Self-reported criminal behavior up to 18 months, official re-arrests up to 24 months, and sentence length on the precipitating case were compared between 1,156 drug court participants from 23 sites and 625 comparison offenders from 6 sites. Slightly smaller sub-samples were retained for follow-up interviews. (Eighty-three percent were retained at the 18-month follow-up.) A "super weighting" strategy was employed to adjust for selection and attrition bias, and hierarchical modeling was employed to adjust for the clustering of outcomes within sites. Drug courts reduced criminal behavior, including a reduction by more than half in the number of criminal acts over 18 months. The magnitude of this effect did not significantly vary across most of 17 offender subgroups. Drug courts did not reduce average sentence length on the precipitating case, whereas program graduates faced little or no incarceration, those failing received much longer sentences than the comparison group. Based on a multi-site design with relatively high external validity as compared with past studies, drug courts appear to reduce future criminal behavior, suggesting that it would be beneficial to expand their reach to more offenders. The discussion addresses key study limitations, including the use of quasi-experimental methods and a follow-up timeframe of less than two years. Abstract published by arrangement with Springer.