Institute for Software Research
School of Computer Science, Carnegie Mellon University


Predicting intentional Tax Error Using
Open Source Literature and Data

Ju-Sung Lee, Kathleen M. Carley

November 2009


Center for the Computational Analysis of Social and Organizational Systems
CASOS Technical Report

Keywords: Tax evasion, non-compliance, intentional error, meta analysis

Intentional non-compliance in providing accurate income tax returns, also known as "tax evasion" or "intentional error", has been studied from both attitudinal and socio-demographic perspectives. A significant portion of previous research employs a common set of indicators, which we can exploit by pooling meta-analytically with the hopes of obtaining a unified, well-predicting model of intentional error. Towards this end, we turn to a large, nationally representative data source, namely the Census Bureau's Public-Use Microdata Samples (PUMS), as our source of covariance between the socio-demographic covariates of interest. Additionally, the same source offers data on potential opportunities of evasion for each PUMS respondent (or agent), in certain line item/taxpayer categories, allowing us to construct distinct error models for these categories. Furthermore, we extend the error model to include attitudinal meta-analysis, by linking the General Social Survey (GSS) to the PUMS through imputation of a GSS covariate that identifies respondents who are more likely to break the law. Our meta-analysis requires an in-depth re-analysis of the selection of previously published results on non-compliance. The result is a comprehensive model of non-compliance that fits historical, published data and that can be applied generically and to specific tax issues.

97 pages

Return to: SCS Technical Report Collection
School of Computer Science homepage

This page maintained by