Subject Area

Industrial/Manufacturing Engineering, Statistics

Abstract

A problem faced by the United States is the ever increasing prison population. There are inmates serving long sentences, new inmates being sentenced for the first time, and those who have previously served prison sentences that reoffend. The third group and the reduction of recidivism are the focus this dissertation. It is estimated that over 80% of inmates released from prison will reoffend within the next ten years. Is there an optimal sentence length that reduces that chance of an ex-convict reoffending? Are there programs or opportunities that some inmates have while incarcerated that reduce the probability they will return to prison? Prison units in states such as Texas can afford inmates the opportunity to receive a high school education, job training, and sometimes work in positions that allow the inmate to acquire career skills for reentry into society. A few of these programs, such as the Texas Corrections Industries and work release, can earn money for not only the inmate but the state. Others help fulfill societal needs such as training service animals and picking up trash on the road side; however, many programs do not benefit the state in a direct monetary sense but can benefit the state by reducing the chance that inmates will reoffend once they are released. A general hypothesis for this research is that optimized sentencing, and spending on education and training of inmates will significantly reduce the overall cost to society. For the purpose of this dissertation, the term "recidivism" refers to re-incarceration after release from prison. This dissertation begins by comparing methods to predict recidivism including logistic regression and classification trees. Then, the second stage uses those results to minimize the overall cost of incarceration considering sentence lengths, opportunities provided to the inmate during their time in prison, and expected cost of possible future crimes. If education or training lowers the lifetime cost, then the final step assigns them to serve in one of the units which has the suggested program. Empirical analysis on the data resulted in logistic regression being one of the most accurate predictors of recidivism. When assessed on the test data subset the accuracy was near 75%. Two hundred thirteen of the inmate cohorts were sentenced to some education, the model predicts that education will save the Texas Department of Corrections over $1,585,000,000 over time.

Degree Date

Summer 8-6-2024

Document Type

Dissertation

Degree Name

Ph.D.

Department

Operations Research and Engineering Management

Advisor

Dr. Eli Olinick

Number of Pages

96

Format

.pdf

Creative Commons License

Creative Commons Attribution-Noncommercial 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License

Download

Share

COinS