Offender risk assessment is the foundation of many interventions in the criminal justice system that affects public safety and the offender’s liberty. The risk of crime that the offender presents influences arrest and sentencing decisions, parole release, how closely the offender is supervised in the community or in prison, and even who receives treatment and what kind of treatment. Treatment within the criminal justice context has the goal of reducing recidivism, and it may or may not involve mental health treatment depending on its link to criminal behavior. There are different approaches to assessing offender risk, and they have been described as generations of risk assessment, ranging from assessments that are the least valid to assessments that are more predictive of criminal behavior and more useful for managing offenders. The four generations are summarized and their key features are highlighted. The entry ends with some emerging issues and an observation of what may soon form fifth-generation assessments.
First Generation: Professional Judgment
First-generation assessment involves a professional who makes a judgment as to the offender’s risk to reoffend and the type of treatment required to reduce the risk. Information about the offender is typically gathered through interviews with the offender, file reviews, and perhaps interviews with people who know the offender. The important characteristic of first generation risk assessment is that the information is collected in an unstructured manner. Although the professional may gather some information that is common to all offenders, the interviewer may deviate and seek information that may be specific to the offender and his or her situation and influenced by the professional’s own training and experience. For example, a psychologist may administer some psychological tests, but the tests chosen are guided by the psychologist’s experience and expertise with particular tests. After collating the information, the professional arrives at a judgment regarding the offender’s risk to the public and the treatment that the offender requires. In other words, this first generation, professional/clinical approach is highly subjective and dependent upon the professional’s training and experience.
Research shows that compared to more structured, empirically validated risk instruments, professional judgments of risk perform no better than chance. One problem with this approach is that using informal, unobservable criteria for making decisions makes it difficult to replicate. For example, clinical experts often disagree on what are the critical factors in a case and, therefore, reach differing judgments on offender risk. Another problem is that the professional may attend to offender characteristics that are unrelated to criminal behavior and ignore the factors that are related to criminal behavior. Because of the poor predictive accuracy of first-generation assessments, this approach has largely been abandoned in favor of second- to fourth-generation assessments.
Second Generation: Actuarial, Static Risk Assessment
Second-generation assessments are actuarial in that the factors assessed are empirically predictive of future criminal behavior. One of the earliest second-generation risk assessment (1928) consisted of 21 items to predict parole failure. Items included factors common to today’s instruments (e.g., drug addiction, prior criminal record) and items that are strange by today’s standards (e.g., hobo, ne’er-do-well). In second-generation risk scales, items are typically assigned a score of 0 for its absence or 1 for its presence. This is called the Burgess method. Another scoring method uses various statistical methods to assign weights to the item. For example, drug addiction may receive a score of 5 and unemployment a score of 2, reflecting the importance of the specific risk factor. Neither scoring approach has demonstrated superiority. Thus, the Burgess 0–1 scoring method is more common because of its ease of use. Essentially, as scores increase then the likelihood of recidivism increases.
Second-generation risk instruments predict criminal outcomes reasonably well and as well as third- and fourth-generation instruments. However, because they consist almost entirely of static historical items, these instruments fail to accommodate offender change and have little use in case planning and risk reduction. For example, someone who scores high on a second-generation assessment early in life will remain high-risk years later even if the person has adopted a prosocial lifestyle and has stopped all criminal activity. The offender who improves his or her life receives little credit for his or her effort. In addition, these assessments do not tell the criminal justice professional what can be done to reduce the offender’s level of risk.
Third Generation: Risk/Need Scales
Second-generation risk scales are used for security, release, and supervision decisions, but the public also expects the criminal justice system to lower the offender’s risk and to reintegrate offenders into society. The lack of risk reduction utility in second-generation assessments is overcome with the introduction of dynamic, changeable risk factors referred to as criminogenic needs. The third-generation risk instruments are known as risk/need scales.
