For a copy of the report, please email the Thompson Center – email@example.com
This is an accordion element with a series of buttons that open and close related content panels.
Cost-Benefit Analysis of the Second Chance Skills Institute
By Maddie Koolbeck, Patrick Landers, Amy Maniola, Noah Roberts, Craig Vruwink
On behalf of Governor Tommy Thompson and the Tommy G. Thompson Center on Public Leadership, we conducted a cost-benefit analysis of a proposal to convert an existing correctional facility into the Second Chance Skills Institute (SCSI). The SCSI would provide educational and other rehabilitative programming to individuals in order to improve post-release employment outcomes to fill labor gaps in Wisconsin as well as reduce rates of reincarceration. We analyze two alternatives for the SCSI and predict that each would have positive net benefits for society.
The SCSI was modeled to be located in the current Racine Youth Offender Correctional Facility (RYOCF) after a facility conversion occurs. RYOCF serves as an ideal location as its current capacity of 450 individuals offers an appropriate scale and its location is near employers and the post-release residencies of a large fraction of people in prison. The SCSI would provide correctional education and employment-based social and emotional development programming similar to programs currently offered by the Wisconsin Department of Corrections (DOC).
To identify the eligible population, a number of requirements would be set. Eligible individuals would have to be near their release date from prison and completed high school or its equivalent. Additionally, those interested would undergo a screening process to determine if their interest and aptitude make them good candidates for the program. Under Alternative 1, the initial programming would be limited to individuals who committed nonviolent crimes.
The SCSI would provide employment-based social and emotional development programming, vocational training, and earned release opportunities. The social and emotional development programming would consist of cognitive behavioral therapy (CBT) and substance use disorder therapy. Individuals with substance use disorders would begin the programming in a one-year long therapeutic community and then proceed to six months of vocational training. Individuals without substance use disorders would start immediately with vocational training and CBT that would last for six months. Program graduates would receive both a vocational training certificate recognized by industry and a rehabilitation certificate, recognizing that they successfully completed these programs. Individuals would then gain three months earned release during which they would be under community supervision. The SCSI would make use of partnerships with educational institutions and major businesses in the area to provide the training and help create employment opportunities for participants upon release. The funding for the SCSI would come from a variety of sources, including current state correctional funds, Pell Grant funds, and eventual repayment by the individuals. To cover the increased costs from this program, each program graduate would be required to pay back the equivalent of their correctional education costs.
Alternative 2 would widen the eligibility requirements to include all people who are incarcerated rather than just individuals who committed nonviolent crimes. In addition, vocational programming would be offered to 75 percent of the population while 25 percent of the individuals would be eligible for postsecondary education. The postsecondary education is expected to take two years to complete and graduates would receive an associate degree diploma. Otherwise, the programming sequence and inclusion of CBT, therapeutic communities, and earned release would all remain the same. Lastly, funding contributions from graduates of the program would not be expected in this alternative to maximize their well-being and incentives to work.
Program Costs and Benefits
To quantify the net benefits to society from the SCSI, we estimated costs and benefits for each alternative over a 50-year lifespan. Both alternatives require upfront costs to convert the facility to an appropriate configuration and ongoing costs to provide each of the relevant services to participants. The programming is expected to generate short-term, long-term, and potential benefits. Short-term benefits occur while participants are in prison or up to one year after release. These benefits include reductions in substance use and suicide, and cost savings from reduced prison time due to earned release. Long-term benefits, which occur more than one year after release, include reduced crime rates. Potential (uncertain) benefits include reductions in prison misconduct, increased earnings, and spillover benefits from increased educational levels.
Using results from the literature on similar programs and Wisconsin-specific statistics when possible, we monetize each of these costs and benefits and calculate net benefits (benefits minus costs). To account for the sensitivity of our results to uncertainty about program impacts, costs, and benefits, we perform a Monte Carlo simulation. In a Monte Carlo simulation, instead of using point estimates for parameters to calculate a single estimate of net benefits, researchers can specify ranges of likely values for variables and draw random values from these distributions. This allows researchers to create a distribution of possible values of net benefits. We report summary statistics of our findings in Table 1. For Alternative 1, we find a mean present value of $40.3 million in net benefits, with 85 percent of trials resulting in a positive value of net benefits. For Alternative 2, we find a mean present value of $65.3 million in net benefits, with over 99 percent of trials resulting in a positive value of net benefits.
Table 1: Summary of Monte Carlo Estimates of Net Benefits
|Percent of Trials with Positive Net Benefits|
By comparing the benefits and costs of individual program components, we observe that correctional education is responsible for the majority of the program’s net benefits, particularly from the predicted increased earnings. However, correctional education’s effect on earnings is under-researched and highly uncertain, so we conducted sensitivity analyses and determined that, in order for the SCSI program to generate a positive net social value, the SCSI program would need to generate permanent or multi-year increases in participant earnings. Correctional education in isolation would likely not have positive net benefits from just reductions in crime and reincarceration. The cognitive behavioral therapy program reduces suicide, misconduct, crime, and reincarceration, benefits which far outweigh the costs of CBT. The ongoing costs of providing substance use treatment are quite high and exceed its benefits, but is invaluable for expanding the program to serve people with substance use problems. Finally, the earned release component has very large net benefits.
While all of the costs would be borne by the state government under the current model, only a portion of the social benefits are also fiscal benefits to the state. The earned release benefits are large and accrue to the state, as does the majority of the crime and reincarceration costs because they are for criminal justice and prison expenses. A portion of the substance use benefits would accrue to the state government because part of this total is for medical costs that are often borne by public payers such as the prison health care system or Medicaid. Finally, the state would recoup a portion of any increases in lifetime earnings and spillover effects from increased tax revenues, reduced public safety-net expenditures, any cost-sharing provisions, and other cost savings for individuals released from prison who stay in Wisconsin.
Because both alternatives yield positive net benefits, we recommend that Wisconsin policymakers pursue implementing a SCSI model. To maximize benefits, we recommend that policymakers consider both including individuals who have committed violent crimes and offering increased amounts of earned release to successful individuals because there are large benefits associated with these policies. Finally, if implemented, we recommend that Wisconsin conduct an extensive evaluation of the SCSI in order to determine its effectiveness and whether it should be replicated elsewhere.
 One limitation of our analysis is that the estimates are drawn from evaluations of programs providing similar services, not the SCSI’s exact combination in a Wisconsin context. In addition, some experts are skeptical that the methods used in many previous evaluations produce unbiased estimates of the impacts of these interventions.
Cost-Benefit Analysis of the Second Chance Skills Institute
Prepared for Governor Tommy Thompson and the Tommy G. Thompson Center on Public Leadership
December 19, 2018
This report would not have been possible without the guidance and resources we received throughout its preparation. First, we would like to thank our clients, the Tommy G. Thompson Center on Public Leadership, Director and UW-Madison Professor, Dr. Ryan Owens, and Former Governor of Wisconsin, Tommy Thompson. We would like to recognize them for their time, insights, and vision. Additionally, we would like to thank the Wisconsin Department of Corrections, particularly Dr. Megan Jones, Dr. Silvia Jackson, and Assistant Deputy Secretary Karley Downing for taking the time to discuss this proposal and project with us. Third, we would like to thank the Wisconsin Public Defender’s Office, particularly State Public Defender Kelli Thompson, Legislative Liaison Adam Plotkin and Communications Director Randy Kraft, for their unique insights on this project. Lastly, we would like to thank Professor Dave Weimer for his guidance, feedback, and patience during the process of creating this report.
Wisconsin faces dual challenges in the increasing costs of corrections and the future availability of workers. The prison population has more than tripled over the last 25 years, eclipsing the correctional system’s functional capacity, and continues to grow at a rate of 35 individuals a month (Cornelius, 2017). Correctional facilities are overcrowded and the correctional budget reached $1.2 billion in 2017 alone (Cornelius, 2017). Simultaneously, Wisconsin’s population is poised to a hit a “workforce cliff” by 2025, under which 65-year-olds will outnumber 18-year-olds for the first time in state history (Thompson, 2018). With an already tight labor market, these changing population demographics pose the potential for large labor shortages in the future.
To address these issues, Former Governor Tommy Thompson has proposed creating a Second Chance Skills Institute (SCSI), a facility within the Wisconsin Department of Corrections focused on providing education to increase the employment odds of released individuals. The SCSI would provide the state with additional trained workers and would reduce prison costs, as studies have shown correctional education significantly decreases the odds of an individual returning to prison after being released (Bozick et al., 2018).
Additionally, rehabilitating, educating, and training individuals who are incarcerated has been shown to have numerous benefits beyond a reduction in reincarceration (Bozick et al., 2018). Research indicates that many individuals struggle to reenter their communities successfully after being released from prison. For example, people who were formerly incarcerated are at a high risk for death in the weeks following release from prison (Binswanger et al., 2007). Additionally, the formerly incarcerated particularly struggle to achieve economic self-sufficiency, with one study estimating that over 27 percent were unemployed in 2008, compared with the national average of less than 6 percent (Couloute & Kopf, 2018). This may contribute to the negative association between a history of incarceration and earnings increases through adulthood, which suggests that incarceration may have negative long-term consequences for employment, productivity, and earnings (Schanzenbach et al., 2016).
The poor outcomes of the formerly incarcerated can have significant spillover effects. For example, when formerly incarcerated parents struggle to achieve self-sufficiency, their children may receive less support, which could lead to long-term consequences such as reduced upward economic mobility (Festen et al., 2002; Western & Pettit, 2010). By making it more difficult for the formerly incarcerated to succeed in the labor market, Wisconsin faces reduced long-term economic output. If individuals struggle to achieve economic self-sufficiency and instead return to crime, then governments like Wisconsin must increase taxes in order to cover the costs of incarceration.
The SCSI provides a way to address the employment gap in Wisconsin as well as the many other negative effects of incarceration. If successful, the SCSI could serve as a model for a complete overhaul of the structure of the criminal justice system. This report assesses the proposal for a SCSI in more depth by completing a cost-benefit analysis. A cost-benefit analysis quantifies in monetary terms the consequences of the program to determine whether its benefits outweigh its costs. The report begins with a description of the programs offered through the SCSI before discussing its various costs and benefits. The report also includes a sensitivity analysis to account for uncertainty in the monetary estimates and provides recommendations concerning the SCSI.
Second Chance Skills Institute Proposal
In the following section, we detail two alternative proposals for the SCSI. The SCSI would operate as a stand-alone facility for 450 eligible individuals with the key goals of providing correctional education and complementary employment-based social and emotional development programming. The programs offered at the SCSI would be similar to current offerings at the Wisconsin Department of Corrections (DOC) but would expand services to additional individuals (see Appendix A: DOC Overview for a more thorough overview of WI DOC current programming).
The SCSI would be located in the current Racine Youth Offender Correctional Facility (RYOCF) after a facility conversion occurs (see Appendix C: Racine Youthful Offender Correctional Facility Overview, for a more thorough overview of this facility). RYOCF serves as an ideal location as its current capacity of 450 people offers an appropriate scale and its location is near employers and the post-release residencies of a large fraction of current inmates—in 2016, 47 percent of people released from DOC custody lived in Racine or one of its four bordering counties (DOC, 2017c).
To determine the eligible population, a number of requirements would be set. Eligible individuals would have to be near their release date from prison and hold a General Education Development (GED) credential or High School Equivalency Diploma (HSED). Specifically, those with substance use disorders would have to be 21 months from release and all others 9 months from release. Additionally, those interested would undergo a screening process to determine if their interest and aptitude make them good candidates for the program. The initial programming would be limited to individuals who committed nonviolent crimes. If additional criteria are needed to limit the initial population to 450 people, then it would be up to the discretion of the DOC to determine the most qualified individuals for the pilot program. Individuals in the SCSI would be subject to the same rules as the general prison population as defined by the DOC’s Code of Conduct.
The SCSI would include employment-based social and emotional development programming. The social and emotional development programming would consist of cognitive behavioral therapy (CBT) and substance use disorder therapy. The SCSI would make use of therapeutic communities for substance use disorder treatment (see Appendix D: Therapeutic Communities for more information on this program) and would make use of Cognitive Behavioral Intervention for Offenders Seeking Employment (CBI-EMP) for CBT programming (see Appendix E: Cognitive Behavioral Interventions for Offenders Seeking Employment for more information on this program). Individuals with substance use disorders, estimated to be 58 percent of the prison population, would begin the programming in a one-year long therapeutic community and then proceed directly to six months of vocational training. The therapeutic community would include CBT training, which removes the need for additional CBT once individuals graduate from the substance use treatment. Individuals without substance use disorders would start immediately with vocational training and CBI-EMP. The vocational training would last for six months, and individuals would receive CBI-EMP simultaneously for the first three months of the program. At the end of the program, graduates would receive both a vocational training certificate recognized by industry and a rehabilitation certificate, recognizing that they successfully completed these programs. (See Appendix F: Second Chance Skills Institute Program Flow and Attrition for an explanation of program flow calculations.)
The last component of the programming would be earned release for those who successfully complete all components of the program. Individuals would gain three months earned release after successful completion of all components of the program, but would still be under community supervision during this time period. Additionally, during the program, the goal would be to help each individual secure an employment opportunity before release.
The vocational programs offered would be based on the current state workforce needs as projected at the time. The SCSI would make use of partnerships with educational institutions in the area, such as UW-Parkside, Blackhawk Technical College, and Gateway Technical College, to deliver vocational training as well as with major businesses in the area to help create employment opportunities for participants upon release.
The funding for the SCSI would come from a variety of sources, including current state correctional funds, Pell Grant funds, and eventual repayment by the individuals. To cover the increased costs from this program, each program graduate would be required to pay back the equivalent of their correctional education costs. The released individuals would have a grace period of one year before they are required to begin paying the state back, would not be charged interest, and a failure to pay would not be subject to a civil judgment.
The second alternative is similar to the first alternative but differs in several ways. To begin with, the eligibility requirements for Alternative 2 would be more inclusive. Most notably, Alternative 2 would widen the eligibility requirements to include all people who are incarcerated rather than just individuals who committed nonviolent crimes, as theory suggests that targeting individuals with higher risk profiles would lead to larger benefits (Andrews & Bonta, 1994; Gendreau, 1996; Latessa, 2011). The inclusion of postsecondary education, defined as an associate degree, would also widen the eligible population to include people 39 months and 27 months from release for those with and without substance use disorders, respectively.
Under Alternative 2, vocational programming would be offered to 75 percent of the population while 25 percent of the individuals would be eligible for postsecondary education. The postsecondary education is expected to take two years to complete and graduates would receive an associate degree diploma. Otherwise, the programming sequence and inclusion of CBI:EMP, therapeutic communities, and earned release would all remain the same. (See Appendix F: Second Chance Skills Institute Program Flow and Attrition for an explanation of program flow calculations.)
Lastly, funding contributions from graduates of the program would not be expected in this alternative. Literature suggests that upon release, individuals already experience high levels of fiscal stress that can be detrimental in efforts to reduce crime and reincarceration (Bannon et al., 2010; DeVuono-Powell et al., 2015; Shafroth et al., 2016). Thus, payback would not be required in this alternative to eliminate the negative impacts from the imposition of additional costs upon release. (To compare the two alternatives, refer to Appendix B: Alternatives Description Chart.)
Program Costs and Benefits
To predict an expected net benefit from the SCSI, we estimate upfront costs, ongoing costs for the programming, and short-term, long-term, potential, and unquantifiable benefits from each of the components of the program. Short-term benefits occur either during the program or within the first year of release, long-term benefits occur within the three years after release, and potential benefits are relatively uncertain benefits. As the flow of SCSI program participants changes over the first four years, we calculate an estimate for the benefits and ongoing costs for the each of first four years and then for the following steady-state, the time at which program enrollment becomes constant for the remaining years in the 50-year lifetime of the program. For simplicity, we report the steady-state estimates in this section. (For more detailed calculations, see Appendices G-R.)
Facility Conversion and Expansion
The SCSI would make extensive use of various correctional programs, from career and technical education to substance use disorder treatment. The broad scope and intensive nature of the programming offered at the SCSI requires an expansion of RYOCF to convert the site to an appropriate configuration and effectively provide these services. The DOC’s 10-year development plan, produced in 2009, provides a general blueprint for locations of potential additions and cost estimates. We use the monetary value estimate, $3 million in 2018 dollars, from the DOC report as the cost to expand the current facility to include additional training spaces (Mead & Hunt, 2009). (See Appendix G: Facility Conversion & Expansion Costs for more details on this calculation.)
Cost of Initial Staff Training in CBI-EMP
All staff at the SCSI would be trained to deploy the CBI-EMP curriculum, an employment- focused cognitive behavioral therapy. While there would be staff designated as facilitators of the group CBI-EMP instruction, CBI-EMP strategies can be utilized in daily interactions across all service types and programming circumstances. Broad staff training can help build a cohesive, system-wide understanding and approach that strengthens the re-entry process (Duran, 2013).
Because of the employment focus of the SCSI, we recommend that all staff receive this training prior to the facility transition. Combining the cost of training and the cost of staff time for the days of training, we estimate the training cost per person to be $1,545. Training the estimated 138 staff members at the facility prior to the commencement of the programming is projected to constitute a one-time cost of $213,000 (See Appendix H: Cost of Initial Staff Training in CBI-EMP).
The following ongoing costs predict the cost to deliver the programs each year. The ongoing costs for correctional education and therapeutic communities rely on a cost per participant and the number of the participants per year. For these programs, Alternative 2 has lower annual costs as the facility would have a smaller number of new entrants, and therefore participants, each year. The smaller number of entrants is because some existing participants would be spending more time in the program participating in postsecondary education, which is only offered in Alternative 2.
Costs of CBI-EMP Recurring Training and Facilitator
As the SCSI experiences staff turnover, there would be a need for ongoing staff training in CBI-EMP. After analyzing multiple research-based estimates, we assume that the SCSI would experience an annual turnover rate of approximately 20 percent. When applied to the number of staff members at the SCSI, this turnover rate projects that 27.5 individuals would be in need of training each year. This translates into $42,000 of recurring annual staff training expenses. This estimate accounts for both the cost of training and the cost to pay staff for the days of training.
