Background Spatial dependencies may influence the success of community action strategies to prevent and reduce harmful alcohol use. This study examined the effectiveness of a multicomponent Responsible Beverage Service (RBS) programme targeting on-licensed premises on police-recorded assaults in Swedish municipalities. It was expected that the implementation of the programme within any given municipality had an indirect effect by reducing violent assaults in adjacent municipalities.

Methods This study was a natural experiment exploiting the temporal and spatial variation in the implementation of the RBS programme to predict change in the rate of violent assaults in all Swedish municipalities during 1996–2009 (n=288; T=14; N=4 032). Yearly police-recorded violent assaults per 100 000 inhabitants aged 15 and above committed on weekend nights were used as a dependent variable. Programme fidelity was identified by means of survey data. A semilogarithmic fixed-effects spatial panel regression model was used to estimate the direct, indirect and total effects of the programme.

Results The direct, indirect and total effects were −1.8% (95% CI −4.4% to 0.8%), −5.8% (95% CI −11.5% to −0.1%) and −7.6% (95% CI −13.2% to −2.2%), respectively. Averaged over time and across all municipalities, implementing one additional programme component in all municipalities will thus reduce violent assaults in one typical municipality by nearly 8%.

Conclusions The indirect effect of the programme was three times larger than its direct effect. Failing to account for such local spillover effects can result in a considerable underestimation of the programme's total impact and may lead to erroneous policy recommendations.