In response to a request by the State of Washington Joint Legislative Audit and Review Committee, University of Maryland criminologists reviewed the research examining the effectiveness of twelve correctional program areas offered by the Washington Department of Corrections: programs for chemically-dependent offenders, sex offender treatment, cognitive skills and moral reconation therapy, anger/stress management, victim awareness, life skills training, adult basic education, correctional industries, vocational education and training, and other work programs. The research included in the review examined the impact of correctional programming on recidivism. Conclusions about "What Works, What’s Promising, What We Don’t Know, and What Doesn’t Work" were drawn on the basis of both assessment of the quality of the research design and the significance, direction, and size of the program effects. The report concluded:

• In-Prison Therapeutic Communities With Follow-Up Community Treatment

• Cognitive Behavioral Therapy: Moral Reconation Therapy and Reasoning and Rehabilitation

• Non-Prison Based Sex Offender Treatment Programs

• Vocational Education Programs

• Multi-Component Correctional Industry Programs

• Community Employment Programs

• Prison-Based Sex Offender Treatment

• Adult Basic Education

• Transitional Programs Providing Individualized Employment Preparation and Services

For High-Risk Offenders

• Acupuncture Within Outpatient Substance Abuse Treatment Programs

• Intensity and Integrity of Substance Abuse Treatment Programs Receiving Referred Offenders

• Anger/Stress Management Programs

• Victim Awareness Programs

• Community Vocational Training Programs

• Success of Programs With Different Types of Sex Offenders

• Life Skills Training Programs

• Work Ethics, In-Prison Work Programs, Halfway Houses with Enhanced Services

• Increased Referral, Monitoring, and Management in the Community

As prison populations grow nationally and corrections makes up increasing proportions of state budgets, many jurisdictions are seeking to determine if their funds are being spent on programs that have some impact on participants criminal behavior. The federal government has also been interested in this question. In 1996, Congress required the Attorney General to provide a "comprehensive evaluation of the effectiveness" of over $3 billion annually in Department of Justice (DOJ) grants to assist state and local criminal justice and community efforts to prevent crime. Congress required that the research included in the evaluation be "independent in nature," and "employ rigorous and scientifically recognized standards and methodologies." The Assistant Attorney General for the Office of Justice Programs asked the National Institute of Justice to commission an independent review of the relevant scientific literature, which exceeded 500 program outcome evaluations. The culmination of this effort was the publication of a more than 600 page report to Congress. The report entitled, "Preventing Crime: What Works, What Doesn’t, What’s Promising," authored by a number of faculty at the University of Maryland. One chapter of the report evaluated the effectiveness of strategies and programs in courts and corrections.

The present report, modeled after the Maryland Report, was requested by the State of Washington Joint Legislative Audit and Review Committee. The report evaluates the effectiveness of correctional programs not examined in the earlier report. The goal of the project is to examine the effectiveness of the type of rehabilitation programs currently offered by the Washington State Department of Corrections (DOC). The DOC offers a variety of work and treatment programming in its institutions, pre-release and work release facilities, and in community corrections offices. This assessment focuses exclusively on specific correctional program areas offered by the DOC: programs for chemically-dependent offenders, sex offender treatment, cognitive skills and moral reconation therapy, anger/stress management, victim awareness, life skills training, adult basic education, correctional industries, vocational education and training, and other work programs. Using the same approach as the Maryland Report, this report will assess the scientific rigor and findings of the program outcome evaluation literature in these twelve program areas. The project addresses the following questions:

• What is known to be effective in adult correctional programming, particularly in the major categories of work, education, and treatment programs?

• What is known about the effectiveness of correctional programs or major program elements that are similar to those currently operating in Washington?

For each of the identified program areas, the report reviews the literature on the identified topic and provides a critical evaluation of the available research studies. The review and the evaluations examine the effectiveness of rehabilitation programs. Effectiveness is defined as the reduction of criminal activities in the community.

The evaluations of each study are based on the model used in the "What Works in Crime Prevention" report. Each study was evaluated and scored for scientific rigor. Within each topic area, the results of the evaluations are summarized in order to draw conclusions about: What works, what doesn’t, what is promising and what we don’t know. The topics selected for review and evaluation were identified in conjunction with members of the Washington State Joint Legislative Audit and Review Committee.

Our review of the correctional programs in this report must be understood within the context of research on rehabilitation and treatment. Rehabilitation strategies focus on changing individual offenders so they will not continue their criminal activities. The research goal is to identify and understand the individual differences that explain criminal behavior and how interventions can be used to change individuals so they will not continue to commit crimes. The work is based on psychological theories of learning, cognition and the general principles of human development applied to the analysis of illegal behavior (Andrews & Bonta, 1994). Research has focused on examining the components of programs that are effective in reducing recidivism.

Today, while there is still some debate about the effectiveness of rehabilitation (e.g., Lab & Whitehead, 1988; Whitehead & Lab, 1989) recent literature reviews and meta-analyses demonstrate that rehabilitation programs can effectively change offenders (Andrews & Bonta, 1994; Andrews et al., 1990; Palmer, 1975; Gendreau & Ross, 1979, 1987; Gendreau, 1981; Gendreau & Ross, 1979; 1981, 1987; Gottfredson, 1979; Cullen & Gilbert, 1982; Greenwood & Zimring 1985; Halleck & Witte, 1977; Van Voorhis, 1987). In general, most literature reviews show positive evidence of treatment effectiveness (Andrews et al., 1990). For example, in a series of literature reviews, the proportion of studies reporting positive evidence of treatment effectiveness varied from near 50 percent to 86 percent: 75 percent (Kirby, 1954), 59 percent (Bailey, 1966), 50 percent (Logan, 1972), 48 percent (Palmer's 1975 re-tabulation of studies reviewed by Martinson in 1974), 86 percent (Gendreau & Ross, 1979) and 47 percent (Lab & Whitehead, 1988). In reviewing these studies, Andrews and his colleagues (1990) conclude that "this pattern of results strongly supports exploration of the idea that some service programs are working with at least some offenders under some circumstances." The important issue is not whether something works but what works for whom.

Based upon the available evidence, some approaches to treatment are clearly better than others. Psychological researchers emphasize that effective treatment programs must follow some basic principles (Gendreau & Ross 1979, 1987; Cullen & Gendreau 1989). First, treatment must directly address characteristics that can be changed (dynamic) and that are directly associated with an individual's criminal behavior (criminogenic factors). There are numerous risk factors associated with criminal activity. Age, gender and early criminal involvement are some examples. In comparison to others, young males who began criminal activities at a young age are higher risks for future criminal activities. But these "static" characteristics such as age, gender and past history, while predictive of recidivism, cannot be changed in treatment. Instead, the "dynamic" or changeable factors should be the target of treatment programs.

Equally as important is the distinction between factors that are criminogenic and those that are not. Criminogenic factors are those that are directly associated with criminal behavior. Research has revealed some dynamic factors that are also criminogenic, i.e., attitudes, cognitions, behavior regarding employment, education, peers, authority, substance abuse and interpersonal relationships that are directly related to an individual’s criminal behavior (Gendreau, Little, & Goggin, 1995). Less promising targets for reducing future criminal behavior include increasing self-esteem without touching antisocial propensity, or increasing the cohesiveness of antisocial peer groups. Treatment programs that target such non-criminogenic factors will not be particularly successful in reducing recidivism. In order to be successful, treatment must address factors that can be changed (e.g. dynamic factors) and that are directly related to an individual's criminal behavior (criminogenic).

A second factor important in determining whether a treatment program will be effective is the therapeutic integrity of the program or the need for effective programs to be delivered as planned and designed. Poorly implemented programs, delivered by untrained personnel, where offenders spend only a minimal amount of time in the program, can hardly be expected to successfully reduce recidivism. A third factor in effective programming is that programs must target offenders who are at sufficient risk for recidivism so that this reduction is measurable. Many offenders are low risk for future recidivism. Treatment programs that provide intensive services for such offenders will show little reduction in future criminal activities because few of these offenders will recidivate anyway. The final principle of effective treatment is the need to deliver treatment in a style and mode that addresses the learning styles and abilities of offenders. For example, more effective programs follow a cognitive behavioral and social leaning approach rather than nondirective relationship-oriented counseling or psycho-dynamic, insight-oriented counseling.

Meta-analyses that have classified studies of treatment as appropriate or inappropriate according to the principles of effective treatment have found support for the importance of the proposed principles. In general, programs that follow the principles are found to reduce recidivism, although the extent of the reduction varies by study and principle being examined (Andrews et. al, 1990; Lipton & Pearson, 1996; Lipsey, 1992).

In summary, there is evidence that rehabilitation is effective in reducing the criminal behavior of at least some offenders. The evidence from the meta-analyses suggests that effective correctional treatment programs appear to follow some basic principles. In order to effectively reduce recidivism, treatment programs appear to need to:

• Be carefully designed to target the specific characteristics and problems of offenders that can be changed in treatment (dynamic characteristics) and those that are predictive of the individual’s future criminal activities (criminogenic) such as antisocial attitudes and behavior, drug use, anger responses;

• Be implemented in a way that is appropriate for the participating offenders and utilizes therapeutic techniques that are known to work (e.g., designed by knowledgeable individuals, programming provided by appropriately educated and experienced staff, use of adequately evaluated programs) and require offenders to spend a reasonable length of time in the program considering the changes desired (deliver sufficient dosage);

• Give the most intensive programs to offenders who are at the highest risk of recidivism;

• Use cognitive and behavioral treatment methods based on theoretical models such as behaviorism, social learning or cognitive-behavioral theories of change that emphasize positive reinforcement contingencies for prosocial behavior and is individualized as much as possible.