Additionally, although correctional facilities already maintain staff to facilitate specialized programming, in order to deliver the intensive CBI-EMP programming offered at the SCSI effectively, our analysis suggests that the SCSI should plan to incorporate a full-time CBI-EMP facilitator. After an assessment of various salaries paid to correctional positions similar to a CBI- EMP facilitator, such as a vocational education teacher or psychological assistant, we estimate total compensation for a full-time CBI-EMP facilitator to be approximately $70,000. This would be a recurring cost that is maintained for the duration of the Second Chance Skills Institute’s life-cycle. These ongoing costs would be the same for both alternatives. (See Appendix I: Cost of Ongoing CBI-EMP Staff Training & Additional CBI-EMP Facilitator for additional details).
To calculate the costs of offering the substance use disorder treatment program, therapeutic communities (TC), we use the marginal annual cost estimate of $2,198 per participant from the Washington State Institute for Public Policy (WSIPP) (WSIPP, 2016c). We calculate our total annual cost by multiplying our cost estimate of $2,198 by the number of TC participants each year. We calculate a recurring annual cost of $630,000 and $500,000 in the steady state for Alternative 1 and 2, respectively. (See Appendix J: Cost of Therapeutic Communities for more details.)
Under Alternative 1, all individuals in the SCSI would take part in a vocational education and training program. The cost of correctional vocational education fluctuates depending on program, location, and partnership opportunities. However, the DOC’s budget reports provide us with a Racine-specific annual cost estimate of $2,700 per enrollee for vocational education (DOC, 2016a). We calculate our total annual cost by multiplying our cost estimate of $2,700 by the number of correctional education participants each year. In Alternative 1, we reach a recurring annual cost of $1.1 million. In Alternative 2, we reach a recurring annual cost of $690,000. (See Appendix K: Cost of Vocational Training for more details on these calculations.)
Our second alternative would provide a select population of individuals in the SCSI the option to participate in postsecondary educational programming. We utilize a WSIPP cost estimate of $1,250 and adjust the cost upward using a WSIPP-DOC cost-differential ratio in order to develop a more Wisconsin-specific cost estimate (WSIPP, 2016b). This adjustment established our marginal annual cost estimate of just over $2,300 per enrollee. We calculated our total annual cost by multiplying our cost estimate by the number of program enrollees each year, yielding a recurring annual cost of $192,000 in the steady state. (See Appendix L: Cost of Postsecondary Education for more details related to postsecondary education programming costs.)
Short Term Benefits
Reduced Suicide in Prison
CBT programming reduces suicide rates for participants in the short term (Brown et al., 2005; Tarrier et al., 2008). This results in significant cost savings due to reduced suicides while in prison. To calculate this benefit, we multiply the percentage point reduction in suicide rate by the number of SCSI participants who complete CBT each year. We use an effect size from Tarrier et al.(2008) to calculate a 0.012 percentage point decrease from the current baseline of 0.00019, or 19 suicides per 100,000 people in Wisconsin state prisons (Noonan, 2016). We multiply this percentage point decrease by the number of SCSI participants who complete CBT each year, 380 people in the steady-state.
To calculate the monetary value of this benefit per participant, we apply the reduction in suicide to an adjusted estimate for the general population of a Value of Statistical Life (VSL) of approximately $11.7 million, which reflects people’s willingness to pay to avoid mortality risk. Because VSL incorporates willingness to pay, some studies adjust VSL when accounting for lower income populations, such as people who are in prison (Hammitt & Robinson, 2011; Robinson & Hammitt, 2015). To be conservative in our estimates, we adjust VSL using the average cost of incarcerating one person for one year as a proxy for the income of people in prison, because people generally earn little to no income while in prison. This adjustment results in a value of $2.66 million, which we vary in our sensitivity analysis. Our final calculations indicate a benefit of $124,000 per year for Alternative 1 and $99,000 per year for Alternative 2. (For more details on the calculations, see Appendix M: Benefit of Reduction in Suicide Rates.)
Earned release, offered for any successful graduate of the program, would generate significant cost savings from the elimination of the need to house individuals. To calculate cost savings, we use an estimate of the cost savings per individual from earned release multiplied by the number of graduates each year. Using a Wisconsin Legislative Fiscal Bureau (LFB) report (n.d.), we calculate the costs savings per individual, mostly from the reduced need of a bed, to be $2,100 to $3,000 for the 3 months of earned release granted. To calculate the benefit, we multiply this amount by the number of graduates each year, which we expect to change before reaching a steady state. We estimate the benefit in the steady state to be $1 million under Alternative 1 and a little over $810,000 under Alternative 2. (For more detailed descriptions of these calculations, refer to Appendix N: Benefits from Early Release.)
Reduced Substance Use
We expect that therapeutic communities would reduce the rates of substance misuse among people released from prison (Pearson & Lipton, 1999; WSIPP 2017c; Mitchell et al., 2018), which would provide direct social benefits. Thus, to calculate this benefit, we estimate the percentage point difference between the baseline rate of substance use and the rate in the first year after completing treatment, which we multiply by the costs of substance use and number of therapeutic community completers per year.
We rely on Mitchell et al.’s (2018) reported impact of therapeutic communities on substance use and a baseline rate of substance use post-release of .43 (Malouf et al., 2013) to calculate a nearly 4.2 percentage point difference. To monetize the benefit, we multiply this difference by the number of TC completers per year and the estimated costs of substance abuse in 2018 dollars, $4,950. The cost of substance abuse is composed of health costs, estimated at $1,550 (Ettner et al., 2006), and the opportunity cost of manufacture and use, $3,400 (Cohen & Piquero, 2009; Kilmer et al., 2014). The estimated annual benefit once the SCSI reaches a steady state is $41,000 and $33,000 under Alternative 1 and Alternative 2, respectively. (For a detailed description of these calculations, see Appendix O: Benefits from Reduced Substance Use.)
Long Term Benefits
Reduced Crime and Reincarceration Rates
The SCSI services of correctional education, cognitive behavioral therapy, and therapeutic communities are expected to reduce crime and reincarceration (Bozick et al., 2018; WSIPP, 2016a, 2016b, 2016c, 2016d). Reduced crime and reincarceration would result in social benefits, by reducing the cost of crime, including the costs to victims of crime and to government for criminal justice system costs. To calculate the total benefits, we multiply the estimated reduction in crimes from participating in SCSI by the cost of crime and the number of graduates per year.
We use estimates from WSIPP for how much each service component would reduce crime and reincarceration, relative to the state’s three-year reincarceration rate of 37.5 percent (WI DOC, 2018c). We estimate that vocational correctional education would reduce crime and reincarceration by nearly 4.6 percentage points, vocational postsecondary education by 6.2 percentage points, cognitive behavioral therapy by 3 percentage points, and therapeutic communities by 2.4 percentage points. While SCSI participants in Alternative 1 only participate in vocational education, for Alternative 2 we take an average of the vocational and postsecondary education impacts, weighted by the numbers of graduates who complete each type of education. To account for potential double- counting, in which individuals who would be helped by one component are the same individuals who would benefit from another component, we assume that therapeutic communities and cognitive behavioral therapy have overlapping effects. As a result, we estimate the SCSI program impacts as only the sum of correctional education and cognitive behavioral therapy impacts.
For the benefits of reduced crime and reincarceration, we draw on estimates of the cost of crime that include victim costs and the costs of the criminal justice system (McCollister et al., 2010; Blincoe et al., 2015; McCollister et al., 2017). To find the average cost of nonviolent crimes committed by individuals included in Alternative 1 and average cost of all types of crimes committed by individuals eligible under Alternative 2, we weight offense-specific costs of crime by the prevalence of those offenses as a share of total arrests in Wisconsin (FBI, 2016).
We use our estimates of the program impacts on reincarceration and costs of crime, which vary by number of graduates, which in turn vary by alternative and year. The estimated annual benefit once the SCSI reaches a steady state is $92,000 and $427,000 under Alternative 1 and Alternative 2, respectively. (For a more detailed description of these calculations, see Appendix P: Benefits from Reduced Crime and Reincarceration Rates.)
The following three sections discuss potential benefits from the SCSI. Estimation of each of these benefits involves a high degree of uncertainty, which is why we classify them as potential benefits. To account for this uncertainty, we use the reported estimates below as an upper bound and vary the benefits from zero to the upper bound estimate in our Monte Carlo simulation.
It is expected that CBI-EMP would lower rates of misconduct within prison (Morgan & Flora, 2002), which we model through reductions in the number of assaults within the SCSI. We estimate both the reduction in assaults against people who are incarcerated and assaults against staff members. To estimate the benefit, we find a percentage point difference between baseline rates of assault and rates of assault in the SCSI, which we multiply by the number of CBT completers each year and the cost of an assault.
For those who are incarcerated, we estimate a percentage point decline of 9.3 from a baseline of 22.3 percent of individuals in prison experiencing an assault in the last six months of incarceration (Wolff et al., 2013). To monetize this benefit, we find a weighted average of the cost of assault of $825, which is based on the medical costs from assault (Miller et al., 1996) and the percent of assaults that result in needed medical attention (Wolff et al., 2013). For staff members, we estimate a percentage point decline of .65 for a completed assault and a percentage point decline of .11 for assaults that result in injury, from baselines of 1.2 percent of individuals completing assaults against staff members and 0.2 percent of individuals completing an assault that injures a staff member. For the cost of assault for staff, we include estimates of lost productivity from assaults that result in injury and quality of life adjustments and medical costs for both types of assault. This results in cost estimates of $3,800 for assaults that do not result in an injury and $36,000 for assaults with injuries.
By adding the impact on staff and individuals who are incarcerated together, we yield yearly monetary benefits, under our steady state, of $54,000 and $43,000 under Alternative 1 and Alternative 2, respectively. (See Appendix Q: Benefits from Reduced Misconduct for moredetails.)
Increased Lifetime Compensation
Human capital theory states that education increases the economic capabilities and productivity of individuals (Schultz, 1971). Therefore, we assume the additional education attained by participants of the SCSI would increase their productivity, which we measure through increases in their earnings capacity for their remaining work career. To estimate the benefits from the SCSI on earnings, we use WSIPP’s estimated return on earnings from an additional year of education multiplied by the baseline expected lifetime compensation for individuals released from prison with a high school diploma. WSIPP’s estimated return is 10 percent for an additional year of education, which we divide by two for the six-month vocational education program and multiply by two for the two-year postsecondary program (WSIPP, 2017).
To estimate the baseline lifetime compensation for people who are released from prison and hold a high school degree or equivalent, we use the average income for formerly incarcerated individuals, $7,600 (Looney & Turner, 2018), an average age of 36 at release (DOC, 2017f) and an estimated curve on how earnings change over a person’s lifetime (Boardman et al., 2018). We estimate a post-release lifetime compensation baseline of $210,000 for a formerly incarcerated person with a high school diploma or equivalent. Finally, using the WSIPP return on education multiplied by this baseline compensation, we calculate that SCSI participants who receive vocational education would earn an additional $10,500 over the course of their lifetime and a SCSI participant who receives postsecondary would earn an additional $41,000 over the course of their lifetime.
We calculate an annual benefit of $3.7 million and $4.8 million under Alternative 1 and Alternative 2, respectively, in the steady state. Although these estimates rely on the average return on education for the overall population, we use these point estimates as an upper bound as we assume that the SCSI graduates would experience a lower return compared to the generalpopulation. (See Appendix Q: Benefits from Increased Lifetime Compensation for additional details.)
Nonmarket Spillovers from Increased Compensation
Increased earnings create external spillover benefits for society, including improved consumption decisions, health, and child development (Haveman & Wolfe, 1984; Haveman & Wolfe, 2002). To calculate this benefit, we use the increase in compensation expected for graduates of the SCSI multiplied by a WSIPP estimate on the external benefit return from increased compensation (calculated in Appendix Q: Benefits from Increased Lifetime Compensation). WSIPP estimates spillover benefits as a fraction of total compensation and reports a modal value for this return of .37 (WSIPP, 2017). We estimate the annual benefits to be $1.37 million and $1.8 million once we reach a steady state of participants under Alternative 1 and Alternative 2, respectively. (See Appendix R: Nonmarket Spillovers from Increased Compensation for additional calculations.)
The literature suggests that rehabilitation certificates may reduce the negative outcomes associated with a criminal record (Doleac, 2018). Specifically, research in Ohio found that rehabilitation certificates improve the likelihood of formerly incarcerated individuals receiving a job interview (Leasure & Anderson, 2016) and landlords accepting housing applications of people with a criminal record (Leasure & Martin, 2017). These two studies suggest that rehabilitation certificates may increase the employment rates of people who graduated from the SCSI as well as their ability to find housing. The two studies relied on audits in one state and only reported likelihood of a positive response. Thus, due to the early stage of this research, small sample sizes, and lack of effect size estimations, we did not monetize these benefits.
We first calculate the present value of the SCSI program, modeling the sensitivity of that result using a Monte Carlo simulation, and find that the program consistently has a positive net social value. We also analyze how different categories of costs and benefits contribute to our results, and identify what categories have implications for government budgets. The increased compensation and related nonmarket spillover benefits are particularly important, but highly uncertain. As a result, we also conduct additional sensitivity analyses in which we model different potential earnings impacts.
Monte Carlo Simulation
To account for the sensitivity of our results to parameter uncertainty, we perform a Monte Carlo simulation. In a Monte Carlo simulation, instead of using single point estimates for parameters to calculate one value of net benefits, researchers can specify distributions and ranges of likely values for variables and draw random values from these distributions. This allows researchers to account for uncertainty in the parameters used to estimate net benefits and create a distribution of possible values of net benefits. We wrote a computer program that completes 10,000 trials by randomly drawing from these distributions. This results in a distribution of 10,000 possible values of net benefits, from which we report summary statistics as shown in Table 1. (For more detailed information about the random variables and code used in the Monte Carlo simulation, see Appendix U: Variable Distributions for Monte Carlo Simulation, and Appendix V: Monte Carlo Stata Code.)
Table 1: Summary of Monte Carlo Estimates of Net Benefits
|Minimum ($ millions)||Maximum ($ millions)||Percent of Trials with Positive Net Benefits|
For Alternative 1, we find a mean present value of $40.3 million in net benefits, with 85 percent of trials resulting in a positive value of net benefits.
For Alternative 2, we find a mean present value of $65.3 million in net benefits, with over 99 percent of trials resulting in a positive value of net benefits.
Benefit and Cost Components
Next, we separate categories of costs and benefits of the SCSI over a 50-year lifetime using the mean values from our Monte Carlo simulation. Table 2: Present Value of Benefits over 50 Year Lifetime shows the present value of total benefits by category, and Table 3: Present Value of Costs Over 50 Year Lifetime show the present value of total costs by category. These costs and benefits allow us to see which categories of costs and benefits are driving our results for each alternative, compare the benefits and costs of individual program components, and discuss which benefits and costs are likely borne by the state of Wisconsin.
Table 2: Present Value of Benefits Over 50 Year Lifetime
|Alternative 1 ($ millions)||Alternative 2 ($ millions)|
|Reduced Substance Use||0.9||0.8|
|Reduced Crime and Reincarceration||2.2||10.0|
|Increased Lifetime Compensation||43.9||56.2|
|Spillovers from Increased Lifetime Compensation||12.1||15.3|
Table 3: Present Value of Costs Over 50 Year Lifetime
|Alternative 1 ($ millions)||Alternative 2 ($ millions)|
- Upfront costs occur at the beginning of the period and are not
- Ongoing costs were calculated using a 3.5 percent discount rate and mid-year
Correctional education is responsible for the majority of the crime and reincarceration benefits and all of the lifetime compensation and resulting spillover effects, which outweigh the upfront and ongoing costs of providing vocational and postsecondary education programming. However, correctional education’s effect on earnings is under-researched, so the compensation and spillover benefits are highly uncertain. Without the lifetime compensation increases and spillover effects we predict for the SCSI individuals, correctional education in isolation would likely not have positive net benefits from just reductions in crime and reincarceration.
The cognitive behavioral therapy program reduces suicide and misconduct, plus contributes to the crime and reincarceration benefits (see Appendix P for more details on how the crime and reincarceration benefits are calculated). These benefits far outweigh the costs of CBT. Therapeutic community treatment is expected to reduce substance use and, along with CBT, contribute to a minority of the recidivism benefits, but the ongoing costs of providing that treatment are quite high. However, TC may be invaluable for expanding the SCSI program to serve people who first need their substance use needs treated before other criminogenic needs can be addressed by correctional education and CBT. Finally, the earned release component has very large benefits.
While all of the costs would be borne by the state government under the current model, only a portion of the social benefits are also fiscal benefits to the state. The earned release benefits are large and accrue to the state, as does the majority of the crime and reincarceration costs becausethey are for criminal justice and prison expenses. A portion of the substance use benefits would accrue to the state government because part of this total is for medical costs that are often borne by public payers such as the prison health care system or Medicaid. Finally, the state would recoup a portion of any increases in lifetime compensation and spillover effects from increased tax revenues, reduced public safety-net expenditures, any cost-sharing provisions, and other cost savings for individuals released from prison who stay in Wisconsin.
Earnings Sensitivity Analysis
The lifetime compensation and spillover benefits are so large that the SCSI would break even by year three under either Alternative 1 or Alternative 2, based on our Monte Carlo simulation.
However, as previously noted, there is a high degree of uncertainty surrounding the benefits of increased lifetime compensation and spillovers from increased lifetime compensation. As shown in Table 2, these benefits are larger than all other benefit categories. Therefore, we present several sensitivity analyses, calculating net benefits under different earnings scenarios, in Appendix W: Earnings Sensitivity Analysis. The spillover effects are a percentage of the compensation benefits, so lower compensation effects simultaneously result in lower spillover benefits. On one hand, these alternative scenarios suggest that, in order for the SCSI program to generate a positive net social value, the SCSI program would need to generate permanent or at least multi-year increases in compensation, either through more individuals being employed, working longer hours, or earning higher hourly wages and compensation. On the other hand, these persistent increases in compensation may not need to be that large in order to quickly add up to a substantial amount.
