The first principle of evidence-based practice (EBP) for corrections populations is to assess the risk level or the probability that a given offender will commit more crimes using an actuarial tool designed and tested for these purposes. It is likewise critical to measure dynamic or changeable risk factors also called criminogenic needs. Needs are factors that have been identified through multiple, rigorous research and evaluation studies as being correlated with criminal behavior. Among the strongest needs are: having poor problem-solving and decision-making skills, harboring anti-social attitudes, hanging around with anti-social people, being restless and impulsive, and being addicted to alcohol and/or drugs. An offender’s risk for general recidivism is often characterized using labels such as “low,” “medium” and “high,” which can be useful at the 30,000 foot level to sort offenders, but can be very misleading and mask all kinds of important contextual information that is necessary to account for in public safety. The following is a discussion of what I believe should be kept in mind when reviewing and discussing risk for offender populations:
It is critical to recognize that there are no universally accepted definitions for risk levels. What this means is that we cannot state that a low risk offender, for example, is one who has a 30 percent or less probability of being arrested for a new crime within a three year period. In order to make sense of these labels, practitioners and policymakers should be aware of:
- The specific actuarial instrument being used and the definition of recidivism associated with that particular tool. For example, recidivism or failure may mean only a new arrest or it could mean only a recommitment to a confinement setting. In the latter category, a particular general risk tool such as the LSI-R or COMPAS, may only “count” a reincarceration for a new offense or it may consider reincarceration for either/both a new crime or/and a technical violation of supervision conditions resulting in reconfinement.
- Next, the scientific validation and norming for individual actuarial instruments is based not only on the definition for failure used, but also on how long offenders are tracked. While many consider the “acceptable” standard to be tracking offenders for three years, there is wide variation in this practice. What’s more, research findings suggest that sub-populations such as sex offenders should be tracked for far longer periods. The same case for tracking offenders over longer time periods can and is often made when trying to predict violence, for example, as the base rate for this offending is much lower and steadier over time than is the case for less serious offenses. Regardless, it is important to know the time period over which offenders are tracked.
- Finally, as there is no absolute definition for any category, it is imperative to keep in mind that the levels only mean something when compared to other groups in the same population studied for validating and norming a particular instrument. So, if you are looking at a state prison population, for example, all of whom have been assessed using the same instrument, you can group offenders into categories. But what this means is that relative to other state prison offenders, the “low” group has the lowest base rates of recidivism, the medium group has a higher probability than the low, but not as much as the high. You get the idea.
Why Does This Matter?
- As so much discussion and decision-making, not to mention every evidence-based practice to follow, is tied to identifying an offender’s risk of committing additional crimes, it seems quite odd that there are no definitions, or in most cases, even questions asked, about what the labels mean in the absolute sense. So, in most instances, there is no discussion that a high risk offender in a particular scenario or population is one for whom the probability of being rearrested at least once within three years is 75% or whatever the data shows it to be. This lack of specificity in discussion can fuel the often held, but equally often mistaken, belief that probationers are less risky than jail inmates who are less risky than prison inmates. Since few jurisdictions use one tool or set of instruments that measure recidivism using the same definition and timeframe, it is virtually impossible to make these kinds of comparisons.
Solution
- For me, it is important to be transparent in all methodology, and data used. I think the lack of specificity and quite frankly, lack of inquisitiveness about these oft-used labels, can lead to inappropriate conclusions and bad policy and practice. One pretty straightforward solution is to state the definitions and timeframes used for the particular instrument under examination. Then the probabilities for each level should be stated. In any discussion of “low,” “medium,” and “high” risk, there should be a corresponding statement that includes the instrument used, the definition of recidivism included, the timeframe over which offenders are tracked and the probabilities associated with each risk level.
