An examination of offense specialization using marginal logit models

Glenn Deane, David P. Armstrong, Richard B. Felson

Research output: Contribution to journalReview article

37 Scopus citations

Abstract

Research on offense specialization has concluded that there is a great deal of versatility in offending. Although the preponderance of evidence supports versatility, some research points to a small but significant tendency to specialize. Beyond this observation there is little consensus over the degree of offense specialization, the similarities and differences between people who commit violent acts and those who engage in other criminal behavior, or the extent to which general causal processes are sufficient to explain variation in diverse forms of crime and delinquency. At the heart of the confusion is the fact that criminal behaviors across a wide spectrum are positively correlated with one another. In our opinion, the conclusion that general offending trumps offense specialization is the result of research designs that predetermined such a conclusion. We propose an alternative method, marginal logit modeling, that supports many desirable features suited to the investigation of offense specialization. We analyze nine self-reported delinquent behaviors (with a tenth category representing "No Offense") from the Add Health study. We show that violent offenders are more likely to engage in additional violent offenses, nonviolent offenders are more likely to engage in additional nonviolent offenses. For some offense types, we find no evidence of a tendency to commit both violent and nonviolent offending. For others, the offense generalization effect is weak compared to the offense specialization effect.

Original languageEnglish (US)
Pages (from-to)955-988
Number of pages34
JournalCriminology
Volume43
Issue number4
DOIs
StatePublished - Nov 2005

All Science Journal Classification (ASJC) codes

  • Pathology and Forensic Medicine
  • Law

Fingerprint Dive into the research topics of 'An examination of offense specialization using marginal logit models'. Together they form a unique fingerprint.

  • Cite this