NCJ Number
85370
Date Published
1981
Length
13 pages
Annotation
This study takes the model of delinquency careers suggested by Wolfgang (1972), that of a simple absorbing Markov chain, and computes the fundamental matrix to determine the predictive utility of the model compared to that of its empirically derived equivalent.
Abstract
The similarities of the predicted and observed mean number of offenses before absorption appear great enough to conclude that the data support the hypothesis that delinquency careers can be modeled by a simple absorbing Markov chain. Other matrix analyses based on this conclusion indicate that, regardless of the type of first offense, the probability of ever being charged in the future for a nonindex offense is about .6, a theft about.3, an injury about.2, a combination of these index components about .2, and a damage offense about .08. The probabilities of multiple repeats of the same kind of offense range from .03 for three nonindex repeats to .007 for three damage or combination repeats. Beyond three repeats, the probabilities drop from .007 for nonindex repeats to .00004 for five damage repeats. Thus, offense specialization of like-offense repetition is not a high incidence phenomenon. The powers of the absorbing matrix P show that offenders rapidly leave the states of offense commission and are not subject to further police contact. The probability of being absorbed into the boundary of no further contact increases from about .5 after the second offense to about .8 after the fifth to .95 after the ninth. When the absorbing state is taken out of the process and a regular chain produced, the probability of commiting an offense rapidly converges to a common vector after the second transition. Thus, the probability of the next offense being a certain type of independent of the kinds of prior offenses committed. Tabular data and two references are provided. (Author summary modified)