More than a dozen cost-benefit analyses have been conducted on drug courts in the last decade. We build on these findings and extend them in several ways: 1) a larger sample allows us to draw inferences from a large sample of individuals (nearly 1,800) and courts (23); 2) survey data on a number of domains which have never been included in past analyses (such as employment, hospital use, homeless shelter use, mental health treatment, and many more) increases the range of program costs and benefits considered; 3) we employ statistical techniques less common in criminal justice cost-benefit analyses, although not new, to identify individual characteristics which make drug court most cost-effective; and 4) we separately analyze each drug courts cost effectiveness to draw inferences about which drug courts are and are not cost effective under different combinations of price structures, program design, and offender population characteristics