Neural Network Analysis of the Employee Classification Problem for Tax Purposes

Thumbnail Image
Official URL
Full text at PDC
Publication Date
Advisors (or tutors)
Journal Title
Journal ISSN
Volume Title
Facultad de Ciencias Económicas y Empresariales. Instituto Complutense de Análisis Económico (ICAE)
Google Scholar
Research Projects
Organizational Units
Journal Issue
Since 1987 the U.S. Internal Revenue Service has relied on twenty common law factors for guidance in determining whether a worker is an emplyoee or an independent contractor. This study presents new evidence on the task of simplifying that complex classification problem. Neural network methodology is used to classify workers using data obtained from Private Letter Rulings issued by the Internal Revenue Service from 1988 through a portion of 1993, a data set not previously used for this purpose. The model is highly accurate in correctly classifying workers as either employees or independent contractors. The overall prediction success rate using sample data was 97.2 percent and drops to 91.4 percent when a holdout sample was used. These findings are robust for each of the years in the study. For comparison purposes, classification results using logistic regression are also included. Results from both methodologies are identical.
Desde 1987 el Servicio de Recaudación Interna de los Estados Unidos ha confiado en veinte factores definidos por ley para guiarse en la clasificación de empleados y trabajadores independientes. Este estudio presenta nueva evidencia para simplificar el complejo problema de dicha clasificación. Mediante el uso de redes neuronales (neural networks), la clasificación se realiza utilizando declaraciones (private Letter Rulings) del Internal Revenue Service desde 1988, hasta una porción de 1993, un banco de datos no utilizado hasta el momento. El modelo es altamente preciso en la clasificación de trabajadores como empleados o trabajadores independientes. El porcentaje de predicciones correctas es de un 97.2 %, Y cae al 91.4% para valores fuera de la muestra. Estos resultados son robustos para cada uno de los años incluídos en el estudio. A los efectos de comparación, también se incluyen resultados de la clasificación utilizando regresión logística. Ambas metodologías producen idénticos resultados.
UCM subjects
Unesco subjects
Belsley. D.A., E. Kuh and R.E. Welsch (1980): Regression Diagnostics: Identifying Influential Data and Sources of Collinearity. John Wiley & Sons, Inc. Church. Pamela H. and Kenneth R. Lambert (1993): "Employee or Independent Contractor?" Management Accounting, 74 (June): 52-55. Coats, Pamela K. and L. Franklin Fant (1993): "Recognizing Financial Distress Patterns Using A Neural Network Tool," Financial Management, 22, 156-162. Collins, E., S. Ghosh, and C. Scofield (1988): "An application of a multiple neuralnetwork learning system to emulation of mortgage underwriting judgements," Proceedings of the IEEE International Conference on Neural Networks. 459-466. Cormmerce Clearing House (1988-93): IRS Letter Rulings Reporter. Commerce Clearing House, Chicago. Microfiche. Dutta, S. And S. Shekhar (1988): "Bond-rating: A non-conservative application of neural networks," Proceedings of the IEEE International Conference on Neural Networks, 443-450. Eynon, G. And K. Stevens (1995): "The Economics of Taxpayer Venue Choice: Effects of Informational Asymmetries." The Journal of the American Taxation Association. Spring, 71-94. Frank, Barry H. (1989): "Independent Contractor vs. Employee: Guidelines for the Practitioner." The Practical Accountant, 22 (December): 17-31. Gold, P.A., and M.D. Esposito (1992): "The Right To Control: Are Your Workers Independent Contractors or Employees," Compensation and Benefits Review, 24 (July-August): 30-37. Green. Gary L. (1993): "Tax Talk", The National Public Accountant, 38 (March): 34-36. Hosmer, D.W. and S. Lemeshow (1989): Applied Logistic Regression. John Wiley & Sons. lne. Hulen. Myron, William Kenny. Jack Robison and Michael Vaughn (1993): "Independent Contractors: Compliance And Classification Issue," in an unpublished White Paper for the American Tax Policy Institute, p. V-7. Kenny. W. and M. Hulen {1989j: "Determining Employee or Independent Contractor Status." The Tax Adviser, 20 (October): 661-669. Maren, Alianna, Craig Harston and Robert Pap (1990): Handbook of Neural Computing Applicatwns. Academic Press. Merritt, J.E. (1991): "Independent Contractor vs. Employee Issues," Twenty-Sixth Annual Conference of The Southern Federal Tax Institute, Atlanta, Ga., October 2. Moore. Walter B. and Mark Turk (1993): "Doggone Dangerous!" The National Public Accountant, 38 (July): 14-15, 34-36. Murphy. M.J. (1989); Acting Commissioner, Internal Revenue Service, in testimony before the House Subcommittee On Commeree, Consumer and Monetary Affairs. Washington, D.C., May 16. O'Rourke, T.J. (1990): "Independent Contractor Or Employee," Management Accounting, 72 (Deeember): 31-36. Porcano, J.L. and T.M. Porcano (1985): "Incorporating Probability Analysis in Taxpayer Appeal Decisions." The Journal of the American Taxation Association, Spring. 18-36. Posner, P.L. (1989): Associate Director, Tax Policy and Administration Issues, General Government Division, General Accounting Office, in testimony before the House Subcommittee On Commerce, Consumer, and Monetary Affairs, Washington, D.C., May 16. Salchenberger, Linda M., E. Mine Cinar and Nicholas A. Lash (1992): "Neural Networks: A New Tool for Predicting Thrift Failures." Decision Sciences, 23, July/ August, 899-916. SAS Institute Inc., SAS/STAT User's Guide, Version 6, Fourth Edition, Vol. 2, Cary, NC: SAS Institute Inc., 1989. p. 846. Stewart, D.N. (1982): "Use Of Logit Analysis To Determine Employment Status For Tax Purposes," Journal of The American Taxation Association, 4 (Summer): 5-12. Sumutka, A.R. (1992): "Employee or Independent Contractor?", The CPA Journal, 62 (April), 48-55. U.S. Department of Labor, Bureau of Labor Statistics, 1985-1991. White, H. (1989): "An Additional Hidden Unit Test for Neglected Nonlinearity in Multilayer Feedforward Networks," Proceedings of the International Joint Conference on Neural Networks, Washington. D.C.