Predicting adult offender recidivism

PDF (25KB)

Research summary
Vol. 2 No. 2
March 1997

Question

What are the best predictors of offender recidivism and, secondly, what type of actuarial measures are best suited for this purpose?

Background

In North America, many prison and probation settings are strained to capacity. For these organizations to protect the public in the most cost-effective manner possible, correctional authorities must concentrate their resources on higher-risk offenders. In order to achieve these goals, it is important that as precise an estimate as possible is obtained regarding the ability of various predictors and actuarial measures to predict recidivism.

Method

A meta-analysis of the literature on the prediction of adult recidivism was conducted. One hundred and thirty-one studies published since 1970 were reviewed producing 1141 correlations (r) between a predictor identified in a study and recidivism. r is a statistical measure of the strength of the association between two variables. Values of .20 and higher are regarded as practically important. The predictors were also grouped into one of three categories: 1) static, 2) dynamic, and 3) actuarial measures (i.e., a combination of factors). Recidivism was defined as a new arrest, conviction, incarceration, or parole violation.

Answer

The best individual predictors were "criminogenic" need variables (i.e., attitudes, values, and behaviours that support a criminal lifestyle, r = .18), criminal history (r = .16), social achievement (employment, education, r = .13), age/gender/race (r = .11), and family factors (r = .09). Weaker predictors were intellectual functioning (r = .07), personal distress (e.g., anxiety, self-esteem, r = .05), and social class (r = .05). Dynamic predictors, those that are changeable (e.g., criminogenic need), predicted recidivism as well as static predictors (e.g., age).

Two types of actuarial measures were assessed - composite measures of risk (e.g., Level of Service Inventory - Revised or LSI-R) and measures of anti-social personality (e.g., Psychopathy Check List). The LSI-R produced the highest correlations with recidivism and these were in the range of r = .30 - .45.

Policy implications

  1. The ideal assessment protocol should include the following constituents.

    Static Predictors
    • age
    • criminal history - both as an adult and juvenile
    • family factors - parental and family criminality, family rearing practices and structure
    Dynamic Predictors
    • anti-social attitudes and values
    • anti-social personality (e.g., psychopathy)
    • companions
    • social achievement
    • substance abuse
  2. Composite actuarial measures of risk outperform individual static and dynamic predictors and therefore, should be used in offender assessment.
  3. The measurement of offender change - assessing the dynamic risk factors of offenders - should be done routinely.
  4. It makes little difference which type of recidivism outcome measure is used for prediction purposes.
  5. Any correctional agency that: a) compares the ability of different actuarial measures to predict recidivism, b) assesses the usefulness of new actuarial techniques, c) generates prediction data on promising predictor domains and distinct groups of offenders (e.g., sex offenders), will enhance our knowledge and benefit practice.

Source

For further information

James Bonta, Ph.D.
Solicitor General Canada
340 Laurier Avenue West
Ottawa, Ontario
K1A 0P8
Tel (613) 991-2831
Fax (613) 990-8295
e-mail jim.bonta@ps-sp.gc.ca

Date modified: