NCJ Number
100350
Date Published
1980
Length
263 pages
Annotation
An analysis of the implications of the use of imperfect data in research on deterrence considers data on Federal income tax violations and concludes that measurement errors produce serious problems in estimating deterrent effects using current macrostatistical models.
Abstract
The results of estimates can be quite sensitive to moderate levels of random or constant error in the measures of offense or sanction rates. Substantial estimation errors can arise regardless of whether single equation or simultaneous equation deterrence models are used. Procedures such as sensitivity analyses and diagnostic plots are needed to help uncover potential error bias. Analysis of data on income tax violations shows several other problems in correctly specifying models for estimating deterrence effects. Using individual data produces associations opposite to those found when aggregate data are used. This finding makes it inadvisable to infer individual behavior on the basis of observed aggregate relationships. In addition, the failure of current deterrence models to control for differences in offense seriousness across jurisdictions or time periods also leads to specification bias. Finally, it is preferable to use indexes of enforcement resources that are standardized by offense counts rather than population counts. Figures, tables, footnotes, and 318 references. (Author abstract modified)