Social Network Analysis and the government fraudster


Detecting and preventing fraud isn't easy.  Many tools and techniques are required, and while the digital age and advanced technologies have made fraud much easier in many ways, those same technologies have opened up new ways to combat it.  One of those comes in the form of social network analysis, or network analytics.  This augments and speeds up what investigators have been doing for a long time, which is link analysis.  Making connections between individuals to find a ring of criminals, or determine connections that shouldn't exist, and if they do, trip the hand of fraud versus legitimate claims.  You've probably seen the old-school version of this dozens of times in movies or cop shows, with photos of gangsters taped up to a bulletin board, with the lines connecting them.

Now imagine having that network built out for the investigator at the time they first receive the lead to investigate.  Better yet, there isn't just a line drawn to show the connection, but details as to how they are connected, such as shared addresses, phone numbers, bank accounts, or maybe just the same shady attorney.  The time that removes from manual work for the investigator is incredible, and in many government programs, the information to make that connection is trapped within their own data.  Other times, it comes from connecting data points from other government programs through information sharing agreements.  A third approach is utilizing text mining and running analytics on free text on everything from free-form case notes by adjudicators or case managers to doctor's reports to Facebook postings.

Now let's turn that from concept to reality to show an example.  Here is a recent posting from a blog called Nailed, which is run from Washington State's Department of Labor and Industries.  It covers topics such as workers' compensation fraud, wage theft and construction contractor fraud.  (Full disclosure: This used to be my blog when I was with Washington State, and is currently run by my successor)  The story is about Mr. Bonilla, who had his workers' compensation coveraged revoked for failure to pay.  Continuing to employ workers after revocation in Washington State is a felony.  From here, I'll let Doric Olson and the Nailed blog tell the story:

L&I investigators learned that one of those employees, under Mr. Bonilla’s guidance, obtained a contractor’s license, opened her own painting business, and hired Mr. Bonilla as a foreman. She understood that Bonilla would teach her the business and use his contacts for jobs.

Bonilla took advantage of that relationship to use the company van and equipment, without the owner’s knowledge, to complete jobs he had begun under his own revoked business.

Because his own contractor registration and workers’ comp account were revoked, Bonilla then had his sister impersonate his employer to set up jobs that Mr. Bonilla would then do without the actual owner’s involvement.

The good news is that an inspector caught him on the jobsite, and this resulted in a conviction in June of 2012.  The bad news is that was a manual process at the time this case started in 2009.  Performing analytics on Mr. Bonilla's network would have shown the interesting relationship when he became an employee of his former worker.  While that isn't unheard of, it is rare, and combined with the problems he was having with his own account, would set off very high fraud scores due to the unlikeliness it is a real situation, rather than perpetration of fraud.  Armed with that knowledge, the agency could have reached out right away and started action.

Another good example is L0s Angeles County's efforts to fight child care benefits fraud. The County's Data Mining Solution delivered more than 200 fraud referrals, and the use of social network analysis uncovered two conspiracy rings comprising 16 cases. You can learn more about LA County's work and many international examples of fraud fighting here.

What information is trapped within your own data?  How could those relationships, possibly even with internal staff, begin to uncover fraud, waste and abuse?  Down the line, we'll look at other cases, and see how proactive approaches and good analytics could shape better outcomes.


About Author

Carl Hammersburg

Manager, Government and Healthcare Risk and Fraud

Carl Hammersburg manages the SAS Government and Healthcare Risk and Fraud team, and has been with SAS since 2012. Prior to that, he spent 20 years in anti-fraud activities for Washington State’s exclusive workers’ comp insurer, the Department of Labor and Industries. In 2004, Carl formed that agency’s comprehensive fraud program, covering tax and premium audit, claim investigation, provider fraud and collections. Data sharing and investigative partnerships with other State and Federal agencies, as well as driving public availability of information and awareness served as cornerstones to the anti-fraud activities of the program. During his stewardship, audit and investigative activities doubled and outcomes tripled, based on a focus on data mining and predictive analytics that improved efficiency and case selection. Program success under Carl’s leadership resulted in awards from two successive Governors of Washington State.

1 Comment

  1. Pingback: Will the last employee please turn off the lights?

Leave A Reply

Back to Top