Purpose The purpose of this paper is to examine the interaction of enforcement and adjudication for general deterrence of drunk-driving. The authors present a triangular feedback model between three domains: police, courts and drunk-driving events. The authors’ deductive approach imposes no structural assumptions beyond the core of general deterrence theory. Design/methodology/approach Using a largely untapped data set for California’s 58 counties from 1990 to 2010, the authors estimate a series of heterogeneous panel Granger non-causality tests. This empirically based evidence is re-organized per the proposed triangular feedback model to objectively categorize local criminal justice systems as active, responsive or reactive (with respect to drunk-driving). Findings Our results suggest that state-level analyses obscure useful variations that empirical panel methods can now handle. The authors provide evidence that research based on empirically derived groupings, rather than inductively based preconceptions, is key to understanding enforcement and compliance. The authors provide a less confounded picture of the relationship between drunk-driving enforcement and adjudication. Research limitations/implications Our study addresses one offense for a particular state in the USA. It is an exploratory analysis. This analytical and empirical approach is new. Practical implications Our approach imposes very few a priori assumptions and requires a minimum of data series to be executed. The method can be broadly applied to a range of topics and observational units. Social implications The authors aim to expand identification of local systems’ effectiveness (or not) and mechanisms of for general deterrence of drunk-driving. The offense is one that can be committed easily and unintentionally; it does not presume anomie. The authors address general communities, not anomalies. Knowing how enforcement and compliance operate is essential to an array of behavioral externalities. Originality/value This is a new empirically based approach for analyzing social systems. It is a marriage of new macroeconomic time-series techniques with an old question, most often addressed by microeconomic research. This study uses an underutilized data source to construct a unique panel data set.