Prior research has shown that victimization incidents are disproportio
nately concentrated among relatively few victims and that prior victim
ization is a consistent predictor of future risk. This paper expands e
xisting knowledge on victimization by describing temporal patterns of
risk and by developing and testing alternative explanatory models of t
he link between past and future risk. Analyses based on panel data fro
m the National Youth Survey support both state dependence and heteroge
neity interpretations of the correlation in risk over time. In other w
ords, prior victimization predicts future risk in part because it alte
rs something about the individual, and because it indicates an unmeasu
red propensity for victimization that persists over time. The theoreti
cal implications of these findings, including the feasibility of a vic
tim labeling perspective, are discussed.