'A Significant Wave of Modernization' Coming to Unemployment Systems

In February 2020, just weeks before the World Health Organization declared Covid-19 a pandemic, 3,000 Kansans filed unemployment claims. By the end of March, that number jumped to 66,000 and continued to skyrocket over the next year.

Approximately 56% of those claims were potentially fraudulent.

The state paid about 30% of the probable bogus claims—to the tune of $700 million—but caught onto the other 70% before the fraudsters collected any money, potentially saving up to $2 billion. And now, at the behest of the state legislature, the Kansas Department of Labor is overhauling the antiquated computer system that processes its unemployment insurance claims.

“I wouldn’t say it’s over,” Matt Etzel, a principal auditor with the Kansas Legislative Division of Post Audit, said of the increasing and increasingly sophisticated attempts to defraud federal and state governments through phony unemployment insurance claims.

Kansas isn’t the only state committed to modernizing the 1970s- and ’80s-era code that runs many UI networks, redoubling the use of data analytics to spot telltale patterns in fraudulent claims, and investing in artificial intelligence and other state-of-the-art technology that might outsmart—or at least outpace—those who would cheat the system.

State-by-state forensic work will uncover just how much money was lost to UI fraud during the worst of the pandemic, although a conservative estimate from the U.S. Department of Labor’s inspector general put it at $87 billion-plus nationwide by September 2021, when the federal government stopped adding $600 a week to UI checks.

Meanwhile, states with auditors and lawmakers who believe future disasters could invite a repeat of the record-setting fraud will invest millions of dollars in technology to prevent it, predicted Shaun Barry, director of fraud and security intelligence for government and health care for data analytics vendor SAS.

“Many of them use … four to five generations of technology removed from state-of-the-art today,” Barry said. “Those systems are fragile and brittle and so old that they don’t have the robust accounting controls that are available today. … States have known for many years that they have fragile [information technology] systems.”

Over the next five years, Barry said, state UI agencies will embark on “a significant wave of modernization in accounting systems. The pandemic has accelerated this modernization.”

It All Started With the Pandemic 

In Kansas and elsewhere, in fact, the pandemic got it started.

To determine how much UI fraud occurred during the 2020 and 2021 months at the heart of the pandemic, Kansas auditors dissected 1,000 claims—out of the million-plus it had processed—to detect patterns that could reveal fraud. What they found were multiple claimants using the same complex password; partially duplicated email addresses; and claims supposedly from state employees whose Social Security numbers did not match human resources records, among other red flags.

In the end, auditors identified 26 indicators of potential fraud, and then “trained” a neural network—a computer modeled after the human brain—to spot them.

“It got really good at replicating our ability,” said Etzel, who noted that the computer made the same decision as the auditors did at least 91% of the time.

Barry said state systems need to overhaul their outdated equipment with that kind of top-shelf technology to prevent future fraud from occurring during another crisis.

Even the work of organized criminal rings of fraudsters, he said, “can be addressed by modernized accounting systems.” He pointed to the success of commercial credit card companies that use artificial intelligence like machine learning, which involves automated systems like the one in Kansas that mimic the decision-making of the human brain, to flag suspicious claims.

As far back as 2013, a U.S. News & World Report survey identified 20 states that were at least dabbling in artificial intelligence to analyze UI claims.

Meanwhile, many states have ramped up their use of data analytics—collecting and analyzing large stores of data to identify suspicious patterns among UI claims. And some are cross-referencing their data with that of the federal government and nearby states to catch serial fraudsters who cross state lines, Barry said.

For example, California Gov. Gavin Newsom in October signed a law requiring the California prison system to share the names and Social Security numbers of inmates—who are not eligible for UI—with the Employment Development Department and ordered cross checking.

Six states—Georgia, Utah, New Jersey, Colorado and Ohio—have created an online data exchange system to share UI claim information. The Alabama Department of Labor, like several in other states, recently rolled out a dual-authentication system that helps to protect the identity of UI claimants.

Paradoxically, Oregon auditors credit the state’s ultra-low UI fraud with its outdated computer system.

“It doesn’t allow for online web claims like in California and Washington,” Ian Green, audit manager for the Secretary of State Audits Division. “It doesn’t immediately process stuff. It’s a lot more manual and cumbersome, so if you tried to defraud in Oregon, you hit roadblocks … in our old system. In California, you’d get a payment the next day.”

Green estimated the state lost “hundreds of thousands of dollars [to fraud during the pandemic] instead of in the hundreds of millions of dollars.”

The downside, he pointed out, is that legitimate claimants often wait weeks or even months to get their checks.

Green’s colleague, Senior Auditor Kathy Davis, called the reduced fraud “a silver lining to a really unfortunate situation.” 

Still, Oregon will update its systems by 2024, Green said.

Optimists Versus Pessimists

Barry predicted that some states will shy away from spending hundreds of thousands of dollars—or even millions—on cutting-edge computer systems.

“There are two competing narratives that are emerging on what the future holds for labor agencies,” he said. “One is an optimistic view and the other is pessimistic.”

The optimists, he said, will view pandemic-induced UI fraud as a one-time hit that will disappear if and when Covid-19 does. The pessimists, on the other hand, will expect a renewed surge of fraud with every future crisis.

“It looks like it will be 50-50 between optimists and pessimists,” he said. “But if there’s anything that state UI agencies agree with, it’s that their UI system is antiquated and needs to be upgraded. But do I replace it with what I had before, or do I replace it and enhance it with UI defenses?”