A review of prior research shows that some theorists argue that time-stable characteristics such as criminal propensity should account for any positive correlation between past and future criminal behavior for all individuals. Others contend that the positive association between offending behaviors at various points in time are partly causal and partly spurious. Still others anticipate that different patterns will emerge for different groups (distinguished by their criminal propensity) of individuals. The current study involved 838 releasees from North Carolina Division of Youth Services training schools during calendar years 1988 and 1989. Background data were used to create a multiwave panel that varied from 4 to 6 years in length. Some individuals were followed for as few as 4 full years (n=47), others were followed for 5 years (n=405), and still others were followed for 6 full years (n=386). The data collected facilitated calculation of the number of arrest events that occurred during each full year of the follow-up period. First, the analysis attempted to identify the effect of prior arrests on future arrests under the restrictive assumption that a single process generated the data for all 838 individuals. Then, using varying onset ages as cutting points, this equal-effects constraint was relaxed to test the hypothesis that prior arrests exerted similar effects on subsequent arrests, regardless of one's membership in an early-onset or late-onset group. The results show that both stability and change have causal implications for a person's offending behavior; and with but one exception, these effects do not vary between high and low criminal propensity groups, as measured by age at the onset of delinquent behavior. 9 tables and 54 references