An increasing number of juveniles are being held in public correctional facilities. The 1-day count of juveniles held in public facilities rose 47.0% from 1983 to 1995 (Sickmund, Snyder, & Poe-Yamagata, 1997). Earlier reports showed that the number of juveniles held in public, long-term institutional facilities increased 10.0% between 1983 and 1991 (Snyder & Sickmund, 1995). These shifts are occurring at a time when overall custody rates - that is, the number of juveniles in custody in public facilities on a given day for every 100,000 juveniles in the population - have increased from 221 per 100,000 juveniles in 1989 (Moone, 1993) to 245 per 100,000 juveniles in 1995 (Sickmund et al., 1997).
These trends underscore the critical importance of effective custody programs. Unfortunately, the role that custody programs play in the rehabilitation of delinquents has not been extensively explored. Decades of criminal justice research have yet to resolve basic questions as to whether specific correctional programs or rehabilitative interventions have consistent, long-term effects on reducing the subsequent criminal activity of adjudicated delinquents. Research on juvenile custody is characterized by an inconsistent set of findings as to which particular settings or programs have beneficial effects.
As a first step in understanding the impact of custody, we need to delineate the types of custody programs to which delinquent youth are admitted. These fall into two broad categories: traditional training schools and a heterogeneous category that we label "alternative" programs. In a 1989 sample of juvenile custody admissions in six states, more than three quarters of all youth (78.0%) were placed in training school programs - secure, restrictive custody programs in institutional settings (Office of Juvenile Justice and Delinquency Prevention [OJJDP], 1993). Smaller proportions of adjudicated youth were admitted to a range of alternative settings: halfway houses, group homes, foster homes, ranches, camping programs, and specialized vocational programs (OJJDP, 1993).
How effective are alternative placements compared with training school placements in rehabilitating delinquent youth? There has been much speculation and debate as to whether alternative programs have beneficial consequences relative to traditional institutions, but few empirical studies have directly compared these basic program types (Geerken & Hayes, 1993; Gottfredson & Barton, 1993). Moreover, recent studies that have examined the relative impact of institutional and alternative programs on outcomes such as recidivism have done so over a short follow-up period (Fendrich, 1991). There have been no recent studies directly evaluating the long-term effectiveness of alternative custody placements.
The current study fills a crucial gap in knowledge about the rehabilitation effects of custody by providing a long-term comparison within the context of a single, statewide juvenile corrections agency, the Texas Youth Commission (TYC). We examine recidivism for two groups of youth remanded to TYC in 1983: a group that was placed in institutional programs (training schools) and a group that was placed in alternative settings. In particular, we compare the relative risk for rearrest for youth placed in institutional and alternative programs over the course of a 12-year follow-up period.
INSTITUTIONS VERSUS ALTERNATIVE PROGRAMS: EVIDENCE FROM PRIOR RESEARCH
Recent decades have seen substantial criticism of training school environments (Feld, 1977; Greenwood & Zimring, 1985; see Shichor & Bartollas, 1990). Greenwood and Turner (1993) suggest that despite this criticism, research studies have not provided a strong justification for a shift toward alternative correctional programs. Indeed, in the early and mid-1980s, the debate about institutionalization was a reaction to empirical evidence underscoring the relative merits of institutional programs for delinquents. In 1979, Murray and Cox (1979) rejected prior notions that institutional training school environments were "schools for crime" that inevitably exacerbated delinquent tendencies among youth who were placed there. They evaluated the impact of a variety of residential and custodial programs on a measure of arrest rate suppression in a sample of adjudicated youth from Chicago. Suppression effects were demonstrated for all programs except probation. The largest suppression effect was demonstrated for youth who were sent to state corrections institutions.
Two subsequent studies challenged these findings. Maltz, Gordon, McDowall, and McCleary (1980) statistically reinterpreted the results from this research. Noting that juveniles in the institutional group had more recommitment offenses, Maltz et al. (1980) concluded that differences in suppression were a consequence of "regression artifacts." Using Murray and Cox's (1979) dependent measure of arrest rate suppression, Lundman's (1986) reanalysis of two quasi-experimental studies showed suppression effects resulting from both community and institutional treatment.
