Use of general population surveys in addition to institutional samples is critical to disentangling the relationship between criminal behavior and alcohol problems or use of illicit drugs. Local area studies can be useful but generalizability of their results is seldom studied. Data from recent US national (n = 2058) and county (n = 3069) general population surveys are used to examine the role of alcohol problem and drug use history in predicting self-reported criminal behavior, arrest and conviction within a logistic regression framework. In the national and county surveys controlling for age, gender, income, marital status, employment, education, race and drug use, lifetime drinking problems significantly predicted current criminal behavior (odds ratios 1.3 and 1.5, respectively) with slightly stronger relationships noted in equivalent models predicting arrest (odds ratios 1.7 and 1.8) and conviction (odds ratios 1.7 and 1.6). Relationships between alcohol, drugs and criminal behavior/justice variables are discussed. Parallels between US and county results suggest that findings from intensive, articulated analyses of community-level population and institutional surveys may be cautiously generalized beyond their geographic locus.