Analytics is at the heart of a state government department's approach to fighting tax return fraud. An added bonus is the new analytics model saves staff time.
The IRS is expecting more than 150 million tax returns to be filed this year. The organization is on high alert for tax fraud on returns, which was estimated at more than $900 million in 2015.
This has prompted state governments to tune up their tax return fraud detection. The Maryland Comptroller Office is one example of the efforts that states are making to improve fraud detection through the use of analytics.
"We incur two types of cost in the war on tax fraud," said Andrew Schaufele, Director at the Bureau of Revenue Estimates at the Maryland Comptroller of the Treasury. "The first is the cost to fight the fraud. The second is the opportunity cost."
Maryland has a team of 25-30 individuals who work on the front lines of tax fraud cases throughout the year, as well as on compliance and policy issues. They rely on advanced tax analytics that assist their efforts.
"Last year, we found $35 million in tax refund fraud, so we feel we are catching the lion's share of fraudulent tax return activity," said Schaufele. "Our analytics modeling is enabling us to do that. We would ask questions like what the tax return's refund was relative to its withholding.... Every year we would uncover new tax fraud schemes and then develop new metrics until we had about 25 different metrics."
Once potentially fraudulent returns were identified, staff had to manually review them all. "If we couldn't confirm fraud, we would let these returns through to get processed," said Schaufele. "And at the end of the day, what we were finding was that our process was only about five percent accurate, and that we were receiving only very small gains."
The state's new process, fueled by analytics, now uncovers fraudulent returns at a rate of 50 to 60% before the returns are even manually reviewed to confirm the findings. This saves staff time.
"We actually began our move to analytics as a tool in fraud detection as early as 2009," said Schaufele. "But we found that there were some steep learning curves in making the transition, and much of the data was 'black box' to us. We also transitioned off legacy systems that were not able to process the analytics, and we performed iterative testing until we felt comfortable with the process. After this, we made a monumental shift to analytics."
What Schaufele's group found was that a significant incidence of fraud was coming from local tax preparers. "We began investigating the returns. Since then, we have stopped accepting e-returns from local tax preparers," he said. "We are excited that we played a role in stopping this fraudulent activity that was taking advantage of other taxpayers."
With the advent of chip-based technology and tighter fraud detection methods for credit cards, state and federal tax bureaus are seeing crime perpetrators redirect their focus on tax returns. This makes the use of effective analytics even more critical.
"We continue to seek out new avenues of fraud that are being perpetrated as we also expand our analytics capabilities," said Schaufele. "We now are adding outside data sources to our internal data that is being used in analytics, and our business partners like Teradata continue to assist our staff and IT department with the development of analytics algorithms and the tax return scoring process."
- Identity scams are up this tax season: How to protect yourself (TechRepublic)
- Big data tools deliver significant ROI in enterprise fraud prevention efforts (TechRepublic)
- Can Big Data Analytics learn some lessons from fraud detection software? (TechRepublic)
- Quick glossary: Business intelligence and analytics (Tech Pro Research)