Michigan MiDAS unemployment insurance fraud detection

Released: October 2013

MiDAS (Michigan Integrated Data Automated System) is an automated information system used by Michigan's Unemployment Insurance Agency (UIA) to collect unemployment taxes from employers and to pay unemployment insurance benefits to eligible claimants.

A new MiDAS system was introduced in 2013 to ensure that unemployment checks went only to people who deserved them, increase UIA’s efficiency and responsiveness to unemployment claims, and reduce UIA’s operational costs by cutting over 400 workers.

Devised, developed and implemented by software companies SAS Institute, Fast Enterprises, and CSG Government Solutions, the new system searched records of employers and claimants in its database, then flagged people for potential unemployment fraud. 

It then sent questionnaires to an email address on the UIA website that 'recipients may not have had reason to monitor', gave them 10 days to respond, and sent a letter informing them they had been charged with fraud. After a 30-day appeal period, the system began dispersing wage and tax refunds, according to Undark.

System databank

Operator: Michigan Unemployment Insurance Agency (UIA)
Developer: SAS Institute; Fast Enterprises; CSG Government Solutions
Country: USA
Sector: Govt - welfare
Purpose: Detect benefits fraud; Notify claimants
Technology: Fraud detection system  
Issue: Accuracy/reliability; Bias/discrimination - economic; Effectiveness/value; Fairness; Oversight/review; Ownership/accountability; Risk management
Transparency: Governance; Black box; Complaints/appeals; Design/usability; Legal 

Wrongful 'robo-adjucation'

Between October 2013 and September 2015, it was discovered that over 34,000 people had been wrongfully accused of unemployment fraud by the system, resulting in significant anxiety, financial loss, bankruptcy, houses foreclosed, homelessness, and reputational damage.

Per risk expert Robert Charette writing for the IEEE, the fraud and fines imposed generated huge amounts of money for the UIA, increasing its coffers from around USD 3 million to over USD 69 million in a little more than a year. 

It was also found (pdf) that a large number of fraud accusations had been algorithmically generated by MiDAS with no human intervention or review of the accusation.

Under political and public pressure, UIA admitted that there had been problems with MiDAS, notably its 'robo-adjudication' process and the lack of human review. 

Legal shenanigans

In January 2017, a lawsuit filed in September 2015 in the Michigan Court of Claims concluded that 85 percent of 40,195 cases of fraud were incorrect, and that another 22,589 cases with some degree of human intervention involved in a fraud determination found a 44 percent false fraud claim rate.

The UIA strongly fought all legal actions, and 'stonewalled all attempts to discover the depth, breadth, and reasons behind the fraudulent fraud accusations', according to Charette.

Legal, regulatory

Investigations, assessments, audits

News, commentary, analysis