It is well documented that there is a strong association between involvemen
t in adolescent delinquency and involvement in adult criminality. However,
the association is not perfect. Some juveniles who offend at high rates do
not go on to offend as adults while some do. Some juveniles who offend at l
ow rates go on to offend as adults while some do not. The reasons for these
behavioral changes are not yet well understood. Some criminologists conten
d that changes in behavior between the adolescent and the adult years are d
ue to processes that occur during the adult years, while others contend tha
t all important adult crime-producing processes operate before the end of a
n adolescence. In this paper, we use data from the Cambridge Study in Delin
quent Development to investigate the empirical merit of the second position
. Specifically, after conditioning on adolescent offending behavior, we ask
whether variation in adult offending is consistent with a conditional rand
om process. Our analysis suggests an affirmative answer to this question. A
lthough our results do not prove that this "conditionally random" explanati
on is the process that generates the data, they do suggest that models maki
ng this prediction cannot be easily dismissed.