Background There are three options in Florida for young people with mental health needs who require out‐of‐home treatment: community‐based group homes, foster families and inpatient group facilities. Prior research has suggested that young people in group settings have poorer criminal justice outcomes, perhaps attributable to negative peer effects in group environments. Aims To compare arrest rates during and after out‐of‐home treatment for youth across the three settings. To test the hypothesis that arrest rates during treatment are independently related to arrest rates after treatment, after allowing for pre‐treatment characteristics and placement type. Methods Florida Medicaid claims data were used to identify children and adolescents in out‐of‐home mental health care from 2003–2007. These were then matched with Florida Department of Juvenile Justice and Law Enforcement records. Propensity score matching was used to allow for observed differences between youth in different treatment settings. Multinomial logistic regression models were used to test relationships between arrest rates. Results Unadjusted arrest rates for youth in therapeutic group home care were higher than in psychiatric inpatient units or foster care during and after treatment. Arrests during the treatment episode accounted for a substantial proportion of the difference in arrests after out‐of‐home treatment. After accounting for differences in arrest rates during treatment, the group versus non‐group nature of the treatment setting per se was not a strong determinant of arrest after the treatment episode. Conclusions and implications for practise Attention to the causes of higher arrest rates in group homes, which may include peer contagion and staff policies, could improve outcomes. For youths without a major psychiatric disorder, therapeutic foster care may be better than community‐based group care. Copyright © 2013 John Wiley & Sons, Ltd.