However, their findings are conflicting and have added to, rather than resolved, the discussion of police effect. These studies suffer from methodological and conceptual weaknesses. First, they sample from populations where system level variance is minimal, and second they analyze models that could either over- or under-estimate effects. The present study addresses these methodological and conceptual problems by first selecting variables for inclusion in the model based on current criminological theories, and then maximizing system level variance by employing a cross-national sample. Regression analyses of the data suggest that previous studies have over-estimated the police effect which is almost non-existent. (Author abstract)