The judicial decisions should made not only on the guilt of the convict and his or her sentencing but also on the way to protect society from his or her recidivism risk. In order to assess the future risk, decision makers generally consider the characteristics of the offences and the offenders' personal character issues. The purpose of this kind of risk assessment is to predict future crime and manage offender risk throughout the course of the criminal justice process. Risk assessment is the process whereby offenders are assessed on several key variables empirically known to increase the likelihood of committing an offense. Literature reports these variables, known as risk factors, are subdivided into static and dynamic factors. Both types of factors are known to share a causal relationship with criminal behaviour, however, static factors are historical and unchangeable, while dynamic factors are current and changeable.
This study aimed to develop an integrated form of assessment tool to measure future risk of offenders with various criminal acts. First, many risk variables were drawn from the literature review. As static risk factors age at first offence, history of prior convictions, and juvenile records were proven to be important predicting recidivism with other minor variables. Also dynamic factors effective to predict recidivism were education, marital status, attitudes supportive of crime, faulty cognitions, sexual deviant preference, alcohol use, substance abuse, and mental problems. During the first year of this study in 2010 the previous form of KORAS-G(Korean Offender Risk Assessment System-General) was developed with 17 risk factors.
In the second year of this study, the validity evidences were gathered for KSORAS-G through a recidivism follow-up study of the inmates who participated in the first year of study. Since an inmate with early release among participants of the first year of the study discharged from the prison in August in 2010, the maximum follow-up duration became 14 months. The recidivism was encoded as the records of indictment and conviction.
Each risk factors was analyzed to be effective predicting recidivism by chisquare, analyses of variances and linear and logistic regressions. As a result, rule violation records, alcohol use and drug abuse, previous criminal records, and antisocial tendency. Repetitive indictments were significantly correlated with education, early delinquency, and juvenile institutionalization.
Interestingly, almost none of risk factors were related to conviction records. This was inferred to happen since more erroneous variables besides criminal act itself were intervened such as employment of an expensive private lawyer. Therefore, indictment records were utilized to apply the final cutoff of total score of KSORAS-G. The ROC curve presented 13 must be the most proper cutoff to differentiate future risk of participants. The sensitivity of the specificity were repectively .952 and .381. The survival analysis presented offenders obtaining KORAS-G below the score of 12 had the accumulated survival rate was 85%. On the other hand the survival rate of offenders obtaining KORAS-G above the score of 13 was 74%. This difference between two survival probabilities was statistically significant(Wilcoxon (Gehan) statistic = 4.945, df = 1, p = .026).