In this study, 216 risk assessments conducted by law enforcement officers in a suburb of Stockholm using the Brief Spousal Assault Form for the Evaluation of Risk (B-SAFER) were followed up and correlated to both recidivism and type of protective actions implemented by the police. The results showed high recidivism rates across all risk categories, except in the highest risk group, where the recidivism rate was significantly lower. This finding suggests a poor, and even negative, predictive power of the police risk assessments: the higher the police-assessed risk, the lower the recidivism rate. However, it was clear that the police did very little, or nothing, in cases not assessed as high risk. All resources appear to have been directed to the high-risk cases, which seems to have had a preventive effect. Our results point to the importance of studying the nature and extent of protective actions taken in response to risk assessment, before drawing conclusions about the predictive validity of risk assessment instruments. Copyright © 2012 John Wiley & Sons, Ltd.