Abstract
The Statistics of Income Panel of Individual Returns is a simple random sample of tax returns that is maintained by the University of Michigan as a part of the Ernst and Young Tax Research Database. Sampling errors during the period 1982–1984 resulted in the inclusion of 66 spurious returns in the Panel. These spurious returns were selected for the Individual Tax Model files, which are stratified random samples. Traditional estimators assume simple random sampling, so inferences about population parameters based on Panel statistics are biased if the dependent variable is related to the stratification. This paper describes the sampling procedures and the impact these spurious returns have on the panel and suggests a weighting procedure that allows the retention of these returns in sample statistics.
Get full access to this article
View all access options for this article.