Every week the U.S. Department of Justice issues news releases announcing the prosecution of companies and individuals for all kinds of federal crimes. Often the crimes involve fraud – people who filed claims for benefits to which they were not entitled, or companies that filed for payment of services they didn’t render.
The federal government spends a lot of its resources, both human and financial, trying to “claw back” these improper payments and punish those who tried to cheat the system. Many agencies also have prevention of fraud as a key part of their core mission.
One such example is the U.S. Department of Agriculture’s Risk Management Agency (RMA). In conjunction with the Federal Crop Insurance Corporation, RMA’s primary responsibility is to provide risk protection through crop insurance for American farmers. But where there is insurance, there is the possibility of insurance fraud.
The Center for Agribusiness Excellence (CAE) at Tarleton State University in Stephenville, Texas, provides data mining and analysis for RMA, in part to root out fraud and avoid improper payments.
“CAE is the thought leader for RMA. They identify additional, impactful analytics for RMA to leverage,” said William Damico, senior solutions architect, Teradata, which provides the industry’s only pervasive data intelligence platform. “It’s a very cost-effective program; the cost avoidance in payments has far surpassed the cost of the analytic environment.”
The advanced analytics mine data from some 170 sources, including satellite images and weather data for more than 3,200 counties across the country, with algorithms to match that wealth of information against crop insurance claims.
“It’s a pretty stable number of data sources, but the sizes of the databases continue to grow,” Damico said. “They are executing queries against upwards of 100TB of data, with more data being added to the system every day.”
The use of geospatial data is four-dimensional, he pointed out – the passage of time changes the data. In an uneventful growing season, a particular area might only need satellite imaging a handful of times. But in the event of disasters – hurricanes and flooding, tornadoes, wildfires, and drought, to name a few – more frequent imaging is needed to identify affected acreage and crops, as well as to track recovery.
There are many kinds of fraud the analytics can detect. For example, there’s “crop shifting” fraud. An Iowa farmer was convicted in November of this year and ordered to pay back more than $30,000 for this type of scheme.
“Crop shifting occurs when a farmer underreports production in a field in order to reach the percentage of loss required to receive an indemnity from their crop insurance policy,” DoJ reported. This farmer “‘shifted’ production from one farm to another by reporting that several thousand bushels of grain were harvested from one farm, when he knew that those bushels were in fact harvested from the other farm. By reporting the bushels of grain from the incorrect farm, or shifting the production, [he] received insurance indemnities to which he was not entitled.”
This is where the use of advanced analytics comes in. By analyzing crop yields, weather conditions, acreage, and a host of other factors, analytics can flag an anomalous insurance claim.
The data also can identify approved crop insurance providers (AIPs), the middlemen in the system, which may be submitting fraudulent claims without farmers’ knowledge.
According to the 2018 President’s Budget for RMA, the 2016 improper payment rate for crop insurance was just 2.02 percent, below the target rate of 2.21 percent set by the Office of Management and Budget. By comparison, in 2014 the improper payment rate was about 6 percent. In its budget, the agency planned to expand its analysis in order to be able to determine improper payment rates for every individual insurance company.
Cost avoidance is much less expensive than cost recovery, of course, and this is another area where the analytics have helped.
“Rather than pay, then have to chase fraudulent claims, RMA is using advanced analytics to determine the likelihood of a claim being fraudulent in the first case,” Damico said. “It’s easier to not pay up front. In this case, RMA’s cost avoidance certainly is over a billion dollars” since the analytics program was initiated.
Using analytics isn’t just about preventing fraud. It also can be used to make sure that claims aren’t denied erroneously.
“There were two adjacent farms that filed claims for hail damage, but weather reports for that county didn’t report any hail, so the claims were flagged as anomalous,” he said. “But when RMA took a more granular look at the weather data, it turned out there had been a microburst right there. The farmers’ claims were paid.”