The study used data from the 1994 National Crime Victimization Survey. The set of explanatory variables comprised individuals' profile and lifestyle as well as characteristics of their household. This article discusses the results of estimated multilevel negative binomial models, which explicitly disentangle the unexplained heterogeneity between individuals and between households. The estimated random effects of household characteristics show that the unexplained heterogeneity for the average number of personal crimes differs across household types. Further, the individual covariates with between-households random effects become less influential the more the base personal crime rates are high. The study claims to have raised more questions than it answered. Having disentangled unexplained heterogeneity of personal crimes between individuals and between households, it suggests the need for research to identify how much heterogeneity exists at the segment or even higher levels of aggregation, such as the Census tract or State. Also, the implications of random effects for victimization theory should be integrated into the current state of the art. Notes, tables, references