale up computing power that typically would be cost prohibitive with legacy hardware-based server systems.
Data can be segmented along business type and size before being analyzed. Engineered features can be added along with outlier and novelty analysis. The full data picture is now used to identify the businesses that exhibit patterns that could be associated with fraud. Novelty detection and outlier analysis can be done in sophisticated ways to highlight patterns that are only available when examining all data within a segment. With high accuracy, these records can be flagged for an audit.
The audit process that requires expert personnel would now receive selections that are far more likely to result in positive findings. Processes that created unacceptable false positive rates and wasted time and resources could be audited more thoroughly, more accurately, and with less waste from false positive audit selections.
Combating benefits fraud by combining sophisticated analytics methods, rather than random selection and whistleblower reports, leads to the improving algorithms that highlight undetected fraud types and early recognition of emerging fraud schemes. Automation and advanced analytics techniques rely on computer systems to do what they do best and crunch data that reduces the upfront burden of audit selection and empowers analysts and auditors to do what humans do best: think creatively and follow leads. Machine learning-enhanced algorithms and statistical methods identify patterns in the data that become profiles of fraud activities, growing in accuracy and efficiency as cases are validated and returned to the pipeline as training data.
Reduce Benefits Fraud
A modern approach to benefit fraud analytics would drastically improve the effectiveness of human auditors by relying on them to do the types of analytics that require human ingenuity. Benefits fraud often relies on individuals working together to defraud the public, resulting in techniques that are more difficult to detect due to their decentralized nature, spreading the culpability and paper trail among multiple people.
Leveraging big data analytics in cloud systems designed for computational scalability opens the door to powerful analytics to solve problems that were previously too large to solve in this way. Human auditors can fully focus on the work that requires their specialized skills the most.
This article does not necessarily reflect the opinion of Bloomberg Industry Group, Inc., the publisher of Bloomberg Law and Bloomberg Tax, or its owners.
Author Information
Steven Purcell is a data scientist and manager at G2Lytics LLC, researching big data analytics and machine learning applications to detect and prevent tax fraud.
We’d love to hear your smart, original take: Write for us.