AIM - A community intervention programme STAD was launched in Stockholm in January 1998, which included training in responsible beverage service and stricter enforcement of existing alcohol laws. An evaluation suggested that during the first 33 months of the programme, the level of police-recorded violence dropped by a striking 29%. We propose to probe the robustness of this estimate, which is often cited as evidence of the effectiveness of these kinds of intervention. In this paper, we reanalyse the underlying data by applying alternative model specifications. DATA AND METHODS - We reanalysed the original data on police-recorded violence from January 1994 to September 2000 by autoregressive integrated moving average (ARIMA) modelling. We estimated models based on raw data and seasonally differenced data; we also varied the definition of control area and applied the statistical technique of difference-in-differences modelling. RESULTS - The estimated intervention effects from these model specifications were all strongly significant statistically, ranging between 21% and 32%. CONCLUSION - Estimates based on a variety of model specifications were generally somewhat lower than those previously reported. However, the new estimates were all strongly statistically significant and fairly uniform with regard to effect size, which suggests that the findings of a substantial impact of the STAD programme are indeed quite robust.