More information is needed regarding: (1) how to ensure that treatment programs have adequate integrity; (2) what should be targeted in the treatment (antisocial attitudes, values, employment behavior, education, etc.); and (3) what method should be used to deliver the treatment (required staff training, outpatient, in-prison programs). In summary, there is relatively strong evidence that some treatment programs work for some offenders. At this point, we need more information about the specific characteristics of the effective programs and the most appropriate target populations.

This report examines the effectiveness of twelve correctional programs that were not assessed in the earlier crime prevention report. We began this project by gathering information on all evaluations on the topics of interest. After preparing the bibliography listing all of the evaluations, we reviewed each study for scientific rigor, calculated effect sizes and summarized the information on tables. As described below, this information was used to draw conclusions about what works, what doesn’t, what is promising, and what is still uncertain.

A program evaluation is the scientific study of the impact of a specific program on its participants. There are different types of evaluations. For example, one type of program evaluation is a process or descriptive evaluation. This type of evaluation would focus on such questions as: Is the program implemented in the way it was designed? How many clients does it serve? What are their characteristics?, and How long does it take participants to complete the program? etc. Process evaluations are geared more toward description of programs and the activities that occur within them, rather than measuring their impact. This type of evaluation does not measure "effectiveness," but can be useful in interpreting results of a second type of program evaluations: outcome or impact evaluations. These evaluations would address such questions as: Is the future behavior of participants impacted as a result of the program? Did completion of the program affect participant attitudes? Do participants succeed in the community following program completion? etc. Outcome evaluations are focused primarily on the effectiveness of the program, but they may also include some process components, as well.

Just as the term "program evaluation" can have various meanings, "effectiveness" can also be defined in a variety of ways. For example, some programs focus attention on increasing participants’ self-esteem. Thus, effectiveness be evaluated by determining whether program participation did indeed positively impact self-esteem. Because correctional programming is primarily intended to reduce criminal behavior, this report defines effectiveness in terms of recidivism. Program outcome evaluation studies that focus on attitudes, skill level, deviate arousal, and other measures are excluded from this assessment because they are not directly addressing the question of recidivism. Instead, they focus on particular mediating factors which may be expected to lead to a reduction in offending behavior, but they do not measure actual behavioral impact. As such, only those outcome evaluations that measure recidivism are included in this report.

There are a variety of different ways that researchers have defined recidivism, e.g. rearrest, reincarceration, community supervision violations, and self-reported offenses, and each definition includes different advantages and disadvantages. For example, defining recidivism as reconviction may be considered a conservative estimate because so much criminal offending escapes detection, arrest, and prosecution. However, reconviction data are useful, in that, they are relatively reliable and assessable. Using self-reported criminal offending to measure recidivism is one alternative to conviction data because it captures may offenses that escape detection and official criminal justice response. However, there are a number of trade-offs with using this type of measure, including deliberate deception and memory lapses. Taken together, all the various methods of defining and measuring recidivism provide a more complete picture of program impact than a single definition alone. Consequently, this report will take into account all the various ways in which researchers might define recidivism in determining program effectiveness.

Only program outcome evaluations measuring recidivism will be assessed in this report. Additionally, the research reviewed will be restricted to evaluations that are relatively recent, i.e. in most cases conducted within the past ten years, because older work is less relevant to present social conditions and correctional programming realities. However, in program areas where research is limited, research conducted since 1980 may be included. Formally published, as well as unpublished research, such as state and local government reports, and research organization studies, are included in this review. While these evaluations are not (in most cases) subject to the quality-control process of peer-reviewed journals, they are valuable in that journals are often biased towards publishing studies of favorable results. Further, many governmental agency reports are not submitted to journals for formal publication. In sum, this report will include published and unpublished program outcome evaluations, largely conducted within the past ten years, that use recidivism as an outcome measure.

How Can Research Results be Combined?

There are several possible methods to combine and summarize research within a particular program area. One such method is a literature review. This is a relatively straight-forward summary of research findings in which the author explains what the collection of results seem to indicate about the effectiveness of the program area. This type of research review generally incorporates a discussion of the quality of the methodology utilized by the various studies. Literature reviews, however, have been criticized because they rely on the author’s subjective assessment of the research and the relative importance of the findings (Wolf, 1986). Another method of combining research results, and one which addresses these problems, is meta-analysis. This is generally viewed as the most sophisticated method of combining research results because it uses statistical techniques to combine the separate studies and draw conclusions from the integrated research. A third means of combining research, the method employed by the Maryland Report to Congress, is a hybrid of a meta-analysis and a literature review. It uses the meta-analytic process of coding research methods, thus providing an objective framework to rate the quality of research methods. Conclusions about the research are based on individual research studies in a manner similar to a literature review, but like meta-analysis, research findings are "weighted" according to the strength of their methodology.

Why Does Research Methodology Matter?

Methodology refers to the way that researchers conduct a study. These methods can be very simple or extremely complex, utilizing highly refined measures and sophisticated statistical models. The quality of the methodology, or research design, used in a study determine the credibility or confidence that can be placed in the results. In scientific research, there are particular types of methods that can be used to increase the credibility, or strength of the evidence produced by studies. In terms of program outcome evaluation research, the goal is to determine if there is a cause-and-effect relationship between program participation and future behavior. In order to produce strong evidence of a cause-and-effect relationship, methodology must be used that incorporates the basic elements of scientific research that minimize the possibility that the results are due to factors other than program participation.

Particularly important in producing strong evidence is the inclusion of a control group for comparison with those who participated in the program. This allows researchers to observe how behavior is impacted by the program. The more similar that the control group is to the program group, the stronger the evidence. For example, a study may only compare the recidivism of program completers to program drop-outs. This design can not provide strong evidence because there are other possible explanations for the findings besides the effect of the program, e.g. completers may be more motivated individuals who would have had less recidivism, even without the program, than those who dropped-out. A better design would be to randomly select individuals to participate in the program and the control group. Random selection is the scientific ideal because it reduces the likelihood that one group will be different for that other. For example, randomly placing individuals in the program and control group would more equally distribute the highly motivated individuals, thus reducing this source of "contamination" of the results.

If program evaluation research is used to draw conclusion about the effectiveness of various programs, then the scientific rigor of the methodology of this research will be an important component in determining what works. The findings of studies employing higher quality methodology should be given more weight in the determination than those utilizing weaker methodology. To objectively rank the quality of research designs and consistently define "what works," this report uses the coding scheme and ranking scale developed for the Maryland Report.

The Maryland Scale rates seven different dimensions of the methods used in each study. The overall rating is based primarily on these factors:

• The study's ability to control extraneous variables (i.e., to eliminate major alternative explanations for the findings). This is accomplished through random assignment, matching treatment and comparison groups carefully, or statistically controlling for extraneous variables that minimize measurement error.

• Other considerations contributing to the overall rating of methodological rigor are the response rate, attrition (dropout) of cases from the study, and the use of appropriate statistical tests.

Studies that show only a relationship between program participation and recidivism, i.e., correlational research, do not provide strong evidence because there are many factors that can produce these results besides program participation. Improvement over this design are studies that use pre- and post-program methods. These studies provide stronger evidence that the "cause" came before the "effect." The strongest evidence is provided by studies that use methods designed to rule out other factors that might impact the results, such as pre-program differences between participants and non-participants. Using the Maryland Scale, studies are ranked according this hierarchy of design quality. Each eligible study was given a "scientific methods score" of one to five, with five providing the strongest scientific evidence. In addition, the use of statistical significance tests is employed as a key criterion in reaching program effectiveness conclusions based on the application of the scores. (Appendix A provides the code sheet used to determine the scientific methods score for each reviewed study.) Studies assigned a particular methods score share the following features:

1. Correlation between the program and measure(s) of recidivism crime and no control group.

2. Temporal sequence between the program and recidivism can be clearly observed (e.g. pre- and post-program research design); or a comparison group is present but lacks a demonstrated comparability to the treatment group.

3. A comparison between two or more groups, one with and one without the program.

4. Comparison between a program group and one or more control groups, controlling for other factors; or a nonequivalent comparison group that is only slightly different from the program group.

5. Random assignment and analysis of comparable program and comparison groups, including controls for attrition.

What Works and What Doesn’t?

Clear conclusions about programs that work and programs that don’t work require a high level of confidence in the research findings. New research results continually add new pieces to the puzzle, and reconsideration of old results in light of the new findings often produces different conclusions. The best one can ever claim to "know" about effectiveness is based only on the quality and quantity of available evidence, pending the results of further research. Given the many consequences of claims about "what works," it is important to use a high threshold for the strength of scientific evidence. The current state of the evaluation research, however, creates a dilemma in attempting to draw conclusions. Using studies that are scored at level five on the Maryland Scale as the "gold standard" of evaluation design, results in the scientific methods scores for most of the available evaluations being relatively low. A very conservative approach to determine what works might require at least two level five studies showing that a program is effective (or ineffective), with the preponderance of the evidence in favor of the same conclusion. Employing a threshold this high, however, would leave very little research upon which to draw conclusions. There is a clear trade-off between the level of certainty in the conclusions that can be drawn and the level of useful information that can be gleaned from the available research. Excluding the findings of moderately rigorous studies would waste a great deal of information that could be useful for policy-making.

