IRS RESPONSE TO RECOMMENDATION: In April 2016, RICS began reporting each Return Review Program Non-Identity Theft model False Detection Rate (FDR) separately and documenting the results. This information will be shared with Taxpayer Advocate Service on a periodic basis. In addition, we are currently creating baseline statistics for the IVO program and will monitor the FDR of each fraud model separately.
For clarification, our non-identity theft model reporting uses the metric of a false detection rate. A false detection rate is the number of false positives divided by the number selected. We believe the false detection rate more accurately reflects the performance of a selection model.
Update: FDR and/or performance for all filters in RRP, DDB and FRE are reviewed throughout the filing season to monitor performance and/or make necessary changes. The final review of filters for IDT and Non-IDT FY17 inventories took place on November 30, 2016. Filter performances were reviewed for identity theft work and planned performance was determined. Non-identity theft selections were discussed on December 6th based upon the November 30th assumptions, with plans for the filing season determined at that time. Filters are currently being programmed in preparation for the filing season.
CORRECTIVE ACTION: Throughout the remainder of Calendar Year (CY) 2016, the FDR metrics for the Non-IDT models will be reviewed by IRS leadership, with in-year model adjustments to be implemented where prudent to minimize the selection of falsely detected returns. At the end of CY 2016, the business team will perform a comprehensive assessment of fraud model performance, and issue recommendations for larger-scale improvement to filters, rules and models to be implemented at the start of the next Filing Season.
TAS RESPONSE: The National Taxpayer Advocate is encouraged by this new development and appreciates the IRS’s recent commitment to begin tracking Non-IDT model false detection rates. However, because the IRS just began tracking this data in April 2016, TAS is currently unable to determine if the IRS can properly identify the major factors that are causing the greater percentage of frozen legitimate refunds and the steps the IRS will take when a problem with a filter or model is identified. TAS looks forward to discussing the results with the IRS and recommends a consistent, collaborative effort moving forward.
ADOPTED, PARTIALLY ADOPTED or NOT ADOPTED: Adopted
OPEN or CLOSED: Closed
DUE DATE FOR ACTION (if left open): N/A