Model-Based Estimation of Offending and Co-Offending Using the Criminal Participation Matrix

Peter J. Carrington, University of Waterloo
Ove Frank, Stockholm University

Police data under-report the numbers of incidents and of offenders, the numbers of offenders participating in individual criminal incidents (incident sizes) and the numbers of incidents in which individual offenders participate (offender activity). We show that co-offending is a concept which underlies, and unifies, all of these phenomena, so that the numbers of incidents and of offenders, and incident size distributions and offender activity distributions, can all be derived from the co-offending (criminal participation) matrix. Two related statistical models are presented which permit the estimation of numbers of incidents and offenders, incident size distributions, offender activity distributions, and co-offending distributions, from police-reported crime data, and data on the reporting of crime to police. These models rely on simple probability mechanisms for incident reporting and identification of offenders. The models are estimated, using Canadian Uniform Crime Reporting Survey and victimization survey data for 1995-99, and offender activity data from criminal careers studies. The implications of the results of fitting the models are discussed, as are other data which might be used.

(Return to Program Resources)

Updated 05/20/2006