This report, like the Maryland Report, takes the middle road between reaching very few conclusions with great certainty and reaching many conclusions with very little certainty. In short, this means eliminating from consideration studies that are so methodologically weak that they are scored as at level one on the Maryland Scale. Studies that score at level two, however, are not eliminated from consideration. While they lack strong scientific rigor and can not provide the sole bases for any conclusion, they do provide some worthwhile information. This is particularly true in program areas that include very scant evaluation research. These admittedly weak studies may be the only information that is available in an area. In program areas that are more well-researched, level two studies become part of the preponderance of evidence, but do not serve as the primary source of evidence. Here, their findings are given reduced weight relative to those of more scientifically rigorous evaluations.

Based on the scientific strength and substantive findings of the available evaluations, this report classifies all local programs into one of four categories: what works, what doesn't, what's promising, and what's unknown.

What Works: These are programs that we are reasonably certain reduce recidivism in the kinds of contexts (and the types of participants) in which they have been evaluated, and for which the findings should be generalizable to similar settings in other places and times. Programs defined as "working" must have at least two "level three" evaluations with statistical significance tests showing effectiveness and the preponderance of all available evidence supporting the same conclusion.

What Doesn't Work: These are programs that we are reasonably certain fail to reduce recidivism in the kinds of contexts (and the types of participants) in which they have been evaluated, and for which the findings should be generalizable to similar settings in other places and times. Programs defined as "not working" must have at least two "level three" evaluations with statistical tests showing ineffectiveness and the preponderance of all available evidence supporting the same conclusion.

What's Promising: These are programs for which the level of certainty from available evidence is too low to support generalizable conclusions, but for which there is some evidence predicting that further research could support such conclusions. Programs are defined as "promising" if they have at least one "level three" evaluation with significance tests showing their effectiveness in reducing recidivism, and the preponderance of all available evidence supports the same conclusion.

What's Unknown: Any program not included in one of the above three categories is defined as having unknown effects. There is simply not enough research, or not enough research employing adequate scientific rigor, upon which to draw even tentative conclusions. Program areas with unknown effects should not be interpreted as ineffective. Succinctly put, the jury is still out.

V. REPORT CONTENT AND LAYOUT

Each of the eleven program areas are reviewed individually in the following chapters. Each chapter is divided into four sections. The first section of each chapter is the introduction, containing a description of the program area, a general discussion of previous research, and evaluation issues. The second section applies Maryland’s "what works" criteria to the research to assess the effectiveness of the eleven individual program areas. The third section is an "Assessment of the Research," containing a discussion of the specific program evaluations reviewed for this report, their scientific rigor, and individual findings. The conclusion section provides a summation and additional issues important for consideration. Additionally, implications of the findings for Washington Department of Corrections programming are discussed. A list of references for each chapter is included in Appendix C.

Appendix B contains one or more tables for each chapter that summarizes the findings of the individual program evaluation. The studies are arranged on the table in order of the strength of scientific evidence they provide, with ranking highest on the scientific methods scale listed first, followed in descending order by less methodologically rigorous program evaluations. The first column of each table contains the study’s author and year of publication or report submission. The second column of each table contains the scientific methods score determined by the Maryland Scale and the total sample size of the study. Jumping ahead momentarily, the fourth and final column in the table contains a synopsis of the findings for each method the study may have used to measure recidivism, such as rearrest, reincarceration, and revocations. In addition to listing the type of outcome measure, this fourth column displays the results of statistical tests (when available) on differences in recidivism between the program and comparison groups.

Returning to the third column, this is a display of "effect sizes" for the recidivism measures listed in the fourth column. Effect size essentially refers to the magnitude of the "effect" of the program on recidivism. Bigger program effects (impacts) are more desirable than smaller program effects. Similarly, bigger effect sizes are more desirable than smaller effect sizes. An arbitrary criteria has been develop to determine what constitute a "big" effect size as opposed to a smaller one (see Cohen, 1977). Effect sizes of .20 are considered small, medium effects sizes are those around .50, and large effects sizes are those .80 or higher. Because there are many factors that influence the magnitude of effect sizes and the limited information provided by the studies allowed only conservative approximations of effect size, this report does not incorporate effect size values into decisions about program effectiveness. They are provided as a conservative estimate of the magnitude of the program effect on the various recidivism measures.

This report is not an audit or a performance evaluation of Washington Department of Corrections (DOC) programs. Prepared materials explaining the content of DOC programs served as the basis for all discussion of the program areas. Therefore, based on this report, no conclusions can be drawn about what specific DOC programs "work" or do not work. Similarities and differences are noted between the reviewed programs evaluated and DOC programs, based on the information available about each. This comparison is intended to demonstrate the degree to which Washington’s correctional programming fits within the context of available evaluation research. While this report might suggest effectiveness or ineffectiveness in particular program areas, methodologically-rigorous evaluations of specific Washington correctional programs are needed to drawn definitive conclusions about the affect of DOC programming on recidivism.

I. INTRODUCTION

Correctional education entered the American scene in 1870 with the presentation of Zebulan R. Brockway’s theory of rehabilitation at the first Conference of the American Prison Association (Reagan & Stoughton, 1976). Central to his theory of rehabilitation was the notion that law abiding behavior was attainable through legitimate industry and education (Reagan and Stoughton, 1976). Almost a decade later, correctional education reached the forefront of penal philosophy. The United States Office of Education’s Adult Education Act became the vehicle through which correctional education, and especially adult basic education, made its entry into the prison environment (Ryan & Mauldin, 1994). This change was fueled by the increased interest in prison education witnessed in the 1980s. Programs such as American Broadcasting Corporation’s (ABC) Project Literacy U.S. (PLUS), Literacy Initiatives, Literacy Volunteers of America, and National Adult Literacy Congress emerged and helped the establishment of education in the penal setting (Ryan & Mauldin, 1994).

The increasing focus on education as a means of rehabilitation prompted researchers to conduct empirical studies analyzing the effectiveness of correctional education programs. Initial findings concluded that poor educational achievement was a risk factor associated with delinquent behavior. An analysis of past studies identifies school attachment, educational achievement and educational aspirations as factors inversely associated with delinquent behavior. By 1994, a low level of personal education was categorized as one of the six major need/risk factors within the general personality and social psychology of criminal conduct (Andrews and Bonta, 1994).

The belief that poor educational achievement was linked to deviant behavior prompted a second wave of research focusing on the rehabilitative effects of correctional education programs. The relevant question was whether education could transform an offender into a non-criminal member of society. An influential study addressing this question was published by Martinson in 1974. In What Works? Questions and Answers About Prison, he asserted that there was no definitive evidence that adult academic education related to lower rates of recidivism. However, a later review of the six studies used in his evaluation concluded that Martinson incorrectly characterized his findings. While three of the studies failed to produce statistically significant results, two others (i.e. Saden, 1962; Schnur, 1948) found a direct correlation between adult education and reduction in recidivism, and the remaining study (Glaser, 1964) was dismissed by the reviewers as unreliable (Gerber & Fritsch, 1995). There remains controversy in the literature about whether adult basic education programming is effective in reducing recidivism among correctional populations.

III. ASSESSMENT OF THE RESEARCH

A number of studies on correctional education made use of methodologies that render their findings questionable. First, most of the research studies analyzing correctional education evaluate the educational program in a vacuum. They did not take into account social and other factors present in the offenders’ release environment. Without proper consideration of the post-release environment, it is difficult to determine whether success or failure is due to the program in question or factors unrelated to the correctional intervention (Gerber & Fritsch, 1995). A second category of concern related to the methodology of available research on correctional education is that self-selection of program participants, rather than random assignment to treatment and control groups, may alter the results and mask the true effects of correctional education on offender behavior. Random assignment is ideal because it maximizes the probability that research results are due to the affects of the program rather than to differences between the participants themselves. A final major concern relating to the correctional education literature is that many evaluation studies do not provide an adequate follow-up period. Longer follow-up periods provide more reliable evidence of a program’s impact on individual behavior. Those that span only a few months do not provide sufficient evidence to draw definitive conclusions about program effects on behavior. Rather, brief follow-up periods (generally those measured in months rather than years) leave room for doubt that behavior has been meaningfully impacted. These methodological considerations are important cautions when considering the current status of correctional education in the empirical literature.

The literature reviews and meta-analyses of the late 1980s indicated that while some studies clearly supported the efficacy of academic educational programs, this sentiment was not unanimous nor overwhelming. It remained an unsettled issue whether Adult Basic Education and/or GED training can effectively reduce recidivism, and if so, under what conditions. As shown in Appendix B, twelve recent studies were identified which address this question. Many of the educational evaluations were rated low (Maryland Scale scores of level one or two) because the studies compared only participants or completers with others. No attempt was made to identify a reasonable comparison group or to compare the characteristics (such as sex, race, age, prior criminal activity, etc.) of the participants to the comparisons. As a results, we do not know what types of individuals entered and/or completed the educational programs. There is a good chance that volunteers for program participation are already at a lower risk for recidivism than others who were not willing or interested in obtaining education. Thus, in these situations, we cannot conclude that the educational program changed the offenders. Of the twelve available evaluations, five (Harer, 1995a, 1995b; Ohio Department of Rehabilitation, 1995; Texas Youth Commission, 1995; Walsh, 1985) studies were reasonably well conducted, rated at level 3 or 4 on the Maryland Scale. However, many of these did not use statistical significance tests, and those that did failed to produce significant findings. Furthermore, generally speaking, the effect sizes were moderate or low. Therefore, we conclude that adult basic education appears to be a promising strategy for reducing recidivism.

