Internal Revenue Service Advanced Analytics
Helping the IRS Close the Tax Gap with Advanced Analytics
Challenge
Recent estimates suggest that the annual “Tax Gap” in the United States is approaching $500 billion. This amount—which represents the difference between the amount of tax owed to the IRS and the amount that is paid voluntarily and on time—reflects various types of noncompliance, including taxpayers who file an inaccurate return, don’t file at all, or fail to pay by the deadline. To recoup as much revenue as possible, the IRS implements a multifaceted, data-driven compliance strategy to identify and treat causes of the Tax Gap.
Solution
Voyatek’s team of data scientists, engineers and analytics experts have partnered with the IRS on a variety of compliance programs to measure and reduce the Tax Gap—leveraging capabilities in program evaluation and analysis, taxpayer behavior analysis, data engineering, artificial intelligence, and machine learning to help the IRS increase compliance and optimize revenue collection. For example:
- Behavioral analytics and risk profiles help auditors, agents and investigators identify high-value cases and select the best enforcement strategy
- Machine learning and advanced analytics detect and prevent fraudulent returns and identity theft
- Performance monitoring tools measure the effectiveness of enforcement activities and automatically disseminate information to internal and external stakeholders
- Research and survey programs examine how taxpayers interact with the IRS and inform the design of improved taxpayer experience
- Internal program evaluation engagements improve internal operations and support the modernization of the IRS organization
Outcomes
Voyatek has developed risk-scoring models used by the IRS to evaluate more than 100 million tax returns each year. These models are a critical part of IRS’s ongoing efforts to close the tax gap. In addition to increasing revenue protection, Voyatek’s machine learning models alert stakeholders about newly identified schemes, emerging threats, and other fraud that has evaded existing detection methods. Recent outcomes include:
- Prevention of more than $10 billion in fraudulent refund payments
- Development of graph-based neural network models to identify tax returns submitted by organized criminal actors
- A 70% decline in identity theft over the last 5 years, with 95% of observable identify theft prevented
- Rapid deployment of algorithms to detect and prevent the fraudulent release of COVID-19 related benefits
