Document Type

Dissertation

Degree

Doctor of Philosophy

Major

Criminology and Criminal Justice

Date of Defense

11-1-2018

Graduate Advisor

Richard Rosenfeld, PhD

Co-Advisor

Janet Lauritsen, PhD

Committee

Janet Lauritsen, PhD

Richard Rosenfeld, PhD

Adam Boessen, PhD

Elizabeth Groff, PhD

Abstract

Spatial crime studies have existed for over a century, but the last 20 years have seen a turn in focus toward micro-spatial units such as street blocks and street segments. A particular subfield of this modern micro-spatial perspective is called crime trajectory analysis, which can isolate patterns of crime at small spatial units over time. Though several crime trajectory analyses have been conducted for coastal cities, the technique has never been applied to Midwestern data. This project fills that research gap by using the group-based trajectory modeling (GBTM) algorithm to uncover patterns of violent crime at street segments in St. Louis, MO. The project addresses four specific issues which have gone either understudied or unexplored in prior research. First, using Uniform Crime Reports (UCR) data, an attempt is made to model violent crime trajectories with GBTMs by finding the best balance of parsimony and model fit, thus producing a model in better concordance with crime theory and having more practical applicability. Second, using data from the UCR, the decennial Census, and the City of St. Louis, multinomial logistic regression models are employed in order to find relationships between the crime trajectories and the various demographic and land use characteristics of the street segments. Third, the demographic factors that influence the amount of segment-level violent crime heterogeneity within a neighborhood are investigated using ordinal logistic regression models. Fourth, negative binomial regression models are used to distinguish which demographic and land use characteristics of a high-crime street segment influence the probability of the segment being located near others of its kind. The results show that although GBTMs can be made simpler by prioritizing parsimony, their ability to link demographic changes to crime changes over time is questionable. It is also demonstrated that high means and low variances in crime risk factors across a neighborhood’s segments produce higher violent crime trajectory heterogeneity in the neighborhood, and that social disorganization, not land use, is the primary cause of high-chronic violent crime clustering. Applications and optimal directions for future research are discussed in the hope that the findings are useful for police departments and practitioners.

Included in

Criminology Commons

Share

COinS