Research by Harer (1995a) examined the impact of general education program participation, rather than focusing specifically on Adult Basic Education. His analysis of 1,205 released offenders found that educational programming most positively impacted those inmates that had already obtained a high school diploma. The rate of rearrest for this group was 31.2 percent, which was lower than the arrest rate of the nonparticipants (38.9%). Inmates possessing only an eighth grade educational level had a higher recidivism rate (46.7%) than those participants with already possessing their high school diploma (31.2%). The author speculated that participants with higher levels of academic achievement prior to beginning the education program could benefit more from the program in terms of reduced recidivism. In a second analysis, Harer (1995b) reproduced these findings with a subset of the offenders from the first study. This second analysis contained only those offenders who were released from Federal prison or halfway houses who had served at least one year in prison. Again, Harer found that education program participants had lower levels of recidivism than non-participants. While these two analyses find positive results for education programs, including Adult Basic Education, neither analyses employs statistical tests. This reduces the confidence that can be placed in the findings. Without statistical tests, it is not possible to estimate that likelihood that the results were due to chance rather than actual differences between the participating and non-participating groups.

Three other studies, while not as methodologically sound (all ranked at level 3 on the Maryland Scale) as Harer’s (1995a, 1995b) two analyses, generally support his research results. The Ohio Department of Rehabilitation and Corrections (1995) collected data on its 1992 released cohort to determine the impact that ABE and GED training had on offender recidivism. It found that GED participants and graduates experienced a lower rate of recommitment (24.1% and 27.1%), than the comparison group (32.3%) during the two years following release. On the other hand, ABE participants returned to prison at a higher rate (32.3%) than the corresponding control group (30.6%). This elevated rate of returns to prison for ABE participants dropped to 26.3% for those inmates who were released after being incarcerated for four or more years. This led the researchers to conclude that ABE programming was especially well-suited for inmates serving long prison terms. As with the previous studies, however, no statistical tests were conducted to support this conclusion. Again, there is no estimate the likelihood that these results might have occurred by chance.

Another evaluation conducted by the Texas Youth Commission (1993) undertook a similar analysis the Ohio study, but this study did employ statistical tests. Utilizing a sample of 1,717 released inmates, the researchers found that the reincarceration and rearrest rates for GED graduates were substantially lower (10.1% and 41.3%, respectively) than those of the comparison group (19.1% and 53.5%). However, these differences were not statistically significant, meaning that the findings could have been due to chance variation among the groups. Comparable findings to those of the Texas Youth Commission were found by Walsh (1988) in another study. This research evaluated the Lucas County Adult Probation Department’s GED program. He found that GED graduates and GED participating noncompleters were rearrested at lower rates (16% and 32%, respectively) than nonparticipants (44%). However, unlike the Texas study, the difference in recidivism rates between graduates and non-participants were found to be statistically significant. These findings led Walsh to conclude that "the crucial variable is not participation per se, but rather whether or not the program was completed" (Walsh, 1985: 73). A further analysis of offender criminal histories showed that while GED completion did impact recidivism more intensely than mere participation, involvement in the program significantly reduced the frequency and/or intensity of the participants’ criminal activity.

Together, these relatively well-designed studies do not provide substantial evidence of the effectiveness of Adult Basic Education Programming. While some find positive results, only one study (Walsh, 1988) statistical significant differences indicating reduced recidivism for program completers. While there are a number of other available studies undertaken to evaluate the effectiveness of Adult Basic Education Programs, they are of such poor methodological quality (all score at level 2 on the Maryland Scale for Scientific Rigor) that little confidence can be placed in the findings. Despite their failings, collectively these studies generally show positive results for education programming for correctional offenders. These studies are discussed below.

Ramsey (1988) utilized an experimental design to test the impact of GED training on reincarceration and rearrest. She found GED program graduates had a lower rate of reincarceration (16%) and rearrest (32%) than nonparticipants (36% and 38%). The effectiveness of the GED training did diminish for those inmates that chose to participate but failed to complete the training. This group of individuals experienced higher reincarceration and rearrest rate (32% and 33%), yet still managed to do better than nonparticipants.

The research conducted by Adams and colleagues (1994) attempted to determine how much participation in these educational programs was enough to trigger the beneficial effects of education. An analysis of 14,411 inmates released from Texas prisons in 1991 and 1992 revealed that recidivism was positively impacted by extensive academic training. Those inmates that engaged in 301 hours or more of course work had a far lower rate of re-incarceration (16.6%) than those who participated for 101-200 hours (20.7%). This latter group, in turn, had a lower recidivism rate than non-participants (23.6%). However, the beneficial effects of education vanished if the inmate had less than 100 hours of training, in that, their rate of reincarceration was actually higher than that of nonparticipants (25%). Adams and colleagues also learned that the positive effect of education on re-incarceration was influenced by the inmates’ initial educational level. This is consistent with Harer’s (1995a) findings. In another study, Porporino and Robinson (1992) found that offender risk level may also influence that impact. These researchers conducted an analysis of data on 1,736 released offenders and found that ABE graduates had the lowest readmission rates (30.1%) when compared to ABE participants who were released before completing the program or those that chose to drop-out (35.7% and 41.6%). Furthermore, offenders in the high risk mandatory supervision category who had obtained a GED while incarcerated, showed a readmission rate that was 10 percent lower than those offenders who had withdrawn prematurely from that program. However, like the research by Adams and colleagues (1994), Porporino and Robinson compare program completers to non-completers making these findings questionable.

Fabelo (1992) used the Reading to Reduce Recidivism Program to examined the effects of institutional and community-based GED training. It found that while recidivism is reduced for community GED graduates and participants, this effect did not occur in community program dropouts. Of the 21 successful field GED participants, only one returned to prison during the 18 months follow-up period. Eighteen of the thirty-one community dropouts were either returned to prison or incarcerated, which amounted to a higher recidivism rate than the control group. However, problems with the research design and small sample size limit the credibility of these findings.

In a study funded by the Adult Probation Department of the Superior Court of Pima County, Siegal and Basta (1997) examined the effectiveness of GED and Principle of the Alphabet Literacy System (PALS) programs on offender rearrest and reconviction rates. They found that participants in both the GED and PALS programs experienced a lower rearrest rate (35% and 24%) than the control groups (46% for both). This difference proved to be statistically significant for the GED students. However, the same positive effect did not translate over to rates of reconviction. While the PALS participants still experienced a slightly lower rate of reconviction (20%) than the control group (22%), the rate of reconviction for GED participants remained unaffected (21% for both).

The New York Department of Correction (1992) found that graduates of its education program experienced a reincarceration rate of 34 percent compared to a rate of 39 percent for noncompleters. This difference was found to be statistical significant. Another study, (Piehl, 1995) reported the same findings with 1,473 Wisconsin inmates. Consistent with the above findings, this research found that completion of an education program was associated with a lower rate of reincarceration (33.5%) than mere participation (40.2%). Those inmates who eligible but failed to complete the training returned to prison at a higher rate than program graduates.

Schumacker, Anderson, and Anderson (1990) study sought to examine what type of educational program affected recidivism the most. They found that programs incorporating elements of vocational and academic training led to a more substantial reduction in recidivism than mere academic course work. The mastering of a trade, along with basic reading and arithmetic skills, resulted in a reconviction rate of 19 percent, which was lower than that experienced by the non-participants (28%). Purely academic training resulted in a 22 percent reconviction rate. Thus, while the completion of a GED increased the offenders post-release success, those inmates that also received vocational training had the highest employment rate and lowest criminal activity rate over the twelve month tracking period.

The evaluations conducted by Tracy & Johnson (1994), Cogburn (1992), Hull (1995), and Correctional Education School Authority (1990) were too methodologically weak to be included in this review (i.e. ranked at level 1 on the Maryland Scale).

This study of correctional education for adult offenders suggests that no one strategy is appropriate for all offenders in all situations. The various and often contradicting results seem to indicate that successful reduction of recidivism depends greatly upon the particular population targeted. Of 12 studies presented, 8 produced results consistent with the notion that correctional education may positively impact the rate of recidivism. While some ABE or GED participants realized a reduction in their criminal involvement, others were either completely unaffected, or negatively influenced by their participation in the educational experience. The one consistent finding that emerged from this research is that completion of, as opposed to some amount of participation in, an educational program is associated with lower recidivism. This finding is most likely a sample selection bias. Those who are at high risk for recidivism drop out of programs, while those who are at lower risk generally stay in programs through completion.

Thus, like the meta-analyses reviewing data from the 1980s, research in the 1990s has failed to reach a definitive conclusion regarding the effectiveness of correctional education on recidivism. The reason for this failure partially resides in the lack of methodological rigor noted in the majority of the studies in this area of correctional programming. Research in the area is riddled with problems, such as a lack of random assignment and statistical analyses. Scientifically-based conclusions cannot emerge from such a limited pool of knowledge. Eventually, as research methods improve, a clearer answer will emerge with respect to the effectiveness of correctional education on recidivism.

The Adult Basic Education programs offered by the DOC are designed to assist offenders in achieving competency in several academic subject fields, including reading, writing, and mathematical skills. The program consists of three levels, through which offenders gradually work to increase their level of skill. The programs also include an emphasis on individual learning styles, vocational skills, and financial management. The program uses a classroom education approach and employs group learning techniques. Our assessment of the literature found that Adult Basic Education Programs appear promising, but the limitations of the research do not allow any stronger statements about effectiveness. While many of the programs evaluated in the literature were described as containing similar components to those offered by the DOC (e.g. reading, writing, mathematical skills, vocational skills) it is impossible to disentangle whether these program components can account for the (admittedly weak and tentative) support of the educational programming. Based upon our assessment of the literature, there is reason to suspect that future evaluations of Adult Basic Education Programs, such as those offered by DOC, might continue to produce promising results.

