Why Benefits Fraud Is So Hard to Detect – and How to Tangibly Improve Fraud Detection

How big a problem is fraud for the government agencies that deliver benefits to citizens? As one indicator, each year well over half of U.S. states inaccurately pay out more than 13% of their unemployment benefits as they fall victim to fraud, waste and abuse. As another, President Biden used his first State of the Union Address in March 2022 to announce appointment of a chief prosecutor to focus on fraud related to COVID-19.

We might expect large, hastily delivered programs like federal pandemic relief to involve claims fraud. But why? Especially when government organizations are increasingly digitized and have ready access to large sets of resident data, why do they have such a hard time preventing benefits fraud?

Knowing the answer can help agencies understand the challenges they face – and recognize the need to dedicate resources to identifying, reducing and proactively preventing fraud.

More Benefits, More Opportunities, More Fraud

These five primary factors have made benefits fraud difficult for agencies to root out:

1. High volumes of benefits applications

At the end of March 2020, just weeks into the COVID-19 crisis, unemployment insurance claims topped 6.6 million. It was the highest level of seasonally adjusted initial claims ever recorded. Months later, the numbers remained high, with 5.9 million filing a “continued claim” in November 2020.

But the pandemic wasn’t unique in driving a high number of claims. At the peak of the Great Recession, in March 2009, 6.5 million people filed for continued claims. In fact, any time state or local residents face unexpected need – during an economic downturn, in the wake of a natural disaster like a hurricane or an earthquake – agencies must grapple with a volume of claims they might not have the staff or compute power to manage. When that happens, routine workflows and controls can break down, and fraudsters can slip past.

 

2. Leadership-mandated speed

When large numbers of residents need relief, politicians face pressure to deliver help quickly. They often pass that pressure on to the agencies that issue the checks. To reassure voters – and in some cases, to help shore up the economy – they might even write prompt issuance of relief into law.

An example is the federal government’s American Rescue Plan. By September 2021, the U.S. Department of the Treasury had delivered more than $450 billion directly to families. That included 170 million Economic Impact Payments totaling more than $400 billion, 106 million Child Tax Credit payments totaling $46 billion, and 1 million Emergency Rental Assistance Program payments totaling more than $5 billion.

When so many checks are being cut so quickly, fraudsters recognize an opportunity. By the time an agency identifies a fraudulent claim, the payment has already been issued.

3. Easy access to personally identifiable information (PII)

In September 2017, the Equifax consumer credit reporting agency announced that a data breach had exposed the PII of 148 million people. The exfiltrated data included personal details such as first and last name, Social Security number, birth data, address and driver’s license number.

The Equifax breach is hardly an anomaly. In fact, 2021 was a record year for breaches. By September, nearly 1,300 breaches had exposed the data of tens of millions of victims.

Once such data is exposed, it can live on the dark web indefinitely. Cybercriminals can sell it on or use it to hijack identities and apply for credit cards, loans and more – including government benefits. In many cases, fraudsters can amass a complete and accurate dataset, with no need to invent details such as marital status, employer and income. To government agencies, the fraudulent benefits application looks legitimate.

4. Criminal sophistication

Fraud can take multiple forms. In the case of unemployment fraud, a claimant might submit false information to continue collecting benefits when they’re no longer eligible. An employer might falsify accounts to avoid tax liability. A criminal might defraud legitimate claimants through websites that mimic a state unemployment insurance portal.

And schemes are growing more sophisticated. In the past, a small-time fraudster working on their own might file a fake claim or two. Today, organized groups around the world have the human, financial and technology resources to access systems, acquire data and commit fraud across multiple channels. In some cases, they might submit large numbers of claims in short periods of time across multiple agencies in multiple states.

Much of this fraud can go undetected by legacy systems armed with traditional business rules and monitoring tools. Agencies need more sophisticated technology and processes to identify, continually detect and get to the root causes of benefits fraud.

5. Cloak of invisibility

Some fraudsters today are better funded than the agencies trying to stop them. They have resources, compute power and speed on their side. They can easily slip past gatekeepers. That puts agencies at a disadvantage.

When claim volumes are high, when benefits are issued quickly, and when claimant data looks legitimate, fraud becomes largely invisible. That’s equally true for fraud committed at scale. When a criminal organization submits numerous claims across numerous agencies in numerous states, it can take time for a criminal pattern to become clear. By the time fraud is suspected, multiple days have passed and many payments have been issued. The fraudsters have collected their payout, covered their tracks and moved on.

The Early Algorithm Catches the Fraud

The good news is that agencies aren’t without recourse. Data analytics, predictive modeling, risk scoring, machine learning (ML) and artificial intelligence (AI) can identify, stop and proactively prevent fraud – efficiently and at scale.

For example, effective data management can identify new sources of data, enable efficient extract-transform-load (ETL) processes, and support ongoing effective data governance. Pattern identification presented in a graphical interface can enable case analysts to identify suspicious behaviors, search graphs for past instances of the pattern, and rank resulting cases.

AI techniques such as neural networks and Beta-skeleton graphs can identify fraud schemes that aren’t revealed by human analysis. These fraud-detection models reduce the number of false positives and increase the efficiency of case analysts. Customized visualizations and dashboards enable efficient reporting by team members and right actions by agency decision-makers.

Stopping benefits fraud calls for a mindset that values an effective level of investment in fraud-prevention data and resources. It also requires the application of technology to that data to uncover and eliminate all forms of fraud, waste and abuse. The outcome is worth the investment and the effort: More government resources quickly and accurately reaching the residents who can benefit from it most.