Criminogenic needs are offender needs that are associated with criminal behavior. Examples are substance abuse, procriminal attitudes, and employment problems. Offenders also have noncriminogenic needs (e.g., poor self-esteem, vague feelings of emotional discomfort). The third- generation offender assessments include both static criminal history items and criminogenic need factors. What further distinguishes third-generation assessments from second-generation assessments is that the majority of the items in third-generation offender instruments consist of dynamic, criminogenic need items.
The most widely used risk/need instrument is the Level of Service Inventory–Revised (LSI-R). Approximately two thirds of the 54 items are dynamic and organized into subcomponents that measure many of the major criminogenic needs of offenders (e.g., education/employment, companions, family/marital). The scoring of the LSI-R uses the Burgess 0–1 method. One of the advantages of the LSI-R, as with all third generation risk assessments, is that it can be used to formulate an offender management plan. By identifying the offender’s criminogenic needs (high scores on the subcomponent), the supervising staff can target reductions in the criminogenic needs and achieve an overall reduction in recidivism. Another advantage is that reassessments can be used to monitor offender changes over time and lead to readjustments in the case plan.
Fourth Generation: Risk/Need Case Management
As just noted, a major feature of third-generation assessments over second-generation instruments is that they offer the promise of guiding offender case management. However, there are a few studies showing that despite administering a risk/need instrument staff fail to use the assessment findings in case planning. More often than not, case plans have not incorporated strategies for addressing the offender’s criminogenic needs. Consequently, when staff behavior is directly observed (e.g., via audio recordings or video tapes), researchers have found staff failing to focus on the criminogenic needs identified by the assessment.
Recognizing the disconnect between administering a validated risk/need instrument and actually using the findings in the day-to-day supervision of offenders has led to the development of a more structured way of linking the assessment and the case management plan. The best validated fourth-generation instrument is the Level of Service/Case Management Inventory (LS/CMI). The LS/CMI evolved from the LSI-R and has added features to better align the instrument to the Risk-Need-Responsivity model of offender classification and treatment. For this reason, the LS/CMI will serve as an illustration of fourth-generation assessment instruments.
The LS/CMI has a number of sections with the first section basically reflecting the LSI-R but reconfigured to represent the eight major risk/need factors. Section 1 of the LS/CMI provides the overall offender risk score. Section 2 has 21 items that have the potential of being criminogenic for a particular offender (e.g., sexual offense with a child victim, stalking behavior). Section 2 does not produce a total score for risk classification. It is intended to ensure that the test administrator has a comprehensive assessment and does not miss aspects of the offender and his or her situation that could affect supervision. Section 5 assesses responsivity factors. Responsivity refers to individual characteristics that may influence how correctional staff engages the offender during supervision and treatment. For example, supervising an offender with a mental disorder would require a different strategy compared to an offender with no mental disorder. The assessment of responsivity factors is relatively cursory with only 10 items (e.g., poor motivation, low intelligence).
The most important feature of the LS/CMI, and what makes it truly a fourth-generation instrument, is the integration of the assessment with case management. In Section 9, the test administrator rank orders the criminogenic needs of the offender, sets treatment targets, and outlines the steps and interventions needed to reach these goals. This is the case management plan. Section 10 records progress and documents whether the goals are reached (the remaining sections focus on the prison experience and other noncriminogenic factors that may influence treatment and supervision). Furthermore, all of the information from risk/need assessment to case planning and monitoring are presented in one booklet. This is done to ensure that staff focuses their activity on addressing the offender’s risk and needs in a structured manner. In summary, fourth-generation offender assessment includes a comprehensive sampling of offender risk, needs, and responsivity factors along with the integration of these factors into a case management plan.
