Pretrial Diversion

The fundamental ambivalence of the goals of diversion was reflected in a number of early evaluations. In addressing the issue of participant recidivism, many programs pointed to a low absolute level of rearrest as a measure of treatment success, seldom recognizing that in the absence of a comparison group this measure did not distinguish the effects of the selection process from actual behavioral change. Many quasi-experimental designs were scarcely more persuasive. Not surprisingly, project evaluators found positive results after comparing participants before and after they were designated as successful, comparing only the successful participant group with all members of an untreated group, or comparing the performance of failures with that of successes. Since each of these designs suffered the obvious bias, it was impossible to determine whether the observed difference was a change attributable to treatment or simply a confirmation that the successful group had been less inclined to criminality in any event.

Even the more comprehensive evaluative efforts were frequently forced to rely on matched group studies in lieu of random assignment procedures as a result of resistance by program personnel to the notion of withholding services from eligible participants. In these designs, defendants who—on paper—appeared to match the characteristics of eligible participants were selected retrospectively from closed case files. Since entrance criteria always required participants to volunteer for the program and to pass more than a paper screening, these matching exercises seldom survived careful scrutiny. In only one study of the Manhattan Court Employment Project, one of the nation's oldest and largest diversion programs, was a satisfactory, retrospective comparison-group selection methodology used. This 1974 reanalysis demonstrated no effect of treatment on recidivism (Zimring). Five years later, in 1979, the same project issued the first definitive experimental study of adult pretrial diversion, which also failed to find any significant effect on recidivism (Baker). Ironically, although resistance to random assignment procedures has typically been founded on legal and ethical objections to the denial of services to eligible groups, the evaluation did not find participants receiving more services or less exposure to the criminal justice system than did members of the control group.

Turning to the literature on the early diversion programs for drug cases, TASC evaluations often repeated the same mistakes observed in evaluations of general deferred prosecution programs. As a result the only firm conclusion that could be drawn is that costs were high and participant impact of any kind, hard to detect.

The drug court successors to TASC (many of which annexed TASC programs) have generated a substantial evaluation literature. A report issued by the U.S. General Accounting Office (GAO) in July 1997 reviewed the results of twenty evaluations of sixteen drug courts and concluded that while positive outcomes were reported, they were typically based on quasi-experimental designs that did not permit any definitive conclusions about impacts on recidivism or drug use. For example, many authors persisted in making invalid comparisons between program graduates and drop-outs, believing that these comparisons demonstrated a positive program effect (Brown). Subsequent reviews in 1998 and 1999 by Belenko were considerably more optimistic than the GAO report, although the author acknowledged the limitations of the available data. In his 1999 review of twelve evaluations that measured post-program recidivism of participants and a comparison sample, he found lower rates of recidivism in seven studies, comparable rates in three, and mixed results in two (Belenko, 1999).

Truitt and Rhodes (2000) examined seven of the most robust evaluations and concluded that even among the best available studies, outcomes ranged widely with four programs finding an impact on recidivism and three (including two based on true experimental designs) showing comparable rates of recidivism. A critical review of those showing positive effects suggested the results were confounded by selection bias that may have caused lower risks to enter drug court. Attempting to correct for that bias in a study of programs in Escambia County, Florida, and Jackson County, Missouri, Truitt and Rhodes found a strong treatment effect in Escambia County when recidivism is defined as an arrest for a felony but no demonstrable effect when both felonies and misdemeanors were considered. In Jackson County, the authors found a sizable treatment effect even when regarding felony and misdemeanor arrests as a single outcome measure.

In the end, however, the best that can be said is that while participants are under the supervision of a drug court (and subject to random drug testing and judicial oversight of their progress in treatment) recidivism appears to be held in check. Thereafter, the results are mixed with the most rigorous evaluations finding no effect on recidivism.