Predicting New Criminal Convictions while on Parole:
The Role of Offense Type
Utah Commission on Criminal & Juvenile Justice
September 2019
Research Question
When controlling for relevant factors do offense type play a part in predicting new criminal convictions while on parole?
Motivation
Figure 1: Utah’s prison population: 2000-19
While risk and needs assessments provide the most accurate information in evaluating an offenders risk to re-offend, both the risk and severity of a particular crime to be repeated are important policy considerations.
Figure 2: Utah’s prison population by offense type
THIS STUDY…examines the role of offense type on the likelihood of a new criminal conviction while on parole and examines the severity of the new criminal convictions that occurred.
Data & Methods
Data
- 3,173 unique parole starts (between 2013-16)
- Predictor variables: Demographics, offense type/severity, and risk-to-reoffend (LSI:RNR)
- Outcome of interest: New criminal convictions (2-year follow-up time)
Methods
Primary analysis: Bayesian Model Averaging (BMA) to identify the variables that are important in predicting a new criminal conviction while on parole
Secondary analysis: Examining the severity of the new criminal convictions that occurred (descriptively)
| Variable | Mean | Sd | Min | Max |
|---|---|---|---|---|
| nc_2 | 0.13 | 0.34 | 0.00 | 1.00 |
| parole_los | 382.53 | 257.11 | 0.00 | 730.00 |
| age | 35.42 | 9.68 | 17.27 | 81.49 |
| male | 0.84 | 0.37 | 0.00 | 1.00 |
| minority | 0.31 | 0.46 | 0.00 | 1.00 |
| high_risk | 0.71 | 0.45 | 0.00 | 1.00 |
| first_parole | 0.29 | 0.45 | 0.00 | 1.00 |
| severity | 1.54 | 0.67 | 1.00 | 3.00 |
| dp_only | 0.09 | 0.28 | 0.00 | 1.00 |
| alch_drug | 0.13 | 0.34 | 0.00 | 1.00 |
| person | 0.19 | 0.39 | 0.00 | 1.00 |
| property | 0.32 | 0.47 | 0.00 | 1.00 |
| driving | 0.07 | 0.26 | 0.00 | 1.00 |
| sex | 0.15 | 0.35 | 0.00 | 1.00 |
| weapons | 0.02 | 0.14 | 0.00 | 1.00 |
| murder | 0.01 | 0.12 | 0.00 | 1.00 |
| other | 0.02 | 0.14 | 0.00 | 1.00 |
Table 1: Summary statistics (N=3,173)
Discussion
with error-bands
(2-year follow-up)
Limitations
- Quasi-experimental design
- Relationship between technical violations and new crimes is complex
- Low re-conviction rate (13%)
Policy Considerations
Policy makers should compare the likelihood and severity of a particular offense to re-occur when finding the appropriate balance between sentence type and length while taking into consideration society’s values and expectations as they relate to justice, public safety, and resource allocation.
Questions?
Thank You!
Email: snystrom@utah.gov; rnpoulson@utah.gov