Correctional industry is a term used to describe a range of offender employment-related activities that occur almost exclusively during an offender’s term of incarceration. While some industrial facilities are located outside of prison walls, correctional industry workers are typically serving time in some type of residential facility (Correctional Industries Association, 1998). In terms of correctional industry participant characteristics, Flanagan (1988) and his colleagues found that offenders in participating in correctional industry programs tended to be older, serving longer sentences, had better pre-incarceration employment records, and were less likely to be drug users than a sample of non-industry participants.

Correctional industries produce a wide range of products and services for both government and private sector consumers, including furniture, health technology, automobile parts, institutional and jail products, signs, printing products, textiles and apparel, traffic paint, and food products. In addition to the production of goods and services, correctional industries can serve other purposes. According to Maguire, Flanagan, & Thornberry (1988), correctional industries reduce operating costs for prisons, produce income for correctional systems through the sale of inmate produced goods, decrease inmate idleness, and provide valuable skills and work experience that inmates can use outside of prison. Correctional industry experience can provide opportunities for inmates to develop better time management skills, self-discipline, and work ethics. However, correctional industries often use outmoded production techniques and equipment, train inmates in areas that already have an excesses of laborers or that require professional licenses that are difficult for ex-offenders to acquire. These and other problems associated with correctional industries make it more challenging for offenders to secure gainful employment outside of the supported work environment of prison. (Maguire, et al, 1988).

In June 1996, Minnesota Department of Corrections Industry Division (MINNCOR) conducted a nation-wide survey of correctional industries, identifying 54 separate jurisdictions operating some type of correctional industry program. While, since mid-century, these programs have largely been a state-run enterprise. However since 1980, private industry involvement in corrections has increased (Henry, 1988), reflecting the historical roots of correctional industries. Early in this century, private correctional industry ventures were common. However, the increases in the power of trade unions in the 1940s and concerns about the humanitarian treatment of inmates brought about the decline of private correctional industry. By the 1970s, however, restrictions on the sale and interstate transportation of prison-made products were lifted, and private industry quickly began to move back into partnerships with correctional industry. By 1984, the National Institute of Justice reported 26 prison-based industries in nine states which had substantial private industry involvement (Henry, 1988). The number of inmates working in these ventures by 1984 was nearly double the number working in private correctional industry in 1980.

III. ASSESSMENT OF THE RESEARCH

Many correctional industry programs contain other intervention components, such as job search or vocational education/apprenticeship training. This makes evaluations of "correctional-industry-only" programs relatively rare. Therefore, evaluations of multi-component programs, including vocational education, correctional industries, and job search assistance or work release services are reviewed in this section. Unfortunately, evaluations of combined approaches make it extremely difficult to assess the effectiveness of correctional industry participation alone. Another problem with correctional industry evaluations is their relatively weak scientific rigor. Nearly all the evaluations reviewed here include some substantial methodological shortcomings. Typically, there is a failure to use random assignment. A related methodological problem in these evaluations is the failure to deal adequately with the issue of subject selection bias. A selection bias most commonly occurs when offenders who are less likely to recidivate (even without the program) are compared to a group of offenders who were initially more likely to engage in future criminal behavior. For example, correctional industries often require minimum levels of academic achievement for potential participants. This can introduce bias because offenders who have more academic education may represent a group of lower risk offenders. Finally, attrition (dropouts) from the intervention program produces another interpretational problem, since many evaluations typically compare program completers to those who received no program services whatsoever. Again this is an inappropriate comparison because those subjects who complete a program may be different (potentially less serious offenders) than those who drop out.

Despite the shortcomings of many studies, several evaluations were available which met at least minimal requirements for scientific rigor. The most rigorous of these evaluations (Maguire et al., 1988) (ranked at a level of 4) did not find a significant reduction in felony rearrest rates for correctional industry participants (29%) compared to non-participants (34%). Two studies of the Federal Bureau of Prison’s UNICOR program by Saylor and Gaes (1992,1996), while somewhat less rigorous (ranked at a level of 3 on the Maryland Scale), do find several significant reductions in recidivism. The program participant group consisted of 57 percent with only correctional industry work experience, 24 percent with vocational and apprenticeship training, and 19 percent with both correctional industry work experience and vocational education/apprenticeship training. The researchers used statistical control for differences between groups. They compared those with at least 6 months of industry participation to those with no industry work experience. After 6 and 12 months of post-release follow-up, correctional industry participants had significantly fewer parole revocations (4.9% and 6.6% respectively) as compared to non-participants (6.6% and 10.1%, respectively). They also report that after a follow-up period of between 8 to 12 years, correctional industry participants had a statistically significant lower risk of being recommitted for a new offense than non-participants in the control group.

Finally, two other studies reviewed state-run industry programs (Anderson, 1995a; Boudouris, 1985). Conclusions based on the results of these evaluations must be interpreted with extreme caution, as neither study employed a more than a minimally sound research design (both ranked at level 2 on the Maryland Scale). Anderson (1995a) conducted an evaluation of a correctional industry program where participants worked at jobs that were divided into five categories based on skill level. High skill level jobs included auto mechanic, drafter, dental lab technician, machinist, and welder. Medium skill level jobs included fabric cutter, graphic arts technician, machine presser, and printer. Low skill jobs included box, chair, or brush factory jobs, as well as cushion maker, wood assembler, and sewing machine operator. Entry level jobs included janitor, sander, and material handler. Finally, clerks’ jobs included payroll clerk, production clerk, and typing clerks (Anderson, 1995a). followed an entire release cohort, and found that correctional industry program completers had lower rates of return to prison after 24 months (28.7%) than those who dropped out of the program (30.1%) and those who did not participate (31.3%). Anderson did not conduct statistical tests for these differences in recidivism rates, therefore it is not possible to tell whether these differences are due merely to chance or if they represent actual reductions in recidivism. Boudouris (1985) reports that after a 24-month follow-up, inmates with correctional industry, vocational, or farm work experience had fewer rearrest, revocations and returns to prison (approximately 9%) than those with education experience alone (approximately 24%). The difference in recidivism rates for participants and nonparticipants was statistically significant.

In light of the serious methodological flaws in many of these evaluations, there are several studies ranked at the bottom of the Maryland Scale for scientific rigor (see Clasby, 1996; Motiuk & Belcourt, 1996; and Pride Enterprises, 1997). As a result, these studies are not included in this review.

IV. CONCLUSION

Based upon the above findings and the methodological rigor of the evaluations, we conclude that multi component correctional industry programs (containing vocational education, apprenticeship, and/or job skills training), are effective in reducing recidivism. The two studies conducted by Saylor and Gaes (1996, 1992) were judged to be of sufficient scientific merit to allow at least preliminary conclusions. The two studies did find significant differences between industry participants and others in most measures of recidivism. However, it should be noted that the differences in recidivism between industry participants and others is relatively small in most of the studies. Most likely, the results are significant because of the large number of subjects in the studies. As such, our conclusion about the effectiveness of such multi modal correctional industry programs must be regarded with caution.

Implications for Washington Department of Corrections Programming

DOC institutions operate and maintain programs that provide in-prison employment positions for inmates. These positions are provided within institutions to contribute to the overall maintenance and management of the institutions. The program is intended to provide offenders with new work skills, while reducing both inmate idleness and the overall operating costs of the institutions. The employment positions offered by the DOC include institutional maintenance jobs, clerical work, equipment operation, and personal services jobs, such as barber and dental assistant. Participants in the correctional industries program also receive an income from their employment, with part of their earnings being applied to crime victim compensation and legal fines and fees. Our assessment of the research concluded that correctional industry programs are effective in reducing recidivism. Much of the research examined, however, evaluated programs that included components other than correctional industry work, so the effects of these programs alone have yet to be isolated. This means that this research is only relevant to DOC programming to the extent that offenders participate in more than DOC correctional industry work positions, but also other vocational and education programming.

For several decades, the relationship between poverty, poor educational achievement and crime has been widely studied by social scientists. The relationship between poverty and crime may be one of the most firmly established in the criminological research (Berk, Lenihan, & Rossi, 1980). The link between school failure and criminality has also been demonstrated (Ward & Tittle, 1994). The prevalence of under-education among offender populations was demonstrated by Freeman (1992), who found that 60 percent of male prisoners between the ages of 25 and 34 had less than 12 years of formal education. Previous research on unemployment and crime has been equally discouraging, with one study reporting that as many as 84 percent of inmates were unemployed at the time of their current offense. The relationship between crime and unemployment has been a consistent research finding over past 60 years (Wolfgang et al., 1972; Sampson & Laub, 1993).

Since lack of education and unemployment tend to be correlated with crime, reduction of these two risk factors has been proposed by many as a means to reduce crime. Many criminal justice system-based programs have been developed in an attempt to increase educational achievement and employability for offenders. Though they share a number of similarities, the primary focus of these programs vary. Some programs, such as the Manhattan Court Employment Project, have attempted to divert offenders from the criminal justice system and into stable employment (Baker & Sadd, 1979). Others, such as the POWER program, project TRADE, and the Wisconsin STEP program, have attempted to provide general education and vocational education to offenders during or following incarceration. Still other programs have attempted to provide inmates with hands-on industrial experience, such as the Canadian CORCAN program and the U.S. Federal Bureau of Prisons’ UNICOR program (See Chapter 11 on Correctional Industries). Pre- and post-release financial and job search assistance have been provided through work release programs, day reporting centers, and halfway houses.

