Burnham and Anderson observed that while a model can never be “truth,” a model might be ranked on a continuum from very useful, to useful, to somewhat useful, to essentially useless. The prevailing Risk-Needs-Responsivity (RNR) model is essentially useless for two reasons: 1) there is no difference in recidivism between Second Chance Grant recipients and non-Second Chance recipients, and 2) our probation and parole numbers have been increasing not decreasing as jurisdictions, unquestioningly, adhere to the RNR model's principles and tenets. The fundamental attribution bias of overestimating the role of person-factors while underestimating the role of each jurisdictional environment is a key aspect of RNR risk assessment algorithms. Thus, the RNR model and its associated risk assessment instruments have no ecological validity. More specifically, neither attends to variations in “Get Tough” jurisdictions policy. Yet, “Get Tough” variables are unacknowledged moderator variables.
TopRisk Assessment Instrument Characteristics
Should one let the defendant go free until their trial date, or should one put them in jail? Detaining defendants will create havoc in their lives and detaining will cost taxpayers thousands. Does it matter if the defendant has been arrested before? Convicted before? Convicted of a felony before? Risk assessments have been a part of the United States criminal legal system for almost century. However, for many decades experts (psychologists, social workers, probation officers) used their judgement to answer the questions above. Yet, nowadays, risk assessment instruments are used at each stage of the bail, sentencing, parole process. And, risk means different things at each stage of the process. This chapter explores the contents of risk assessments and argues there are critical components missing from them.
The Council for State Governments funded a review by Desmarais, Singh, and Johnson of risk assessment instruments. After reviewing 53 studies (some journal articles, some government reports, and some master’s theses and dissertations), researchers identified 19 assessment instruments designed for predicting the risk of general recidivism (Desmarais et al., 2016, p. 208).
Table 1
indicates that most instruments were designed to assess the risk of new offenses. Estimated administration time was reported in the manuals ranged from 5–10 minutes for the Ohio Risk Assessment System – Community Supervision Screening Tool (ORAS-CSST) up to 60 minutes for COMPAS (Desmarais et al., 2016, p. 210). The number of items on the various risk assessment instruments ranged from four for the ORAS-CSST to 130 for the Inventory of Offender Risks, Needs, and Strengths (IORNS), with an average of 41.
Table 1. Risk instruments characteristics
Source: Adapted from Desmarais, Singh, & Johnson (2016, p. 209)
Eight risk assessment domains were identified by Bonta, Andrews, and their colleagues; these are indicated in red in Table 2.
Table 2. Risk and need assessment instruments content domains
Source: Adapted from Desmarais, Singh, & Johnson (2016, p. 211)