The Criminogenic Embeddedness Model

A life-course model of strain, control, and social learning

CEM is a life-course model of offending that integrates strain, social control, and social learning theories into a single relational framework, rather than treating them as competitors.

The core idea

The model’s organizing premise is that each existing tradition in criminology has identified one genuine component of offending:

Tradition Component it identifies
Strain theory Motivation
Social learning theory Transmission
Control theory Constraint
Developmental and life-course theories Trajectories
Network theories Position

CEM’s contribution is to model the interactional structure these components form when they operate together in a person’s life, rather than adding one more component to the list.

The model is built around four offending pathways and a set of formal propositions, including a provisional measurement agenda, the Criminogenic Embeddedness Assessment Score (CEAS), for eventually operationalizing embeddedness directly. It was first developed in my 2026 book, Criminogenic Embeddedness and Criminal Trajectories: A Life-Course Model of Strain, Control, and Social Learning (DOI: 10.5281/zenodo.21961664).

Why do people exposed to the same conditions (the same strain, the same networks, the same disadvantage) so often arrive at different outcomes?

Empirical testing

CEM’s propositions are being tested against independent, publicly available longitudinal datasets, chosen to check whether findings generalize across a justice-involved sample and a general-population sample:

A third paper, focused on the model’s desistance-related propositions (an asymmetric “ratchet” pattern in how embeddedness accumulates and predicts relapse into offending), is in planning. See Publications for current status.

Where it began

One strand of this work traces back further than the book itself. My 2019 doctoral dissertation on unemployment, poverty, and recidivism in West Virginia found that macro-level economic strain did not predict recidivism differences, while individual factors such as age and supervision status did: an early, unlabeled version of the question CEM was eventually built to answer.