Some programs combine elements of several separate approaches, sometimes integrating vocational education, trade apprenticeship, prison industrial work experience, and pre-release job search assistance. Some researchers (Johnson, 1984; Beck, 1986; Gendreau, 1993; Bushway & Reuter, 1997) have suggested that combining interventions, and thus targeting multiple offender needs, would increase program effectiveness. While such multifaceted programs may be desirable, they often present a challenge for program evaluation, in that, interpretation of results can be complicated. Some aspects of a multifaceted program may be more effective than others, but when they are simultaneously provided to participants, it is difficult (if not impossible) to disentangle the effects of individual program components. Consequently, interpretation of program evaluation literature is not an easy task and must be considered in the context of program ingredients.

General vocational educational programs involve the provision of classroom opportunities to acquire basic work-related knowledge, such as the basic math skills needed for automotive mechanics or construction tasks. This type of educational program is widely offered and is typically a prerequisite for the more hands-on vocational or job training experiences. The Vocational Delivery System of the North Carolina Department of Corrections (Lattimore, Witte, & Baker, 1990) is an example of a vocational training program without an in-prison "apprenticeship" type component. Participants in this program are offered vocational classes, as well as academic and self improvement courses (i.e. Alcoholics Anonymous). Some vocational programs, such as those offered by the Wisconsin Department of Corrections, offer accreditation to those offenders who complete the program, therefore making it easier for an offender to obtain a trade license (Piehl, 1995). More hands-on types of vocational or apprenticeship training are sometimes offered in conjunction with outside organizations, such as the Home Builders Institute’s (1996) Project Trade, which offers a series of construction job training programs for both juvenile and adult offenders. Project Trade attempts to increase offenders basic knowledge of mathematics and vocational skills related to construction trades, as well as improve post-release employment possibilities, through a partnership between the correctional institution and local tradesman. In addition, Project Trade instructors provide community work projects where offenders can gain actual work experience.

III. ASSESSMENT OF THE RESEARCH

Despite the serious methodological limitations of many evaluations in this area, there are also several that are relatively well designed. One study that ranked at the top of the Maryland Scale for research design (level 5) was conducted by Bloom and colleagues (1994) examining the program created by the Job Training Partnership Act (JTPA). JTPA programs were evaluated in several cities and varied in the amount and type of services actually provided. For the most part, JTPA programs provided classroom training in occupational skills, on-the-job skills training, job search services, basic education, real world work experiences, as well as other related services. The programs can, however, vary widely in content. While JTPA programs do not specifically target offender populations, the programs do serve young people with prior arrest records. Among this group, the researchers found that, after a 36 month follow-up period, male JTPA participants had a higher arrest rate (59.2%) than non-participants (55.7%). No statistical tests were conducted on this difference, however. Among males with no arrest history, JTPA participation was associated with a significantly higher arrest rates (25.8%) 36 months after completing the program, than non-participants (18.7%). Finally, after 36 months, female youth JTPA participants had higher arrest rates (7%) after program participation than controls (5.3%), but this difference was not statistically significant. The researchers concluded that the JTPA approach to job training for youths, with and without prior records, may actually increase recidivism.

We were able to uncover only one study (Lattimore et al., 1990) that used a random assignment design to examine the effectiveness of a vocational education program for prison inmates. Components of the program included an assessment of vocational interests and aptitude, specific skills training and post-release employment assistance. While largely methodologically rigorous (ranked at a level of 4 on the Maryland Scale), the evaluation had some problems with attrition and program implementation. The researchers randomly assigned 16 to 22 year old property offenders to a vocational education program or a control group. Findings revealed that vocational education participants had significantly lower new crime reconviction rates (36%) than controls (46%). The authors concluded that the vocational education program was effective in reducing recidivism among its participants.

Many of the studies of vocational education programs compare the recidivism of those who participated in a prison-based vocational education program with those who did not. Groups are identified after they leave the prison, so can not they are not randomly assigned to the vocational education programs and comparisons. Because of this, we can never be sure that the participants were not different than the non-participants before they entered the vocational programs. The better studies attempt to equate the groups in some way (e.g., by matching) or statistically control for characteristics that may result in differences between groups.

For example, Harer (1995a) (ranked at a level of 4) conducted an evaluation of educational programs in general, which included vocational education. The study used statistical techniques to control for attrition and selection effects. Using the entire sample of 1,205 subjects, Harer found that there was an approximate 20 percent reduction in recidivism (statistically significant) due to education program participation. Education course completers (those completing at least one-half of a course in a six month period) had a lower rate of rearrest or parole revocation (30.1%) than those who completed less than one-half of a course in a six month period (39%) and those who completed no courses (44.5%). In a second study, Harer’s (1995b) sample consisted of 619 offenders (drawn from the first study) who had spent at least one year in prison. The results were that education program participation reduced the likelihood of rearrest or parole revocation by about 50 percent. Additionally, among those who were rearrested or experienced a parole revocation, education program participants remained in the community longer without failing (65%) than those who did participate. While the results of this study are somewhat promising, it is impossible to determine if the positive outcomes are the result of vocational education, per se, or of some other type of educational program included in the research.

Four other studies (Van Stelle et al., 1995; Adams et al, 1994; Downes et al., 1989; Saylor & Gaes, 1996) with weaker scientific rigor (all scored at level 3 on the Maryland Scale), generally fail to find statistically significant differences in recidivism between those who complete the program and those who drop out or do not participate.

One of these studies (Van Stelle, 1995) evaluated a comprehensive program in Wisconsin (the STEP Program-Specialized Training and Employment Project). This program involved three phases, institutional training, institution/community transition, and a community component. The institutional training component required participants to maintain a full-time job, participate in STEP curriculum classes, attend adult basic education classes (if necessary), and maintain acceptable conduct within the institution. The STEP classroom curriculum involved meeting five times a week for 1.5 hours over a six month period. These meetings focused on developing an attitude conducive to employment, developing life skills, reflecting on long-term work related plans, operationally defining short-term objectives, and developing a work history indicating motivation to complete a project. Training modules in this STEP curriculum included 34.5 hours of employability skills training, 42 hours of cognitive interventions training, 15 hours of money management training, 19.5 hours of relationships training, 15 hours of assertiveness training, 15 hours of anger management training, and 15 hours of pre-release preparation. In the institution/community transition component included transfer to a minimum security facility and meetings between with community staff to assist with employment placement. STEP participants were contacted by an outside employment agency representative prior to release to assess each student’s interests, abilities, and circumstances, build a rapport with the agency contact, and set realistic job expectations. Finally, the community component included employment assistance, intensive supervision, and support services. STEP participants were required to maintain full-time employment, comply with supervision conditions, attend criminality relapse prevention counseling, and continue to address their educational needs. The findings revealed that STEP program appeared to increase recidivism rather than decrease it. Program completers had a significantly higher average number of parole violations (M = 6) than controls (M = 1). In addition, they found that STEP program completers had more rearrests (24%) than controls (19%), although this result was not statistically significant. Van Stelle and colleagues examined the differences between those who completed and those who dropped out of the program. As a result, no data are available on who STEP completers might have compared to those who never entered the program.

Another study (Adams et al., 1994) was a large scale evaluation of the effects of education programs on an entire group of inmates released from prison in Texas. This evaluation (ranked at a level of 3) had substantial shortcomings in terms of its control for the effects of program drop-outs and the selection of subjects for each group. The authors found that those who participated in a vocational education program had fewer returns to prison (20.9%) than controls (25.1%), but the difference was not significant. Another study (Downes et al., 1989) of similar scientific rigor also found no significant differences between program participants and non-participants. This study assigned offenders to program and comparison groups by matching them on a few characteristics. Program participants were involved in vocational and/or adult basic education programs. All instruction was individualized, competency-based courses that lead to either certification or associates degrees in specific vocational areas. Participants could also enroll in several college-level courses offered by the state’s higher education system. In addition to academic and vocational education, the program included a career services component which helped students with job placement services, job preparation and other life skills, (thinking skills, good work habits, and social/interpersonal skills). The results of the study indicated that participants had more parole failures (24%) than non-participants (20%), but the difference was not statistically significant. The research design was rated a level 3 on the Maryland Scale because it was relatively well-conducted, but, among other drawbacks, small sample sizes detract from the confidence that can be placed in these results.

Finally, Saylor and Gaes (1996) assigned offenders to the treatment and comparison group by matching them on particular characteristics, but like the previous study did not use statistical control for program drop-outs. After a follow-up period ranging from 8 to 12 years, the researchers found that those who participated in the vocational training or apprenticeship program had a 33 percent lower risk of new offense returns to prison than the matched control group. This difference was statistically significant. During the same follow-up period, those who had prison industry, vocational training, or an apprenticeship had a 23 percent lower risk of new offense recommitment than the matched controls. This difference, however, was not significant.

Five studies (McGee, 1997; Ryan, 1997; Anderson, 1995b; Piehl, 1995; Schumacker et al., 1990) were judged to be so weak in scientific rigor (level 2) that little confidence be placed in their results. Collectively, however, they still provide some useful information, in that, all four consistently show positive results for vocational education programs.

