Statement of Shawn D. Bushway
Associate Professor of Criminal Justice
University at Albany (SUNY), School of Criminal Justice
- Current Practice for assessing risk of offending.
- Individual decisions made by human resources
- In most cases, individual decision making is worse than formal decision making based on statistical rules.
- Criminal history records are complicated documents intended for informed members of the criminal justice system.
- Statistical risk assessment using samples of ex-offenders.
- Large literature on the use of statistical rules to conduct risk assessment of ex-offenders
- Sentencing
- In prison management
- Parole
- Four best predictors of risk in samples of ex-offenders: age, criminal history record, race and sex
- Most tools only use aspects of the criminal history record
- Caveats
- Most tools only use criminal history record
- Ignoring age, sex and race could still lead to bias because criminal history records are correlated with age, sex and race.
- Evidence is based on official criminal history records with fingerprint quality matches.
- Public Criminal History record searches with finger print matches have a number of well known problems centered on dispositions.
- Private criminal history record searches with name and date of birth matches are substantially worse, and may miss up to half of all individuals with criminal history records.
- Why is recidivism the measure for employment decisions?
- Overall quality of prediction is modest at best
- Particularly hard to predict rare, high-risk offenders.
- Any decision rule will include substantial numbers of people who were not in fact high risk.
- Criminal history records
- Type of offense largely irrelevant
- Offenders are generalists not specialists.
- Past offense modest predictor of the next offense
- High rate offenders will tend to have more violence, even without specialization
- Property offenders actually recidivate at higher rates than violent offenders
- Criminal history record DOES NOT indicate individual always at a higher risk for recidivism than an individual with no record
- KEY QUESTION: How many years of non-offending does it take before an ex-offender looks like a non-offender
- Active area of research
- Risk differential is very high – up to 10 times higher -- immediately after arrest/contact with the criminal justice system
- Risk of offending for young first-time offenders declines to the same level as those with no record by roughly 6-8 years after the first arrest.
- KEY INSIGHT: Non-offenders do not have zero risk
- Definition of non-offenders very important in employment realm
- What is the level of risk of people for people who currently work for the company?
- Nature of criminal history record check
- Caveats
- Time since conviction might not be as relevant as time since release from prison for those incarcerated.
- Number of prior offenses MIGHT be good predictor of relative risk even after a number of years of release
- This risk assessment is being done in an environment where criminal history record has costs. But, risk assessment does not take current practices into account.
- Time since last conviction may not be causal – could just be good signal.
- Potential mitigators for ex-offenders besides years since last conviction.
- Abstinence from drugs/alcohol decreases risk dramatically for ex-offenders
- Stable employment (a year on same job) is rare, and predictive of low risk
- Age: 40 year old ex-offender much less risky than 20 year old.
- Marriage