AI Case Study
New York State increased its collections from delinquent revenue by 8% using reinforcement learning
The New York State Department of Taxation and Finance (NYS DTF) has collaborated with IBM Research and IBM Global Business Services to optimise tax collection. The developed solution is based on data analytics and the constrained Markov decision processes (C-MDP) framework. It considers taxpayer background information and collections actions by the DTF to generate a case-specific suggested course of action commensurate with each debtor. From 2009 to 2010, the State achieved an 8% increase in delinquent revenue collection, equal to $83 million, while the solution's expected benefit far exceeded the initial target reaching up to $120 to $150 million over a period of three years.
Public And Social Sector
"In a collaborative work, NYS DTF, IBM Research, and IBM Global Business Services developed a novel tax collections optimization solution to meet this chal- lenge. The solution combines data analytics and opti- mization using the unifying framework of constrained Markov decision processes (C-MDP). It optimizes collection actions to maximize long-term returns, while considering other complex dependencies."
"From 2009 to 2010, New York State increased its collections from delinquent revenue by $83 million (8 percent) using the same set of resources. Given a typical annual increase of 2 to 4 percent, the system’s expected benefit is approximately $120 to $150 million over a period of three years, far exceeding the initial target of $90 million."
"The operations research-based solution combines data analytics and optimization using the unifying framework of constrained Markov decision processes (C-MDP)."
Budgeting And Forecasting
"The New York State Department of Taxation and Finance (NYS DTF) collects over $1 billion annually in assessed delinquent taxes. The mission of DTF’s Collections and Civil Enforcement Division (CCED) is to increase collections, but to do so in a manner that respects the rights of citizens by taking actions commensurate with each debtor’s situation. CCED must accomplish this in an environment with limited resources.
New York State has always been aggressive in applying technology to facilitate collections; however, critical resource decisions about “who will work the cases and when” have essentially relied on manual rules based on gut decisions. The result was a one-size-fits-all approach that was both inefficient and inequitable in treating tax debtors. NYS DTF sought a more flexible decision model that would customize its collection activities to an individual debtor’s situation."
"Given data consisting of taxpayer background information, complete taxpayer history of transactions (i.e., payments) and actions (e.g., contact and collections actions) taken onto them by the DTF, we generate a time-stamped sequence of feature vectors at multiple sampling (or evaluation) time steps for each taxpayer case; we use these as training data in the C-RL procedure. In the deployed system, we used approximately 200 modeling features. Below, we list some examples that we have grouped into categories based on their features.
(1) Taxpayer: number of nonrestricted financial sources, sales tax inactive indicator, number of bankruptcy filings;
(2) liability: total liability balance, sum of collectible assessments, sum of assessments available to warrant;
(3) transactional: tax paid previous year, number of payments since last action, number of payments to date, sum of payments during previous year;
(4) collections: number of open perfected warrants,
days since last warrant perfected."