NCJ Number
85340
Date Published
1981
Length
23 pages
Annotation
This study discusses the methodological procedures for the modeling and analyses of autocorrelated data from intervention experiments, and they are integrated in an evaluation framework.
Abstract
The procedures and their associated computational considerations are illustrated by their application to evaluation of changes in crime rates associated with the policy intervention of gun control. Analysis of the impact upon a system's performance after an intervening action compared to its performance before this action is called intervention analysis. Data used to measure a system's behavior in such analyses often take the form of autocorrelated time series. Two methodological approaches, one that measures an instantaneous shift in the level of an autocorrelated time series and one that uses dynamic models to measure such changes over time (Box and Tiao, 1975, respectively) are useful approaches in the evaluation of change. In the former approach, the detection of a shift or change in the level of a nonstationary time series is possible. The second method determines how any shift in process level behaves after the initial change. An application of these methods to an analysis of the impact of gun control on crime rates in Boston shows a constant reduction in the activity level of gun-related offenses. This constant reduction was determined to be a permanent change in level that existed for the 2 1/2-year postintervention period. Tabular and graphic data and nine references are provided.