Date of Graduation


Document Type


Degree Type



Eberly College of Arts and Sciences


Geology and Geography

Committee Chair

Gregory Elmes

Committee Co-Chair

Jamison Conley

Committee Member

Jamison Conley

Committee Member

Trevor Harris

Committee Member

James J. Nolan

Committee Member

Emily H. Griffith


The Uniform Crime Reporting (UCR) Program is a law enforcement statistical system open to unreported information due to its voluntary nature. As such, there need to be a valid and accepted means to estimate official reports of crime for those different levels of geography where reporting may be incomplete. Current methods of imputing and modeling UCR data, which have not been updated since the 1960s, are based upon conceptualizations of law enforcement agencies that may no longer be valid. These older models do not appropriately represent the law enforcement assessment of space and place and its effects on discretionary recording behavior. The number of specialized agencies that share jurisdiction and population with primary law enforcement agencies has increased since early data modeling techniques were developed around the 1960s. This study explores the connection between the policing and the collection of crime data to advance our understanding of how differences among types of law enforcement may impact the discretionary decision to record data. To explore this topic, I have divided this study into three papers touching on differing dimensions of place, scale, and uncertainty connected to the recording of law enforcement data. The data for these papers includes national UCR Program data, as well as calls for service and recorded incident data from two law enforcement agencies in the mid-South—Knoxville Police Department and the University of Tennessee Police Department.

Firstly, this research explores the influence of agency attributes to assess their possible impact on the treatment of missing data. The coefficient of variation (CV) is used to measure the internal variation of reported crime within various groups of agencies. The average CVs calculated with and without specialized agencies are compared using a Jackknifing technique to test whether the presence of specialized agencies increases the internal variation within the group or not. The comparison demonstrates that eliminating specialized agencies from the strata has a statistically significant effect on reducing internal variation for property crimes. For violent offenses, however, the results are more modest. While the average CV for violent crime does decrease with the elimination of specialized agencies, the improvements are not statistically significant. The results from this research point to a greater need to address the changing circumstances to incorporate the diversity of law enforcement agency type.

Secondly, although there is an interest by researchers to use calls for service (CFS) as a useful proxy for recorded incident information by law enforcement as more of this type of data is made available in open data initiatives, the assumption that CFS could serve as a proxy for incident information in spatial analysis is not supported by the evidence. Instead, there is some indication that law enforcement activities are mediated by the agency’s goals for its data, such as intelligence-led policing or fulfillment of Clery Act reporting, thus affecting the recording of incident information. Using data from two different types of law enforcement agency within the same community, CFS and incident reports for property crimes in April 2014 were tested for spatial association using both the Cross-K function and the Co-location Quotient. Findings from this study show there is a modest amount of detectable clustering of CFS for the agency that fits a model of traditional municipal law enforcement. However, the law enforcement agencies serving a large university campus did not show any detectable spatial association for these events. The findings suggest that in the movement towards using open data researchers will need to take greater care in the selection of data to understand if underlying spatial assumptions about the data can be supported.

Thirdly, an increasing quantity of data is currently being made available by law enforcement agencies, but frequently that data is not a consistent level of areal aggregation and scale. Factors such as the Modifiable Areal Unit Problem (MAUP) and the Uncertain Geographic Context Problem (UGCoP) make rectifying differing scales problematic. Central to this problem are the dynamics of recording crime data and whether law enforcement activity—specifically the concept of the patrol officer in a boundary role—is a key influence that should be accounted for in crime data models. With data from a midsized, southern municipal police department, two dasymetric allocation techniques using street networks and street networks weighted by calls for service are used to test potential improvements on the scale and aggregation problem through the introduction of law enforcement activity into allocation models for recorded crime data. Results demonstrate that the introduction of law enforcement activity—especially officer-initiated activity—improves the overall fit of the allocation of recorded crime into smaller subjurisdictional units. In addition, there is modest evidence to advocate for the use of law enforcement-generated subjurisdictional units (such as a precinct or beat) as opposed to population-based Census Tracts. These findings suggest that the production of crime statistics is subject to influences originating from law enforcement agency policy and the recording behavior of its officers.

The findings of the three studies inform important discussions in the geographic community on the heterogeneous nature of law enforcement. More explicitly, combining the conclusions of these three papers contributes to an evolving understanding of the representations of place by geographic information science (GISc) and criminology, and the construction of place through the roles and behaviors of individuals, and the increasing use of “Big Geodata”. Future research with data collected from official police activities should consider the degree of uncertainty introduced by the nature of the activities themselves — especially considering the growing use, influence and reliance on georeferenced data produced by individuals not particularly informed about the nuances of geography.