Outstanding Issues and Future Directions
Although second-generation risk assessments predict recidivism, they lack the advantages of the later risk/need instruments. One would expect that second-generation instruments would have been replaced with the newer instruments, but they remain in widespread use. The reasons for their continued use are 2-fold. First, because second-generation scales have fewer items than third- and fourth-generation risk scales they are easier to use, more efficient, and less costly. However, second-generation instruments fail to assess criminogenic needs. This failure then relates to the second reason for the continued use of the static risk scales: the belief that offenders cannot be rehabilitated. If one accepts this view or the view that it is not the business of corrections to correct except through the application of punishment, then third- and fourth-generation assessments add little value.
Another issue is the increasing popularity of what is called structured professional judgment (SPJ). SPJ instruments such as the Historical- Clinical-Risk Management-20 structure the factors to be considered but do not link actuarially scores to outcomes. For example, a test administrator may be given guidelines to evaluate substance abuse problems; relationship instability; and attitudes in terms of low, medium, or high risk, but there is no scoring of items or adding them to produce a total score that is associated with a probability of recidivism. SPJ has elements of first-generation (clinical judgment), second-generation (structure), and third-generation (criminogenic needs) assessments. The research on the predictive validity of SPJ has found SPJ to perform better than first-generation professional judgment. However, one review has found it to not predict as well as actuarial assessments for sex offenders. Other reviews of SPJ have mostly focused on the Historical-Clinical-Risk Management-20 where the effect size estimates have ranged from moderate to high. The reason why the Historical-Clinical-Risk Management-20 and its variants may do reasonably well in predicting recidivism is that those completing the instrument have a high level of professional expertise in assessment technologies to begin with and thus are more committed to the assessment process. The SPJ instruments are highly appealing to psychologists, psychiatrists, and other forensic workers because they fit well with a professional training that highly values clinical decision-making.
Related to the continued use of professional judgment is the issue of professional override in third- and fourth-generation risk assessments. A professional override alters the risk level calculated by the instrument. Sometimes a professional override is dictated by policy (e.g., all sex offenders are considered high risk even if they score low risk), or the test administrator is given personal discretion to override the test results. The third- and fourth-generation instruments that have an override feature advise to use the override sparingly and with justification. However, research on the application of overrides has found that they are excessively used and that they increase prediction error.
Progress in offender assessment will continue and, just as each successive generation of offender assessment is replaced by more comprehensive and useful assessment instruments, fourth-generation assessments will eventually lead to a fifth generation. What exactly will fifth-generation instruments look like is difficult to predict. Two possibilities exist. First, there may be an insertion of acute, fast-changing dynamic risk factors such as the sudden loss of employment or intoxication into fourth-generation assessment instruments. Second, considerable research has accrued on the neurophysiological correlates of criminal behavior, particularly violent behavior. The results from these studies could also be added to the present risk/need tool kit. The important lesson is that offender assessment development must be guided by empirical research.
References:
- Andrews, D. A., Bonta, J., & Wormith, S. J. (2006). The recent past and near future of risk and/or need assessment. Crime & Delinquency, 52, 7–27. doi:10.1177/0011128705281756
- Bonta, J. (1996). Risk-needs assessment and treatment. In A. T. Harland (Ed.), Choosing correctional options that work: Defining the demand and evaluating the supply (pp. 18–32). Thousand Oaks, CA: Sage.
- Bonta, J. (2002). Offender risk assessment: Guidelines for selection and use. Criminal Justice and Behavior, 29, 355–379.
- Bonta, J., & Andrews, D. A. (2017). The psychology of criminal conduct (6th ed.). New York, NY: Routledge.
- Olver, M. E., Stockdale, K. C., & Wormith, S. J. (2014). Thirty years of research on the level of service scales: A meta-analytic examination of predictive accuracy and sources of variability. Psychological Assessment, 26, 156–176. doi:10.1037/a0035080
- Smith, P., Cullen, F. T., & Latessa, E. J. (2009). Can 14,373 women be wrong? A meta-analysis of the LSI-R and recidivism for female offenders. Criminology & Public Policy, 8, 183–208. doi:10/1111/j.1745–9133.2009.00551.x