Skip to main content

Read our report on six communities’ experiences with pandemic funding and programs, which provides valuable lessons learned to improve federal emergency response programs.

<Case Study>

We created a risk scoring model that rates CARES Act state and local spending.

PACE data scientists helped the Department of Treasury (Treasury) Office of Inspector General (OIG) on a project involving the Coronavirus Relief Fund (CRF). Under this program, over 960 prime recipients received $150 billion directly from the federal government. These prime recipients paid some of the money they received to sub-recipients.  

Given the large number of primes and subrecipients—about 79,000 in all—advanced data analytics techniques were essential for streamlining and elevating Treasury OIG’s review process.  

Our Solution: 

PACE data scientists developed a risk scoring model to prioritize the risks associated with prime and sub-recipients using an ensemble technique This technique aims to improve the accuracy of results by combining multiple models instead of relying on a single model alone. . They used CRF data and other third-party datasets to create a risk model incorporating 27 risk indicators, calculating four different risk scores (risk indicator score, risk amount score, sub propagation score, and a final composite risk score calculated using the first three scores) to help prioritize reviews.  

Why this Matters: 

The CRF Risk Model dashboard is now available for 73 Treasury OIG users. The dashboard provides clear insights with easy-to-navigate summary and detail pages. Users are also able to export data for desk reviews. As of April 2022, Treasury OIG has opened 31 new desk reviews using the insights from the CRF Risk Model dashboard. We have also shared the entity resolution scripts with eight OIGs and the Government Accountability Office for their own data analytics work in pandemic programs across the government. 


Page last modified: 11/06/2023
Thank you for your feedback!
Would you tell us more? Feedback
Was this page helpful?