Risk-Based Collection Model Development and Testing

The IRS Strategic Plan in part calls for “increasing compliance among small business and self-employed taxpayers.” In 2000, an SB/SE (Small Business/Self-Employed) Design Team report focused on the need for the IRS to do the following: adopt an integrated compliance strategy and shift the emphasis toward risk-based compliance; include profiling major customer segments; and develop multifunctional treatment strategies in order to change compliance behavior patterns. A risk-based model prioritizes collection cases by risk of nonpayment. A key component in this Collection Reengineering process was the formation of a Collection Strategy Team to identify potential improvements in the collection process and to suggest treatments. One of the objectives developed by the team was the use of predictive models to characterize aspects of the open SB/SE collection modules. The models would indicate a higher probability of a productive closure and conversely a low probability of a negative resolution. Predictive models are used by financial institutions, underwriters, and credit card companies to assess credit risk and collectibility of accounts. The SB/SE Research staff was asked to conduct a modeling effort to develop such a system. Model filters were identified for those modules with a balance due that are likely to be unproductive “CNC” (Currently Not Collectible) or productive “FP” (Full Pay).1 Accounts that have been routed through both filters, but cannot be identified as either CNC or FP criteria, are designated as “Other” accounts. Several benefits can be expected by routing cases to the most effective treatment: 2 Through early intervention, cases that would otherwise digress from potential Full Pay into a CNC can be treated and cured.