Citation

Rountree PW, Land KC. Soc. Sci. Res. 2000; 29(2): 284-305.

Copyright

(Copyright © 2000, Elsevier Publishing)

DOI

10.1006/ssre.2000.0670

PMID

unavailable

Abstract

Recent criminal victimization studies have emphasized the interrelationship between individual and structural effects in explaining individuals' risks of victimization. But extant contextual victimization studies have yet to incorporate cross-city comparisons of individual- and neighborhood-level mixed-effect models. As a result, questions remain regarding the generalizability of the extant multilevel victimization models across cities or regions with different economies, histories, cultures, etc. The present study addresses the potential generalizability of empirical relationships from multilevel (individual- and neighborhood-level) models of burglary victimization across three cities by comparing the effects of individual-level sociodemographic and routine-activity variables, neighborhood-level social disorganization and concentration-of-poverty variables, and micro-macro interactions using data from Rochester, St. Louis, and Tampa-St. Petersburg. Contextual findings from multilevel burglary victimization models appear generalizable in several respects. First, mean burglary victimization risk varies significantly across neighborhoods in all cities examined, but the individual-level covariates of risk do not vary in their effects across neighborhood or city contexts. Secondly, much of the variability in burglary victimization risk across neighborhoods is accounted for by the inclusion of neighborhood-level covariates. Third, while many of these contextual effects reported appear city-specific in that associated p values vary and represent different levels of statistical significance, further inspection and comparison across cities of the confidence intervals of the effects reveals substantive consistency. In short, we conclude that our findings make a case for the generalizability of victimization theory across a variety of urban settings and provide a good illustration of possible limitations of relying solely on significance tests for conclusions.