NCJ Number
85332
Date Published
1981
Length
17 pages
Annotation
This essay clarifies the analysis of panel data, observations or measurements collected for different units of analysis at two or more points in time, so that the usefulness of the panel model in criminological research can be seen.
Abstract
A panel model specifies explicitly the mathematical structure of the relationships among the variables for which data are collected in a panel study. Panel studies are appearing with increasing frequency in the criminological literature, notably in the study of delinquency, the study of the impact of arrests on crime rates, and in examining the deterrent effect of imprisonment on crime. Despite the growing popularity of panel studies, a great deal of confusion remains about the analysis of panel data. This confusion has led to poorly conceived research decisions, findings based on estimation procedures that cannot be justified by statistical theory, and failure to extract as much information from the data as is present. Panel models do not eliminate the need to make assumptions about the magnitudes of certain causal forces to identify a model as a whole. A choice of models should be based on the plausibility of the assumptions needed to identify the remaining free parameters. With additional waves of data, it is easy to demonstrate that multiwave panel models are useful, because they can be identified with assumptions far less restrictive than in either the two-wave or the cross-sectional models. This is shown by using the work of Greenberg et al. (1979) on the relationship between crime rates and arrest rates. Tabular data and mathematical equations are provided, along with 12 references.