In addition to the uncertainty detailed in our discussion of the Monte Carlo simulation, there are several additional limitations to our analysis. First, a number of our estimates, such as baseline rates for substance use or the average income of people released from prison, come from national or WSIPP estimates rather than Wisconsin-specific data. The inability to rely solely on Wisconsin- specific numbers limits our capacity to precisely predict the impact of our alternatives in Wisconsin.
A second limitation is that our analysis fails to account for how the cost-sharing requirement in Alternative 1 would affect our benefit estimates. Requiring individuals to pay money back to the state as reimbursement for participating in SCSI would be equivalent to placing a tax or debt obligation on individual’s future earnings, therefore lowering their returns from working. Research finds that income taxes reduce labor supply, and that child support debt is also associated with reduced formal employment (CBO, 2012; Miller & Mincy, 2012). In addition to likely reducing labor supply and lifetime earnings, the cost-sharing policy would mean that a portion of individuals’ earnings would be transferred to the state. These reductions and transfers of individuals’ compensation would also lower nonmarket spillover benefits because less income would end up in the hands of the released individuals and their families. Finally, given the hypothesized link between employment and recidivism, reduced labor supply might lead to smaller recidivism reductions. In sum, a cost-sharing requirement seems likely to reduce SCSI benefits in the areas of lifetime compensation, nonmarket spillovers, and recidivism, relative to not having a cost-sharing requirement. Although we can predict the estimated direction of the effect, modeling the size of this effect on present net benefits would require specifying the cost-sharing policy in greater detail than is possible in this analysis.
Another limitation regards the potential spillover effects from removing select individuals from the general prison population and placing them at RYOCF. The population in the SCSI would be made up of individuals who took the initiative to express interest in this program and who passed all the eligibility screening requirements. For that reason, this population would likely be different from the overall prison population and the separation of these individuals might affect outcomes in other prisons, most notably behavioral outcomes. For example, the creation of the SCSI might concentrate “bad apples” in other prisons and increase rates of misconduct there, or the opportunity to participate in this program might motivate individuals in other prisons to behave better in order to have a chance to participate in the SCSI. We have not found any other state with a program similar to the proposed SCSI in terms of concentrating services for everyone in a facility, so there is no research or program experiences with what happens in other prisons. As a result, this analysis was unable to assess how this separation might affect statewide outcomes.
The Monte Carlo simulation also did not model variation in attrition rates, but rather used fixed population projections derived from our life table analysis. Although we reviewed the literature and WI DOC’s existing programs extensively to determine the program attrition rates, some uncertainty around these estimates exists. In a more extensive study, it would be desirable to also vary the attrition rates as part of the Monte Carlo analysis.
Lastly, a number of the studies used to estimate effect sizes relied on propensity-score and other matching methods. Although these methods control for observable differences in characteristics between program and comparison group members, many experts are skeptical about the effectiveness of these designs in controlling for selection bias (Doleac, 2018; King, in press).
Selection bias may inflate the impacts from the intervention, which would mean that the effect sizes used in our analysis overestimate the impact from the services provided by SCSI.
We recommend that policymakers implement the SCSI model because we project that the program would generate positive social benefits. Alternative 2 yields larger net social benefits, suggesting that it would be a superior option to Alternative 1. However, both options yield positive net benefits, and the majority of the difference between the two is from highly uncertain earnings benefits. Policymakers may wish to consider alternative factors, such as political feasibility or implementation concerns, when comparing these alternatives.
Existing research suggests that interventions should be targeted to individuals with a higher risk of reincarceration and at risk of committing more costly crimes. As a result, this cost-benefit analysis finds that including individuals who have committed violent offenses in the SCSI program also has significant benefits, which contributes to why Alterative 2 yields larger social benefits than Alternative 1. Policymakers should consider these benefits when deciding whether to provide services to violent offenders who are scheduled to be released.
Similarly, the benefits from earned release opportunities are quite large and relativelycertain. As this component creates significant cost savings for the state, policymakers may wish to consider providing high-achieving participants such as those who successfully complete the SCSI model with the opportunity to access even more months of earned release. Earned release opportunities may also help incentivize participants to persist in the SCSI’s programming, and incentivize individuals in other prisoners to behave better in order to increase their likelihood of being able to enroll in SCSI. These types of behavioral responses from the offer of increased earned release would further increase the net benefits of theprogram.
While this cost-benefit analysis predicts that the SCSI model and specific program elements have sizable social benefits, the research evidence underlying these predictions is of moderate quality, the SCSI program represents a unique combination of components, and the quality of the program’s implementation in the Wisconsin context would affect the program’s effects. As a result, the state should conduct impact, outcomes, ex post cost-benefit, and implementation evaluations of the Second Chance Skills Institute. Each of these evaluations would provide the state with valuable information that can help ensure that state resources are being effectively spent. The impact evaluation would be important for providing rigorous evidence on the causal effect of the SCSI’s unique combination of services.
The outcomes evaluation would align with the state’s traditional approach to performance management reporting and capture key data on participants and service components. An ex post cost-benefit analysis would be able to use the effects from the impact evaluation, participant data from the outcome evaluation, and Wisconsin-specific cost data to assess whether the program has positive net social and fiscal benefits in practice. Finally, the implementation evaluation could use qualitative and quantitative data to identify lessons from the SCSI pilot that could be applied if the state tried to bring the program to scale, or if other states wished to adapt the model to their incarceration systems.
In view of the likelihood that there would be more individuals eligible to participate than SCSI has the capacity to serve, a lottery or finely-detailed point system could be used to determine who gets to participate and enable random assignment or regression-discontinuity impact designs, respectively (Shadish et al., 2002). Those designs are more likely to produce unbiased causal estimates than the propensity-score matching designs researchers have primarily used to date to evaluate prison rehabilitation program such as correctional education.
A.) DOC Overview
B.) Alternatives Description Chart,
C.) Racine Youthful Offender Correctional Facility Overview
Appendix A – DOC Overview
The DOC operates 37 prison facilities across the state, housing 23,519 people, 93 percent of which are male, at a total cost of $1.2 billion (Cornelius, 2017). The population is 48 percent first- time offenders and 52 percent second-time offenders (DOC, 2018d). Of the male population, around 30 percent do not hold a high school diploma or equivalent and an additional 47 percent only have a high school equivalency degree or GED (DOC, 2018d). DOC currently offers a variety of educational and socio-emotional programs for people in prison, which are discussed below.
Of the $1.2 billion WI DOC budget, only a little over $20 million is allocated towards educational programs (Haynes, 2018). The top educational priority for people who are incarcerated are high school equivalency diplomas or GED programs, which are offered at all prisons except one. These programs typically focus on basic academic skills, such as reading, writing, and mathematics. The Department of Corrections also provides career and technical training, also known as vocational education. Due to the demand for skilled workers, the Department of Corrections, in conjunction with the Department of Workforce Development, has been able to offer vocational training in prisons through partnerships with technical colleges and private firms. Individuals who are incarcerated have the opportunity to participate in training in over 24 occupational areas. As examples, the Racine Correctional Institution partners with Gateway Technical College to provide individuals with CNC mobile lab training, the Oregon Correctional Center partners with Madison College to provide training in industrial maintenance, and the Powers Correctional Center works with Northeast Wisconsin Technical College to provide machine operation training. The DOC also notes that after an individual completes the training, they would be linked with a staff member who would help the participant make connections with potential employers and track their success post-graduation (DOC, 2016a). Only 1,485 inmates were enrolled in CTE programs as of FY 2014, a fraction of the state’s incarcerated population (WI DOC, 2016b).
Lastly, the DOC currently offers individuals who are incarcerated the opportunity to take courses for college credit through various University of Wisconsin system campuses. The expenses for these courses are fully covered by the individual.
Cognitive Behavioral and Substance Use Disorder Programming
The DOC currently provides cognitive behavioral therapy (CBT) to a number of individuals residing in prison. CBT is a widely used therapeutic practice that helps individuals identify and change thinking processes that lead to maladaptive behaviors and negative outcomes (What Works for Health, 2018). It can be provided through both individual and group therapy and by either trained facilitators or licensed professionals. In a prison setting, the programs emphasize personal accountability, help reduce individuals’ anger and alleviate their criminogenic thinking patterns, and teach alternative thinking patterns and behaviors.
The DOC currently provides multiple CBT programs, including Cognitive Intervention Program (CGIP), Thinking for a Change, Anger Management (AM), and more general anger management/impulse control programming (DOC, 2017d). The DOC also uses CBT to target specific behaviors, including substance abuse, domestic violence, and employment skills (DOC, 2018c). These programs typically last about 15 weeks and target individuals who are deemed to be at a moderate to high-risk of offending once they are eventually released from prison (Tatar,2015).
The programs often incorporate additional elements, such as cognitive and interpersonal skills training (e.g., anger management, communication, social). The length of sessions (e.g., 90 minutes to two hours), frequency (e.g., one or more times a week), and number of participants (e.g., eight to 15) varies by program. The DOC’s programming is funded through the Becky Young Community Corrections Recidivism Reduction Appropriation; approximately $1,087,000 was allocated for these programs during fiscal year 2017 (What Works for Health, 2018). In 2018, the CBT programs in Wisconsin served 3,431 people in prison (DOC, 2018c).
The WI DOC also currently offers five forms of substance use disorder programming for persons who are incarcerated: Alcoholics Anonymous (AA), Narcotics Anonymous (NA), Self- Management and Recovery Training (SMART), Dual Diagnosis, and the Drug Abuse Correctional Center (DACC) in Winnebago County. In 2016, the Department of Corrections estimated that nearly 400 individuals received direct substance use treatment (DOC, 2016b).
Appendix B: Alternatives Description Chart
Table B.1: Comparison of Alternatives
|Alternative 1||Alternative 2|
Racine Youthful Offender Correctional Facility (convert existing facility)
Racine Youthful Offender Correctional Facility (convert existing facility)
● Limited to non-violent offenders
● Additional screening process by DOC staff (based on current criteria from DOC*)
● Limit program to 450 individuals
● All individuals eligible for programming
● Additional screening process by DOC staff (based on current criteria from DOC*)
● Limit program to 450 individuals
● Social/emotional development (includes substance use treatment and cognitive-behavioral therapy)
● Vocational training sponsored by specific employers, provided by local colleges
● Rehabilitation and vocational trainingcertifications
● 3 month early release
● Social/emotional development (includes substance use treatment and cognitive-behavioral therapy)
● Vocational training (75 percent of population) sponsored by specific employers, provided by local colleges
● Postsecondary education (associate’s degree) options (25 percent of population)
● Rehabilitation, vocational training, and associate degree diplomas
● 3 month early release
● DOC Code of Conduct
● Cost-sharing: participants contribute equivalent of correctional education cost
● DOC Code of Conduct
*The Wisconsin Department of Corrections uses the Correctional Offender Management Profiling for Alternative Sanctions tool to assess risk of reoffending (see https://doc.wi.gov/Pages/AboutDOC/COMPAS.aspxfor more information).
Appendix C: Racine Youthful Offender Correctional Facility Overview
The Racine Youthful Offender Correctional Facility is located in Southeastern Wisconsin (see Figure C.1 and C.2). It was originally designed as a 450-bed medium-security facility and, when it opened in 1988, it was Wisconsin’s first urban prison facility. The Racine Youthful Offender Facility has two housing units containing 120 cells each and contains a segregation unit with 27 beds. The facility currently houses males between the ages of 15 and 24 who have been tried in adult court.
The facility also maintains a visiting room, a multi-purpose room that also serves as a chapel, a health services area, meeting rooms, staff offices, an education area containing six classrooms, a recreation field, a gymnasium, and a support facility that contains food service and maintenance prison industries.
There are multiple features that make the Racine Youthful Offender Correctional Facility (RYOCF) a quality site to house the Second Chance Skills Institute (SCSI). The dual housing facilities allow for the SCSI to operate both the cognitive behavioral therapy (CBI-EMP) programming and the substance use disorder treatment (therapeutic community) with fidelity to the program design and functionality. Specifically, as therapeutic community programming recommends using a separate housing unit to facilitate the supportive “community” aspect of the rehabilitative program, the existence of two facilities allows for the recommended delivery of this service.
Furthermore, the location of RYOCF would allow for easy facilitation of private and public partnerships to help the SCSI deliver these programs, reduce costs, and provide employment, education, and training opportunities to program participants.
D.) Therapeutic Communities
E.) Cognitive Behavioral Interventions for Offenders Seeking Employment
F.) Second Chance Skills Institute Participant Flow and Attrition
Appendix D: Therapeutic Communities
Therapeutic communities are a group-based program designed to address mental illness, personality disorders, and substance use disorders. Through cognitive behavioral therapy and interventions, as well as a broader focus on positive lifestyle changes, this program seeks to guide individuals to recovery while also teaching participants how to proactively guard against relapse in the future. The overall goal of therapeutic communities is treat the entire person through the use and interaction of a peer group. This form of treatment typically is cited as the most effective in reducing rates of substance (Inciardi et al., 2004).
In prison, therapeutic community participants live in a separate housing unit to better facilitate the community support aspects of the treatment. Treatment usually progresses through a series of phases, each of which increase responsibility for an individual. Typical activities are morning meetings, individual and group therapy, life skills groups, and participation in psychotherapy with well-specified roles, privileges, and responsibilities. The first phase, lasting between one and three months, revolves around orienting the participant to TC philosophy and concepts, making an initial diagnosis and allowing them to assimilate into the process. The second phase, lasting between three and seven months, increases an individual’s responsibility and involvement by allowing them to teach new members and assist in the day to day operation. The final stage, lasting between one to three months, focuses on strengthening planning and decision- making skills. Individuals develop a relapse prevention plan and design an individual release plan, both of which are signed off on by the person who is incarcerated and the treatment specialist. After completion of the one-year period, there is typically an aftercare component provided.
Appendix E: Cognitive Behavioral Interventions for Offenders Seeking Employment
Cognitive Behavioral Interventions for Offenders Seeking Employment (CBI-EMP), designed by University of Cincinnati Corrections Institute, combines cognitive behavioral interventions with traditional employment approaches (DOC, 2017f). Participants engage in activities that aim to build cognitive, social, and emotional skills tailored to the work environment. The main goal of the program is to teach individuals how to secure employment, maintain employment, and manage workplace situations. The CBI-EMP program is a 31-session curriculum consisting of five main modules.
Table E.1: CBI-EMP Programming Modules
|Motivational Enhancement: Getting Them Ready for Work||Introduce CBI-EMP Clarify values
|Cognitive Restructuring: Thinking Right about Work||Behavior is a Choice
Recognize, Change, Replace Risky Thinking
|Social Skills/Emotional Regulation Skills: Skills for Work||Controlling Emotions
Learning and Using Self-Control
Dealing with Anger, Rejection, Accusation Asking Permission
Giving Feedback Answering a Complaint
Working through Challenges at Work
|Introduction to Problem-Solving Identifying Your Problem and Goal Brainstorming Options
Planning and Trying Your Solution
Being Successful at Work
|Developing a Plan Getting to the Source Reinventing my Life Staying on Track
Responding to a Roadblock Rehearsing and Presenting My Plan
Appendix F: Second Chance Skills Institute Participant Flow and Attrition
The Second Chance Skills Institute (SCSI) is expected to serve 450 participants per quarter and provide differing combinations of services to participants based on participants’ criminogenic needs and the SCSI’s expected capacity to deliver certain services. Alternative 2 includes an additional service option, postsecondary correctional education, relative to Alternative 1.
Participants require certain amounts of time to complete each of the primary service components: vocational or postsecondary correctional education, cognitive behavioral therapy, and therapeutic communities. Participants who start a service component may withdraw and fail to complete the component (attrition), which means they also exit from the SCSI.
These factors mean that the number of people entering the SCSI, starting or completing various service components, and graduating from the SCSI would ebb and flow on a quarterly basis. Modeling these factors is critical for a cost-benefit analysis because costs are often determined by the number of individuals receiving services, while benefits are determined by the number of participants who successfully complete the program and can be expected to have improvedoutcomes as aresult.
We use a modified version of life table analysis to project and analyze these demographic changes (Rowland, 2003). The projected population characteristics vary over the first few years as a result of the large initial cohort of 450 participants moving through the newly-opened SCSI. As participants reach different service components with differing attrition rates or graduate from the program each quarter, smaller cohorts of new participants enter the SCSI to maintain the population of 450 participants. By year five, the SCSI’s population is projected to be nearly stationary. The projected steady state for Alternative 1 is approximately 494 entrants, 142 withdrawals, and 352 graduates per year. The projected steady state for Alternative 2 is approximately 393 entrants, 121 withdrawals, and 272 graduates per year.
Institute Capacity, Participant Needs, and Service Component Attrition Rates
The current plan is to locate the SCSI at the Racine Youth Correctional Facility (with many of the youth currently housed there being transferred to other facilities in the state, and many individuals housed in other facilities in the state being transferred to the Second Chance Skills Institute). This facility currently has a daily population of 450; therefore, we assume that the SCSI’s quarterly population would be 450 (RYOC, 2018). The primary conditions for entering the SCSI are
1) having already obtained a high school diploma or its equivalent, and 2) being identified by Department of Corrections staff or contractors as being good candidates for needing, successfully completing, and benefiting from the SCSI services. We consider two alternative versions of theSCSI model; one of the major distinctions is that Alternative 1 assumes that only individuals who have not committed a violent offense can participate. Under Alternative 2, people who have committed violent crime can also be eligible to participate.
The SCSI is expected to provide participants with increased levels of therapeutic communities (TC), cognitive behavioral therapy (CBT), and correctional education (CE) services, in addition to the typical levels of services already typically available in Wisconsin state prisons (e.g., health services). Therapeutic communities is a substance use treatment model that we expect any inmate who meets the criteria for drug dependence or abuse to complete before beginning correctional education. The most recent, publicly available data from the federal Bureau of Justice’s National Inmate Surveys find that 58 percent of state prisoners meet these criteria (Bronson et al., 2017). We assume that 58 percent of SCSI entrants would have to complete substance use treatment before moving onto other services; the other 42 percent can immediately begin other types of services. Based on existing therapeutic community programs, SCSI’s program is expected to last 12 months and include extensive cognitive behavioral programming. As a result, we assume that any participants who complete TC would simultaneously fulfill the CBT element of the SCSI model.
