NCJ Number
229342
Journal
Journal of Quantitative Criminology Volume: 25 Issue: 4 Dated: December 2009 Pages: 391-417
Date Published
December 2009
Length
27 pages
Annotation
Using data from the National Longitudinal Survey of Youth for 1997 (NSLY97), this study examined whether "dynamic selection" (individual's changing criminal tendency across time points due to personal decisions) could have a significant effect on whether treatment is effective in reducing criminality.
Abstract
A key issue that arises regarding the assessment of treatment effectiveness is that the allocation of the treatment is not completely random. Rather, the individuals who receive a particular treatment at a given point in time may be systematically different in characteristics from those who do not receive the treatment. These systematic differences might be correlated with the individual's underlying criminal tendencies at that point in time, a process often called "selection." "Dynamic selection" refers to the changing characteristics and motivations of individuals regarding criminal behavior. The current study found that the "dynamic selection" factors of motivation to join gangs and the motivation to marry at any given point in time were related to an individual's criminal behavior. Researchers should be concerned that whatever change an individual is experiencing at a given point in time that causes him/her to join a gang or get married might also be influencing his/her underlying tendency toward criminality. The extent to which this is true will bias the fixed-effects estimates of each treatment on criminality away from zero. This analysis of marriage and gang membership from the NSLY97 shows that such "dynamic selection" could potentially be quite large in its effect on criminality, and it might even be magnified when using more advanced fixed-effects methods, such as Inverse Probability of Treatment Weighting. Given these findings, the author advises researchers to be cautious when interpreting results derived from fixed effects methods that do not account for changes in individuals that influence their decisions at a given time. 6 tables and 59 references