The research used the bivariate probit model, a statistical methodology that allowed for the explicit recognition that an initial victimization must occur prior to any repeat event. A multinomial logit model was used as a preliminary step to the bivariate probit. Results revealed that repeat victims lack distinctive characteristics. However, households with characteristics that protect from victimization, in the sense of having a low initial risk, have this protection reduced for a subsequent event. In addition, comparing two households with different risk characteristics reveals that their probabilities of repeat victimization are more similar than were those for the initial crime. Tables and 35 references (Author abstract modified)