After completing TC and CBT, participants can move on to correctional education. For participants who are not identified as meeting the criteria for substance dependence or abuse, they can immediately advance to correctional education. Because an extensive body of research finds that CBT has positive impacts, participants who have not already completed CBT as part of TC would instead participate in CBT sessions for the first three months at SCSI—a standard course of CBT treatment.
The amount of time participants require to complete correctional education depends on whether they participate in vocational or postsecondary education. In Alternative 1, participantsonly have the opportunity to complete vocational correctional education. In Alternative 2, we assume that 75 percent of participants would have the opportunity to complete correctional education and 25 percent would participate in postsecondary education. This split is meant to reflect the likelihood that not all individuals would succeed in or benefit from postsecondary correctional education, and that there may be limits to how much capacity there is to provide postsecondary education at the SCSI. We expect that participants would take six months to successfully complete a vocational correctional education program resulting in a labor market credential and two years to successfully complete a postsecondary correctional education program resulting in an Associate’sdegree.
Individuals who start SCSI and its various service components may not complete a service component, therefore withdrawing from the SCSI. For example, individuals may fail to comply with substance use treatment, apply the lessons from cognitive behavioral therapy to areas such as prison misconduct, or determine that they are not suited for additional education in a correctional setting.
Based on our review of therapeutic community research, we assume that 92.5 percent of TC participants would persist each quarter of the service component – i.e., 7.5 percent of those who start services in a quarter would withdraw within the next three months (Hiller et al., 1999; De Leon et al., 2000; Pelissier, 2004; Jensen & Kane, 2012; Welsh et al., 2014). Based on an average of the 2018 fiscal year retention rates for Wisconsin’s existing career and technical education and community corrections employment programs, we assume that 88.5 percent of vocational correctional education participants who have already completed TC (and its CBT elements) would persist through the first three months of vocational correctional education (DOC, 2018c). The relatively low first-quarter retention rate reflects how expanding correctional education to serve many more participants than are currently participating in correctional education in Wisconsin’s prisons would likely result in more individuals starting correctional education who are unable to complete the program. After the first quarter, we use the quarterly retention rate for Wisconsin’s existing career and technical education services as the basis for our assumption that 98 percent of participants would persist through each subsequent quarter of vocational correctional education. For those who participate in postsecondary education under Alternative 2, we lack information on postsecondary correctional education attrition rates, so we use the same quarterly retention rates as vocational correctional education (88.5 percent for the first quarter, 98 percent for the subsequent quarters through the two years needed to obtain an associate degree).
Individuals who do not participate in TC complete CBT during their first quarter of correctional education, which should result in a lower retention rate in that quarter. Based on an average of the 2018 fiscal year retention rates for Wisconsin’s existing career and technical education and CBT programs, we assume that 84 percent of vocational correctional education participants who have not already completed TC (and its CBT elements) would persist through their first three months when they are participating in vocational correctional education and CBT (DOC, 2018c). After the first quarter, the retention rate for correctional education for people who did not participate in TC rises to 98 percent. As we did for those participated in TC, we assume that the same retention rates for vocational correctional education apply to postsecondary education forthose who did not participate in TC services (84 percent for the first quarter, 98 percent for all subsequent quarters).
Life Table Results
The combination of different services, the time required to complete services, and attrition from those services means that the number of people entering the SCSI, starting or completing various service components, and graduating from the SCSI would vary over time. We project how these factors influence SCSI population characteristics using a form of life table analysis. Summary results from Alternatives 1 and 2 are presented in Table F.1: SCSI Life Table.
The projected SCSI population characteristics vary over the first few years as a result of the large initial cohort of 450 participants moving through the newly-opened SCSI. As participants withdraw or graduate, smaller cohorts of new participants enter the SCSI to maintain the Institute’s population of 450 participants. By year five, the SCSI’s projected population is nearly stationary. The steady state for Alternative 1 is approximately 494 entrants, 142 withdrawals, and 352 graduates per year. The steady state for Alternative 2 is approximately 393 entrants, 121 withdrawals, and 272 graduates per year.
Table F.1: SCSI Life Table
# of SCSI
|Total # of SCSI
withdrawals (all services)
# of TC completers
|# of CBT completers (including TC)||# of SCSI
graduates [Voc./PSE completers]
|Year 1||692||112||191||436||167 [167/0]|
|Year 2||347||128||131||254||298 [263/35]|
|Year 3||404||152||179||321||246 [186/60]|
|Year 4||396||91||157||297||265 [208/57]|
|Steady State||393||121||167||306||272 [210/62]|
G.) Facility Conversion and Expansion Costs
H.) Cost of Initial Staff Training in CBI-EMP
I.) Cost of Ongoing CBI-EMP Staff Training & Additional CBI-EMP Facilitator
Appendix G: Facility Conversion and Expansion Costs
In order to effectively house the SCSI, the Racine Youthful Offender Facility likely needs additional space for vocational training and other correctional programming, such as classrooms for cognitive behavioral therapy and therapeutic communities. As a part of the Ten-Year Correctional Facility System Development Plan, the DOC developed plans to expand current facilities to provide additional space for both vocational education and training (Mead & Hunt, 2009). Figure G.1 demonstrates the DOC’s plans to expand the RYOCF on the current site. Item G marks a new vocational education building. While the report did not suggest an expansion of the facility to increase bed capacity, it did recommend expanding the facility to increase the capacity for programming by constructing new training spaces, vocational education space, and work programs space.
Absent a DOC-developed, detailed inventory of facility alterations or expansions needed to transition the Racine Youthful Offender Correctional Facility into a Second Chance Skills Institute, these preliminary facility expansion cost estimates serve as our best approximation of the upfront costs required for appropriate facility conversion.
Using the DOC estimates, we predict that the construction of additional vocational education space and work program space would cost $2.7 million in 2018 dollars (Mead & Hunt, 2009).5 As this value seems potentially low, we use it as a lower bound and vary the cost upward in our Monte Carlo.
Cost adjustments to account for inflation were calculated using the Bureau of Labor Statistics’ inflation calculator (2018a) and cross-referenced using construction-industry specific inflation estimates.
Appendix H: Cost of Initial Staff Training in CBI-EMP
Staff at the SCSI would be trained to deploy the Cognitive Behavioral Interventions for Offenders (CBI-EMP) curriculum. Over a three day period, facilitators are trained on how to integrate cognitive behavioral interventions with conventional employment approaches. In the past, the Department of Corrections has contracted with the University of Cincinnati Corrections Institute for CBI-EMP training. In fiscal year 2018, the Department of Corrections trained 18 individuals over a three-day period at a total cost of $13,750. This provides a cost of $764 for each individual who participates in CBI-EMP training.
Our recommendation is that the SCSI train as many individuals as possible. At a minimum, corrections staff that interact with participants on a daily basis should be trained in CBI-EMP. This includes, but is not limited to, the following departments: Education, Security, Housing, Program Services, Health Services, Psychological Services, Workforce Development, and any university, technical college, and industry partners. Many organizations require every staff member to complete the training, including leadership, in order to build a common understanding of the objectives and strategies that would be deployed by staff on a daily basis. As such, our estimate assumes that every staff member at the SCSI would be trained in CBI-EMP.
In order to estimate the number of staff that would be employed at the SCSI, we took the average of various corrections facilities inmate-to-staff ratios. Corrections facility inmate-to-staff ratios generally ranged from 2:1 to 4:1 (DOC Annual Reports). Our generated average ratio is 3.26:1. As a result, we estimate that the SCSI would maintain approximately 138 staff members.
To estimate the cost to pay staff members, or the replacement staff for individuals while they are at training, we used the median hourly wage for correctional officers in the state of Wisconsin, which was estimated by the Bureau of Labor Statistics to be $20.61. In order to capture the full cost, we multiplied the median wage by 1.58 to account for total compensation, including fringe benefits, for state and local government employees (Bureau of Labor Statistics, 2018b). To fully monetize this cost, we take the compensation per hour multiplied by the number of hours of training required and the number of staff members. Given these estimates, we assume that the total upfront cost related to training, accounting for both marginal cost of training as well as cost to pay staff for training, amounts to approximately $213,267. See full estimates in Table H.1: Upfront Staff Training Costs.
Table H.1: Upfront Staff Training Costs
|Marginal Cost of Staff Training ($)||Total Cost of Training for 138 Staff Members ($)|
|Marginal Cost of Training||764||105,415|
|Cost of Loss in Productivity||781||107,850|
|Total Cost of Training||1,545||213,265|
Appendix I: Cost of Ongoing CBI-EMP Staff Training & Additional CBI-EMP Facilitator
While we recommend that all staff be trained in CBI-EMP, staff turnover would naturally lead to situations where some staff members are not trained in CBI-EMP at any given time. We recognize that this must be factored into ongoing cost calculations related to staff training.
Therefore, we recommend that the SCSI provide CBI-EMP professional development every other year on a recurring basis.
We analyzed multiple turnover rate point-estimates and ranges in order to establish an appropriate turnover rate that can be accounted for in our analysis of ongoing training costs. Between 2015 and 2016, Wisconsin experienced a state employee turnover rate of 11.4 percent, which is likely most applicable to non-correctional office staff. Peer-reviewed research has shown that correctional officer turnover rates usually fall between 15 and 25 percent (Lambert et al., 2009; Minor et al., 2011). Additionally, literature on turnover rates for correctional officers in states that consistently struggle to keep their correctional staff suggest that the annual rates generally fluctuate between 18 percent and 30 percent (Fifield, 2016).
Table I.1: Staff Turnover Rates
Source of Estimate
|Annual Turnover Rate Range (%)||Annual Turnover Rate Point-Estimate (%)|
|State Employee Turnover Rate (Milwaukee Journal Sentinel)||
|Kansas Correctional Officer Turnover Rate (Pew)||20.8 – 29.7||25.3|
|Nebraska Correctional Officer Turnover Rate (Pew)||18.5 – 30.8||24.65|
|Lambert & Hogan (2009)||15 – 25||20|
|Minor et al. (2011)||18.1||18.1|
|Average Turnover Rate||19.9|
By offering CBI-EMP professional development opportunities every other year to untrained staff at the SCSI, the Department of Corrections is able to combat the effects of staff turnover. Using our estimate of 19.9 percent, the SCSI would be training a cohort of approximately 55 employees every other year. In order to calculate the annual cost of the program, we divided this cohort by two. By assuming we train 27.5 employees each year, we reach a yearly cost estimate of $42,500.
Correctional facilities throughout the state of Wisconsin maintain staff for specialized programming, such as vocational training and substance use disorder treatment. The cost of expanding these programs are included within our cost estimates for those programs, however, our training costs for CBI-EMP do not account for any additional staff that may be needed to provide CBI-EMP to these participants. Since CBI-EMP group instruction can be delivered by existing staff positions, we have determined that in order to provide the one hour a week, 31-week programming to all participants not enrolled in Therapeutic Community, the SCSI would likely only need to hire one additional CBI-EMP facilitator.
We found full-time equivalent cost-estimates using a variety of point-estimates. A correctional education teacher’s starting salary is approximately $45,000. An advanced youth counselor’s starting salary is $35,600 and the starting salary for a correctional vocational education teacher is $45,000. Employment coordinator salaries range from $36,608 to $60,382 and a DOC budget request provides us with an estimated salary of $41,000 for youth counselors (DOC, 2016a). These estimates are similar to the average correctional officer salary of $41,770 (Beck, 2016). When averaged, these estimates provide us with a cost estimate for a 1.0 FTE CBI-EMP facilitator with a starting salary of $43,750 and, including benefits, an annual compensation cost of $69,078.66 to calculate the full compensation cost, we use the same method as explained in Appendix H and multiply the estimated salary by 1.58 (BLS, 2018c).
J.) Cost of Therapeutic Communities
K.) Cost of Vocational Training
L.) Cost of Postsecondary Education
Appendix J: Cost of Therapeutic Communities
Individuals at the SCSI that have been identified as needing substance use disorder treatment would participate in therapeutic communities. The Washington State Institute for Public Policy conducted a cost-benefit analysis that determined the annual marginal cost per participant was
$2,198 (WSIPP, 2016c).The DOC does not have cost estimates for therapeutic communities, so we utilize WSIPP’s point estimate for the marginal cost of the program per enrollee to calculate the cost of delivering the program at the SCSI.
Table J.1: Therapeutic Community Costs illustrates the fluctuating participation and program costs over the first four years of its operation for Alternative 1 and Alternative 2, after which we provide the stabilized long-term annual cost of therapeutic communities.
Table J.1: Therapeutic Community Costs
# of SCSI
Marginal Daily Cost ($)
Total Daily Cost ($)
Marginal Annual Cost ($)
Total Annual Cost ($)
Appendix K: Cost of Vocational Training
The SCSI would provide participants the chance to take advantage of vocational programming, however, this would require a sizable expansion of the programming. Using multiple data sources and analyses, we derived estimates of the costs related to expanding these programs in the SCSI. The Washington State Institute of Public Policy conducted a cost-benefit analysis of offering vocational training to individuals who are incarcerated, finding a marginal cost of $1,495 per enrolled participant (2016d).
Using data provided in various DOC documents, such as the DOC 2017-19 Biennial Budget Issue Paper and a 2017 Legislative Fiscal Bureau Report, we have calculated a Wisconsin-specific average marginal cost of vocational training expansion. These reports have indicated that the marginal cost of expanding vocational programs at various sites ranges between approximately $2,500 to $5,000 (DOC, 2016a; LFB, 2017). Our average marginal cost estimates of expanding vocational training were calculated by dividing the total cost of expansion provided in the DOC and LFB reports by the total number of additional participants. For example, the total cost of expanding CNC training to 30 individuals at the Racine Correctional Institution is $83,000. By dividing $83,000 by 30, we yield a marginal cost per participant of $2,766.
Due to the location and nature of the vocational training at the SCSI, we have decided to use the Racine-specific cost estimate for expanding vocational training. As such, $2,766 is our point estimate for annual marginal cost (DOC 2016a; RYOCF, 2018). However, it is worth noting that our estimates are potentially conservative because they do not take into account the cost savings associated with a concentration of services provided at one site, rather than spread out to a select few individuals throughout Wisconsin’s correctional system. The cost estimates are shown in Table K.1: Vocational Education Estimates.
Table K.1: Vocational Education Estimates
Source of Estimate
|Marginal Annual Cost ($)||Marginal Daily Cost ($)|
|WSIPP CTE Analysis (2016)||1,495||4.10|
|DOC 17-19 Budget Request (Racine Specific)||2,766||7.58|
|DOC 17-19 Budget Request (General)||4,545||12.45|
Table K.2: Vocational Education Costs illustrates the fluctuating participation and program costs over the first four years of its operation for Alternative 1 and Alternative 2, after which we provide the stabilized long-term annual cost of vocational education.
Table K.2: Vocational Education Costs
|# of SCSI
Marginal Daily Cost ($)
Total Daily Cost ($)
Marginal Annual Cost ($)
Total Annual Cost ($)
Appendix L: Cost of Postsecondary Education
Under Alternative 2, the SCSI provides the option for postsecondary education, which would impose an additional cost to the WI DOC.
Our research has provided us with only one peer-reviewed point-estimate for the marginal cost of providing an incarcerated individual with postsecondary education programming. The Washington State Institute for Public Policy (WSIPP) conducted research on the cost of providing postsecondary education programming in correctional facilities and published a cost-benefit analysis that specifies a marginal annual cost of $1,249 per participant per year.
As we have only one plausible estimate for the cost of delivering postsecondary education programming to individuals that are incarcerated, we attempted to calculate a Wisconsin specific point-estimate by multiplying our WSIPP estimate by the cost-differential ratio between WSIPP’s vocational education cost-estimate and the Wisconsin DOC’s vocational education cost-estimate. When we divide our DOC vocational education cost estimate, $2,766, by WSIPP’s vocational education cost estimate, $1,495, we receive a cost-differential ratio of 1.85. Using this ratio, we can estimate that a Wisconsin DOC marginal cost estimate for delivering postsecondary education to a single participant in the SCSI would amount to approximately $2,310.7 In addition to calculating this point estimate, we also use the cost-differential between WSIPP’s vocational education estimate and our Wisconsin-specific vocational education upper bound estimate to calculate an upper bound for postsecondary education in Wisconsin of $3,797.8 We believe that this DOC-WSIPP ratio adjusted estimate will provide a more cautious, Wisconsin-specific cost-estimate. For full detail see Table L.1: Postsecondary Education Cost Estimates.
($1,249 x 1.85 = $2,310)
(1,249 x 3.04 = $3,797)
Table L.1: Postsecondary Education Cost Estimates
Source of Estimate
|Marginal Annual Cost ($)||Marginal Daily Cost ($)|
|WSIPP PSE Analysis (2016)||1,249||3.42|
|DOC-WSIPP Low CTE
Cost-Differential Ratio Adjusted
|DOC-WSIPP High CTE
Cost-Differential Ratio Adjusted
Table L.2: Postsecondary Education Costs illustrates the fluctuating participation and program costs over the first four years of its operation, after which we provide the stabilized long- term average annual cost of the postsecondary education program. We only provide estimates for Alternative 2 as we do not include this component in Alternative 1.
Table L.2: Program Costs Postsecondary Education
Year of Operation
|Annual # of Program Enrollees||Marginal Daily Cost ($)||
Total Daily Cost ($)
|Marginal Annual Cost ($)||Total Annual Cost ($)|
M.) Benefits from Reduced Suicide in Prison
N.) Benefits from Earned Release
O.) Benefits from Reduced Substance Use
Appendix M: Benefits from Reduced Suicide in Prison
While researching possible impacts of SCSI programming on overall health of people in prison, we identified a strong impact of CBT on suicide rates (Brown et al., 2005; Tarrier et al., 2008). To calculate the impact of the SCSI on reducing suicides, we multiply the reduction insuicide by the count of people who complete the full CBT program in the SCSI in one year. This gives us the total reduction in suicides resulting from the SCSI programming. We monetize the benefit by multiplying this value by an adjusted Value of a Statistical Life.