McGee (1997) evaluated post-secondary education programs offered in prison, including both academic and vocational education. However, the study only compared completers and non-completers. The findings revealed that, after an average follow-up period of three years, male vocational education program completers had fewer returns to prison (13.4%) than matched comparison non-completers (38%). For females, after the same average follow-up period of three years, vocational education program completers had fewer returns to prison (5%) than the matched comparison non-completers (25%). Overall, of vocational education program completers who were employed had fewer returns to prison (8.7%) than matched comparison non-completers (38%). These results are appear positive for vocational education programming, the study utilized a very weak design and did not conduct statistical significance tests. Schumacker and colleagues (1990) conduct an evaluation of a vocation education program that incorporated the use of random assignment to the program and control groups. However, the researchers failed to use any statistical controls for the detrimental effects of program drop-out. The researchers found that the vocational education participants had fewer parole violations (21%) than controls (28%). No statistical tests were conducted so there is no way to detect whether this difference may be due to chance or an actual difference between the groups.

Ryan (1997) conducted an evaluation of an in-prison carpentry skills training program, which included work on community construction jobs. The results were supportive of vocational education. After a follow-up period ranging from 36 to 84 months, job training program participants had fewer "non-positive parole outcomes" (25%) than non participants (41%). This difference was statistically significant. However, this evaluation has substantial methodological shortcomings and results should therefore be interpreted cautiously. For example, Ryan compared all program completers to a control group generated by creating a pool of offenders matched on age, race and type of crime. From this pool, a random sample was drawn to serve as the control group. This procedure is not a truly random assignment because program participants were not also randomly assigned.

Two other methodologically weak studies also found supportive evidence for vocational education programs. Anderson (1995b) collected data on an entire release cohort, and found that vocational education program completers had fewer returns to prison (28.7%) than the control group (31.3%). There were, however, no tests of statistical significance reported and statistical controls for pre-existing group differences or program drop-outs (attrition) were not used. In another study, Piehl (1995) compared those who completed education and vocational education programs to a group of offenders who were eligible for, but did not participate in each program. Positive results were found for program participation. Vocational and education program completers had significantly lower recidivism (approximately 35%) than the controls who were eligible for vocational or educational programs (approximately 40%). However, there were no statistical controls for or even description of subjects who dropped out of the programs.

Several studies ranked at the bottom (level 1) of the Maryland Scale for scientific rigor (Coburn, 1987; Craig and Rogers, 1993; Gainous, 1991; Green et al., 1993; Home Builders Institute, 1996; Hull, 1995; Jenkins et al., 1995; Maryland Department of Public Safety and Correctional Services, 1989; Tracey and Johnson, 1994a; Winterton, 1995). No reasonable conclusions can be based on the results of these studies due to their weak methodology.

Many authors have reviewed research studies examining the effectiveness of vocational education programs in reducing recidivism, with some concluding that the programs are at least minimally effective under certain conditions or with some groups (Gerber and Fritsch, 1993, Bushway and Reuter, 1997, Tracy and Johnson, 1995). Others conclude that these program do not have any effect on recidivism (Beck, 1978). Most of these authors do agree that there is conflicting evidence. The current review is no different in that regard. While some studies of fairly high scientific rigor have shown positive effects (Harer, 1995, Lattimore et al., 1990) others demonstrated negative impacts on recidivism (Van Stelle, 1995) in some cases even increasing recidivism (Bloom et al., 1994). However, overall there is sufficient evidence to conclude that some vocational education programs appear to be effective in reducing recidivism. However, employment programs, such as JTPA, do not appear to be effective. More research, however, is required before any definitive conclusions can be drawn about this type of vocational training.

Some of the vocational education programs offered by the DOC are labeled as Crime Related programs. They are offered by several sources, including institutions and local community colleges. The programs vary in length and employ an educational model. The vocational education programs offer training in skills necessary for particular types of employment, such as computer literacy, automotive repair, and carpentry. Also, some programs offer employment-related skills, including interview etiquette, résumé construction, and job search methods. Our assessment of the research revealed that vocational education programs are effective in reducing recidivism. The research focused on programming within institutions or programs with both a prison and community component. In short, the research assessed seems to be applicable to the institution-based programming offered by the DOC. The research on institution-based programming was not designed in a way that would allow a separation of what components may be more effective than other components. As such, it is not clear whether job search assistance, specific skills training, or other activities might be important features to include in vocational education programs.

Correctional work programs often involve the use of inmate labor to perform tasks essential to the operation of a prison facility. Examples of work performed by inmates include cleaning, maintenance, painting, grounds keeping and basic carpentry or electrical work. The expressed goal of these programs is often to reduce inmate idleness, as well as prison operating costs (Maguire, et al. 1988). However, few if any of these programs are specifically designed with recidivism reduction as a primary goal. In light of these stated goals, and the fact that these work activities tend to be relatively unstructured, providing little beyond menial or semi-skilled work experience, it is unlikely that they have measurable impacts on recidivism. It may not be accurate to describe many of these activities "programs" in the sense that they are not as structured, intense, or based on inmate needs, as are most correctional interventions. As a result, there is a general lack of research evaluating these activities in terms of recidivism reduction.

One in-prison work activity program is the Washington State McNeil Island Work Ethic Camp. This unique program is a four-month alternative sentencing option for adult male and female offenders. The camp, opened in 1993, is designed to provide an intensive employment experience within the prison environment that gives offenders the "opportunity to develop a positive work ethic and crime-free lifestyle through a regimented program" (Campbell, 1996). Offenders with no history of violence or sex crimes are eligible to participate. According to initial estimates, approximately 600 offenders would enter the program per year. The program itself consists of two components, classroom activities and work. Participants receive an initial needs assessment, including educational level, substance abuse needs, and job interests. Classroom activities are then tailored to those individual needs. Participants are also assigned to work crews where they are supervised by correctional staff. During the first 30 days, they are assigned to "clean up crews" and in the following months are assigned to other details which perform specific tasks, such as working in a meat plant, repairing cars, and so on. Basic job skills and work ethics are reinforced in the course of these work experiences. Transitional aftercare services are also part of the program. Participants are screened for transitional needs and plans are developed to meet those needs, while under intensive community supervision.

Other programs provide employment opportunities and/or services for offenders in community settings. For example, a work release program may allow participants to leave the facility during their last few months in prison to either seek employment or, more commonly, to participate in some type of work program (Turner & Petersilia, 1996). The goal of this type of program is to ease the participants’ transition from institution to community and promote productive, stable employment following release. According to Turner and Petersilia (1996), work release has been a part of many correctional systems for nearly 80 years and, while 43 states have statutes allowing work release programs, only about one third of these states actually operate such programs. The researchers estimate that fewer than three percent of all prison inmates participate in work release. Some of the reluctance of states to utilize such programs no doubt stems from the fear that inmates released to work may pose a public safety risk. In addition, federal funding for these programs is limited.

Halfway house programs are another community-based intervention which has been used in various forms since the 1800s (Latessa & Travis, 1991). In the 1960s, when the move toward deinstitutionalization emerged, halfway houses became more prevalent. More recently, they have seen a re-emergence due to the need of correctional systems to alleviate prison crowding. Halfway house programs have been tailored to fit both the security level and particular needs of offenders. These type programs have specialized on drug treatment, counseling, education or employment (Latessa & Travis, 1991). Halfway houses, in which a substantial program component revolves around the offender seeking and maintaining stable employment, have been evaluated by several researchers (Hartmann, Friday, & Minor, 1994).

Other types of community employment programs attempt to ease the transition from prison to the community by assisting participants to obtain information and assistance in employment while they are in prison or upon release. This section focuses on programs that attempt to provide work experiences, work ethics or employment services to offenders.

III. ASSESSMENT OF THE RESEARCH

There are several studies using strong research designs that enable us to draw conclusions about the effectiveness of community employment. From this work, we conclude that some community employment programs are effective in reducing recidivism. Again, as we have noted in other sections of this report, all types of employment programs are not successful. Clark and colleagues’ (1992) study, incorporating very strong research methods (rated level 5 on the Maryland Scale) of enhanced employment searches suggests that this is an effective method for reducing recidivism. Similarly, while not as strong methodologically, Menon and colleagues’ (1992) study of employment programs suggests that this might be effective for offenders who are at high risk for recidivism. However, there is little overall consistency in the results of the research. In some studies, no differences in recidivism were found between program participants and nonparticipants (Uggen, 1997; Latessa & Travis, 1991) and, in others, differences in effectiveness varies by site (e.g. see Milkman, 1985) or by measure of recidivism (Turner and Petersilia, 1996). We hypothesize that some of these differences are the result of the implementation of the programs and/or the type of offenders who participate in the programs. In other words, as is evident throughout this report, all programs do not appear to work for all offenders. Some of the exploratory research suggests that older offenders may benefit more than younger offenders (Uggen, 1997) or that offenders at higher risk of recidivism may benefit more from employment programs (Menon et al., 1992). Furthermore, the results of Turner & Petersilia’s (1996) study of work release suggests that the increased surveillance of offenders in these programs could lead to more returns to prison for technical violations.