Using meta-analytic summaries, effect size estimates derived from studies of programs delivered in training school or custodial settings can be compared with estimates derived from studies of programs delivered in alternative settings. Andrews et al. (1990) provide data suggesting significant main effects for their effect size estimate of an indicator of program setting contrasting community programs with residential programs. They found that residential settings "dampen the positive effects of appropriate service while augmenting the negative impact of inappropriate service" (p. 384). Unfortunately, the indicator used by Andrews et al. lumped both residential treatment centers (a type of alternative program) and training schools into a single category. Lipsey (1992) used a hierarchical regression model to evaluate the impact of the "nature of treatment," including a measure of whether studies were conducted in "custodial institutions." Lipsey concluded that "as a category, treatment in an institutional context seems to be associated with smaller effect sizes" (p. 122).
This ongoing skepticism regarding the effectiveness of institutions in rehabilitating juvenile offenders underlies a substantial body of work that has evaluated specific programs within the context of different types of custody settings (Greenwood & Turner, 1993; see meta-analyses by Andrews et al., 1990; Lipsey, 1992), but these studies have not clearly established the relative effectiveness of either alternative or institutional programs. Only a few studies directly address this comparison. The work of Greenwood and Turner (1993) directly compares these two types of correctional intervention strategies. The authors found no significant differences in recidivism risk between a sample of serious delinquent youth randomly assigned to a comprehensive residential treatment program or a traditional training school. Directly comparing postrelease adjustment among youth assigned to institutional and alternative programs, Fendrich (1991) found some benefit for alternative placement; this work underscored the potential importance of interactions between participant characteristics and program type in predicting postrelease behavior.
In another recent study, Gottfredson and Barton (1993) compared recidivism among an institutionalized and noninstitutionalized population of youthful offenders in Maryland. The noninstitutionalized group spent less time than the institutionalized group in large, secure, state-run custodial facilities and more time in community-based and psychiatric settings. The authors found recidivism to be significantly higher among the noninstitutionalized group, and the difference was large: After 1 year, 72.0% of the noninstitutionalized group were recidivists compared to about 45.0% of the institutionalized group. Although their study is limited by both a short follow-up period and the lack of a contemporaneous comparison between program groups, the authors rightly conclude that the implementation as well as the design of treatment interventions determines whether a specific alternative program will be effective at reducing recidivism.
In sum, despite claims that alternative programs are more effective than institutions in reducing recidivism among juvenile offenders, empirical research has produced inconsistent findings. Alternative programs vary a great deal across time and location in terms of available services and interventions and the quality and implementation of services (Gottfredson & Barton, 1993; Shichor & Bartollas, 1990). Thus, what is most useful as a starting point is a focus on broad comparisons of traditional and alternative approaches to juvenile corrections rather than more limited comparisons of specific alternative programs. This is especially the case if, as some critics have claimed, the advantages of alternative programs are attributable less to specific treatments and interventions than to the absence of the negative impact of institutions (Geerken & Hayes, 1993). This article addresses the broad issue of whether alternative programs represent an overall improvement, relative to institutions, in terms of reducing recidivism among juvenile offenders.
DETERMINANTS OF JUVENILE RECIDIVISM
Several studies have documented high rates of recidivism among juvenile offenders, especially among chronic offenders and those who have committed serious crimes, in which recidivism rates may approach two thirds or more (e.g., Beck & Shipley, 1987; Bureau of Justice Statistics, 1987; Duncan, Kennedy, & Patrick, 1995; Gruenewald & West, 1989; Palmer & Wedge, 1989; Speirs, 1988; Visher, Lattimore, & Linster, 1991). For example, Gruenewald and West (1989), in their 5-year study of recidivism among chronic juvenile offenders in several cities, found a 73.0% recidivism rate at 30 months for serious, chronic offenders. Visher et al. (1991) found an 88.0% recidivism rate for a sample of high-risk male juvenile and young adult offenders released by the California Youth Authority. In a study of 129 juveniles released from a state training school in Florida, Duncan et al. (1995) found that 52.0% were rearrested within a period of only 6 months.
A review of the literature (Fendrich, 1991; Hanson, Henggeler, Haefele, & Rodick, 1984; Hepburn & Albonetti, 1994; Visher et al., 1991) suggests that factors …