Tarrier et al. (2008) conduct a meta-analysis of studies on the effect of various cognitive- behavioral interventions on suicidal behavior. The CBT programs that the authors analyzed were comparable to the SCSI in terms of length and number of sessions. We use the Hedges’ g effect size of CBT of -0.562 from Tarrier et al. (2008) in our analysis, which translates to a 0.012 percentage point decrease in suicide from the current baseline.
Importantly, the effect size from the literature is taken from within three months after CBT programming is completed. Based on our timeline, all participants would still be incarcerated and completing SCSI programming within three months of completing CBT, so we are specifically reducing the number of suicides that occur in prison. To account for this, we use the suicide rate in Wisconsin state prisons as the baseline in our analysis. The Bureau of Justice Statistics reports that the average annual mortality rate by suicide in Wisconsin state prisons between 2001-2014 was 19 suicides per 100,000 people in prison, or 0.00019 (Noonan, 2016).
To calculate a monetary value of this benefit per participant, we apply the reduction in suicide rates to an adjusted estimate of the Value of a Statistical Life (VSL). VSL specifically refers to a person’s willingness to pay to reduce their own mortality risk, so some argue that VSL should account for individuals’ income because this affects individuals’ ability and willingness to pay (Robinson & Hammitt, 2015). To be conservative in our estimates, we use an adjusted VSL to account for the fact that people who have been in prison generally have lower income than people who have not been in prison. We first calculate a lower bound for an adjusted VSL using the median income of people before they are incarcerated. Looney and Turner (2018) find that the median income of people in prison three years prior to incarceration is $6,251, conditional on employment, with 49 percent employed. We adjust this value to account for those who are not employed (annual income of $0), and end up with a median annual income of $3,063. Second, we calculate a point estimate for an adjusted VSL using the average cost of incarcerating one person for one year as a proxy for the income of people in prison.
We use the following equations from Hammitt & Robinson (2011) to calculate a minimum value and point estimate for an adjusted VSL:
𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑𝑉𝑆𝐿=𝑉𝑆𝐿 ∗(𝑐𝑜𝑠𝑡 _ 𝑝𝑟𝑖𝑠𝑜𝑛/𝑖𝑛𝑐𝑜𝑚𝑒_𝑈𝑆)^𝑒𝑙𝑎𝑠𝑡𝑖𝑐𝑖𝑡y
𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑𝑉𝑆𝐿 = 11,740,000 ∗ (32,450/62,955.57)^2.24
Where VSL is the median value for VSL found in the literature by Boardman et al. ($11.74 million); income_prison is the median annual income of a person who has been in prison, conditional on employment ($10,090); income_US is the median household income in the United States ($62,955); cost_prison is the average cost of incarcerating one individual for one year in Wisconsin according to the WI DOC ($32,450) and elasticity is the income elasticity of VSL in the lowest decile of income (2.24), based on analysis by Kniesner et al. (2010). Using quantile regressions, Kniesner et al. (2010) find that the income elasticity of VSL generally decreases as income increases, with the income elasticity of VSL for the lowest income decile being the highest elasticity. This implies that the VSL is more strongly affected by income at lower levels of income than at higher levels of income. All dollar values are reported in 2018 dollars, with any necessary adjustments made using the CPI inflation calculator (Bureau of Labor Statistics, 2018a).
We ultimately conduct a sensitivity analysis using adjustedVSL as our point estimate, min(adjustedVSL) as a lower bound and the full VSL ($11.74 million) as the upper bound. We discuss this in more detail in our description of our Monte Carlo analysis in Appendix U: Variable Distributions for Monte Carlo Simulation.
Table M.1: Variables Calculating the Adjusted VSL
|VSL ($)||11,740,000||Boardman et al. (2018)|
|income_prison ($)||3,063||Looney & Turner (2018)|
|cost_prison ($)||32,450||WI DOC|
|income_US ($)||62,955||Semega et al. (2018)|
|elasticity||2.24||Kniesner et al. (2010)|
We assume that there would not be any partial effects of CBT on suicide for people who dropped out of the program during the CBT portion, so our estimates only apply to participants who complete the entire CBT portion of the SCSI programming. We therefore multiply by the count of SCSI participants who complete CBT each year.
We use the following equation to calculate the benefits of reduced suicide in prison:
Where suicide_base is the current suicide rate in Wisconsin prisons, suicide_CBT is the suicide rate for people in the SCSI programming, cbt_completers is the number of SCSI participants who complete CBT programming in a year, and adjustedVSL is the adjusted VSL that we calculated to account for people in prison having lower incomes on average. The number of CBT completers varies depending on the alternative and year of the program, so we calculate benefits of reduced suicide for each alternative and year.
Table M.2: Variables Estimating the Impact on Suicide Rate
|suicide_CBT (%)||0.006857||Tarrier et al. (2008)|
|suicide_base (%)||0.019||Noonan (2016)|
|adjustedVSL ($)||2,660,434||Boardman et al. (2018); Hammitt & Robinson (2011); Kniesner et al. (2010)|
We report the count of CBT completers for the first five years along with the benefits for that number of participants in Table M.3: Benefits from Reduced Suicide in Prisons
Table M.3: Benefits from Reduced Suicide in Prison
|Year||CBT Completers||Benefits ($)|
Appendix N: Benefits from Earned Release
As the SCSI provides earned release to each individual who successfully completes each part of the programming, we estimate the benefits from the avoided cost of housing those individuals. To calculate this benefit, we multiply the marginal cost of early release by the number of days of incarceration avoided per graduate and the number of graduates each year.
To calculate the marginal cost of earned release, we use estimates from the Wisconsin Legislative Fiscal Bureau (LFB) (n.d) on the cost savings expected from expanding earned release to additional individuals. Their estimate includes the cost of hiring new staff to administer the program, as well as the savings generated from the reduced need for a contract bed. The LFB estimate for the annual marginal cost savings per individual from earned release is between $8,338 and $11,950 (2017-2018 and 2018-2019 estimates). We divided the yearly estimate by four quarters, as each of the graduates are only granted three months early release rather than a full year, yielding estimates ranging between $2,084 to $2,987 per quarter. We use $2,987 as the point estimate, as the number of additional individuals added under the 2018-2019 estimate (237) is identical to the number of graduates we expect in year one.
To create a range of estimates for our sensitivity analysis, we take the $2,084 from the LFB report estimates as our lower bound and use the percent difference between $2,084 and $2,987 to calculate a symmetrical upper bound of $4,281.10 As our number of graduates ultimately increases above 237, we assume that the cost savings would increase as well, explaining the creation of upper bound above the current estimates. We expect the savings from earned release to grow higher as the number of participants increase, as this is how the LFB estimated the savings between their 2018-19 and 2017-18 estimates.
The percent difference is calculated as (2987-2084)/2084, giving us a difference of 43.3%. We then take 1.43 (1+percent difference) multiplied by 2987 to create an upper bound of 4281.
We multiply this calculated savings per graduate by the number of graduates per year in the SCSI under each alternative. We use the following equation to fully quantify this benefit:
Where b_er is the fully quantified benefit for each year, total_grads is the total number of graduates per year, and marginal_savings_erare the cost savings from early release expected per individual.
As we expect a different number of graduates until the SCSI reaches a steady-state, we calculate the benefits for each of the first four years and for yearly benefit from that point on. These results are summarized in Table N.1 Earned Release Benefits.
Table N.1: Earned Release Benefits
Appendix O: Benefits from Reduced Substance Use
Based on evidence from the literature, the therapeutic community component of SCSI programming would likely reduce the odds of substance use post-release (Pearson & Lipton, 1999; Mitchell et al., 2018). To estimate these benefits, we calculate a percentage point decline in substance use post-release and multiply this by the number of therapeutic community completers per year and the one-year cost of substance misuse per individual.
A meta-analysis by Mitchell et al. (2018) examined the effect of therapeutic communities within prison on a number of outcomes, including substance use rates one year after release.
Mitchell et al. (2018) reported an odds ratio of 1.33 for the impact of therapeutic communities on substance use rates, which means that individuals who participated in therapeutic communities had a lower likelihood of substance use compared to the control group. As Mitchell et al. (2018) only report outcomes for one-year post release, we assume that this impact only lasts for one year after release from the SCSI. Additionally, WSIPP’s reported effect sizes for substance use disorder treatments support this assumption as they report that an effect size of zero after the first year. Using this odds ratio and a baseline rate of substance dependence in the first year of release of 43 percent, we calculate a 4.16 percentage point difference.
To monetize this benefit, we take the percentage point difference multiplied by the number of therapeutic completers per year and the estimated economic and social cost of substance use disorder. To quantify the cost of substance use disorder, we rely on two studies on costs related to substance abuse. The first study (Ettner et al., 2006) surveyed participants of substance use treatment programs on medical costs pre- and post-treatment (9 months post) in order to calculate the reduction in costs that occur as a result of treatment. The study reports on the cost of hospital nights for medical problems, the cost of emergency room visits, and the cost of inpatient and outpatient mental health services. Together, the cost reductions after treatment equal $1,545 in 2018 dollars.
For more details on this type of calculation, see Appendix T: Turning Effect Sizes into Percentage PointDifferences.
The second cost of substance use is the opportunity cost from the manufacture and sale of substances. Cohen and Piquero (2009) estimate this cost by using a fraction of the average lifetime purchases on substance use by a heavy drug user. The authors use a range of 50 to 75 percent of total purchases as they assume that the average purchase price includes a substantial risk premium to account for the opportunity cost of drug distribution. We replicate this method, but use annual estimates of substance use purchases as we only expect our treatment to have an impact for the first year after release. We use an estimate of $4517 for the annual retail expenditure on substances (Kilmer et al., 2014), which we multiply by the average of the risk premium range used in Cohen (62.5 percent) to get an adjusted value of $2,936. After inflating to 2018 dollars, our estimate on the opportunity cost of substance use is $3,426. The medical costs and substance use costs give us a total value for one year of avoided substance use of $4971.
To fully quantify this benefit, we use the following equation:
𝑏_𝑠𝑢𝑏𝑢𝑠𝑒=(𝑠𝑢𝑏𝑢𝑠𝑒_𝑏𝑎𝑠𝑒−𝑠𝑢𝑏𝑢𝑠𝑒_𝑡ℎ𝑒𝑟) ∗𝑠𝑢𝑏𝑢𝑠𝑒_𝑐𝑜𝑠𝑡∗𝑇𝐶_𝑐𝑜𝑚𝑝𝑙𝑒𝑡𝑒𝑟𝑠Where b_subuse is the quantified benefit, subuse_base is the current substance use rates for people released from prison, subuse_ther is the substance use rates for people released from the Second Chance Skills Institute programming, subuse_cost is the economic cost of substance use per individual for one year, TC_completers is the number of Second Chance Skills Institute participants who complete therapeutic communities in a year.
Table O.1: Substance Use Variables
|subuse_base (%)||43||Malouf et al (2013)|
|subuse_ther (%)||38.9||Mitchell et al. (2018)|
|subuse_cost ($)||4791||Kilmer et al. (2014); Ettner et al. (2006)|
As we expect a different number of graduates until the SCSI reaches a steady-state, we calculate the benefits for each of the first four years based on the changing number of graduates and then report the annual benefit for the steady state of the SCSI. These results are summarized in Table O.2: Substance Use Reduction Benefits.
Table O.2: Substance Use Reduction Benefits
|Year||TC Completers||Benefits ($)|
P.) Benefits from Reduced Crime and Reincarceration Rates
Q.) Benefits from Reduced Misconduct
R.) Benefits from Increased Lifetime Compensation
Appendix P: Benefits from Reduced Crime and Reincarceration Rates
To calculate the benefits from reduced crime and reincarceration, we multiply our projections of SCSI graduates by estimates of program impacts on crime and the cost of crime. Because there is no empirical research on the impact of the exact combination of services modeled for the Institute, our estimate is based on findings from past evaluations of individual service components that are adjusted for the possibility that component effects might be overlapping and not additive. The cost of crime accounts for the wide variety of tangible and intangible social costs of crime, including the costs to victims of crime and to government for criminal justice costs.
Intervention Component Impacts on Crime and Reincarceration
The Washington State Institute for Public Policy provides estimates for the effect sizes on crime of vocational and postsecondary correctional education included in Alternatives 1 and 2.
WSIPP finds that vocational correctional education has an effect size of .167, which translates into a nearly 4.6 percentage point reduction in the predicted rate of crime for Second Chance participants, relative to the baseline level of reincarceration within three years of release in Wisconsin, 37.5 percent (2016d; WI DOC, 2018).13 WSIPP (2016b) finds that postsecondary correctional education has an effect size of .227, which translates into a more than 6.2 percentage point difference relative to the baseline level of reincarceration in Wisconsin. Under Alternative 2, however, participants would receive either vocational or postsecondary education. Therefore, we take an average of the vocational and postsecondary education impacts, weighted by respective number of graduates from the two types of education, to produce a weighted correctional education impact under Alternative 2. For example, at the steady state of Alternative 2 in which there are 210 graduates from vocational education and 62 graduates from postsecondary education, the average impact would be 4.9 percentage points.
For more details on this type of calculation, see Appendix T: Turning Effect Sizes into Percentage Point Differences.
WSIPP (2016a) finds that cognitive-behavioral therapy has an effect size of .109, which translates to a 2.95 percentage point difference relative to the baseline level of reincarceration in Wisconsin. WSIPP (2016c) estimates that therapeutic communities for substance use disorder have an effect size of .089, which translates into a 2.4 percentage point difference relative to the baseline level of reincarceration in Wisconsin. Research to date has not found any impact on reincarceration for the rehabilitation certifications that would be provided to SCSI graduates (Leasure & Stevens- Anderson, 2016; Leasure & Martin, 2017). As these intervention components would be delivered to the same individuals, we have to account for potential double counting of impacts, in which individuals who would be helped by one component are the same individuals who would benefit from another component. Because therapeutic community interventions often incorporate CBT, we assume that the CBT and therapeutic community intervention effects should not be added in order to avoid double-counting, with the therapeutic community intervention effects being completely subsumed by the larger effects found in the literature for CBT. However, we assume that the vocational and postsecondary correctional education effects do not overlap with those of CBT (and therapeutic communities), so these effect sizes can be simply added. In summary, we estimate that the mixture of services provided under Alternative 1 at the steady state would reduce the three-year rate of reincarceration for SCSI graduates from 37.5 percent to 30 percent, a reduction of 7.5 percentage points. For Alternative 2’s steady state, there is a reduction of 7.9 percentage points, resulting in a reincarceration rate of 29.6 percent. Table P.1: Service Component Impact Variables at Steady State summarizes these various service component combinations. We make the conservative assumption that individuals who complete the TC and CBT service components but do not graduate from SCSI (i.e., withdraw during the education component) experience no reductions in reincarceration as a result of that completion when they are released from prison.
Table P.1: Service Component Impact Variables at Steady State
|Component(s)||Point Estimate (%)||Source|
|recid_base||37.5||(WI DOC, n.d.)|
Cost of Crime
To calculate the cost of crime, we make the assumption that revocations are unaffected by any of the Second Chance Institute’s intervention components. For new offenses, the shadow price of crime includes the victim costs (tangible and intangible victim pain and suffering costs) and the costs of the criminal justice system (including police protection costs, legal and adjudication costs, and corrections costs) (McCollister et al., 2012). To find the average cost of crime across offense types, we use the 2016 FBI Crime Data for Wisconsin and McCollister et al.’s (2017) estimates of the cost of crime for various offense types (CJISD, 2016). The FBI Crime Data allows us to create a distribution of arrests by type of offense, which we assume is equivalent for recidivating individuals.14 As Alternative 1 does not include people who have committed violent offenses, we subtract the number of violent offenses from the total number of arrests and do not include those offenses in our calculation of the distribution of arrests. We then multiply the percent of arrests for each offense type by the cost of crime, using estimates from McCollister et al. (2012) and Blincoe et al. (2015).15 Cost of crime estimates are not available for each type of offense documented by the FBI Crime Data, so we group several offense types into Other categories and use representative cost estimates based on length of time typically served or cost of crime for similar offenses.16 Table P.2: Social Costs of Crime – Policy Alternative One shows the estimates for each type of crime, which were summed to calculate the cost of crime for all non-violent offenses committed by adult offenders, $3,645.
The FBI Crime Statistics report total arrests and number of arrests by type of crime for individuals under the age of 18 and all individuals. To calculate the percent of arrests by type of offense for adult populations, we took the total number of arrests by crime type minus the number of arrests by crime type for those under 18 divided by the total number of arrests.
These cost of crime estimates are inflated using CPI from 2008 dollars to October 2018 dollars.
For Public Order-Other and Other, we use the cost of a alcohol driving offense, without injury, as it is the most typical public order offense. For drug abuse violations, we found the average sentence time for a drug offense and compared it to average sentence times for other crimes. Based on that comparison, we use the cost of lacrency for the cost of drug abuse violations as the time served are closest.
Table P.2: Social Cost of Crime – Policy Alternative One
|Type of Crime||Percent of Arrests||Cost of Crime ($)||Total ($)|
|Motor vehicle theft||.003809||11,958||45|
|Forgery and counterfeiting||.00513||5,517||28|
|Drug abuse violations||.124475||4,024||500|
|Time-Discounted Total (cost_nvcrime)||3,488|
The second alternative broadens the eligibility requirements to include violent offenders. For thecost of crime, we used the same method as above to get an estimate of the total cost of crime including violent offenses.17 Table P.3: Social Costs of Crime – Policy Alternative Two shows the estimates for each type of crime, which were summed to calculate the total cost of crime committed by adult offenders,$20,792.