Several studies have investigated the impact of transitional work programs designed to help offenders find employment upon release from prison. Turner and Petersilia’s 1996 evaluation of Washington State’s work release program used a generally high-quality random assignment design (rated level 4 on the Maryland Scale). The researchers attempted to minimize the possible biases resulting from attrition (dropouts) and used other statistical techniques to further reduce the impact of possible pre-existing group differences on recidivism. Despite some shortcomings, the evaluation was generally well designed and conducted. The Washington DOC work release program includes requirements that participants remain in the work release facility at all times (except as approved to go out to work) that they remain drug- and alcohol-free, be employed or have resources to meet their financial needs, report all income to program staff, and obey all laws. Offenders are required to pay a small amount towards their room and board. Dependent on court order, they must also pay all required support to their families and restitution to the court. Residents are responsible for finding work with the assistance of work release program staff. The work release participants were found to have lower rearrest rates (22.3%) and reconvictions rates (7.1%) than nonparticipants (30.2% rearrests, and 7.6% reconvictions) but these results were not statistically significant. None of the nonparticipants were returned to jail, as opposed 3.4 percent of the work release participants (statistically significant). While not statistically significant, the work release participants were also returned to prison for new crime at a higher rate (4.7%) than the comparison group (3.6%). Overall, while somewhat mixed, most of the results of this study were not significant. The two exceptions are that work release participants had higher rates of return to prison and jail for program infractions. This may be attributable to increased supervision or tighter restrictions on work release participants than those offenders in the comparison group. The authors conclude that the program did not appear to succeed in reducing either recidivism or correctional costs.

In another investigation of a transitional program, Menon and colleagues (1992) examined the effectiveness of a Texas in-prison program, called Project RIO. Project RIO, as evaluated by Menon includes the development of an Employability Development Plan, by both a counselor and soon-to-be-released inmate. This plan is used by the parole officer at the release site. In addition, offenders are helped in obtaining necessary work documents, such as social security cards and birth certificates. These counselors also encourage clients to participate in ongoing vocational, apprenticeship, secondary and post-secondary training, as appropriate, while still in the prison. In addition, staff provide job preparation and job search services after release. Where available, usually in more urban areas, participants attend a 20 hour job search workshop focusing on basic skills, such as filling out an application, preparing a résumé, and performing a mock interview. Releasees may also be referred for community services such as inmate support groups, food stamps, bus tokens for job interviews, and on-the-job training programs, as available. Staff also make sure that prospective employers are aware of hiring incentives that may apply to their companies. The evaluation design included a control group of non-Rio participants. Findings reveal that "high risk" participants had significantly fewer rearrests (48%) and reconvictions (23%) than "high risk" non-participants (57% rearrests, 38% reconvictions). Likewise, "low risk" participants had fewer rearrests (16%) and reconvictions (0.6%) than low risk comparisons (18% rearrests, 1% reconvictions), but statistical tests of significance are not reported for these results.

Another type of transitional program is the halfway house. We located two studies of halfway houses that provided offenders with enhanced employment services. Hartmann et al. (1994) evaluated a halfway house program, which included classes aimed at teaching employment skills, job seeking and retention skills, basic life skills, as well as preparation for the GED. The study examined the recidivism rates of subjects who were successful program completers and those who were unsuccessful at completing the program. Findings indicated that halfway house employment program completers had significantly lower rearrest rates (52%) and lower felony rearrest rates (44%) than non-completers (75.3% rearrests; 68.8% felony rearrests). The lack of a separate control group, however, makes it difficult to attribute any reduction in recidivism to program effectiveness. Latessa and Travis (1991) examined the impact another halfway house program that included enhanced services and employment assistance. Halfway house participants received services such as vocational training, employment, educational, budgeting, drug, alcohol, mental health, welfare services, as well as family, group, and individual counseling. A comparison group was randomly drawn from the probation population but no information was provided about how the halfway house groups were selected. The study found that halfway house participants had fewer new crime convictions (29.5%) than the comparison probation subjects (30.7%), although this difference was not significant.

Some studies have investigated the effectiveness of enhanced employment services provided to offenders in the community. Uggen (1997) conducted a particularly rigorous evaluation of a community employment program. Subjects were randomly assigned to the program and control groups, in order to control for differences between the groups. After assignment, program participants were offered a minimum-wage, subsidized job and were assigned to work crews of 8 to 10 people led by a supervisor who acted as a counselor. The program drop-outs and completers were included in the study. One possible limitation of this research was the reliance on self-reported behavior. Other than this possible limitation, the study is strong methodologically. The results indicated that after three years, employment program participants (of all ages) had the same rearrest rate (43%) as those not participating in the program (43%). Follow-up analyses suggested that older offenders benefitted more from the program than the younger offenders. It is not clear whether this short term reduction in recidivism among older offenders would be maintained over a longer follow-up period.

In a study of enhanced employment search assistance, Clark and colleagues (1992) matched a sample of offenders in pairs who had volunteered to participate in an enhanced employment services program. One of each matched pair was assigned to the program and the other became a member of the control group. While 52.3% of the program participants dropped out of the program, the researchers did compare recidivism of the total participant group to the control group. Also, subjects were included as participants if they attended only one meeting. In addition, the enhanced employment services provided to the program group were very similar to the services provided to those in the comparison group. Findings indicated that 56.7% of the program group participants had misconduct violations compared to 66.7% of the comparison group. Similarly, the program participants averaged 1.03 misconduct violations for the six month follow-up period, while the comparisons averaged 1.23 misconduct violations. Neither of these differences were significant. Program participants did have significantly lower rates of return to prison (13.3%) compared to matched controls (33%). Similarly, program participants went a significantly more time on average before their first misconduct violation (125.2 days) than matched controls (81.8 days).

In another evaluation, Milkman (1985) randomly assigned subjects to a program and comparison groups to assess the effectiveness of a program that included job development classes, job counseling, job readiness training, orientation, needs assessment, screening and evaluation, and referral to other agencies for additional services, as needed. This program was evaluated in three cities; Boston, Chicago, and San Diego. Milkman found that participants who received program services and were placed in jobs had more rearrests for any reason, but also for serious crimes, than nonparticipants in Boston and Chicago, but not San Diego. Finally, those who received program services and were placed in jobs had more "income producing crime rearrests" in all cities than comparisons. In general, the results of the analysis do not indicate that the program was successful in reducing recidivism. Comparison of the recidivism rates themselves showed no significant results, in some cases increasing and in others decreasing with various recidivism rates.

Only one evaluation of the Work Ethic Camp has been conducted to date, the one year follow-up report prepared by the Washington State Department of Corrections. The evaluation focused mainly on a descriptive analysis of the Work Ethic Camp participants. However, an attempt was made to compare the recidivism of participants with a matched comparison group. Findings from this preliminary evaluation reveal that Work Ethic Camp (WEC) participants had a lower average number of "community custody inmate" violations per offender (2.2) than the control group (2.9). No statistical significance tests were reported. WEC participants also had a higher rate of returns to prison (25%) than controls (24%), again with no significance tests reported. Finally, 35% of both WEC participants and controls were rearrested. This research design was methodologically weak (rated at level 2 on the Maryland Scale), reducing the confidence that can be placed in its findings.

One study identified in the area of community employment rated at the bottom of the Maryland Scale for scientific rigor, meaning that no reliable conclusions could be drawn from its results (Gray and Wren, 1996).

Based upon these findings and the quality of the research, we conclude that there is sufficient evidence to say that some employment programs are effective in reducing the recidivism of offenders. The enhanced employment search program evaluated by Clark et al. (1992) did appear to reduce the number of misconduct violations and returns to prison for the participants. Similarly, the Menon et al. (1992) evaluation suggests that employment programs targeting offenders at high risk for recidivism may be successful in its reduction. Additionally, there is some evidence to suggest that work release programs may have an effect on recidivism. However, these results are not consistent in the research, making it difficult to draw definitive conclusions on this issue. Using the methodological rigor and significance tests as guidelines, we conclude that the program had the effect of reducing rearrests and return-to-jails for the participants. But, it also had the effect of increasing the returns to prison for violations of conditions of supervision.

The evidence relating to halfway house programs is more consistent -- there is no evidence that the employment interventions provided by the halfway house program evaluated by Latessa and Travis (1991) had an impact on recidivism. The Uggen (1997) study of pre-release or in the community employment services did not find an effect on recidivism. However, in the follow-up study focusing on offender age, the research suggested that programs may be more effective for older offenders. Similarly, the Milkman study (1984) failed to find convincing evidence that these programs are successful at reducing recidivism.

Overall, the research leaves us with many questions about the implementation and target population for successful community employment programs. We do believe that there is enough consistency in the findings of several methodologically-sound studies to conclude that some programs are effective in reducing recidivism.

Preliminary results from this first evaluation of the McNeil Island Work Ethic program allow for no particular conclusions to be drawn. While it appears that WEC participants had fewer supervision violations, more frequent returns to prison, and rearrest rates equal to those of offenders not participating in the program, without tests of statistical significance there is no way to assess whether these are meaningful findings. In other words, there has not been adequate research to determine whether the WEC program has an impact on recidivism.

The DOC Work Release Program (described above) was the focus of a recent evaluation by Petersilia and Turner (1996). The research was of fairly high quality and concluded that the Work Release Program was not effective in reducing recidivism. However, the recidivism rates are surprisingly low leading us to question whether the offenders eligible for program participation are already at relatively low risk for recidivism. Rehabilitation research consistently recommends that intensive rehabilitative programs should target offenders at high risk for recidivism to achieve noticeable effects. The recidivism rates for some offenders are already so low that there is little room for change. This may be what occurred in the work release study. As an illustration of this point, one study (Menon et al., 1992) separated the high and low risk participants and found significant difference only with the high risk offenders.

The Washington Work Ethic Camp has been the focus of a specific evaluation. However, the evaluation is mostly descriptive and only includes limited information on recidivism. Since there is some evidence that community programs offering services to participants may reduce recidivism, it might be helpful for the Work Ethic Camp to enhance the community phase of the program. Furthermore, since both prison industries and vocati