Estimating the Number of Tax Returns Required to Be Filed: Nonparametric Methods to Address Respondent Rounding

• Activity Number: 64 - Nonparametric Modeling of Survey Data
• Type: Topic Contributed
• Date/Time: Sunday, July 30, 2017 : 4:00 PM to 5:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Minsun Riddles* and Sharon Lohr and J. Michael Brick and Patrick T. Langetieg and J. Mark Payne and Alan H. Plumley
• Companies: Westat and Westat and Westat and Internal Revenue Service and Internal Revenue Service and Internal Revenue Service
• Keywords: Heaping ; Imputation ; Income distribution ; Kernel density estimation ; Maximum likelihood ; Measurement errors

Abstract

The Internal Revenue Service uses the Current Population Survey (CPS) to estimate the number of individual income tax returns that are required to be filed. The filing requirement depends on filing status, age, and income levels; in 2015, for example, a single person under age 65 was required to file a tax return if his or her gross income was above $10,300. Many CPS respondents, however, report values for wage and other types of income that are rounded. A respondent reporting $10,000 in wage income may have rounded his or her data and may in reality have had wage income above the filing threshold. We compare three methods for addressing respondent rounding: parametric density estimation with imputation, kernel smoothing, and weight redistribution. We perform a simulation study to investigate the methods under a variety of "true" distributions and also apply the methods to public use data from the CPS.