- There was one violent crime offense, other assaults, that did not have a cost of crime. We used the lowest cost of crime for a violent offense, the cost of robbery, as an estimate of this offense.
Table P.3: Social Cost of Crime – Policy Alternative Two
|Type of Crime||Percent of Arrests||Cost of Crime ($)||Total ($)|
|Motor vehicle theft||.003466||12,271||42|
|Forgery and counterfeiting||.004668||5,517||25|
|Drug abuse violations||.113263||4,024||455|
|Time-Discounted Total (cost_crime)||19,902|
These estimates are the average present costs of crime committed very shortly after someone is released from prison. However, we expect the program’s reductions in crime to occur over the three-year period after release, so the costs of crime have to be discounted for the fact that they are going to occur in the future (at some point over the three-years following release). In order to estimate when the average avoided crime would occur, we take a weighted average of the 1-year (15.8 percent), 2-year (25.7 percent), and 3-year (33.8 percent) reincarceration rates, but assuming that each year’s reincarceration is uniformly distributed within that year, to estimate that the average avoided crime occurs 1.27 years post-release. Assuming a discount rate of 3.5 percent and the need to discount for 1.27 years, the cost of crime for Alternatives 1 and 2 are $3,488 and $19,902, respectively.
To estimate the benefits of reduced crime from a single-year’s cohort, we use the following models for Alternative 1 and Alternative 2 graduates, respectively:
Where b_recid_alt1 is the benefit under Alternative 1 and b_recid_alt2 is the benefit under Alternative 2, cost _nvcrime is the average cost of crime for nonviolent crimes, cost_crime is the average cost of crime for all offenses, total_grads_alt1 is the number of graduates in a year for Alternative 1, total_grads_alt2 is the number of graduates in a year for Alternative 2, recid_base is the baseline reincarceration rate in Wisconsin, recid_scsi_alt1is the average reincarceration rate for SCSI graduates in Alternative 1 andrecid_scsi_alt2 is the average reincarceration rate for SCSI graduates in Alternative 2.
As we expect a different number of graduates until the SCSI reaches a steady-state, we calculate the benefits for each of the first four years based on the changing number of graduates and then report the annual benefit for the steady state of the SCSI. These results are summarized in Table P.4: Benefits of Reduced Crime and Reincarceration.
Table P.4: Benefits of Reduced Crime and Reincarceration
|Year||SCSI Graduates [Vocational/Postsecondary]||Benefits ($)|
|Year 1||167 [167/0]||249,393|
|Year 2||298 [263/35]||456,663|
|Year 3||246 [186/60]||387,321|
|Year 4||265 [208/57]||414,697|
|Steady State||272 [210/62]||426,813|
Appendix Q: Benefits from Reduced Misconduct
Reduced misconduct during incarceration is a well-documented benefit from correctional education (Delaney et. al 2016; Winterfield et al. 2009; Brazzell et al. 2009; Gerber & Fritsch, 1995; Lahm, 2009), cognitive behavioral interventions (Di Placido et al., 2006; Baro et al., 1999), and therapeutic communities (Dietz et al. 2003; Zhang et al. 2009; Prendergast et al., 2001). To calculate the impact of the Second Chance Institute on misconduct, we multiply the percentage point decline in institutional misconduct expected from SCSI programming by the cost of misconduct. As limited data on the cost of institutional misconduct exists, we focus on the reductions in institutional violence, specifically assaults. We estimate the impact on people who are incarcerated and staff separately as the base rates of assault differ among the groups.
Although it is likely that each of the programs would reduce misconduct, the quantitative findings on misconduct are less straightforward. To begin with, we were not able to identify any quantitative studies that estimate usable effect sizes for two of the main programming components – therapeutic communities and correctional education. For correctional education, there is one article by Pompoco et al. (2017) that reports mixed results on institutional misconduct within Ohio’s correctional system. Similarly, the articles on therapeutic communities report lower rates of misconduct within individual prisons (Dietz et al. 2003; Zhang et al. 2009), but do not translate these findings into effect sizes. For this reason, we focus only on the impacts from CBT.
To estimate the impact of CBT, a number of studies attempt to measure the effect on violence in different ways. Beck and Fernandez (1998) completed a meta-analysis on the impacts of CBT on anger among all populations and report an effect size of .70. Similarly, Saini (2009) reports an effect size of .60 for a meta-analysis on how CBT reduces anger within individuals. Lastly, Morgan and Flora (2002) completed a meta-analysis on the impacts of psychotherapy on institutional adjustment in prison (measured through disciplinary records). Morgan and Flora report an effect size of .43 for institutional adjustment after adjusting for outlier studies and their analysis shows that CBT studies report the largest effect sizes. Although this study does not directly measure CBT’s effect on violence, its results are specific to prison populations and it provides a measure for institutional misconduct, a closer approximation to rates of violent misconduct. For those reasons, we use the institutional adjustment effect size, .43, in our estimations, but we do recognize that the estimate may be high as it accounts for all disciplinary misconduct while our final quantification is only on violent misconduct. For this reason, along with the small sample sizes used in the studies, we are highly uncertain about these effect sizes, which we account for in the Monte Carlo.
People in Prison
To fully calculate the point estimates from reduced violent misconduct for people in prison, the effect size must be converted to a reduction in assaults. The effect size of .43 from Morgan and Flora (2002) translates to a 9.3 percentage point difference in the number of people that would experience assault while in prison, based on a 22.3 percent baseline as reported in Wolff et al. (2009). We expect this baseline rate of assault to be an overestimation as the SCSI requires a screening to determine eligibility for the program and thus, we think that the individuals in the SCSI population would be less likely to commit misconduct or assault compared to the general prison population.
In order to monetize the reduced assaults, we use the cost to a victim from an assault as we could not find the cost of assault or violent misconduct within prison. There are a number of sources that calculate the victim cost of assault. For this estimation, we use Miller et al. (1996) as they separate the victim costs by productivity losses, medical expenses, and quality of life adjustments.
This allows us to only include the medical expenses for this population, which is a conservative estimate. We do not include additional monetary estimates for two reasons: (1) these individuals likely do not see a large loss in productivity and (2) quality of life adjustments are based on a willingness to pay or median income, which we are unable to estimate for this population. Miller et al. also report costs for assault with an injury and without an injury. The total medical costs amount to $1,736 in 2018 dollars for assaults that result in an injury and $358 in 2018 dollars for assaults without injury. Based on Wolff and Shi (2009), 40 percent of assaults in prison result in injury, 33 percent of the assaults result in medical attention, and 21 percent of assaults in prison result in required hospitalization. As the injury versus non-injury cost estimates vary due to medical care costs, we use the 33 percent estimate to find a weighted average of the cost of assault on prisoners of $822.
Table Q.1: Misconduct Variables (Inmates)
|ar_inmate_base (%)||22.3||Wolff et al. (2013)|
|ar_inmate_CBT (%)||9.3||Morgan & Flora (2002)|
|cost_inmate ($)||822||Miller et. al (1996)|
To monetize the benefit, we use the following equation:
𝑏_𝑚_𝑖𝑛𝑚𝑎𝑡𝑒𝑠=(𝑎𝑟_𝑖𝑛𝑚𝑎𝑡𝑒𝑠_𝑏𝑎𝑠𝑒−𝑎𝑟_𝑖𝑛𝑚𝑎𝑡𝑒𝑠_𝐶𝐵𝑇)∗𝐶𝐵𝑇_𝑐𝑜𝑚𝑝𝑙𝑒𝑡𝑒𝑟𝑠∗𝑐𝑜𝑠𝑡_𝑖𝑛𝑚𝑎𝑡𝑒Where b_m_inmates is the quantified benefit, ar_inmate_base is the baseline percent of people who experience assault, ar_inmate_CBT is the percent of people who experience assault in the SCSI, CBT_completers is the number of people who complete CBT (either through CBI:EMP or through therapeutic communities) per year, and cost_inmate is the cost to a victim for assault.
To calculate the benefit of reduced assaults for staff, we use the same method, but rely on baseline assault rates on staff, and account for both quality of life adjustments and lost productivity for staff members in our cost of assault measures.
The Wisconsin Department of Corrections publishes the total number of (1) completed assaults and (2) injuries from assaults each year for staff (DOC, 2016). To find the rate of staff assaults per year, we take the number of completed assaults and the number of injuries divided by the point in time prison population for that year to arrive at estimated baseline rates for both. The four-year average is 1.2 percent of people in prison complete assaults against staff and .2 percent of people in prison complete assaults that injure staff. For Alternative 1, these numbers are potentially high as the eligible population for the SCSI only includes nonviolent offenders.
Using the baseline rate from the DOC as our point estimate, the percentage point difference for completed assaults is .65 and the percent point difference for injuries is .11. To monetize the impact of this reduction in assault, we use the cost of assault with injury ($36,079) and cost of assault without injury ($3,862) for each estimate, respectively. The estimate for the cost of assault with injury includes losses in productivity and both estimates include quality of life adjustments.
Table Q.2: Misconduct Variables (Staff Assaults)
|arc_staff_base (%)||1.2||WI DOC|
|arc_staff_CBT (%)||.6||Morgan & Flora (2002)|
|ari_staff_base (%)||.2||WI DOC|
|ari_staff_CBT (%)||.09||Morgan & Flora (2002)|
|cost_staff ($)||3,862||Miller et. al (1996)|
|cost_staff_inj ($)||36,079||Miller et. al (1996)|
The following equation is used to calculate the benefits from reduced staff assault:
Where b_m_staff is the quantified benefit, arc_staff_base is the baseline percent of inmates who complete an assault against a staff member, arc_staff_CBT is the percent of inmates who complete an assault against a staff member in the Second Chance Institute, ari_staff_base is the baseline percent of inmates who injure a staff member, ari_staff_CBT is the percent of inmates who injure a staff member in the SCSI, cost_assault is the cost to a victim for assault without injury, and cost_assault_inj is the cost to a victim for assault with injury.
As we expect a different number of CBT completers per alternative and for each individual year (see Appendix F), Table Q.3 estimates the benefits for the first four years of the programming and the expected benefit for all the remaining years once it reaches a steady state. As discussed above, the effect size and the baseline rate of assaults used are both highly uncertain, and for that reason, the estimates in Table Q.3: represent the upper bound for the monetary value of this benefit. In our Monte Carlo (see Appendix U), we vary these benefit from this upper bound to zero to account for the high levels of uncertainty.
Table Q.3 Benefits from Reduced Misconduct
|Year||CBT Completers||Benefits ($)|
Appendix R: Benefits from Increased Lifetime Compensation
The human capital theory states that education increases the economic capabilities of people (Schultz, 1971). Therefore, we assume that graduates of the SCSI and its correctional education programming would have higher earnings capacity throughout their remaining lifetimes. We multiply the estimated return in earnings from an additional year of education by our projection of the average earnings for individuals in each year following release from prison to estimate the lifetime benefits on earnings for graduates from SCSI. We calculate separate estimates of the lifetime benefits in earnings for vocational education graduates and postsecondary education graduates because they engage in education for different lengths of time and therefore have different returns on education.
To estimate the return on education from an additional year of schooling, we use the WSIPP (2017) estimate that an additional year of education increases annual income in each year, equivalent to lifetime earnings, by 10 percent. As the vocational education programming would be six months in length we assume a return on wages of five percent. For postsecondary correctional education, we assume a 20 percent return on wages, as participants are receiving a two-year associatedegree.
To calculate baseline earnings per individual per year, we rely on a method from Boardman et al. (2018) for projecting the compensation of individuals with a high school diploma over their lifetime, from age 18 to age 65. We use the projected earnings pattern for a person with a high school diploma because each participant in the SCSI is required to have already completed a GEDor HSED. Boardman et al. multiply earnings by an adjustment factor to estimate total compensation, account for mortality (decreasing probability of survival with advancing age), and vary compensation in subsequent years based on a worker’s years of experience and the observed pattern of real-world changes in compensation for the average worker with that level ofeducation.
To adapt this model for our purposes, we start with the average earnings for people in their first full-year of release from prison, $7,639. This estimate comes from a report that uses tax data on formal reported earnings for people who were recently released from prison, as identified in a recent quasi-census of the prison population in the United States (Looney & Turner, 2018).18 We also adjust the Boardman et al. method to account for the fact that the average person released from prison in Wisconsin is age 36 (DOC, 2017f).19 After these adjustments, we follow the Boardman et al. approach of using the projected annual earnings for individuals released from prison (age 36 through age 65) and a discount rate of 3.5 percent to calculate a net present value of the projected earnings for the average individual released from prison at the time of their release. As a result, we estimate that the net present value of lifetime earnings for an individual released from prison without any additional education is$210,944.
To calculate our expected benefit for vocational education we take the estimated lifetime earnings for someone released from prison ($210,944) and multiply by 5percent to get a return of
$10,547 per graduate for vocational education. For postsecondary education we take $210,944 and multiply by 20 percent, getting a value of $42,180. These calculations mean that SCSI participants who receive vocational education would make an additional $10,547 over the course of their lifetime and a SCSI participant who receives postsecondary would make an additional $42,180 over the course of their lifetime. We expect that these estimates are an upper bound of the earnings benefits from education as it is likely that people who were formerly incarcerated experience a lower return on education compared to the general population. We vary the benefits from zero to theseestimates in our Monte Carlo (Appendix U: Monte Carlo Equations andEstimates).
- To calculate average earnings per person in year one post-release, we multiply the reported mean earnings conditional on work ($13,889) by the percentage of the sample population who have any reported earnings (55 percent).
- We modify the Boardman et al. estimation in two ways. First, we divide our estimate of the average earnings for a person released from prison by the calibration factor associated with age 36 to come up with a base earnings value for the model that is comparable to the model’s existing base value of the average earnings of a high school graduate at age 24. In essence, this produces a projected lifetime earnings curve for the average individual being released from prison that has the same lifetime earnings pattern as a high school graduate, but is uniformly scaled down (across the lifetime) to account for the lower average earnings of individuals who are sent to and released from prison, such that the projected earnings for a 36-year-old from the adjusted model closely matches the observed earnings for the average individual released from prison as estimated by Looney and Turner (2018). Second, when we calculate the net present value of expected earnings for an individual released from prison, we only include the projected earnings from age 36 to age 65.
Table R.1: Earnings Benefit of SCSI Education Programming
|Estimated Life Earnings ($)||Benefit of Education ($)|
|Formerly Incarcerated – Non-SCSI||210,944||–|
|SCSI Participant – Vocational Education||221,491||10,547|
|SCSI Participant – Postsecondary Education||253,124||42,180|
Table R.2: Return on Education Variables
|lifetime_earnings ($)||210,944||Boardman et. al (2018); Looney & Turner (2018)|
To fully quantify this benefit, we use the following equations:
𝑏_𝑒𝑎𝑟𝑛𝑖𝑛𝑔𝑠_𝑎𝑙𝑡2=(𝑙𝑖𝑓𝑒𝑡𝑖𝑚𝑒_𝑒𝑎𝑟𝑛𝑖𝑛𝑔𝑠∗𝑒𝑎𝑟𝑛𝑖𝑛𝑔𝑠_𝑟𝑒𝑡𝑢𝑟𝑛_𝑝𝑜𝑠𝑡) ∗𝑝𝑜𝑠𝑡_𝑔𝑟𝑎𝑑𝑠 + (𝑙𝑖𝑓𝑒𝑡𝑖𝑚𝑒_𝑒𝑎𝑟𝑛𝑖𝑛𝑔𝑠∗𝑒𝑎𝑟𝑛𝑖𝑛𝑔𝑠_𝑟𝑒𝑡𝑢𝑟𝑛_𝑣𝑜𝑐) ∗𝑣𝑜𝑐_𝑔𝑟𝑎𝑑𝑠
Where b_earnings_alt1 is the quantified benefits under Alternative 1, b_earnings_alt2is the quantified benefit under Alternative 2, voc_grads is the number of graduates from vocational training, post_grads is the number of postsecondary graduates, earnings_return_vocis the return from vocational training, earnings_return_postis the return from postsecondary education, and lifetime_earnings is the baseline lifetime compensation of people who are released from prison.
As we assume a different number of graduates for the first four years before hitting a steady state, we estimate the benefits for years one through four as well as expected benefits each year in the steady state.
Table R.3: Total SCSI Graduate Lifetime Earnings Benefits from SCSI Education.
|Year||Vocational Graduates||Postsecondary Graduates||Benefits ($)|
S.) Non-market Spillovers from Increased Compensation
T.) Turning Effect Sizes into Percentage Point Differences
U.) Variable Distributions for Monte Carlo Simulation
Appendix S: Non-market Spillovers from Increased Compensation
The estimated increases in wages, as calculated in Appendix R: Benefits from Increased Lifetime Compensation, are expected to create external spillover benefits for society, including improved consumption decisions, improved health, and improved child development (Haveman & Wolfe, 1984, 2002). To calculate the value of these additional spillover benefits, we use estimates from WSIPP (2017) on the external benefit return from increased compensation and the estimated increase in total lifetime compensation from SCSI education programming.
To calculate the impact of spillover benefits, WSIPP estimates external benefits as a fraction of total compensation. They document an estimated range of .13 to .42 from a variety of sources and use .37 as the modal value (Boardman et al. 2018). To calculate our impacts, we use this value (.37) multiplied by the increase in total lifetime compensation arising from the SCSI programming (calculated in Appendix R) to determine the spillover benefits.
To fully quantify this benefit, we use the following equation:
𝑏 _ 𝑒𝑎𝑟𝑛𝑖𝑛𝑔𝑠 _ 𝑠𝑝𝑖𝑙𝑙𝑜𝑣𝑒𝑟= 𝑏 _ 𝑒𝑎𝑟𝑛𝑖𝑛𝑔𝑠∗ 𝑒𝑑𝑢 _ 𝑛𝑜𝑛𝑚𝑎𝑟𝑘𝑒𝑡
Where b_earnings_spillover_1 are the spillover benefits under Alternative 1, b_earnings is the total benefit from increased earnings, and edu_nonmarket is the return on increased compensation from external benefits.
