Risk Assessment and Rehabilitative Potential

Mark Zelig, Ph.D., ABPP:

4325 Laurel Street, Suite 297
Anchorage, Alaska 99508

6925 Union Park Center, Suite 550
Cottonwood Heights, Utah 84047

fax: 866-907-2822
email: markzelig@markzelig.com

For Further Information Contact Us using our Online Request Form.

Concepts utilized in evaluating risk and rehabilitation potential

As pointed out by Douglas & Guy (2012), forensic psychologists “don’t predict risk, we assess it.”  This means we best serve the Courts and other decision makers by providing a sound evaluation of risk.  Indeed, a scientifically-based evaluation of risk is likely to lead to interventions that decrease the probability of a prediction coming true.

Risk, and one’s prospects for rehabilitation, are intimately tied. There are three main risk assessment techniques:

•    Unstructured Clinical Judgment,
•    Actuarial Risk Assessment, and
•    Structured Professional Judgment.

Unstructured Clinical Judgment is solely based on professional experience.  Research indicates it is better than chance when evaluating short-term risk, but does not provide reliable estimates of long-term risk.  In fact, there is considerable research indicating unaided clinical evaluation, based solely on interviews with the subject, fares worse than predictions from clinicians who do not interview the subject and instead, rely on file information and test scores (e.g., Grove 2000). Unstructured Clinical Judgment is exemplified by the witness who bases their opinion “based on my clinical experience . . .” without reference to empirically-established measures or decision-making guides.

Actuarial Risk Assessment evaluates risk based on factors present in a given individual.   A voluminous literature indicates that Actuarial Risk Assessment far surpasses Unstructured Clinical Judgment.  Actuarial approaches are particularly helpful in discriminating between groups of low, moderate, and high risk individuals over the long haul.  These instruments typically rely on static risk factors, or variables that do not readily change for a particular person.  Using the same science as an auto insurance carrier employs to determine the risk of a driver, these instruments assign the offender to a group, and then indicate the probability that members of that group will recidivate within a given time frame.  One advantage of most actuarial risk assessments is that many can be reliably scored by nonprofessionals – as long as they have access to sufficient information to determine if a given risk factor is present or absent.  One disadvantage is that outcome based on an actuarial result cannot be overridden by common sense: For example, consider a rapist assigned to a high risk group based on the accumulation of  various static risk factors.  If he subsequently becomes a paraplegic  he would remain in the same risk recidivism group despite his obvious incapacitation.

Instruments that rely upon Structured Professional Judgment (SPJ) use criteria based on scientific research, which allow a trained professional to assign risk to a particular person.  Such instruments include static factors, but often supplement the evaluation with dynamic or changeable risk factors.  So for example, if an offender’s offense history is limited to occasions in which she used alcohol, an intervention that successfully curtailed or eliminated alcohol consumption would lower risk; the recurrence of alcohol use would increase risk.   SPJ is used as a model when an actuarial approach is not possible or practical.  Parenthetically, the efforts to develop an instrument to identify those at risk for terrorism have focused exclusively upon an SPJ approach:  We will never have an opportunity to follow a group of prospective terrorists for the sake of determining the factors that differentiate the low versus high risk individuals.

In the following sections I will discuss three commonly used instruments to predict risk in sex offenders – all of which are designed using either actuarial approaches or SPJ.

Psychopathy Checklist Revised, 2nd Edition  (PCL).

This instrument was developed to measure the construct of psychopathy, not  risk.  However, when subsequent use demonstrated that the construct of psychopathy and recidivism were closely connected, practitioners added the PCL to their armamentarium of forensic instruments.  Indeed, a survey of forensic psychologists found that most recommended that the PCL be included within a battery designed to evaluate the risk of recidivism (Lally, 2003).

The PCL developer, Robert Hare, views psychopathy as manifesting in “three broad categories . . . . [1] Interpersonally, psychopaths are grandiose, egocentric, manipulative, dominant, forceful, exploitative, and coldhearted. [2] Affectively, they display shallow and labile emotions, are unable to form long-lasting bonds to people, principles, or goals, and are lacking in empathy and genuine guilt and remorse. [3] Their lifestyle is impulsive, unstable, and sensation-seeking.  They readily violate social norms and fail to fulfill social obligations and responsibilities, both explicit and implied” (PCL-R Test Manual, p. 5).

The PCL includes 20 items, whose ratings are based on information obtained from the subject’s history and clinical observation.  The test manual requires that collateral sources be used in addition to information obtained from the subject, especially when the subject denies involvement in any of the 20 areas.

Although Dr. Hare views psychopathy as manifesting in three categories, the present edition of this test yields scores for two factors, each of which has two facets.  Summing the raw scores of each of the four facets, plus two additional items, yields the total score.

•    Factor 1: Interpersonal/ Affective – Selfish, Callous, and Remorseless Use of Others (Aggressive Narcissism). Facets include:

•    Facet 1: Interpersonal
•    Facet 2: Affective

•    Factor 2, Social Deviance is described as “Chronically Unstable, Antisocial, and Socially Deviant Lifestyle” (Chronic Antisocial Behavior).    Facets include:

•    Facet 3: Lifestyle
•    Facet 4: Antisocial

This scale rates the subject on characteristic life patterns.  Therefore, any behaviors exhibited in the course of committing the instant offense would not be scored on this measure, unless the behavior was consistent with historical life patterns. For example, if I knew with certainty that a subject’s denial of criminal liability was deceptive, I would not score him on the Pathological Lying item unless deception was a fairly consistent life long characteristic.

