Chapter 2 The Empirical Basis to the Psychology of Criminal Conduct

Overview

The understanding of criminal behavior sought by PCC is empirical, theoretical, and practical. In this chapter, the empirical methods used to understand criminal behavior are described and summarized. Four major research designs are presented: 1) cross-sectional, 2) longitudinal, 3) multi-wave longitudinal, and 4) randomized experiments. The first three types of designs provide information about the correlates of criminal conduct (simple correlation, predictors, dynamic risk factors). However, we ultimately seek the causes of criminal behavior and for this we need randomized experimental research.

No single study, even a randomized experiment, can provide the “true” value of research concerning criminal behavior. Meta-analysis, because it involves many investigations, sometimes numbering in the hundreds, provides an unbiased and quantitative assessment of a body of literature. This chapter provides a primer on the importance of meta-analysis and the common statistics used in building knowledge.

Technical Notes 2.1 and 2.2

Worth Remembering

1. PCC seeks to understand the personal, interpersonal, social, and situational variables associated with criminal conduct. There are different types of covariates: simple correlates, predictors, dynamic risk factors (criminogenic needs), and causal variables.
2. Different research designs yield different types of covariates. Cross-sectional research designs identify correlates; longitudinal designs yield information on predictors; multi-wave longitudinal designs identify dynamic risk factors; and randomized experiments provide the highest level of knowledge: causal variables. Knowledge of the causes of criminal behavior has the potential to develop interventions that may reduce criminal conduct.
3. Testing the null hypothesis through statistical significance is falling out of favor.
4. A very handy and powerful way of describing the strength of the covariates of criminal behavior is the Pearson Product Moment Coefficient, or r. The r may be also interpreted through the Binomial Effect Size Display (BESD). BESD is the difference in the percentage of cases criminal in one condition (e.g., high risk) compared to the percent criminal in another condition (e.g., low risk).
5. The Pearson Product Moment Coefficient (r) is not without limitations, especially when base rates and selection ratios are at the extremes. Therefore, the Area Under the Curve (AUC) measure is often used in studies of risk scales because it is unaffected by base rates and selection ratios.
6. A meta-analysis is a quantitative review of the literature. The results from primary studies are converted into a common metric or effect size and averaged. With the addition of the Confidence Interval (CI), we come closer to estimating the “true” relationship.
7. The covariates of criminal behavior, regardless of their type (e.g., predictors, causal variables), will often depend on other variables. These are called moderator variables.
8. There is a direct connection between individual differences in criminal behavior and aggregated crime rates, but one must be cautious when interpreting findings at the aggregate level with reference to individual differences.

Quiz

Further Reading

Campbell, Donald T., & Stanley, Julian C. (1966). Experimental and Quasi-experimental Designs for Research. Boston: Houghton Mifflin Company. https://www.sfu.ca/~palys/Campbell&Stanley-1959-Exptl&QuasiExptlDesignsForResearch.pdf

Gendreau, P., & Smith, P. (2007). Influencing the “people who count”: some perspectives on the reporting of meta-analytic results for prediction and treatment outcomes with offenders. Criminal Justice and Behavior, 34, 1536–1559.

Hanson, R. K. (2022). Prediction statistics for psychological assessment. Washington, DC: American Psychological Association.

Lambdin, C. (2012). Significance test as sorcery: Science is empirical—significance tests are not. Theory & Psychology, 22, 67–90.