As we expect a different number of graduates from each educational programming per alternative and for each individual year, the following table estimates the benefits for the first four years of the programming and the expected benefit for each of the remaining years once it reaches a steady state in year five. To calculate this number, we take the estimated values from Table R.3: Total SCSI Graduate Lifetime Earnings Benefits from SCSI Education, which is based on number of graduates per year, multiplied by .37 to get the total benefits for each year. Table S.1: Lifetime Non-Market Benefits of SCSI Education Programming shows the full results.
Table S.1: Lifetime Non-Market Benefits of SCSI Education Programming
|Year||Lifetime Earnings Increase ($)||Expected Benefits ($)|
Appendix T: Turning Effect Sizes into Percentage Point Differences
Meta-analyses and single-study evaluations of correctional education and other programs often report impacts in terms of standardized effect sizes or odds ratios. Cost-benefit analyses rely on estimates of the percentage point differences in outcomes between individuals participating in the program versus the alternative. This appendix uses vocational correctional education as an example of how we convert effect sizes or odds ratios from the program evaluation literature into percentage point differences for the purposes of cost-benefit analysis.
We use an estimate that vocational correctional education has an effect size of .167 on crime (WSIPP, 2016d). To turn this into a percentage point impact, we first convert the effect sizereported by WSIPP to an odds ratio using theformula:
𝑂𝑅 =𝑒* ((𝜋∗𝐸𝑆)/√3)
where OR = odds ratio and ES = standard mean difference effect size (Hasselblad & Hedges, 1995). In this case, an effect size of .167 equals an odds ratio of 1.35. Because the odds ratio of 1.35 indicates how correctional education increases the odds of not being incarcerated, we have to take the inverse to find the change in the odds of committing a crime (1/1.354 = .739).
Next, we convert the odds ratio to a relative risk ratio, using the formula:
𝑅𝑅 = 𝑂𝑅/(1−𝑃0) +(𝑃0∗ 𝑂𝑅)
where RR = relative risk and P0 = the incidence of the outcome of interest in the nontreated group. Because the effect sizes and odds ratios are taken from meta-analyses that do not report estimates for the average values of P0 in the included studies, we use our estimates of the baselines rates for the outcomes of interest in Wisconsin as the values for P0. In this case, we use the inverted odds ratio of .739 and the current three-year reincarceration rate in Wisconsin, .375, to calculate a relative risk of 0.819 (WI DOC, 2018c).
Then, we use the relative risk to calculate the reincarceration rate for the group graduating from SCSI after completing vocational education, given the baseline reincarceration rate, with the formula:
𝑃𝑆𝐶𝑆𝐼 = (𝑅𝑅∗𝑃0)/1+(𝑅𝑅∗𝑃0) −𝑃0
In this case, the formula would take on the values (.819 * .375) / (1 + (.819 x .375) – .375) = 0.329. Thus, the predicted reincarceration rate for the population receiving vocational correctional education is 32.9 percent, compared with the baseline rate in Wisconsin of 37.5 percent, a difference of 4.6 percentage points.
Appendix U: Variable Distributions for Monte Carlo Simulation
For our Monte Carlo simulation, we defined distributions of our uncertain parameters. We use different methods of defining these parameters depending on our estimates. For costs, we use multiple point estimates from different sources in the literature and create a distribution around these estimates. We find three different estimates of marginal annual costs for vocational education: one from WSIPP (2016), and two from Wisconsin Department of Corrections (DOC, 2016a; RYOCF, 2018). One of the Wisconsin DOC estimates is a general estimate for the state, and one is specific to the Racine facility. We use an asymmetric triangular distribution with the Racine-specific number ($2,766) as our mode because this value is the most relevant for our project, and the WSIPP and general DOC estimates as lower and upper bounds, respectively. Similarly, we find three estimates of the marginal annual costs for postsecondary education from a combination of WSIPP and DOC. However, none of our estimates are Racine-specific, so we assume all values are equally likely and we use a uniform distribution between the low and high values.
We account for uncertainty in the upfront construction costs by taking our point estimate as a lower bound and varying the cost upward by 20 percent. We use a uniform distribution between these two values. For the costs of crime (as part of our calculations of crime and reincarceration benefits), we calculate values to use as point estimates and then subtract/add twenty percent to these point estimates and use these values as a minimum and maximum for a uniform distribution. We use the same method to calculate a minimum and maximum value for a uniform distribution for the cost of substance use and assaults within prison.
In situations in which we use a point estimate of an effect size or odds ratio from the literature and converted this to a percentage point difference, we assume a normal distribution with the standard deviation reported in the literature. This applies to our estimates of all effect sizes of various programs on crime and reincarceration, the effect size of programming on suicide, and the odds ratio for the effect of programming on reduced substance use. Refer to appendices P, M, and O, respectively.
When calculating the adjusted VSL, we use a minimum value that is based on the income of people who are in prison and a maximum value of the full VSL reported in Boardman et al. (2018). For our point estimate, we use an adjusted VSL that is based on the average annual cost of incarcerating one person. We assume that this point estimate has a higher likelihood of occurrence than values towards the minimum and maximum values, so we use an asymmetric triangular distribution. Refer to Appendix M for more details on the calculations of these values.
For the misconduct benefit, we vary the effect size from zero to the reported point estimate of 0.43 as we are less certain about the number used in our calculations. We are uncertain about the effect size for multiple reasons: it does not directly measure violence within prison, the sample sizes used in the literature are small, and our population is not fully representative of the entire prison population (See Appendix Q for more information about these point estimates and calculations.)
For earnings return values and nonmarket compensation benefits, we use a similar approach as the return to education for the SCSI graduates is highly uncertain. We use a uniform distribution from zero to the point estimates. (See Appendix R for more information about these point estimates and calculations.) For nonmarket spillover benefits, we use a symmetric triangle distribution because that is the distribution used by our source (WSIPP). (See Appendix S for more details on these calculations.)
Lastly, we calculate an asymmetric triangular distribution for the marginal savings of earned release using a point estimate from the literature. (Refer to Appendix N for more detailed calculations, and see Table U.1: Monte Carlo Variables for a summary of variables included in the Monte Carlo simulation.)
|Table U.1: Monte Carlo Ranges
[name in Stata code]
|Vocational Training ($)
|Postsecondary Education ($)
|Upfront Cost – Construction ($)
|2.7 million||2.7 million||3.24 million||Uniform||G|
|Crime and Reincarceration Benefits|
|Cost – Nonviolent Crime ($)
|Cost – Crime ($)
|Effect Size (Voc.)
|Effect Size (Post.)
|Effect Size (CBT)
|Reduced Suicide Benefits|
|Adjusted VSL ($)
|2.66 million||13,452||11. 74 million||Asymmetric Triangle||M|
|Reduced Substance Use Benefits|
|Cost of Substance Use ($)
|Reduced Misconduct Benefits|
|Cost of Assault-Inmates ($)
|822||13,452||11. 74 million||Uniform||Q|
|Cost of Assault-Staff ($)
|Cost of Assault-Staff Injury ($)
|Earned Release Benefits|
|Marginal Savings ($)
|Earnings Return (Voc.)
|Earnings Return (Post.)
|Nonmarket Compensation Benefits|
V.) Monte Carlo Stata Code
W.) Earnings Sensitivity Analysis
Appendix V. Monte Carlo Stata Code
- Contact Thompson Center: firstname.lastname@example.org
Appendix W: Earnings Sensitivity Analysis
There is a high degree of uncertainty surrounding the benefits of increased lifetime compensation, and these benefits are larger than all other benefit categories. For that reason, we complete several alternative calculations to further explore the impact of this benefit on the overall benefits of the SCSI.
In the original Monte Carlo simulation, we randomly draw compensation effects from a uniform distribution spanning zero (no effect) to 5 percent for six months of vocational education or 20 percent for two years of postsecondary education, based on WSIPP’s finding that a 10 percent return for each additional year of education is a good rule of thumb for the average population. By drawing from a uniform distribution that falls to zero impact, we allow for the effect of education to be a lot less for individuals being released from prison, given this population’s much lower levels of employment and earnings before and after prison. The uniform distribution from zero to 5 or 20 percent means that the mean values in Table 2 for the benefit category of increased lifetime compensation are equivalent to the benefits from assuming 2.5 and 10 percent rates of increased lifetime compensation from vocational or postsecondary education, respectively.
However, even this reduced effect might be too optimistic. Table 4 compares the mean net present value and percent of trials with positive net benefits for both Alternatives 1 and 2 fromthe original Monte Carlo analysis and the corresponding values under different earningsassumptions.
In the first alternative, we cut the return to education to half that of the original Monte Carle simulation, in effect assuming mean lifetime compensation increases of 1.25 and 5 percent from vocational and postsecondary education, respectively. Under this scenario for Alternative 1, we find a mean present value of $12.2 million, with nearly 72 percent of trials resulting in a positive value of net benefits. For Alternative 2, we find a mean present value of $29.9 million, with over 98 percent of trials resulting in a positive value of net benefits. If we instead cut the return to education by a fifth relative to the original analysis, which is equivalent to respective lifetime compensation increases of 0.5 percent and 2 percent for vocational and postsecondary education, we find a mean present value of -$4.6 million for Alternative 1, with only 30 percent of trials resulting in a positive value of net benefits. For Alternative 2, we find a mean present value of $8.7 million in net benefits, with nearly 90 percent of trials resulting in a positive value of net benefits.
Alternatively, we model the impact of SCSI increasing the likelihood of employment by about 2 percentage points for just the first year post-release, based on the results of a RAND meta- analysis of the available evidence on vocational correctional education’s impact on short-term employment outcomes.20 Under that scenario for Alternative 1, we find a mean present value of -$12.7 million, with zero trials resulting in a positive value of net benefits. For Alternative 2, we find a mean present value of -$3.1 million, with less than 21 percent of trials resulting in a positive net benefit. If we assume correctional education and the SCSI program has no impact on post-release employment, then Alternative 1 would have a mean present value of -$15.8 million and zero trials resulting in positive net benefits, and Alternative 2 would have a mean present value of -$5.5 million and just over six percent of trials resulting in a positive net benefit.
Bozick et al. (2018) report an odds ratio of 1.22 for vocational education, meaning that individuals who receive correctional education have a higher rate of employment in the 12 months post-release. We make the conservative assumption that postsecondary education has the same effect, and convert both into percentage point increases in employment using the approach outlined in Appendix T. We assume baseline employment rate and mean earnings within the first year of release of 55 percent and $13,889, respectively, based on Looney and Turner (2018). To calculate the monetized benefit from increased employment rates, we multiply the percentage point difference in employed by the number of graduates each year and the mean earnings for those within their first year of release from prison.
Table W.1: Net Benefits under Alterative Earnings Assumptions Over 50 Year Lifetime
|Alternative 1||Alternative 2|
|Earnings Assumption||Mean ($ millions)||Percent of Trials with Positive Net Benefits||Mean ($ millions)||Percent of Trials with Positive Net Benefits|
|Original Monte Carlo||40.3||85||65.3||10021|
|Half of original||12.2||72||29.9||98|
|One-fifth of original||-4.6||30||8.7||90|
|One-year employment increase||-12.7||0||-3.1||21|
Andoh, S., Wysock, R., Provencher, A., Hill, B., Clark, A., Lize, S., Dube, S., (2013). Connecticut Results First Model Summary. Retrieved from https://www.cga.ct.gov/app/tfs%5C20141215_Juvenile%20Justice%20Policy%20and%20Oversight%20Committee%5C20141016/ResultsFirstpresentationJJPOC%20PRINT.pdf.
Andrews, A. & Bonta, J. (1994). The Psychology of Criminal Conduct. Cincinnati, OH: Anderson.
Bannon, A., Nagrecha, M., & Diller, R. (2010). Criminal justice debt: a barrier to reentry. New York: Brennan Center for Justice at New York University School of Law.
Baro, A. L. (1999). Effects of a cognitive restructuring program on inmate institutional behavior. Criminal Justice and Behavior,26(4), 466-484.
Beck, R., & Fernandez, E. (1998). Cognitive-behavioral therapy in the treatment of anger: A meta- analysis. Cognitive therapy and research, 22(1), 63-74.
Beck, M. (June, 2016). Median pay for state’s prison staff dropped nearly 13% since 2009.
Wisconsin State Journal. Retrieved from https://madison.com/wsj/news/local/govt-and-politics/median-pay-for-state-s-prison-staff-dropped-nearly-percent/article_89605fb7-9fab-5e9f-a692-1e1c2e4beab6.html
Binswanger, I. A., Stern, M. F., Deyo, R. A., Heagerty, P. J., Cheadle, A., Elmore, J. G., & Koepsell,
- D. (2007). Release from prison—a high risk of death for former inmates. New England Journal of Medicine, 356(2), 157-165.
Blincoe, L. J., Miller, T. R., Zaloshnja, E., & Lawrence, B. A. (2015). The economic and societal impact of motor vehicle crashes, 2010. (Revised) (Report No. DOT HS 812 013).
Washington, DC: National Highway Traffic Safety Administration.
Boardman, A. E., Greenberg, H. D, Vining R. A., Weimer, D. L., (2018). Cost-Benefit Analysis: Concepts and Practice. New York, NY: Cambridge University Press.
Bozick, R., Steele, J., Davis, L. M., Turner, S. (2018). Evaluating the
effectiveness of correctional education: A meta-analysis of programs that provide education to incarcerated adults. RAND Corporation.
Brazzell, D., Crayton, A., Mukamal, D. A., Solomon, A. L., & Lindahl, N. (2009). From the Classroom to the Community: Exploring the Role of Education during Incarceration and Reentry. Urban Institute (NJ1).
Bronson, J., Stroop, J., Zimmer, S., & Berzofsky, M. (2017). Drug use, dependence, and abuse among state prisoners and jail inmates, 2007-2009. Washington, DC: United States Department of Justice, Office of Juvenile Justice and Delinquency Prevention.
Brown, G. K., Ten Have, T., Henriques, G. R., Xie, S. X., Hollander, J. E., & Beck, A. T. (2005).
Cognitive therapy for the prevention of suicide attempts: a randomized controlled trial. Jama, 294(5), 563-570.
Cornelius, T. (2017). Wisconsin’s Prison Price Tag: Still Growing.
Wisconsin Budget Project. Retrieved from http://www.wisconsinbudgetproject.org/wisconsins-prison-price-tag-still-growing
Cohen, M. A., & Piquero, A. R. (2009). New evidence on the monetary value of saving a high risk youth. Journal of Quantitative Criminology, 25(1), 25-49.
Congressional Budget Office. (2012). How the Supply of Labor Responds to Changes in Fiscal Policy. Washington, DC: Congressional Budget Office.
Couloute, L., & Kopf, D. (2018). Out of Prison & Out of Work. Prison Policy Initiative. Retrieved from https://www.prisonpolicy.org/reports/outofwork.html.
Delaney, R. E., Subramanian, R., & Patrick, F. (2016). Making the grade: Developing quality postsecondary education programs in prison. Vera Institute of Justice.
De Leon, G., Melnick, G., Thomas, G., Kressel, D., & Wexler, H. K. (2000). Motivation for treatment in a prison-based therapeutic community. The American journal of drug and alcohol abuse, 26(1), 33-46.
DeVuono-Powell, S., Schweidler, C., Walters, A., & Zohrabi, A. (2015). Who pays? The true cost of incarceration on families. Oakland, CA: Ella Baker Center.
Dietz, E. F., O’Connell, D. J., & Scarpitti, F. R. (2003). Therapeutic communities and prison management: An examination of the effects of operating an in-prison therapeutic community on levels of institutional disorder. International Journal of Offender Therapy and Comparative Criminology, 47(2), 210-223.
Di Placido, C., Simon, T. L., Witte, T. D., Gu, D., & Wong, S. C. (2006). Treatment of gang members can reduce recidivism and institutional misconduct. Law and Human Behavior, 30(1), 93-114.
Doleac, J. L. ( 2018). Strategies to Productively Reincorporate the Formerly-Incarcerated into Communities: A Review of the Literature. (IZA Institute of Labor Economics Discussion Paper No. No. 11646). Retrieved from https://www.iza.org/publications/dp/11646/strategies-to-productively-reincorporate-the-formerly-incarcerated-into-communities-a-review-of-the-literature.
Duran, L., Plotkin, M., Potter, P., & Rosen, H. (2013). Integrated Reentry and Employment Strategies: Reducing Recidivism and Promoting Job Readiness. The Council of State Governments: Justice Center. Retrieved from http://csgjusticecenter.org/wp-content/uploads/2013/09/Final.Reentry-and-Employment.pp_.pdf
Ettner, S. L., Huang, D., Evans, E., Rose Ash, D., Hardy, M., Jourabchi, M., & Hser, Y. I. (2006). Benefit–cost in the California Treatment Outcome Project: Does substance abusetreatment “pay for itself”? Health Services Research, 41(1),192-213.
Festen, M., Waul, M., Soloman, A., Travis, J. (2002) From Prison to Home: The Effect of Incarceration on Children, Families and Communities. Urban Institute – Prepared for U.S. Department of Health and Human Services.
Fifield, J. (2016). States Face Dire Shortage of Prison Guards. Pew Trusts. Retrieved from https://www.pewtrusts.org/en/research-and-analysis/blogs/stateline/2016/03/01/many-states-face-dire-shortage-of-prison-guards.
Gendreau, P. (1996). The principles of effective intervention with offenders. In (Alan T. Harland, ed.), Choosing Correctional Options That Work: Defining the Demand and Evaluating the Supply. Thousand Oaks, CA: Sage.
Gerber, J., & Fritsch, E. J. (1995). Adult academic and vocational correctional education programs: A review of recent research. Journal of Offender Rehabilitation, 22(1-2), 119-142.
Hammitt, J. K., & Robinson, L. A. (2011). The income elasticity of the value per statistical life: transferring estimates between high and low income populations. Journal of Benefit-Cost Analysis, 2(1), 1-29.
Haveman, R. H., & Wolfe, B. L. (1984). Schooling and economic well-being: The role of nonmarket effects. Journal of human Resources, 377-407.
Henrichson, C. (2015). The Price of Prisons. Senate Presidents’ Forum. Retrieved from https://www.senpf.org/price-of-prisons.html.
Hiller, M. L., Knight, K., & Simpson, D. D. (1999). Risk factors that predict dropout from corrections-based treatment for drug abuse. The Prison Journal, 79(4), 411-430.
Huebner, B. M., & Cobbina, J. (2007). The effect of drug use, drug treatment participation, and treatment completion on probationer recidivism. Journal of Drug Issues, 37(3), 619-641.
Illinois Sentencing Policy Advisory Council. (2018). 2018 Update: Dynamic Marginal Costs in Fiscal Impact Analyses. Retrieved from http://www.icjia.state.il.us/spac/pdf/Dynamic_Marginal_Costs_2018_Update.pdf.
Inciardi, J. A., Martin, S. S., & Butzin, C. A. (2004). Five-year outcomes of therapeutic community treatment of drug-involved offenders after release from prison. Crime & Delinquency, 50(1), 88-107.
Jensen, E. L., & Kane, S. L. (2012). The effects of therapeutic community on recidivism up to four years after release from prison: A multisite study. Criminal Justice and Behavior, 39(8), 1075-1087.
Johnson, Shawn. (March 2018). Groups Fight Plan To Build New Prison In Wisconsin. Wisconsin Public Radio. Retrieved from: https://www.wpr.org/groups-fight-plan-build-new-prison-wisconsinhttps://www.wpr.org/groups-fight-plan-build-new-prison-wisconsin
Kilmer, B., Everingham, S., Caulkins, J., Midgette, G., Pacula, R., Reuter, P., Burns, R., Han, B., Lundberg, R. (2014). What America’s Users Spend on Illegal Drugs: 2000-2010. RAND Corporation.
King, G., & Nielsen, R. (in press). Why Propensity Scores Should Not Be Used for Matching.
Kniesner, T. J., Viscusi, W. K., & Ziliak, J. P. (2010). Policy relevant heterogeneity in the value of statistical life: New evidence from panel data quantile regressions. Journal of Risk and Uncertainty, 40(1), 15-31.
Lahm, K. F. (2009). Educational participation and inmate misconduct. Journal of Offender Rehabilitation, 48(1), 37-52.
Lambert, E. G., Barton-Bellessa, S. M., & Hogan, N. L. (2015). The consequences of emotional burnout among correctional staff. Sage Open, 5(2).
Latessa, E. (2011). Why the risk and needs principles are relevant to correctional programs (even to employment programs. Criminology & Public Policy, 10(4), 973-977.
Leasure, P. & Martin, T. (2017). Criminal records and housing: an experimental study. Journal of Experimental Criminology, 13: 527–535.
Leasure, P. & Stevens-Andersen, T. (2016). The Effectiveness of Certificates of Relief as Collateral Consequence Relief Mechanisms: An Experimental Study. Yale Law and Policy Review Inter Alia.
Looney, A. & Turner, N. (2018). Work and opportunity before and after incarceration. The Brookings Institution.
Mason, K. (2016). Milwaukee Tech College To Offer Pell Grants to Prisoners. Wisconsin Public Radio. Retrieved from https://www.wpr.org/milwaukee-tech-college-offer-pell-grants-prisoners.
McCollister, K. E., French, M. T., & Fang, H. (2010). The cost of crime to society: new crime- specific estimates for policy and program evaluation. Drug and alcohol dependence, 108(1- 2), 98-109.
McCollister, K., Yang, X., Sayed, B., French, M. T., Leff, J. A., & Schackman, B. R. (2017).
Monetary conversion factors for economic evaluations of substance use disorders. Journal of substance abuse treatment, 81, 25-34.
Mead & Hunt Inc. (2009).Ten-Year Correctional Facility System Development Plan. (Division of State Facilities Project No. 06F1F). Retrieved from https://doa.wi.gov/DFDM_Documents/State-Building-Program/DOC-Dev-Plan.pdf.
Minor, K. I., Wells, J. B., Angel, E., & Matz, A. K. (2011). Predictors of early job turnover among juvenile correctional facility staff. Criminal Justice Review, 36(1), 58-75.
Miller, T. R., Cohen, M. A., & Wiersema, B. (1996). Victim costs and consequences: A new look.
Washington, DC: US Department of Justice, Office of Justice Programs, National Institute of Justice.
Miller, D. P., & Mincy, R. B. (2012). Falling further behind? Child support arrears and fathers’ labor force participation. Social Service Review, 86(4), 604–635.
Mitchell, O., Wilson, D., & Layton MacKenzie, D. (2018). The effectiveness of incarceration-based drug treatment on criminal behavior. Society for Social Medicine Annual Scientific Meeting & European Congress of Epidemiology.
Morgan, R. D., & Flora, D. B. (2002). Group psychotherapy with incarcerated offenders: A research synthesis. Group Dynamics: Theory, Research, and Practice, 6(3), 203-218.
N/A. (2016). MATC Chosen for U.S. Department of Education’s New Second Chance Pell Pilot Program. MATC News. Retrieved fromhttps://www.matc.edu/matc_news/MATCSelectedforUSDepartmentofEducationProgram.cfm.
Noonan, M. E., & Bureau of Justice Statistics. (2016). Mortality in State Prisons, 2001-2014— Statistical Tables. Bureau of Justice Statistics, U.S. Department of Justice.
Pearson, F. S., Lipton, D. S. (1999). A Meta-Analytic Review of the Effectiveness of Corrections- Based Treatments for Drug Abuse. The Prison Journal, 79(4).
Pelissier, B. (2004). Gender differences in substance use treatment entry and retention among prisoners with substance use histories. American Journal of Public Health, 94(8), 1418-1424.
Pompoco, A., Wooldredge, J., Lugo, M., Sullivan, C., & Latessa, E. J. (2017). Reducing inmate misconduct and prison returns with facility education programs. Criminology & Public Policy, 16(2), 515-547.
Prendergast, M., Farabee, D., & Cartier, J. (2001). The impact of in-prison therapeutic community programs on prison management. Journal of Offender Rehabilitation, 32(3), 63-78.
Racine Youthful Offender Correctional Facility. (2018). Annual Report July 1, 2017 – June 30, 2018.Retrieved from https://doc.wi.gov/Documents/OffenderInformation/AdultInstitutions/RYOCFAnnualReport.pdf.
Robinson, L. A., & Hammitt, J. K. (2015). Research synthesis and the value per statistical life. Risk Analysis, 35(6), 1086-1100.
Rowland, D. (2003). Demographics Methods and Concepts. Oxford University Press. Schanzenbach, D. W., Nunn, R., Bauer, L., Breitwieser, A., Mumford, M., & Nantz, G. (2016).
Twelve facts about incarceration and prisoner reentry. Hamilton Project.
Schultz, T. W. (1971). Investment in human capital: The role of education and research.
New York: The Free Press.
Semega, J., Fontenot, K., Kollar, M., & United States Census Bureau. (2018). Income and Poverty in the United States: 2017 – Data United States Census Bureau. Retreived from https://www.census.gov/data/tables/2018/demo/income-poverty/p60-263.html.
Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and Quasi-Experimental Designs for Generalized Causal Inference (2nd ed.) Boston, MA: Houghton-Mifflin Company.
Shafroth, A., Seligman, D., Kornya, A., Taylor R., Allen N. (2016). Confronting Criminal Justice Debt: A Guide for Litigation. National Consumer Law Center. Retrieved from http://cjpp.law.harvard.edu/assets/CJD-LitigationGuide.pdf.
Siani, M. (2009). A Meta-analysis of the Psychological Treatment of Anger: Developing Guidelines for Evidence-Based Practice. Journal of American Academy Psychiatry Law, 37, 473-488.
Speckhard Pasque, L. (2017).Outside Chance: Wisconsin has stepped up job training in prison, but critics say it’s not enough. The Cap Times. Retrieved from https://madison.com/ct/news/local/govt-and-politics/outside-chance-wisconsin-has-stepped-up-job-training-in-prison/article_8f03a9c3-8d7e-5b93-af5e-3622e6db4dec.html.
State of Wisconsin Building Commission (2017). 2017-2019 State of Wisconsin Capital Budget Recommendations. Retrieved from https://doa.wi.gov/DFDM_Documents/State-Building-Program/CapBud/2017-19%20Capital%20Budget%20SBC%20Recommendations.pdf.
Stein, Jason. (June 26, 2017). More state workers leaving for other jobs: Stagnant wages driving employees to private sector. Milwaukee Journal Sentinel.
Tatar, J. R. (2015). Wisconsin Results First Initiative Adult Criminal Justice Program Inventory: Initial report. Madison, WI: State of Wisconsin Results First Initiative.
Tarrier, N., Taylor, K., & Gooding, P. (2008). Cognitive-behavioral interventions to reduce suicide behavior: a systematic review and meta-analysis. Behavior modification, 32(1), 77-108.
Thompson, T. (2018). Help solve Wisconsin’s coming labor shortage by rehabilitating prisoners.
Milwaukee Journal Sentinel. Retrieved from https://www.jsonline.com/story/opinion/contributors/2018/04/20/tommy-thompson-solve-labor-shortage-rehabilitate-prisoners/529305002/.
U.S. Department of Labor, Bureau of Labor Statistics. (2016). Number, incidence rate, and median days away from work for nonfatal occupational injuries and illnesses involving days away from work by selected worker occupation and ownership, 2015. Retrieved fromhttps://www.bls.gov/news.release/osh2.t03.htm.
U.S. Department of Labor, Bureau of Labor Statistics. (2018a). CPI Inflation Calculator. Retrieved fromhttps://www.bls.gov/data/inflation_calculator.htm.
U.S. Department of Labor, Bureau of Labor Statistics. (2018b). May 2017 State Occupational Employment and Wage Estimates Wisconsin – Data Tables. Retrieved fromhttps://www.bls.gov/oes/current/oes_wi.htm#21-0000.
U.S. Department of Labor, Bureau of Labor Statistics. (2018c). State and local government, by occupational and industry group. Retrieved fromhttps://www.bls.gov/news.release/ecec.t04.htm.
U.S. Department of Transportation, National Highway Traffic Safety Administration. (2015). The Economic and Societal Impact of Motor Vehicle Crashes, 2010 (Revised) (Publication No. DOT HS 812 013). Retrieved fromhttps://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/812013.
U.S. Federal Bureau of Investigation, Criminal Justice Information Services Division. (2016). Table 22: Arrests by State, 2016. Retrieved fromhttps://ucr.fbi.gov/crime-in-the-u.s/2016/crime-in-the-u.s.-2016/topic-pages/tables/table-22.
Washington State Institute for Public Policy. (2016a). Cost-Benefit Analysis: Cognitive behavioral therapy (for individuals classified as high- or moderate-risk). Retrieved from http://www.wsipp.wa.gov/BenefitCost/Program/10.
Washington State Institute for Public Policy. (2016b). Cost-Benefit Analysis: Correctional Education (Post Secondary Education). Retrieved fromhttp://www.wsipp.wa.gov/BenefitCost/Program/735.
Washington State Institute for Public Policy. (2016c). Cost-Benefit Analysis: Therapeutic Communities (during incarceration) for Individuals with Substance Use Disorders. Retrieved from http://www.wsipp.wa.gov/BenefitCost/Program/187.
Washington State Institute for Public Policy. (2016d). Cost-Benefit Analysis: Vocational Education in Prison. Retrieved from http://www.wsipp.wa.gov/BenefitCost/Program/6.
Washington State Institute for Public Policy. (2017). Benefit-Cost Technical Documentation.
Welsh, W. N., & Zajac, G. (2013). A multisite evaluation of prison-based drug treatment: Four-year follow-up results. The Prison Journal, 93(3), 251-271.
Welsh, W. N., Zajac, G., & Bucklen, K. B. (2014). For whom does prison-based drug treatment work? Results from a randomized experiment. Journal of Experimental Criminology, 10(2), 151-177.
Western, B., & Pettit, B. (2010). Incarceration & social inequality. Daedalus, 139(3), 8-19.
What Works for Health. (2018). Cognitive-Behavioral Therapy (CBT) for Offenders. Madison, WI: Population Health Institute, School of Medicine and Public Health, University of Wisconsin.
Winterfield, L., Coggeshall, M., Burke-Storer, M., Correa, V., & Tidd, S. (2009). The Effects of Postsecondary Correctional Education:” Final Report”. Urban Institute (NJ1).
Wisconsin Department of Corrections. (n.d.a.) Council On Offender Reentry 2018 Annual Report. Retrieved from https://doc.wi.gov/Documents/AboutDOC/Reentry/2018CouncilOffenderReentryAnnualReport.pdf.
Wisconsin Department of Corrections. (n.d.b.) Racine Correctional Institution 2018 AnnualReport. Retrieved from https://doc.wi.gov/Documents/OffenderInformation/AdultInstitutions/RCIAnnualReport.pdf.
Wisconsin Department of Corrections. (n.d.c.) Recidivism and Reincarceration. Retrieved from https://doc.wi.gov/Pages/DataResearch/RecidivismReincarceration.aspx.
Wisconsin Department of Corrections. (n.d.d.) Staff Assaults and Injuries. Retrieved from https://doc.wi.gov/DataResearch/DataAndReports/StaffAssaultInjuryReport.pdf.
Wisconsin Department of Corrections. (2016b). Council on Offender Reentry 2016 Annual Report. Retrieved from https://doc.wi.gov/Documents/AboutDOC/Reentry/2016CouncilOffenderReentryAnnualReport.pdf.
Wisconsin Department of Corrections. (2016c). Division of Adult Institutions Correctional Facilities. Retrieved from https://doc.wi.gov/Documents/OffenderInformation/AdultInstitutions/DAIFacilities.pdf.
Wisconsin Department of Corrections. (2016d). Reentry at a Glance. Retrieved from https://doc.wi.gov/Documents/AboutDOC/Reentry/161001ReentryAtGlance.pdf.
Wisconsin Department of Corrections. (2017a). 2015-17 Biennial Report. Retrieved from https://doa.wi.gov/budget/SBO/2015-17%20410%20DOC%20Biennial%20Report.pdf.
Wisconsin Department of Corrections. (2017b). Becky Young Community Corrections Recidivism Reduction Fiscal Year 2017 Report. Retrieved from https://doc.wi.gov/Documents/AboutDOC/Reentry/2017BeckyYoungReport.pdf.
Wisconsin Department of Corrections. (2017c). Data Trend Series Prison Releases: 2000-2016. Retreieved from https://doc.wi.gov/DataResearch/InteractiveDashboards/DAIReleases2000to2016.pdf.
Wisconsin Department of Corrections. (2017d). Opportunities and Options Resource Guide.
Wisconsin Department of Corrections. (2017e). Oshkosh Correctional Institution. Retrieved from https://doc.wi.gov/DataResearch/DataAndReports/OSCIInstitutionalFactSheet.pdf.
Wisconsin Department of Corrections. (2017f). Prairie Du Chien CorrectionalInstitution.
Wisconsin Department of Corrections. (2017g). Prison Releases 2000-2016. Retrieved from https://doc.wi.gov/DataResearch/InteractiveDashboards/DAIReleases2000to2016.pdf.
Wisconsin Department of Corrections. (2017h). Racine Correctional Institution. Retrieved from https://doc.wi.gov/DataResearch/DataAndReports/RCIInstitutionalFactSheet.pdf.
Wisconsin Department of Corrections. (2017i). Racine Youthful Offender Correctional Facility. Retrieved from https://doc.wi.gov/DataResearch/DataAndReports/RYOCFInstitutionalFactSheet.pdf.
Wisconsin Department of Corrections. (2018a). 2018-2019 Reentry Business Plan. Retrieved from https://doc.wi.gov/Documents/AboutDOC/Reentry/ReentryBusinessPlan.pdf.
Wisconsin Department of Corrections. (2018c). Becky Young Community Corrections Recidivism Reduction Fiscal Year 2018 Report. Retrieved from https://doc.wi.gov/Documents/AboutDOC/Reentry/2018BeckyYoungReport.pdf.
Wisconsin Department of Corrections. (2018d). Inmate Profile 2017. Retrieved from https://doc.wi.gov/DataResearch/DataAndReports/InmateProfile.pdf.
Wisconsin Jobs. (2018). Employment Opportunities at the Department of Corrections. Retrieved from http://wisc.jobs/public/job_search_list.asp?advsrch=1
Wisconsin Legislative Fiscal Bureau. (n.d..) Department of Corrections Budget Summary. Retrieved from https://docs.legis.wisconsin.gov/misc/lfb/budget/2017_19_biennal_budget/034_comparative_summary_of_provisions_2017_act_59_by_agency/corrections.pdf.
Wisconsin Legislative Fiscal Bureau. (2017). Windows to Work and Vocational Training (Corrections — Adult Corrections) Report prepared for the Joint Committee on Finance. Retrieved from https://docs.legis.wisconsin.gov/misc/lfb/budget/2017_19_biennal_budget/050_budget_papers/237_corrections_adult_corrections_windows_to_work_and_vocational_training.pdf
Wisconsin Results First Initiative. (June, 2015). Wisconsin Results First Initiative Adult Criminal Justice Program Inventory: Initial Report. Retrieved from https://www.wicourts.gov/courts/programs/docs/ejs_crimjustprograminventory.pdf.
Wolfe, B. L., & Haveman, R. H. (2002). Social and nonmarket benefits from education in an advanced economy. Conference series-federal reserve bank of Boston (Vol. 47, pp. 97-131).
Wolff, N., & Shi, J. (2009). Contextualization of physical and sexual assault in male prisons: Incidents and their aftermath. Journal of correctional health care, 15(1), 58-77.
Wolff, N., Shi, J., & Siegel, J. A. (2009). Patterns of victimization among male and female inmates: Evidence of an enduring legacy. Violence and victims, 24(4), 469.
Zhang, S. X., Roberts, R. E., & McCollister, K. E. (2009). An economic analysis of the in-prison therapeutic community model on prison management costs. Journal of Criminal Justice, 37(4), 388-395.