For most items, a score of 0 is awarded for No; 1= Maybe;  2= Yes.  This means that if collateral or reliable psychological test data was lacking for a particular item, I would award the subject a score of 1.


Psychopathy is a construct that is relevant for identifying risk in all criminal populations.  The more that a person possesses psychopathic traits, the less reservations they have about exploiting others.  Moreover, there is a large literature that finds that psychopathic individuals do not learn from their mistakes, profit from experience, nor are deterred by the possibility of sanctions for their exploitive behaviors.  A recent meta analysis conducted by Hawes and his associates, pooled the results of 20 different studies, involving a total of 5239 subjects.  The outcome verified the ability of the PCL to inform decision makers about the risk a particular person possesses.

In terms of predicting sexual recidivism, the authors found that Factor 2 and Facet 4 were more predictive than Factor 1and the Total Score.  To the extent that an offender had elevated scores on the PCL and a measure of sexual deviance, they were “more likely to reoffend sexually than other offenders” with odds ratios ranging from 2.8 to 3.2 (p. 233).  In their conclusion, Hawes et al. emphasized “in terms of effect size, findings from this meta-analysis suggest that the strongest rationale for administering the PCL-R in sex offender risk assessment may not be the predictive effects alone but, rather, the predictive value of high PCL-R scores and sexual deviance” (p. 241).

Static 99R

The Static 99R is an actuarial rating instrument composed of items that reflect the consistent finding that recidivism in sex offenders is most strongly related to a given offender’s (1) general criminal persistence, (2) sexual persistence, (3) young age, (4) lack of of involvement in intimate relationships, and (5) selecting male victims (see Barbaree et al. 2006 for a discussion of broad and robust risk factors).

The Static 99R represents the third revision of an instrument developed on the basis of  scientifically-supported predictors of recidivism, and has demonstrated moderate levels of accuracy in ranking offenders according to their relative risk for sexual recidivism, across a wide variety of samples, countries, and unique settings.

Research indicates that the relative risk associated with different Static-99R scores remains constant, “even though the base rate of recidivism varies across different correctional and geographic samples. Risk ratios for Static-99R scores were based on 8 samples of sexual offenders from Canada, the United States, the United Kingdom, Austria, and Sweden (n = 4,037). The analyses were based on routine (i.e., relatively unselected) correctional samples.”  Accordingly, in most cases, the bottom line is expressed in terms of the offender’s relative risk.

Modifiers to consider in regard to the Static-99 results.

1.    Undetected recidivism.  All actuarially-based risk instruments report likelihood of recidivism based on detected offenses.  Hanson, Morton, and Harris (2003) suspect that actual recidivism is 10-15% higher than detected rates.

2.    Denial and acceptance of responsibility.  It is no accident that the offender’s avowed willingness to take responsibility or deny responsibility is not contained in the Static-99.   Since this statement may be counterintuitive to many readers, I will discuss research findings relevant to this topic in a separate information sheet.

Sexual Violence Risk – 20

The Sexual Violence Risk 20 (SVR-20), published in 1997, is a commonly used instrument that relies upon SPJ to determine the presence/absence of 20 risk factors in a given offender. Analysis of these risk factors is used to classify a given offender into Low, Moderate, or High risk categories.   As originally designed, there are no absolute cut offs that classify a particular offender into one of these risk categories.  In addition, the professional determines the relative weight of a factor for a particular offender.

Overview of research.   The strengths of this instrument is that it is able to accommodate dynamic or changeable risk factors (e.g., attitude toward treatment, abstinence from alcohol) in addition to static (unchangeable) factors (e.g., the offender’s criminal history).  It has demonstrated much better accuracy than Unstructured Clinical Judgment, and is equivalent to actuarial instruments in terms of long-term predictive validity.  The SVR-20 has held up well in validation studies involving different cultures and countries  (e.g., Rittenberger, Boer, & Eher, 2011).

The authors define sexual violence as the: [A]ctual, attempted, or threatened sexual contact with a person who is nonconsenting or unable to give consent. The definition has two major elements. The first concerns the nature of the act. We construe sexual contact broadly to include acts such as sexual battery (e.g., rape, sexual touching), communications of a sexual nature (e.g., exhibitionism, obscene letters or phone calls, distribution of pornography), and violating property rights for sexual purposes (e.g., voyeurism, theft of fetish objects). The second element concerns the victim of the act. In most cases of sexual violence, victims are aware of the acts perpetrated against them, but do not consent to the acts. That is, the sexual contact is coerced. In other cases, victims may assent to the acts, but are unable to give consent (i.e., true, full, legal consent) by reason of youth or mental disability. Finally, some victims are unable to give consent because they are not aware of the acts perpetrated against them (e.g., victims of voyeurism)” (p. 9).

The following ratings are awarded to each risk factor:

Yes     = Evidence of a history of the trait
?         = Possible/ partial evidence
No      = No evidence of history of the trait

Research support.    The Rettinger et al. 2011 article was discussed earlier.  Another authority, commenting on this instrument stated,  “Meta-analytic results (Hanson & Morton-Bourgon, 2004) suggest that a slightly lower, but broadly similar level of predictive accuracy is typically achieved with this method. The smaller number of studies involved means that predictive accuracy is known with less precision than is possible for the actuarial instruments. The SVR-20 itself has a relatively high average predictive accuracy, but with results varying widely between individual studies” (Knight & Thornton, 2007, p. 17)

Comments are closed.